JNEUROSCI.1757-23.2024.full-html.html

Research Articles | Behavioral/Cognitive

7-Tesla evidence for columnar and rostral–caudal
organization of the human periaqueductal gray
response in the absence of threat: a working memory
study.

https://doi.org/10.1523/JNEUROSCI.1757-23.2024

Received: 12 September 2023
Revised: 1 March 2024
Accepted: 8 April 2024

This is an open-access article distributed under the terms of the

Creative Commons

Attribution 4.0 International license

, which permits unrestricted use, distribution and

reproduction in any medium provided that the original work is properly attributed.

This Early Release article has been peer reviewed and accepted, but has not been through
the composition and copyediting processes.The final version may differ slightly in style or
formatting and will contain links to any extended data.

Alerts: Sign up at

www.jneurosci.org/alerts

to receive customized email alerts when the fully

formatted version of this article is published.

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Title page

Title:

7-Tesla evidence for columnar and rostral

–caudal organization of the human

periaqueductal gray response in the absence of threat: a working memory study.

Abbreviated title:

Human periaqueductal gray in working memory

Author Names and Affiliations:

Alexandra K. Fischbach

, Ajay B. Satpute

, Karen Quigley

Philip A. Kragel

, Danlei Chen

, Marta

Bianciardi

Larry Wald

Tor D. Wager

, Ji-Kyung Choi

Jiahe Zhang

Lisa Feldman Barrett

1 **

Jordan E. Theriault

3**

**Shared senior authorship

Author Affiliations:

Department of Psychology, Northeastern University, Boston, MA, 02115

Department of Psychology, Emory University, Atlanta, GA, 30322

Department of Radiology, Athinoula A. Martinos Center for Biomedical Imaging,

Massachusetts General Hospital and Harvard Medical School, Charlestown, MA, 02129

Department of Psychological and Brain Sciences, Dartmouth College, Hanover, NH,

03755

Department of Surgery, University of California, San Francisco, CA, 94143

Author contributions:

P.A.K., M.B., K.Q., L.W., T.D.W., L.F.B., and A.B.S. designed research;

A.K.F. and J.E.T., analyzed data; A.K.F. and J.E.T. wrote the first draft of the paper; A.K.F.,
P.A.K., A.B.S., K.Q., L.F.B., and J.E.T. edited the paper; A.K.F., A.B.S., K.Q., L.F.B., and J.E.T.
wrote the paper; D.C., J.-K.C., and J.Z. performed research.

Corresponding author:

Jordan Theriault (email: JTHERIAULT2@mgh.harvard.edu)

Number of figures: 6

Number of words

Abstract: 209

Introduction: 1,375

Discussion: 919

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Abstract

The periaqueductal gray (PAG) is a small midbrain structure that surrounds the cerebral

aqueduct, regulates brain

–body communication, and is often studied for its role in “fight-or-flight”

and “freezing” responses to threat. We used ultra-high field 7-Tesla fMRI to resolve the PAG in

humans and distinguish it from the cerebral aqueduct, examining its in vivo

function in humans

during a working memory task (N = 87). Both mild and moderate cognitive demand elicited

spatially similar patterns of whole brain BOLD response, and moderate cognitive demand

elicited widespread BOLD increases above baseline in the brainstem. Notably, these brainstem

increases were not significantly greater than those in the mild demand condition, suggesting

that a subthreshold brainstem BOLD increase occurred for mild cognitive demand as well. PAG

response was group-aligned and examined with subject-specific masks. In PAG, both mild and

moderate demand elicited a well-defined response in ventrolateral PAG (vlPAG), a region

thought to be functionally related to anticipated painful threat in humans and non-human

animals

—yet, the present task posed only the most minimal (if any) “threat”, with the cognitive

tasks used being approximately equivalent to remembering a phone number. These findings

suggest that the PAG may play a more general role in visceromotor regulation, even in the

absence of threat.

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Significance statement

The periaqueductal gray (PAG) is thought to control survival-related behavior, and is typically

studied using experiments that manipulate threat. Others have proposed that the PAG plays a

more general role in bodily regulation, but studies examining PAG function outside of threat-

based experimental contexts are rare. We used high-resolution fMRI to examine PAG response

in humans during a working memory task, which involves minimal threat. Moderate cognitive

demands elicited a well-defined response in ventrolateral PAG, a functional subregion thought

to coordinate a

“freezing” response to threat. A task where threat is minimal elicited a clear fMRI

response in one of the most well-known survival circuits in the brain, which suggests the PAG

supports a more general function in brain

–body coordination.

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Introduction

Historically, the periaqueductal gray (PAG) has been identified as a key brain structure

mediating the “fight-or-flight” and “freezing” responses (for review, see Bandler, 1988; Bandler,

Carrive, & Depaulis, 1991), and this functional association of PAG with threat responses

continues to be articulated in recent reviews (Mobbs et al., 2015, 2020; C. Silva & McNaughton,

2019). Direct chemical stimulation of the PAG elicits a diverse range of motor behaviors and

physiological responses

—including, in cats, immobility, pupil dilation, vocalization (e.g., hissing

or howling), alert positions (Bandler & Carrive, 1988), running and jumping (S. P. Zhang et al.,

1990), and hypertension and tachycardia (Abrahams et al., 1960), all of which are generally

understood as part of the

“fight-or-flight” and “freezing” responses (Bandler, Carrive, & Zhang,

1991a; Bandler & Depaulis, 1991; Fanselow, 1994). Likewise, in humans PAG is most

commonly studied for its role in threat (Coker-Appiah et al., 2013; Hermans et al., 2013; Lindner

et al., 2015; Wang et al., 2022; C. Weis et al., 2020; C. N. Weis et al., 2022; Zhou et al., 2021).

Deep-brain stimulation of the human PAG elicits a desire to escape (Amano et al., 1978),

feelings of fear and impending death, and autonomic changes including hyperventilation and

increased heart rate (Nashold et al., 1969). Brain imaging studies in humans observed blood-

oxygen level-dependent (BOLD) signal intensity increases in PAG when participants were

chased through a virtual maze (i.e., approached by a simulated predator; Mobbs et al., 2007),

and when exposed to a learned cue signaling the onset of an unpleasant respiratory restriction

(Faull et al., 2016, 2019). The interpretation of the PAG as a primary region for coordinating

survival-based responses has been used to unify these many observations under a single

coherent function (Fanselow, 1991; B. A. Silva et al., 2016)

in which the PAG was conceived of

as playing a central role in an evolutionarily ancient “fear circuit” (LeDoux & Daw, 2018; B. Lu et

al., 2022; McNaughton & Corr, 2018; Panksepp, 2011; B. A. Silva et al., 2016; Vázquez-León et

al., 2023).

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

The hypothesis that PAG is a center for survival-based responses unifies a great deal of

evidence, but not all evidence. For example, PAG has been observed to play a functional role in

bodily processes at moments when threat is absent (for review, see Carrive, 1993), including

control of the heart (M. A. Gray et al., 2009; Hermans et al., 2013; Hohenschurz-Schmidt et al.,

2020; Pereira et al., 2010; Wager, Waugh, et al., 2009), blood pressure control (Green et al.,

2005), cardiorespiratory coordination (Dampney et al., 2013), bladder control (B. F. Blok et al.,

1997; B. F. M. Blok & Holstege, 1998), rapid-eye-movement (REM) sleep (J. Lu et al., 2006),

body temperature (Chen et al., 2002; Y. H. Zhang et al., 1997), trigeminovascular contributions

to migraine headache (Knight & Goadsby, 2001), communication (e.g., the vocalization of

Songbirds; Haakenson et al., 2020), mating behavior (e.g., the control of male copulatory

behavior in Quails; Carere et al., 2007), and feeding behaviors (e.g., the motivational drive to

hunt in mice; Mota-Ortiz et al., 2012; also see Tryon & Mizumori, 2018). Furthermore, the PAG

has extensive connectivity with regions across the entire neuraxis from spinal cord and

brainstem to cerebral cortex, placing it at a critical integration point in brain

–body

communication (for review, see Carrive & Morgan, 2012). Given this evidence, we hypothesized

that PAG involvement in response to predatory threat is consistent with a more fundamental

function: the PAG may coordinate and regulate internal bodily (visceromotor) systems in service

of efficient regulation of the body (called allostasis). On this interpretation, threat responses

represent only one of many contexts where the autonomic nervous system, the endocrine

system, the immune system and tissues of the body must be coordinated with each other and

with motor movements (for review, see Benarroch, 2012; Motta et al., 2017). If this is correct,

and the PAG does play a more general regulatory role (rather than a threat/survival-specific

role) then organized patterns of PAG response may be observed in task-based contexts

involving effort, but minimal threat.

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Critically, well-known columnar functional distinctions within the PAG set some

expectations for how its response is organized. Within the PAG, prior work has identified

dorsomedial (dmPAG), dorsolateral (dlPAG), lateral (lPAG), and ventrolateral (vlPAG) functional

columns (Bandler, Carrive, & Depaulis, 1991; Carrive, 1993). In the context of threat/survival-

based research, dl/lPAG is thought to support active coping

“fight-or-flight” responses to

predatory threats, whereas dmPAG is thought to support similar responses to aggressive

conspecifics, and vlPAG is thought to support passive coping or

“freezing” responses to painful

threats (Gross & Canteras, 2012; Morgan et al., 1998; S. P. Zhang et al., 1990). For example,

stimulating dl/lPAG, in rodents and cats increases cardiac and respiratory rates (Bandler,

Carrive, & Zhang, 1991b; Carrive & Bandler, 1991; Depaulis et al., 1992; Subramanian et al.,

2008; W. Zhang et al., 2005) and elicits explosive running and jumping (Bandler, Carrive, &

Zhang, 1991b; Fardin et al., 1984; Morgan et al., 1998; S. P. Zhang et al., 1990). By contrast,

stimulating vlPAG decreases cardiac and respiratory rates (Bandler, Carrive, & Zhang, 1991b;

Carrive & Bandler, 1991; Depaulis et al., 1992; Lovick, 1992; Subramanian et al., 2008), and

induces an absence of movement in cats (S. P. Zhang et al., 1990)

and “freezing” behavior in

rats (Carrive et al., 1997; Morgan et al., 1998). In humans, anticipated breathlessness increases

the BOLD signal in vlPAG (Faull et al., 2016, 2019; Tinoco Mendoza et al., 2023), whereas

subcutaneous (deep muscle) and cutaneous pain increases the BOLD signal in vlPAG and

100

lPAG respectively (Tinoco Mendoza et al., 2023). Observations that correspond to this

101

functional localization along PAG columns would increase our confidence in an observed

102

pattern of results, given that these functional distinctions have been strongly established in prior

103

work.

104

In the present study, we examined changes in PAG activity in the context of a mild-to-

105

moderately challenging N-back working memory task, which aims to simulate energetic,

106

cognitive, and physiological demands within the range of those typically encountered in daily

107

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

life. Participants observed a sequence of letters and responded with a key press when a letter

108

was repeated, either from the prior trial (1-back) or three trials prior (3-back; Fig. 1). This task

109

required that participants simultaneously attend to and remember target stimuli while inhibiting

110

or completing task-relevant motor responses. The 1-back task could be conceptually described

111

as easier than remembering a phone number, and so we consider it to fall well outside the

112

traditional survival-based contexts in which PAG function has previously been tested.

