Physiologic studies of spinal inhibitory circuits in patients with stiff-person syndrome

Stiff-person syndrome (SPS) is a syndrome of fluctuating muscle stiffness with superimposed spasms.¹ Muscle stiffness has improved with diazepam²; this observation has led to the hypothesis that the disorder impairs inhibitory circuits which use the neurotransmitter gamma-amino butyric acid (GABA). Supporting this hypothesis is the finding that two thirds of SPS patients have circulating antibodies against GABAergic neurons,^3,4 which react against the intracellular enzyme glutamic acid decarboxylase (GAD) in most SPS patients. However, whether antibodies against GABAergic neurons play a causative role or are a secondary phenomena in a more generalized autoimmune disorder is unclear. Patients often have other autoimmune abnormalities, such as diabetes mellitus, thyroiditis, and pernicious anemia.^3,5,6 Furthermore, circulating antibodies against GABAergic neurons and GAD may be present in asymptomatic patients, particularly those with diabetes mellitus.^3,7

Nevertheless, the clinical features of SPS are compatible with a dysfunction of central inhibitory mechanisms. The muscle stiffness results from involuntary firing of motor units in a pattern that resembles a normal contraction but persists even during contraction of antagonist muscles,⁸ an action that normally activates reciprocal inhibition. This "continuous motor unit activity" is suppressed by epidural or peripheral nerve block⁵ as well as by drugs with central actions such as diazepam, baclofen, and tizanidine.⁸ Such findings support the CNS origin for the excessive motor unit activity. Central loss of inhibition is also suggested by the exaggerated muscle contractions in response to sound, electrical stimulation, and touch.^8-13 There is no evidence for increased motoneuron excitability; measures of motoneuron pool excitability such as the H-reflex amplitude(Hmax)/maximal muscle compound action potential (Mmax) ratio have been normal,^8,9,14 suggesting that the origin of the enhanced motoneuron activity lies in underactivity of the premotor inhibitory interneurons. Within the spinal cord there are several classes of interneurons that mediate inhibition of motoneurons and are believed to be either GABAergic or glycinergic. We examined several of these inhibitory spinal circuits in patients with SPS to test the hypothesis that the disorder results from selective dysfunction of presumptive GABAergic interneurons.

Methods. Patients. Eleven SPS patients and 14 healthy subjects (age range, 20 to 72 years; mean = 47) participated in studies of spinal reflex circuits. The protocol was approved by the Institutional Review Board and all subjects gave their written informed consent for the study. Patients with a diagnosis of SPS were recruited through advertisements in medical journals and letters to physicians between 1990 and 1997. Patients selected for the study fulfilled the clinical criteria for diagnosis of SPS,¹⁵ did not have other neurologic conditions, and were willing to participate in physiologic studies. Nine patients were studied at the National Institutes of Health, Bethesda, MD, and two patients were studied at the Hospital Clinic, Barcelona, Spain. Clinical criteria consisted of an initial presentation with axial muscle stiffness that progressed to involve proximal leg muscles, a clinical response to benzodiazepines, and a history of superimposed spasms, often precipitated by noise, touch, or anxiety. Electromyography (EMG) was performed to exclude myotonic disorders or neuromyotonia. Evidence of an autoimmune disorder was not required for the clinical diagnosis, but serum was screened for evidence of diabetes, thyroid abnormality, or pernicious anemia and tested for reactivity against GABAergic neurons on tissue sections.¹⁵ Patients with abnormal results on motor and sensory examination (excepting neuropathy attributable to diabetes) or abnormal intellect were excluded. The diagnosis was supported by findings of symmetric spine deformity, typically lordosis, but in some cases loss of the normal spinal curvature. Clinical characteristics of the SPS patients are given in the table. One patient (Patient 10) had symptoms and exhibited signs only in one leg.¹⁶ All patients except Patient 3 were receiving chronic treatment with benzodiazepines. The usual dose of benzodiazepines was withheld for 1 to 2 days beforehand for all but one patient who was unable to tolerate this.

Normal subjects were volunteers from the local community who participated in variable numbers of reflex studies. There was no selection of individual subjects to participate in different reflex studies.

