Variations on a theme: species differences in synaptic connectivity do not predict central pattern generator activity

This research shows that the synaptic connectivity between homologous neurons exhibits species-specific variations on a basic theme. The neurons vary in the extent of electrical coupling and reciprocal inhibition. They also exhibit different patterns of activity during rhythmic motor behaviors that are not predicted by their circuitry. The circuitry does not map onto the phylogeny in a predictable fashion either. Thus neither neuronal homology nor species behavior is predictive of neural circuit connectivity.

Abstract

A fundamental question in comparative neuroethology is the extent to which synaptic wiring determines behavior vs. the extent to which it is constrained by phylogeny. We investigated this by examining the connectivity and activity of homologous neurons in different species. Melibe leonina and Dendronotus iris (Mollusca, Gastropoda, Nudibranchia) have homologous neurons and exhibit homologous swimming behaviors consisting of alternating left-right (LR) whole body flexions. Yet, a homologous interneuron (Si1) differs between the two species in its participation in the swim motor pattern (SMP) and synaptic connectivity. In this study we examined Si1 homologs in two additional nudibranchs: Flabellina iodinea, which evolved LR swimming independently of Melibe and Dendronotus, and Tritonia diomedea, which swims with dorsal-ventral (DV) body flexions. In Flabellina, the contralateral Si1s exhibit alternating rhythmic bursting activity during the SMP and are members of the swim central pattern generator (CPG), as in Melibe. The Si1 homologs in Tritonia do not burst rhythmically during the DV SMP but are inhibited and receive bilaterally synchronous synaptic input. In both Flabellina and Tritonia, the Si1 homologs exhibit reciprocal inhibition, as in Melibe. However, in Flabellina the inhibition is polysynaptic, whereas in Tritonia it is monosynaptic, as in Melibe. In all species, the contralateral Si1s are electrically coupled. These results suggest that Flabellina and Melibe convergently evolved a swim CPG that contains Si1; however, they differ in monosynaptic connections. Connectivity is more similar between Tritonia and Melibe, which exhibit different swimming behaviors. Thus connectivity between homologous neurons varies independently of both behavior and phylogeny.

NEW & NOTEWORTHY This research shows that the synaptic connectivity between homologous neurons exhibits species-specific variations on a basic theme. The neurons vary in the extent of electrical coupling and reciprocal inhibition. They also exhibit different patterns of activity during rhythmic motor behaviors that are not predicted by their circuitry. The circuitry does not map onto the phylogeny in a predictable fashion either. Thus neither neuronal homology nor species behavior is predictive of neural circuit connectivity.

behavior and neural mechanisms have been proposed to represent two independent levels of biological organization (Katz 2016; Sommer 2008; Striedter and Northcutt 1991). If so, it suggests that two behaviors can be homologous while having divergent neural mechanisms. It also implies that independently evolved behaviors can have different neural mechanisms and that homologous neural circuitry can underlie divergent behavior. In this article we directly examine the relationship between neural circuitry, behavior, and phylogeny. Understanding these relationships can lead to insights into the evolution of neural circuits and behavior (Katz 2011, 2016).

The nudibranchs Melibe leonina and Dendronotus iris exhibit similar swimming behaviors that consist of flattening their bodies in the sagittal plane and rhythmically flexing from side to side (Fig. 1, A and B). This left-right (LR) swimming behavior is also seen in several other nudibranch species (Newcomb et al. 2012). A recent molecular phylogeny of nudibranchs shows that Dendronotus and Melibe fall within a clade containing only genera that exhibit LR swimming behavior (Fig. 1E; Goodheart et al. 2015), suggesting that LR swimming was present in the most recent common ancestor of Melibe and Dendronotus and is therefore homologous (Sakurai and Katz 2017).

Fig. 1.

Species in this study and their phylogenetic relationships. A–D: images of the species in this study. Melibe leonina (A), Dendronotus iris (B), and Flabellina iodinea (C) are left-right (LR) flexion swimmers. Tritonia diomedea (D) is a dorsal-ventral (DV) swimmer. E: phylogenetic tree shows the subjects of this study highlighted in bold. The type of swimming behavior each genus exhibits is noted next to its name. The most recent common ancestor of Melibe and Dendronotus to have LR swimming is marked with a red circle. LR swimming evolved independently in Flabellina (magenta circle) and Bornella (green circle). Because other species within the genus Flabellina do not swim, it is possible the F. iodinea and F. telja evolved swimming independently; the phylogeny of this genus is not resolved well enough to distinguish between these different scenarios. Tritonia evolved DV swimming (orange circle). Most of the 2,000 species of nudibranchs are nonswimming (NS). Phylogenetic tree is based on Goodheart et al. (2015) and Newcomb et al. (2012a).

Although the LR swimming behaviors exhibited by Melibe and Dendronotus are homologous, their neural mechanisms are quite different. The central pattern generator (CPG) underlying the rhythmic LR swimming behavior in Melibe consists of four bilaterally represented neurons Si1, Si2, Si3, and Si4 (Sakurai et al. 2011, 2014a; Thompson and Watson 2005). In contrast, the LR swim CPG in Dendronotus is much simpler, containing just two neurons (Si2 and Si3); Si1 is not part of the CPG in Dendronotus and a homolog of Si4 has not been identified (Sakurai and Katz 2017).