113

Imaging resolution in 1.5 Tesla and 3 Tesla MRI scanners is insufficient to resolve small

114

brainstem nuclei, both due to low signal-to-noise ratios (SNR) and voxel resolution. By contrast,

115

ultra-high resolution 7 Tesla (7T) imaging, used in the present work, provides a substantial

116

increase in SNR (Linnman et al., 2012; Satpute et al., 2013). An earlier report included the first

117

24 participants from the current data set (all the participants who were available for analysis at

118

that time) and used univariate and multivariate methods to distinguish PAG BOLD responses to

119

1-back (mild cognitive demand) and 3-back (moderate cognitive demand) conditions (Kragel et

120

al., 2019). This earlier study reported BOLD signal intensity greater increases in the 3-back (vs.

121

the 1-back) condition, which were localized to the right rostral vlPAG. The present work builds

122

on this earlier report, with a sample size (N = 87) that affords the power to identify functional

123

distinctions among PAG columns. We also improved our localization procedure, identifying PAG

124

columnar subregions by their radial degree. As in prior work (Kragel et al., 2019), we aligned

125

subject-specific PAG masks to a group template to estimate the voxel-wise BOLD response

126

during the N-back task. For PAG columns, tract-tracing techniques in non-human animals

127

typically produce symmetrical results (Carrive et al., 1987; Carrive & Bandler, 1991; Holstege et

128

al., 1997; Jansen et al., 1998; Li et al., 1993; Paredes et al., 2000; Sandkühler & Herdegen,

129

1995; S. P. Zhang et al., 1990), and so we expected that PAG activity would be symmetric

130

along the PAG midline. We recorded cardiac interbeat interval (IBI) and respiratory rate (RR) to

131

examine whether columnar organization supported visceromotor functions, and whether this

132

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

visceromotor support could possibly account for any task-elicited changes observed during the

133

working memory task.

134

Methods

135

Participants.

One hundred forty participants were recruited from the Greater Boston area and

136

provided informed consent in accordance with guidelines set forth b

y the Partners’ Healthcare

137

Institutional Review Board (IRB). Consented participants completed two scanning sessions and

138

were paid $150 for each scan (i.e., $300 upon study completion). Eighteen participants either

139

withdrew or ended the MRI scan session before starting the working memory task. Nineteen

140

participants were excluded due to poor image quality (e.g., large artifacts compromising

141

analysis), as assessed by visual inspection in Mango version 4.1 (RRID: SCR_009603). Sixteen

142

participants were excluded for excessive head motion (> 0.5 mm framewise displacement in >

143

20% of TRs in the run; see

MRI preprocessing

). Our final sample consisted of 87 participants

144

(

age

= 27.0 years,

age

= 6.1 years; 39 female, 47 male, 1 non-response). For self-reported

145

ethnicity, 13% of participants identified as Hispanic. For self-reported race, 62% identified as

146

White, 25% identified as Asian, and 9% identified as Black. Of the 87 participants, three did not

147

provide a response for race, and one did not provide a response for ethnicity. The participant

148

sample was generally highly-educated (24% had some graduate education, 25% had completed

149

college/university, 37% had some college/university, 10% had completed high school, and 2%

150

had not completed high school). Peripheral physiological recordings were inspected for artifacts

151

(see

Peripheral Physiological Recording and Quality Assessment

) and only non-artifactual

152

recordings were included in analyses. Of the 87 participants with usable fMRI data, 59 had

153

artifact-free cardiac data, and 47 had artifact-free respiratory data.

154

All participants were between 18 and 40 years old, right-handed, fluent English

155

speakers, with normal or corrected to normal vision, and no known neurological or psychiatric

156

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

illnesses. Participants were excluded from participating if they were pregnant, claustrophobic or

157

had any metal implants that could cause harm during scanning.

158

Experimental Paradigm.

Participants completed a visual N-back working memory task during

159

fMRI scanning, based on the experimental design of prior studies (Gevins & Cutillo, 1993; J. R.

160

Gray et al., 2003; Kirchner, 1958; van Ast et al., 2016; Wager & Smith, 2003). On each trial, a

161

letter (Q, W, R, S, or T) was presented at the center of the visual field for 1-second followed by

162

a blank screen for 1-second (Fig. 1). Participants were instructed to respond with a button press

163

when the letter on the screen matched the letter presented

trials ago where

was either 1 or

164

3. The task was administered during a single scanning run, across 12 blocks, each consisting of

165

10 trials (120 trials total). Each block began with a 3-second cue indicating the current trial type

166

(1-back or 3-back). 1-back and 3-back blocks were presented in ABBA or BAAB order

167

(counterbalanced across participants) and each block was followed by 25-seconds of resting

168

fixation. Stimuli were classified into 2 cate

gories; a “target” was a stimulus that met the N-back

169

criteria for a match (e.g., N-back either 1 or 3 trials ago), and all other stimuli were classified as

170

“non-targets”. The task had fixed proportions of 20% targets and 80% non-targets. Of the non-

171

targets, 12.5% were classified as lure trials

—i.e., 2-back matches in the 1-back blocks and 2 or

172

4-back matches in the 3-back blocks. The proportion of lure trials was the same for both 1-back

173

and 3-back blocks.

174

The task was administered in MATLAB (The MathWorks; RRID:SCR_001622), using the

175

Psychophysics Toolbox extensions (Kleiner, 2007; RRID:SCR_002881). Stimuli were made

176

visible to participants by projecting them onto a mirror attached to the head coil. Responses

177

were recorded using an MR-compatible button box.

178

To familiarize participants with the task, participants completed a practice N-back task

179

session on a laptop immediately before the scan session. These practice sessions mirrored the

180

task used during scanning, including blocks of 1-back and 3-back trials, for a total of 48 trials.

181

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

MRI Data Acquisition.

Gradient-echo echo-planar imaging BOLD-fMRI was performed on a 7-

182

Tesla (7T) Siemens MRI scanner at the Athinoula A. Martinos Center for Biomedical Imaging at

183

Massachusetts General Hospital (MGH), Boston, MA. The scanner was built by Magnex

184

Scientific (Oxford, UK), with the MRI console, gradient and gradient drivers, and patient table

185

provided by Siemens. Functional images were acquired using GRAPPA-EPI sequence: echo

186

time = 28 ms, repetition time = 2.34 s, flip angle = 75°, slice orientation = transversal (axial),

187

anterior to posterior phase encoding, voxel size = 1.1 mm isotropic, gap between slices = 0 mm,

188

number of axial slices = 123, field of view = 205 × 205 mm

, GRAPPA acceleration factor = 3;

189

echo spacing = 0.82 ms, bandwidth = 1414 Hz per pixel, partial Fourier in the phase encode

190

direction = 7/8. A custom-built 32-channel radiofrequency coil head array was used for

191

reception. Radiofrequency transmission was provided by a detunable band-pass birdcage coil.

192

Structural images were acquired using a T1-weighted EPI sequence, selected so that

193

functional and structural images had similar spatial distortions, which facilitated co-registration

194

and subsequent normalization of data to MNI space (Renvall et al., 2016). Structural scan

195

parameters were: echo time = 22 ms, repetition time = 8.52 s, flip angle = 90°, slice orientation =

196

transversal (axial), voxel size = 1.1 mm isotropic, gap between slices = 0 mm, number of axial

197

slices = 126, field of view = 205 × 205 mm

, GRAPPA acceleration factor = 3; echo spacing =

198

0.82 ms, bandwidth =1414 Hz per pixel, partial Fourier in the phase encode direction = 6/8.

199

MRI Preprocessing.

Functional and structural MRI data were preprocessed using the fmriprep

200

pipeline, version 1.1.2 (Esteban et al., 2019, 2020; RRID: SCR_016216), a Nipype-based tool

201

(Esteban et al., 2020; Gorgolewski et al., 2011; RRID: SCR_002502). Full pipeline details can

202

be found at (https://fmriprep.org/en/1.1.2/workflows.html). Spatial normalization of anatomical

203

data to the 2009c ICBM 152 Nonlinear Asymmetrical template (Fonov et al., 2009; RRID:

204

SCR_008796) was performed through nonlinear registration with the antsRegistration tool of

205

ANTs version 2.1.0 (Avants et al., 2008; RRID: SCR_004757), using brain-extracted versions of

206

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

both T1w (T1-weighted) volume and template. Brain tissue segmentation of cerebrospinal fluid

207

(CSF), white-matter (WM) and gray-matter (GM) was performed on the brain-extracted T1w

208

using FAST in FSL version 5.0.9 (Y. Zhang et al., 2001; RRID: SCR_002823). Functional data

209

were slice time corrected using 3dTshift from AFNI version 16.2.07 (Cox, 1996; RRID:

210

SCR_005927) and motion corrected using MCFLIRT (FSL version 5.0.9 (Jenkinson et al.,

211

2002). This was followed by co-registration to the corresponding T1w using boundary-based

212

registration (Greve & Fischl, 2009) with 9 degrees of freedom, using FLIRT

(FMRIB’s Linear

213

Image Registration Tool; FSL version 5.0.9; Jenkinson et al., 2002; Jenkinson & Smith, 2001).

214

Motion correcting transformations, BOLD-to-T1w transformation, and T1w-to-template (MNI)

215

warp were concatenated and applied in a single step using antsApplyTransforms (ANTs version

216

2.1.0) using Lanczos interpolation. Physiological noise regressors were extracted using the

217

aCompCor method (Muschelli et al., 2014), taking the top five principle components from

218

subject-specific CSF and WM masks, where the masks were generated by thresholding the

219

WM/CSF masks derived from FAST at 99% probability, constraining the CSF mask to the

220

ventricles (using the ALVIN mask; Kempton et al., 2011), and eroding the WM mask using the

221

binary_erosion function in SciPy version 1.6. Frame-wise displacement (Power et al., 2014) was

222

calculated for each functional run using the implementation of Nipype. Many internal operations

223

of fmriprep use Nilearn (Abraham et al., 2014; RRID: SCR_001362), principally within the

224

BOLD-processing workflow. For all participants, the quality of brain masks, tissue segmentation,

225

and MNI registration was visually inspected for errors using the html figures provided by the

226

fmriprep pipeline.

227

fMRI Analysis.