Immunocytochemical studies. Rats were perfused with 10% neutral formalin. Brains were postfixed for 30 minutes at room temperature and cryoprotected in 20% sucrose in phosphate buffered saline (PBS) at 4 °C for 2 days. Twenty-millimeter sections were cut on a cryostat and stored on subbed slides at -70 °C until staining.

SPS patients' sera were diluted 1:1000 in PBS containing 1% goat serum and 0.3% Triton X-100. Slides with rat brain sections were incubated in the serum solution for 2 days at 4 °C, rinsed twice in 0.2% Triton X/PBS, and incubated for 30 minutes with an FITC-labeled goat anti-human secondary antibody (Jackson Immunoresearch, Westgrove, PA) diluted 1:100. Sections were washed twice in 0.2% Triton/PBS, coverslipped with glycerol p-phenylene-diamine, and viewed with a fluorescent microscope. Rat brain sections immunostained with serum of an antibody-positive SPS patient showed heavy labeling of synaptic boutons in the brainstem nuclei, basal ganglia, and cerebellum, matching the pattern seen with an anti-GAD antiserum control.

Physiologic studies. Nerve conduction studies of the upper and lower extremities were performed using surface electrodes and constant current stimulation. Concentric needle electrodes and a Nicolet Viking electromyograph (Madison, WI) were used to sample EMG of leg and paraspinal muscles. In most patients, antagonist muscle pairs were recorded simultaneously in two channels. When motor unit firing was observed despite attempted relaxation, patients were asked to activate antagonist muscles. For studies of spinal reflexes, EMG activity from selected muscles was recorded using paired surface electrodes with a belly tendon configuration. At the National Institutes of Health, we used a Dantec Counterpoint Electromyograph(Allendale, NJ) and in Barcelona we used a Nihon Koden Neuropack-8 electromyograph (Tokyo, Japan) with amplifier filter settings of 20 Hz and 5 kHz. Both machines could deliver paired stimulation of different intensities through one or two constant current outputs. In some studies, an additional second Grass S-11 stimulator (W. Warwick, RI) with a stimulus isolation unit was used. Vibration was applied with three different mechanical systems(Panasonic, Secaucus, NJ; Wahl 4300, Sterling, IL; and Bruel and Kjaer 4810, Worthington, OH) capable of delivering repetitive displacement stimuli at a frequency from 50 to 100 Hz. The displacement motor, stimulators, and electromyograph were triggered using an external pulse generator (AMPI Master 8 or Grass S11, Jerusalem, Israel). Records were sampled at 10 kHz using a Macintosh (Cupertino, CA) computer with LabView software (National Instruments, Austin, TX). Traces were printed and stored individually for later review and averaged online.

H/M ratios. The soleus H-reflex amplitude (Hmax) and maximal muscle compound action potential (Mmax) were measured from peak to peak for averages of 3 to 10 H-reflexes at each stimulus intensity. H-reflexes were expressed as a percentage of Mmax to allow comparisons between subjects.

Vibration-induced inhibition of the soleus H-reflex. The procedure was modified from that of Bour et al.¹⁷ Subjects lay prone and H-reflexes were recorded from the soleus muscle. The tibial nerve was stimulated with 1 ms duration pulses in the popliteal fossa using a monopolar ball electrode with an anode on the patella. Blocks of trials with 12 to 20 H-reflexes were collected, interleaving trials with and without a 10-second vibration of the Archilles tendon. The tibial nerve was stimulated 10 ms after the vibration ended, with 15 to 30 seconds between stimuli. For successive blocks of trials, the stimulus intensity was increased from the minimum intensity needed to evoke an H-reflex to the intensity needed to evoke Mmax. Typically, 6 to 12 intensity levels were used, with at least four stimulus levels between the minimum intensity and the intensity evoking Hmax. EMG activity was monitored over a loudspeaker throughout as a feedback mechanism for relaxation. Tonic vibration reflexes did not occur, but some patients had a low level of tonic muscle contraction in the soleus muscle at all times. Peak-peak amplitudes of the H-reflex and the M waves at each intensity, with and without vibration, were plotted to obtain an amplitude-stimulus intensity curve, normalizing all amplitudes as a percentage of Mmax. A "cumulative vibratory index" (CVI) score was calculated as the ratio of the areas under the curves for vibrated/control H-reflexes up to the intensity evoking Hmax.¹⁸ Trapezoidal approximation was used to calculate area.

