Animals and dissection

All experiments were carried out in isolated abdominal nervous systems of the signal crayfish Pacifastacus leniusculus, Dana 1852, of both sexes. Animals were acquired from a local fisherman in the river Wupper, Germany, and kept in appropriate water tanks at 14°C. Dissection of the nervous system was done according to Seichter and colleagues [30]. Briefly, animals were anesthetized in ice, exsanguinated, decapitated, and the nervous system carefully dissected from thoracic ganglion 4 (T4) to the terminal abdominal ganglion (A6). The first segmental nerves (N1) of the abdominal ganglia A2 – A5 were preserved as long as possible for recording, the other nerves were cut short. The ganglia chain was pinned out dorsal side up in a petri dish lined with Sylgard (Dow Corning, Midland, Michigan, USA) and filled with normal saline (concentrations in mM: 195.0 NaCl, 5.4 KCl, 2.6 MgCl₂, 15.8 CaCl₂, buffered with 1 M Trizma base and 0.5 M maleic acid at pH 7.4 in aerated saline). All ganglia were dorsally desheathed using fine scissors to facilitate uptake of chemicals and to ensure better penetration of antibodies.

Electrophysiology

Activity of power-stroke (PS) motor neurons was recorded with stainless steel pin electrodes from the posterior branch of N1 on both sides of ganglia A2 to A5. Nerve and recording electrode were insulated from the bath with petroleum jelly. The reference electrode was placed in the bath close to the recording electrode. If preparations did not show spontaneous activity, it was induced with 2 μM – 4 μM carbachol (Sigma, St. Louis, Missouri, USA). Sharp microelectrodes pulled from borosilicate capillaries (o.d. 1.0 mm, i.d. 0.5 mm) filled with 1% dextran Texas Red (dTR; Invitrogen, Carlsbad, CA, USA) in 1 M KAc + 0.1 M KCl (electrode resistance 30 MΩ – 45 MΩ) were used to record intracellularly from the Commissural Interneuron 1 (ComInt 1) at the ganglion midline, or the Coordinating Neurons in the LN. After impaling a neuron, it was physiologically identified by its spiking activity and effect on the motor output. After the experiment, its identity was confirmed by morphology with iontophoretic dye injection for approximately 15 min using 250 ms +1 nA pulses at 2 Hz.

Extracellular electrodes were connected to a differential amplifier (Model MA 102, Animal Physiology Electronics Lab, University of Cologne, Cologne, Germany). Intracellular signals were amplified with an intracellular amplifier SEC-05X (npi Electronic Instruments, Tamm, Germany). All signals were digitized with a Digidata 1400 (Molecular Devices; Sunnyvale, CA, USA) and recorded with Clampex (Molecular Devices).

During electrophysiological experiments, preparations were constantly perfused with saline. For blocking experiments, we made a large petroleum jelly well around ganglion A4 and perfused it with one of the following solutions: low Ca²⁺ / high Mg²⁺ saline (concentrations in mM: 118.0 NaCl, 5.33 KCl, 52.0 MgCl₂, 2.4 CaCl₂, buffered with 10 mM Trizma base and 4.7 mM maleic acid at pH 7.4 in aerated saline), 6 μM – 120 μM 6,7-dinitroquinoxaline-2,3-dione (DNQX, glutamate antagonist, BioTrend, Cologne, Germany), 100 μM – 5 mM picrotoxin (PTX, GABA antagonist; Sigma), or 0.5 mM - 1 mM serotonin (5-HT, Sigma). For wash out, we perfused the treated segment with normal saline for at least one hour. Focal application of drugs was mediated by a PDES 2L pressure ejection system (npi Electronic Instruments) connected to a capillary positioned close to ComInt 1’s recording site at the ganglion midline. 0.8 mM DNQX was applied for 4 s each 5 s, 5 mM PTX was applied for 10 s each 20 s with 0.5 bar. When we had to induce rhythmic activity with carbachol, we mixed the drugs into saline with the same carbachol concentration to exclude effects which could have been caused by a lack of cholinergic agonists.

Analysis of electrophysiological data

Coordination between segments was analyzed as follows: The beginning of each PS burst in A5 served as reference to calculate phase lags. Phase lag was calculated as the latency from beginning of the reference burst to beginning of another burst, e.g. PS4, divided by cycle period. Means and SD of at least 10 consecutive bursts were used in this analysis. To reveal rhythmic modulation of apparently tonic PS activity, we rectified and smoothed the voltage recordings with Spike2 (CED, Cambridge, UK), using a time constant of 0.1 s for smoothing.