To estimate BOLD signal intensity during the working memory N-back task, a

228

general linear model was applied to preprocessed functional time-series (GLM; FSL version

229

5.0.9). Subject-level models provided beta weight estimates for 1-back and 3-back conditions

230

within each participant, relative to baseline fixation. All trial regressors were aligned to trial onset

231

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

(duration = 1 second), and convolved with a double gamma hemodynamic response function

232

(HRF). Because trials were closely spaced, the convolved 1-back and 3-back regressors

233

approximated a block-based design, meaning that observed PAG responses should be

234

interpreted as an average across the 1-back/3-back block period (as opposed to more time-

235

sensitive responses to the letter presented on each trial). Subject-level nuisance regressors

236

included a run intercept, motion (i.e., translation/rotation in x/y/z planes), aCompCor

237

components (5 cerebral spinal fluid, 5 white matter; Muschelli et al., 2014), a discrete cosine

238

transformation set with a minimum period of 264 seconds, and spike regressors (> 0.5mm

239

framewise displacement (Satterthwaite et al., 2013). To assess multicollinearity variance

240

inflation factors (VIFs), were inspected for 1-back (VIF

mean

= 1.33; VIF

= 0.16) and 3-back

241

regressors (VIF

mean

= 1.48; VIF

= 0.28), and were considered to be within an acceptable range

242

(i.e., < 5; Mumford et al., 2015). Group-level analyses included whole-brain contrasts and

243

analyses within PAG voxels (see

PAG Identification and Alignment

). Whole-brain contrasts

244

computed

-scores relative to 0 for subject-level beta weights after smoothing with a 1 mm

245

FWHM Gaussian kernel. Thresholds for

score maps were corrected for multiple comparisons

246

using threshold-free cluster-enhancement (TFCE)

(α < .05; Smith & Nichols, 2009).

247

PAG Identification and Alignment.

The PAG was identified and aligned across subjects using

248

a semi-automated iterative procedure (for details, see Kragel et al., 2019), which built on

249

manual PAG alignment methods previously used in 7T studies of the PAG (for details, see

250

Satpute et al., 2013). For each subject, a subject-specific mask of the cerebral aqueduct was

251

created by identifying high-variance voxels in the region. Following this, subject-specific PAG

252

masks were created by dilating the aqueduct mask (2 voxels, i.e., 2.2 mm) and masking (a)

253

voxels in the original subject-specific aqueduct mask, (b) voxels not in a subject-specific gray-

254

matter segmentation (> 50% probability), and (c) any voxels outside the target range [MNI:

−22

255

> y >

−42; z > −14]. This created hollow and cylindrical PAG masks for each subject, which

256

were aligned and warped to a group-aligned mask in DARTEL (Ashburner, 2007). These

257

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

transformations were then applied to whole-brain contrast images to align subject-level

258

functional data to a common PAG-aligned space. For all analyses within the PAG, voxels were

259

resampled to 0.5 mm isotropic space (from 1.1 mm isotropic native space), masked with the

260

group-aligned PAG mask, and smoothed 1 mm.

261

Each voxel within the group-aligned PAG mask was tagged according to (a) its radial

262

degree from the PAG midline, (b) its position along the rostral

–caudal PAG axis, and (c) its

263

radial distance from the center of the cylinder.

Figure 2b depicts the “unrolled” PAG, with radial

264

degrees on the x-axis, and rostral-caudal position on the y-

axis. Figure 2c depicts the “unrolled”

265

PAG “sliced” according to radial distance from the center of the cylinder. Rostral–caudal position

266

was calculated using principal components analysis (PCA) on voxelwise x/y/z coordinates, and

267

radial degree and distance were calculated using a geometric transformation of those PCA

268

coordinates, giving degree relative to the volumetric y-axis (at 0°) and cylinder center (at radial

269

distance = 0). For visualization, Figure 2 depicts the functional columns within the PAG (Fig. 2);

270

however, all group-level analyses were performed using either voxelwise data (Figs. 2, 5a), or

271

bilateral ROIs aggregating across radial degrees or the rostral

–caudal axis (Fig. 5b; see

Linear

272

Mixed Effects Model

). The ventral most 90° of the PAG was excluded from analysis, given that it

273

contains other subcortical structures (e.g., dorsal raphe).

274

Linear Mixed-Effects Model.

Voxel beta weights

within the PAG group-aligned mask were

275

submitted to a linear mixed-effects model to summarize the spatial distribution of BOLD signal

276

changes. PAG voxel beta weights were averaged within four ranks of rostral

–caudal position

277

and ten ranks of radial degree (Fig. 5b). Each rostral

–caudal rank consisted of a 3.5 mm

278

segment along the 14 mm rostral

–caudal positional axis, and each radial degree rank consisted

279

of a

bilateral average

.5° sections of the PAG

—i.e., the first radial degree rank included 0-

280

13.5° in left and right PAG, while the last radial degree rank included 121.5-135° in left and right

281

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

PAG. Given this, our mixed effects model assumes bilaterality in any effect along PAG radial

282

degree (an assumption we visually first confirmed by visual inspection of the voxels (Fig. 5a).

283

The mixed effect model was specified to include fixed effects of condition (1-back/3-

284

back), rostral

–caudal rank, PAG degree rank, and all interactions. The model included by-

285

participant random intercepts, and by-participant random slopes for all fixed effects (Barr et al.,

286

2013). Analyses were conducted in R (R version 3.6.2;R Core Team, 2016) using the lmer4

287

package (Bates et al., 2015), with degrees of freedom approximated by the Satterthwaite

288

method, as implemented in the lmerTest package (Kuznetsova et al., 2017). The code was

289

entered as follows:

290

model <- lmer(averaged PAG BOLD ~ condition * rostral

–caudal position * PAG degree +

291

(condition * rostral

–caudal position * PAG degree | participant), control =

292

lmerControl(optimizer = "bobyqa", optCtrl = list(maxfun = 2e5)))

293

Peripheral Physiological Recording and Quality Assessment.

Peripheral physiological

294

measures were collected at 1kHz using an AD Instruments PowerLab data acquisition system

295

with MR-compatible sensors and LabChart software. Data were continuously acquired

296

throughout the entire scan session and partitioned for alignment with fMRI data using

297

experimenter annotations in LabChart and scanner TR events. A piezo-electric pulse transducer

298

(PowerLab, AD Instruments, Colorado Springs, Colorado, USA) measured heart rate from the

299

left index fingertip. A respiratory belt with a piezo-electric transducer (UFI) measured respiratory

300

rate and was placed around the chest at the lower end of the sternum.

301

Physiological time-

series data were visually inspected for quality in Biopac’s

302

AcqKnowledge data acquisition and analysis software (Biopac Systems Inc., Goleta, California,

303

USA) and in custom visualizations using R (R version 3.6.2;R Core Team, 2016) and the

304

ggplot2 package (version 3.2.1; Wickham, 2009). Time-series were classified as high-quality

305

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

(i.e., few or no motion artifacts), noisy (i.e., IBI: motion artifacts affected < 50% of each 1-minute

306

segment; RR: large, frequent motion artifacts or RRs outside of a typical range of 7-20 breaths

307

per min), and unusable (i.e., IBI: containing excessive artifacts with few reliably detectable

308

peaks; RR: containing low amplitudes that could not be distinguished from noise). Among the 87

309

participants with usable fMRI data, cardiac data was high-quality (i.e., all 1-minute segments

310

were sufficiently free of motion to easily identify peaks in the IBI signal) for 59 participants, noisy

311

for 12, unusable for 14, and missing for two. Among the 87 participants with usable fMRI data,

312

respiratory data was high-quality for 47 participants, noisy for 20, unusable for 14, and missing

313

for six. Only high-quality data were used for cardiac (IBI) and respiratory (RR) analyses.

314

Peripheral Physiological Preprocessing and Analysis.

Physiological time-series data were

315

analyzed using MATLAB toolboxes and custom R scripts. Heart rate was calculated using the

316

PhysIO MATLAB toolbox (Kasper et al., 2017a), which used a 0.3 to 9 Hz band-pass filter, and

317

identified peaks using a two-pass process. The first-pass estimated run average heart rate

318

(inverse of IBI), detecting peaks exceeding 40% of normalized amplitude, assuming a minimum

319

peak spacing < 90 beats per minute. First-pass peaks were used to create an averaged

320

template, and the first-pass estimated average heart rate was used as a prior to detect peaks on

321

the second-pass (for more details, see Kasper et al., 2017b). PhysIO pipeline-detected peaks

322

were compared with peaks identified in Biopac by trained coders. Discrepancies between the

323

two methods were rare (0.5% of peaks across all runs and participants), and occurred very

324

rarely in the dataset scored by a more experienced coder (0.323% of peaks), compared to the

325

dataset scored by a less experienced coder (0.878% of peaks). To reduce drift in the cardiac

326

signal, heart rates were smoothed with a 6-second rolling average and down-sampled into scan

327

repetition time. Upon completion of pre-processing, we then converted heart rate into IBI, as IBI

328

displays a more linear relationship with underlying autonomic cardiac inputs than heart rate

329

(Berntson et al., 1995). After converting HR to IBI, cardiac data were analyzed exclusively using

330

IBI. Respiratory rates were calculated using custom R scripts. A 1 Hz low-pass filter was applied

331

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

to the respiratory time-series, and local peaks were identified in a sliding 500 ms window. Small

332

excursions (those < 0.5 SD of average respiratory belt tension across the entire run) were

333

deemed too small to meet criteria as a respiratory peak or trough. To maintain consistency

334

between pre-processing of IBI and RR, RRs were also smoothed with a 6-second rolling

335

average and down-sampled into scan repetition time. Down-sampling IBI and RR into scan

336

repetition time resulted in a total of 263 captured instances of physiological activity for each

337

variable (IBI, RR) in every participant; averages were then computed for each participant and for

338

each resting fixation, 1-back and 3-back task epoch (see

Experimental Paradigm

). Using the

339

averaged task epoch values, change scores were computed as the difference in physiological

340

activity (IBI, RR) between resting fixation and 1-back and 3-back task epoch.

341

Results

342

Performance and reaction time increased in the 1-back vs. 3-back working memory task.

343

As expected, and consistent with the earlier work using a small subset of this sample,

344

performance (i.e., hit rates) were higher for 1-back trials (

= 0.99,

= 0.02) than for 3-back

345

trials (

= 0.89,

= 0.08;

(85)

= 12.18,

< 1e

−16

= 1.31). Likewise, reaction times for correct

346

responses were higher for 1-back trials (

= 0.76 sec,

= 0.16 sec) than 3-back trials (

347

0.99 sec,

= 0.22 sec;

(85)

= 13.75,

< 1e

−16

= 1.48). Reaction times for incorrect

348

responses did not significantly differ between 1-back and 3-back conditions.

349

Peripheral physiology was affected by 1-back and 3-back working memory conditions.

350

Group-level changes in peripheral physiology (IBI and RR) were compared to resting baseline

351

during 1-back and 3-back task epochs. Although previous analyses using a subset of the

352

present sample did not observe any change in peripheral physiology from baseline (Kragel et

353

al., 2019), we did observe changes in both 1-back and 3-back conditions (Fig. 3). In the 1-back

354

condition, across participants, IBI increased (i.e., cardio-deceleration occurred) and reparatory

355

rate increased compared to resting baseline (IBI:

(58)

= 3.64,

< 0.0006,

= 0.47; RR:

(46)

356

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

4.75,

< 1e

−5

= 0.69). In the 3-back condition IBI decreased (i.e., cardio-acceleration

357

occurred) and respiratory rate increased compared to resting baseline (IBI:

(58)

= 3.82,

358

0.0003,

= 0.50; RR:

(46)

= 9.07,

< 1e

−12

= 1.32). Relative to the 1-back condition, 3-back

359

IBI was shorter (

(58)

= 5.21,

< 1e

−6

= 0.68), and 3-back respiration rate was faster (

(46)

360

6.07,

< 1e

−7

= 0.89). Thus, both mild and moderately difficult cognitive tasks elicited

361

physiological changes: they both elicited increased respiratory rates, and they elicited opposite

362

cardiac responses, with the mildly difficult 1-back reducing heart rate, and moderately difficult 3-

363

back increasing it.

