Functionally asymmetric motor neurons coordinate locomotion of Caenorhabditis elegans

Invertebrate nervous systems are valuable models for fundamental principles of the control of behavior. Ventral nerve cord (VNC) motor neurons in Caenorhabditis elegans represent one of the best studied locomotor circuits, with known connectivity and functional information about most of the involved neuron classes. However, for one of those, the AS motor neurons (AS MNs), no physiological data is available. Combining specific expression and selective illumination, we precisely targeted AS MNs by optogenetics and addressed their role in the locomotion circuit. After photostimulation, AS MNs induce currents in post-synaptic body wall muscles (BWMs), exhibiting an initial asymmetry of excitatory output. This may facilitate complex regulatory motifs for adjusting direction during navigation. By behavioral and photo-inhibition experiments, we show that AS MNs contribute to propagation of the antero-posterior body wave during locomotion. By Ca2+-imaging in AS MNs and in their synaptic partners, we also reveal that AS MNs play a role in mediating forward and backward locomotion by integrating activity of premotor interneurons (PINs), as well as in coordination of the dorso-ventral body wave. AS MNs do not exhibit pacemaker properties, but potentially gate VNC central pattern generators (CPGs), as indicated by ceasing of locomotion when AS MNs are hyperpolarized. AS MNs provide positive feedback to the PIN AVA via gap junctions, a feature found also in other locomotion circuits. In sum, AS MNs have essential roles in coordinating locomotion, combining several functions, and emphasizing the compressed nature of the C. elegans nervous system in comparison to higher animals. Highlights A class of motor neurons with unidentified function – AS cholinergic motor neurons - was characterized in C. elegans. AS neurons show asymmetry in both input and output and are specialized in coordination of dorso-ventral undulation bends. AS neurons mediate antero-posterior propagation of the undulatory body wave during locomotion. AS neurons integrate signals for forward and reverse locomotion from premotor interneurons and may gate ventral nerve cord central pattern generators (CPGs) via gap junctions.