The size and shape indices of the EPSPs in ComInt 1 were measured with Clampfit (Molecular Devices). First, at least 30 EPSPs were triggered on a specific ASC_E or DSC recording and averaged. From this, we used the build-in Clampfit tool “Statistics” to calculate the EPSP shape indices.

For the blocking experiments we averaged and compared amplitudes and spike indices of at least 10 EPSPs under control condition, and at the beginning and end of DNQX and PTX application using custom-written MATLAB (MathWorks, Natick, MA, USA) scripts. For simplification, we focus on the large and medium EPSPs in this analysis as they come from different types Coordinating Neurons (ASC_E and DSC).

“N” denotes number of experiments/animals; “n” the number of trials. Results from different treatments were compared in MATLAB with repeated-measures ANOVA (RM-ANOVA) with Tukey’s post-test, or paired t-test if only two dependent conditions were tested. Shape indices of EPSPs from ASC_E and DSC were compared with unpaired t-test. If the assumption of sphericity was violated for RM-ANOVA, Greenhouse-Geisser (GG) correction was used. Data and results of the statistic tests can be found in the supplementary spreadsheets. Because we could not detect any obvious differences in results obtained from experiments with bath or focal application we pooled those results for statistical analyses.

Neuron labeling by dye injection and immunohistochemistry

We used whole-mount preparations of abdominal ganglia A2 – A5 to examine if Coordinating Neurons contained 5-HT. Before immunohistochemistry, Coordinating Neurons were iontophoretically filled with 1% dTR (dextran Texas Red) for at least 30 min using 250 ms +1 nA pulses at 2 Hz. For dye transportation back into the soma and along its axon, preparations were kept at 4°C for up to three days.

After that, ganglia were fixed for 2 h at room temperature in Roti Histofix (Carl Roth, Karlsruhe, Germany) or 4% paraformaldehyde (Merck, Darmstadt, Germany) + 0.5% glacial acetic acid in 0.1 M phosphate buffered saline (PBS; concentrations in M: 0.1 Na₂HPO₄, 0.01 NaH₂PO₄, 0.15 NaCl). After fixation, samples were washed 3x10 min in PBS. For antibody labeling, ganglia were first preincubated in PBS containing 1% Triton-X-100 (PBST) and 5% normal goat serum + 0.1% NaAc (NGS) and then incubated for 12 h – 36 h at 4°C using rabbit anti-5-HT (Sigma) diluted 1:400 in PBST-NGS. Subsequently, ganglia were washed 3x2 h in PBST and incubated overnight at 4°C in goat anti-rabbit Alexa Fluor 488 conjugated (abcam, Cambridge, UK), diluted 1:200 in PBST-NGS. Afterwards, ganglia were washed 4x1 h in PBS, dehydrated in ascending ethanol concentrations, and mounted on microscope slides in methylsalicylate (Carl Roth).

Samples were visualized using a confocal laser scanning microscope (LSM 510 Meta, Zeiss, Oberkochen, Germany) with 10x (Plan-APOCHROMAT 10x/0.45) or 40x (Plan-NEOFLUAR, 40x/1.30 oil) magnification. The microscope was equipped with a primary dichroic beamsplitter HTF 488/543 for excitation. Emission of Alexa Fluor 488 was filtered by a 505–530 nm band pass filter, emission of dTR by a 650 nm long pass filter. Samples were scanned as z-stacks (10μm step size) of which maximum intensity or depth-coded projections were generated with ZEN 2011 black edition (Zeiss). Brightness and contrast were adjusted with Photoshop (CS 5, Adobe System, San Jose, CA, USA). Videos were prepared with ImageJ/FIJI [31, 32].

