Decreased cerebrospinal fluid concentrations of substance P in treatment-resistant depression and lack of alteration after acute adjunct vagus nerve stimulation therapy☆

1. Introduction 

Substance P (SP), an undecapeptide, was first discovered in crude form in 1931 by von Euler and Gaddum (von Euler and Gaddum, 1931), but its study was limited until its isolation by Leeman and colleagues in 1971 (Hokfelt et al., 2001). SP is found heterogeneously distributed in both the CNS and the peripheral nervous system (PNS). In the periphery, SP is located in the primary sensory neurons as well as in neurons within the gastrointestinal tract (Pernow, 1953).

SP belongs to the tachykinin family, which also includes neurokinin A and neurokinin B. Tachykinins originate from the preprotachykinin I (PPT-I) and preprotachykinin II (PPT-II) genes (Stout et al., 2001). Upon its release, SP binds to a family of neurokinin (NK) receptors, preferentially acting on the metabotropic NK1 receptor (Harrison and Geppetti, 2001).

Preclinical and clinical studies have suggested a potential antidepressant and anxiolytic effect of NK1 receptor antagonism (De Felipe et al., 1998, Bondy et al., 2003). For this reason, considerable attention has been given to the pharmacological development of SP antagonists (McLean, 2005). The rationale for this investment is multifold. One key set of preclinical research findings indicate that SP is released in response to stressful stimuli (Culman and Unger, 1995, Ebner et al., 2004). Further, SP-mediated activation of the NK1 receptor produces behavioral changes suggestive of “emotion modulation” in experimental animals (Elliott, 1988, Aguiar and Brandao, 1996, Teixeira et al., 1996). A second line of evidence supporting a role for SP in depression comes from data demonstrating that SP is co-localized within serotonergic and noradrenergic pathways (Santarelli and Saxe, 2003), and is found in regions that are involved in the regulation of mood and emotion, such as the amygdala, raphe nuclei, locus coeruleus, hypothalamus (Arai and Emson, 1986), limbic areas (Mantyh et al., 1984) and periaqueductal grey (Aguiar and Brandao, 1996). Thirdly, a growing body of results from preclinical and clinical investigations suggests that SP and the NK1 receptor are intimately linked with a diverse array of biological markers and neurotransmitter system abnormalities that have traditionally been associated with mood and anxiety disorders (Commons et al., 2003, Hwang et al., 2005). For example, behavioral and physiological effects produced by genetic or pharmacologic inactivation of the NK1 receptor in laboratory animals suggest anxiolytic- and antidepressant-like effects of SP antagonists (De Felipe et al., 1998, Ebner et al., 2004), which may be partially mediated by the alteration of neurofilaments and synaptic remodeling (Guest et al., 2004). NK1 antagonists have also been shown to increase the firing rate of serotonergic (Haddjeri and Blier, 2001, Santarelli et al., 2001), noradrenergic (Millan et al., 2001), and dopaminergic (Lejeune et al., 2002) neurons. Finally, SP circuits have long been implicated in nocioception, and recently in controlled studies the comorbidity of major depression and somatic pain (e.g., back pain, headache, abdominal pain, pelvic pain) has been unequivocally demonstrated (Ohayon and Schatzberg, 2003, Ohayon, 2004).

The measurement of CSF SP in humans has been a relatively neglected area. Rimon and colleagues (Rimon et al., 1984) reported that depressed patients had a fourfold increase in basal CSF SP-like immunoreactivity relative to patients with schizophrenia and healthy controls. However, two research groups (Berrettini et al., 1985, Deuschle et al., 2005) did not replicate this result. Our group recently reported elevated basal CSF SP concentrations in medication-free patients with major depressive disorder (MDD) relative to healthy control subjects (Geracioti et al., 2006). In the same report we also described elevated basal CSF SP in a group of medication-free adults with post-traumatic stress disorder (PTSD) and dynamic changes in CSF SP concentrations in response to acute psychological stress during serial sampling of CSF via indwelling lumbar intrathecal catheter.