Supplementary Videos S10, S11). A similar increase was found in the AS8 neuron, and both AS3 and AS8 showed a synchronized increase of activity (Supplementary Fig. S4A, B). Thus, signaling from  Recent observations revealed that AVA coupling to A-type MNs via gap junctions is strongly 3 2 2 rectifying towards AVA (Liu et al., 2017). We thus wanted to explore the role of the electrical 3 2 3 synapses between the PINs and the AS MNs in more detail, e.g. whether AS MN photostimulation 3 2 4 could lead to depolarization of the PINs. We generated strains expressing the ratiometric Ca 2+ indicator cameleon (bearing CFP and YFP 3 2 6 moieties; Miyawaki et al., 1997) in the command interneurons (driven by sra-11 and nmr-1 promoters, 3 2 7 for expression in AVB and AVA, respectively) together with ChR2 expressed in the AS MNs ( Fig.   3 2 8 6BI). Both promoters express in several head neurons, yet we could identify AVB and AVA by their 3 2 9 position with respect to anatomical landmarks and with respect to other (known) fluorescent neurons. as the sole innexins. INX-3 is widely expressed in multiple tissues, and AVA and AVB also express 3 3 5 UNC-7 (Altun et al., 2009;Starich et al., 2009). Thus, we used an unc-7(e5) null mutant, in which no 3 3 6 electrical coupling should occur between AS MNs and AVA or AVB, and repeated the above MNs and AVA. For AVB, we did not observe any significant effect in the unc-7(e5) mutant. wave. 3) AS MN activity during locomotion is oscillatory, and is correlated more with forward than   physiological information, they were missing in many models representing the locomotor circuit 3 5 9 function in C. elegans (Von Stetina et al., 2005;Zhen and Samuel, 2015). We showed that AS MNs MNs counteract neurons providing a ventral bias, e.g. the VA and VB MNs, or even the VC neurons 3 6 5 (Faumont et al., 2011;White et al., 1986). This corresponds to recent computational studies, which integrating forward and backward locomotion motifs, e.g. by providing an electrical sink (or source) 3 7 2 for the PINs of the respective opposite direction (Fig. 7A, B). Similar functions were shown for A-type MNs and AVA (Kawano et al., 2011;Liu et al., 2017)   In the lamprey, lateral bends were shown to be caused by asymmetry in stimulation of the 3 8 4 mesencephalic locomotor region (Sirota et al., 2000), and in the freely moving lamprey even  Asymmetry in contralateral motifs of complex locomotor circuits is also known from vertebrate  Varshney et al., 2011;White et al., 1986), predominance is apparent in excitatory neuromuscular 3 9 3 junctions from A-and B-type MNs to ventral muscles, as well as in the corresponding contralateral 3 9 4 synapses to inhibitory DD MNs, which innervate dorsal muscle. Therefore, tonic activity of B-or A-3 9 5 type MNs would be expected to generate a bias towards ventral bending, and this could be balanced by 3 9 6 excitation of AS MNs. In addition, VC neurons may contribute in counteracting AS MN function (see 3 9 7 above). However, the compressed nature of the C. elegans nervous system, in which single neurons 3 9 8 fulfill multiple tasks that in higher animals are executed by layers of different cells, may not always 3 9 9 allow for the direct comparison to vertebrate systems.   Recently, ability of MNs to modulate activity of PINs was shown in several animal models: for B-type the mouse, such activity was suggested for MNs, changing the frequency of rhythmic CPG activity  CPGs are dedicated neural circuits with intrinsic rhythmic activities (Grillner, 2006;Guertin, 2013). In   The previously uncharacterized class of AS motor neurons is specialized in coordination of dorso-  Strains and Genetics 4 6 0 C. elegans strains were maintained under standard conditions on nematode growth medium (NGM) and fed by 4 6 1 E. coli strain OP50-1 (Brenner, 1974). Transgenic lines were generated using standard procedures (Fire and 4 6 2 Pelham, 1986) by injecting young adult hermaphrodites with the (plasmid-encoded) transgene of interest and a 4 6 3 marker plasmid that expresses a fluorescent protein. In some cases, empty vector was included to increase the 4 6 4 overall DNA concentration to 150-200 ng/µl. 4 6 5 The following strains were used or generated for this study: N2 (wild type isolate, Bristol strain), CB5: unc- respectively. The punc-4::ChR2 sense and antisense construct was generated as follows: The punc-4 promoter 4 9 1 from plasmid pCS139 was subcloned into pCS57 (Schultheis et al., 2011) by SphI and NheI, 1 kb antisense 4 9 2 sequence was amplified from the ligated construct pOT1 punc-4::ChR2::YFP using primers (5'-4 9 3 GGGGTTTAAACAGCTAGCGTCGATCCATGG-3' and 5'-4 9 4 1 6 CCCGCGGCCGCCCAGCGTCTCGACCTCAATC-3') and subcloned into the same construct by NotI and 4 9 5 PmeI to get pOT2. To silence ChR2 expression under the pacr-5 promoter we used a sense and antisense strands 4 9 6 approach (Esposito et al., 2007) as follows: The pacr-5 promoter was amplified from genomic DNA using 4 9 7 primers (5'-TTATGATGCGAAAGCTGAATCGAGAAAGAG-3', 5'-4 9 8 CCATGCTTACTGCACTTGCTTCCCATACTTC-3', nested 5'-4 9 9 GGGGCATGCATCGAGAAAGAGAAGCGGCG-3', 5'-CCCGCTAGCAAAGCATTGAAACTGGTGAC-3') 5 0 0 and subcloned into pCS57 with SphI and NheI to yield pOT3 (pacr-5::ChR2::YFP). The sense and antisense 5 0 1 strands were amplified from this construct using primers (for the coding region of ChR2: 5'-5 0 2 ATGGATTATGGAGGCGCCC-3', 5'-CCAGCGTCTCGACCTCAATC-3'; for the promoter sense: 5'-5 0 3 GGCGGAGAGTAGTGTGTAGTG-3' and 5'-5 0 4 GGGCGCCTCCATAATCCATCAAAGCATTGAAACTGGTGACGAG-3'; for the promoter antisense: 5'-5 0 5 GGCGGAGAGTAGTGTGTAGTG-3' and 5'-5 0 6 GATTGAGGTCGAGACGCTGGCAAAGCATTGAAACTGGTGACGAG-3'; for fusion of sense strand: 5'-5 0 7 GCGGTTTCACGCTCTGATGAT-3' and 5'-CTCAGTGCCACCAATGTTCAA-3'; and for fusion of the 5 0 8 antisense strand: 5'-GCGGTTTCACGCTCTGATGAT-3' and 5'-GCGCGAGCTGCTATTTGTAA-3').

