Striatal and Thalamic Auditory Response During Deep Brain Stimulation for Essential Tremor: Implications for Psychosis

Abstract

Introduction:

The P50, a positive auditory-evoked potential occurring 50 msec after an auditory click, has been characterized extensively with electroencephalography (EEG) to detect aberrant auditory electrophysiology in disorders like schizophrenia (SZ) where 61–74% have an auditory gating deficit. The P50 response occurs in primary auditory cortex and several thalamocortical regions. In rodents, the gated P50 response has been identified in the reticular thalamic nucleus (RT)—a deep brain structure traversed during deep brain stimulation (DBS) targeting of the ventral intermediate nucleus (VIM) of the thalamus to treat essential tremor (ET) allowing for interspecies comparison. The goal was to utilize the unique opportunity provided by DBS surgery for ET to map the P50 response in multiple deep brain structures in order to determine the utility of intraoperative P50 detection for facilitating DBS targeting of auditory responsive subterritories.

Materials and Methods:

We developed a method to assess P50 response intraoperatively with local field potentials (LFP) using microelectrode recording during routine clinical electrophysiologic mapping for awake DBS surgery in seven ET patients. Recording sites were mapped into a common stereotactic space.

Results:

Forty significant P50 responses of 155 recordings mapped to the ventral thalamus, RT and CN head/body interface at similar rates of 22.7–26.7%. P50 response exhibited anatomic specificity based on distinct positions of centroids of positive and negative responses within brain regions and the fact that P50 response was not identified in the recordings from either the internal capsule or the dorsal thalamus.

Conclusions:

Detection of P50 response intraoperatively may guide DBS targeting RT and subterritories within CN head/body interface—DBS targets with the potential to treat psychosis and shown to modulate schizophrenia-like aberrancies in mouse models.

INTRODUCTION

Auditory evoked potentials (AEPs) are fundamental to understanding auditory electrophysiology and mechanisms of gating, a filtering process that attenuates response to redundant sensory stimuli that may prevent sensory overload. A classic mid-latency AEP, the “P50,” is the first large positive wave, P1, with a latency of 50 ms following a click. P50 is typically measured with electroencephalogram (EEG) (1,2). In the auditory gating paradigm, P50 responses are measured to identical, paired auditory click stimuli typically presented 500 ms apart. The magnitude of auditory gating is determined by calculating the P50 ratio = $\frac{\text{P50 test response, second click, S2}}{\text{P50 conditioning response, first click, S1}}$. Normal gating in healthy controls is over 50% (P50 ratio<0.5). A few reports identify auditory gating deficits, defined as a P50 ratio ≥ 0.5, in bipolar disorder, autism spectrum disorder, Alzheimer’s disease, and movement disorders like Parkinson’s and Huntington’s disease warranting further investigation (3-9). Auditory gating deficits have been most extensively investigated in schizophrenia (SZ), where patients have self-reported deficits in sensory filtering and experience sensory overload (10-12). The auditory evoked P50 gating deficit is found in 61–74% of SZ patients compared to only 20% of healthy subjects without mental illness (2,13-17).

Several animal and human studies show that P50 originates in the primary auditory cortex and that the medial geniculate mediates thalamocortical responses in parallel (1,18,19). Consistent with parallel cortical sensorimotor integration, robust P50 response has been mapped to several cortical structures including primary motor, supplementary sensorimotor, and supplementary without motor response using electrocorticography (ECoG) localized with both imaging and functional mapping (20). In addition, gated response has been clearly identified with single unit activity recordings from reticular thalamic nucleus (RT) neurons in rodents (21).

DBS surgery provides a unique opportunity to further characterize the auditory physiology in humans using microelectrode recording (MER) conducted during DBS surgery in patients with essential tremor (ET) to identify electrophysiological signatures of deep brain structures to inform optimal DBS electrode placement. For this research investigation, MER was also used to precisely localize P50 response in deep brain structures, like RT, traversed along the trajectory to the ventral intermedius (VIM), the motor nucleus of the thalamus targeted. MER allows sampling of thousands of neurons via local field potentials (LFP), in a range similar to EEG.

DBS is a safe and effective therapeutic option for treatment-refractory ET. ET characterized by kinetic tremor is found at increased incidence with age (22). ET is a motor circuit disorder within the ventral intermediate thalamus (VIM)-motor cortex-cerebellum circuit (23). Therapeutic DBS targeting VIM works through neuromodulation of low frequency (4–7 Hz), high amplitude synchronous oscillatory activity within the VIM, the inferior olive, cerebellum, and motor cortex (24). Following the widespread successful DBS therapy demonstrated in both Parkinson’s disease (PD), and ET, DBS is being investigated as a therapeutic option for new indications including treatment-refractory SZ (24-29). The DBS therapy requires identification of disease relevant: 1) circuitry, 2) targets, and 3) electrophysiological biomarkers to guide implantation. Like PD, and ET, SZ is a circuit disorder with multiple independent electrophysiological and neuroimaging observations of aberrancy noted in the cortico-basal ganglia-thalamo-cortical (CBGTC) circuit. Although evidence remains very preliminary, several prospective DBS targets for SZ have been identified including the head/body interface of the caudate (CN) and reticular thalamic nucleus (RT) both investigated here with an electrophysiological biomarker, the auditory evoked P50 response (30-40).