Several groups have examined the effects of antidepressant treatment on concentrations of SP. In rodents, chronic administration of mianserin or one of several tricyclic antidepressants reduced SP content in the striatum, substantia nigra, and amygdala (Shirayama et al., 1996). However, in a study of humans, Deuschle et al. (2005) reported that pharmacological treatment of depressed patients with paroxetine or amitriptyline did not significantly alter CSF or plasma SP concentrations. Similarly, Martensson et al. (1989) found that successful treatment of depression with fluoxetine did not alter CSF SP levels. Both of those studies were limited by small numbers of subjects (n=9 and n=11, respectively). Two groups have described a relationship between clinical outcomes and serum SP concentrations. Bondy et al. (2003) noted a significant correlation between depressive symptoms scores and serum SP levels following antidepressant treatment. Patients who demonstrated a decrease from high baseline serum SP concentrations with treatment were more likely to experience a positive clinical response than those patients whose serum SP concentrations increased from their relatively low baseline values (Bondy et al., 2003). Lieb et al. (2004) also reported that antidepressant responders were characterized by high baseline serum SP concentrations that declined significantly with pharmacotherapy. Deuschle et al. (2005) did not confirm a difference in serum SP concentrations between antidepressant responders and nonresponders. It should be noted, however, that there are no data to suggest that serum SP concentrations reflect relevant CNS concentrations of the peptide (Landgraf et al., 1983, Freed et al., 2002).

In this report, we first compared basal CSF SP concentrations in a group of depressed patients on stable psychotropic medication regimens with those of a matched group of healthy, unmedicated controls. Based on previous work by our group and others, we hypothesized that CSF SP concentrations would be elevated in patients with treatment-resistant depression relative to healthy controls. Next we evaluated the effect of 10–12 weeks of adjunctive treatment with cervical vagus nerve stimulation (VNS) therapy on CSF concentrations of SP in the group of depressed patients. The vagus nerve projects to brain regions involved in mood regulation and stress reactivity, including the locus coeruleus, and raphe nuclei, hypothalamus, and amygdala (Henry, 2002). As mentioned above, SP is also present in these serotonergic, noradrenergic, and limbic brain regions. As the mechanism of action of VNS is unknown, we reasoned that evaluation of the effects of VNS on CSF SP might help clarify the effects of this novel treatment.

2. Methods 

Twenty-one adults with recurrent or chronic treatment-resistant depression participated in a protocol which included standardized lumbar puncture (LP) for collection of 12 cc of CSF on three separate but identical procedure days (Carpenter et al., 2004) during participation in the VNS “D02” multi-center clinical trial (George et al., 2005, Rush et al., 2005) (Fig. 1). Detailed demographic and clinical characteristics of these subjects, including a table listing each subject's regimen of stable psychotropic medications, are described in detail elsewhere (Carpenter et al., 2004). Briefly, participants were required to present with a recurrent or chronic major depressive episode that had failed to respond to at least two but not more than six adequate trials of standard antidepressant treatment. A score of 20 or higher on the 24-item Hamilton Rating Scale for Depression (Hamilton, 1960) was required. All patients were on stable doses of psychotropic medications for at least 4 weeks prior to participation. Eighteen participants were diagnosed with unipolar major depression and three had bipolar depression. Data were not obtained for systematic diagnosis of Axis I or II psychiatric comorbidities. CSF specimens were successfully obtained at baseline (immediately before initiating a course of adjunct VNS) in 19 of the depressed subjects; 18 had CSF from both a pre- and post-VNS time point. A separate group of healthy control subjects, free of current or past psychiatric illness and on no medications, underwent a single, similarly standardized LP procedure, but did not subsequently enter any treatment condition or have further CSF sampling. From the available pool of healthy subjects, 19 were selected to match sex and age (as closely as possible) with the depressed group.

View Large Image Download to PowerPoint

Fig. 1. Lumbar punctures for cerebrospinal fluid (CSF) collection in relation to vagus nerve stimulation (VNS) D-02 study timeline. LP=lumbar puncture; B=pre-stimulation baseline; E=post-stimulation endpoint (following 10–12 weeks active VNS); EE=extended endpoint, following 22 weeks active stimulation. Analysis of effects of VNS on CSF substance P concentrations for 18 subjects in this report included specimens from B and E from all individuals, regardless of initial assignment to active or sham conditions. Assessment of depressive symptoms at B and E took place on concurrent LP procedure days, prior to CSF collection.

All subjects provided written informed consent on forms approved by the designated Institutional Review Board (IRB) and were remunerated for participating in the study. Subjects were required to be free of clinically significant medical disorders, based on history, physical examination, and laboratory studies, which included a complete blood count, serum chemistries, liver and thyroid function tests, urine toxicology, electrocardiogram, and pregnancy test for women. Control subjects were medication-free for at least 2 weeks by the time of participation, and depressed patients were not permitted to have psychotropic medication changes (unless medically necessary) for 4 weeks prior to the first LP or during the 24 weeks of the VNS D02 trial (Rush et al., 2005).