In Melibe, Si1 can be identified by three neuroanatomical features. 1) Its soma is in the cerebral ganglion, near the cerebral serotonergic posterior (CeSP) cluster, a group of about five neurons, which includes homologs of the three dorsal swim interneurons (DSIs). 2) It is immunoreactive for the neuropeptide FMRFamide. 3) Its axon has a characteristic posterior bend before it continues to the ipsilateral pedal ganglion and then through the large pedal commissure (PP2) to the contralateral pedal ganglion (Sakurai et al. 2011; Thompson and Watson 2005). Si1 was identified in Dendronotus on the basis of the same three anatomical and neurochemical criteria that identify Si1 in Melibe (Sakurai et al. 2011). Because the same criteria uniquely identify this neuron in both species, it is mostly likely homologous (Croll 1987).

Despite being homologous, the connectivity of Si1 differs in the two species (Fig. 2, A and B). In Melibe, the contralateral Si1 neurons monosynaptically inhibit each other (Fig. 2C) and exhibit a small degree of electrical coupling (Fig. 2A and Table 1; Sakurai et al. 2011; Thompson and Watson 2005). In addition, Si1 is strongly coupled to the ipsilateral Si2 with a coupling coefficient approximately ten times more than that of the Si1-Si1 connection (Fig. 2A; Sakurai et al. 2011). However in Dendronotus, Si1 does not inhibit its contralateral counterpart (Fig. 2, B and D). Instead, those two neurons show a higher degree of electrical coupling than do their homologs in Melibe (Fig. 2D and Table 1). Furthermore, in Dendronotus, Si1 exhibits a lower coupling coefficient with the ipsilateral Si2 (Sakurai et al. 2011).

Fig. 2.

Comparison of Si1 connectivity and activity in Melibe and Dendronotus. A and B: species differences in synaptic connectivity. In Melibe (A), the left (L) and right (R) Si1 are reciprocally inhibitory (circles; Thompson and Watson 2005). They have strong electrical coupling to the ipsilateral Si2 (resistor symbol) and are reciprocally inhibitory with the contralateral Si2 (Thompson and Watson 2005). In Dendronotus (B), Si1 exhibits only electrical coupling with its contralateral counterpart and with both Si2s (Sakurai et al. 2011). The thicker lines correspond to higher coupling coefficients. C and D: simultaneous intracellular recordings show that Si1 has opposite effects in Melibe and Dendronotus. In Melibe (C), injection of depolarizing current (1 nA; arrows) into the R-Si1 hyperpolarized the contralateral Si1. In Dendronotus (D), injection of depolarizing current (2 nA) into the R-Si1 depolarized the contralateral Si1. The resting potential of the L-Si1 was about −45 mV in recordings in both C and D. E and F: simultaneous intracellular microelectrode recordings from L- and R-Si1 and Si2 show differences in the activity of Si1 in both species. In Melibe (E), Si1 fired bursts of action potentials in alternation with its contralateral counterpart and roughly synchronously with bursts in the ipsilateral Si2. In Dendronotus (F), Si1 fired slowly and irregularly while the two Si2s fired bursts in alternation. C–E show original recordings that replicate previous results from Sakurai et al. (2011).

Table 1.

Study species and characteristics

Genus	Type of Swimming	Si1 Part of Swim CPG	Si1 Reciprocal Inhibition	Si1–Si1 Electrical Coupling
Melibe	LR*	Yes	Monosynaptic and polysynaptic	0.02†
Dendronotus	LR*	No	Absent	0.06†
Flabellina	LR	Yes	Polysynaptic only	0.10
Tritonia	DV	No	Monosynaptic	0.04

^*LR swimming in Melibe and Dendronotus is homologous.

^†Sakurai et al. (2011).

In Melibe, Si1 is rhythmically active with consistent phase relationships to the other CPG neurons (Fig. 2E), motor efferents, and swimming movements (Thompson and Watson 2005). Prolonged hyperpolarization or depolarization of Si1 halts swimming, whereas brief stimulation resets the rhythm, consistent with its membership in the swim CPG (Sakurai et al. 2011; Thompson and Watson 2005). Si1 directly excites efferent neurons in the ipsilateral pedal ganglion, thereby conveying the motor pattern to the effectors (Thompson and Watson 2005). In Dendronotus, however, Si1 is not rhythmically active during the swim motor pattern (Fig. 2F). Neither hyperpolarization nor depolarization of Si1 in Dendronotus causes the swim motor pattern to cease, indicating that it is not a member of the swim CPG (Sakurai et al. 2011). Thus, although both Melibe and Dendronotus have homologous behaviors and homologous neurons, those neurons have different connectivity and play distinct roles in the production of the behavior.

On the basis of just two species, it is not possible to determine if there is an evolutionary trend in circuit organization. For example, it is ambiguous as to whether the Dendronotus Si1 lost reciprocal inhibition or whether the Melibe Si1 gained it. Therefore, we identified the homologs of Si1 in two other nudibranch species. One of the species, Flabellina iodinea, is another LR swimmer (Fig. 1C). However, it is likely that Flabellina evolved LR swimming independently from Melibe and Dendronotus, based on its position in the nudibranch phylogeny compared with Melibe and Dendronotus (Fig. 1E; Carmona et al. 2013; Goodheart et al. 2015) and the absence of swimming in various other lineages (Newcomb et al. 2012). Of the dozens of species in the genus Flabellina, only F. iodinea (Farmer 1970; MacFarland 1966) and F. telja (Farmer 1970; Ferreira and Bertsch 1972; Marcus and Marcus 1967) have been reported to swim. Although the absence of reports on swimming does not mean that other Flabellina species do not swim, we tested and reported that F. trophina does not swim (Newcomb et al. 2012). Thus it is likely that LR swimming arose within the Flabellina lineage. The other species examined in this study, Tritonia diomedea (synonymous with T. tetraquetra Pallas, 1788), swims with dorsal ventral (DV) whole body flexions (Fig. 1D; Willows 1967; Wyeth and Willows 2006). The DV swim CPG in Tritonia consists of a set of neurons that are not homologous to any of the LR swim CPG interneurons (Newcomb et al. 2012).