Sample preparation for single cell MALDI-TOF mass spectrometry

For cell identification, Coordinating Neurons were labeled intracellularly with 1% dTR for up to two hours and processed immediately if the soma was visible, or after overnight incubation at 4°C to allow additional dye diffusion. Motor neurons were backfilled via N1 with 1% dTR overnight. For single cell dissection, abdominal ganglia were first washed in single cell saline (concentrations in mM: 46 NaCl, 4.0 KCl, 7.5 CaCl₂ at pH 7.4) adapted from Jullien and Ripplinger [33]. Then the ganglion was fixed with microneedles in a preparation dish and covered with fresh single cell saline containing 33% glycerol to stabilize the fluorescent dye. Under a stereo fluorescence microscope (SteREO Lumar V12, Zeiss) equipped with a Lumar 43 HE Cy3 filter, the soma was isolated and dissected from the ganglion using ultra-fine scissors, and transferred manually with a glass capillary to a stainless steel sample plate for matrix assisted laser desorption/ionization (MALDI)—time of flight (TOF) mass spectrometry. The saline around the soma was completely removed with the same glass capillary. Without any additional rinsing steps, the soma was covered with 10–20 nl saturated alpha-cyano-4-hydroxycinnamic acid (CHCA) as matrix diluted in 50% methanol/water in a ratio of 1:2.

MALDI-TOF mass spectrometry

Mass spectra were acquired under manual control in positive ion mode on an UltrafleXtreme TOF/TOF mass spectrometer (Bruker Daltonics, Billerica, MA, USA). The instrument settings were optimized for mass ranges of m/z 0 – 300 Da and calibrated using a mixture of synthetic GABA and acetylcholine, respectively. Laser fluency was adjusted to provide optimal signal-to-noise ratio. MS/MS was performed with LIFT technology by an acceleration set at 1 kV. The number of laser shots used to obtain a spectrum varied from 1000 to 5000 depending on ion signal quality. The putative acetylcholine ion signal was verified using MS/MS fragmentation pattern and compared using fragmentation data provided by Scripps METLIN Center of Metabolomics [34]. All data obtained in these experiments were processed with the FlexAnalysis 3.4 software package (Bruker Daltonics).

To compare MALDI-TOF data of single ASC_E, DSC, and motor neuron preparations, we used ClinPro Tools software 3.0 (Bruker Daltonics) for comparison of signal intensities in a mass range of m/z 100–300. For that, MALDI MS data were uploaded and the following settings for PCA analysis were used: resolution 800, top hat baseline at 10.0% minimal baseline width, enable smoothing at 0.1 width (m/z) with 5 cycles using Savitzky-Golay filter type, recalibration of 1000 ppm maximal peak shift, 30% match to calibrant peaks, and a peak picking on total average spectrum at 4.00 signal to noise threshold.

Neurotransmitters affecting the motor output

The activity of Coordinating Neurons modulates the motor output of their target ganglia by increasing or decreasing the spiking frequency and number of recruited motor neurons, and / or altering the phase relationship between segments [7, 11, 12]. Coordinating information is integrated by one neuron in each hemisegment: Each spike from the Coordinating Neurons elicits a distinct EPSP in their target ComInt 1, which, in turn, is electrically coupled to one of the segment’s CPG neurons [14, 15]. This information transfer has to be reliable in order to maintain the phase relationship between segments. One step towards a deeper understanding of this connectivity is the identification of the neurotransmitter involved in transferring this information. Because of the vast abundance of neurotransmitters, we needed a preselection in the swimmeret system for our screening purposes. Hence, we first identified neurotransmitters and neuromodulators that changed the swimmeret system’s motor activity, specifically the phase relationship between segments, because they could possibly act by influencing the information transmission between the Coordinating Neurons and ComInt 1.

ACh is present in many sensory neurons, and cholinergic agonists are known to activate and modulate the swimmeret rhythm and have been used for decades to stimulate quiescent preparations [20, 37]. Furthermore, application of cholinergic agonists to only one half of a split-bath preparation changes the phase relationships between the segments at the boundary [23]. Therefore, ACh was one of our neurotransmitter candidates. To find further possible neurotransmitters that influenced the coordination between segments, we used an electrophysiological approach. For this, we focused on neuroactive substances that have been identified in other neurons in the crustacean ventral nerve cord. Glutamatergic and GABAergic ascending interneurons have been described in the terminal ganglion A6 [38–40]. Additionally, glutamate and GABA are synthesized by excitatory and inhibitory swimmeret motor neurons, respectively [16–18]. Furthermore, those motor neurons share some characteristics with the Coordinating Neurons, e.g. soma location, area of dendritic arborization, phase of activity, and innervation by the CPG (cf. Fig 1 and [41]). With respect to morphological similarities, Coordinating Neurons are also present in the locations of the four serotonergic fiber bundles identified in lobster [19], allowing the possibility that they contain 5-HT and are part of those fiber bundles. Hence, glutamate, GABA, and 5-HT were considered neurotransmitter candidates in addition to ACh.