Participating subjects agreed to comply with instructions for a modified diet and activity schedule in the 24 h preceding the LP. Routine measures were taken to reduce anxiety and arousal associated with the LP procedure (Carpenter et al., 2004). A total of 12 cc of CSF was collected from the L3–L4 or L4–L5 interspace with a fine-gauge spinal needle during each LP. All LP procedures were standardized to yield CSF samples at 12:00 noon±15 min. Data for nine (50%) of the subjects in this analysis had pre- and post-VNS CSF data obtained from LPs performed at week 2 and week 12, respectively, during the “Acute Phase” portion of the blinded study (subjects randomized to receive active VNS; Fig. 1). The other half had pre- and post-VNS CSF data obtained from CSF collected at week 12 and week 24 (Cross-over Phase from sham stimulation to open-label VNS; Fig. 1). Collected CSF was immediately separated into aliquots and frozen at −80 °C until assay. For depressed patients, the Hamilton Rating Scale for Depression (HAM-D) 24-item version (Hamilton, 1960) was used to measure depression severity. Semi-structured clinical interviews were used to determine absence of major depression or any other Axis I psychiatric disorder in control subjects, but HAM-D ratings were not obtained.

CSF SP immunoreactivity was determined in blinded CSF samples by solid phase radioimmunoassay (RIA) using a highly specific SP antibody. Details of the assay procedure are described elsewhere (Geracioti et al., 2006). The sensitivity of the assay was 1.5 fmol/ml. Intra-assay variability was 8.2% and inter-assay variability was 9.7%.

Descriptive statistics were used to characterize the two subject groups. Student's t-tests were used to compare mean basal CSF SP concentrations in depressed and control subjects and Pearson correlation coefficients were generated to examine the associations between CSF SP, age, and depression severity (among depressed subjects). To examine the effects of subsequent adjunctive VNS on CSF SP concentrations within the 18 depressed patients, baseline (pre-VNS) and endpoint (post-VNS) CSF SP concentrations were examined with paired t-tests (within-subjects comparison of baseline and week 12 CSF SP). Pearson correlations were generated for a matrix including baseline-to-endpoint change in CSF SP during VNS, baseline depression severity, endpoint depression severity, and change in depressive symptoms. All tests were two-tailed with p-value set at 0.05 for significance.

3. Results 

The clinical characteristics of the two main subject groups are described in Table 1. As expected with matching procedures, depressed and control subjects did not significantly differ in sex composition (57% female in both groups) or in mean±S.D. age (47.8±8.8 and 43.4±9.0 years, respectively). The depressed group was characterized by a HAMD-24 mean±S.D. total score of 26.7±7.2, reflecting a high degree of depressive symptomatology despite psychotropic medication treatment. As seen in Fig. 2, CSF SP concentrations were significantly lower in patients with treatment-resistant depression as compared with controls (26.4±21.8 versus 49.8±26.2 fmol/ml, t=3.0, df=36, P=0.005; Fig. 2). No significant correlations were detected between CSF SP and age or HAMD-24 score. Baseline characteristics for the 18 depressed subjects who had pre- and post-VNS CSF samples available were similar to those reported for the entire group of 19 (which included a nonbipolar, depressed female subject 59 years of age, with a baseline HAMD-24 score of 35, who did not have a post-VNS CSF sample).

Table 1. Clinical characteristics of subject groups

	Subjects	Healthy control (n=19)	Treatment-resistant major depressive episode (n=19)
	Age (yrs, mean±S.D.)	43.3±9.0	47.9±8.8
	Sex (n, % female)	11; 58%	11; 58%
	Psychotropic medication use	None	Multiple; chronic; stable doses
	CSF SP (fmol/ml, mean±S.D.)	49.8±26.2	26.4±21.8
	HAMD-24 score (mean±S.D.)	NA	26.7±7.2

View Large Image Download to PowerPoint

Fig. 2. Basal (mean and S.D.) CSF concentrations of substance P in patients with treatment-resistant depression and matched healthy controls.

Mean change in HAMD-24 scores corresponding with the pre-stimulation baseline and post-stimulation endpoint (26.3±7.1 and 21.9±12.5) did not reflect significant depressive symptom improvement associated with VNS (t=1.6, df=17, P=0.14) for the group; eight (44%) experienced 30% or greater decrease (improvement) from baseline HAMD-24 score, four (22%) experienced 30% or greater increase (symptom worsening), and the remaining six (33%) had little (less than 30%) or no change from baseline HAMD-24 score. No significant change was noted in CSF SP concentrations from collection before (27.3±22.0 fmol/ml) and after (34.3±33.8 fmol/ml) a 10–12 week course of adjunct VNS therapy (t=0.98, df=17, P=0.34; Fig. 3). Clinical response, calculated both as absolute change and as percent change in HAMD-24 score, was not significantly correlated with baseline-to-endpoint change in CSF SP. Post-hoc application of several different categorical response and remission criteria also did not produce significant relationships with baseline, endpoint, or change value for CSF SP.