Vibration-induced inhibition of FCR H-reflex. We applied electrical stimuli to the median nerve at the cubital fossa to record an H-reflex from the flexor carpi radialis (FCR) muscle. The H-reflex was identified according to the criteria established by Fuhr et al.¹⁹: latency of 16 to 24 ms, enhancement with voluntary activation, elicitation of the response with stimuli submaximal for M waves of the same or greater amplitude, and clear separation from late components of the M wave. H-reflexes of an amplitude of about 200 µV were elicited with single stimuli delivered at intervals of at least 30 seconds. Control H-reflexes and H-reflexes obtained during a 100-Hz vibration of the FCR tendon were randomly alternated. Two or three test stimuli were given during the vibration, which was applied to the FCR tendon for approximately 1 minute. The series was repeated several times to obtain 5 to 12 reflexes in each condition. The mean peak-peak amplitude of the H-reflexes obtained during vibration were expressed as a percentage of control H-reflexes.

Reciprocal inhibition of wrist flexors and extensors. Reciprocal inhibition between forearm antagonist muscles was measured using methods previously described.^19-21 The median nerve was stimulated percutaneously at the elbow, with pulse durations of 0.5 to 1.0 ms, at 1.1 times the threshold for evoking an H-reflex with a 50-µV peak-peak amplitude in the FCR. The radial nerve was stimulated above the elbow with 0.5 to 1.0 ms duration pulses at 1.3 times the threshold for a 100 µV muscle potential in the forearm extensor muscles. For some intervals, stimulus intensities at 0.95 to 1.0 threshold were also tested. FCR H-reflexes were recorded stimulating the median nerve alone (control) or in combination with the radial nerve (conditioning) at intervals from -2 ms to 50 ms to assess the first two phases of reciprocal inhibition. Control and conditioning trials were interleaved with 10 to 20 seconds between trials, averaging 8 to 12 reflexes for each condition at each interval. Peak-to-peak amplitudes were measured. The ratio of conditioned reflexes/control reflexes was compared.

Reciprocal inhibition of knee flexors and extensors.²² The H-reflex of the vastus medialis(VM) muscle was obtained by electrically stimulating the femoral nerve at the inguinal fold with pulses of 1-ms duration. The stimulus intensity was adjusted to produce an H-reflex in the VM 5 to 10% of the M wave. Conditioning volleys were induced in the sciatic nerve with magnetic stimulation applied to the proximal thigh, with the inner edge of the coil touching the sacrum and the current flowing rostro-caudally. In control trials, the femoral nerve was stimulated alone; in test trials, the stimulus of the femoral nerve was paired with the conditioning stimulus at interstimulus intervals ranging from -10 to +10 ms in steps of 1 to 2 ms. Control and test trials were randomly alternated, allowing at least 10 seconds between consecutive stimuli. Mean peak-to-peak amplitude of the H-reflexes obtained in test trials were expressed as a percentage of those of the control trials.

Recurrent inhibition of ankle extensors.²³ To elicit the control soleus H-reflex (termed H1 in this method), the tibial nerve was stimulated in the popliteal fossa with 1 ms duration submaximal stimulation pulses (S1), using monopolar stimulation with the remote anode on the patella. In conditioned trials, the S1 stimulus was followed 13 ms later by a supramaximal stimulus (SM) through the same electrode. This pairing results in a second H-reflex delayed by 13 ms, referred to as H′, which is generated by those motoneurons that fired in response to S1, because collision of orthodromic and antidromic volleys occurs in these motoneurons, rendering them free from antidromic invasion and refractoriness. At low intensities of S1, H1 equals H′, but with higher intensities of S1, recurrent inhibition is more effectively activated, and H1 > H′. Blocks of trials at each intensity of S1 were collected, interleaving 5 to 12 control and conditioning trials with 10 to 20 seconds between trials. The intensity of S1 was progressively increased for successive blocks until H1 attained Hmax. Peak-peak amplitudes of H1 and H′ were measured at each S1. Recurrent inhibition was graded as present (+) if the amplitude of the H′ declined at stimulus intensities where H1 had not yet achieved Hmax.