Primary to implicating the CBGTC circuit in SZ is the fact that all of the many antipsychotics developed since the discovery of chlorpromazine share the same key psychosis-specific mechanism— modulation of neuronal signaling in the striatum—through antagonistic action at the dopamine D₂ receptor (41-45). The striatum, including the CN head/body interface investigated here, has been independently implicated in SZ through multiple large imaging studies showing a 14% increase in presynaptic dopamine, increased glutamate, and decreased activation during tasks requiring coordination with the frontal cortex (31-33,37,39). Central within the implicated SZ circuitry, the CN head/body interface has the potential to treat psychosis through neuromodulation of 1) the striato-nigro-striatal ascending spiral involved in striatal dopamine regulation; 2) the dorsal lateral prefrontal cortex (DLPFC); and 3) the auditory cortex where spontaneous activity is associated with auditory hallucinations (46-48).

Finally, RT has an essential role in filtering sensory information to suppress distraction and focus attention through neuromodulation of thalamo-cortical processes that are aberrant in schizophrenia (49,50). DBS of RT modulated aberrant oscillatory activity associated with positive, negative, and cognitive symptoms in the neonatal ventral hippocampal lesion (NVHL), a rat model of schizophrenia (51). RT is a thin layer surrounding the thalamus requiring intraoperative biomarkers to facilitate DBS targeting. Conversely, the CN is large necessitating intraoperative biomarkers to define subterritories as has been done electrophysiologically for the thalamus to distinguish dorsal, ventral, lateral motor (ventro oral posterior/anterior [VOP/A] and VIM), and somatosensory (ventrocaudalis internus [VC]), thalamic nuclei to guide VIM localization for DBS treatment of ET (52).

This investigation is the first demonstration of P50 response within deep brain structures traversed along the trajectory to the VIM target. P50 response measured with LFP was recorded from CN head/body interface, internal capsule, RT, dorsal and ventral thalamus during DBS surgery. P50 response was chosen for this investigation for multiple reasons including: 1) the P50 paradigm was conducive to being recorded intraoperatively; 2) P50 response has been found in RT in rodents, and may assist in DBS targeting RT in humans if present (21,51); and 3) overlap between P50 circuitry and SZ circuitry may be relevant to treating SZ with DBS. We developed a novel method to allow intraoperative recording of P50 response in humans during awake DBS surgery. In this study, P50 response is the key element being assessed and intraoperative detection in subjects during awake neurosurgery precludes reliable assessment of gating of the P50 response, which is characterized during a relaxed state. Results from this investigation expand our understanding of how P50 response is relayed through different brain structures and is the first step in assessing P50 as an intraoperative biomarker to guide DBS targeting in humans.

RESULTS

Characteristics of Significant Conditioning P50 Responses Across Deep Brain Structures

Using the P50 paradigm, 177 LFP recordings were obtained during 30 pauses along trajectories to the VIM in seven subjects ranging from 12 to 30 recordings/subject (Table 2). Of the 177 recordings, 40 conditioning P50 responses were identified as being significantly more positive than background and 137 of the nonsignificant (NS) recordings were considered background including 115 where nonsignificant peaks were noted and 22 where no peaks could be identified. The 40 P50 responses identified with LFP were not followed by the higher amplitude N100 response sometimes characteristic of scalp EEG recordings, which are not specific to auditory responses (Fig. 1) (53). The P50 responses were found at similar frequencies in CN (22.7%), RT (26.7. %), and VTH (26.5%) compared to both ALIC and DTH where no P50 response was found (Fig. 2a). The higher frequency of P50 responses in VTH reached significance only when compared with ALIC and DTH (Fig. 2a). In addition, overall peak heights were significantly lower in ALIC (2.0 ± 5.0 μV), a white matter tract, compared to CN (7.19 ± 4.8 μV), RT (6.9 ± 5.3 μV), and VTH (7.2 ± 5.2 μV; Fig. 2B).

Table 2.

Significant Conditioning P50 Response to Paired Auditory Stimuli From LFP (10–100 Hz; n = 40).