First, we tested whether any of those candidates affected the coordination and the motor output of the swimmeret system. For this, we recorded the motor activity of the isolated abdominal ganglia chain and perfused only ganglion A4 with the glutamate antagonist DNQX (6 μM– 60 μM), the GABA antagonist PTX (100 μM), or 5-HT (0.5 μM– 1 μM) (Fig 2). Both antagonists have been shown to be effective in crustaceans, also specifically in crayfish [16, 42–44].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Glutamate antagonists, GABA antagonists, and 5-HT affected intersegmental coordination and rhythmic activity.

Raw PS recordings are shown in black, rectified and smoothed recordings are superimposed as red lines. The pharmacological agents were applied to one segment only, shown here with a grey box. A: Phase lag of power-stroke (PS) activity in ganglion A4 without and with DNQX application (grey box). Ai: PS activity in three ganglia of the same preparation. In control condition (left) all motor bursts were properly coordinated. When the glutamate antagonist DNQX (6μM) was perfused over ganglion A4 (grey box) its PS was delayed with respect to PS5 (right). Blue triangles indicate 25% phase lag with respect to PS5. Aii: Mean phase lag of PS activity in ganglion A4 and A3 before, during, and after application of the glutamate antagonist DNQX to A4 (grey background) in eight different preparations. Individual experiments are connected with lines. During DNQX perfusion, the phase lag of A4 increased in 6 of 8 experiments (F(2,12) = 8.095, p < 0.01). After wash-out, the phase lag recovered in 5 of 7 experiments. No rhythmic activity could be recorded in 1 of 8 experiments after washing. The phase lag between the untreated ganglia A5-A3 did not change significantly. Filled circles highlight data from the experiment in Ai. See also S1 Table. B: PS activity recorded from three different ganglia of the same preparation without and with PTX application. In control condition (left), the bursting activity in each segment is clearly distinguishable. The coordinated activity starts in the posterior segment and the other follow with approximately 25% phase lag. When the GABA antagonist PTX (400μM) was perfused over ganglion A4 (grey box) its PS became tonic (right). This tonic activity was still modulated on a cycle-by-cycle basis (red line). Coordination between untreated ganglia was unaffected. C: PS activity recorded from three different ganglia of the same preparation without and with 5-HT application. In control condition (left), the bursting activity in each segment is clearly distinguishable. The coordinated activity starts in the posterior segment and the others follow with approximately 25% phase lag. When 5-HT (500 μM) was perfused over ganglion A4 (grey box) bursting activity ceased in all ganglia (right). All data and statistics for these experiments are in S1 Table.

https://doi.org/10.1371/journal.pone.0197781.g002

The phase lag between neighboring ganglia was approximately 25% in control condition. In comparison, when DNQX blocked glutamatergic transmission in A4, the phase lag was significantly greater in that ganglion with respect to A5 (Fig 2A; S1 Table). PS activity of the perfused ganglion was delayed: the phase lag increased in 6 of 8 experiments (Fig 2Aii). DNQX applied specifically on A4 did not affect the phase lag between the untreated ganglia A5-A3 (Fig 2Aii). Some experiments seemed to reflect in A3 a change in phase lag seen in A4, which can be explained by the closed loop of coordinating information transfer in the swimmeret system. We hypothesized that DNQX might block synaptic transmission from Coordinating Neurons to ComInt 1, disturbing the coordination between segments.

In contrast to DNQX, perfusion of PTX over A4 (N = 6, Fig 2B) resulted in tonic motor activity with recruitment of larger units in that segment. Similar results were reported by Sherff and Mulloney [25] for application of PTX to the whole chain of abdominal ganglia. Rectifying and smoothing the raw recording (red lines in Fig 2) revealed that the tonic activity was still modulated on a cycle-by-cycle basis. We therefore assume that PTX application disinhibited motor neurons, while the CPG continued to oscillate. Because of the increased tonic activity, we were unable to calculate phase relationships of the PTX-treated ganglion.

When applying 5-HT to A4 (N = 6, Fig 2C), PS activity became either tonic on all recorded motor nerves (4 of 6 experiments), or only in the treated ganglion A4 (2 of 6 experiments). After washing with normal saline, rhythmic activity recovered, regardless of the applied substance. Because of their ability to corrupt the coordinated motor pattern, glutamate, GABA, and 5-HT qualified as neurotransmitter candidates in addition to ACh and were further investigated.