364

Whole brain contrast elicits expected pattern of cortical results, and widespread

365

brainstem activity in 3-back condition.

Whole-brain group-level analyses for 1-back, 3-back,

366

and 3-back > 1-back contrasts were performed to ensure that cortical results were consistent

367

with observations in prior work (Kragel et al., 2019). Group-level analyses were performed on

368

first-level contrasts after applying a transformation to align the subject-specific PAG masks (see

369

PAG Identification and Alignment, in Methods

), The 3-back > 1-back contrast was consistent

370

with prior work, with bilateral responses observed in the inferior frontal junction, the intraparietal

371

sulcus, dorsal anterior insula, supplemental motor area, posterior parietal cortex, and

372

dorsolateral prefrontal cortex (Fig. 4a). Notably, unlike in Kragel et al., (2019), we did not

373

observe a supra-threshold response in PAG for the 3-back > 1-back contrast. However, the 3-

374

back > baseline contrast did elicit supra-threshold activity throughout the brainstem, including

375

within the PAG (Fig. 4b).

376

1-back and 3-back (vs. baseline) contrasts elicited similar patterns of BOLD response

377

across the cortex (Fig. 4a), meaning that the 3-back > 1-back contrast estimates were generally

378

caused by a difference in the intensity of the BOLD signal increase/decrease between the

379

conditions. In other words, a similar set of brain regions responded to 1-back and 3-back

380

conditions, but the magnitude of increase/decrease was greater for the 3-back condition (with

381

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

the exceptions of posterior parietal cortex and posterior insula, where no supra-threshold

382

differences between 1-back and resting baseline were observed; Fig. 4a). The fact that the 3-

383

back condition elicited brainstem and PAG responses that exceeded baseline, but that did not

384

differ from the 1-back response, suggested the presence of a subthreshold PAG response in the

385

1-back < baseline contrast as well. Given that the cortical response in 1-back and 3-back

386

conditions were similar in their topographic organization, but different in their intensities, it was

387

possible that a similar topographic similarity would be observed in the PAG as well. Extracting

388

and analyzing the voxelwise topography of the BOLD response in PAG allowed us to test this

389

hypothesis.

390

The voxelwise topography of PAG BOLD response for 1-back and 3-back working

391

memory conditions is strongly correlated.

Voxelwise

-scores for the 1-back and 3-back <

392

baseline contrasts were plotted on a 2D surface to visualize the topography of the PAG BOLD

393

response (Fig. 5a). Visual inspection of these results suggested that the 1-back and 3-back

394

contrasts did elicit a PAG BOLD response that was similar spatially, but different in intensity, as

395

in the cortical pattern of results. Supporting this topographic similarity, voxelwise

-scores were

396

highly correlated between 1-back and 3-back contrasts,

= 0.76 [0.75, 0.77; bootstrapped 95%

397

CI, 10,000 samples; Fig. 5a]. To model the topographic pattern of PAG response, we assigned

398

two ordinal ranks to each PAG voxel: a rostral

–caudal rank (4 levels, from caudal to rostral

399

poles of the PAG; Fig. 5b), and a radial degree rank (10 levels, from dorsomedial to

400

ventrolateral PAG, averaging bilaterally; Fig. 5b; see

Linear Mixed Effects Model

in Methods

401

These ordinal ranks, along with N-back condition (1-back, 3-back) and all interactions, were

402

submitted to a linear mixed effects model (including subject-level intercepts and all random

403

slopes; Barr et al., 2013) to model the spatial topography of the PAG BOLD beta weights in both

404

N-back conditions.

405

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

The main effect of N-back condition was non-significant, [

(1, 86)

= 2.12,

< 0.15],

406

consistent with failure to observe any 3-back > 1-back difference within the PAG in the whole-

407

brain contrast (Fig. 4b). The main effect of rostral

–caudal rank was significant, [

(1, 86)

= 52.61,

408

< 1e

-9

], but was qualified by a two-way interaction with N-back condition, [

(1, 86)

= 7.88,

409

.007], suggesting that the linear increase along the rostral

–caudal axis of the PAG occurred in

410

both N-back conditions, but was greater in the 3-back condition than in the 1-back condition

411

(Fig. 5c; y-axis increases from pink to green categories). The main effect of degree rank was

412

also significant, [

(1, 86)

= 41.92,

< 1e

-8

], and was also qualified by a two-way interaction with N-

413

back condition [

(1, 86)

= 4.62,

< 0.035], suggesting that the linear increase from dorsomedial to

414

ventrolateral PAG occurred in both N-back conditions, but was greater in the 3-back condition

415

than in the 1-back condition (Fig. 5c; y-axis increases from left to right along the x-axis). The

416

magnitude of the linear increase from dorsomedial to ventrolateral PAG also increased from the

417

caudal to rostral end of the PAG (Fig. 5c), as indicated by a two-way rostral

–caudal rank by

418

degree rank interaction, [

(1, 86)

= 26.03,

< 1e

-5

], but two-way interaction was not qualified by

419

an additional 3-way interaction with N-back condition [

(1, 86)

= 3.64,

> .05], meaning that

420

spatial topography of the BOLD response in 1-back and 3-back conditions was similar along the

421

dimensions of the rostral

–caudal axis and radial degrees, but the conditions did differ in the

422

magnitude of change BOLD response along these dimensions (Fig. 5c).

423

Task-elicited changes in peripheral physiological response correlate with PAG voxels,

424

but the voxelwise topography of correlation is weakly related to task-elicited BOLD

425

responses.

The PAG is implicated in physiological regulation, and we expected that task-

426

elicited changes in PAG BOLD signal would correlate with changes (from baseline) in cardiac

427

interbeat interval (IBI) and respiratory rate (RR). To identify voxels that were associated with the

428

magnitude of task-elicited physiological change from baseline, in each voxel subject-level beta

429

weights for 1-back and 3-back > baseline contrasts were correlated with subject-level changes

430

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

in IBI and RR (Fig. 6a). The voxelwise topography of the correlation coefficients was less clearly

431

distinguishable along the expected organizational pattern of functional PAG columns than in the

432

1-back and 3-back > baseline contrasts (Fig. 5a). To directly compare these two patterns

433

voxelwise topography (i.e., voxelwise correlations with physiological change, Fig. 6a, and

434

voxelwise task-elicited BOLD increases/decreases, Fig. 5a), we plotted the estimates against

435

each other (Fig. 6b). We observed moderate correlations across PAG voxels between task-

436

elicited BOLD and its correlation with changes in cardiac interbeat interval (Fig. 6b, top), but

437

weaker or negative correlations with change in respiratory rate (Fig. 6b, bottom). In all cases,

438

the similarity in topography was much weaker than the similarity between PAG BOLD signal

439

elicited in the 1-back and 3-back tasks.

440

Discussion

441

Our results suggest that, in humans, non-painful and non-threatening tasks that elicit

442

mild physiological changes elicit a spatially organized pattern of BOLD response in PAG.

443

Further, the moderate and mild working memory demands of the 3-back and 1-back tasks

444

elicited nearly identical patterns of BOLD response across the PAG

—the 3-back response

445

magnified the intensity of response, but did not change its voxelwise topography. In both 1-back

446

and 3-back conditions, BOLD increases were observed along the entire rostral

–caudal axis in

447

vlPAG, suggesting that the BOLD response to the cognitive task is organized along previously

448

the known anatomical boundaries of PAG functional columns (for review, see Carrive & Morgan,

449

2012).

450

It was surprising that light cognitive exertion would elicit a localized BOLD response in

451

vlPAG, as previous studies have associated the function of vlPAG with survival-related stimuli,

452

including stimuli eliciting

“freezing” behavior (in rats; Carrive et al., 1997; Morgan et al., 1998),

453

and anticipated breathlessness or subcutaneous pain in humans (Faull et al., 2016; Tinoco

454

Mendoza et al., 2023). Given that the present work observed a clearly defined pattern of BOLD

455

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

increase in vlPAG in response to a task approximately as easy as remembering a phone

456

number, we posit that the PAG plays a more general functional role than just supporting

457

survival-related behavior

(e.g., housing “fear” or “survival” circuits; LeDoux & Daw, 2018; B. Lu

458

et al., 2022; McNaughton & Corr, 2018; Panksepp, 2011; B. A. Silva et al., 2016; Vázquez-León

459

et al., 2023)

—consistent with accumulating evidence that the PAG helps coordinate and

460

regulate the internal (visceromotor) systems of the body (for similar perspectives, see

461

Benarroch, 1993, 2001; Faull et al., 2019). In other words, although prior work has observed

462

associations between PAG activity

“fight-or-flight” or “freezing” behavior in the context of threat,

463

the overall function of the PAG may be better described by its support for autonomic regulation

464

many situations

, even in contexts where survival-related threat is absent, like the working

465

memory context used in the present work.

466

Within the past decade, significant progress has been made in imaging the human PAG

467

(e.g., Coker-Appiah et al., 2013; Hermans et al., 2013; Lindner et al., 2015; Tinoco Mendoza et

468

al., 2023; Wang et al., 2022; C. Weis et al., 2020; C. N. Weis et al., 2022; Zhou et al., 2021);

469

however, because the PAG has been associated with survival-related behavior, most

470

experiments continue to explore its function in these threat-based contexts. By contrast, the N-

471

back task used in the present work involves cognitive demands that are within the range of what

472

is experienced in daily life, and provided a context in which to probe PAG BOLD response

473

during less affectively evocative task conditions. Successful performance in a N-back working

474

memory task requires integrating exteroceptive (e.g., visual stimuli) and interoceptive sensory

475

signals (e.g., signals of energy status from the body), maintaining those signals across trials

476

(i.e., working memory), mobilizing energy to meet changing task demands (e.g., as reflected in

477

changes in cardiorespiratory reactivity (Berntson et al., 2017; Dampney et al., 2013; Hoemann

478

et al., 2020; Thayer et al., 2012), maintaining goal-directed behavior (e.g., allocating attentional

479

resources), and finally, generating and executing a motor plan (e.g., pushing a button to make a

480

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

response). These brain functions support successful working memory performance, but they

481

also support other complex, goal-directed sensorimotor behaviors. Indeed, these are the basic

482

operations that are needed to perform many laboratory psychological tasks and engage in real-

483

world motivated performance (i.e., situations that are goal-relevant to the performer, require

484

instrumental cognitive responses, and are active rather than passive (Blascovich & Mendes,

485

2000). From this perspective, the working memory task is a motivated performance task, similar

486

to other motivated performance tasks that have elicited a PAG BOLD response under more

487

stressful conditions (e.g., social stress during preparation for public speaking; Wager, Waugh, et

488

al., 2009; Wager, van Ast, et al., 2009; Theriault et al., under review). The N-back task, then, in

489

addition to being regarded as a prototypical

‘cognitive’ task, can also be considered as a task

490

that drives a brain to mobilize resources to meet task demands

—i.e., to engage in allostatic

491

regulation (Barrett, 2017; Sennesh et al., 2022). The PAG is anatomically well-positioned to

492

contribute to this allostatic function

—especially given the its anatomical position at the

493

convergence of ascending viscerosensory signals from the body and descending visceromotor

494

signals for controlling the body (for review, see Carrive & Morgan, 2012).