Structure/Electrode	Tip dist. (mm)	Epochs	Cluster	DBS con	CSD	Conditionng		Test		P50 ratio
						peak μV (SD)	Lat (ms)	peak (μV)	Lat (ms)
Subject 1
CN+ m1	24.7	49			–	1.5 (6.1)	55.6	0.5	50.9	0.34
CN + μ1	23.7	37	1		–	23.3 (47.3)	40.4	0	49.5	0
CN + μ2	23.7	36	1		–	19.5(48.8)	41.1	0	53.1	0
CN + μ3	23.7	39	1		–	24.6(48.9)	40.4	1.4	51.6	0.06
CN+ m1^	22.6	39	2		0	1.8 (5.5)	39.6	0.7	37.1	0.40
RT+ > m3^	22.6	40	2		–	1.7 (5.7)	41.8	1.1	41.5	0.62
RT+ > μ1	19.6	44	2		–	15.6 (55.3)	57.0	10.2	58.2	0.65
VOP/A? + <> m2^	7.5	35	3		0	4.1 (6.7)	51.3	3.5	49.5	0.85
VOP/A? + <> μ2	4.5	49	3		–	24.6 (52.8)	46.2	11.13	47.3	0.45
VIM+<> m1^	7.5	33	3		–	4.9 (6.5)	48.4	3.7	49.5	0.75
VIM+<> m3^	7.5	33	3	2	0	4.2 (6.5)	48.4	4.2	49.5	0.98
VIM+<> μ1	4.5	47	3		–	23.3 (54.6)	45.5	9.3	46.5	0.40
VIM+<> μ3	4.5	45	3	1*	–	30.5 (53.2)	46.2	13.8	47.3	0.45
Subject 2
VOP/A? + <> m1	9.0	44	4		–	2.3 (6.6)	56.4	1.63	56.7	0.70
VOP/A? + <> m1	7.0	32	5		–	1.9 (6.1)	60.7	1.82	51.6	0.96
VOP/A? + <> μ1	6.0	49	4		–	3.7(10.2)	46.9	2.3	59.6	0.61
VOP/A? + <> μ1	3.0	47	6		–	2.4 (9.2)	54.9	2.4	56.7	0.99
VOP/A? + <> μ1	1.0	44	7		–	6.4 (10.3)	42.6	0.2	53.1	0.03
VC+<> m3	7.0	26	5		–	6.8(20.3)	42.6	–	–	0
VIM+<> μ2	6.0	42	4	1–2*	–	2.8 (10.6)	53.5	0.6	52.4	0.21
VIM+<> m2	4.0	50	7	1*	–	5.0 (13.5)	41.1	–	–	0
VIM μ2	0.0	50	6		–	3.8 (15.4)	53.5	0.5	53.1	0.14
Fpz EEG	All 5	215				3.2	45.5	2.8	46.5	0.90
T4 EEG	All 5	173				2.8	44.7	2.6	44.4	0.92
Pz EEG	All 5	190				3.1	46.2	3.6	46.5	1.17
Subject 3
VIM+ m1	8.0	36	8	2*	–	2.0 (5.7)	51.3	1.2	50.9	0.59
VIM+ m3	8.0	37	8		–	1.8 (6.2)	52.0	0.7	50.9	0.36
VIM+ μ1	5.0	48	8	1*	–	2.4 (8.2)	51.3	1.5	50.2	0.65
VIM+ μ3	5.0	48	8		–	3.2 (11.3)	44.7	3.1	50.9	0.98
VIM+ μ2	3.0	50			–	3.9 (12.7)	57.1	–	–	0
VIM+ m1	3.0	34	9	0–1*	–	3.0 (7.2)	60.0	0.3	68.4	0.09
VIM+ μ1	2.1	41	10	0	–	2.7 (8.3)	57.8	0.7	58.2	0.28
VIM+ μ3	2.1	48	10		–	3.2 (12.4)	59.3	0.5	61.1	0.15
VC+ μ3	0.0	46	9		–	4.5 (10.3)	60.0	4.7	67.6	1.04
Subject 4
VIM+ m2	8.9	35	11		–	2.1 (5.7)	68.0	0.02	75.6	0.01
VIM+ μ1	5.9	50	11		–	2.7 (6.9)	60.7	1.2	51.6	0.44
VIM+ μ3	5.9	46	11	1–2*		3.4 (9.7)	41.8	3.8	50.2	1.12
Subject 5
RT+ m1	20.0	28			–	4.2 (10.3)	59.3	6.4	59.6	1.51
VIM+ m3	8.6	32			–	3.8 (12.3)	53.5	1.6	51.6	0.43
Subject 6
RT+ μ1	17.5	38			–	3.0 (8.5)	68.7	3.6	59.6	1.19
VIM+ m1	6.4	37	12		–	1.4 (4.6)	53.5	2.7	49.5	1.99
VOP/A+ μ2	5.3	50			–	3.3 (12.6)	57.1	1.6	59.6	0.50
VOP/A+ μ2	3.4	46	12		–	2.8 (10.5)	58.6	–	–	0
Fpz EEG	All 4	154				1.8	48.4	3.0	46.5	1.73
T4 EEG	All 4	123				2.1	49.1	1.5	44.4	0.72
Pz EEG	All 4	167				1.6	49.1	1.9	45.8	1.18
Cz EEG	All 4	194				1.9	47.6	2.2	48.0	1.19
Subject 7**

1, 2, 3 = electrodes; μ = micro contact; m = macro contact; cluster = concurrently measured; ? = less certain; all LFP P50s were reverse polarity; mm = millimeters; Tip dist. = distance from DBS macroelectrode tip; DBS con = recordings at the contact on the implanted DBS lead

^*at therapeutic DBS contact

^**= no significant P50 responses were detected

^{^}=rhythmic LFP; VOP/A = presumptive ventrooralis posterior/ventrooralis anterior; VIM = ventral intermediate nucleus; CDS = current density source; − = CD source; 0 = CD sink; VC = ventrocaudalis; structure based on microelectrode recording = +; motor testing = < or MRI=>; Lat = latency.

Figure 1.