Nonreciprocal (Ib) inhibition.^24,25 The tibial nerve was stimulated in the popliteal fossa while recording from the soleus muscle and the medial gastrocnemius (MG) muscle. A soleus H-reflex of approximately 10% Mmax was used for the control. Conditioning stimuli were applied to MG nerve at the lower part of the popliteal fossa, using an intensity 0.95 times threshold to obtain a motor response in the MG muscle. The interstimulus interval between tibial and MG stimulation was varied from 2 to 8 ms, and five conditioned and five control H-reflexes were obtained at each interval. Mean peak-peak amplitudes of the H-reflexes obtained during conditioning stimulation were expressed as a percentage of control H-reflexes.

Statistics. Student's t-test was used to compare numeric results between groups of patients and control subjects. To test whether abnormalities in reflex studies were independent of the presence of continuous motor unit firming, a chi-square statistic was computed with a three-dimensional contingency table,²⁶ comparing abnormal results versus motor unit firing versus each interneuron pathway studied.

Results. A summary of the clinical features and findings is shown for individual patients in the table. The mean patient age was 56 years (range 45 to 70), and the mean duration of symptoms was 7 years. All patients but one were receiving chronic treatment with benzodiazepines, and the usual doses are noted in the table. Five patients had diabetes mellitus and were taking insulin. Nerve conduction studies showed evidence of mild sensory neuropathy in two patients. All patients had positive staining for antibodies against GABAergic neurons. EMG showed involuntary motor unit firing at rest in clinically affected muscles in eight patients. In six of these patients, contraction of antagonist muscles failed to induce motor unit relaxation in leg muscles (figure 1; denoted [+] in the table). In four of the six, marked axial rigidity was present. In some patients, continuous motor unit activity in paraspinal muscles at lumbar levels resembled a full interference pattern. In others, regular motor unit activity was recorded at multiple levels. In two other patients, motor unit relaxation was only mildly impaired (±) and could be silenced by contracting the antagonist muscles. In three patients, no motor unit firing was seen at rest (-) at the time of study. Lack of motor unit firing at rest could be attributed to residual effects of long-acting benzodiazepines, as most patients only withheld medication for 1 to 2 days for these studies.

A battery of test was performed, but not all studies could be performed in every patient. Many patients had difficulty tolerating the fairly long testing sessions, which required tolerating the fairly long testing sessions, which Some patients did not have H-reflexes reliable enough for the studies, particularly in the arm. The specific number of patients studied for each test is indicated in the table and reported in the following results.

Hmax/Mmax. The ratio Hmax/Mmax provides a measure of the relative excitability of the motoneuron pool to peripheral nerve stimulation. The soleus Hmax/Mmax was measured in eight SPS patients and was not significantly different from eight normal subjects (SPS = 0.45 ± 0.23[mean ± SD]; control = 0.51 ± 0.18; p = 0.26).

Inhibition by tendon vibration. Vibration-induced inhibition of the soleus H-reflex was studied in eight SPS patients and eight normal control subjects. Vibration of the Achilles tendon caused significantly less inhibition of the soleus H-reflex in the patients than in control subjects. The mean CVI of SPS patients was higher than that of controls (0.91 ± 0.10 for SPS, 0.62 ± 0.2 for controls; p = 0.01, t = 3.79). For individual control subjects, the range of CVI was 0.22 to 0.80. All but one of the SPS patients had a CVI higher than this range, as noted by boldface in the table. Amplitude-intensity plots for three SPS patients and one normal subject are shown (figure 2).

The effects of vibration of the FCR tendon on the FCR H-reflex were tested in four SPS patients. Vibration failed to inhibit the FCR H-reflex in three of them.