The synapse from the Coordinating Neurons onto ComInt 1 is mainly chemical

After the preselection of neurotransmitter candidates, we focused our investigation on the synapse between the Coordinating Neurons and ComInt 1. First, we had to examine whether the synapse is chemical or electrical. Each spike of the Coordinating Neurons elicits a distinct and fast EPSP in their target ganglia’s ComInt 1. Largest EPSPs are elicited by the Coordinating Neurons originating in immediate neighboring ganglia, and EPSPs elicited by the posterior ASC_E are always larger than those of the anterior DSC (Fig 3A and 3B) [14]. Therefore, the largest EPSPs in A4’s ComInt 1 are elicited by ASC_E spikes originating in A5 (Fig 3A), medium EPSPs by DSC spikes in the neighboring A3. Smallest EPSPs are always elicited from the most distal Coordinating Neuron, in this case DSC in A2

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The synapse between Coordinating Neurons and ComInt 1 is mainly chemical.

A: Schematic of activity in a chain of three connected ganglia and the input ComInt 1 receives in the middle ganglion. PS motor neuron activity is shown in black, ASC_E in cyan, and DSC in violet. ComInt 1 is depicted as intracellular recording where color-coded EPSPs elicited by the posterior ASC_E (from A5) are the largest, and those elicited by the anterior DSC (from A3) are medium-sized. Black EPSPs come in this example from the most anterior DSC in A2. The rhythmic activity starts with a burst of action potentials of PS motor neurons in the most posterior ganglion (A5), the anterior ones follow with a phase lag of approximately 25%. Because of this phase lag, EPSPs elicited by spikes from the posterior neighboring ASC_E (A5, in phase with PS) and anterior DSC (A3, out of phase with PS) arrive almost simultaneously in ComInt 1 in A4 (see also [14]). B: Overdraw of one ComInt 1 recording in A3 when triggered on ASC_E (Bi) or DSC spikes (Bii) from neighboring ganglia in normal saline (control, wash) and low Ca²⁺ / high Mg²⁺ saline. The time of the coordinating spike used for triggering the EPSP overdraw is indicated by the colored bar. Values in mV give average EPSP amplitude. C: Change in the normalized amplitude of both large (left panel, cyan) and medium sized (right panel, violet) EPSPs recorded in ComInt 1 in normal saline (ctrl) and lowCa²⁺ / high Mg²⁺ saline (N = 13). Amplitudes were significantly reduced in lowCa²⁺ / high Mg²⁺ saline (paired t-test). A: Abdominal ganglion; ASC_E: Ascending Coordinating Neuron; DSC: Descending Coordinating Neuron; ComInt 1: Commissural Interneuron 1; PS: Power-stroke. All raw data and statistics for these experiments are in S2 Table.

https://doi.org/10.1371/journal.pone.0197781.g003

Because EPSP amplitude strongly depends on the recording site, which could differ between experiments, we refer to the EPSPs elicited by different Coordinating Neurons as large, medium, and small, rather than giving the absolute amplitude.

We replaced the saline in the bath around one ganglion with low Ca²⁺ / high Mg²⁺ saline to block neurotransmitter release at chemical synapses (N = 13), while the activity of the neighboring segments was preserved. Therefore, ASC_E and DSC from the neighboring ganglia were still active and sent spikes to the ganglion bathed in low Ca²⁺ / high Mg²⁺ saline [8]. In all cases, EPSP size in ComInt 1 was significantly reduced (Fig 3B and 3C, S2 Table). In four of these experiments, we subsequently replaced the low Ca²⁺ / high Mg²⁺ saline with normal saline and demonstrated a recovery in EPSP size (Fig 3B). The average reduction of EPSP amplitude by 50% in low Ca²⁺ / high Mg²⁺ saline and subsequent recovery shows that the synapse between Coordinating Neurons and ComInt 1 has a chemical component. Electrical connections between neurons are unaffected by low Ca²⁺ / high Mg²⁺ saline and might account for the residual, diminished EPSPs. The actual proportion of the chemical component could be underestimated in our experiments because the calcium concentration we used in the low Ca²⁺ / high Mg²⁺ saline was shown to be not as effective as other concentrations [45].

