Myogenic activity and serotonergic inhibition in the chromatophore network of the squid Dosidicus gigas (family Ommastrephidae) and Doryteuthis opalescens (family Loliginidae)

Chromatophores are responsible for colorful and dynamic patterning of the skin in coleoid cephalopods, a subclass that includes squid, octopus and cuttlefish. Chromatophores in these taxa are unique in that they are composed of an elastic pigment sac surrounded by a ring of radial muscle fibers that shorten when excited, thereby expanding the pigment sac and making it visible. Although all species of squid have chromatophores, our understanding of the mechanisms that control activity in the chromatophore network of this group derives from studies on a handful of species in the family Loliginidae (Messenger, 2001). Chromatophore displays of living squid (chromogenic behaviors) have been described for other families (Moynihan, 1983a,b; Bush et al., 2009; Burford et al., 2014; Rosen et al., 2015), but comparative work on anatomy and control mechanisms is lacking.

Chromogenic behaviors are subject to strong descending motor control, but coordinated chromatophore activity also occurs when central control is absent, most notably in the ‘wandering clouds’ display (‘wolkenwandern’ of Hofmann, 1907) that is readily observed in chronically denervated skin of squid and octopuses (Packard, 1992a,b, 1995a,b). Post-denervation activity of this sort has been studied in the loliginid squid Loligo vulgaris (Packard, 2001, 2006, 2011) and Doryteuthis opalescens (Packard, 1995a). Although waves of chromatophore activity have not been described in intact loliginids, wave-like activity is prominent in many species of cuttlefish and octopus (Hanlon and Messenger, 1996) as well as in conjunction flickering in the oceanic squid Dosidicus gigas (Rosen et al., 2015).

Coupling of chromatophores within the network itself has been proposed to permit peripheral ‘horizontal’ control over myogenic activity – in addition to the more widely recognized, descending ‘vertical’ control by the nervous system (Packard, 2001). Mechanistic details of how these two control systems might actively work in concert to coordinate chromogenic behaviors are largely unknown (Packard, 2011). Vertical control is hierarchical and top-down. Horizontal coupling is a network property, and connections between individual chromatophores must occur at the cellular level – yet little experimental attention has been paid to relevant mechanisms or structures underlying coupling in any species. Wave-like patterning does not have identical features in different coleoid taxa, as suggested by differences in the wandering clouds display cited above versus the repetitive ‘passing clouds’ of cuttlefish (Laan et al., 2014), and the balance of control between central and peripheral processes in vivo, as well as basic mechanisms themselves, may differ. This paper focuses on the chromatophore network in squid.

Because the chromatophore is the effector organ for chromogenic behaviors, regardless of the control pathways involved, properties of radial muscle fibers and the axons that innervate them are important to consider. Synaptic excitation generates post-synaptic potentials in radial muscle fibers of D. opalescens and ‘graded spike potentials’ that are thought to involve voltage-gated calcium channels occur in fibers of spontaneously pulsating chromatophores (Florey and Kreibel, 1969). Such graded Ca²⁺-based excitability is a common feature in invertebrate muscle cells (Zachar, 1971; Schwartz and Stühmer, 1984), but comparative work on radial muscle fibers in other teuthid species has not been reported.

Endogenous neurotransmitters of the chromatophore network have been examined in several loliginids (Messenger, 2001); again, comparative data are not available. Immunohistochemistry has revealed that L-glutamate (L-Glu) is contained in axons associated with radial muscle fibers (Messenger et al., 1997), application of L-Glu results in tonic expansion of chromatophores (Bone and Howarth, 1980; Florey et al., 1985; Messenger et al., 1997), and iontophoretic delivery of L-Glu results in excitatory post-synaptic currents in isolated radial muscle fibers (Lima et al., 2003). This body of evidence has led to the acceptance of L-Glu as the excitatory neurotransmitter in the chromatophore network in loliginids.