MATERIALS AND METHODS

Animal collection and dissection.

Flabellina iodinea (2.5–6 cm in body length) were obtained as adults from Marinus Scientific (Long Beach, CA). Adult Melibe leonina (3–10 cm) and Dendronotus iris (6–20 cm) were obtained from Monterey Abalone (Monterey, CA). Tritonia diomedea (5–20 cm) were provided by Living Elements (Vancouver, BC, Canada). All animals were maintained in recirculating artificial seawater (Instant Ocean, Blacksburg, VA) tanks on a fixed 12:12-h light-dark cycle at 10–13°C.

Flabellina, Melibe, and Dendronotus were anesthetized by injecting the body cavity with 0.33 M magnesium chloride solution. Tritonia were anesthetized by chilling in a 4°C refrigerator. To remove the brain, a cut was made on the dorsal body wall near the esophagus. The brain, consisting of the paired cerebral, pleural, and pedal ganglia, was extracted from the body by cutting all nerve roots. The brain was transferred and pinned to a Sylgard-lined dish, where it was superfused with physiological saline or artificial seawater (Instant Ocean, Mentor, OH) at a rate of 1 ml/min at 10–12°C. Saline composition was (in mM) 420 NaCl, 10 KCl, 10 CaCl₂, 50 MgCl₂, 10 d-glucose, and 10 HEPES, pH 7.6. The surrounding connective tissue and the sheath immediately encasing the brain were manually removed using fine forceps and scissors. The brain was kept at 4–6°C during the dissection procedure.

Electrophysiology.

Intracellular recordings were made using glass microelectrodes (12–30 MΩ) filled with 3 M KCl or a mixed solution of 2 M K-acetate and 0.2 M KCl and connected to an Axoclamp 2B amplifier (Molecular Devices, Sunnyvale, CA). Extracellular nerve recordings were made by gently drawing nerves of interest into polyethylene tubing or a heat-polished glass pipette filled with normal saline. These suction electrodes were connected to an A-M Systems Differential AC Amplifier (model 1700; A-M Systems, Carlsborg, WA). Both intra- and extracellular recordings were digitized (>2 kHz) with a 1401Plus or Micro1401 analog-to-digital converter from Cambridge Electronic Design (CED; Cambridge, UK).

The effects on the swim motor pattern of current injection to Si1 homologs were examined by injecting positive or negative current (−9 to 5 nA) through a bridge-balanced microelectrode. To measure electrical connections between neurons, the amplifier was set to discontinuous current-clamp (DCC) mode. Data acquisition and analysis were performed with Spike2 software (CED). Digital filters were applied to remove high-frequency noise resulting from DCC.

To test for monosynaptic connectivity and direct electrical coupling between the swim interneurons, high-divalent cation (Hi-Di) saline was used to raise the threshold for spiking and reduce spontaneous firing. The composition of the Hi-Di saline was (in mM) 285 NaCl, 10 KCl, 25 CaCl₂, 125 MgCl₂, 11 d-glucose, and 10 HEPES, pH 7.6.

Tracer injections, immunohistochemistry, and imaging.

In preparations where neurons were injected with a tracer, the cell body was impaled by a microelectrode filled with a 2% solution of biocytin (Invitrogen, Carlsbad, CA) dissolved in 0.75 M KCl. The biotinylated tracer was injected via iontophoresis for 15–30 min (bipolar current pulses from −10 n to +3 nA, 1 Hz, 50% duty cycle). The preparations were maintained in normal saline for 1–4 h before fixation overnight in paraformaldehyde-lysine-periodate fixative: 4% paraformaldehyde, 1.85% lysine monohydrochloride, and 0.22% sodium periodate in cacodylate buffer (0.2 M cacodylic acid in 0.3 M NaCl, pH 7.4–7.6). The tissue was quickly rinsed several times with phosphate-buffered saline (PBS) (50 mM Na₂HPO₄ in 140 mM NaCl, pH 7.2) followed by two longer PBS rinses (3 h each). The tissue was then washed twice with 4.0% Triton X-100 in PBS (3 h each) and incubated in antiserum diluent (ASD; 0.5% Triton X-100, 1% normal goat serum, and 1% bovine serum albumin in PBS) for 1–2 h. The brains were incubated for 3–5 days in either primary rabbit polyclonal anti-serotonin (ImmunoStar, Hudson, WI) or anti-FMRFamide (ImmunoStar) antiserum diluted 1:1,000 in ASD. To visualize the biotinylated tracer, streptavidin-Alexa Fluor 594 conjugate at a dilution of 1:200 was added. The brains were washed six times (1 h each) with 0.5% Triton X-100 in PBS and then incubated overnight in goat anti-rabbit antiserum conjugated to Alexa Fluor 488 (Invitrogen) diluted to 1:100 in ASD. Finally, the tissue was washed six times (1 h each) with 0.5% Triton X-100 in PBS, dehydrated in an ethanol series (70%, 80%, 2 × 90%, 95%, and 3 × 100%, 20 min each), cleared in methyl salicylate, and mounted on a slide with Cytoseal 60 (Richard-Allan Scientific, Kalamazoo, MI). The tissue was kept at 4°C under gentle agitation for the entire immunohistochemistry protocol, except for dehydration and clearing.