495

In sum, our observation that

both mild and moderately challenging

cognitive tasks

496

elicited near similar patterns of PAG BOLD response is consistent with the emerging

497

perspective that the PAG serves a general integrative function

—as opposed to a survival-

498

specific function

—and may support the integration of bottom-up sensory signals from the body

499

with top-down visceromotor control signals from the brain (for review, see Bandler et al., 2000;

500

Benarroch, 2012; Gianaros & Wager, 2015; Koutsikou et al., 2017).

This perspective was

501

hindered by the rarity of studies probing PAG function outside of threat-related contexts, and so

502

future work would benefit from exploring PAG function in naturalistic conditions, and in other

503

experimental contexts, like the N-back task, that drive the brain to mobile resources for action.

504

Such studies could identify subcortical responses that might be missed in experimental contexts

505

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

References

Abraham, A., Pedregosa, F., Eickenberg, M., Gervais, P., Mueller, A., Kossaifi, J., Gramfort, A.,

Thirion, B., & Varoquaux, G. (2014). Machine learning for neuroimaging with scikit-learn.

Frontiers in Neuroinformatics

. https://doi.org/10.3389/fninf.2014.00014

Abrahams, V. C., Hilton, S. M., & Zbrożyna, A. (1960). Active muscle vasodilatation produced

by stimulation of the brain stem: Its significance in the defence reaction.

The Journal of

Physiology

154

(3), 491

–513.

Amano, K., Tanikawa, T., Iseki, H., Kawabatake, H., Notani, M., Kawamura, H., & Kitamura, K.

(1978). Single Neuron Analysis of the Human Midbrain Tegmentum.

Stereotactic and

Functional Neurosurgery

–4), 66–78. https://doi.org/10.1159/000102402

Ashburner, J. (2007). A fast diffeomorphic image registration algorithm.

NeuroImage

(1), 95

–

113. https://doi.org/10.1016/j.neuroimage.2007.07.007

Avants, B. B., Epstein, C. L., Grossman, M., & Gee, J. C. (2008). Symmetric diffeomorphic

image registration with cross-correlation: Evaluating automated labeling of elderly and

neurodegenerative brain.

Medical Image Analysis

(1), 26

–41.

https://doi.org/10.1016/j.media.2007.06.004

Bandler, R. (1988). Brain mechanisms of aggression as revealed by electrical and chemical

stimulation: Suggestion of a central role for the midbrain periaqueductal grey region. In

Progress in psychobiology and physiological psychology, Vol. 13

(pp. 67

–154).

Academic Press.

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Bandler, R., & Carrive, P. (1988). Integrated defence reaction elicited by excitatory amino acid

microinjection in the midbrain periaqueductal grey region of the unrestrained cat.

Brain

Research

439

–2), 95–106. https://doi.org/10.1016/0006-8993(88)91465-5

Bandler, R., Carrive, P., & Depaulis, A. (1991). Emerging Principles of Organization of the

Midbrain Periaqueductal Gray Matter. In A. Depaulis & R. Bandler (Eds.),

The Midbrain

Periaqueductal Gray Matter: Functional, Anatomical, and Neurochemical Organization

(pp. 1

–8). Springer US. https://doi.org/10.1007/978-1-4615-3302-3_1

Bandler, R., Carrive, P., & Zhang, S. P. (1991a). Integration of somatic and autonomic reactions

within the midbrain periaqueductal grey: Viscerotopic, somatotopic and functional

organization. In

Progress in Brain Research

(Vol. 87, pp. 269

–305). Elsevier.

https://doi.org/10.1016/S0079-6123(08)63056-3

Bandler, R., Carrive, P., & Zhang, S. P. (1991b). Integration of somatic and autonomic reactions

within the midbrain periaqueductal grey: Viscerotopic, somatotopic and functional

organization.

Progress in Brain Research

, 269

–305. https://doi.org/10.1016/s0079-

6123(08)63056-3

Bandler, R., & Depaulis, A. (1991). Midbrain Periaqueductal Gray Control of Defensive Behavior

in the Cat and the Rat. In A. Depaulis & R. Bandler (Eds.),

The Midbrain Periaqueductal

Gray Matter: Functional, Anatomical, and Neurochemical Organization

(pp. 175

–198).

Springer US. https://doi.org/10.1007/978-1-4615-3302-3_11

Bandler, R., Keay, K. A., Floyd, N., & Price, J. (2000). Central circuits mediating patterned

autonomic activity during active vs. Passive emotional coping.

Brain Research Bulletin

(1), 95

–104. https://doi.org/10.1016/S0361-9230(00)00313-0

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random effects structure for

confirmatory hypothesis testing: Keep it maximal.

Journal of Memory and Language

(3), 255

–278. https://doi.org/10.1016/j.jml.2012.11.001

Barrett, L. F. (2017). The theory of constructed emotion: An active inference account of

interoception and categorization.

Social Cognitive and Affective Neuroscience

(1), 1

–

23. https://doi.org/10.1093/scan/nsw154

Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting Linear Mixed-Effects Models

Using lme4.

Journal of Statistical Software

(1), Article 1.

https://doi.org/10.18637/jss.v067.i01

Benarroch, E. E. (1993). The central autonomic network: Functional organization, dysfunction,

and perspective.

Mayo Clinic Proceedings

(10), 988

–1001.

https://doi.org/10.1016/s0025-6196(12)62272-1

Benarroch, E. E. (2001). Pain-autonomic interactions: A selective review.

Clinical Autonomic

Research

(6), 343

–349. https://doi.org/10.1007/BF02292765

Benarroch, E. E. (2012). Periaqueductal gray: An interface for behavioral control.

Neurology

(3), 210

–217. https://doi.org/10.1212/WNL.0b013e31823fcdee

Berntson, G. G., Cacioppo, J. T., & Quigley, K. S. (1995). The metrics of cardiac chronotropism:

Biometric perspectives.

Psychophysiology

(2), 162

–171.

https://doi.org/10.1111/j.1469-8986.1995.tb03308.x

Berntson, G. G., Quigley, K. S., Norman, G. J., & Lozano, D. L. (2017). Cardiovascular

psychophysiology. In

Handbook of psychophysiology, 4th ed

(pp. 183

–216). Cambridge

University Press.

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Blascovich, J., & Mendes, W. B. (2000). Challenge and threat appraisals: The role of affective

cues. In

Feeling and thinking: The role of affect in social cognition

(pp. 59

–82).

Cambridge University Press.

Blok, B. F. M., & Holstege, G. (1998). The central nervous system control of micturition in cats

and humans.

Behavioural Brain Research

(2), 119

–125.

https://doi.org/10.1016/S0166-4328(97)00184-8

Blok, B. F., Willemsen, A. T., & Holstege, G. (1997). A PET study on brain control of micturition

in humans.

Brain: A Journal of Neurology

120 ( Pt 1)

, 111

–121.

https://doi.org/10.1093/brain/120.1.111

Carere, C., Ball, G. F., & Balthazart, J. (2007). Sex differences in projections from preoptic area

aromatase cells to the periaqueductal gray in Japanese quail.

The Journal of

Comparative Neurology

500

(5), 894

–907. https://doi.org/10.1002/cne.21210

Carrive, P. (1993). The periaqueductal gray and defensive behavior: Functional representation

and neuronal organization.

Behavioural Brain Research

–2), 27–47.

https://doi.org/10.1016/0166-4328(93)90088-8

Carrive, P., & Bandler, R. (1991). Viscerotopic organization of neurons subserving hypotensive

reactions within the midbrain periaqueductal grey: A correlative functional and

anatomical study.

Brain Research

541

(2), 206

–215. https://doi.org/10.1016/0006-

8993(91)91020-2

Carrive, P., Dampney, R. A., & Bandler, R. (1987). Excitation of neurones in a restricted portion

of the midbrain periaqueductal grey elicits both behavioural and cardiovascular

components of the defence reaction in the unanaesthetised decerebrate cat.

Neuroscience Letters

(3), 273

–278. https://doi.org/10.1016/0304-3940(87)90395-8

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Carrive, P., Leung, P., Harris, J., & Paxinos, G. (1997). Conditioned fear to context is associated

with increased Fos expression in the caudal ventrolateral region of the midbrain

periaqueductal gray.

Neuroscience

(1), 165

–177. https://doi.org/10.1016/S0306-

4522(97)83047-3

Carrive, P., & Morgan, M. M. (2012). Periaqueductal Gray. In G. Paxinos & J. K. Mai (Eds.),

The

Human Nervous System (Second Edition)

(pp. 367

–400). Elsevier.

https://doi.org/10.1016/B978-012547626-3/50013-2

Chen, X.-M., Nishi, M., Taniguchi, A., Nagashima, K., Shibata, M., & Kanosue, K. (2002). The

caudal periaqueductal gray participates in the activation of brown adipose tissue in rats.

Neuroscience Letters

331

(1), 17

–20. https://doi.org/10.1016/S0304-3940(02)00757-7

Coker-Appiah, D. S., White, S. F., Clanton, R., Yang, J., Martin, A., & Blair, R. J. R. (2013).

Looming animate and inanimate threats: The response of the amygdala and

periaqueductal gray.

Social Neuroscience

(6), 621

–630.

https://doi.org/10.1080/17470919.2013.839480

Cox, R. W. (1996). AFNI: Software for Analysis and Visualization of Functional Magnetic

Resonance Neuroimages.

Computers and Biomedical Research

(3), 162

–173.

https://doi.org/10.1006/cbmr.1996.0014

Dampney, R. A. L., Furlong, T. M., Horiuchi, J., & Iigaya, K. (2013). Role of dorsolateral

periaqueductal grey in the coordinated regulation of cardiovascular and respiratory

function.

Autonomic Neuroscience

175

–2), 17–25.

https://doi.org/10.1016/j.autneu.2012.12.008

Depaulis, A., Keay, K. A., & Bandler, R. (1992). Longitudinal neuronal organization of defensive

reactions in the midbrain periaqueductal gray region of the rat.

Experimental Brain

Research

(2), 307

–318. https://doi.org/10.1007/BF00227243

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Esteban, O., Markiewicz, C. J., Blair, R. W., Moodie, C. A., Isik, A. I., Erramuzpe, A., Kent, J. D.,

Goncalves, M., DuPre, E., Snyder, M., Oya, H., Ghosh, S. S., Wright, J., Durnez, J.,

Poldrack, R. A., & Gorgolewski, K. J. (2019). fMRIPrep: A robust preprocessing pipeline

for functional MRI.