As a consequence, the EPSPs in ComInt 1 have to be elicited by neurotransmitter release. ASC_E and DSC are two distinct Coordinating Neurons, therefore it is feasible that they might use different neurotransmitters at the synapse to ComInt 1. Interestingly, EPSP shape indices, i.e. rise-time and half-width, did not differ between those elicited by ASC_E or by DSC, regardless of ComInt 1 location (i.e. ComInt 1 located in A3 or A4; see Table 1, S3 Table). Therefore, we hypothesized that ASC_E and DSC use the same neurotransmitter.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. EPSP shape indices (in milliseconds) from Coordinating Neurons recorded in ComInt 1 in ganglia A3 and A4.

https://doi.org/10.1371/journal.pone.0197781.t001

Coordinating Neurons do not release glutamate or GABA at their synapse to ComInt 1

After confirming that the synapse between the Coordinating Neurons and ComInt 1 is chemical, we investigated whether Coordinating Neurons released the same neurotransmitters as the motor neurons (glutamate and GABA). This seemed a valid hypothesis because DNQX application delayed the motor bursts of the affected ganglion and PTX application corrupted the coordinated motor activity. We recorded intracellularly from ComInt 1 and applied a glutamate antagonist or GABA antagonist either to the bath or focally on the neuron. We expected that, if these antagonists would be effective, the EPSP shape and size recorded in ComInt 1 should change.

Neither bath nor focal application of the glutamatergic antagonist DNQX affected EPSP amplitude, rise-time, or half width in ComInt 1 (Fig 4A; S4 Table). ComInt 1’s membrane potential did not change during the application and it continued to oscillate phase-locked to its home ganglion’s PS bursts. As the membrane potential did not change, this factor could be excluded as influence on EPSP size here. Amplitude and shape of all EPSPs were similar to control in normal saline (N = 6). Similarly, application of PTX had no effect on EPSP size or shape (Fig 4B; N = 2; S4 Table). We could therefore exclude glutamate and GABA as neurotransmitters at the synapse between the Coordinating Neurons and ComInt 1, despite the puzzling effect their antagonists had on the coordination between segments.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Glutamate antagonists and GABA antagonists did not change EPSP amplitude or shape in ComInt 1.

Largest and medium-sized EPSPs are shown in cyan and violet, respectively. See also S4 Table. Ai: Intracellular recording of ComInt 1 in normal saline and during bath application of 0.8 mM DNQX (glutamate antagonist). Aii: Average waveform of largest and medium-sized EPSPs from one experiment (Ai) in normal saline and during DNQX application (grey background) at the beginning (t₁ = 14 min) and later (t₂ = 30 min) in the recording. Shape and amplitude of both EPSP types did not differ between control and DNQX application. Aiii: Data for individual experiments with DNQX application (grey box). Individual experiments are connected by lines. Data points for the experiment shown in Ai and Aii are highlighted by black fills with colored lines. Neither the absolute amplitude of EPSPs [F(2,8) = 0.017, p = 0.982 for large EPSPs; F(1.12,5.62) = 5.803, p = 0.053 for medium EPSPs], nor the normalized amplitudes were significantly different. No significant difference was found for the rise-time [F(1.04,4.18) = 1.652, p = 0.268 for large EPSPs; F(1.11,5.57) = 1.947, p = 0.219 for medium EPSPs] or half-width [F(2,8) = 0.167, p = 0.849 for large EPSPs; F(2,10) = 0.156, p = 0.857 for medium EPSPs]. Bi: Intracellular recording of ComInt 1 in normal saline and during focal application of 5 mM PTX (GABA antagonist). Bii: Average waveform of largest and medium-sized EPSPs from one experiment in normal saline and during focal PTX application (grey background) at the beginning (t₁ = 0 s) and later (t₂ = 160 s) in the recording. Shape and amplitude of both EPSP types did not differ between control and PTX application. Biii: Data for individual experiments with PTX application (grey box). Individual experiments are connected by lines. Data points for the experiment shown in Bi and Bii are highlighted by black fills with colored lines. Paired t-test p ≥ 0.5 for amplitude, rise-time, and half-width of large and medium EPSPs shows no statistical difference in the shape indices. All data and statistics for these experiments are in S4 Table.

https://doi.org/10.1371/journal.pone.0197781.g004

Coordinating Neurons are not serotonergic

In lobster, Beltz and Kravitz [19] described two medial (MFB) and lateral fiber bundles (LFB) with 5-HT-immunoreactive labeling that projected into the LN of each abdominal ganglion, the site of the swimmeret CPG connections in crayfish. Beltz and Kravitz did not identify the origins of all 5-HT-immunoreactive fiber bundles. Because of the branching in the LN and two fiber bundles running close to the midline we hypothesized that the Coordinating Neurons might be serotonergic.