The tissue was imaged using a Zeiss LSM 700 Axio Examiner D1 confocal microscope (Carl Zeiss, Oberkochen, Germany) with ×5–×20 objectives. Fluorophores were excited with two lasers (488 and 555 nm), and fluorescence emissions were passed through a 505- to 550-nm bandpass filter to visualize Alexa Fluor 488 and through a 560-nm long-pass filter to visualize Alexa Fluor 594. The LSM 700 software ZEN was used to acquire the images. Maximal projections of confocal stacks were made and exported as TIFF files to Adobe Photoshop CS where montages were assembled. Brightness and contrast of all images were adjusted when necessary.

RESULTS

Flabellina swims with LR body flexions.

The Flabellina LR swimming behavior is analogous to that of Melibe and Dendronotus (Lawrence and Watson 2002; Sakurai et al. 2011). Flabellina swims by flattening its whole body in the sagittal plane and bending from side to side, making a semicircle shape with its body on each flexion (Fig. 1C and Supplementary Videos S1 and S2). (Supplemental material for this article is available online at the Journal of Neurophysiology website.) The cycle period of the Flabellina swim, taken as the time between the start of a body flexion and the start of the next flexion on the same side, was 2.3 ± 0.18 s (n = 14). This is slightly faster than that reported period for the Melibe swimming behavior (2.7 ± 0.2 s; Lawrence and Watson 2002) or that of Dendronotus (2.9 ± 0.1 s; Sakurai et al. 2011). The LR swimming behavior occurred spontaneously or could sometimes, but not reliably, be triggered by noxious stimuli such as a tube foot of a sea star or a high-molarity salt solution. Swim bouts were highly variable in length, ranging from seconds to many minutes.

The Si1 homolog in Flabellina is identified by neuroanatomical features.

We identified a putative homolog of Si1 in the cerebral ganglia of Flabellina. It displayed the same anatomical and neurochemical characteristics that uniquely identify Si1 in Melibe (Thompson and Watson 2005; Watson et al. 2001) and its homolog in Dendronotus (Sakurai et al. 2011). Si1 in Flabellina had a nonpigmented soma located on the dorsal surface of the cerebral ganglion near the cerebral commissure (Fig. 3, A and B). Intracellular injection of biocytin revealed that its axon projected ipsilaterally to the pedal ganglion (n = 7; Fig. 3, A and B). The axon of Si1 had a slight bend in the ventral direction (Fig. 3C) as opposed to a posterior bend as in Melibe and Dendronotus (Sakurai et al. 2011; Watson et al. 2001). It should be noted that the overall shape of the brain is different in Flabellina, and this might account for differences in axon orientation. As in Melibe and Dendronotus, Si1 had large neurites that spread toward PP2 (Fig. 3, A and B) and fine neurites that emanated from the length of axon near the soma (Fig. 3C). In addition, a small neurite extended from the soma in the opposite direction from the axon (Fig. 3D).

Fig. 3.

Identification of Si1 in Flabellina iodinea. A: schematic of the Flabellina brain shows the location of the soma and axon of Si1 (pink) relative to the locations of the cerebral serotonergic neurons (green). The cerebral, pleural and pedal ganglia form a ring, which in the animal surrounds the esophagus. In the schematic, the pedal ganglion is twisted to be above the cerebral ganglion. Because of the torus-like orientation of the ring ganglia, traditional metrics of dorsal (D) and ventral (V) are difficult to apply. We refer to things closer to the surface in this orientation as “dorsal.” B–D: negative fluorescence images of biocytin fills of Si1. The Si1 soma is in the cerebral ganglion, and the axon projects into the ipsilateral pedal ganglion (B). The axon of Si1 showed a characteristic bend (arrow) in the cerebral ganglia in the D-V plane (C); arrow points to fine neurites emerging from the proximal axon. A fine neurite emerged from the soma opposite the axon (arrow in D). E: Si1 was labeled by intracellular injections of biocytin (pink) and was next to the serotonin-immunoreactive cerebral neurons (5-HT-ir; green). F: Si1 (pink) showed immunoreactivity to the neuropeptide FMRFamide (FMRF-ir; green), resulting in a white appearance. PP1 and PP2, pedal commissures 1 and 2.

The Si1 soma was located near the previously identified and highly conserved CeSP cluster (n = 3; Fig. 3E), which includes the three DSIs (Newcomb et al. 2006). Si1 was immunoreactive to the neuropeptide FMRFamide (n = 6; Fig. 3F). This suite of neuroanatomical and neurochemical characteristics uniquely identified Si1 in Flabellina, just as they do in Melibe and Dendronotus, and distinguished it from all other neurons in the brain, providing support for the hypothesis that it is homologous to Si1 in Melibe and Dendronotus.

Si1 is a member of the LR swim CPG in Flabellina.

A motor pattern consisting of alternating bursts in the left and right Si1 was reliably evoked in isolated brain preparations by brief electrical stimulation of a body wall nerve (BWN; Fig. 4A). Left and right pedal efferent neurons (PdN) also fired alternating bursts at the same periodicity as the Si1 bursts. The average burst period was 2.5 ± 0.51 s (n = 10), which is similar to the period of the swimming behavior in vivo. In semi-intact preparations in Melibe, it was shown that such bursting activity caused the LR swimming movements (Watson et al. 2002); therefore, it is likely to underlie the swimming movements in Flabellina, as well.