View Large Image Download to PowerPoint

Fig. 3. CSF substance P concentrations in 18 patients with treatment-resistant depression before and after receiving 10–12 weeks treatment with adjunctive vagus nerve stimulation (VNS). The difference between pre-VNS baseline and post-VNS endpoint is not statistically significant (mean±S.D. 27.3±22.0 versus 34.3±33.8 fmol/ml; t=0.98, P=0.34).

4. Discussion 

The main finding of this study was a difference in basal CSF SP concentrations between the depressed and control groups in the opposite direction to that which was hypothesized. One possibility for this finding is that the significantly lower CSF concentrations of SP we observed in the depressed group are simply due to the effects of chronic antidepressant or other psychotropic medication administration. However, two published reports showing no significant CSF SP change during antidepressant treatment (Martensson et al., 1989, Deuschle et al., 2005) provide support for the notion that the relatively lower CSF SP levels we observed in depressed subjects are not necessarily secondary to pharmacotherapy. Nevertheless, the depressed patients we studied were typically taking high but stable doses of multiple psychotropic medications for months or years prior to participating in the study, and the effects of chronic, high-dose pharmacotherapy on CSF SP concentrations may be fundamentally different from those measured in the context of short-term clinical trials with a single antidepressant drug. The data are therefore suggestive, but not conclusive, regarding the possibility that decreased CSF SP may reflect neurobiological changes related to the more severe and refractory illness in this population of depressed patients.

One could speculate that the association we found between low CSF SP concentrations and poor prior response to antidepressant therapies is akin to the finding of Lieb et al. (2004) that individuals with low baseline serum SP levels were less likely than their high baseline serum SP counterparts to experience clinical benefits from paroxetine. While a “basement effect” may have precluded the opportunity to examine such a relationship in our depressed population, we did not find any significant association between magnitude of CSF SP baseline concentration and clinical response to a 10–12 week course of adjunct VNS. Overall response rate was not robust in the subgroup of 18 subjects we examined (i.e, 44% met a criterion of 30% or greater symptom reduction from baseline, with no significant change in mean HAMD-24 for the group as a whole), and our within-subjects analysis did not show any significant alteration of CSF SP associated with VNS therapy. A lack of VNS-induced change in CSF SP concentrations is consistent with the negative findings described in two other published studies of antidepressant medication and CSF SP (Martensson et al., 1989, Deuschle et al., 2005), but study of a larger sample with more positive responders could address this issue more definitively.

The present study was limited by a fairly small number of patients and by the concurrent use of multiple psychotropic medications within an investigation aimed at measuring basal CSF SP concentrations and the effects of a specific antidepressant (albeit adjunctive) therapy. The extent to which each LP procedure induced subjective or objective psychological or physiological stress was not systematically recorded, but we have recently described evidence of acute increases (up to 169%) in CSF SP induced by strong emotional arousal in PTSD patients (Geracioti et al., 2006). Preclinical research suggests that SP is involved in terminating the hypothalamus–pituitary–adrenal (HPA) axis stress response (Jessop et al., 2000) and moreover that NK1 receptors influence limbic sources of corticotropin-releasing hormone (CRH) by modulating the activity of CRH/serotonin neurons in the dorsomedial dorsal raphe nucleus (Commons et al., 2003). A recent report described a study of HPA axis function in patients with chronic, treatment-resistant depression who underwent 12 weeks of VNS treatment (O'Keane et al., 2005). In comparison to matched controls, the depressed patients had elevated baseline adrenocorticotropin (ACTH) responses to CRH stimulation which normalized after VNS treatment. Changes in ACTH concentrations were highly correlated with reductions in symptoms of atypical depression that occurred following VNS treatment (O'Keane et al., 2005). In the present study, it is possible that the decreased concentrations of CSF SP were associated with HPA axis dysregulation related to chronic, treatment-resistant depression. In addition, it is possible that effects of concurrent psychotropic medications on HPA regulation or acute HPA axis responses of the subjects during CSF collection procedures may have contributed to our results in ways which are still poorly understood. However, our finding that depressed patients who have failed to experience substantial benefit from multiple antidepressant therapies are characterized by relatively low basal lumbar CSF SP concentrations, taken together with other emerging data regarding correlates of chronic, treatment-resistant depression, may suggest a novel avenue for increasing our understanding of the complex pathogenesis of this condition.