bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

1 TITLE. Freely behaving mice can brake and turn during optogenetic stimulation of 2 the Mesencephalic Locomotor Region. 3 4 ABBREVIATED TITLE. MLR, speed and direction. 5 6 AUTHORS. Cornelis Immanuel van der Zouwen1, Joël Boutin1, Maxime Fougère1, 7 Aurélie Flaive1, Mélanie Vivancos1, Alessandro Santuz2,3,4, Turgay Akay2, Philippe 8 Sarret1,5,6,7, Dimitri Ryczko1,5,6,7*. 9 10 AFFILIATIONS. 1Département de Pharmacologie-Physiologie, Faculté de médecine 11 et des sciences de la santé, Université de Sherbrooke, Sherbrooke, QC, Canada. 12 2Atlantic Mobility Action Project, Brain Repair Center, Department of Medical 13 Neuroscience, Dalhousie University, Halifax, Nova Scotia, Canada. 3Department 14 of Training and Movement Sciences, Humboldt-Universität zu Berlin, Berlin, 15 Germany. 4Berlin School of Movement Science, Humboldt-Universität zu Berlin, 16 Berlin, Germany. 5Centre de recherche du Centre Hospitalier Universitaire de 17 Sherbrooke, Sherbrooke, QC, Canada. 6Centre d’excellence en neurosciences de 18 l’Université de Sherbrooke. 7Institut de Pharmacologie de Sherbrooke, 19 Sherbrooke, QC, Canada. 20 21 *CORRESPONDING AUTHOR. 22 Dr. Dimitri Ryczko 23 Département de Pharmacologie-Physiologie 24 Faculté de médecine et des sciences de la santé 25 Université de Sherbrooke, 26 Sherbrooke (Québec) Canada J1H 5N4 27 Tel: 1 (819) 821 8000 (ext. 75347) | Email: [email protected] 28 29 CONFLICT OF INTEREST. The authors declare no competing financial interests. 30 31 ACKNOWLEDGMENTS. We thank Jean Lainé for technical assistance with the 32 microscopy platform, Florian Bentzinger for providing access to the genotyping 33 equipment. 34 35 FUNDING. This work was supported by the Canadian Institutes of Health Research 36 (407083 to D.R. and FDN-148413 to P.S.); the Fonds de la Recherche - Québec 37 (FRQS Junior 1 awards 34920 and 36772 to D.R.); the Natural Sciences and 38 Engineering Research Council of Canada (RGPIN-2017-05522 and RTI-2019- 39 00628 to D.R.); the Canada Foundation for Innovation (39344 to D.R.), the Centre 40 de Recherche du Centre Hospitalier Universitaire de Sherbrooke (start-up funding 41 and PAFI grant to D.R.), the Faculté de médecine et des sciences de la santé 42 (start-up funding to D.R.), the Centre d’excellence en Neurosciences de 43 Sherbooke (to D.R.). P.S. is the holder of the Canada Research Chair Tier 1 in the 44 Neurophysiopharmacology of Chronic Pain 45 46 KEYWORDS 47 Locomotion, speed, braking, turning, Mesencephalic Locomotor Region, 48 cuneiform nucleus, Vglut2.

1/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

49 HIGHLIGHTS

50 - Mice brake and turn when approaching the arena’s corner during MLR-evoked

51 locomotion

52 - Speed decrease is scaled to speed before the turn during MLR-evoked

53 locomotion

54 - Turn angle is scaled to turn speed during MLR-evoked locomotion

55 - Gait and limb kinematics are similar during spontaneous and MLR-evoked

56 locomotion

2/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

57 ABSTRACT

58 Background: Stimulation of the Mesencephalic Locomotor Region (MLR) is

59 increasingly considered as a target to improve locomotor function in Parkinson's

60 disease, spinal cord injury and stroke. A key function of the MLR is to control the

61 speed of forward symmetrical locomotor movements. However, the ability of freely

62 moving mammals to integrate environmental cues to brake and turn during MLR

63 stimulation is poorly documented.

64 Objective/hypothesis: We investigated whether freely behaving mice could brake

65 or turn based on environmental cues during MLR stimulation.

66 Methods: We stimulated the cuneiform nucleus in mice expressing

67 channelrhodopsin in Vglut2-positive neurons in a Cre-dependent manner (Vglut2-

68 ChR2-EYFP) using optogenetics. We detected locomotor movements using deep

69 learning. We used patch-clamp recordings to validate the functional expression of

70 channelrhodopsin and neuroanatomy to visualize the stimulation sites.

71 Results: Optogenetic stimulation of the MLR evoked locomotion and increasing

72 laser power increased locomotor speed. Gait diagram and limb kinematics were

73 similar during spontaneous and optogenetic-evoked locomotion. Mice could brake

74 and make sharp turns (~90⁰) when approaching a corner during MLR stimulation

75 in an open-field arena. The speed during the turn was scaled with the speed before

76 the turn, and with the turn angle. In a reporter mouse, many Vglut2-ZsGreen

77 neurons were immunopositive for glutamate in the MLR. Patch-clamp recordings

78 in Vglut2-ChR2-EYFP mice show that blue light evoked short latency spiking in

79 MLR neurons.

3/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

80 Conclusion: MLR glutamatergic neurons are a relevant target to improve locomotor

81 activity without impeding the ability to brake and turn when approaching an

82 obstacle, thus ensuring smooth and adaptable navigation.

4/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

83 INTRODUCTION

84 Coordination of speed, braking and steering is essential to navigate the

85 environment [1]. In the brainstem, the Mesencephalic Locomotor Region (MLR)

86 plays a key role in initiating and controlling locomotion ([2], for review [3]). MLR

87 glutamatergic neurons control locomotor speed from basal vertebrates to

88 mammals (e.g. lamprey [4, 5, 6], salamanders [7], mice [8, 9, 10, 11, 12]). A key

89 function of the MLR is to elicit forward symmetrical locomotion by sending bilateral

90 glutamatergic inputs to reticulospinal neurons that project to the spinal central

91 pattern generator for locomotion (cat [13], lamprey [14, 15], zebrafish [16, 17],

92 salamander [18], mouse [19, 20, 10]). In mammals, MLR commands are relayed

93 by reticulospinal neurons in the lateral paragigantocellular nucleus (LPGi) [10].

94 Steering movements are induced by asymmetrical reticulospinal activity.

95 Increased reticulospinal activity on one side induces ipsilateral steering

96 movements in lamprey [21, 22, 23, 24], zebrafish [25, 26], salamander [27] and rat

97 [28]. In mammals, steering commands are relayed by reticulospinal neurons in the

98 gigantocellular nucleus (Gi) that express the molecular marker Chx10. Their

99 unilateral activation evokes ipsilateral braking and turning [29, 30, 31]. These

100 neurons receive no input from the MLR, but a major glutamatergic input from the

101 contralateral superior colliculus (SC), a region involved in visuomotor

102 transformations ([32, 33, 34], for review [35]).

103 The interactions between brainstem substrates controlling speed (MLR-

104 LPGi) and those controlling braking and turning (SC-Gi) are unknown. Whether

105 braking or turning can be done during MLR stimulation is poorly documented in

5/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

106 mammals. In mice, MLR-evoked locomotion was recorded on a trackball [8, 9],

107 treadmill [11] or in a linear corridor [12], but the ability to integrate environmental

108 cues that modify the direction of motion was not studied. In decerebrated cats held

109 over a treadmill oriented in various directions, MLR stimulation generated well-

110 coordinated locomotion only when the treadmill was going in front-to-rear direction,

111 suggesting that the MLR could only generate forward motion [36].

112 Here, we examined whether freely behaving mice can brake or turn by

113 integrating environmental cues during optogenetic stimulation of the MLR. In mice

114 expressing channelrhodopsin in neurons positive for the vesicular glutamatergic

115 transporter 2 (Vglut2-ChR2-EYFP mice), we targeted the cuneiform nucleus

116 (CnF), the MLR subregion that controls the largest range of speeds [11, 12]. We

117 used deep learning to detect locomotor movements in a linear corridor and in an

118 open field arena. It is relevant to determine whether MLR-evoked locomotion can

119 be dynamically adapted to the environment, as MLR stimulation is explored to

120 improve locomotor function in Parkinson’s disease [37, 38, 39, 40] and in animal

121 models of spinal cord injury [41, 42] and stroke [43]. We focused on the CnF, which

122 is increasingly considered as the optimal subregion to target within the MLR [44].

6/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

123 MATERIALS AND METHODS

124 The procedures were in accordance with the guidelines of the Canadian Council

125 on Animal Care and approved by the animal care and use committees of the

126 Université de Sherbrooke.

127

128 Animals. Thirteen mice were used. We used Vglut2-Cre mice (Jackson

129 laboratories, #028863, Vglut2-ires-cre knock-in (C57BL/6J)) [45] (Fig. 1A-B),

130 ChR2-EYFP-lox mice (Ai32 mice, Jackson laboratory, #024109, B6.Cg-

131 Gt(ROSA)26Sortm32(CAG-COP4*H134R/EYFP)Hze/J) [46] (Fig. 1B), and ZsGreen-lox mice

132 (Ai6 mice, Jackson laboratory, #007906, B6.Cg-Gt(ROSA)26Sortm6(CAG-

133 ZsGreen1)Hze/J) [46] (Fig. 1A). We crossed homozygous Vglut2-Cre mice with

134 homozygous ChR2-EYFP-lox mice to obtain the double heterozygous Vglut2-

135 ChR2-EYFP mice. We crossed homozygous Vglut2-Cre mice with homozygous

136 ZsGreen-lox mice to obtain the double heterozygous Vglut2-ZsGreen mice.

137 Animals had ad libitum access to food and water, with lights on from 6 AM to 8 PM.

138 Mice were 16-36 weeks old for in vivo optogenetics (3 males, 2 females), 10-18

139 weeks old for neuroanatomy (1 male, 5 females), and 15-23 days old for patch-

140 clamp experiments (1 male, 1 undetermined).

141

142 Optical fiber implantation. Mice were anesthetized using isoflurane (induction:

143 5%, 500 mL/min; maintenance: 1.5-2.5%, 100 mL/min) delivered by a SomnoSuite

144 (Kent Scientific, Torrington, CT, USA). Mice were placed in a Robot Stereotaxic

145 instrument coupled with StereoDrive software (Neurostar, Tübingen, Germany) to

7/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

146 perform unilateral implantation of an optical fiber (200 µm core, 0.22 NA, Thorlabs,

147 Newton, NJ, USA) held in a 5 mm ceramic or stainless-steel ferrule 500 µm above

148 the right CnF at -4.80 mm anteroposterior, +1.10 mm mediolateral, -2.40 mm

149 dorsoventral relative to bregma [11, 12]. The ferrule was secured on the cranium

150 using two 00-96 x 1/16 mounting screws (HRS scientific, QC, Canada) and dental

151 cement (A-M Systems, Sequim, WA, USA).

152

153 In vivo optogenetic stimulation. The optical fiber was connected using a pigtail

154 rotary joint (Thorlabs) to a 470 nm laser (Ikecool, Anaheim, CA, USA) or a 589 nm

155 laser (Laserglow, ON, Canada) driven by a Grass S88X that generated the

156 stimulation trains (10 s train, 10 ms pulses, 20 Hz) [11, 12]. To visualize

157 optogenetic stimulation, a small (diameter 0.5 cm), low-power (0.13 W) red LED

158 that received a copy of the stimulation trains was placed in the camera’s field of

159 view. The 470 nm light source was adjusted to 6-27% of laser power and the 589

160 nm to 40-53% of laser power. The corresponding power measured at fiber tip with

161 a power meter (PM100USB, Thorlabs) was 0.1-16.0 mW for the 470 nm laser and

162 1.7-9.4 mW for the 589 nm laser.

163

164 Open-field locomotion. Locomotor activity was filmed from above in a 40 × 40

165 cm arena at 30 fps using a Logitech Brio camera coupled to a computer equipped

166 with ANY-maze software (Stoelting Co., Wood Dale, IL, USA) or using a Canon

167 Vixia HF R800 camera. Locomotor activity was recorded during trials of 15 min

168 during which 10 stimulation trains were delivered every 80 s at various laser

8/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

169 powers. Video recordings were analyzed on a computer equipped with

170 DeepLabCut (version 2.1.5.2), a software based on deep learning to track user-

171 defined body parts [47, 48], and a custom Matlab script (Mathworks, Natick, MA,

172 USA). We tracked frame by frame the body center position, the corners of the

173 arena for distance calibration and the low-power LED to detect optogenetic

174 stimulations. Timestamps were extracted using Video Frame Time Stamps (Matlab

175 File Exchange). Body center positions and timestamps were used to calculate

176 locomotor speed in cm/s. To compare and average speed over time for different

177 stimulations, the data was downsampled to 20 Hz. Body center positions were

178 excluded if their likelihood of detection by DeepLabCut was < 0.8, if they were

179 outside of the open field area, or if body center speed exceeded the maximum

180 locomotor speed recorded in mice (334 cm/s, [49]).