Serotonin [5-hydroxytryptamine (5-HT)] is also found in a subset of axons that are apposed to radial muscle fibers in loliginid squid (Messenger et al., 1997). Topical application of 5-HT results in retraction of expanded chromatophores (Florey, 1966; Florey and Kreibel, 1969; Messenger et al., 1997) but has relatively minor effects on nerve-induced contraction/relaxation cycles (Florey and Kreibel, 1969). Inhibitory postsynaptic potentials have not been observed in radial muscle fibers, nor does 5-HT affect membrane potential or ionic permeability of the muscle fibers, consistent with the idea that it does not act as a classical inhibitory transmitter (Florey and Kreibel, 1969). These observations, along with the complex nature of the intracellular, G protein-coupled pathways that mediate effects of 5-HT in other systems (Hannon and Hoyer, 2008), make it reasonable to conclude that 5-HT acts as an inhibitory neuromodulator in the chromatophore network of loliginid squid.

This paper provides the first characterization of the chromatophore network in an ommastrephid squid, D. gigas, and compares waves of chromatophore activity with those in denervated D. opalescens. Activity in both cases is resistant to tetrodotoxin (TTX) and eliminated by 5-HT. Waves thus occur in the absence of neuronal control through any known pathway. Immunohistochemical data confirm that L-Glu and 5-HT are endogenous to the chromatophore networks of both species. Both innervations in D. opalescens degenerate completely after denervation, and both are less developed in D. gigas. We propose that myogenic wave activity in these cases is facilitated by weak (or no) inhibitory control exerted by 5-HT, relative to the strong control in intact D. opalescens. Putative connections observed between branching distal ends of radial muscle fibers of neighboring chromatophores suggest that these structures may be relevant to the propagation of such activity.

Dosidicus gigas (d'Orbigny 1835) was captured in the Gulf of California within 3 km of Santa Rosalia, Baja California Sur, Mexico (27.339N, 112.267W) during June and July, 2015. All specimens were of the small-size-at-maturity phenotype that has persisted in this area since 2010 (Hoving et al., 2013; Robinson et al., 2016). Squid with a mantle length (ML) of 14–23 cm were caught using light fishing line and small jigs, and smaller individuals (8–10 cm ML) were caught at the surface with a dip net. Captured squid were immediately placed in coolers filled with ambient seawater (28–30°C) on the vessel and transported to an onshore laboratory where they were maintained for up to 24 h in a round tank (1.5×1 m) with recirculating seawater at 19–20°C. Squid were killed via rapid decapitation.

Small specimens of D. opalescens (Berry 1911) (8–9 cm ML) were caught in Monterey Bay, California, from the Monterey public wharf (36.604N, 121.889W), and larger squid (13–15 cm ML) were caught within 1 km of the shoreline near Hopkins Marine Station, CA, USA. All squid were caught using P-Line Squid Slayer Finesse (G. Pucci and Sons, Inc., Brisbane, CA, USA) squid jigs. Animals were maintained in holding tanks at Hopkins Marine Station with flow-through seawater at ambient temperature (12–15°C). Within 24 h of capture, squid were anesthetized in seawater containing 1% ethanol, and denervation was carried out by transecting the left pallial nerve just before its entry into the stellate ganglion. This was done through the mantle aperture without any other surgery. Squid were then returned to the holding tank for 7 days and regularly fed live goldfish or fathead minnows in accordance with Stanford University's Institutional Animal Care and Use Committee (IACUC) vertebrate-animal protocol. After visual confirmation of spontaneous wave activity in the denervated area, the squid was killed by decapitation for physiological and anatomical studies.

For pharmacological studies with D. gigas, an entire fin from small squid or a piece (2×2 cm) of fin from larger squid was removed and pinned out, ventral side up, on the Sylgard (Dow Corning, Midland, MI, USA) bottom of a dish filled with filtered seawater at room temperature (∼24°C). The epidermis was left intact because it was not possible to remove it without damaging the underlying chromatophores. For studies of (denervated) D. opalescens, a piece of mantle (2×2 cm) with skin attached and containing equal portions of the denervated area and contralateral intact side was pinned out in a dish filled with filtered seawater at room temperature (20–22°C), and a small window (5×5 mm) of epidermis was removed in the center of both denervated and intact portions to facilitate penetration of bath-applied reagents (TTX, 5-HT and L-Glu).