Electrophysiological identification of neurotransmitters usually requires stable intracellular recordings over long periods of time to allow thorough application and washing of suitable, specific agonists and antagonists. In some cases, deteriorating recording quality can lead to misinterpretation of results, especially when focusing on low-amplitude, passively propagated EPSPs. Hence, our second approach to identify putative neurotransmitters produced by specific neurons was double labeling with intracellular dye application in the neuron of interest and immunohistochemistry using antibodies against a suspected neurotransmitter, in this case 5-HT.

Like in lobster, two MFBs and two LFBs were present in all investigated ganglia (N = 17, Fig 5, S1 Video and S2 Video). LFBs branched in the area of the LN. At the posterior border of the ganglion’s core region, fibers crossed between the MFBs and LFBs. Additionally, MFBs crossed to the contralateral side at the ganglion’s mid region. Four small cell bodies were present in the lateral ventral region: two anterior and two posterior (Fig 5, arrowheads in Aiii and Diii). Depending on the preparation, up to three large cell bodies could be located ventrally at the midline (Fig 5, arrows in Diii; S1Ai, S1Bi and S1Ci Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. The Coordinating Neurons were not 5-HT-immunoreactive.

A: Dorsal view of a whole-mount preparation with intracellular staining of ASC_E (Ai) and 5-HT antibody labeling (Aii) in A4. Merge shows no co-labeling for dye-filled ASC_E (magenta) and 5-HT-immunoreactive neurons and fiber bundles (green) in ASC_E’s home ganglion (A4, Aiii) and one of ASC_E’s target ganglia (A3, B and C; the area shown is indicated in the schematic). D: Dorsal view of a whole-mount preparation with intracellular staining of DSC (Di) and 5-HT antibody labeling (Dii) in A3. Merge shows no co-labeling for dye-filled DSC (magenta) and 5-HT-immunoreactive neurons and fiber bundles (green) in DSC’s home ganglion (Diii). E: Soma (Ei) and axon (Eii) of the Coordinating Neuron were located in different planes compared to 5-HT-immunoreactive neurons and fiber bundles. Any apparent double-labeling in the figures is due to the 2D z-stack projection (10μm step size). Scale bar (100μm) applies to all figures. Arrowheads indicate small lateral 5-HT-immunoreactive somata; arrows indicate large ventral 5-HT-immunoreactive somata. For better view please also see supplementary movies of the confocal scans of these whole mounts (S1 Video, S2 Video) A: Abdominal ganglion; ASC_E: Ascending Coordinating Neuron; DSC: Descending Coordinating Neuron.

https://doi.org/10.1371/journal.pone.0197781.g005

ASC_E’s and DSC’s somata were located very close but dorsal to the posterior or anterior small 5-HT-immunoreactive cell bodies, respectively (Fig 5A and 5D, S1 Fig, S1 Video, S2 Video). Both ASC_E’s and DSC’s dendrites branched in the 5-HT-immunoreactive neuropil region and the axons projected dorsally in close proximity to the 5-HT-immunoreactive MFBs along the midline. Nevertheless, we could not observe any co-labeling of 5-HT-immunoreactive and the Coordinating Neurons in their home ganglion (Fig 5A, 5D and 5E) or target ganglia (Fig 5B and 5C) (ASC_E: N = 7; DSC: N = 9). If Coordinating Neurons released 5-HT at their synapse to ComInt 1 we would have expected co-labeling along the coordinating axons at the midline of their target ganglia. But the Coordinating Neurons’ somata and axons were located in different planes than the 5-HT-immunoreactive fibers (Fig 5E, S1 Fig, S1 Video, S2 Video). Therefore, we could also exclude 5-HT as a neurotransmitter of the Coordinating Neurons. Any apparent double-labeling in the figures is due to the 2D z-stack projection (see also S1 Video and S2 Video).

Both Coordinating Neurons contain ACh

With the elimination of glutamate, GABA, and 5-HT, we could narrow down the list of putative neurotransmitters. Antibodies are very specific to their epitopes and usually designed for vertebrates. Hence, some structures in invertebrates cannot be labeled easily with standard antibodies. To continue our neurotransmitter screening and ascertain whether ACh could be a neurotransmitter of the Coordinating Neurons, we used the highly sensitive and robust method of single cell profiling by matrix-assisted laser desorption/ionization (MALDI)–time of flight (TOF) mass spectrometry.