181 For offline analysis of turning movements in the arena’s corners, we defined

182 regions of interest (ROIs) as circles (radius 20 cm) centered on each corner. The

183 turning point was defined as the intersection of the mouse’s trajectory with the

184 bisector of each corner (i.e. diagonal of the corner, Fig. 5D). The coordinates of

185 the turning point were calculated using Curve intersections (Matlab File

186 Exchange). Within an ROI, a turn was defined as a trajectory that started at least

187 5 cm away from the turning point, crossed the diagonal, and ended at least 5 cm

188 away from the turning point. Turns were excluded if the mouse crossed the

189 diagonal more than once without leaving the ROI. The turn angle was measured

190 between the first point of the trajectory, the turning point, and the last point of the

191 trajectory.

9/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

192 Locomotor speed during the start of the turn (“entry speed”), around the

193 turning point (“turn speed”), and during the end of the turn (“exit speed”) were

194 measured using the distance of each point of the trajectory to the turning point.

195 These distance values were binned (width: 1 cm) and speed values were averaged

196 per bin. Entry speed was averaged from the four most distal distance bins before

197 the turning point and at least 5 cm away from the turning zone (2 cm radius around

198 turning point). Turn speed was averaged from the four distance bins located within

199 the turning zone. Exit speed was averaged from the four most distal distance bins

200 after the turning point and at least 5 cm away from the turning zone. Turns were

201 removed from the analysis if fewer than four bins were available to calculate entry,

202 turn, or exit speed.

203

204 Footfall patterns and limb kinematics. To label hindlimb joints for offline tracking

205 with DeepLabCut, mice were anesthetized with isoflurane (induction: 5%, 500

206 mL/min; maintenance: 1.5-2.5%, 100 mL/min), the hindlimb was shaved and white

207 dots (diameter ~2 mm) were drawn on the iliac crest, hip, knee, ankle, and

208 metatarsophalangeal (MTP) joints, and toe tip using a fine-tip, oil-based paint

209 marker (Sharpie). For footfall pattern tracking, no labeling of paw underside was

210 needed. Animals recovered for 20 min after anesthesia and were placed in a 1 m

211 long, 8 cm wide transparent corridor. The footfall pattern and hindlimb kinematics

212 were recorded at 300 fps using two high speed Genie Nano Camera M800

213 cameras (Teledyne DALSA, Waterloo, ON, Canada) coupled to a computer

214 equipped with Norpix Streampix software (1st Vision, Andover, MA, USA).

10/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

215 Hindlimb kinematics were recorded with a camera placed on the side of the

216 corridor. Footfall patterns were recorded with a camera placed on the side and

217 directed toward a 45-degree mirror placed below the corridor. For distance

218 calibration, 4 markers (diameter 0.5 cm) were distributed 5 cm apart and placed in

219 the field of view of each camera. To detect optogenetic stimulation, a low-power

220 LED that received a copy of the stimulation trains was placed in the field of view of

221 both cameras. Animals were recorded during spontaneous locomotion evoked by

222 a gentle touch of the animal’s tail, and during optogenetic-evoked locomotion.

223 For the footfall pattern, videos recorded from below were used to track the

224 position of the MTPs of the four paws with DeepLabCut [47, 48, 50]. Paw speeds

225 were calculated and smoothened with a moving average (on 5 frames) using a

226 custom Matlab script. Touchdown and lift-off were defined for each paw as the time

227 points at which each MTP speed respectively fell below or rose above 15 cm/s.

228 The touchdown and lift-off time points of each limb were identified using Curve

229 intersections (Matlab File Exchange) and were normalized to the step cycle of the

230 left hindlimb to generate normalized gait diagrams [11]. A step cycle was defined

231 as the time between two consecutive touchdowns of the left hindlimb [12]. Cycle

232 duration, stance phase duration, swing phase duration and stride length were

233 calculated [12].

234 For hindlimb kinematics, the positions of the joints and toe tip were detected

235 using DeepLabCut. The moving average of the MTP speed was used to determine

236 the stance and swing phases by detecting the touchdown and lift-off times with a

237 speed threshold of 9 cm/s. The joint positions were used to extract the angles of

11/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

238 the hip, knee and ankle joints (Fig. 4C). The angular variations as a function of

239 time were normalized to step cycle duration using MTP touchdown times as a

240 reference [51].

241 Frames were excluded from the analysis if the MTPs of any paw (for the

242 footfall pattern) or any limb joints or the toe tip (for limb kinematics) had a likelihood

243 of detection < 0.8 by DeepLabCut. Frames were excluded from the analysis if any

244 paw’s or joint’s speed exceeded 400 cm/s, i.e. the maximum locomotor speed of

245 a mouse with a 20% margin to account for increased speed of individual body parts

246 [49].

247

248 Statistical analysis. Data are presented as mean ± standard error of the mean

249 (SEM) unless stated otherwise. Statistical analyses were done using Sigma Plot

250 12.0. Normality was assessed using the Shapiro-Wilk test. Equal variance was

251 assessed using the Levene test. Parametric analyses were used when

252 assumptions for normality and equal variance were respected, otherwise non-

253 parametric analyses were used. To compare the means between two dependent

254 groups, a two-tailed paired-t test was used. For more than two dependent groups,

255 a parametric one-way analysis of variance (ANOVA) for repeated measures or a

256 non-parametric Friedman ANOVA for repeated measures on ranks was used.

257 ANOVAs were followed by a Student Newman-Keuls post-hoc test for multiple

258 comparisons between groups. Statistical differences were assumed to be

259 significant when P < 0.05.

12/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

260 For genotyping, patch-clamp recordings, histology and

261 immunofluorescence, patch-clamp recordings, specificity of the antibodies and of

262 the transgenic mice, and details on DeepLabCut networks, please see the

263 supplementary material.

13/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

264 RESULTS

265 We targeted glutamatergic cells that expressed Vglut2 in the CnF for

266 optogenetic stimulation. We examined the presence of such cells by crossing mice

267 expressing the Cre-recombinase under control of the Vglut2 promoter (Vglut2-Cre

268 mouse) with mice expressing a green fluorescent protein in a Cre-dependent

269 manner (ZsGreen-lox mice) (Fig. 1A). In the offspring (Vglut2-ZsGreen mice),

270 many cells were positive for ZsGreen in the CnF (n = 3 mice, Fig. 1C-E). Many of

271 these cells were immunopositive for NeuN (n = 3 mice, Fig .1I-J) and for glutamate

272 (n = 3 mice, Fig. 1F-H). To stimulate these cells with blue light, we crossed Vglut2-

273 Cre mice with mice expressing ChR2 in a Cre-dependent manner (ChR2-EYFP-

274 lox) (Fig. 1B). Using patch-clamp recording in slices of the offspring (Vglut2-ChR2-

275 EFYP mice), we validated that blue light elicited spiking at short latency in MLR

276 neurons (n = 2 neurons from 2 mice, Fig 1K).

277 We then activated CnF neurons in freely moving Vglut2-ChR2-EYFP mice

278 in an open-field arena (Fig. 2A). We implanted an optic fiber 500 µm above the

279 right CnF and verified the implantation sites (n = 5 mice, Fig. 2B-C). Optogenetic

280 stimulation of the CnF with blue light increased locomotor speed as shown by

281 single animal data (Fig. 2D,F) and data pooled from 5 mice (Fig. 2H). Statistical

282 analysis confirmed that speed was increased during optogenetic stimulation (P <

283 0.01 vs prestim, Student Newman-Keuls after a one-way ANOVA for repeated

284 measures, P < 0.01) and decreased after light was switched off (P < 0.01 vs opto

285 stim, Fig. 2J). Replacing the 470 with a 589 nm laser did not increase locomotion

14/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

286 as shown by single animal data (Fig. 2E,G) and data pooled from 4 mice (P > 0.05

287 one-way ANOVA for repeated measures, Fig. 2I,K).

288 Next, we compared spontaneous and optogenetic-evoked locomotion in a

289 transparent linear corridor. We tracked the movements of each paw frame by

290 frame using DeepLabCut [47, 48] (Fig. 3A). The footfall pattern was similar during

291 spontaneous and optogenetic-evoked locomotion (Fig. 3B-C). We normalized the

292 cycle duration as a function of the left hindlimb movements and observed again

293 similar gait diagrams during spontaneous and optogenetic-evoked locomotion as

294 shown by single animal data (Fig. 3D-E) and data pooled from 4 mice (Fig. 3F-G).

295 We noticed, however, that mice were stepping faster during optogenetic-evoked

296 locomotion as the cycle duration was shorter (P < 0.05 vs. spontaneous, paired

297 test, n = 4 animals, Fig. 3H) while stride length did not differ (P > 0.05, Fig. 3I).

298 This was associated with a shorter stance duration (P < 0.01, Fig. 3J), but no

299 modification of swing duration (P > 0.05, Fig. 3K), consistent with the specific

300 modulation of stance duration when speed increases during natural locomotion

301 [52]. Altogether, this indicated that optogenetic CnF stimulation evoked a normal

302 footfall pattern in Vglut2-ChR2-EYFP mice.

303 We then compared the limb kinematics by tracking each hindlimb joint (iliac

304 crest, hip, knee, ankle, MTP) and the toe tip using DeepLabCut (Fig. 4A). The stick

305 diagrams were similar during spontaneous and optogenetic-evoked locomotion

306 (Fig. 4B). We compared the angular variations of the hip, knee, ankle and MTP

307 joints as a function of time (Fig. 4C) and cycle duration was normalized relative to

308 MTP movements (Fig. 4D). The angular variations were similar during

15/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

309 spontaneous and optogenetic-evoked locomotion as shown by single animal data

310 and data pooled from 4 mice (Fig. 4E). Statistical analysis revealed no difference

311 of the angle amplitude of the four joints between the two conditions (P > 0.05 vs

312 spontaneous, paired tests, n = 4 animals, Fig. 4F). Altogether, this indicated that

313 optogenetic CnF stimulation evoked normal limb kinematics.

314 We then examined whether freely behaving mice could brake or turn during

315 optogenetic CnF stimulation in the open-field arena. Inspection of the speed as a

316 function of time uncovered oscillations during optogenetic stimulation (Fig. 5A-B).

317 We plotted the speed as a function of the location of the animal in the arena and

318 found that speed decreased in the corners of the arena, where the animal was

319 performing turning movements (Fig. 5C). This suggested that during CnF

320 stimulation, the animal dynamically controlled speed as a function of environmental

321 cues. We furthered studied this phenomenon by analyzing locomotor movements

322 in each corner of the arena during CnF stimulation (Fig. 5D). We defined ROIs as

323 circles (20 cm radius) centered on each corner. The trajectories of a single mouse

324 within the four ROIs during CnF stimulation are illustrated in Fig. 5E. We defined

325 the turning point as the intersection between the mouse’s trajectory and the

326 corner’s bisector (i.e. arena’s diagonal). Plotting the speed relative to the distance

327 from the turning point indicated that speed was lower around the turning point in

328 single animal data (Fig. 5G) as in data pooled from 5 mice (Fig. 5F,H). Statistical

329 analysis showed that speed decreased by ~61% during the turn (P < 0.05 vs. entry

330 speed, Student Newman-Keuls after a one-way ANOVA for repeated measures, P

331 < 0.01, Fig. 5I). Speed increased when exiting the corner (P < 0.01 vs. turn speed,

16/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

332 Fig. 5I) to values that were not different from the entry speed (P > 0.05 vs entry

333 speed, Fig. 5I). This indicated that the slowdown was transient and linked to the

334 turn, after which ongoing MLR stimulation regained control over speed.

335 We examined the relationships between speed, braking and turn angle. We

336 found a strong positive linear relationship between the entry speed, and the speed

337 decrease between entry and turning zone in 5 mice (P < 0.0001, R = 0.90, Fig. 5J),

338 and a strong positive linear relationship between the exit speed, and the speed

339 increase between turning zone and exit (P < 0.0001, R = 0.92, Fig. 5K). We found

340 a weak but significant positive linear relationship between entry speed and turn

341 speed (P < 0.05, R = 0.27, Fig. 5L) and a positive linear relationship between turn

342 speed and turn angle (P < 0.0001, R = 0.44, Fig. 5M). This last relationship is

343 visible when looking at the trajectories color coded as a function of turn speed (Fig.

344 5E-F). This indicated that during CnF stimulation, wider angles were easier to

345 negotiate at high speed than sharp ones, as reported during natural locomotion in

346 mammals [1].

347 We examined the robustness of such scaled control of speed during turning

348 when increasing CnF stimulation. Increasing the laser power applied to the CnF

349 increased locomotor speed as shown by single animal data (Fig. 6A-B) and data

350 pooled from 5 mice (Fig. 6C). We expressed the laser power and speed as a

351 function of their maximal values per animal, and we found a strong positive

352 sigmoidal relationship between laser power and speed (P < 0.01, R = 0.99, Fig.