Fig. 4.

In Flabellina, Si1 is a member of the LR swim CPG. A: electrical stimulation of a body wall nerve (arrow) initiated alternating bursts of action potentials in the L- and R-Si1. B: injection of hyperpolarizing current into the R-Si1 during an ongoing motor pattern halted bursting in a contralateral pedal ganglion neuron (L-PdN). Bursting resumed after current injection ceased. C: injection of depolarizing current into the R-Si1 during a quiescent period initiated bursting in both neurons. D: a brief depolarizing current pulse into the R-Si1 reset the motor pattern. Circles indicate the expected times of bursts in a PdN. The phase relationship of the swim motor pattern remained stable before and after the reset (shaded bars). Recordings in B–D are from the same animal. The resting membrane potentials of both neurons are between −40 and −55 mV.

Although Si1 was rhythmically active in the Flabellina swim motor pattern as in Melibe, its duty cycle was longer. In Flabellina, Si1 fired action potentials for 44.2 ± 0.9% of the period (n = 10). In contrast, Si1 in Melibe was active for only 31 ± 10% of the cycle (see Fig. 2E) (Sakurai et al. 2014a).

Perturbing Si1 activity strongly affected the LR swim motor pattern in Flabellina. Prolonged hyperpolarization of Si1 halted rhythmic activity (Fig. 4B). Conversely, depolarization of Si1 when the swim motor pattern was quiescent could initiate and maintain bursting (Fig. 4C). Si1 was capable of resetting the motor pattern; a brief current pulse shifted the subsequent bursts (Fig. 4D; n = 10). Thus Si1 is a member of the LR swim CPG, making it similar to Si1 in Melibe but different from Dendronotus (Table 1).

The contralateral Si1s in Flabellina are electrically coupled and recruit polysynaptic inhibition onto each other.

When the swim motor pattern was quiescent, injection of depolarizing current into one Si1 had variable effects on the contralateral Si1; sometimes it would enhance spiking (Fig. 5Ai), but other times, even in the same preparation, it would decrease spiking (Fig. 5A, ii and iii). The contralateral Si1s sometimes exhibited synchronous burstlets consisting of only two or three spikes (Fig. 5A, asterisks).

Fig. 5.

Synaptic connectivity of Si1 in Flabellina. A: in normal saline, Si1 had variable effects on its contralateral counterpart. Injection of depolarizing current (0.5 nA) into the R-Si1 (i) initially excited the L-Si1, but a short time later (ii) caused it to stop firing. Injection of depolarizing current (0.5nA) into the L-Si1 (iii) similarly had a weak inhibitory effect on the R-Si1. When they were spontaneously active, the contralateral Si1s sometimes fired synchronous burstlets consisting of 2 or 3 spikes (*). The resting potentials of both cells were −45 to −50 mV. B: in Hi-Di saline, injection of weak depolarizing current into the R-Si1 (0.7 nA) depolarized the contralateral Si1 (i). Riding on top of the depolarization were small fluctuations (gray arrowheads) that were time-locked to the presynaptic spikes (dotted lines). When the R-Si1 was stimulated with progressively more current (ii, 1 nA; iii, 2nA), it recruited IPSPs (black arrows) that were not time-locked to presynaptic spikes. The resting membrane potentials were −64 mV for L-Si1 and −45 mV for R-Si1. C: the small membrane potential fluctuations shown in Bi were time-locked with almost no delay to the presynaptic spikes and not blocked in Ca²⁺-free saline. D: injection of hyperpolarizing currents (1.0, 1.5, and 2.0 nA) into the R-Si1 caused proportional hyperpolarization of the contralateral Si1 membrane potential. The resting membrane potentials were −45 mV for L-Si1 and −60 mV for R-Si1. E: proposed synaptic connectivity for Si1 in Flabellina has electrical coupling between the contralateral counterparts and recruited inhibition, possibly through an unidentified Si2-like neuron.

When the brain was bathed in Hi-Di saline to minimize recruitment of polysynaptic connections, injection of depolarizing current into one Si1 caused a depolarization of the contralateral Si1 with small membrane potential fluctuations that were time-locked to spikes in the stimulated Si1 (Fig. 5Bi, gray arrowheads). These small potentials persisted in Ca²⁺-free saline (Fig. 5C), indicating that they were electrotonic in nature. Injection of hyperpolarizing current into one Si1 caused its contralateral counterpart to hyperpolarize (Fig. 5D). The electrical coupling coefficient was 0.10 ± 0.03 (n = 6; range 0.08–0.15), which is larger than that observed in either Melibe (0.02) or Dendronotus (0.06; Table 1). The electrical connection was symmetric and exhibited no rectification.

As more depolarizing current was injected into one Si1, inhibitory postsynaptic potentials (IPSPs) appeared in the contralateral Si1, which were not time-locked to the presynaptic spikes (Fig. 5B, ii and iii, black arrows). The summation of these IPSPs in the contralateral Si1 reduced the amplitude of the electrotonic potential. The IPSPs increased in frequency with increasing spiking of the presynaptic Si1 (Fig. 5B). The effect of Si1 on its contralateral partner is consistent with it being coupled to a neuron that produces IPSPs in the contralateral Si1 (Fig. 5E). Thus, although Si1 in Flabellina functionally inhibits its contralateral counterpart in normal saline, the inhibition is weak, leading to variability. The inhibition is mediated through a polysynaptic connection and partially counteracted by electrical coupling. This is unlike Melibe, which evokes both monosynaptic and polysynaptic inhibition of the contralateral Si1 (Fig. 2, A and C, and Table 1).