Nature Methods

(1), 111

–116. https://doi.org/10.1038/s41592-018-

0235-4

Esteban, O., Markiewicz, C. J., Goncalves, M., DuPre, E., Kent, J. D., Salo, T., Ciric, R.,

Pinsard, B., Blair, R. W., Poldrack, R. A., & Gorgolewski, K. J. (2020).

fMRIPrep: A

robust preprocessing pipeline for functional MRI

[Computer software]. Zenodo.

https://doi.org/10.5281/zenodo.4252786

Fanselow, M. S. (1991). The Midbrain Periaqueductal Gray as a Coordinator of Action in

Response to Fear and Anxiety. In A. Depaulis & R. Bandler (Eds.),

The Midbrain

Periaqueductal Gray Matter

(pp. 151

–173). Plenum Publishing Corp.

https://doi.org/10.1007/978-1-4615-3302-3_10

Fanselow, M. S. (1994). Neural organization of the defensive behavior system responsible for

fear.

Psychonomic Bulletin & Review

(4), 429

–438.

https://doi.org/10.3758/BF03210947

Fardin, V., Oliveras, J.-L., & Besson, J.-M. (1984). A reinvestigation of the analgesic effects

induced by stimulation of the periaqueductal gray matter in the rat. I. The production of

behavioral side effects together with analgesia.

Brain Research

306

(1), 105

–123.

https://doi.org/10.1016/0006-8993(84)90360-3

Faull, O. K., Jenkinson, M., Ezra, M., & Pattinson, K. T. (2016). Conditioned respiratory threat in

the subdivisions of the human periaqueductal gray.

eLife

, e12047.

https://doi.org/10.7554/eLife.12047

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Faull, O. K., Subramanian, H. H., Ezra, M., & Pattinson, K. T. S. (2019). The midbrain

periaqueductal gray as an integrative and interoceptive neural structure for breathing.

Neuroscience & Biobehavioral Reviews

, 135

–144.

https://doi.org/10.1016/j.neubiorev.2018.12.020

Fonov, V., Evans, A., McKinstry, R., Almli, C., & Collins, D. (2009). Unbiased nonlinear average

age-appropriate brain templates from birth to adulthood.

NeuroImage

, S102.

https://doi.org/10.1016/S1053-8119(09)70884-5

Gevins, A., & Cutillo, B. (1993). Spatiotemporal dynamics of component processes in human

working memory.

Electroencephalography and Clinical Neurophysiology

(3), 128

–

143. https://doi.org/10.1016/0013-4694(93)90119-g

Gianaros, P. J., & Wager, T. D. (2015). Brain-Body Pathways Linking Psychological Stress and

Physical Health.

Current Directions in Psychological Science

, 313

–321.

https://doi.org/10.1177/0963721415581476

Gorgolewski, K., Burns, C. D., Madison, C., Clark, D., Halchenko, Y. O., Waskom, M. L., &

Ghosh, S. S. (2011). Nipype: A Flexible, Lightweight and Extensible Neuroimaging Data

Processing Framework in Python.

Frontiers in Neuroinformatics

https://doi.org/10.3389/fninf.2011.00013

Gray, J. R., Chabris, C. F., & Braver, T. S. (2003). Neural mechanisms of general fluid

intelligence.

Nature Neuroscience

(3), 316

–322. https://doi.org/10.1038/nn1014

Gray, M. A., Rylander, K., Harrison, N. A., Wallin, B. G., & Critchley, H. D. (2009). Following

One’s Heart: Cardiac Rhythms Gate Central Initiation of Sympathetic Reflexes.

Journal

of Neuroscience

(6), 1817

–1825. https://doi.org/10.1523/JNEUROSCI.3363-08.2009

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Green, A. L., Wang, S., Owen, S. L. F., Xie, K., Liu, X., Paterson, D. J., Stein, J. F., Bain, P. G.,

& Aziz, T. Z. (2005). Deep brain stimulation can regulate arterial blood pressure in

awake humans.

NeuroReport

(16), 1741

–1745.

https://doi.org/10.1097/01.wnr.0000183904.15773.47

Greve, D. N., & Fischl, B. (2009). Accurate and robust brain image alignment using boundary-

based registration.

NeuroImage

(1), 63

–72.

https://doi.org/10.1016/j.neuroimage.2009.06.060

Gross, C. T., & Canteras, N. S. (2012). The many paths to fear.

Nature Reviews Neuroscience

(9), Article 9. https://doi.org/10.1038/nrn3301

Haakenson, C. M., Balthazart, J., & Ball, G. F. (2020). Effects of Inactivation of the

Periaqueductal Gray on Song Production in Testosterone-Treated Male Canaries

(Serinus canaria).

eNeuro

(4), ENEURO.0048-20.2020.

https://doi.org/10.1523/ENEURO.0048-20.2020

Hermans, E. J., Henckens, M. J. A. G., Roelofs, K., & Fernández, G. (2013). Fear bradycardia

and activation of the human periaqueductal grey.

NeuroImage

, 278

–287.

https://doi.org/10.1016/j.neuroimage.2012.10.063

Hoemann, K., Khan, Z., Feldman, M. J., Nielson, C., Devlin, M., Dy, J., Barrett, L. F.,

Wormwood, J. B., & Quigley, K. S. (2020). Context-aware experience sampling reveals

the scale of variation in affective experience.

Scientific Reports

(1), Article 1.

https://doi.org/10.1038/s41598-020-69180-y

Hohenschurz-

Schmidt, D. J., Calcagnini, G., Dipasquale, O., Jackson, J. B., Medina, S., O’Daly,

O., O’Muircheartaigh, J., de Lara Rubio, A., Williams, S. C. R., McMahon, S. B.,

Makovac, E., & Howard, M. A. (2020). Linking Pain Sensation to the Autonomic Nervous

System: The Role of the Anterior Cingulate and Periaqueductal Gray Resting-State

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Networks.

Frontiers in Neuroscience

https://www.frontiersin.org/articles/10.3389/fnins.2020.00147

Holstege, G., Kerstens, L., Moes, M. C., & VanderHorst, V. G. J. M. (1997). Evidence for a

periaqueductal gray

–nucleus retroambiguus–spinal cord pathway in the rat.

Neuroscience

(2), 587

–598. https://doi.org/10.1016/S0306-4522(97)00061-4

Jansen, A. S. P., Farkas, E., Mac Sams, J., & Loewy, A. D. (1998). Local connections between

the columns of the periaqueductal gray matter: A case for intrinsic neuromodulation.

Brain Research

784

(1), 329

–336. https://doi.org/10.1016/S0006-8993(97)01293-6

Jenkinson, M., Bannister, P., Brady, M., & Smith, S. (2002). Improved Optimization for the

Robust and Accurate Linear Registration and Motion Correction of Brain Images.

NeuroImage

(2), 825

–841. https://doi.org/10.1006/nimg.2002.1132

Jenkinson, M., & Smith, S. (2001). A global optimisation method for robust affine registration of

brain images.

Medical Image Analysis

(2), 143

–156. https://doi.org/10.1016/S1361-

8415(01)00036-6

Kasper, L., Bollmann, S., Diaconescu, A. O., Hutton, C., Heinzle, J., Iglesias, S., Hauser, T. U.,

Sebold, M., Manjaly, Z.-M., Pruessmann, K. P., & Stephan, K. E. (2017a). The PhysIO

Toolbox for Modeling Physiological Noise in fMRI Data.

Journal of Neuroscience

Methods

276

, 56

–72. https://doi.org/10.1016/j.jneumeth.2016.10.019

Kasper, L., Bollmann, S., Diaconescu, A. O., Hutton, C., Heinzle, J., Iglesias, S., Hauser, T. U.,

Sebold, M., Manjaly, Z.-M., Pruessmann, K. P., & Stephan, K. E. (2017b). The PhysIO

Toolbox for Modeling Physiological Noise in fMRI Data.

Journal of Neuroscience

Methods

276

, 56

–72. https://doi.org/10.1016/j.jneumeth.2016.10.019

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Kempton, M. J., Underwood, T. S. A., Brunton, S., Stylios, F., Schmechtig, A., Ettinger, U.,

Smith, M. S., Lovestone, S., Crum, W. R., Frangou, S., Williams, S. C. R., & Simmons,

A. (2011). A comprehensive testing protocol for MRI neuroanatomical segmentation

techniques: Evaluation of a novel lateral ventricle segmentation method.

NeuroImage

(4), 1051

–1059. https://doi.org/10.1016/j.neuroimage.2011.06.080

Kirchner, W. K. (1958). Age differences in short-term retention of rapidly changing information.

Journal of Experimental Psychology

(4), 352

–358. https://doi.org/10.1037/h0043688

Kleiner, M. (2007).

What’s new in Psychtoolbox-3?

89.

Knight, Y. E., & Goadsby, P. J. (2001). The periaqueductal grey matter modulates

trigeminovascular input: A role in migraine?11Presented in part at the Xth International

Headache Congress, Barcelona, Spain, 24

–26 June 1999 [Knight and Goadsby,

Cephalalgia 19 (1999) 315].

Neuroscience

106

(4), 793

–800.

https://doi.org/10.1016/S0306-4522(01)00303-7

Koutsikou, S., Apps, R., & Lumb, B. M. (2017). Top down control of spinal sensorimotor circuits

essential for survival.

The Journal of Physiology

595

(13), 4151

–4158.

https://doi.org/10.1113/JP273360

Kragel, P. A., Bianciardi, M., Hartley, L., Matthewson, G., Choi, J.-K., Quigley, K. S., Wald, L. L.,

Wager, T. D., Feldman Barrett, L., & Satpute, A. B. (2019). Functional Involvement of

Human Periaqueductal Gray and Other Midbrain Nuclei in Cognitive Control.

The

Journal of Neuroscience

(31), 6180

–6189. https://doi.org/10.1523/JNEUROSCI.2043-

18.2019

Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017). lmerTest Package: Tests in

Linear Mixed Effects Models.

Journal of Statistical Software

(1), Article 1.

https://doi.org/10.18637/jss.v082.i13

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

LeDoux, J., & Daw, N. D. (2018). Surviving threats: Neural circuit and computational

implications of a new taxonomy of defensive behaviour.

Nature Reviews Neuroscience

(5), Article 5. https://doi.org/10.1038/nrn.2018.22

Li, Y.-Q., Takada, M., Shinonaga, Y., & Mizuno, N. (1993). Direct projections from the midbrain

periaqueductal gray and the dorsal raphe nucleus to the trigeminal sensory complex in

the rat.

Neuroscience

(2), 431

–443. https://doi.org/10.1016/0306-4522(93)90264-G

Lindner, K., Neubert, J., Pfannmöller, J., Lotze, M., Hamm, A. O., & Wendt, J. (2015). Fear-

potentiated startle processing in humans: Parallel fMRI and orbicularis EMG assessment

during cue conditioning and extinction.

International Journal of Psychophysiology

(3,

Part 2), 535

–545. https://doi.org/10.1016/j.ijpsycho.2015.02.025

Linnman, C., Moulton, E. A., Barmettler, G., Becerra, L., & Borsook, D. (2012). Neuroimaging of

the periaqueductal gray: State of the field.