Microelectrode P50 recordings. (a) MRI T2 sagittal and coronal imaging showing positions where 13 significant conditioning P50 responses were identified of 27 recordings performed in subject 1. (b) P50 response recorded from EEG. (c) Diagram of the microelectrode used for recording both local field potential (LFP) from macro and micro contacts 3 mm apart referenced with the cannula. (d) Significant P50 responses measured with local field potentials intraoperatively. Green dots = caudate; purple dot = presumptive reticular thalamic nucleus (RT); yellow dots = ventrooralis posterior/ventrooralis anterior (VOP/A);orange dots = ventral intermediate nucleus (VIM). Reverse polarity recordings of P50 response;arrow = timing of click; position in mm relative to tip of implanted DBS macroelectrode. Blue = macro contact recordings; black = micro contact recordings. (e) Current source density (CSD) analysis of averaged LFP recordings from three microelectrode recordings (MERs) at the tip identifying current sources in dark blue corresponding with conditioning P50 latencies. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 2.

Frequency of P50 response across caudate (CN), anterior limb of the internal capsule (ALIC), presumptive reticular thalamic nucleus (RT), dorsal thalamus (DTH), and ventral thalamus (VTH). Refined P50 localization within VTH nuclei including ventrooralis anterior/ventrooralis posterior (VOP/A), ventrocaudalis internus (VC), and ventral intermediate nucleus (VIM) and possible P50 defined sub-territories include medial VTH (MVTH), lateral VTH (LVTH). Comparison of P50 response at the therapeutic DBS contact (DBS) with the rest of VTH. (a) Frequency in percentages of significant P50 responses over total recordings by deep brain structure, or sub-territory. * indicates significant differences in the frequency of P50 response between sub-territories MVTH and LVTH (p = 0.04). (b) Comparison of both normalized P50 and non-significant (NS) P50 peaks by deep brain structure. * indicates significantly lower P50 peaks found in ALIC compared to CN, RT, and VTH (p = 0.04). (c) Frequency in percentage of gated (P50 ratio ≤ 0.5) and nongated (P50 ratio > 0.5) P50 responses. (d) 3D scatter plot of both NS recordings and P50 responses collapsed to the right side and normalized. Corresponding centroids within identified deep brain structures. (e) VTH 3D scatter plot of both NS recordings and P50 responses within MVTH and LVTH sub-territories with corresponding centroids. VOP/A, VIM, and VC indicated by shape. (f) Sagittal and (g) Coronal T1 MRI showing 21 trajectories of the microelectrodes where the recordings were made with centroids from deep brain structures traversed. [Color figure can be viewed at wileyonlinelibrary.com]

The majority, 92.5%, of the 40 conditioning P50 responses identified were local neuronal current sources based on current source density (CSD) estimates (Fig. 1e and Table 2). The presence of two or more P50 responses recorded with LFP simultaneously from different contacts or electrodes was characterized as clusters. The significant P50 responses in or bordering the head/body interface of the CN included two clusters (1 and 2) of large amplitude, concurrently recorded, P50 responses (Table 2; subject 1; Fig. 1). Cluster 1 conditioning responses had similar latencies (40.4–41.8) and were similarly gated (Table 1). Cluster 2 included concurrent P50 recordings from both CN and RT. The P50 latencies (41.8–68.7 ms) and gating (P50 ratio = 0.40–0.65) varied by contact as well in RT. The centroid of the positions of P50 response (P50 centroid) from CN recordings was medial to positions where no P50 response was identified (NS centroid) and both anterior and superior to the P50 centroid for RT (Table 3A and Fig. 3c). In RT, the P50 centroid was posterior to the NS centroid (Table 3 A and B). The NS centroids for ALIC and DTH were distinct from other NS centroids in neighboring deep brain structures.

Table 1.

Participants With Essential Tremor Undergoing DBS Surgery Targeting VIM.

Subject	Age	Sex	Disease duration	Side implanted	Track implanted	Contacts used
#1	66	M	35	L VIM	#3 posterior	0, 1
#2	65	M	41	R VIM	#2 center	1
#3	71	M	N/A	L VIM	#1 center	1, 2
#4	57	M	39	L VIM	#3 posterior	1
#5	69	M	20	L VIM	#4 anterior, lateral	2
#6	68	F	20	R VIM	#3 posterior	1, 2
#7	72	M	57	L VIM	#4 posterior	1, 2, 3

Table 3.

P50 and NSP50 Centroids by Deep Brain Structure.

Structure	x	y	z
A. Centroids of the positions of P50 response
CN	19.2 (0.10)	5.2 (0.65)	18.6 (0.24)
RT	19.1 (0.47)	3.8 (0.30)	15.2 (0.92)
VTH	13.8 (0.16)	−2.8 (0.59)	2.6 (0.41)
VTH Nuclei
VC	13.4 ((0.25)	−7.2 (3.2)	0.1 (2.89)
VIM	14.0 (0.21)	−3.7 (0.52)	2.9 (0.45)
VOA/VOP	13.6 (0.26)	0.4 (0.73)	2.4 (0.75)
VTH sub-territory
MVTH	13.6 (0.13)	−2.7 (0.66)	2.3 (0.45)
LVTH	15.4 (0.40)	−3.3 (1.33)	4.5 (0.56)
B. Centroids of the positions of nonsignificant response (NSP50)
CN	20.2 (0.26)*	5.8 (0.47)	18.3 (0.23)
ALIC	18.5 (0.12)	4.2 (0.38)	15.4 (0.32)
RT	18.9 (0.35)	2.0 (0.25)*	13.9 (0.42)
DTH	17.3 (0.18)	2.2 (0.80)	10.3 (0.54)
VTH	14.7 (0.15)*	−3.0 (0.26)	3.4 (0.28)*
VTH Nuclei
VC	13.2 (0.23)	−5.1 (0.53)	1.1 (0.64)
VIM	14.9 (0.17)*	−3.4 (0.26)	3.6 (0.31)
VOA/VOP	14.2 (0.36)	0.1 (0.49)	2.6 (0.72)
VTH subterritory
MVTH	13.5 (0.09)	−3.3 (0.33)	2.1 (0.31)
LVTH	15.7 (0.16)†	−2.8 (0.38)	4.3 (0.38)†