For both species, chromatophore activity was recorded at 30 frames s^–1 (720×480 pixels) with a Sony camcorder mounted on a tripod. Spontaneous activity was recorded for 2 min, and then electrical stimuli (1 ms duration) were applied to the skin using a tungsten bipolar electrode of 0.5 mm diameter (Rhodes Medical Instruments, Woodland Hills, CA, USA) and a Grass SD-9 stimulator (Grass Technologies, Quincy, MA, USA). After identifying threshold, chromatophore activity for each stimulus was recorded for five additional stimuli increasing at 1 V increments. This tended to give a near maximal response for the strongest stimulus. The same stimuli were then delivered in descending order. Each voltage was delivered a total of five times, alternating between ascending and descending order. The procedure was then repeated at four additional spots on the 2×2 cm skin sample (D. gigas) or 5×5 mm window (D. opalescens), one at each remaining corner and one in the center.

Following a stimulation experiment, the seawater was removed from the dish and replaced with an equal volume of test solution containing varying concentrations of 5-HT creatinine (H7752, Sigma-Aldrich, St Louis, MO, USA), 200 µmol l⁻¹ (D. gigas) or 100 µmol l⁻¹ (D. opalescens) L-glutamic acid (G1626, Sigma-Aldrich), or 200 nmol l⁻¹ (D. gigas) or 400 nmol l⁻¹ (D. opalescens) TTX (1069, Tocris, Bristol, UK). These concentrations of 5-HT and L-Glu were chosen based either on published studies (Florey and Kreibel, 1969; Cornwell and Messenger, 1995; Messenger et al., 1997) or, in the case of L-Glu and TTX, by adjusting concentrations until an effective dosage was found. The stimulation procedure described above was then repeated. In the case of TTX, 10–40 min were allowed between the time of application and experimentation for the toxin to show a maximal effect. In some cases the test solution was removed from the dish after the stimulation procedure and replaced with seawater to verify reversibility.

Recorded video clips were converted to image stacks using VirtualDub (v1.9.11, Avery Lee, Cambridge, MA, USA) and the images were analyzed using MATLAB R2014b (MathWorks, Natick, MA, USA). In both species, analysis was conducted over a 5×5 mm square. In D. gigas, the region of interest (ROI) was centered on the electrode for analysis of stimulated activity. The ROI for spontaneous activity was selected as the area with the most frequent activity (as judged visually) when the tissue was in seawater. This same ROI, which was assumed to be representative, was used when the skin sample was tested in the experimental solution. In D. opalescens, the ROI was always the 5×5 mm² window lacking the epidermis.

The RGB values for all pixels within the ROI were summed for each frame throughout the time course of the response that accompanied the strongest electrical stimuli (i.e. 5 V above threshold). Summed pixel values were used as a proxy for chromatophore activity, with darker pixels of a higher value representing expanded chromatophores. A 25 mm² area was equal to roughly 100×100 pixels, yielding a maximum pixel value of 3×255×10⁴, or 7.65×10⁶. In order to control for inherent variation of chromatophore response within the same field, an average of the five responses for each spot was treated as one replicate. These replicates were averaged to yield a canonical control (mean±s.d., total n=25) response for that preparation that could be compared with a test response computed in the identical manner under different experimental conditions. Spontaneous activity was analyzed over a 30 s time course, and summed pixel values for the ROI were computed for each frame to define the time course of the activity. The summed variance of pixel values over the 30 s sampling period was used as a measurement of spontaneous transient activity.

Skin samples from D. gigas were prepared for immunohistochemistry by first separating the skin from the ventral side of a fin from the underlying muscle. Skin samples were then pinned out at their natural size in filtered seawater. The iridophore layer was removed by dissection, and the combined chromatophore layer and epidermis were then fixed for 4 h at room temperature in either 4% paraformaldehyde or 4% paraformaldehyde plus 0.1% glutaraldehyde in sterile-filtered seawater. Addition of this low concentration of glutaraldehyde was found to be necessary for successful glutamate labeling, but higher concentrations interfered with detection of 5-HT in double-labeling experiments due to an increase in non-specific background labeling.