353 6D). Such precise control of speed confirmed that we successfully targeted the

354 CnF (Fig. 2C). Mice made to walk at increasing speeds imposed by increasing CnF

17/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

355 stimulation were able to maintain successful braking and turning as shown by

356 single animal data (Fig. 6E-F) and data pooled from 5 mice (Fig. 6G-H). The

357 relationships describing the scaling of speed relative to the turn properties were

358 conserved within this range of speeds (Fig. 6I-L). This indicated that CnF

359 stimulation controls locomotor speed, without preventing the animal from precisely

360 regulating braking and turning, likely through dynamic integration of environmental

361 cues.

18/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

362 DISCUSSION

363 We show in Vglut2-ChR2-EYFP mice that optogenetic stimulation of the

364 CnF with blue light evoked locomotion and that increasing laser power increased

365 speed. Replacing the blue laser with a red laser evoked no locomotion. In a linear

366 corridor, footfall patterns and limb kinematics were largely similar during

367 spontaneous and optogenetic-evoked locomotion. In the open-field arena, mice

368 could brake and perform sharp turns (~90⁰) when approaching a corner during CnF

369 stimulation. Speed decrease during the turn was scaled to speed before the turn,

370 and turn speed was scaled to turn angle. We verified the stimulation sites in the

371 CnF and showed that many Vglut2-ZsGreen cells in the CnF were positive for

372 NeuN and glutamate. Using patch-clamp recordings in brainstem slices we

373 showed that blue light evoked short latency spiking. Altogether, our study indicates

374 that the CnF controls locomotor speed without preventing the animal from

375 integrating environmental cues to perform braking and turning movements, and

376 thereby smoothly navigate the environment.

377

378 Methodological considerations. We cannot exclude that some non-

379 glutamatergic neurons were stimulated in the CnF of Vglut2-ChR2-EYFP mice. We

380 crossed Vglut2-Cre with ChR2-EYFP-lox mice. In the offspring (Vglut2-ChR2-

381 EYFP), if Vglut2 is expressed during cell lifetime, ChR2 is expressed permanently

382 under control of the CAG promoter even if Vglut2 is not expressed anymore [53,

383 54]. Mainly glutamatergic and GABAergic neurons are present in the CnF (for

384 review [3]). Although the expression of Vglut2 was detected in some GABAergic

19/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

385 neurons in the mammalian brain (for review [55]), it is unlikely that we stimulated

386 GABAergic neurons. Two arguments indicating that we successfully targeted CnF

387 glutamatergic neurons are the normal gait diagrams and limb kinematics, and the

388 precise control of speed when increasing laser power. These effects are consistent

389 with results obtained in Vglut2-Cre mice optogenetically stimulated in the CnF

390 following injection of an AAV encoding for ChR2 in a Cre-dependent manner [11,

391 12].

392

393 Brainstem control of speed. Our results support the idea that MLR glutamatergic

394 neurons play a key role in the initiation of forward symmetrical locomotion and in

395 the control of speed by sending input to reticulospinal neurons (lamprey [4, 3],

396 salamander [7, 18], mouse [20, 8, 9, 10, 11, 12]) that send input to excitatory

397 neurons of the locomotor Central Pattern Generator (lamprey [14], zebrafish [16,

398 17], mouse: [19, 10], for review [56]). In mice, the reticulospinal neurons relaying

399 MLR locomotor commands are in the LPGi [10].

400 In addition, we show that MLR stimulation does not prevent mice from

401 braking and turning following integration of environmental cues. The turning and

402 braking movements recorded here displayed the same characteristics as those

403 shown by mammals during natural locomotion. At high speed, mice used fewer

404 sharp turns (i.e. higher angles), consistent with observations in wild northern

405 quolls, which reduce their locomotor speed more during turns with a smaller radius

406 [1].

407

20/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

408 Brainstem control of braking and turning. Our observations support the idea

409 that distinct reticulospinal neurons control speed and turning/braking movements

410 [29, 30, 31]. First, our data indicate that a substrate for turning movements is

411 activated transiently during MLR stimulation when approaching a corner. This

412 motor signature closely matches that previously recorded when selectively

413 activating the brainstem circuit for turning [30, 31]. Indeed, unilateral activation of

414 Chx10-positive reticulospinal neurons in the Gi produces an ipsilateral turn

415 together with a decrease in speed [30, 31]. Gi-Chx10 reticulospinal neurons

416 receive no input from the MLR, but a major input from the contralateral superior

417 colliculus (SC), a region involved in visuomotor transformations [30]. Such

418 connectivity is relevant to the behavioral task mice had to solve here in the open-

419 field arena, i.e. integrating visual cues to avoid the arena’s corner during

420 locomotion evoked by MLR stimulation (Fig. 7). Altogether, this suggests that the

421 brainstem substrates for braking and turning ([29, 30, 31], see also [57]) can be

422 recruited during MLR stimulation, therefore allowing the animal to smoothly

423 navigate the environment (Fig. 7).

424 Interestingly, the speed decreased to zero during some turns, i.e. mice

425 transiently halted during optogenetic MLR stimulation (Fig. 5G). Future studies

426 should examine which level of the locomotor circuitry is involved in this effect. At

427 the reticulospinal level, “stop cells” could increase their activity to stop locomotion

428 as shown in in basal vertebrates (lamprey [58, 59]). In mammals, halt is induced

429 by a bilateral recruitment of Gi Chx10-positive reticulospinal neurons, i.e. the same

430 neurons that induce turning when activated unilaterally [29, 30, 31] (see also [57]).

21/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

431 Locomotor pauses and rhythm resetting were also reported when photoactivating

432 Vglut2-positive neurons in the Gi in mice [60]. At the MLR level, local GABAergic

433 neurons could stop locomotion likely by inhibiting MLR glutamatergic neurons [9,

434 12]. In lamprey, stimulation of the MLR, at lower stimulation intensity values than

435 the ones evoking locomotion, stop locomotion by recruiting reticulospinal stop cells

436 [59]. Two incoming inputs to the MLR could be involved. A transient increase in

437 the GABAergic tone from the output stations of the basal ganglia could stop

438 locomotion (lamprey [61, 62], mouse [63, 9]). Alternatively, increased activity from

439 the output station of the basolateral amygdala could be involved, since activation

440 of this region is synchronized with locomotor arrests in familiar places during

441 exploratory behavior ([64], for review [65]).

442

443 Conclusions. We show that optogenetic stimulation of the CnF in Vglut2-ChR2-

444 EYFP mice controls locomotor speed without preventing braking and turning

445 movements following integration of environmental cues. This supports the idea that

446 distinct brainstem circuits control speed (“MLR-LPGi pathway”, [8, 9, 10, 11, 12])

447 and braking/turning movements in mammals (“SC-Gi pathway”, [29, 30, 31]) (Fig.

448 7). This also suggests that CnF glutamatergic neurons are a relevant target to

449 improve navigation adaptable to the environment in conditions where locomotion

450 is impaired such as Parkinson’s disease [37, 38, 39, 40] spinal cord injury [41, 42]

451 and stroke [43] (for review [44]).

22/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

452 AUTHOR CONTRIBUTIONS. CIvdZ: Conceptualization; Data curation; Formal

453 analysis; Investigation; Methodology; Software; Validation; Visualization;

454 Roles/Writing - original draft; Writing - review & editing. JB: Data curation;

455 Investigation; Methodology; Validation; Visualization; Writing - review & editing.

456 MF: Data curation; Investigation; Methodology; Validation; Visualization; Writing -

457 review & editing. AF: Data curation; Investigation; Methodology; Validation;

458 Visualization; Writing - review & editing. MV: Methodology; Writing - review &

459 editing. AS: Methodology; Software; Writing - review & editing. TA: Methodology;

460 Software; Writing - review & editing. PS: Methodology; Resources; Writing - review

461 & editing. DR: Conceptualization; Data curation; Formal analysis; Funding

462 acquisition; Methodology; Project administration; Resources; Supervision;

463 Validation; Visualization; Roles/Writing - original draft; Writing - review & editing.

23/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

464 REFERENCES

465 [1] Wynn ML, Clemente C, Nasir AFAA, Wilson RS. Running faster causes

466 disaster: trade-offs between speed, manoeuvrability and motor control when

467 running around corners in northern quolls (Dasyurus hallucatus). J Exp Biol

468 2015;218:433–9. https://doi.org/10.1242/jeb.111682.

469 [2] Shik ML, Severin FV, Orlovskiĭ GN. [Control of walking and running by means

470 of electric stimulation of the midbrain]. Biofizika 1966;11:659–66.

471 [3] Ryczko D, Dubuc R. The multifunctional mesencephalic locomotor region. Curr

472 Pharm Des 2013;19:4448–70. https://doi.org/10.2174/1381612811319240011.

473 [4] Sirota MG, Di Prisco GV, Dubuc R. Stimulation of the mesencephalic locomotor

474 region elicits controlled swimming in semi-intact lampreys. Eur J Neurosci

475 2000;12:4081–92. https://doi.org/10.1046/j.1460-9568.2000.00301

476 [5] Brocard F, Dubuc R. Differential contribution of reticulospinal cells to the control

477 of locomotion induced by the mesencephalic locomotor region. J Neurophysiol

478 2003;90:1714–27. https://doi.org/10.1152/jn.00202.2003.

479 [6] Le Ray D, Brocard F, Bourcier-Lucas C, Auclair F, Lafaille P, Dubuc R. Nicotinic

480 activation of reticulospinal cells involved in the control of swimming in lampreys.

481 Eur J Neurosci 2003;17:137–48. https://doi.org/10.1046/j.1460-

482 9568.2003.02417.x.

483 [7] Cabelguen J-M, Bourcier-Lucas C, Dubuc R. Bimodal locomotion elicited by

484 electrical stimulation of the midbrain in the salamander Notophthalmus

485 viridescens. J Neurosci 2003;23:2434–9.

24/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

486 [8] Lee AM, Hoy JL, Bonci A, Wilbrecht L, Stryker MP, Niell CM. Identification of a

487 Brainstem Circuit Regulating Visual Cortical State in Parallel with Locomotion.

488 Neuron 2014;83:455–66. https://doi.org/10.1016/j.neuron.2014.06.031.

489 [9] Roseberry TK, Lee AM, Lalive AL, Wilbrecht L, Bonci A, Kreitzer AC. Cell-Type-

490 Specific Control of Brainstem Locomotor Circuits by Basal Ganglia. Cell

491 2016;164:526–37. https://doi.org/10.1016/j.cell.2015.12.037.

492 [10] Capelli P, Pivetta C, Soledad Esposito M, Arber S. Locomotor speed control

493 circuits in the caudal brainstem. Nature 2017;551:373–7.

494 https://doi.org/10.1038/nature24064.

495 [11] Josset N, Roussel M, Lemieux M, Lafrance-Zoubga D, Rastqar A, Bretzner F.

496 Distinct Contributions of Mesencephalic Locomotor Region Nuclei to Locomotor

497 Control in the Freely Behaving Mouse. Curr Biol 2018;28:884-901.e3.

498 https://doi.org/10.1016/j.cub.2018.02.007.

499 [12] Caggiano V, Leiras R, Goñi-Erro H, Masini D, Bellardita C, Bouvier J, et al.

500 Midbrain circuits that set locomotor speed and gait selection. Nature

501 2018;553:455–60. https://doi.org/10.1038/nature25448.

502 [13] Orlovskiĭ GN. [Work of reticulo-spinal neurons during locomotion]. Biofizika

503 1970;15:728–37.

504 [14] Buchanan JT, Grillner S. Newly identified “glutamate interneurons” and their

505 role in locomotion in the lamprey spinal cord. Science 1987;236:312–4.

506 https://doi.org/10.1126/science.3563512.

507 [15] Brocard F, Ryczko D, Fénelon K, Hatem R, Gonzales D, Auclair F, et al. The

508 transformation of a unilateral locomotor command into a symmetrical bilateral

25/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

509 activation in the brainstem. J Neurosci 2010;30:523–33.

510 https://doi.org/10.1523/JNEUROSCI.3433-09.2010.

511 [16] Kinkhabwala A, Riley M, Koyama M, Monen J, Satou C, Kimura Y, et al. A

512 structural and functional ground plan for neurons in the hindbrain of zebrafish.

513 Proc Natl Acad Sci USA 2011;108:1164–9.

514 https://doi.org/10.1073/pnas.1012185108.

515 [17] Kimura Y, Satou C, Fujioka S, Shoji W, Umeda K, Ishizuka T, et al. Hindbrain

516 V2a neurons in the excitation of spinal locomotor circuits during zebrafish

517 swimming. Curr Biol 2013;23:843–9. https://doi.org/10.1016/j.cub.2013.03.066.