The Si1 homolog is identified in Tritonia on the basis of neuroanatomical features.

In contrast to Melibe, Dendronotus, and Flabellina, Tritonia swims by flattening its body in the horizontal plane and making whole body DV flexions (Fig. 1D; Newcomb et al. 2012; Willows and Hoyle 1969). We identified a putative homolog of Si1 in Tritonia. Its soma was located on the dorsal surface of the cerebral ganglion, and its axon made a characteristic posterior bend (n = 8) near the soma and projected ipsilaterally to the pedal ganglion and into PP2 (Fig. 6, A and B). Although we were unable to visually observe the axon in PP2, we confirmed its presence by simultaneous recording in the soma of Si1, a motor follower neuron, and extracellularly on PP2; spikes generated in the soma were time-locked with excitatory postsynaptic potentials in ipsilateral pedal neurons and spikes on PP2 (Fig. 6C).

Fig. 6.

Identification of Si1 in Tritonia diomedea. A: schematic of the Tritonia brain shows the location of the soma and axon of Si1 (pink) relative to the locations of the serotonergic DSI somata (green). Unlike the Flabellina brain, the ganglia in Tritonia all sit on the dorsal side of the esophagus. The pedal commissures (PP1 and PP2) are longer and stretch around the ventral side of the esophagus. B: a negative fluorescence image of an Si1 biocytin fill shows the location of the soma in the cerebral ganglion and axon projection into the ipsilateral pedal ganglion. C: Si1 projects into the large pedal commissure (PP2) as shown by overlaid recordings of action potentials in the L-Si1 soma, EPSPs in a neuron in the ipsilateral pedal ganglion (L-PdN), and a time-locked impulse recorded on PP2. D: the axon of Si1 showed a characteristic bend in the cerebral ganglia. A neurite exited the soma opposite the axon (filled arrow), and fine neurites emanated from the proximal axon (open arrows). E: Si1 was labeled by intracellular injection of biocytin (pink) and was surrounded by the serotonin-immunoreactive (5-HT-ir) DSIs (green). F: Si1 was labeled by intracellular injection of biocytin (pink) and showed double labeling (pink-white) with antiserum against FMRFamide (FMRF-ir; green).

Similar to Melibe, Dendronotus, and Flabellina, we observed fine neurites emanating from the axon near the soma of Si1 and a neurite that emerged directly from the soma, opposite the axon (Fig. 6D). Si1 was located next to the serotonergic DSIs (n = 3; Fig. 6E) and was immunoreactive for the neuropeptide FMRFamide (n = 6; Fig. 6F). This suite of neuroanatomical and neurochemical characteristics identified the Si1 homolog in Tritonia, just as they do in Melibe, Dendronotus, and Flabellina, and distinguished it from all other neurons in the brain.

Si1 was not active during a DV swim motor pattern in Tritonia.

In Tritonia, the DV swim motor pattern does not exhibit left-right alternation; rather, neurons on the left and right sides fire in phase with each other. We found that Si1 did not fire rhythmic bursts of action potentials during a DV swim motor pattern (Fig. 7). Instead, the neuron decreased its firing, often stopping altogether (n = 11). Although it did not participate in the DV swim motor pattern, it received subthreshold synaptic input, phase-locked to the DV swim motor pattern. The left and right Si1 received synchronous inputs.

Fig. 7.

In Tritonia, Si1 does not participate in the motor pattern. Brief electrical stimulation of a body wall nerve (arrow) elicited a DV swim motor pattern consisting of bursts of action potentials in CPG neuron C2 and a pedal neuron (PdN). The L- and R-Si1 both became hyperpolarized during the swim motor pattern. They received simultaneous synaptic input in phase with the motor pattern and recovered spontaneous spiking ~30 s after the cessation of bursting in C2. Action potentials in both Si1s were truncated to better show synaptic inputs.

The contralateral Si1s in Tritonia exhibit reciprocal inhibition and electrical coupling.

In normal saline, Si1 caused a delayed inhibition in its contralateral counterpart (Fig. 8A). When examined in Hi-Di saline, the Si1 pair interacted with each other through electrical connections and inhibitory synapses (Fig. 8, B–D). Depolarization of one Si1 with a constant-current pulse produced unitary IPSPs that rode on top of a slow depolarization in the contralateral Si1 (Fig. 8B). The IPSPs corresponded one-for-one and had a constant latency with spikes in the presynaptic Si1 (Fig. 8C), suggesting that they are monosynaptic (Table 1). Si1 was electrically coupled to its contralateral counterpart; injection of hyperpolarizing current pulses into one Si1 hyperpolarized the contralateral Si1 (Fig. 8D). The electrical coupling coefficient was 0.04 ± 0.02 (n = 4; range 0.02–0.06), which is less than that seen in Flabellina but similar to that in Dendronotus and slightly higher than in Melibe (Table 1). Thus, in Tritonia, Si1 monosynaptically inhibits its contralateral partner while being electrically coupled to it (Fig. 8E).

Fig. 8.