For immunohistochemistry in D. opalescens, a 2×3 cm piece of skin centered on the dorsal midline was removed from the mantle, thereby yielding two samples of both denervated and intact skin, each 1.0×1.5 cm. The intact side provided an internal control for the effects of denervation in that animal. The epidermis and iridophore layers of the skin were removed by dissection, leaving only the chromatophore layer. Samples were either fixed for 2 h in 4% paraformaldehyde in filtered seawater at room temperature to be labeled for 5-HT, or for 4 h in 4% paraformaldehyde plus 0.1% glutaraldehyde to be labeled for L-Glu. After fixation, three 0.5×0.5 cm pieces were taken from the center of each sample from both the intact and denervated sides of the fixed tissue to avoid any damaged edges.

All fixed samples were washed once in seawater, then in seawater mixed 1:1 with phosphate-buffered saline (PBS), and finally in PBS (10 mmol l⁻¹ Na₂HPO₄, 137 mmol l⁻¹ NaCl, 1.5 mmol l⁻¹ KH₂PO₄, 5 mmol l⁻¹ KCl). Samples were stored at 4°C in PBS with 0.05% sodium azide for a maximum of 4 days.

Prior to carrying out antibody labeling procedures, all samples were incubated in 5 mg ml⁻¹ type 1 collagenase (17100, Gibco, Carlsbad, CA, USA) in PBS for 30 min at 37°C. Samples were then blocked overnight in PBS with 0.1% Triton X-100, 6.25% goat serum (G9023, Sigma-Aldrich) and 0.1% bovine serum albumin (15561020, Invitrogen, Carlsbad, CA, USA) prior to and during incubation in primary antibodies to minimize nonspecific labeling. Samples fixed in paraformaldehyde only were labeled with a polyclonal anti-5-HT antibody (raised in rabbit) at a dilution of 1:200 (S5545, Sigma-Aldrich) for 3 h at room temperature. Samples fixed in paraformaldehyde plus glutaraldehyde were simultaneously labeled with anti-5-HT at a dilution of 1:500 and a monoclonal anti-glutamate antibody (raised in mouse) at a dilution of 1:10,000 (G9282, Sigma-Aldrich) for 3 h at room temperature. Samples were then washed in PBS with 0.1% Triton X-100, 6.25% goat serum and 0.1% bovine serum albumin, and incubated for 2 h in the secondary antibodies at a dilution of 1:250. Secondary antibodies used were anti-rabbit with a 546 nm fluorophore (A11010, Life Technologies, Carlsbad, CA, USA) to target anti-5-HT primary antibodies and anti-mouse with a 647 nm fluorophore (A21235, Life Technologies) to target anti-L-Glu antibodies.

Incubation with secondary antibodies also contained Alexa-Fluor-488–phalloidin (A12379, Life Technologies) at a 1:125 dilution to label actin filaments. After labeling with the secondary antibodies and phalloidin, the samples were incubated in 1:50 DAPI (D1306, Life Technologies) for 5 min and mounted on glass slides in 100% glycerol.

Controls were performed by incubating samples in secondary antibodies without first exposing them to any primary antibodies. No labeling was observed in such experiments. Controls for double-labeling experiments were carried out by labeling with one antigen at a time (Wessendorf and Elde, 1985), and no change in individual labeling patterns was noted in the double-labeling experiments. Both primary antibodies used are commercially produced and specificities are confirmed by the manufacturers. In addition, specificity of the anti-L-Glu antibody has been confirmed in several peer-reviewed publications, one of which compared the labeling patterns produced by different anti-glutamate antibodies (Kolodziejczyk et al., 2008). Specificity of the anti-5-HT antibody was also checked by mixing the antiserum with 500 µmol l⁻¹ 5-HT creatinine overnight and then using these ‘blocked’ antibodies to incubate samples from D. opalescens to be labeled as described above. This procedure greatly reduced 5-HT labeling in axons of the chromatophore layer.

Imaging of skin samples employed a Zeiss LSM 700 confocal microscope with a 20× dry objective and a 40× oil objective, and all images were taken from areas with minimal chromatophore damage. Under the 20× objective, tile scans were taken with each tile covering a 0.32×0.32 mm area in a z-stack with a thickness of 23 µm that spanned the entire chromatophore layer. Individual tiles for each layer of the z-stack were stitched together during processing to create an image representing a 1.6×1.6 mm region. Scans using the 40× objective covered an area of 0.16×0.16 mm.