518 [18] Ryczko D, Auclair F, Cabelguen J-M, Dubuc R. The mesencephalic locomotor

519 region sends a bilateral glutamatergic drive to hindbrain reticulospinal neurons in

520 a tetrapod. J Comp Neurol 2016a;524:1361–83.

521 https://doi.org/10.1002/cne.23911.

522 [19] Hägglund M, Borgius L, Dougherty KJ, Kiehn O. Activation of groups of

523 excitatory neurons in the mammalian spinal cord or hindbrain evokes locomotion.

524 Nat Neurosci 2010;13:246–52. https://doi.org/10.1038/nn.2482.

525 [20] Bretzner F, Brownstone RM. Lhx3-Chx10 reticulospinal neurons in locomotor

526 circuits. J Neurosci 2013;33:14681–92.

527 https://doi.org/10.1523/JNEUROSCI.5231-12.2013.

528 [21] Deliagina TG, Zelenin PV, Fagerstedt P, Grillner S, Orlovsky GN. Activity of

529 reticulospinal neurons during locomotion in the freely behaving lamprey. J

530 Neurophysiol 2000;83:853–63. https://doi.org/10.1152/jn.2000.83.2.853.

26/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

531 [22] Fagerstedt P, Orlovsky GN, Deliagina TG, Grillner S, Ullén F. Lateral turns in

532 the Lamprey. II. Activity of reticulospinal neurons during the generation of fictive

533 turns. J Neurophysiol 2001;86:2257–65.

534 https://doi.org/10.1152/jn.2001.86.5.2257.

535 [23] Kozlov AK, Kardamakis AA, Hellgren Kotaleski J, Grillner S. Gating of steering

536 signals through phasic modulation of reticulospinal neurons during locomotion.

537 Proc Natl Acad Sci U S A 2014;111:3591–6.

538 https://doi.org/10.1073/pnas.1401459111.

539 [24] Suzuki DG, Pérez-Fernández J, Wibble T, Kardamakis AA, Grillner S. The role

540 of the optic tectum for visually evoked orienting and evasive movements. Proc

541 Natl Acad Sci U S A 2019;116:15272–81.

542 https://doi.org/10.1073/pnas.1907962116.

543 [25] Huang K-H, Ahrens MB, Dunn TW, Engert F. Spinal Projection Neurons

544 Control Turning Behaviors in Zebrafish. Curr Biol 2013;23:1566–73.

545 https://doi.org/10.1016/j.cub.2013.06.044.

546 [26] Thiele TR, Donovan JC, Baier H. Descending control of swim posture by a

547 midbrain nucleus in zebrafish. Neuron 2014;83:679–91.

548 https://doi.org/10.1016/j.neuron.2014.04.018.

549 [27] Ryczko D, Thandiackal R, Ijspeert AJ. Interfacing a salamander brain with a

550 salamander-like robot: Control of speed and direction with calcium signals from

551 brainstem reticulospinal neurons. 2016b 6th IEEE International Conference on

552 Biomedical Robotics and Biomechatronics (BioRob), 2016, p. 1140–7.

553 https://doi.org/10.1109/BIOROB.2016.7523785.

27/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

554 [28] Oueghlani Z, Simonnet C, Cardoit L, Courtand G, Cazalets J-R, Morin D, et

555 al. Brainstem Steering of Locomotor Activity in the Newborn Rat. J Neurosci

556 2018;38:7725–40. https://doi.org/10.1523/JNEUROSCI.1074-18.2018.

557 [29] Bouvier J, Caggiano V, Leiras R, Caldeira V, Bellardita C, Balueva K, et al.

558 Descending Command Neurons in the Brainstem that Halt Locomotion. Cell

559 2015;163:1191–203. https://doi.org/10.1016/j.cell.2015.10.074.

560 [30] Cregg JM, Leiras R, Montalant A, Wanken P, Wickersham IR, Kiehn O.

561 Brainstem neurons that command mammalian locomotor asymmetries. Nat

562 Neurosci 2020. https://doi.org/10.1038/s41593-020-0633-7.

563 [31] Usseglio G, Gatier E, Heuzé A, Hérent C, Bouvier J. Control of Orienting

564 Movements and Locomotion by Projection-Defined Subsets of Brainstem V2a

565 Neurons. Curr Biol 2020. https://doi.org/10.1016/j.cub.2020.09.014.

566 [32] Felsen G, Mainen ZF. Neural substrates of sensory-guided locomotor

567 decisions in the rat superior colliculus. Neuron 2008;60:137–48.

568 https://doi.org/10.1016/j.neuron.2008.09.019.

569 [33] Shang C, Liu Z, Chen Z, Shi Y, Wang Q, Liu S, et al. BRAIN CIRCUITS. A

570 parvalbumin-positive excitatory visual pathway to trigger fear responses in mice.

571 Science 2015;348:1472–7. https://doi.org/10.1126/science.aaa8694.

572 [34] Zingg B, Chou X-L, Zhang Z-G, Mesik L, Liang F, Tao HW, et al. AAV-

573 Mediated Anterograde Transsynaptic Tagging: Mapping Corticocollicular Input-

574 Defined Neural Pathways for Defense Behaviors. Neuron 2017;93:33–47.

575 https://doi.org/10.1016/j.neuron.2016.11.045.

28/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

576 [35] Oliveira AF, Yonehara K. The Mouse Superior Colliculus as a Model System

577 for Investigating Cell Type-Based Mechanisms of Visual Motor Transformation.

578 Front Neural Circuits 2018;12:59. https://doi.org/10.3389/fncir.2018.00059.

579 [36] Musienko PE, Zelenin PV, Lyalka VF, Gerasimenko YP, Orlovsky GN,

580 Deliagina TG. Spinal and Supraspinal Control of the Direction of Stepping during

581 Locomotion. J Neurosci 2012;32:17442–53.

582 https://doi.org/10.1523/JNEUROSCI.3757-12.2012.

583 [37] Plaha P, Gill SS. Bilateral deep brain stimulation of the pedunculopontine

584 nucleus for Parkinson’s disease. Neuroreport 2005;16:1883–7.

585 https://doi.org/10.1097/01.wnr.0000187637.20771.a0.

586 [38] Hamani C, Aziz T, Bloem BR, Brown P, Chabardes S, Coyne T, et al.

587 Pedunculopontine Nucleus Region Deep Brain Stimulation in Parkinson Disease:

588 Surgical Anatomy and Terminology. Stereotact Funct Neurosurg 2016a;94:298–

589 306. https://doi.org/10.1159/000449010.

590 [39] Hamani C, Lozano AM, Mazzone PAM, Moro E, Hutchison W, Silburn PA, et

591 al. Pedunculopontine Nucleus Region Deep Brain Stimulation in Parkinson

592 Disease: Surgical Techniques, Side Effects, and Postoperative Imaging.

593 Stereotact Funct Neurosurg 2016b;94:307–19.

594 https://doi.org/10.1159/000449011.

595 [40] Goetz L, Bhattacharjee M, Ferraye MU, Fraix V, Maineri C, Nosko D, et al.

596 Deep Brain Stimulation of the Pedunculopontine Nucleus Area in Parkinson

597 Disease: MRI-Based Anatomoclinical Correlations and Optimal Target.

598 Neurosurgery 2019;84:506–18. https://doi.org/10.1093/neuros/nyy151.

29/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

599 [41] Bachmann LC, Matis A, Lindau NT, Felder P, Gullo M, Schwab ME. Deep

600 brain stimulation of the midbrain locomotor region improves paretic hindlimb

601 function after spinal cord injury in rats. Sci Transl Med 2013;5:208ra146.

602 https://doi.org/10.1126/scitranslmed.3005972

603 [42] Richardson M. Deep brain stimulation for locomotor recovery following spinal

604 cord injury. Neurosurgery 2014;74:N18-19.

605 https://doi.org/10.1227/01.neu.0000442979.07078.ac.

606 [43] Fluri F, Malzahn U, Homola GA, Schuhmann MK, Kleinschnitz C, Volkmann

607 J. Stimulation of the mesencephalic locomotor region for gait recovery after

608 stroke. Ann Neurol 2017;82:828–40. https://doi.org/10.1002/ana.25086.

609 [44] Chang SJ, Cajigas I, Opris I, Guest JD, Noga BR. Dissecting Brainstem

610 Locomotor Circuits: Converging Evidence for Cuneiform Nucleus Stimulation.

611 Front Syst Neurosci 2020;14:64. https://doi.org/10.3389/fnsys.2020.00064

612 [45] Vong L, Ye C, Yang Z, Choi B, Chua S, Lowell BB. Leptin action on GABAergic

613 neurons prevents obesity and reduces inhibitory tone to POMC neurons. Neuron

614 2011;71:142–54. https://doi.org/10.1016/j.neuron.2011.05.028.

615 [46] Madisen L, Mao T, Koch H, Zhuo J, Berenyi A, Fujisawa S, et al. A toolbox of

616 Cre-dependent optogenetic transgenic mice for light-induced activation and

617 silencing. Nat Neurosci 2012;15:793–802. https://doi.org/10.1038/nn.3078.

618 [47] Mathis A, Mamidanna P, Cury KM, Abe T, Murthy VN, Mathis MW, et al.

619 DeepLabCut: markerless pose estimation of user-defined body parts with deep

620 learning. Nat Neurosci 2018;21:1281–9. https://doi.org/10.1038/s41593-018-

621 0209-y.

30/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

622 [48] Nath T, Mathis A, Chen AC, Patel A, Bethge M, Mathis MW. Using

623 DeepLabCut for 3D markerless pose estimation across species and behaviors.

624 Nat Protoc 2019;14:2152–76. https://doi.org/10.1038/s41596-019-0176-0.

625 [49] Garland T, Gleeson TT, Aronovitz BA, Richardson CS, Dohm MR. Maximal

626 sprint speeds and muscle fiber composition of wild and laboratory house mice.

627 Physiol Behav 1995;58:869–76. https://doi.org/10.1016/0031-9384(95)00148-c.

628 [50] Santuz A, Akay T. Fractal analysis of muscle activity patterns during

629 locomotion: pitfalls and how to avoid them. J Neurophysiol 2020;124:1083–91.

630 https://doi.org/10.1152/jn.00360.2020.

631 [51] Leblond H, L’Esperance M, Orsal D, Rossignol S. Treadmill locomotion in the

632 intact and spinal mouse. J Neurosci 2003;23:11411–9.

633 [52] Herbin M, Hackert R, Gasc J-P, Renous S. Gait parameters of treadmill versus

634 overground locomotion in mouse. Behav Brain Res 2007;181:173–9.

635 https://doi.org/10.1016/j.bbr.2007.04.001.

636 [53] Steinkellner T, Yoo JH, Hnasko TS. Differential Expression of VGLUT2 in

637 Mouse Mesopontine Cholinergic Neurons. ENeuro 2019;6.

638 https://doi.org/10.1523/ENEURO.0161-19.2019.

639 [54] Fougère M, van der Zouwen CI, Boutin J, Ryczko D. Heterogeneous

640 expression of dopaminergic markers and Vglut2 in mouse mesodiencephalic

641 dopaminergic nuclei A8-A13. J Comp Neurol 2020.

642 https://doi.org/10.1002/cne.25020.

643 [55] Root DH, Zhang S, Barker DJ, Miranda-Barrientos J, Liu B, Wang H-L, et al.

644 Selective Brain Distribution and Distinctive Synaptic Architecture of Dual

31/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

645 Glutamatergic-GABAergic Neurons. Cell Rep 2018;23:3465–79.

646 https://doi.org/10.1016/j.celrep.2018.05.063.

647 [56] Ryczko D, Simon A, Ijspeert AJ. Walking with Salamanders: From Molecules

648 to Biorobotics. Trends Neurosci 2020;43:916–30.

649 https://doi.org/10.1016/j.tins.2020.08.006.

650 [57] Liang F, Xiong XR, Zingg B, Ji X, Zhang LI, Tao HW. Sensory Cortical Control

651 of a Visually Induced Arrest Behavior via Corticotectal Projections. Neuron

652 2015;86:755–67. https://doi.org/10.1016/j.neuron.2015.03.048.

653 [58] Juvin L, Grätsch S, Trillaud-Doppia E, Gariépy J-F, Büschges A, Dubuc R. A

654 Specific Population of Reticulospinal Neurons Controls the Termination of

655 Locomotion. Cell Rep 2016;15:2377–86.

656 https://doi.org/10.1016/j.celrep.2016.05.029.

657 [59] Grätsch S, Auclair F, Demers O, Auguste E, Hanna A, Büschges A, et al. A

658 Brainstem Neural Substrate for Stopping Locomotion. J Neurosci 2019;39:1044–

659 57. https://doi.org/10.1523/JNEUROSCI.1992-18.2018.