x = medial-lateral

y = posterior-anterior

z = inferior-superior.

^†LVTH NS axis distinct from P50 MVTH.

^*NSP50 axis distinct from P50 centroid (standard error).

Figure 3.

(a) Diagram of DBS macroelectrode with four 1.5 mm contacts (0, 1, 2, and 3) spaced 1.5 mm apart. (b) Postoperative T2 MRI (inverse) probe view oblique sagittal image of left side of subject 1. Position of significant P50 relative to contacts 0, 1, 2, and 3 on the implanted DBS macroelectrode. (c) Postoperative T2 MRI (inverse) probe view oblique sagittal image of right side of subject 2. Position of significant P50 relative to contacts 0, 1, 2, and 3 on the implanted DBS macroelectrode. [Color figure can be viewed at wileyonlinelibrary.com]

A gating deficit or nongated response has been defined as a P50 ratio ≥ 0.5 using the auditory gating paradigm, where P50 responses were measured to identical, paired auditory click stimuli typically presented 500 ms apart. The significant P50 responses were gated more than background NS responses (p < 0.05; P50 ratio = 0.42 ± 0.46; NS ratio = 9.09 ± 65.92) and the frequency of gating (P50 ratio < 0.5) was significantly higher in CN and VTH compared to RT (Fig. 2c). Both EEG and LFP recordings were collected in two subjects and gating was variable (Table 2). In subject 2, four of nine LFP P50 ratios were ≥ 50%, whereas all three of the EEG-recorded P50 ratios were ≥ 50%. In subject 6, two of the four LFP significant P50 responses and all four EEG P50 ratios were ≥ 50%.

Characteristics of Significant Conditioning P50 Response Within VTH Nuclei and Sub-Territories

An effort was made to identify the presumptive P50 source within VTH nuclei, within P50-defined sub-territories, and relative to therapeutic DBS contacts. Of the 117 LFP recordings from the ventral thalamus, the frequencies of the 31 VTH P50 responses did not vary significantly between VC (18.1%), VIM (23.5%), and VOP/A (42.9%) nuclei. There were 10 clusters of two or more concurrently recorded P50 responses from five subjects (Table 2). The most robust P50 responses were recorded at 4.5 and 7.5 mm from target from subject 1 at all six contacts with amplitudes of 7.7–8.93 μV at latencies of 48.4–51.3 ms from the macrocontact recordings and amplitudes of 26.88–34.10 μV at latencies of 45.5–46.2 ms from the microcontacts (Fig. 1 and Table 2, cluster 3; Fig. 3a). For these six simultaneous P50 responses, the P50 ratios were similarly gated among recordings from the micro contact (Table 2: P50 ratio = 0.4–0.45) and the P50s detected from the macro contact were similarly not gated (P50 ratio = 0.75–0.98). The P50 responses within each of the nine other clusters detected appeared distinct from each other with variable peaks, latencies, and P50 ratios (Table 2, clusters 4–12).

Within the VTH, the P50 centroid was distinctly medial and inferior to the NS centroid (Table 3). In addition, a VTH sub-territory was readily distinguished based on the 3D plot of P50 responses that were more medial and inferior to nonsignificant recordings (Fig. 2a,e). The P50 and NS centroids for both MVTH and LVTH were distinct from each other (Table 3 and Fig. 2e). While the P50 and NS centroids for VC, VIM, and VOP/A were distinct on the y (anterior-posterior axis) as expected, there were no significant differences in the frequency, peaks, or amplitudes or gating of P50 across these VTH nuclei. Of the 27 LFP recordings that occurred along the DBS implanted track, 21 occurred near therapeutic DBS contacts. There were seven significant P50 responses of the 21 recordings at or between therapeutic DBS contacts (within 0.5 mm) (Figs. 2a and 3 ).