Synaptic connectivity of Si1 in Tritonia. A: in normal saline, injection of depolarizing current (3 nA; arrows) in the L-Si1 caused a delayed inhibition of the contralateral Si1 (i). The connections were reciprocal (ii); injection of 3 nA in the R-Si1 caused a delayed inhibition in the L-Si1. The resting membrane potentials of both neurons were about −50 mV. A 0.8-nA current pulse was injected when the synapses were tested to show the inhibition of spiking. B: in Hi-Di saline, injection of depolarizing current (1.5 nA) into the L-Si1 caused the contralateral Si1 to depolarize (i). The same was observed by injecting 1 nA into the R-Si1 (ii). Riding on top of the depolarization were IPSPs. C: the IPSPs in the R-Si1were time-locked with a short delay to spikes in the L-Si1. D: injection of hyperpolarizing current into the L-Si1 caused a proportional hyperpolarization of the contralateral Si1. E: results indicate that in Tritonia, Si1 is electrically coupled to its contralateral counterpart and monosynaptically inhibits it.

DISCUSSION

Although Melibe and Dendronotus have homologous LR swimming behaviors and contain homologous Si1 neurons, the role of Si1 differs in the production of the behavior, as does its connectivity. It was of interest therefore to compare the role of Si1 and its synapses in other species. In this study we identified Si1 homologs in Flabellina iodinea, a nudibranch species that evolved LR swimming independently from Melibe and Dendronotus, and in Tritonia diomedea, a nudibranch species that swims with DV body flexions. We found that Flabellina represents a third circuit variation that produces rhythmic left-right alternation. Furthermore, the connectivity of Si1 in Tritonia was similar to that in Melibe, but Si1 did not produce alternating left-right bursting. The results are summarized in Table 1.

The connectivity of Si1 is different in each of the three LR swimmers.

Si1 in Flabellina is a member of the LR swim CPG; it is rhythmically active in phase with the motor pattern. Its activity is both necessary and sufficient to produce bursting. Brief perturbations of its firing reset the motor pattern. In this respect, Flabellina is more like Melibe than Dendronotus. However, there are differences between Melibe and Flabellina. First, the duty cycle of Si1 in Flabellina is greater than in Melibe, suggesting differences in how the circuit functions. The synaptic connectivity of Si1 in Flabellina differs from that in both Melibe and Dendronotus. Unlike in Melibe, Si1 in Flabellina does not monosynaptically inhibit its contralateral counterpart, but rather recruits inhibition. The recruited inhibition could be through a homolog of Si2, which we have not yet identified. Si1 is electrically coupled to its contralateral counterpart, more so even than Si1 in Dendronotus. This increase in coupling strength could arise from an increase in the expression of innexin genes, but it might also arise from the fact that Flabellina is much smaller than Dendronotus, and thus the site of coupling might be electrotonically closer to the recording and stimulating sites in the cell bodies. Thus the connectivity of Si1 in Flabellina differs from both Melibe and Dendronotus, representing a third neural mechanism that produces an LR swim motor pattern. We have not yet identified other members of the LR swim CPG in Flabellina, so we do not know the full extent to which these CPGs resemble or differ from each other.

Multiple solutions for neural circuits may allow dissociation of neural circuit from behavior during evolution.

Theoretical studies have suggested that the same motor pattern can be produced by different patterns of connectivity or cellular properties (Prinz et al. 2004). This was noted in the crustacean stomatogastric ganglion (STG), where individual identified neurons differ from preparation to preparation in the magnitude of particular ionic conductances (Golowasch et al. 2002). Many other differences have been observed between neurons and synapses within the STG that do not have an impact on the function of the system (Hamood and Marder 2014; Marder et al. 2015). In the leech heartbeat CPG, synaptic strength can vary up to sevenfold without an impact on the motor pattern (Roffman et al. 2012). In Tritonia, the connection between two swim CPG interneurons (C2 and VSI-B) varies between individuals with no obvious effect on the motor pattern under normal conditions (Sakurai et al. 2014b).

If CPG outputs are robust to differences in connectivity of neurons, then it is possible that individuals within a species could have heritable differences in neural circuitry, yet be behaviorally similar. If selection favored one circuit configuration over another, then it could become fixed in a population, leading to evolution of neural circuits without a change in behavior. This might be the case with Dendronotus and Melibe, which share a common ancestor that exhibited LR swimming but which differ in Si1 connectivity.

The potential to have multiple circuit configurations that produce the same output also suggests that convergent evolution of behavior could arise through different neural mechanisms (Katz 2016). Si1 connectivity in Flabellina represents a third type of configuration that produces LR swimming. The swimming behavior was not found in the most recent common ancestor of Flabellina or in the clade containing Melibe and Dendronotus, indicating that Flabellina independently incorporated Si1 into a swim CPG using polysynaptic inhibition instead of monosynaptic inhibition.

Si1 connectivity in a DV swimmer is similar to that in an LR swimmer.

In Tritonia, the connectivity of Si1 more closely resembled that of Melibe than Dendronotus; Si1 was electrically coupled to its contralateral counterpart and monosynaptically inhibited it. There was no evidence for recruited inhibition as was seen in Flabellina.