660 [60] Lemieux M, Bretzner F. Glutamatergic neurons of the gigantocellular reticular

661 nucleus shape locomotor pattern and rhythm in the freely behaving mouse. PLoS

662 Biol 2019;17:e2003880. https://doi.org/10.1371/journal.pbio.2003880.

663 [61] Ménard A, Auclair F, Bourcier-Lucas C, Grillner S, Dubuc R. Descending

664 GABAergic projections to the mesencephalic locomotor region in the lamprey

665 Petromyzon marinus. J Comp Neurol 2007;501:260–73.

666 https://doi.org/10.1002/cne.21258.

32/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

667 [62] Stephenson-Jones M, Samuelsson E, Ericsson J, Robertson B, Grillner S.

668 Evolutionary conservation of the basal ganglia as a common vertebrate

669 mechanism for action selection. Curr Biol 2011;21:1081–91.

670 https://doi.org/10.1016/j.cub.2011.05.001.

671 [63] Kravitz AV, Freeze BS, Parker PRL, Kay K, Thwin MT, Deisseroth K, et al.

672 Regulation of parkinsonian motor behaviours by optogenetic control of basal

673 ganglia circuitry. Nature 2010;466:622–6. https://doi.org/10.1038/nature09159.

674 [64] Botta P, Fushiki A, Vicente AM, Hammond LA, Mosberger AC, Gerfen CR, et

675 al. An amygdala circuit mediates experience-dependent momentary exploratory

676 arrests. BioRxiv 2019:797001. https://doi.org/10.1101/797001.

677 [65] Roseberry T, Kreitzer A. Neural circuitry for behavioural arrest. Philos Trans

678 R Soc Lond B Biol Sci 2017;372. https://doi.org/10.1098/rstb.2016.0197.

33/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

679 FIGURE LEGENDS

680 Figure 1. Cre-dependent expression of channelrhodopsin or ZsGreen in

681 neurons of the Mesencephalic Locomotor Region (MLR) expressing the

682 vesicular glutamate transporter 2 (Vglut2). A. For anatomical experiments,

683 homozygous mice expressing Cre-recombinase under the control of the Vglut2

684 promoter (Vglut2-Cre, see Methods) were crossed with homozygous mice with

685 ZsGreen preceded by a STOP cassette flanked by loxP sites preventing ZsGreen

686 expression. In the resulting heterozygous mice (Vglut2-ZsGreen), if Vglut2 is

687 expressed during cell lifetime, Cre-dependent recombination removes the STOP

688 cassette, allowing permanent expression of ZsGreen under control of the CAG

689 promoter. B. For optogenetic experiments, mice homozygous for Vglut2-Cre were

690 crossed with mice homozygous for channelrhodopsin (ChR2) and enhanced

691 yellow fluorescent protein (EYFP) preceded by a STOP cassette flanked by loxP

692 sites preventing their expression. In the resulting heterozygous mice (Vglut2-

693 ChR2-EYFP), if Vglut2 is expressed during cell lifetime, Cre-dependent

694 recombination removes the STOP cassette, allowing permanent expression of

695 ChR2-EYFP under control of the CAG promoter. C. Photomicrographs of

696 transversal brain slices from Vglut2-ZsGreen mice at the MLR level. D. Schematic

697 representation of a brain slice at the MLR level. E. Higher magnification of the brain

698 slice in C at the level of the CnF. F-H. Epifluorescence images taken in the same

699 field of view in the CnF, showing typical examples of cells expressing ZsGreen

700 (green, F), cells immuno-positive for glutamate (red, G), and the two markers

701 merged (H). I-J. Images taken in the CnF, showing that cells expressing ZsGreen

34/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

702 (green) are immunopositive for the neuronal marker NeuN (red). K. Whole cell

703 patch-clamp recording of a neuron recorded in a brainstem slice of a Vglut2-ChR2-

704 EYFP mouse at the level of the MLR. The neuron spikes an action potential at

705 short latency in response to a 10 ms blue light pulse. IC, inferior colliculus PAG,

706 periaqueductal grey, PPN, pedunculopontine nucleus.

35/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

707 Figure 2. Optogenetic stimulation of the cuneiform nucleus (CnF) increases

708 locomotor speed in the open field arena in Vglut2-ChR2-EYFP mice. A. An

709 optic fiber implanted in the right CnF was connected to a blue laser (470 nm) or

710 red laser (589 nm). Animals were placed in an open-field arena (40 × 40 cm) and

711 their movements were recorded using a camera placed above. B.

712 Photomicrograph showing the position of the optic fiber (dashed blue line) ~500

713 μm above the target site. The cholinergic neurons of the pedunculopontine nucleus

714 (PPN) (choline acetyltransferase positive, ChAT, red) and the expression of EYFP

715 (green) are visible. C. Location of the optic fibers (blue circles) after histological

716 verification as illustrated in B, with the relative position to the bregma. D-E. Raw

717 data showing the effects of 10 optogenetic stimulations with a 470 nm laser (E,

718 light blue lines, 10 s train, 20 Hz, 10 ms pulses, 11% of laser power) or a 589 nm

719 laser (F, red lines, 10 s train, 20 Hz, 10 ms pulses, 53% of laser power). A time

720 interval of 80 s was left between two trains of stimulation. The position of the

721 animal’s body center was tracked frame by frame with DeepLabCut (see methods).

722 Dark blue dots (D) and orange dots (E) illustrate the onset of each stimulation. F-

723 G. Locomotor speed (mean ± sem) as a function of time before, during and after a

724 10 s optogenetic stimulation (onset at t = 0 s) with a 470 nm laser (G) or 589 nm

725 laser in a single animal (same animal as in D,E). H-I. Locomotor speed (mean ±

726 sem) before during and after optogenetic stimulation with a 470 nm laser in 5

727 animals (H, 10 stimulations per animal, 10-24% of laser power) and with the 589

728 nm laser in 4 animals (I, 10 stimulations per animal, 40-53% of laser power). J-K.

729 Locomotor speed (mean ± sem) before (-10 to 0 s), during (0 to +10s), and after

36/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

730 optogenetic stimulation (+10 to +20 s and +20 to +30 s) with the 470 nm laser in 5

731 animals (J) and with the 589 nm laser in 4 animals (K) (10 stimulations per animal,

732 ** P < 0.01, n.s. not significant, P > 0.05, Student-Newman-Keuls test after a one

733 way ANOVA for repeated measures, P < 0.01 in J and P > 0.05 in K).

37/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

734 Figure 3. Gait diagrams during spontaneous locomotion and locomotion

735 evoked by optogenetic stimulation of the cuneiform nucleus (CnF) in Vglut2-

736 ChR2-EYFP mice in a linear corridor. A. Mouse forelimbs (FL) and hindlimbs

737 (HL) were filmed from below at 300 fps in a transparent linear corridor and the

738 position of each limb was tracked frame by frame with DeepLabCut (see methods).

739 The four panels show the movement speed of each paw as a function of time.

740 Cycle duration was defined as the time duration between two touchdowns of the

741 left hindlimb (HL) using a speed threshold of 15 cm/s to define the transitions

742 between swing and stance phases. Full circles are touchdowns; empty circles are

743 lift-offs. B-C. Gait diagram for each limb obtained during a single spontaneous

744 locomotor bout (B) and a locomotor bout evoked by optogenetic stimulation in the

745 same animal (470 nm laser, 10 s train, 20 Hz, 10 ms pulses, 8 % of laser power).

746 D-E. Gait diagrams during a normalized locomotor cycle, showing the stance

747 phase start (mean ± SD) and end (mean ± SD) during 16 spontaneous locomotor

748 bouts (45 steps) and during 8 locomotor bouts evoked by optogenetic stimulation

749 in the same animal (48 steps, same stimulation parameters as in E). Cycle has

750 been normalized to the left HL’s touchdown. F-G. Normalized gait diagram

751 showing the touchdown (mean ± SD) and lift-off (mean ± SD) pooled from 4 mice

752 during a total of 55 spontaneous locomotor bouts (8-16 trials per animal, 13-45

753 steps per animal) and from 4 mice during a total of 30 locomotor bouts (6-8 bouts

754 per animal, 8-48 steps per animal) evoked by CnF optogenetic stimulation (470

755 nm laser, 10 s train, 20 Hz, 10 ms pulses, 8-15% of laser power). The data from

756 each animal are illustrated with a different symbol. H-K. Comparison of cycle

38/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

757 duration (H), stride length (I), stance duration (J) and swing duration (K) in 4

758 animals during spontaneous optogenetic-evoked locomotion (same animals as in

759 F,G). *P < 0.05, **P < 0.01, n.s. not significant, P > 0.05, paired t tests).

39/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

760 Figure 4. Hindlimb kinematics during spontaneous locomotion and

761 locomotion evoked by optogenetic stimulation of the cuneiform nucleus

762 (CnF) in Vglut2-ChR2-EYFP mice in a linear corridor. A. Six hindlimb joints

763 were labeled with a white paint marker and were filmed from the side at 300 fps in

764 a transparent linear corridor and the trajectory of each joint was extracted with

765 DeepLabCut and plotted in a different color (see methods). B. Side view of the

766 hindlimb joints during spontaneous locomotion (top) and optogenetic-evoked

767 locomotion (470 nm laser, 10 s train, 20 Hz, 10 ms pulses, 8% of laser power)

768 (bottom). Total time elapsed from first to last frames is 700 ms (top) and 500 ms

769 (bottom). C. Joint angles at the hip, knee, ankle and metatarsophalangeal joint

770 (MTP) levels were calculated frame by frame using the position of the joint of

771 interest and those of two proximal joints. D. Cycle duration was defined as the time

772 duration between two consecutive touchdowns of the MTP using a speed threshold

773 of 9 cm/s to define the transitions between swing and stance phases. Full circles

774 are touchdowns; empty circles are lift-offs. E. Joint angles (mean ± SD) at the hip,

775 knee, ankle and MTP levels plotted for a normalized locomotor cycle during

776 spontaneous locomotion (29 steps) and locomotion evoked by optogenetic

777 stimulation (31 steps) (470 nm laser, 10 s train, 20 Hz, 10 ms pulses, 8% of laser

778 power). For the pooled data, joint angles (mean ± SD) of 4 animals plotted for a

779 normalized locomotor cycle during spontaneous locomotion (12-29 steps per

780 animal) and locomotion evoked by optogenetic stimulation are shown (1-31 steps

781 per animal) (470 nm laser, 10 s train, 20 Hz, 10 ms pulses, 8-15% of laser power).

782 F. Amplitude of the hip, knee, ankle and MTP angles measured in 4 animals during

40/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

783 spontaneous and optogenetic evoked locomotion (same data as in E). n.s., not

784 significant, P > 0.05, paired t tests.

41/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

785 Figure 5. Freely behaving mice brake and turn during locomotion evoked by

786 optogenetic stimulation of the cuneiform nucleus (CnF) in Vglut2-CHR2-

787 EYFP mice in the open field arena. A. Raw data showing that locomotor speed

788 is modulated during optogenetic stimulation of the CnF (470 nm laser, 10 s train,

789 20 Hz, 10 ms pulses, 11% of laser power). B. Magnification of A showing rhythmic

790 speed decrease (white circles) during CnF stimulation (blue solid line). C. Color

791 plot illustrating the locations of the speed decreases during CnF stimulation in the

792 open-field arena (same data as in A-B). Colder colors (blue) illustrate slower

793 speeds. D. Animal’s speed was measured when moving in 20 cm circles centered

794 on each corner of the arena. The speed at the entry of the corner, during the turn

795 (in a 2 cm circle centered on the location where the animal crossed the corner’s

796 diagonal) and at the exit of the corner were calculated (see methods). The turn

797 angle was measured between the positions of the animal i) at the entry of the

798 corner, ii) during the turn, and iii) when exiting the corner (see methods). E-F. Raw

799 data showing the extracted locomotor trajectories in the corners of the arena during

800 optogenetic-evoked locomotion in a single animal (E, 470 nm laser, 10 s train, 20

801 Hz, 10 ms pulses, 11% of laser power) and in 5 animals (F, 470 nm laser, 10 s

802 train, 20 Hz, 10 ms pulses, 10-24% of laser power). Warmer colors (red) illustrate

803 trajectories with higher speeds during the turn. Triangles illustrate movement

804 onsets. Dots illustrate diagonal crossings. G. Locomotor speed as a function of the

805 distance to the corner’s diagonal during each turn (grey) shown in E for a single

806 animal. In orange, the averaged speed (± sem) is shown. H. Locomotor speed as

807 a function of the distance to the diagonal during the turns of the 5 animals (An)

42/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

808 illustrated in F. In black, the mean speed (± sem in grey) is shown. I. Entry speed,

809 turn speed and exit speed for 5 animals (10 stimulations per animal, *P < 0.05, **P

810 < 0.01, n.s., not significant, P > 0.05, Student-Newman-Keuls test after a one way

811 ANOVA for repeated measures, P < 0.01). J. Relationship between the speed at

812 corner entry, and the difference between entry speed and turn speed in 5 mice

813 (linear fit, P < 0.0001, R = 0.90, n = 87 turns pooled from 50 stimulations, 10

814 stimulations per animal). K. Relationship between the speed at corner exit, and the

815 difference between exit speed and turn speed (linear fit, P < 0.0001, R = 0.92). L.