NeuroImage

(1), 505

–522.

https://doi.org/10.1016/j.neuroimage.2011.11.095

Lovick, T. A. (1992). Inhibitory modulation of the cardiovascular defence response by the

ventrolateral periaqueductal grey matter in rats.

Experimental Brain Research

(1),

133

–139. https://doi.org/10.1007/BF00229010

Lu, B., Fan, P., Wang, Y., Dai, Y., Xie, J., Yang, G., Mo, F., Xu, Z., Song, Y., Liu, J., & Cai, X.

(2022). Neuronal Electrophysiological Activities Detection of Defense Behaviors Using

an Implantable Microelectrode Array in the Dorsal Periaqueductal Gray.

Biosensors

(4), 193. https://doi.org/10.3390/bios12040193

Lu, J., Sherman, D., Devor, M., & Saper, C. B. (2006). A putative flip

–flop switch for control of

REM sleep.

Nature

441

(7093), Article 7093. https://doi.org/10.1038/nature04767

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

McNaughton, N., & Corr, P. J. (2018). Survival circuits and risk assessment.

Current Opinion in

Behavioral Sciences

, 14

–20. https://doi.org/10.1016/j.cobeha.2018.01.018

Mobbs, D., Hagan, C. C., Dalgleish, T., Silston, B., & Prévost, C. (2015). The ecology of human

fear: Survival optimization and the nervous system.

Frontiers in Neuroscience

https://www.frontiersin.org/articles/10.3389/fnins.2015.00055

Mobbs, D., Headley, D. B., Ding, W., & Dayan, P. (2020). Space, Time, and Fear: Survival

Computations along Defensive Circuits.

Trends in Cognitive Sciences

(3), 228

–241.

https://doi.org/10.1016/j.tics.2019.12.016

Mobbs, D., Petrovic, P., Marchant, J. L., Hassabis, D., Weiskopf, N., Seymour, B., Dolan, R. J.,

& Frith, C. D. (2007). When Fear Is Near: Threat Imminence Elicits Prefrontal-

Periaqueductal Gray Shifts in Humans.

Science

317

(5841), 1079

–1083.

https://doi.org/10.1126/science.1144298

Morgan, M. M., Whitney, P. K., & Gold, M. S. (1998). Immobility and flight associated with

antinociception produced by activation of the ventral and lateral/dorsal regions of the rat

periaqueductal gray.

Brain Research

804

(1), 159

–166. https://doi.org/10.1016/S0006-

8993(98)00669-6

Mota-Ortiz, S. R., Sukikara, M. H., Bittencourt, J. C., Baldo, M. V., Elias, C. F., Felicio, L. F., &

Canteras, N. S. (2012). The periaqueductal gray as a critical site to mediate reward

seeking during predatory hunting.

Behavioural Brain Research

226

(1), 32

–40.

https://doi.org/10.1016/j.bbr.2011.08.034

Motta, S. C., Carobrez, A. P., & Canteras, N. S. (2017). The periaqueductal gray and primal

emotional processing critical to influence complex defensive responses, fear learning

and reward seeking.

Neuroscience & Biobehavioral Reviews

, 39

–47.

https://doi.org/10.1016/j.neubiorev.2016.10.012

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Mumford, J. A., Poline, J.-B., & Poldrack, R. A. (2015). Orthogonalization of Regressors in fMRI

Models.

PLOS ONE

(4), e0126255. https://doi.org/10.1371/journal.pone.0126255

Muschelli, J., Nebel, M. B., Caffo, B. S., Barber, A. D., Pekar, J. J., & Mostofsky, S. H. (2014).

Reduction of motion-related artifacts in resting state fMRI using aCompCor.

NeuroImage

, 22

–35. https://doi.org/10.1016/j.neuroimage.2014.03.028

Nashold, B. S., Wilson, W. P., & Slaughter, D. G. (1969). Sensations evoked by stimulation in

the midbrain of man.

Journal of Neurosurgery

(1), 14

–24.

https://doi.org/10.3171/jns.1969.30.1.0014

Panksepp, J. (2011). The basic emotional circuits of mammalian brains: Do animals have

affective lives?

Neuroscience & Biobehavioral Reviews

(9), 1791

–1804.

https://doi.org/10.1016/j.neubiorev.2011.08.003

Paredes, J., Winters, R. W., Schneiderman, N., & McCabe, P. M. (2000). Afferents to the central

nucleus of the amygdala and functional subdivisions of the periaqueductal gray:

Neuroanatomical substrates for affective behavior.

Brain Research

887

(1), 157

–173.

https://doi.org/10.1016/S0006-8993(00)02972-3

Pereira, E. A. C., Lu, G., Wang, S., Schweder, P. M., Hyam, J. A., Stein, J. F., Paterson, D. J.,

Aziz, T. Z., & Green, A. L. (2010). Ventral periaqueductal grey stimulation alters heart

rate variability in humans with chronic pain.

Experimental Neurology

223

(2), 574

–581.

https://doi.org/10.1016/j.expneurol.2010.02.004

Power, J. D., Mitra, A., Laumann, T. O., Snyder, A. Z., Schlaggar, B. L., & Petersen, S. E.

(2014). Methods to detect, characterize, and remove motion artifact in resting state fMRI.

NeuroImage

, 320

–341. https://doi.org/10.1016/j.neuroimage.2013.08.048

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

R Core Team. (2016).

R: A language and environment for statistical computing

[Computer

software]. R Foundation for Statistical Computing. https://www.R-project.org

Renvall, V., Witzel, T., Wald, L. L., & Polimeni, J. R. (2016). Automatic cortical surface

reconstruction of high-resolution T1 echo planar imaging data.

NeuroImage

134

, 338

–

354. https://doi.org/10.1016/j.neuroimage.2016.04.004

Sandkühler, J., & Herdegen, T. (1995). Distinct patterns of activated neurons throughout the rat

midbrain periaqueductal gray induced by chemical stimulation within its subdivisions.

The Journal of Comparative Neurology

357

(4), 546

–553.

https://doi.org/10.1002/cne.903570406

Satpute, A. B., Wager, T. D., Cohen-Adad, J., Bianciardi, M., Choi, J.-K., Buhle, J. T., Wald, L.

L., & Barrett, L. F. (2013). Identification of discrete functional subregions of the human

periaqueductal gray.

Proceedings of the National Academy of Sciences

110

(42),

17101

–17106. https://doi.org/10.1073/pnas.1306095110

Satterthwaite, T. D., Elliott, M. A., Gerraty, R. T., Ruparel, K., Loughead, J., Calkins, M. E.,

Eickhoff, S. B., Hakonarson, H., Gur, R. C., Gur, R. E., & Wolf, D. H. (2013). An

improved framework for confound regression and filtering for control of motion artifact in

the preprocessing of resting-state functional connectivity data.

NeuroImage

, 240

–

256. https://doi.org/10.1016/j.neuroimage.2012.08.052

Sennesh, E., Theriault, J., Brooks, D., van de Meent, J.-W., Barrett, L. F., & Quigley, K. S.

(2022). Interoception as modeling, allostasis as control.

Biological Psychology

167

108242. https://doi.org/10.1016/j.biopsycho.2021.108242

Silva, B. A., Gross, C. T., & Gräff, J. (2016). The neural circuits of innate fear: Detection,

integration, action, and memorization.

Learning & Memory

(10), 544

–555.

https://doi.org/10.1101/lm.042812.116

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Silva, C., & McNaughton, N. (2019). Are periaqueductal gray and dorsal raphe the foundation of

appetitive and aversive control? A comprehensive review.

Progress in Neurobiology

177

, 33

–72. https://doi.org/10.1016/j.pneurobio.2019.02.001

Smith, S. M., & Nichols, T. E. (2009). Threshold-free cluster enhancement: Addressing

problems of smoothing, threshold dependence and localisation in cluster inference.

NeuroImage

(1), 83

–98. https://doi.org/10.1016/j.neuroimage.2008.03.061

Subramanian, H. H., Balnave, R. J., & Holstege, G. (2008). The Midbrain Periaqueductal Gray

Control of Respiration.

Journal of Neuroscience

(47), 12274

–12283.

https://doi.org/10.1523/JNEUROSCI.4168-08.2008

Thayer, J. F., Åhs, F., Fredrikson, M., Sollers, J. J., & Wager, T. D. (2012). A meta-analysis of

heart rate variability and neuroimaging studies: Implications for heart rate variability as a

marker of stress and health.

Neuroscience & Biobehavioral Reviews

(2), 747

–756.

https://doi.org/10.1016/j.neubiorev.2011.11.009

Theriault, J. E., Fischbach, A., Kragel, P. A., Wager, T. D., Wald, L. L., Bianciardi, M., Quigley,

K. S., Barrett, L. F., & Satpute, A. B. (under review).

Distinct spatial patterns of neural

and physiology-related activity in the human periaqueductal grey during anticipatory

social stress.

Tinoco Mendoza, F. A., Hughes, T. E. S., Robertson, R. V., Crawford, L. S., Meylakh, N.,

Macey, P. M., Macefield, V. G., Keay, K. A., & Henderson, L. A. (2023). Detailed

organisation of the human midbrain periaqueductal grey revealed using ultra-high field

magnetic resonance imaging.

NeuroImage

266

, 119828.

https://doi.org/10.1016/j.neuroimage.2022.119828

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Tryon, V. L., & Mizumori, S. J. Y. (2018). A Novel Role for the Periaqueductal Gray in

Consummatory Behavior.

Frontiers in Behavioral Neuroscience

, 178

–178. PubMed.

https://doi.org/10.3389/fnbeh.2018.00178

van Ast, V. A., Spicer, J., Smith, E. E., Schmer-Galunder, S., Liberzon, I., Abelson, J. L., &

Wager, T. D. (2016). Brain Mechanisms of Social Threat Effects on Working Memory.

Cerebral Cortex (New York, N.Y.: 1991)

(2), 544

–556.

https://doi.org/10.1093/cercor/bhu206

Vázquez-León, P., Miranda-Páez, A., Valencia-Flores, K., & Sánchez-Castillo, H. (2023).

Defensive and Emotional Behavior Modulation by Serotonin in the Periaqueductal Gray.

Cellular and Molecular Neurobiology

(4), 1453

–1468. https://doi.org/10.1007/s10571-

022-01262-z

Virtanen, P., Gommers, R., Oliphant, T. E., Haberland, M., Reddy, T., Cournapeau, D.,

Burovski, E., Peterson, P., Weckesser, W., Bright, J., van der Walt, S. J., Brett, M.,

Wilson, J., Millman, K. J., Mayorov, N., Nelson, A. R. J., Jones, E., Kern, R., Larson, E.,

… van Mulbregt, P. (2020). SciPy 1.0: Fundamental algorithms for scientific computing

in Python.

Nature Methods

(3), 261

–272. https://doi.org/10.1038/s41592-019-0686-2

Virtanen, P., Gommers, R., Oliphant, T. E., Haberland, M., Reddy, T., Cournapeau, D.,

Burovski, E., Peterson, P., Weckesser, W., Bright, J., Van Der Walt, S. J., Brett, M.,

Wilson, J., Millman, K. J., Mayorov, N., Nelson, A. R. J., Jones, E., Kern, R., Larson, E.,

… Vázquez-Baeza, Y. (2020). SciPy 1.0: Fundamental algorithms for scientific

computing in Python.