Reciprocal inhibition. Reciprocal inhibition between wrist forearm flexor and extensor muscles was tested in seven SPS patients, with mixed results, and in nine normal control subjects. In normal subjects, the FCR H-reflex was inhibited to approximately 50% of control values when conditioning stimulation of the radial nerve was delivered with no delay and with a delay of about 10 ms, with a range from 27 to 76%. There was no difference between the group of patients and normal subjects at either delay(0 ms: t = 0.57, p = 0.28; 10 ms: t = 0.95, p = 0.36), although two SPS patients exceeded the range for control subjects in the first period of inhibition and one in the second period, as noted by boldface in the table. However, in most patients, the second period of inhibition, which is presumed to be produced through presynaptic inhibition of the FCR H-reflex, was normal.

In SPS patients, stiffness in the arms is much less common than stiffness in the legs, particularly stiffness of the proximal leg muscles. For this reason, reciprocal inhibition between knee flexor and extensor muscles was assessed in four SPS patients and seven normal control subjects. In normal subjects, magnetic stimulation of the sciatic nerve inhibited the Vastus medialis (VM) H-reflex to approximately 80% of control, an effect attributed predominantly to presynaptic inhibition of VM Ia afferents.²² The VM H-reflex was reduced to 30 to 80% of control in the four patients studied. These findings suggest that presynaptic contributions to reciprocal inhibition were mostly normal in SPS patients, whether studied between antagonist muscle pairs in the arm or the leg.

Nonreciprocal (Ib) inhibition. Nonreciprocal inhibition between synergist muscles was studied in four control subjects and five SPS patients, with mixed results. In the four control subjects, the amplitude of the soleus H-reflex was inhibited to 54 to 73% of control by electrical stimulation of the medial gastrocnemius nerve at intervals between 4 and 5 ms. As a group, the SPS patients did not differ from control subjects (SPS mean = 77% control reflex, normal subjects 67% control reflex, p = 0.35), although two SPS patients had higher values than any of the control subjects, indicated by boldface in the table.

Recurrent inhibition. Recurrent inhibition of the soleus H-reflex was tested in five SPS patients and six control subjects and was present (normal) in all. In normal subjects and patients, the amplitude of H′ and H1 initially increased with increasing stimuli, and H′ subsequently declined; H1 continued to increase until Hmax.

Motor unit activity effects on spinal reflex circuit studies. The severity of muscle stiffness and ability to relax motor unit firing in the limbs differed among patients. Of the 11 patients, six were unable to silence motor units by contraction of antagonist muscles, whereas three had no difficulty with relaxation at the time of study. We presume that these differences reflect disease severity and the effectiveness of the benzodiazepine treatment in each patient. Abnormalities in vibration-induced inhibition and other reflex studies were seen in patients regardless of the degree of motor unit firing at rest. Abnormal results did not significantly depend on motor unit firing at rest (χ² = 13.5;p > 0.05, n = 27).

Discussion. We tested inhibitory spinal circuits in patients with SPS to elucidate whether the clinical symptoms of muscle stiffness correspond to a selective dysfunction of GABAergic neurons in the spinal cord. The working hypothesis was that spinal inhibition mediated by GABAergic interneurons would be abnormal, and that inhibition mediated by glycinergic neurons would be spared. The results of our study are mixed. At least one inhibitory circuit, vibration-induced inhibition, was fairly consistently abnormal in SPS patients. This inhibition has commonly been attributed to presynaptic inhibition through GABAergic neurons in humans,²⁷ although the true mechanism may be considerably more complex, as discussed in the following. Another inhibitory circuit, reciprocal inhibition, is also presumed to be presynaptic in part, with specific components produced by GABAergic neurons.^20,28 The presynaptic component of reciprocal inhibition was normal in most SPS patients. This was true for measures of reciprocal inhibition between forearm muscles, which were not usually affected in patients, as well as between the proximal leg muscles, which were usually clinically stiff. Fluctuating co-contraction of antagonist muscles is a common finding in SPS patients, and was apparent during needle EMG in half of the patients examined. Reciprocal inhibition represents a series of reflex mechanisms that hamper co-contraction, and it is paradoxical that it was mostly normal in SPS patients.