DISCUSSION

The goal of this study was to measure the auditory evoked P50 response with LFP during awake DBS surgery in patients with ET to identify deep brain structures involved in P50 auditory-evoked response circuitry and potentially sensory processing. While P50 response measured with EEG occurs through volume conductance, the aim here was to identify regions (subterritories) where P50 response may be locally generated. Results of this study showed that the NS responses were detectable in all deep brain structures tested. These were not distinct from background variance and could not be considered true auditory evoked responses. Focusing on the significant P50 responses provided some evidence of potential local P50 sources in the VTH, CN, and RT based on CSD analyses and more frequent detection of multiple significant P50 responses both in several patients and in clusters where multiple P50 responses were detected simultaneously. The finding of no significant P50 response of recordings from DTH and ALIC—a white matter tract—was consistent with P50 response having some neuroanatomic specificity. Background overall was significantly lower in recordings from ALIC as demonstrated with NS peaks (Fig. 2b). Initial stereotactic coordinates were determined for both significant P50 centroids and NS responses for each relevant structure within the data collection areas (Fig. 2). Ultimately, identification of regions involved in generating P50 response/gating may be useful for DBS targeting subterritories of deep brain structures to either avoid DBS induced sensory processing side effects or treat deficits in auditory gating found in disorders like SZ.

Our detection of robust P50 signal, precisely at the contacts shown to alleviate tremor upon stimulation, is consistent with the hypothesis that sensory processing and motor circuitry are sometimes integrated within VIM—a finding independently supported with electrocorticographic (ECoG), functional connectivity, and diffusion tensor tractography findings reported in the literature (20,54,55). Similarly, large maximum P50 responses have been recorded with ECoG from both midfrontal (including primary/supplementary motor) and lateral temporoparietal (including auditory) cortex identified functionally using electrical stimulation mapping (20,56). Independent approaches demonstrate connectivity between VIM and premotor and motor cortices (23,57,58). Functional connectivity has been demonstrated by both detection of VIM DBS-induced cortical evoked potentials, and diffusion tensor imaging (DTI) analysis showing that white matter tracts seeded in VIM and VOP extend to premotor and motor cortices (23,58). Some studies have shown changes in VIM connectivity in ET (23,58). Therefore, there may be ET disease-related effects on P50 response indirectly related to either the aberrant oscillatory activity or connectivity in motor circuitry. Additional evidence supporting P50 and motor circuitry integration comes from evidence that P50 gating shares circuitry with prepulse inhibition (PPI) of the acoustic startle response—a paradigm of auditory sensorimotor gating (59,60).

While variability in concurrent recordings and gating from different EEG sites is known, and the most robust P50 response (subject 1, from 4.5 mm) concurrently recorded highly similar P50 responses and gating from three microcontacts only 2 mm apart (P50 ratios of 0.40, 0.45, and 0.45), it was unexpected that most of the other P50 ratios were highly variable (61). Increasing the number of epochs for estimating P50 within deep brain structures would likely decrease variability and may be necessary for investigations into gating. The P50 responses (subject 1 P50 4.5 and 7.5 mm) measured with LFP at the microcontact at the tip were 3.6 times larger and appeared distinct from the P50 responses at the macro contacts (a band 3 mm superior to the tip) on each microelectrode (p < 0.05) likely due to differences in impedances, shape and positions of these contacts. It is possible that the macro and micro contacts record from different populations of cells even at the same depth. Overall, 55% (22/40) of the P50 ratios were gated in the seven ET patients (Table 2). Some subjects may have been acutely stressed during the awake DBS surgery and this state of heightened awareness was expected to impede gating. To determine the gating status of individuals with ET, P50 should be measured in a quiet environment (outside the OR).

Finally, while P50 gating deficits have been described preliminarily for several other brain disorders and these findings may help determine if there is overlap with circuitry implicated these brain disorders, the P50 paradigm has been particularly relevant to illnesses like SZ that are associated with sensory deficits (5,7,9,62). P50 gating has enabled electrophysiological-informed development of novel therapeutic interventions (63). With the goal of identifying novel therapies that normalize gating deficits, P50 auditory gating has been employed as a treatment-responsive biomarker used to generate titration curves to demonstrate target engagement for: 1) pharmaceutical intervention; 2) prenatal choline supplementation to overcome SZ-related developmental delay; and 3) the development of transcranial direct current therapy targeting dorsolateral prefrontal cortex (64-68). In addition, auditory gating measured with P50 may be a useful biomarker in the development of deep brain stimulation (DBS) therapy to treat SZ (30).

In light of our goal to identify P50 circuitry relevant to SZ, our finding of significant P50 response in VOP/A and in the head/body interface of CN is of particular interest because of their connectivity with DLPFC—a structure both implicated in P50 gating through fMRI analysis and with notable aberrancies in SZ patients (40,55,67,69-71). In addition to showing connectivity with premotor and motor cortices, VOP-seeded tractograms have a high intensity in the DLPFC (20). Impairment in DLPFC function is thought to contribute to cognitive deficits in working memory and psychosis in patients with SZ (42,72-74). The CN has been proposed as a DBS target for treating symptoms of SZ (30,39). DBS specifically targeting the sub-territory of the CN at the head/body interface is effective at treating tinnitus, a circuit disorder hypothesized to result from aberrant auditory gating (3,47,48). Together this convergent evidence supports DBS targeting the CN head/body interface also being relevant to treating auditory hallucinations that are found in the majority of subjects with SZ.