Animal tracking and behavioral analysis 5 3 4
For worms moving freely on NGM, locomotion parameters were acquired with a previously described worm 5 3 5 tracker (Stirman et al., 2011) allowing to precisely target illumination of identified segments of the worm body 5 3 6 by a modified off-the-shelf liquid crystal display (LCD) projector, integrated with an inverted epifluorescence 5 3 7 microscope. Light power was measured with a powermeter (PM100, Thorlabs, Newton, NY, USA) at the 5 3 8 specimen focal plane. Animals used in all the optogenetics experiments were raised in the dark at 20˚C on NGM 5 3 9 plates with E. coli OP50-1 and all-trans-retinal. The OP50-retinal plates were prepared 1-2 days in advance by 5 4 0 seeding a 6 cm NGM-agar plate with 250 µl of OP50 culture and 0.25 µ l of 100 mM retinal dissolved in ethanol.

4 1
Young adults were transferred individually on plain NGM plates under red light (>600 nm) in a dark room and 5 4 2 kept for 5 minutes in the dark before transfer to the tracker.

4 7
For optogenetic ablation experiments (Fig. 3A, Supplementary Fig. S2), AS MNs were ablated in animals 5 4 8 expressing PH-miniSOG by 2.5 min exposure to 470 nm light of 1.8 mW/mm 2 intensity; segments 3-10 out of 5 4 9 11 were illuminated. Animals were analyzed after a 2 h resting period for 60 s without illumination. Wild type 5 5 0 worms were used as a control with the same illumination protocol. Ablation was verified by fluorescence 5 5 1 microscopy in strain ZX2110 expressing green fluorescent protein (GFP) in all cholinergic neurons, in addition 5 5 2 to PH-miniSOG in AS MNs.

3
For experiments of AS MN hyperpolarization using the histamine-gated Cl --channel HisCl1 (Fig. 3B, 5 5 4 Supplementary Fig. S2), worm locomotion was measured on NGM plates with 10 mM histamine 4 minutes after 5 5 5 transfer from plates without histamine, for 60 s without illumination. The same strain on NGM without 5 5 6 histamine served as a control.

7
For experiments in which MNs were hyperpolarized with natural Cl --conducting anion channel rhodopsin 5 5 8 (ACR1; Fig. 4), due to the high operational light sensitivity of the channel, the system was modified as 5 5 9 described (Steuer Costa et al., 2017). An additional band pass filter (650 ± 25 nm) was inserted in the 5 6 0 background light path and a mechanical shutter (Sutter Instrument Company, Novato, USA), synchronized to 5 6 1 the light stimulation, was placed between projector and microscope. Control animals were tested for the 5 6 2 background light stimulation and showed no response. The light stimulation protocol was 20 s without 5 6 3 illumination, 20 s in 70 µW/mm² 470 nm light and 20 s without illumination. The worms' body was divided into 5 6 4 11 segments, and segments 3-10, 3-4, 5-6 or 9-10 were illuminated, respectively. As the experiment in unc-5 6 5 47(e307) background was performed with a different transgene injected into unc-47(e307) mutants, we tested 5 6 6 1 8 the extrachromosomal array after outcrossing into wild type background, where it evoked contraction of the 5 6 7 animals, as expected ( Supplementary Fig. S3C).

6 8
Tracks were automatically filtered to exclude data points from erroneously evaluated movie frames with a 5 6 9 custom-made workflow in KNIME (KNIME Desktop version 3.5, KNIME.com AG, Zurich, Switzerland; 5 7 0 (Warr, 2012). Our constraints were that animals do not move faster than 1.25 mm/s and their length does not 5 7 1 show a discrepancy above 25 % to the mean first five seconds of the video. Videos were excluded from analysis 5 7 2 when more than 15 % of the data points had to be discarded by our constraints. Behavior data passed the 5 7 3 Shapiro-Wilk normality test.