816 Relationship between the speed at corner entry and the turn speed (linear fit, P <

817 0.05, R = 0.27). M. Relationship between the turn speed and the turn angle (linear

818 fit, P < 0.0001, R = 0.44).

43/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

819 Figure 6. Robustness of scaled control of speed during turning at different

820 speeds controlled by the level of optogenetic stimulation of the cuneiform

821 nucleus (CnF) in Vglut2-ChR2-EYFP mice in the open field arena. A. Color plot

822 illustrating the increase in locomotor speed (cm/s) evoked by increases in laser

823 power (6-13%) in a single animal (470 nm laser, 10 s train, 20 Hz, 10 ms pulses).

824 Each line illustrates the speed as a function of time for a given laser power

825 expressed as a percentage of maximal laser power used for this animal. White

826 dotted lines indicate the onset and offset of optogenetic stimulation. Warmer colors

827 (red) indicate higher speeds. B. Locomotor speed (1-42 cm/s) as a function of laser

828 power (6-13%) in one animal. Each dot represents the speed (mean ± sem) over

829 3 stimulations. Speed and laser power were expressed as a function of their

830 maximal values. C. Relationship between locomotor speed (0.2-42.0 cm/s) and

831 increasing laser power (6-27%) for all animals. Data from each mouse are

832 illustrated with a different color. Each dot represents the speed (mean ± sem) over

833 3 stimulations. Speed and laser power were expressed as a percentage of their

834 maximal values per animal. D. Relationship between locomotor speed (mean ±

835 sem) and laser power in the same animals as in C, this time with data binned as a

836 function of laser power used per animal with a bin size of 10%. The data followed

837 a sigmoidal function (solid black line, P < 0.01, R = 0.99). The dotted lines illustrate

838 the 95% prediction intervals. E, G. Raw data showing the extracted locomotor

839 trajectories in the corners of the arena during optogenetic-evoked locomotion in a

840 single animal (E) and in 5 animals (G). Triangles illustrate movement onsets. Dots

841 illustrate diagonal crossings. Warmer colors (red) illustrate higher turn speeds. F,

44/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

842 H. Locomotor speed as a function of the distance to the diagonal during the turns

843 for increasing power of optogenetic stimulation of the CnF in a single animal (F,

844 470 nm laser, 10 s train, 20 Hz, 10 ms pulses, 6-13% of laser power) and in 5

845 animals (H, 470 nm laser, 10 s train, 20 Hz, 10 ms pulses, 6-27% of laser power).

846 Warmer colors (red) indicate stronger optogenetic stimulation of the CnF. Laser

847 powers were normalized as a percentage of their maximal value used per animal

848 and were binned in H (bin width: 10%). In F, each curve was obtained from 1-14

849 turns in a single animal. In H, each curve was obtained from 2-73 turns pooled

850 from 5 animals. I. Relationship between the speed at corner entry, and the

851 difference between entry speed and turn speed (linear fit, P < 0.0001, R = 0.87, N

852 = 157 turns pooled from 150 stimulations, 30 stimulations per animal). J.

853 Relationship between the speed at corner exit, and the difference between exit

854 speed and turn speed (linear fit, P < 0.0001, R = 0.93). K. Relationship between

855 the speed at corner entry and the turn speed (linear fit, P < 0.0001, R = 0.47). L.

856 Relationship between the turn speed and the turn angle (linear fit, P < 0.0001, R =

857 0.49).

45/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

858 Figure 7. Coordinated control of locomotor speed and turning movements

859 during optogenetic stimulation of the Mesencephalic Locomotor Region

860 (MLR). Illustration of the relation between motion direction, locomotor speed, and

861 MLR optogenetic stimulation level observed during the present study. The neural

862 circuits likely involved before, during, and after the turn are illustrated. Before the

863 turn, forward locomotion is evoked by unilateral stimulation of glutamatergic

864 neurons of the cuneiform nucleus (CnF, part of the MLR) that provides bilateral

865 activation of reticulospinal (RS) neurons located in the lateral paragigantocellular

866 nucleus (LPGi) that project to the spinal neurons of the central pattern generator

867 (CPG) for locomotion (mouse [20, 8, 9, 10, 11, 12]. Also see corresponding studies

868 in lamprey [5, 6, 14, 15], zebrafish [16, 17], and salamander [7, 18]). During the

869 turn, the visual inputs conveying the approach of the corner are relayed by the

870 superior colliculus (Sup colli) that sends projections to contralateral reticulospinal

871 neurons of the gigantocellularis nucleus (Gi) that evoke ipsilateral braking and

872 turning movements [29, 30, 31]; also see studies on the role of reticulospinal

873 neurons in steering control in lamprey [21, 22, 23, 24], zebrafish [25, 26],

874 salamander [27], and rat [28]). After the turn, the sensory inputs generated by the

875 corner disappear, Gi neurons are deactivated, speed increases back to the value

876 set by the steady MLR command, and forward symmetrical locomotion is restored.

46/46 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

A Vglut2-ZsGreen

Vglut2 Cre CAG STOP ZsGreen Vglut2 Cre loxP loxP

Vglut2 Cre CAG STOP ZsGreen CAG ZsGreen loxP loxP loxP B Vglut2-ChR2-EYFP

Vglut2 Cre CAG STOP ChR2 EYFP Vglut2 Cre loxP loxP

Vglut2 Cre CAG STOP ChR2 EYFP CAG ChR2 EYFP loxP loxP loxP C D E IC

IC PAG CnF PAG CnF

PPN ZsGreen 500 µm 500 µm IF G HJ

Scale ZsGreen 50 µm 50 µm 50 µm ZsGreen Glutamate Glutamate I J K Vglut2-ChR2-EYFP

20 mV 10 ms -58 mV ZsGreen NeuN 50 µm NeuN 50 µm Opto

Figure 1 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

A 470 or D E 589 nm Camera Laser

Open- field

470 nm OFF 589 nm OFF 470 nm ON 589 nm ON Vglut2-ChR2-EYFP mice F Single animal G Single animal 60 60 B IC 470 nm 589 nm 40 40 PAG CnF 20 20 Speed (cm/s) Speed (cm/s)

PPN 0 0 -10 0 10 20 30 -10 0 10 20 30 EYFP 500 µm ChAT Time (s) Time (s) C -4.72 mm H N=5 mice I N=4 mice 40 40 470 nm 589 nm

20 20

-4.96 mm Speed (cm/s) Speed (cm/s)

0 0 -10 0 10 20 30 -10 0 10 20 30 Time (s) Time (s) J N=5 mice K N=4 mice 40 40 -5.02 mm ** ** n.s. n.s. ** n.s.

20 20 Speed (cm/s) Speed (cm/s) 0 0 -10s Stim +10s +20s -10s Stim +10s +20s Time bin Time bin

Figure 2 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

A Gait analysis B Single trial (spontaneous) C Single trial (optogenetic) Left HL Left HL Left FL Left FL Right FL Right FL Left HL Right HL Right HL 100 200 ms 200 ms

D Single animal (spontaneous) E Single animal (optogenetic)

Speed (cm/s) Left HL Left HL Left FL Left FL 0 Right FL Right FL Left FL Right HL 100 Right HL 0 Cycle 1 0 Cycle 1 F N=4 mice (spontaneous) G N=4 mice (optogenetic)

Speed (cm/s) Left HL Left HL 0 Left FL Left FL Right FL Right FL Right FL 100 Right HL Right HL 0 Cycle 1 0 Cycle 1 H N=4 I N=4 J N=4 K N=4

Speed (cm/s) mice mice mice mice n.s. n.s. 0 0.25 * 12 0.08 ** 0.12 Right HL 100 (s) (cm) Speed (cm/s) Swing (s) Stance (s) Stride length 0 Cycle duration 0 0 0 0 0 Time (s) 1 Spont Opto Spont Opto Spont Opto Spont Opto

Figure 3 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

A B Progression Spontaneous

Iliac crest Hip Optogenetic-evoked 1 cm Knee Ankle MTP Toe tip

C D Touchdown Lift-off E Spontaneous Optogenetic-evoked 90 Single animal N=4 mice Single animal N=4 mice Hip 200 Knee 60 100 Ankle 30 Hip MTP Angle (°) 0 0 200

Met speed (cm/s) 0 Time (s) 1.2 100 Knee F Hip Knee Ankle MTP Angle (°) 0 n.s. n.s. n.s. n.s. 200 120 100 Ankle 80 Angle (°) 0 200 40 100 MTP Angle (°) Angle amplitude (°) 0 0 0 Cycle 1 0 Cycle 1 0 Cycle 1 0 Cycle 1 Spontaneous Optogenetic-evoked (%) (%) (%) (%)

Figure 4 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

A Opto stim B Opto stim C Open-field arena 140 120

D 80 70 40 Speed (cm/s) Speed (cm/s) 0 0 5 s 1 s 5 cm D Analysis of the turns E Single animal F N=5 mice r = 20 cm 0 Distance (cm) 20 0 Distance (cm) 20 r = 0 0 2 cm Turn Entry angle

Diagonal Distance (cm) Exit Distance (cm)

6 Turn speed 38 3 Turn speed 38 20 (cm/s) 20 (cm/s) G Single animal H N=5 mice I N=5 mice 100 * ** An 1 60 120 An 2 n.s. 80 An 3 An 4 80 60 An 5 40

Speed (cm/s) 40 Speed (cm/s) 20 Speed (cm/s)

0 0 0 -20 -10 0 10 20 -20 -10 0 10 20 Entry Turn Exit Distance to turning point (cm) Distance to turning point (cm) J N=5 mice K N=5 mice L N=5 mice M N=5 mice 40 160 80 80 30 120 40 40 20 80 10 40 0 P < 0.0001 0 P < 0.0001 P < 0.05 P < 0.0001 R = 0.90 R = 0.92 R = 0.27 Turn angle (°) R = 0.44 0 0

20 60 100 20 60 100 Turn speed (cm/s) 0 40 80 120 0 10 20 30 40

Speed decrease (cm/s) Entry speed (cm/s) Speed increase (cm/s) Exit speed (cm/s) Entry speed (cm/s) Turn speed (cm/s)

CnF stim (% max)

Figure 5 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

A 100 90 B Single animal C N=5 mice D N=5 mice 100 100 70 (5) An 1 P < 0.01 (5) 80 An 2 R = 0.99 80 An 3 50 (5) An 4 60 60 An 5 (5) 30 (3) 40 40 (% max) (3) Laser power Speed (cm/s) 20 10 20 Speed (% max) Speed (% max) Speed (% max) 45 0 0 -10 0 40 60 80 100 40 60 80 100 40 60 80 100 470 nm 5 s Laser power (% max) Laser power (% max) Laser power (% max)

E Single animal F Single animal G N=5 mice H N=5 mice 0 Distance (cm) 20 Laser power (% max) 0 Distance (cm) 20 Laser power (% max) 0 0 64% 100% 40% 100% 80 80 60 60 40 40 20 20 Distance (cm) Distance (cm) Speed (cm/s) Speed (cm/s) 5 34 0 3 35 0 Turn speed -20 -10 0 10 20 Turn speed -20 -10 0 10 20 20 (cm/s) Distance to turning point (cm) 20 (cm/s) Distance to turning point (cm)

I N=5 mice J N=5 mice K N=5 mice L N=5 mice 60 P < 0.0001 80 P < 0.0001 40 P < 0.0001 140 R = 0.87 60 R = 0.93 R = 0.47 40 30 40 100 20 20 20 60 0 10 Turn angle (°) 0 P < 0.0001 R = 0.49 -20 Turn speed (cm/s) 0 20 0 20 40 60 80 020 40 60 80 0 10 20 30 40

Speed increase (cm/s) 0 20 40 60 80 100 Speed decrease (cm/s) Entry speed (cm/s) Exit speed (cm/s) Entry speed (cm/s) Turn speed (cm/s)

Figure 6 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

DIRECTION STRAIGHT TURN STRAIGHT

Speed

Sensory input MLR SPEED opto activation

MLR OPTOGENETIC STIMULATION LEVEL

NEURAL CIRCUITS INVOLVED Sensory input

Sup colli

CnF CnF CnF MLR

RS LPGi Gi LPGi LPGi

SC CPG CPG CPG

Figure 7 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

1 SUPPLEMENTARY MATERIAL

2 Genotyping. Mice were genotyped as previously described [1]. Briefly, DNA was

3 extracted from ear punches using the Taq DNA polymerase (NEB, Ipswich, MA,

4 United States). Genotyping was performed using the primers recommended by the

5 supplier (Jackson laboratory). Vglut2-ires-Cre mice were genotyped using mixed

6 primer PCR employing Vglut2-ires-Cre-Com-F (AAGAAGGTGCGCAAGACG),

7 Vglut2-ires-Cre-Wt-R (CTGCCACAGATTGCACTTGA) and Vglut2-ires-Cre-Mut-R

8 (ACACCGGCCTTATTCCAAG). Amplification of wild-type genomic DNA yielded a

9 245 bp PCR product whereas amplification from the mutant locus yielded a 124 bp

10 PCR product. ChR2-lox mice and ZsGreen-lox mice were genotyped using mixed

11 primer PCR employing ZsGreen-ChR2-lox-Wt-F (AAGGGAGCTGCAGTGGAG

12 TA), ZsGreen-ChR2-lox-Wt-R (CCGAAAATCTGTGGGAAGTC), ZsGreen-ChR2-

13 lox-Mut-R (GGCATTAAAGCAGCGTATCC), and either ChR2-lox-Mut-F

14 (ACATGGTCCTGCTGGAGTTC) or ZsGreen-lox-Mut-F

15 (AACCAGAAGTGGCACCTGAC). Amplification of wild-type genomic DNA yielded

16 a 297 bp PCR product whereas amplification from the mutant ChR2-lox locus

17 yielded a 212 bp PCR product and amplification of the mutant ZsGreen-lox locus

18 yielded a 199 bp PCR product.