Stereology was carried out manually on the 1.6×1.6 mm region stitched together from tiles taken using the 20× objective in order to quantify several features. All radial muscle fibers labeled with phalloidin in each image stack were counted. A muscle fiber was counted only if the proximal wedge that contained the nucleus could be identified and if the fiber was longer than 20 µm. Another count was made of the length of contact between muscle fibers and associated antibody-labeled axons. When determining the effects of denervation in D. opalescens, a muscle fiber with an associated axon was defined by a continuous line of positive labeling of the selected neurotransmitter that maintained direct contact with a radial muscle fiber for more than 10 µm. From these measurements, the percentage of radial muscle fibers with associated axons was calculated.

Two muscle fibers were considered to be ‘connected’ at the distal ends if both muscle fibers could be traced over their entire length from the proximal nucleated ends to the distal connection point with no detectable gap in the phalloidin labeling. Muscle fibers that simply overlapped were excluded by observing the samples on multiple planes using a z-stack with a vertical resolution of 3 µm using a 20× objective. From these measurements, the percentage of radial muscle fibers with putative connections to other muscle fibers was calculated.

When alive, D. gigas typically displays a ‘resting’ coloration of reddish brown on the dorsal surface and paler red to white on the ventral surface, a common countershading pattern in pelagic organisms (Packard, 1995c) (Fig. 1A). Although all chromatophores of D. gigas are reddish brown, subtle differences in skin color are evident. For example, the dorsal surface of the mantle shortly after death typically has a purplish hue, when most chromatophores are expanded (Fig. 1B), whereas the ventral side tends to be lighter red with an orange tint (Fig. 1C). Such differences are at least partially due to the degree of chromatophore expansion and/or different size classes of chromatophores, but these features cannot be easily distinguished in chromatophores of one color (insets in Fig. 1B,C). In addition, numerous iridophores confer a golden metallic-like sheen to the ventral (Fig. 1C) and lateral mantle. Differences between the dorsal and ventral surfaces of the fins, which contain few iridophores, are much less noticeable (compare Fig. 1B and C). Chromatophore density on the ventral surface of the fin (1344±273 cm⁻², n=4; this study) is comparable to that on the dorsal surface of the mantle (1480±174 cm⁻²) or head (1332±155 cm⁻²) (Rosen et al., 2015).

Electrical stimulation with a single brief shock can elicit two types of chromatophore activity in D. gigas: a rapid, localized response and, in approximately 12% of stimulations (29/243), a delayed wave that propagates outward from the stimulated area. The localized response is characterized by a brief latency (<33 ms=1 frame) and the synchronous expansion of a field of chromatophores around the site of stimulation ranging in size from a few units to an area >1 cm in diameter (Fig. 2A). Regardless of the size of the stimulated field, peak expansion of responsive chromatophores generally occurs within 200 ms and relaxation is fairly complete within 1–2 s. This type of response presumably results from electrical stimulation of motor axons in the skin that innervate a defined field of radial muscle fibers.

In approximately 12% of the applied stimulations, the localized response was immediately followed by a propagating wave of activity that spread from the excited area at a rate of ∼0.9 cm s⁻¹ (Fig. 2B). The direction of propagation was typically not symmetrical with respect to the localized response. No pattern of stimulation could be found that reliably produced waves, and identical stimuli at the same spot were also inconsistent. Relaxation time of chromatophores participating in a wave is 0.5–1.5 s, slightly faster than for chromatophores involved in the localized response.

Propagating waves also occur spontaneously in fresh skin preparations from D. gigas, with activity showing an irregular pattern with the wave front spreading with a velocity of ∼0.9 cm s⁻¹ (Fig. 2C, Movie 1). In some cases, similar waves repeat at irregular intervals of several seconds, but the exact pattern is generally variable. Based on the nature and speed of propagation, this spontaneous wave activity is similar to the stimulated waves described in the preceding paragraph, and it is likely that a common mechanism underlies propagating activity in both cases.