7 4
For determination of the ratio of dorso-/ventral angles (Fig. 1BIII, IV), the second out of 11 three point angles, 5 7 5 measured from head to tail, was registered for animals for which the vulva position was previously indicated by 5 7 6 manually indicating this to the software. For each track, values of the second three-point angle were averaged 5 7 7 for dorsal and ventral bends individually, and the ratio was calculated.

7 8
To calculate the frame-to-frame difference of bending angles (Fig. 4E), data on each of eleven 3-point angles 5 7 9 were extracted, smoothed by running an average of 15 frames, and the ∆ of absolute values between two 5 8 0 subsequent frames were calculated and averaged for before and during illumination conditions for each angle.

8 1
The light -no light ∆ ∆ of bending angles (Fig. 4F) were calculated by subtracting the value of the no light from 5 8 2 the light condition. They were then averaged for the bending angles 1-5 (anterior) and 6-11 (posterior). Binarized videos of freely crawling animals were used to segment the animals' body, and analyzed as described Massachusetts). Briefly, grey scale worm images were binarized with a global image threshold using Otsu's 5 8 7 method (Otsu, 1979). Objects encompassing border pixels were ignored and only the largest object was assumed 5 8 8 to be the worm. The binary image was further processed (by thickening, removing spur pixels, flipping pixels 5 8 9 by majority and filling holes). Worm skeletonization was achieved by thinning to produce an ordered vector of 5 9 0 100 body points and corresponding tangent angles (theta) from head to tail. Images that could not be analyzed or 5 9 1 where the skeleton of the animal was unusually small were considered as missing data points. Head and tail 5 9 2 assignment was checked manually. The theta angles were smoothened by a simple moving average with a 5 9 3 window of 15 centered data points. The mean of these angles was then compared to the Eigenworms computed 5 9 4 from previously published data on N2 videos (Stephens et al., 2008). The Eigen projections obtained were taken 5 9 5 as a measure of worm posture and plotted. 5 9 6 Electrophysiology 5 9 7 Recordings from dissected dorsal BWM cells of strain ZX2221 (used to avoid unspecific excitation of AS MNs 5 9 8 via PINs were conducted as described previously (Liewald et al., 2008). Only the left side of the worm was cut 5 9 9 to preserve commissural connections from the ventral nerve cord where AS MN cell bodies reside. Light 6 0 0 activation was performed using an LED lamp (KSL-70, Rapp OptoElektronik, Hamburg, Germany; 470 nm, 8 6 0 1 1 9 mW/mm²) and controlled by an EPC10 amplifier and Patchmaster software (HEKA, Germany). 6 0 2 Ca 2+ imaging microscope setup 6 0 3 Fluorescence measurements were carried out on an inverted fluorescence microscope (Axiovert 200, Zeiss, 6 0 4 Germany) equipped with motorized stage MS 2000 (Applied Scientific Instrumentation, USA) and the 6 0 5 PhotoTrack quadrant photomultiplier tube (PMT; Applied Scientific Instrumentation, USA). Two high-power 6 0 6 light emitting diodes (LEDs; 470 and 590 nm wavelength, KSL 70, Rapp Optoelektronik, Germany) or a 100 W 6 0 7 HBO mercury lamp were used as light sources. A Photometrics DualView-Λ beam splitter was used to obtain 6 0 8 simultaneous dual-wavelength acquisition; these were coupled to a Hamamatsu Orca Flash 4.0 sCMOS camera 6 0 9 operated by HCImage Live (Hamamatsu) or MicroManager (http://micro-manager.org). Light illumination 6 1 0 protocols (temporal sequences) were programmed on, and generated by, a Lambda SC Smart shutter controller 6 1 1 unit (Sutter Instruments, USA), using its TTL output to drive the LED power supply or to open a shutter when 6 1 2 using the HBO lamp. 6 1 3