19

20 Patch-clamp recordings. Coronal brainstem slices were obtained from 15-23-day

21 old mice as previously described [2]. Briefly, mice were anesthetized with

22 isoflurane (0.5-1 mL of isoflurane in a 1.5 L induction chamber) and decapitated

23 with a guillotine. The cranium was opened and the brain removed to be dipped in

1/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

24 an ice-cold sucrose-based solution (in mM: 3 KCl, 1.25 KH2PO4, 4 MgSO4, 26

25 NaHCO3, 10 Dextrose, 0.2 CaCl2, 219 Sucrose, pH 7.3–7.4, 300-320 mOsmol/kg)

26 bubbled with 95% O2 and 5% CO2. MLR slices (350 μm thick) were prepared with

27 a VT1000S vibrating blade microtome (Leica Microsystems, Concord, ON,

28 Canada) and stored at room temperature for 1 hour in artificial cerebrospinal fluid

29 (aCSF) (in mM: 124 NaCl, 3 KCl, 1.25 KH2PO4, 1.3 MgSO4, 26 NaHCO3, 10

30 Dextrose, and 1.2 CaCl2, pH 7.3–7.4, 290–300 mOsmol/kg) bubbled with 95% O2

31 and 5% CO2. Whole-cell patch-clamp recordings were done in a chamber perfused

32 with bubbled aCSF under an Axio Examiner Z1 epifluorescent microscope (Zeiss,

33 Toronto ON Canada), differential interference contrast (DIC) components, and an

34 ORCA-Flash 4.0 Digital CMOS Camera V3 (Hamamatsu Photonics, Hamamatsu,

35 Japan). Patch pipettes were pulled from borosilicate glass capillaries (1.0 mm

36 outside diameter, 0.58 mm inside diameter; 1B100F-4, World Precision

37 Instruments, FL, USA) using a P-1000 puller (Sutter Instruments). Pipettes with a

38 resistance of 6–12 MΩ were filled with a solution containing (in mM) 140 K-

39 gluconate, 5 NaCl, 2 MgCl2, 10 HEPES, 0.5 EGTA, 2 Tris ATP salt, 0.4 Tris GTP

40 salt, pH 7.2–7.3, 280–300 mOsmol/kg, 0.05 Alexa Fluor 594 or 488, and 0.2%

41 biocytin). Positive pressure was applied through the glass pipette and neurons

42 were approached using a motorized micromanipulator (Sutter instruments). A

43 gigaseal was established and the membrane potential was held at -60 mV. The

44 membrane patch was suctioned, and the pipette resistance and capacitance were

45 compensated electronically. Neurons were discarded when action potentials were

46 less than 40 mV or when the resting membrane potential was too depolarized (>-

2/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

47 45 mV). Patch-clamp signals were acquired with a Multiclamp 700B coupled to a

48 Digidata 1550B and a computer equipped with PClamp 10 software (Molecular

49 Devices, Sunnyvale, CA, USA). Optogenetic stimulations (475 nm, 10 ms pulses,

50 2.5-5% of LED power) were applied using the 475 nm LED of a Colibri 7

51 illumination system (Zeiss).

52

53 Histology and immunofluorescence. Procedures were as previously reported

54 [1]. Briefly, mice were anaesthetized using isoflurane (5%, 2.5 L per minute) and

55 transcardially perfused with 50 mL of a phosphate buffer solution (0.1M) containing

56 0.9% NaCl (PBS, pH = 7.4), followed by 40 to 75 mL of a PBS solution containing

57 4% (wt/vol) of paraformaldehyde (PFA 4%). Post-fixation of the brains was

58 performed in a solution of PFA 4% for 24 h at 4°C. Then, the brains were incubated

59 in a PBS solution containing 20% (wt/vol) sucrose for 24 h before histology. Brains

60 were snap frozen in methylbutane (-45°C ± 5°C) and sectioned at -20°C in 40 µm-

61 thick coronal slices using a cryostat (Leica CM 1860 UV). Floating sections of the

62 MLR were collected under a Stemi 305 stereomicroscope (Zeiss) and identified

63 using the atlas of Franklin and Paxinos (2008) [3].

64 For immunofluorescence experiments, all steps were carried out at room

65 temperature unless stated otherwise. The sections were rinsed in PBS for 10 min

66 three times and incubated for 1h in a blocking solution containing 5% (vol/vol) of

67 normal donkey serum and 0.3% Triton X-100 in PBS. The sections were then

68 incubated at 4°C for 48 h in a PBS solution containing the primary antibody against

69 choline acetyltransferase (ChAT) (goat anti-choline acetyltransferase, Sigma

3/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

70 AB144P, lot 3018862 (1:100), RRID AB_2079751), the neuronal marker NeuN

71 (rabbit anti-NeuN, Abcam AB177487, lot GR3250076-6 (1:1,000), RRID

72 AB_2532109) or glutamate (rabbit anti-glutamate, Sigma G6642, lot 079M4802V

73 (1:3,000), RRID AB_259946) and agitated with an orbital shaker. The sections

74 were washed three times in PBS and incubated for 4 h in a solution containing the

75 appropriate secondary antibody to reveal ChAT (donkey anti-goat Alexa 594,

76 Invitrogen A11058, lot 1975275 (1:400), RRID AB_2534105) NeuN or glutamate

77 (with a donkey anti-rabbit Alexa Fluor 594, Invitrogen A21207 lot 1890862 (1:400),

78 RRID: AB_141637; or a donkey anti-rabbit Alexa Fluor 647, ThermoFisher A31573

79 lot 2083195 (1:400), RRID: AB_2536183). The slices were rinsed three times in

80 PBS for 10 min and mounted on Colorfrost Plus glass slides (Fisher) with a medium

81 with DAPI (Vectashield H-1200) or without DAPI (Vectashield H-1000), covered

82 with a 1.5 type glass coverslip and stored at 4°C before observation. Brain sections

83 were observed using a Zeiss AxioImager M2 microscope equipped with

84 StereoInvestigator 2018 software (v1.1, MBF Bioscience). Composite images

85 were assembled using StereoInvestigator. The levels were uniformly adjusted in

86 Photoshop CS6 (Adobe) to make all fluorophores visible and avoid pixel saturation,

87 and digital images were merged.

88

89 Specificity of the antibodies. The AB177487 anti-NeuN has been widely used to

90 label the neuronal marker NeuN (also called Fox-3, see [4, 5]) in mouse brain

91 tissues by us [1] and others [6, 7]. According to the supplier, this monoclonal

92 purified antibody (clone EPR12763) is directed towards a synthetic peptide of the

4/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

93 residues 1-100 of the human NeuN. NeuN is present in most mouse neurons, but

94 not in cerebellar Purkinje cells, olfactory bulb mitral cells, and retinal photoreceptor

95 cells [4]. According to the supplier, AB177487 labels NeuN in HeLa cell lysates

96 and in brains of mice, rats and humans. It detects two bands at 45-50 kDA in

97 Western blots performed on mouse, rat or human brain tissues.

98 The G6642 anti-glutamate polyclonal antibody was used to label

99 glutamatergic neurons in zebrafish, rat and mouse (see below). The dot-blot

100 immunoassays carried out by the supplier and by Terada et al. (2009) [8] indicated

101 that this antiserum recognizes L-glutamate, glutamate conjugated to keyhole

102 limpet hemocyanin (KLH), glutamate conjugated to bovine serum albumin (BSA),

103 KLH and shows no cross-reactivity with L-aspartate, L-glutamine, L-asparagine, L-

104 alanine or BSA, and only weak cross-reactivity with glycyl-L-aspartic acid, GABA,

105 β-alanine, glycine and 5-aminovaleric acid. Pre-incubation of spinal cord slices

106 with glutamate eliminates G6642 immunoreactivity in zebrafish [9].

107 Immunostaining with G6642 labels the majority of neurons expressing a

108 fluorescent protein under control of the Vglut2 promoter [9] or under control of the

109 Chx10 promoter in zebrafish (i.e. V2a neurons, a population of spinal glutamatergic

110 interneurons that generate the locomotor rhythm from fish to mammals, [10]).

111 Double labeling with Neurotrace fluorescent Nissl stain indicated that cells labelled

112 by G6642 are neurons in zebrafish spinal cord [10]. The G6642 antibody was

113 successfully used to stain glutamate in photoreceptor cells in mouse retina [8],

114 glutamatergic neurons in rat brainstem [11, 12] or to distinguish glutamatergic from

5/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

115 glycinergic and/or GABAergic populations in zebrafish spinal cord [10], rat auditory

116 brainstem circuits [13] and mouse amygdala [14].

117 The AB144P ChAT antibody has been widely used to label cholinergic

118 neurons in lamprey [15, 16, 17], salamander [18, 19, 20], rat [21], human [22, 21]

119 and mouse brain tissues [23]. This affinity purified polyclonal antibody is raised

120 against the human placental enzyme. The supplier has tested its specificity in

121 human placenta lysates and using western blots on mouse brain lysates, where it

122 detects a band of 68-70 kDA. It labels neurons expressing a fluorescent protein

123 under control of the ChAT promoter in mice [24].

124

125 Specificity of the transgenic mice. Vglut2-ires-Cre mouse. These mice are

126 widely used to express Cre-recombinase in glutamatergic Vglut2-positive neurons

127 without interfering with Vglut2 gene expression [25]. When Vglut2-ires-Cre are

128 crossed with a lox-GFP mouse, GFP-positive neurons are found in glutamatergic

129 regions (positive for Vglut2 mRNA) and absent from GABAergic regions (positive

130 for the vesicular GABA transporter mRNA) [25]. When Vglut2-ires-Cre are crossed

131 with a lox-tdTomato mouse, the cells labelled in the dorsal horn of the spinal cord

132 are immuno-positive for NeuN and immuno-negative for Pax2 and for Wilm’s tumor

133 1, two markers of inhibitory neurons [26, 27]. Chemogenetic activation of Vglut2-

134 Cre neurons increases the frequency of synaptic excitatory currents recorded with

135 patch-clamp in spinal cord slices [26] and evokes short latency excitatory

136 responses in periaqueductal gray neurons [28]. Excitatory postsynaptic responses

137 are evoked in the striatum when stimulating thalamic terminals in mice obtained by

6/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

138 crossing the Vglut2-ires-Cre with lox-channelrhodopsin (ChR2) mice [29]. The

139 Vglut2-ire-Cre mouse was used to study the role of reticulospinal neurons in

140 locomotor control [30].

141 ZsGreen-lox mouse. The Ai6 mouse has been widely used to label cells

142 expressing the Cre-recombinase [31]. After exposure to Cre-recombinase, the

143 floxed STOP cassette is removed, and this results in the expression of the

144 ZsGreen fluorescent protein under control of CAG promoter. Cells display intense

145 labelling with ZsGreen as demonstrated by us [1] and others [32, 23]. In our

146 previous study, we compared ZsGreen-lox mouse (Cre-negative) and Vglut2-

147 ZsGreen (Cre-positive) brain sections and confirmed that prior to introduction of

148 Cre-recombinase, only a very low baseline level of fluorescence was present in

149 brain slices of homozygous ZsGreen-lox mice [1]. This is classical for reporter lines

150 based on CAG promoter-driven expression (e.g. Ai9, tdTomato-lox mouse) as

151 mentioned by the supplier (Jackson laboratory).