Nature Methods

(3), 261

–272. https://doi.org/10.1038/s41592-

019-0686-2

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Wager, T. D., & Smith, E. E. (2003). Neuroimaging studies of working memory: A meta-analysis.

Cognitive, Affective & Behavioral Neuroscience

(4), 255

–274.

https://doi.org/10.3758/cabn.3.4.255

Wager, T. D., van Ast, V. A., Hughes, B. L., Davidson, M. L., Lindquist, M. A., & Ochsner, K. N.

(2009). Brain mediators of cardiovascular responses to social threat, Part II: Prefrontal-

subcortical pathways and relationship with anxiety.

NeuroImage

(3), 836

–851.

https://doi.org/10.1016/j.neuroimage.2009.05.044

Wager, T. D., Waugh, C. E., Lindquist, M., Noll, D. C., Fredrickson, B. L., & Taylor, S. F. (2009).

Brain mediators of cardiovascular responses to social threat: Part I: Reciprocal dorsal

and ventral sub-regions of the medial prefrontal cortex and heart-rate reactivity.

NeuroImage

(3), 821

–835. https://doi.org/10.1016/j.neuroimage.2009.05.043

Wang, S., Veinot, J., Goyal, A., Khatibi, A., Lazar, S. W., & Hashmi, J. A. (2022). Distinct

networks of periaqueductal gray columns in pain and threat processing.

NeuroImage

250

, 118936. https://doi.org/10.1016/j.neuroimage.2022.118936

Weis, C., Bennett, K., Huggins, A., Parisi, E., & Larson, C. (2020). Dynamic Functional

Connectivity of the Periaqueductal Grey in Response to Predictable and Unpredictable

Threat Using 7-Tesla MRI.

Biological Psychiatry

(9), S164

–S165.

https://doi.org/10.1016/j.biopsych.2020.02.436

Weis, C. N., Bennett, K. P., Huggins, A. A., Parisi, E. A., Gorka, S. M., & Larson, C. (2022). A 7-

Tesla MRI study of the periaqueductal gray: Resting state and task activation under

threat.

Social Cognitive and Affective Neuroscience

(2), 187

–197.

https://doi.org/10.1093/scan/nsab085

Wickham, H. (2009).

ggplot2: Elegant Graphics for Data Analysis

. Springer-Verlag.

https://doi.org/10.1007/978-0-387-98141-3

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GREY IN WORKING MEMORY

Zhang, S. P., Bandler, R., & Carrive, P. (1990). Flight and immobility evoked by excitatory

amino acid microinjection within distinct parts of the subtentorial midbrain periaqueductal

gray of the cat.

Brain Research

520

–2), 73–82. https://doi.org/10.1016/0006-

8993(90)91692-a

Zhang, W., Hayward, L. F., & Davenport, P. W. (2005). Respiratory muscle responses elicited

by dorsal periaqueductal gray stimulation in rats.

American Journal of Physiology-

Regulatory, Integrative and Comparative Physiology

289

(5), R1338

–R1347.

https://doi.org/10.1152/ajpregu.00828.2004

Zhang, Y., Brady, M., & Smith, S. (2001). Segmentation of brain MR images through a hidden

Markov random field model and the expectation-maximization algorithm.

IEEE

Transactions on Medical Imaging

(1), 45

–57. https://doi.org/10.1109/42.906424

Zhang, Y. H., Hosono, T., Yanase-Fujiwara, M., Chen, X. M., & Kanosue, K. (1997). Effect of

midbrain stimulations on thermoregulatory vasomotor responses in rats.

The Journal of

Physiology

503 ( Pt 1)

, 177

–186. https://doi.org/10.1111/j.1469-7793.1997.177bi.x

Zhou, F., Zhao, W., Qi, Z., Geng, Y., Yao, S., Kendrick, K. M., Wager, T. D., & Becker, B.

(2021). A distributed fMRI-based signature for the subjective experience of fear.

Nature

Communications

(1), Article 1. https://doi.org/10.1038/s41467-021-26977-3

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GRAY IN WORKING MEMORY

Figure captions:

Figure 1.

Schematic overview of the working memory N-back task. An example of a 1-back

(mild cognitive demand) and 3-back (moderate cognitive demand) task showing a sequence of
trials. For every trial, a single letter was presented at the center of the visual field for 1-second.
Following the 1-second stimuli presentation, participants were given an additional 1-second to
respond before the next trial began. Participants were instructed to respond with a button press
if the target stimulus matched the stimulus

(either 1 or 3) positions back.

Figure 2.

PAG identification, segmentation, and visualization.

(

) Sagittal and coronal

perspectives on the human PAG. For the purposes of visualization, functional columns within
the PAG are colored according to the legend. The ventral 90° of the cylinder contains the dorsal
raphe and was excluded from analysis (

) The PAG is a 3D structure, and is difficult to

visualize. To simplify the presentation of results, we labeled each voxel with its position on the
rostral

–caudal axis, and its radial degree from the volumetric y-axis. Plotting rostral–caudal

position (y-axis) by radial degree (x-axis) flattens the PAG into a 2D surface. This convention
will be used in later figures. (

) The flattened map of PAG voxels can be simplified further by

taking radial slices within the PAG. Each voxel was labeled with its radial distance from the
center of the cylinder, and radial distance is plotted panels depicting 1 mm bins. For example,
the 0-1 mm panel shows voxels abutting the cerebral aqueduct, and the 3-4 mm panel shows
voxels at the outer periphery of the PAG. As in (b), the y-axis illustrates the longitudinal (rostral

–

caudal) axis of the PAG. Within each bin, radial degrees are plotted on the x-axis.

Figure 3.

Task-dependent physiological changes from baseline. Group means and 95%

confidence intervals are plotted in red for IBI (interbeat interval) and RR (respiration rate; see

Peripheral Physiological Recording & Quality Assessment

). Black dots illustrate individual mean

change across all 1-back (left) or 3-back (right) task periods. Note that a decrease in IBI reflects
an increase in heart rate. ***

< 0.001.

Figure 4.

Group-level patterns in whole-brain and brainstem contrasts.

(a)

Whole-brain

contrasts for 1-back > fixation, 3-back > fixation, and 3-back > 1-back contrasts. Subject-level
estimates were smoothed with a 1 mm FWHM Gaussian kernel and submitted to a group-level
GLM. Thresholding was performed using threshold-free cluster-enhancement

(α < .05; Smith &

Nichols, 2009). 1-back and 3-back > fixation contrasts generally showed a similar pattern within
the cortex, with the exception of in posterior insula and posterior parietal cortex, where no 1-
back response exceeding the TFCE threshold was observed. Cortical results for the 3-back > 1-
back contrast were similar to those observed in prior work (Kragel et al., 2019), except that no
3-back > 1-back difference was observed in PAG.

(b)

The 3-back > fixation contrast elicited

widespread BOLD activity in the brainstem, and in a number of voxels (marked in black) within
the group-aligned PAG mask (marked in white). Notably, there was no 3-back > 1-back
difference observed in the PAG or brainstem, suggesting that subtle changes above baseline

JNeurosci Accepted Manuscript

HUMAN PERIAQUEDUCTAL GRAY IN WORKING MEMORY

were present in the 1-back condition as well. Voxel-wise analyses within the PAG (Fig. 5)
allowed us to examine this further. All subject-level contrasts were warped to a common PAG
mask template to align PAG voxels at the group level (see

PAG Identification and Alignment in

Methods

Figure 5.

Patterns of group-level PAG BOLD signal intensity elicited by the 1-back/3-back task.

(a)

Left: Voxelwise topography of PAG BOLD response (

-scored) during 1-back (left) and 3-

back (right) conditions. Upper panels display radial slices of the PAG cylinder (each 1 mm thick;
see Fig. 2). The y-axis illustrates voxel position along the rostral

–caudal axis (14 mm in length).

The x-axis illustrates voxel position in +/

− 135 degrees from the brain midline. The heat map

illustrates group-level

-scored BOLD signal intensity relative to baseline fixation between task

blocks. Voxels were interpolated to 0.5 mm isotropic, from 1.1 mm native space, and smoothed
with a 1mm FWHM Gaussian kernel. Right: Beta weights for all 1-back PAG voxels (y-axis)
plotted against beta weights for all 3-back PAG voxels (x-axis). A Pearson correlation coefficient
and 95% confidence interval was calculated using 10,000 bootstrapped samples. Across voxels,
1-back and 3-back

-scores were highly correlated, consistent with similar in 1-back and 3-back

cortical responses, reported in Fig. 4. (

) To model PAG topography in the 1-back and 3-back

conditions, ordinal ranks were assigned to voxels according to their position on the rostral

–

caudal axis (4 ranks, left), and radial degree (10 ranks, right). Radial degree ranks were
bilateral, meaning that our analysis assumed symmetry in any PAG response (which is visually
supported by the topography in panel (a). In each subject, voxels in each combination of
rostral

–caudal and degree rank were averaged to produce 40 estimates for each subject to be

used in the mixed effects model.

(c)

Subject-level beta weights were averaged and plotted (with

95% confidence intervals) for each combination of rostral

–caudal and degree ranks. The y-axis,

unlike in plots above, depicts group-averaged beta weights (in arbitrary units, a.u.), and the x-
axis depicts radial degree rank (i.e., bilateral dorsomedial to ventrolateral regions, moving from
left to right). In both 1-back (left) and 3-back (right) conditions, the rostral end of the PAG
showed a general BOLD increase above fixation, whereas the caudal end of the PAG showed a
dorsomedial decrease, and increased BOLD response in ventrolateral areas. These plotted
averages are consistent with the results of the mixed effects model (see Results).

Figure 6.

Voxelwise correlation with task-elicited physiological changes.

(a)

Voxelwise

topography of between-subject correlations between either cardiac interbeat interval (top) or
respiratory rate (bottom) and estimated voxel-wise PAG BOLD signal intensity (i.e., beta weight)
elicited by the 1-back (left) and 3-back (right) task. Upper panels display radial slices of the PAG
cylinder (each 1 mm thick; see Fig. 2). The y-axis illustrates voxel position along the rostral

–

caudal axis (14 mm in length). The x-axis illustrates voxel position in +/

− 135 degrees from the

brain midline. The heat map illustrates Pearson correlation coefficients. Correlations were
calculated across all subjects with artifact-free physiological data (cardiac interbeat interval =
69; respiratory rate = 47).

(b)

Voxelwise BOLD-physiological change correlation coefficients for

interbeat interval change (y-axis, top) and respiratory rate change (y-axis bottom) are plotted
against contrast beta weights for all 1-back PAG voxels (x-axis, left) and 3-back PAG voxels (x-
axis, right). Pearson correlation coefficients and 95% confidence intervals were calculated using
10,000 bootstrapped samples. Correlations across voxels were non-zero, but substantially lower

JNeurosci Accepted Manuscript