Loss of vibration-induced inhibition of the H-reflex was the most consistent abnormality in our patients. Several mechanisms are likely to contribute to this inhibition, which is typically produced by vibrating the tendon of a muscle for several seconds to minutes. Because vibration activates muscle spindles, motoneurons receive a sustained barrage of Ia afferent input.^29,30 In the technique used here, the effects of vibration on motoneurons were assessed using a reflex elicited by electrical stimulation of the same Ia inputs after a slight delay to avoid axonal refractoriness. Part of the decline of the H-reflex could occur through the action of Ia afferents on inhibitory interneurons that cause postsynaptic inhibition of motoneurons, including the Ib inhibitory interneurons that mediate nonreciprocal inhibition.³¹ However, because motoneurons remain excitable to cutaneous inputs during vibration,³² postsynaptic inhibition cannot be the predominant cause of the vibration-induced inhibition. There are two possible presynaptic mechanisms: "classic presynaptic inhibition," mediated by interneurons that form axo-axonic synapses on presynaptic terminals; and the presynaptic phenomenon of "postactivation depression."³³ Classic presynaptic inhibition, produced by interneurons activated by antagonist Ia and Ib afferents, is mediated by last-order GABAergic interneurons in animals.³⁴ Vibration of the Achilles tendon would activate classic presynaptic inhibition through the spread of vibration to antagonist muscles.³⁵ Postactivation depression follows activation of spindle afferents through any mechanism-passive stretch, tendon tap, or a preceding H-reflex-depressing the amplitude of an H-reflex for as long as 10 to 15 seconds afterward. The duration of postactivation is much longer than classic presynaptic inhibition in humans^36,37 and is thought to be produced by mechanisms intrinsic to presynaptic terminals rather than through GABAergic interneurons.³⁸ Vibration is likely to produce postactivation depression as well as activating classic presynaptic inhibition. The deficit observed in vibration-induced inhibition in SPS patients could reflect the loss of contributions from both mechanisms.

The differing findings in the two presumptive GABAergic inhibitory reflexes could suggest that not all spinal GABAergic neurons are uniformly affected. Differences between patients could suggest that SPS may have heterogeneous pathophysiologic mechanisms. Clinical variants of SPS, for example involving legs alone,^16,39 may represent different examples of this heterogeneity. Clinical heterogeneity could have its foundation in the autoimmune response in a number of different ways. First, antibodies may not be directed against the same target in all patients. Antibodies directed against GAD are associated with some, but not all, of the clinical variants. The clinical syndrome of SPS has also been reported in patients with antibodies against other central nervous system antigens. In some patients, the anatomic distributions of the antigen were similar to GAD,⁴⁰ but other antigens (for example, amphiphysin, a synaptic vesicle protein) may have somewhat broader distributions.^41,42 Furthermore, the antigens targeted may not be uniformly expressed in all classes of GABAergic neurons. The proportions of two isoforms of GAD, GAD65 and GAD67, vary somewhat in GABAergic neurons and have a different cellular localization.⁷ The isoform GAD65 predominates at synaptic terminals, and most SPS patients have antibodies against a specific region of the N-terminal domain of GAD65 that confers the ability to associate with the cytoplasmic surface of the vesicular membrane.^7,43 The accessibility of antigen to extracellular antibodies presents an additional variable, as GAD65 has a limited exposure on the cell surface.⁴⁴ Finally, antibodies in different patients may exert their pathophysiologic effects in different ways: for example, antibodies to GAD65 blocked GAD activity in only two of five SPS patients in whom it was tested.⁴⁵ These features-variability in antibodies produced by patients, differential expression of GAD65 in different neurons, and variable ability of antibodies to block GAD activity-could all contribute to heterogeneity in GABAergic dysfunction among different cell populations and in different patients. Antibodies against GAD were recently reported in three patients with cerebellar ataxia and polyendocrine abnormalities without features of SPS,⁴⁶ illustrating the need for a better understanding of how a similar autoimmune response targets different populations of GABAergic cells. These details were not explored in our patients, but may provide further insights into pathophysiologic mechanisms in future studies.