The finding of P50 responses in presumptive RT was expected based on convergent neurophysiological, pharmacological, molecular expression, and functional imaging investigations (21,60,75-78). However, targeting RT—a thin nuclear rim comprised of inhibitory GABAergic cells surrounding part of the thalamus—was based on the first detection of characteristic thalamic SUA activity (increased background and detection of cellular activity at one or more of the three microelectrodes) and stereotactic MRI. RT integrates corticothalamic circuitry by receiving excitatory projections from both the thalamus and the cortex while transmitting through inhibitory efferent output to dorsal thalamus. Finally, aberrancies in thalamocortical connectivity and coordination through oscillatory activity identified in SZ implicate RT as a DBS target capable of modulating relevant signaling in a neonatal ventral hippocampal lesion rat model of SZ (51).

The method described here is part of an approach we are taking to inform DBS therapeutic interventions that may treat sensory processing deficits such as those found in patients with SZ. Characterizing P50 response in deep brain structures will help identify and localize the circuitry that may be relevant to aberrancies in sensory processing. To inform DBS clinical trials aimed at alleviating symptoms related to sensory processing, intraoperative stimulation may indicate when the intended target is engaged with altered P50 response. Successful DBS targeting to treat disease-specific symptoms will likely depend on identifying a target in which multiple circuits related to disease pathology intersect, and the P50 response may assist in identifying the specific subterritories to target.

We conclude that the auditory evoked P50 responses measured with LFP are readily detectable in awake DBS patients in specific anatomical structures. The LFP analysis for P50 response was validated by: 1) detecting multiple significant conditioning P50s concurrently; 2) identifying significant P50 responses in the same deep brain structures across multiple cases; and 3) the finding of P50 peak responses that were up to 10-fold higher than EEG-recorded P50 (both in the literature and in our study). When used in combination with stereotactic imaging and MER, significant P50 response from LFP recordings showed anatomic specificity by mapping to CN, RT and VTH, but not ALIC or DTH. Several of the P50 responses in VIM were at DBS contacts that were therapeutic for ET. These findings were consistent with the P50 circuitry implicated in the literature.

MATERIALS AND METHODS

Participants

The research protocol was approved by Colorado Multiple Institutional Review Board (COMIRB #14–0358). Seven patients with ET undergoing DBS surgery and capable of hearing the auditory clicks provided informed consent (Table 1; one female and six males). All DBS surgeries were performed without complications. Demographics are summarized in Table 1. All the subjects had severe symptoms that were refractory to treatment. The average duration of ET was 35 years and average age was 67 years old. All were treated with staged bilateral DBS but participated in this study for either the left (five), or right (two) VIM to treat ET.

Imaging

Stereotactic imaging included 1.5 Tesla magnetic resonance imaging (MRI; Siemens 64-slice unit, Medical Solutions, Malvem, PA, USA) and computed tomography (CT; Philips 64-slice unit, Pittsburg, PA) scans with a stereotactic head/body interface frame (Integra™ CRW^® System, Integra Life Sciences Corporation, Plainsboro, NJ, USA or Leksell Stereotactic System^®, Elekta, Stockholm, Sweden). Cortical entry site and trajectory to thalamic VIM target were identified on preoperative stereotactic MRI images fused to CT, using stereotactic preoperative planning software and platform (iPlan 3.0 Stereotaxy; Brainlab, Germany).

DBS Surgery, Identifying Deep Brain Structures

DBS trajectories were identified entirely based on clinical parameters used to safely and effectively treat symptoms of ET. Intraoperatively, patients were placed supine, with stereotactic frame secured to the operating room (OR) table. After prepping the skin with a mild abrasive (Nuprep^®, Weaver and Company, Aurora, CO, USA), EEG electrodes were positioned with conductive paste (Ten20^®, Weaver and Company) including Cz, Pz, Fpz, T4, right superior orbit, right lateral orbit, right ear (reference), and left ear (ground). Impedance at each EEG electrode was <10 kΩ, except Pz on subject 2 which was 28.7 kΩ. Ear buds, were placed to test hearing at a volume set to the 90% setting (80–85 dB). The use of MER for VIM DBS targeting is routine for ET cases and no changes were made to the surgical procedure to accommodate this research study except for the time to perform the P50 recordings. During surgery, three microelectrodes (NeuroProbe Sonus™; STR-009080-10; Alpha Omega, Nazareth, Israel) were advanced from 25 mm above target to 2 mm below target, to validate exit from the inferior border of thalamus, as per standard clinical protocol. Auditory clicks were delivered when caudate (CN) and thalamic nuclei (RT, VIM, ventrooralis posterior/anterior/ [VOP/A], and ventrocaudalis [VC]) were identified using intraoperative monitoring of SUA to identify high background levels, tremor cells (VOP/A and VIM), kinesthetic neurons (VIM), firing during voluntary movement (VOP/A), tactile neurons (VC) and/or induction of paresthesias with macrostimulation (VC) (52,79). During recording of P50, extraneous noise in the operating room was silenced. Patients were instructed to fix their gaze on an object and avoid blinking and eye movement within trials. Intraoperative SUA and LFP were recorded with microelectrodes referenced to the cannula (Neuro Omega™, Alpha Omega). SUA was recorded from the tips of the microelectrodes, and LFP were recorded both at the microelectrode tip and at the semi-macroelectrode 3 mm proximal to the tip (Fig. 1). Pre/post-operative imaging was used to confirm and refine recording positions.