Spontaneous waves similar to those described above were also observed in denervated skin samples from D. opalescens (Movie 2), but this activity never propagated into the contralateral control side of the same sample. Similar activity was also evident in the denervated field of the living animal (not illustrated), but activity of this sort does not occur in intact living D. opalescens or in fresh skin samples from intact squid.

Stimulated localized responses in both D. gigas (Fig. 3A) and intact D. opalescens (Fig. 3B) are greatly reduced by exposure to 200–400 nmol l⁻¹ TTX. Because TTX has no known action other than blocking voltage-gated sodium channels, these results are consistent with the idea that electrical stimulation of the skin excites branches of motor axons. In contrast, the amplitude of spontaneous wave activity is not diminished by TTX in either D. gigas (Fig. 3C) or denervated D. opalescens (Fig. 3D).

In both species, application of 2.5 µmol l⁻¹ 5-HT leads to relaxation of any chromatophores that were tonically active and a pallid appearance of the skin (not illustrated). In D. gigas, 5-HT greatly reduces responses to electrical stimulation (Fig. 4A) but has no detectable effect on stimulated activity in intact D. opalescens (Fig. 4B). Spontaneous activity in both D. gigas (Fig. 4C) and denervated D. opalescens (Fig. 4D) is essentially eliminated by the same concentration of 5-HT. Preliminary studies on D. gigas with lower concentrations of 5-HT suggest that spontaneous activity is more sensitive than stimulated activity. In these studies spontaneous activity was eliminated by 0.9 µmol l⁻¹ 5-HT (n=3), but stimulated activity was reduced by only 12.8% (n=15).

In all preparations of D. opalescens and D. gigas skin, bath application of L-Glu leads to tonic expansion of most chromatophores, but it has no obvious effect on the transient responses of relaxed chromatophores to electrical stimulation or on spontaneous wave activity (not illustrated). Effects of the pharmacological agents tested on the overall temporal variation of spontaneous activity were quantified by calculating the variance of summed chromatophore activity during a 30 s observation, and only 5-HT has a significant inhibitory effect in either species (Fig. 5).

Samples from D. gigas contained axons labeled by anti-5-HT (Fig. 6A) and anti-L-Glu (Fig. 6B) that were associated with radial muscle fibers in all material examined. Both glutamatergic and serotonergic axons tend to run obliquely with respect to muscle fibers in their proximal region, with an average length of potential contact between an axon and a muscle fiber of 12±15 µm (mean±s.d., n=263 muscle fibers). The discrete nature of the labeling and low background, in conjunction with control experiments (see Materials and methods), suggest that the primary antibodies used are highly specific. Thus, both endogenous 5-HT and L-Glu are contained in axons within the chromatophore layer, and the pharmacological data presented above are consistent with L-Glu being an excitatory neurotransmitter and 5-HT being either an inhibitory neurotransmitter or neuromodulator.

In samples from intact D. opalescens, serotonergic axons are frequently associated with radial muscle fibers (Fig. 7A), with labeling by anti-5-HT occurring over the proximal region of the muscle fiber, consistent with previous findings (Messenger et al., 1997). The length of potential contact between a serotonergic axon and a muscle fiber is highly variable, with an average of 95±65 µmol l⁻¹ (n=316 muscle fibers) and therefore much longer than in D. gigas. In denervated samples there were no detectable serotonergic axons associated with radial muscle fibers (Fig. 7B and Table 1). Glutamatergic axons were also frequently observed in association with radial muscle fibers (Fig. 7C), and glutamatergic axons were also eliminated by the denervation procedure (Fig. 7D and Table 1). These findings confirm that transecting the pallial nerve results in complete degeneration of both glutamatergic motor axons and serotonergic axons associated with the chromatophore network.

Double-labeling of 5-HT and L-Glu in both D. gigas and D. opalescens reveals the presence of glutamatergic axons that do not appear to contain 5-HT, but the small diameter of these axons complicates analysis by confocal microscopy. It was often difficult to discern definitively whether 5-HT and L-Glu are colocalized in the same axon (Fig. 8A,B) because the axons run tightly wrapped around each other. However, when 5-HT and L-Glu labeling are viewed separately in the same sample, there are cases in which glutamatergic axons exist that do not appear to contain 5-HT (Fig. 8C–F). Although this does not rule out the possibility that a subset of axons contains both 5-HT and L-Glu, it strongly suggests that some, and probably most, glutamatergic axons do not contain 5-HT.