SPS is a disorder with fluctuating symptoms, and some variability in inhibitory circuits could reflect the clinical status of patients at the time when different inhibitory circuits were tested. Each reflex study typically takes several hours to complete and most of our patients were tested in several sessions over a period of several days. Most patients were receiving chronic treatment with benzodiazepines, and differences in residual medication levels may have contributed to variability. Benzodiazepines are effective for reducing involuntary motor unit firing in SPS.^2,8 Although this effect alone did not account for whether a patient exhibited abnormalities in reflex studies, we cannot exclude the possibility that benzodiazepines may have affected some inhibitory reflex circuits, for example reciprocal inhibition, more than others. Unfortunately, the clinical condition of patients in our study did not allow them to be completely taken off medications for physiologic studies. In future studies, we plan to follow changes in abnormal inhibitory circuits during therapy, and thereby gain a better understanding of how these correlate with the severity of clinical symptoms.

Recurrent inhibition, nonreciprocal (Ib) inhibition, and the first period of reciprocal inhibition have all been attributed to postsynaptic inhibition of motoneurons through glycinergic neurons.^34,47,48 We did not find a consistent pattern of abnormalities for these three inhibitory reflex circuits in SPS patients as a group, but some individual patients exhibited less inhibition than normal controls. In two patients, nonreciprocal inhibition was diminished, and in two other patients, the first period of reciprocal inhibition was diminished. These scattered abnormalities in presumptive glycinergic pathways could indicate that the abnormalities in SPS extend beyond GABAergic pathways. One mechanism by which this could occur would be expansion of the initial antibody response against GAD to other surface components of GABAergic neurons or a switch from humoral to cellular-mediated immune responses, as has been suggested for the immune response against GAD in pancreatic islet cells in diabetes.^7,49 Co-expression of GABA and glycine has been reported in some spinal interneurons⁵⁰ and such cells could provide a source of common antigen. Another explanation may be that the autoimmune injury in SPS affects GABAergic neurons in supraspinal structures, leading to loss of descending inhibitory controls on a number of pharmacologically diverse spinal interneuron circuits. It is also unknown whether supraspinal structures develop mechanisms to compensate for losses in spinal inhibition.

Changes in spinal inhibitory circuits have been observed in a number of disorders characterized by muscle co-contraction, including dystonia, spasticity, and hereditary hyperekplexia (startle disease). An abnormality in a single spinal circuit is not diagnostic for a particular disease, although the pattern of abnormalities in spinal reflex circuits is somewhat different in each of these disorders. In dystonia, vibration-induced inhibition may also be decreased, but is quite variable¹⁷; the second (presynaptic) component of reciprocal inhibition is diminished^21,51; the first period of reciprocal inhibition is most often normal.⁵¹ In spasticity, vibration-induced inhibition^17,52 and the first period of reciprocal inhibition are diminished.^53,54 Findings are mixed in spasticity for recurrent⁵⁵ and nonreciprocal (Ib) inhibition^56,57 and for the second period of reciprocal inhibition, with differences between hemispheric stroke and paraplegia, and between the arm and the leg.^53,58 Loss of vibration-induced inhibition, therefore, may be seen in a patient with SPS, spasticity, or dystonia. In contrast, vibration-induced inhibition is normal in patients with hyperekplexia, whose muscle stiffness is episodic rather than sustained as in the other disorders. In hyperekplexia, the first period of reciprocal inhibition is diminished, as in spasticity, and the second component of reciprocal inhibition is normal.⁵⁹ Hereditary hyperekplexia results from a mutation in a subunit of the glycine receptor,⁶⁰ and it is interesting that in these patients, as in SPS patients, recurrent inhibition was normal. These findings add support to the suggestion that Renshaw inhibition is mediated by populations of GABAergic and glycinergic neurons.⁶¹

Acknowledgments

We thank Quentis Scott for coordinating patient visits and Joy Kopyto for assistance with the manuscript.