While a MALDI-TOF MS-based method for the detection of neuropeptides on the single cell level has been established (e.g. [46]), an equivalent approach for the detection of smaller neurotransmitters such as ACh from individual dissected single cells is lacking so far. In our study, we modified a strategy published from Persike and colleagues [47] to fill this gap. The advantage of detecting ACh with MALDI-TOF is that, compared to other neurotransmitters, no further derivatizing steps are necessary because the molecule is positively charged.

We dissected individual ASC_E (N = 5) and DSC (N = 3) somata using a glass capillary after filling the neuron with fluorescent dye (Fig 6C). For both cell types, we found an ion signal at m/z 146.12 in the resulting mass spectra, which corresponds to the theoretical mass of ACh (Fig 6A). To verify the structure of the putative ACh, the same sample spots were used for tandem mass spectrometry (MS/MS, Fig 6B) experiments. Resulting mass spectra generated an ion pattern of two distinct ion signals at m/z 60.07 and m/z 87.00, which corresponds to ACh data provided by the metabolite database METLIN. For experimental structure validation we confirmed our findings with MS/MS experiments using synthetic ACh (Fig 6B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. MALDI-TOF MS revealed ACh in ASC_E and DSC somata.

A: Representative mass spectra from a single ASC_E (upper panel, cyan) and DSC soma preparation (middle panel, violet). In both, an ion signal at m/z 146.12 was detectable, which is clearly distinct from the ion signal at m/z 146.05 assigned to the matrix of cyano-4-hydroxycinnamic acid (CHCA, grey) (lower panel and inset in ASC_E panel). B: Tandem mass spectrometry (MS/MS) experiments confirmed the ion at m/z 146.12 (arrows in A) as ACh from ASC_E’s (upper panel, cyan) and DSC’s soma preparations (middle panel, violet). As control MS/MS experiments were performed for synthetic ACh (lower panel, orange). C: Image series of a single ASC_E soma (arrow) dissection. ASC_E: Ascending Coordinating Neuron; DSC: Descending Coordinating Neuron.

https://doi.org/10.1371/journal.pone.0197781.g006

To screen for contaminations during the dissection procedure of ASC_Es and DSCs, we dissected individual somata from motor neurons, which are glutamatergic or GABAergic and do not synthesize ACh, as control [17, 18, 48, 49]. In all preparations (N = 3; n = 9), resulting mass spectra did not show an ion signal at m/z 146.12 but at m/z 146.06 (Fig 7A). Fragmentation experiments demonstrated that the ion signal at m/z 146.06 corresponded to CHCA, which was used as matrix (Fig 7B). Therefore, we conclude that both ASC_E and DSC contain ACh.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. MALDI-TOF MS of individual non-cholinergic motor neuron was negative for ACh.

A: Mass spectrum of a single motor neuron soma (green). Inset illustrates ion signal at m/z 146.06. B: Fragments of this ion (arrows in A) corresponded to CHCA (grey) and not to fragments generated from synthetic ACh (orange). C: Principal Component Analysis of mass spectra from individual soma preparations of ASC_E (N = 5, cyan), DSC (N = 3, violet), and motor neurons (n = 9, all from different ganglia from three different animals, green). The first three principal components (PC) are shown. Motor neurons formed a distinct cluster which is separate from Coordinating Neurons for PC1 vs. PC2 (Ci) and PC1 vs. PC3 (Cii). ASC_E: Ascending Coordinating Neuron; DSC: Descending Coordinating Neuron; MN: Motor neuron.

https://doi.org/10.1371/journal.pone.0197781.g007

To determine the differences in the molecular composition of Coordinating Neurons and motor neurons, we performed a principal component analysis (Fig 7C). Mass spectra of individual motor neurons contained the most homogenous group of data points and formed a distinct cluster, followed by ASC_E data. The most variable data were obtained from the three individual DSC somata. The differences in the cluster spread of ASC_E and DSC could result from differences in the molecular composition of these two neuron types. However, both ASC_E and DSC were clearly separated from the motor neuron cluster.

With these results, we demonstrated that there was no cross contamination of the sample with ACh positive tissue that surrounded the extracted neurons, since it is known that the motor neurons use glutamate and GABA at their neuromuscular end plate [17,25,48]. Therefore, this strengthened our result that the Coordinating Neurons ASC_E and DSC indeed contain ACh, which they might release at the synapse to ComInt 1.