Measurement of Ca 2+ in muscles and AS MNs in immobilized worms 6 1 4
For measurements of GCaMP3 (Fig. 2) and RCaMP (Fig. 3B) in muscles and GCaMP6 in AS MNs (Figs. 5, 6), 6 1 5 the following light settings were used: GFP/mCherry Dualband ET Filterset (F56-019, AHF Analysentechnik, 6 1 6 Germany), was combined with 532/18 nm and 625/15 nm emission filters and a 565 longpass beamsplitter (F39-6 1 7 833, F39-624 and F48-567, respectively, all AHF). ChR2 stimulation was performed using 1.0-1.2 mW/mm 2 6 1 8 blue light, unless otherwise stated. To measure RCaMP or mCherry fluorescence, 590 nm, 0.6 mW/mm 2 yellow 6 1 9 light was used. The 2x binned images were acquired at 50 ms exposure time and 20 fps. Animals were 6 2 0 immobilized on 2 or 4% M9 agar pads with polystyrene beads (Polysciences, USA) and imaged by means of 6 2 1 25x or 40x oil objective lenses. 5 s of yellow light illumination and 15 s of blue light illumination protocols 6 2 2 were used. For RCaMP imaging 20 s yellow light illuminations were used. Measurements of control animals 6 2 3 (i.e. raised without ATR, or without histamine) were conducted the same way as for animals kept in the 6 2 4 presence of ATR, or exposed to histamine. 6 2 5 Image analysis was performed in ImageJ (NIH). For Ca 2+ -imaging in muscles, regions of interest (ROIs) were 6 2 6 selected for half of the BWM cells in the field of view, or around the neuron of interest for Ca 2+ -imaging in AS 6 2 7 MNs. Separate ROIs were selected for background fluorescence with the same size. Mean intensity values for 6 2 8 each video frame were obtained and background fluorescence values were subtracted from the fluorescence 6 2 9 values derived for GCaMP or RCaMP. Subtracted data was normalized to Δ F/F = (F i -F)/F, where F i represents 6 3 0 the intensity at the given time point and F represents the average fluorescence of the entire trace. 6 3 1

Measurement of Ca 2+ in muscles and AS motor neurons in moving animals 6 3 2
Measurements of GCaMP6 and mCherry were performed using the same filter and microscope settings as for 6 3 3 immobilized worms. Moving worms were assayed on 1% agar pads in M9 buffer. Tracking was based on the 6 3 4 PhotoTrack system (Applied Scientific Instrumentation, USA) that uses the signals from a 4-quadrant 6 3 5 photomultiplier tube (PMT) sensor for automated repositioning of a motorized XY stage to keep a moving 6 3 6 2 0 fluorescent marker signal in the field of view (Faumont et al., 2011). For this purpose, an oblique 80% 6 3 7 transmission filter was inserted in the light path to divert 20% of the light to the PMT quadrants. A 535/30 6 3 8 bandpass filter (F47-535, AHF) was used to narrow the emission spectrum prior to detection for improved 6 3 9 tracking performance. A fluorescent marker GFP was expressed in vulval muscle cells, strains PD4665 and 6 4 0 ZX2012.Video files containing data of both fluorescent channels (for GCaMP6 and mCherry) were processed 6 4 1 with custom written Wolfram Mathematica notebooks. Both color channels were virtually overlaid to accurately 6 4 2 correct the spatial alignment. Images were first binarized to identify the centroid of the moving neuronal cell 6 4 3 bodies throughout all frames. Mean intensity values of a circular ROI (18 pixel radius) centered on this centroid 6 4 4 were measured and subtracted with the mean intensity values of a surrounding donut shaped background ROI (5 6 4 5 pixel width). Coordinates of two AS neurons of interest were recorded relative to the vulva and to each other to 6 4 6 obtain their relative distance and the angle between the vulva and the two neurons of interest. The traces were 6 4 7 normalized to Δ F/F = (F i -F)/F, where F represents the average of the entire trace, and were used for correlation 6 4 8 analysis.

6 0
elegans is modified by a dominant mutation in the GLR-1 ionotropic glutamate receptor. Neuron 24, 347-361. designed experiments, performed data interpretation, and edited the manuscript.