152 ChR2-EYFP-lox mouse. The Ai32 mouse [31] has been widely used to activate

153 cells expressing the Cre-recombinase using optogenetics (e.g. [33, 34]). When

154 exposed to Cre-recombinase, the floxed STOP cassette is removed, and this

155 results in the expression of the ChR2(H134R)-EYFP fusion protein under control

156 of CAG promoter.

157

158 DeepLabCut networks. For open-field locomotion analysis, we labelled 6

159 landmarks on 520 frames taken from 20 videos of 8 different animals assigning the

160 95% of those images to the training set without cropping. The landmarks were the

7/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

161 body center, the four corners of the arena, and the low-power LED to visualize

162 optogenetic stimulation. We used a ResNet-50-based neural network [35, 36] with

163 default parameters for 1,030,000 training iterations. We validated with one shuffle

164 and found the test error was 2.28 pixels and the train error 1.85 pixels.

165 For footfall pattern analysis, we labelled 9 landmarks 489 frames taken from

166 27 videos of 9 animals assigning the 95% of those images to the training set

167 without cropping. The landmarks were the four paws, the four distance calibration

168 markers, and the low-power LED to visualize optogenetic stimulation. We used a

169 ResNet-50-based neural network [35, 36] with default parameters for 1,030,000

170 training iterations. We validated with one shuffle and found the test error was 2.31

171 pixels and the train error 1.76 pixels.

172 For limb kinematics analysis, we labelled 11 landmarks on 906 frames taken

173 from 44 videos of 7 different animals assigning the 95% of those images to the

174 training set without cropping. The landmarks were the 5 joints and the toe tip, the

175 four distance calibration markers, and the low-power LED to visualize optogenetic

176 stimulation. We used a ResNet-50-based neural network [35, 36] with default

177 parameters for 1,030,000 training iterations and one refinement of 1,030,000

178 iterations. We validated with one shuffle and found the test error was 2.03 pixels

179 and the train error 1.87 pixels.

8/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

180 SUPPLEMENTARY REFERENCES

181 [1] Fougère M, van der Zouwen CI, Boutin J, Ryczko D. Heterogeneous expression

182 of dopaminergic markers and Vglut2 in mouse mesodiencephalic dopaminergic

183 nuclei A8-A13. J Comp Neurol 2020. https://doi.org/10.1002/cne.25020.

184 [2] Ryczko D, Hanini-Daoud M, Condamine S, Bréant BJB, Fougère M, Araya R,

185 et al. Astrocytic modulation of information processing by layer 5 pyramidal

186 neurons of the mouse visual cortex. BioRxiv 2020:2020.07.09.190777.

187 https://doi.org/10.1101/2020.07.09.190777.

188 [3] Franklin, K., & Paxinos, G. (2008). The mouse brain in stereotaxic

189 coordinates, compact (3rd ed. Cambridge, Massachusetts, USA). Academic

190 Press.

191 [4] Mullen RJ, Buck CR, Smith AM. NeuN, a neuronal specific nuclear protein in

192 vertebrates. Development 1992;116:201–11

193 [5] Kim KK, Adelstein RS, Kawamoto S. Identification of neuronal nuclei (NeuN) as

194 Fox-3, a new member of the Fox-1 gene family of splicing factors. J Biol Chem

195 2009;284:31052–61. https://doi.org/10.1074/jbc.M109.052969.

196 [6] Saito K, Koike T, Kawashima F, Kurata H, Shibuya T, Satoh T, et al.

197 Identification of NeuN immunopositive cells in the adult mouse subventricular

198 zone. J Comp Neurol 2018;526:1927–42. https://doi.org/10.1002/cne.24463.

199 [7] Joy MT, Ben Assayag E, Shabashov-Stone D, Liraz-Zaltsman S, Mazzitelli J,

200 Arenas M, et al. CCR5 Is a Therapeutic Target for Recovery after Stroke and

201 Traumatic Brain Injury. Cell 2019;176:1143-1157.e13.

202 https://doi.org/10.1016/j.cell.2019.01.044.

9/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

203 [8] Terada N, Ohno N, Saitoh S, Saitoh Y, Ohno S. Immunoreactivity of glutamate

204 in mouse retina inner segment of photoreceptors with in vivo cryotechnique. J

205 Histochem Cytochem 2009;57:883–8. https://doi.org/10.1369/jhc.2009.953851.

206 [9] Berg EM, Bertuzzi M, Ampatzis K. Complementary expression of calcium

207 binding proteins delineates the functional organization of the locomotor network.

208 Brain Struct Funct 2018;223:2181–96. https://doi.org/10.1007/s00429-018-

209 1622-4.

210 [10] Pedroni A, Ampatzis K. Large-Scale Analysis of the Diversity and Complexity

211 of the Adult Spinal Cord Neurotransmitter Typology. IScience 2019;19:1189–

212 201. https://doi.org/10.1016/j.isci.2019.09.010.

213 [11] Lin LH, Emson PC, Talman WT. Apposition of neuronal elements containing

214 nitric oxide synthase and glutamate in the nucleus tractus solitarii of rat: a

215 confocal microscopic analysis. Neuroscience 2000;96:341–50.

216 https://doi.org/10.1016/s0306-4522(99)00560-6.

217 [12] Liang C-L, Quang Nguyen T, Marks GA. Inhibitory and excitatory amino acid

218 neurotransmitters are utilized by the projection from the dorsal deep

219 mesencephalic nucleus to the sublaterodorsal nucleus REM sleep induction

220 zone. Brain Res 2014;1567:1–12.

221 https://doi.org/10.1016/j.brainres.2014.04.016.

222 [13] Fredrich M, Reisch A, Illing R-B. Neuronal subtype identity in the rat auditory

223 brainstem as defined by molecular profile and axonal projection. Exp Brain Res

224 2009;195:241–60. https://doi.org/10.1007/s00221-009-1776-7.

10/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

225 [14] Xu P, Cao X, He Y, Zhu L, Yang Y, Saito K, et al. Estrogen receptor-α in medial

226 amygdala neurons regulates body weight. J Clin Invest 2015;125:2861–76.

227 https://doi.org/10.1172/JCI80941.

228 [15] Pombal MA, Marín O, González A. Distribution of choline acetyltransferase-

229 immunoreactive structures in the lamprey brain. J Comp Neurol 2001;431:105–

230 26. https://doi.org/10.1002/1096-9861(20010226)431:1<105::aid-

231 cne1058>3.0.co;2-p.

232 [16] Le Ray D, Brocard F, Bourcier-Lucas C, Auclair F, Lafaille P, Dubuc R.

233 Nicotinic activation of reticulospinal cells involved in the control of swimming in

234 lampreys. Eur J Neurosci 2003;17:137–48. https://doi.org/10.1046/j.1460-

235 9568.2003.02417.x.

236 [17] Quinlan KA, Buchanan JT. Cellular and synaptic actions of acetylcholine in

237 the lamprey spinal cord. J Neurophysiol 2008;100:1020–31.

238 https://doi.org/10.1152/jn.01157.2007.

239 [18] Marín O, Smeets WJ, González A. Distribution of choline acetyltransferase

240 immunoreactivity in the brain of anuran (Rana perezi, Xenopus laevis) and

241 urodele (Pleurodeles waltl) amphibians. J Comp Neurol 1997;382:499–534.

242 https://doi.org/10.1002/(sici)1096-9861(19970616)382:4<499::aid-

243 cne6>3.0.co;2-y.

244 [19] Cabelguen J-M, Bourcier-Lucas C, Dubuc R. Bimodal locomotion elicited by

245 electrical stimulation of the midbrain in the salamander Notophthalmus

246 viridescens. J Neurosci 2003;23:2434–9.

11/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

247 [20] Ryczko D, Auclair F, Cabelguen J-M, Dubuc R. The mesencephalic locomotor

248 region sends a bilateral glutamatergic drive to hindbrain reticulospinal neurons

249 in a tetrapod. J Comp Neurol 2016;524:1361–83.

250 https://doi.org/10.1002/cne.23911.

251 [21] Ryczko D, Cone JJ, Alpert MH, Goetz L, Auclair F, Dubé C, et al. A descending

252 dopamine pathway conserved from basal vertebrates to mammals. Proc Natl

253 Acad Sci USA 2016;113:E2440-2449.

254 https://doi.org/10.1073/pnas.1600684113.

255 [22] Massouh M, Wallman M-J, Pourcher E, Parent A. The fate of the large striatal

256 interneurons expressing calretinin in Huntington’s disease. Neurosci Res

257 2008;62:216–24. https://doi.org/10.1016/j.neures.2008.08.007.

258 [23] Steinkellner T, Yoo JH, Hnasko TS. Differential Expression of VGLUT2 in

259 Mouse Mesopontine Cholinergic Neurons. ENeuro 2019;6.

260 https://doi.org/10.1523/ENEURO.0161-19.2019.

261 [24] Bloem B, Schoppink L, Rotaru DC, Faiz A, Hendriks P, Mansvelder HD, et al.

262 Topographic mapping between basal forebrain cholinergic neurons and the

263 medial prefrontal cortex in mice. J Neurosci 2014;34:16234–46.

264 https://doi.org/10.1523/JNEUROSCI.3011-14.2014.

265 [25] Vong L, Ye C, Yang Z, Choi B, Chua S, Lowell BB. Leptin action on GABAergic

266 neurons prevents obesity and reduces inhibitory tone to POMC neurons.

267 Neuron 2011;71:142–54. https://doi.org/10.1016/j.neuron.2011.05.028.

12/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

268 [26] Wang L, Chen S-R, Ma H, Chen H, Hittelman WN, Pan H-L. Regulating

269 nociceptive transmission by VGluT2-expressing spinal dorsal horn neurons. J

270 Neurochem 2018;147:526–40. https://doi.org/10.1111/jnc.14588.

271 [27] Haque F, Rancic V, Zhang W, Clugston R, Ballanyi K, Gosgnach S. WT1-

272 Expressing Interneurons Regulate Left-Right Alternation during Mammalian

273 Locomotor Activity. J Neurosci 2018;38:5666–76.

274 https://doi.org/10.1523/JNEUROSCI.0328-18.2018.

275 [28] Falkner AL, Wei D, Song A, Watsek LW, Chen I, Chen P, et al. Hierarchical

276 Representations of Aggression in a Hypothalamic-Midbrain Circuit. Neuron

277 2020;106:637-648.e6. https://doi.org/10.1016/j.neuron.2020.02.014.

278 [29] Johnson KA, Voyvodic L, Loewinger GC, Mateo Y, Lovinger DM. Operant self-

279 stimulation of thalamic terminals in the dorsomedial striatum is constrained by

280 metabotropic glutamate receptor 2. Neuropsychopharmacology 2020.

281 https://doi.org/10.1038/s41386-020-0626-y.

282 [30] Capelli P, Pivetta C, Soledad Esposito M, Arber S. Locomotor speed control

283 circuits in the caudal brainstem. Nature 2017;551:373–7.

284 https://doi.org/10.1038/nature24064.

285 [31] Madisen L, Mao T, Koch H, Zhuo J, Berenyi A, Fujisawa S, et al. A toolbox of

286 Cre-dependent optogenetic transgenic mice for light-induced activation and

287 silencing. Nat Neurosci 2012;15:793–802. https://doi.org/10.1038/nn.3078.

288 [32] Steinkellner T, Zell V, Farino ZJ, Sonders MS, Villeneuve M, Freyberg RJ, et

289 al. Role for VGLUT2 in selective vulnerability of midbrain dopamine neurons. J

290 Clin Invest 2018;128:774–88. https://doi.org/10.1172/JCI95795.

13/14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.30.404525; this version posted December 2, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license. Van der Zouwen et al. MLR, speed and direction 30 November 2020

291 [33] Josset N, Roussel M, Lemieux M, Lafrance-Zoubga D, Rastqar A, Bretzner F.

292 Distinct Contributions of Mesencephalic Locomotor Region Nuclei to Locomotor

293 Control in the Freely Behaving Mouse. Curr Biol 2018;28:884-901.e3.

294 https://doi.org/10.1016/j.cub.2018.02.007.

295 [34] Lemieux M, Bretzner F. Glutamatergic neurons of the gigantocellular reticular

296 nucleus shape locomotor pattern and rhythm in the freely behaving mouse.

297 PLoS Biol 2019;17:e2003880. https://doi.org/10.1371/journal.pbio.2003880.

298 [35] He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition.

299 Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit. 2016:770–778

300 (2016). https://doi.org/10.1109/CVPR.2016.90

301 [36] Insafutdinov E, Pishchulin L, Andres B, Andriluka M, Schiele B. DeeperCut: A

302 Deeper, Stronger, and Faster Multi-person Pose Estimation Model. in European

303 Conference on Computer Vision (eds. Leibe, B., Matas, J., Sebe, N. & Welling,

304 M.) vol. 9910 34–50 (Springer International Publishing, 2016).

14/14