Generating Auditory Clicks

A custom MATLAB (v.7.14. R2015a, MathWorks, Natik, MA, USA) application was constructed to deliver auditory clicks (square wave pulse of 0.04 ms duration), and P50 responses were recorded. Paired clicks were spaced 500 ms apart, with 2.8 s intervals between paired clicks, and with an additional random variable interval from 1–100 ms. Seventeen sets of paired clicks were delivered in each of three trials, with a 0.5-min rest interval between each trial. At the start of each run, following 100 ms of baseline recording and concurrent with the first audio click, a TTL signal was relayed through a toolbox allowing triggers (DATAPixx™ Lite Data acquisition system, Vpixx Technologies, Saint-Bruno, QC, Canada) to the MER system (NeuroOmega™, Alpha Omega) through a DB25 to 2BNC cord. For EEG, LFP and SUA, a total of 50 epochs were analyzed at each of two to five stops along the trajectory to the VIM.

Offline EEG and LFP Processing, Artifact Removal, and Analysis

Analysis occurred offline by converting. map data files, generated by the neural signaling recording system (NeuroOmega™, Alpha Omega), into MATLAB compatible files (.mat) by a program called Converter (Alpha Omega). Epochs were automatically rejected if artifact, greater than three standard deviations from the mean, occurred within or between paired P50 detection windows. Artifact in all EEG channels due to electrooculographic (EOG) activity recorded from both the right superior and lateral orbits was eliminated from further analysis. Data were baseline normalized. For the primary analysis of all data, a 10 Hz high-pass filter, 100 Hz low-pass filter, and 60 Hz notch filter were applied to all data using Signal Processing Toolbox^® (v.7.14. R2015a, MathWorks).

Analyzed offline, LFP recordings were amplified 1000 times. Conditioning P50 responses were identified as the peak of the most positive wave, P1, with a latency of 40–75 ms following a click. Eleven recordings without peaks within the window and eight conditioning P50s that were <0.5 μV were excluded from analysis. Significant conditioning P50s were consistently identified with inverted (reverse) polarity (Table 2). A 13 ms conditioning-testing interval between P50s latencies was used in order to include three concurrently measured robust and similar P50 responses (CN [n = 2] and VOP/A [n = 1]; Table 2); the conditioning-testing interval may vary based on site of P50 measurement (61). If test P50s were not identified it was considered complete gating (n = 11). Finally, significant differences were identified in the variance of baseline recordings between patients. Therefore, both conditioning and test P50 amplitudes were normalized across subjects to an average baseline variance for comparison across subjects. The P50 amplitude was measured from the preceding trough to peak.

CSD analyses were used to estimate whether extracellular currents were a source or a sink at the latencies of peak P50 signals. A source is where the current is produced from surrounding neurons and a sink is where the current is absorbed by surrounding neurons. The inverse (delta) CSD method was employed using: 1) averaged LFP recordings used for P50 detection; 2) 25,000 Hz sampling frequency; 3) 0.002 m spacing between three microelectrodes; 4) 1 mm radius of the microelectrode, and 5) 0.3 siemans/meter conductivity of the extracellular matrix (the default) (80,81). Concurrently recorded LFPs from parallel, equally spaced, linear microelectrodes were used to estimate CSD.

Mapping Position of P50 Detection in Deep Brain Structures

Postoperative trajectories of the three MER tracks and positions of P50 assessments were derived based on position relative to the implanted DBS macroelectrode (DBS macroelectrode model 3387; Medtronic, Minneapolis, MN, USA) visualized with computed tomography (CT) and magnetic resonance imaging (MRI) using stereotactic imaging software (iPlan 3.0 Stereotaxy; Brainlab, Germany). The trajectory of the implanted DBS macroelectrode was identified (hyperintense signal) with postoperative axial CT scans. Three (anterior, center, and posterior) of five possible tracks in the neurosurgical guiding tool (“bengun” array) were used in the “+” configuration for MER. Microelectrode track positioning was confirmed by identifying edema on postoperative MRIs fused to postoperative axial CT scans. Intraoperative positions of the P50 recording were noted relative to final DBS lead placement. Positions were identified with intraoperative assessment of SUA was primarily used in combination with stereotactic MRI imaging-based positioning (CN, anterior limb of the internal capsule [ALIC], and RT) within deep brain structures (Table 2).

Statistics

Peak μV P50 values were used to assess significance (Table 2). Significant P50 responses were identified using a paired t-test to compare conditioning P50 for each epoch at the latency of the average peak conditioning P50 response compared to the baseline averaged over 100 ms for each epoch (p < 0.05).

The frequency of P50 responses in different structures was analyzed by Fisher’s Exact Test (FET) with post hoc analysis of individuals pairs. P50 peaks and amplitudes means were analyzed across structures with a one-way ANOVA and Duncan’s Multiple Range Test (Statistical Analysis Software [SAS], Cary, NC, USA). For each point of MER and P50 assessment, 3D coordinates were identified, collapsed to the right side, normalized across patients by both anterior commissure and posterior commissure (AC/PC) lengths (to the average 24.85 mm) and to placement of the bottom of the DBS lead (to average x = 12.26 mm; y = −5.29 mm; z = −1.90). Centroids were determined with P50 coordinates weighted by P50 peak responses for each structure or subterritory (SAS). Centroids generated from NS P50 coordinates were not weighted.