During the DV swim motor pattern, Si1 in Tritonia became hyperpolarized but received periodic depolarizing synaptic potentials near the end of the dorsal flexion phase of each cycle of the motor pattern. Thus Si1 was not part of the DV swim CPG. Furthermore, the reciprocal inhibition between the contralateral Si1s did not cause these neurons to burst in left-right alternation, indicating that this connectivity itself is not sufficient to account for the LR swim CPG. It has been shown in artificially created networks containing reciprocally inhibitory neurons that neurons can exhibit a variety of different firing states besides rhythmic alternation (Kleinfeld et al. 1990; Sharp et al. 1996). Furthermore, modeling studies have suggested that reciprocally inhibitory neurons could fire synchronously depending on the amplitude and time course of the synapses (Jalil et al. 2010, 2012; Van Vreeswijk et al. 1994).

Although it is not involved in swimming in Tritonia, Si1 likely plays a role in other motor behaviors. It strongly excites pedal neurons and induces large impulse activity in the body wall nerve (data not shown), indicating some motor function. The A4 neuron in Pleurobranchaea californica (another DV swimmer) may be homologous to Si1; it is located next to the DSI homologs and has an ipsilaterally projecting axon with a small bend (Jing and Gillette 2003), which is characteristic of Si1. Although A4 functionally inhibits its contralateral counterpart, unlike Tritonia, the inhibition is not monosynaptic. During a DV swimming motor pattern, A4 receives synaptic input but only fires weakly. A4 was characterized as a turning interneuron in Pleurobranchaea, being both necessary and sufficient for avoidance turns (Jing and Gillette 2003). Si1 in Tritonia might play a similar role in avoidance turns. Its inhibition during swimming would then suggest a behavioral hierarchy, where turning is suppressed during swimming. It would be of interest to see if Si1 causes turning in LR swimmers. Perhaps LR swimming evolved from an avoidance turning mechanism that became rhythmic.

The DV swim CPG in Tritonia is composed of the neurons, DSI, C2, and VSI (Getting 1989b; Katz 2009). Although homologs of DSI and C2 have been identified in Melibe, Dendronotus, and Flabellina, they are not part of the LR swim CPG in those species (Lillvis et al. 2012; Newcomb and Katz 2007, 2009; Newcomb et al. 2012). Despite not being part of the swim CPG in Melibe, DSI homologs share aspects of connectivity with those in Tritonia. In Melibe, as in Tritonia, the DSIs are electrically coupled to each other and recruit inhibition on to each other (Newcomb and Katz 2007). However, in Melibe, the electrical coupling is equal among the DSIs, whereas in Tritonia, DSI-A is most strongly coupled to its contralateral counterpart and DSI-B,C are strongly coupled both ipsi- and contralaterally. Thus there are differences in microcircuitry in other neurons, as well.

Reciprocal inhibition is not predictive of motor function.

In Tritonia as well as in Flabellina, the contralateral Si1s were reciprocally inhibitory. Reciprocally inhibitory neurons are a common circuit motif for network-based CPGs (Getting 1989a; Marder and Calabrese 1996). Although the presence of reciprocal inhibition between the Si1 homologs in Flabellina and Melibe and its absence in Dendronotus are consistent with the hypothesis that reciprocal inhibition is necessary for alternating bursting in Si1, the presence of reciprocal inhibition in Tritonia suggests that by itself, it is neither predictive nor sufficient to cause alternating firing in Si1. This indicates that one cannot predict function of neurons simply by examining their synaptic connections.

Variations in connectivity.

Neuronal homologies can be recognized across distant members of a phylogenic tree. For example, a giant serotonergic cerebral neuron is found in most gastropods (Pentreath et al. 1982; Weiss and Kupfermann 1976). This probably derives from shared developmental programs that generate the nervous system, as has been observed in insects (Thomas et al. 1984) and vertebrates (Finlay et al. 1998; Karten 2015). Although major brain areas and neurons are highly conserved, this and other studies suggest that microcircuitry is phylogenetically variable. For example, insect photoreceptors have undergone extensive changes in synaptic connectivity while retaining their neuronal identities (Shaw and Moore 1989). In the mammalian retina, species differences in the locations of input synapses are responsible for differences in velocity tuning of starburst amacrine cells (Ding et al. 2016). In the leech, differences in synaptic connectivity among homologous neurons contribute to different reflex responses (Baltzley et al. 2010). Homologous neurons can be recognized within the STG of decapod crustaceans on the basis of their connections to muscles, but the details of the microcircuitry within the STG differ across species (Katz and Tazaki 1992; Weimann et al. 1991).

In the present study, we found that the microcircuitry differed between the most closely related species (Melibe and Dendronotus), which share homologous behavior. We also found that circuitry could be similar even though the behavior varied. It should be noted, however, that all of the variability that we saw maintained a common theme: electrical coupling and reciprocal inhibition. The electrical coupling ranged from strong in Flabellina to weak in Melibe (Table 1). Similarly, the reciprocal inhibition was monosynaptic in Melibe and Tritonia, indirect in Flabellina, and nonexistent in Dendronotus (Table 1). The phylogenetic variability of connectivity within a theme suggests that these features have contributed to the evolution of neural circuits and behavior (Katz 2011).

GRANTS

This work was supported by National Science Foundation Grant IOS-1455527 and a Sigma Xi Grant-in-Aid of Research.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

C.A.G., A.S., and P.S.K. conceived and designed research; C.A.G. and A.S. performed experiments; C.A.G., A.S., and P.S.K. analyzed data; C.A.G., A.S., and P.S.K. interpreted results of experiments; C.A.G., A.S., and P.S.K. prepared figures; P.S.K. drafted manuscript; C.A.G., A.S., and P.S.K. edited and revised manuscript; C.A.G., A.S., and P.S.K. approved final version of manuscript.