Labeling of actin filaments with phalloidin reveals prominent branching at the distal ends of the muscle fibers in D. opalescens (circled regions in Fig. 9A), as previously described in this species (Florey and Kreibel, 1969) and in Doryteuthis pealei (Bell et al., 2013). Similar distal branching also occurs in D. gigas (circles in Fig. 9B) and, in many cases, radial muscle fibers from different chromatophores appear to be connected through their distal ends (arrowhead in Fig. 9B). Examination of sequential optical sections (z-stacks) of 3 µm thickness showed that these muscle fibers do not appear to simply overlap, but rather intersect within the same plane. A total of 1037 muscle fibers in D. gigas (four squid) were examined and, of those, 8.0±2.3% (mean±s.d.) had a possible connection of this sort with a muscle fiber belonging to a different chromatophore. In most cases the two chromatophores were not directly adjacent but were separated by a third chromatophore.

Putative connections of this sort are rare in D. opalescens. Examination of 1625 muscle fibers in six samples of innervated skin showed that only 0.3±0.4% of radial muscle fibers appeared to be connected. In denervated skin this figure was higher (0.8±0.8%, n=1437 in six samples), but the difference between denervated and innervated skin is not significant (Wilcoxon rank-sum test, P=0.21). It again appears that connected chromatophores may not be directly adjacent to one another, suggesting that connections may primarily occur between chromatophores of the same color class (or size class in the case of D. gigas).

This study examines the structure and function of chromatophores in Dosidicus gigas, an oceanic ommastrephid squid, and compares results to parallel experiments on both intact and denervated Doryteuthis opalescens, a coastal loliginid species. We find that, as in loliginid squid, L-Glu and 5-HT are contained in axons associated with the chromatophore network in D. gigas, with L-Glu probably functioning as an excitatory neurotransmitter and 5-HT as an inhibitory modulator (or conceivably neurotransmitter). Although the roles of these compounds are similar in these two squid families, some major differences in the pattern of innervation of chromatophores and in potential connectivity between chromatophores merit consideration. Our results suggest that 5-HT plays an important neuro-modulatory role in inhibiting spontaneous chromatophore activity, but to a varying degree in the two species.

Striking differences exist in the innervation pattern of the chromatophore layer of D. gigas compared to that in loliginid squid. Our results on D. opalescens confirm that motor axons in loliginids run directly apposed to radial muscle fibers, with multiple synapses distributed along the proximal region of the muscle fiber (Florey and Kreibel, 1969; Reed, 1995a; Messenger et al., 1997). This differs from the situation in D. gigas, where both glutamatergic and serotonergic axons pass more obliquely across muscle fibers. In this case, both types of axons would be able to form at most only a few en passant synapses with any given muscle fiber, in contrast to the multiple neuromuscular junctions in loliginids.

The pattern of motor innervation in D. gigas is actually more similar to that reported in the octopus Eledone cirrhosa (Dubas, 1987). Spatial–temporal patterning by octopus chromatophores involves coordinated wave activity (see figure 15 of Packard and Sanders, 1971), and these animals also have a sparser innervation. Additional comparative studies of structure and function in the chromatophore network of additional cephalopod species would be extremely worthwhile in furthering our understanding of the control of chromogenic behaviors in these animals.

Prominent wave activity through the chromatophore network is readily apparent in the intact skin of D. gigas and in denervated D. opalescens. In D. gigas, these waves are associated with a sparse innervation pattern and, in D. opalescens, they are accompanied by a complete degeneration of both serotonergic and glutamatergic axons associated with the radial muscle fibers. In both species these waves were not affected by TTX at concentrations that greatly reduce electrically stimulated activity. This observation confirms that these waves are not controlled by the known neural control pathway that descends from the brain. TTX-resistant sodium channels have not been reported in either D. gigas or loliginid species based on a large body of work on giant axons and on cell bodies of the giant fiber lobe neurons that give rise to these axons (Rojas and Armstrong, 1971; Brismar and Gilly, 1987; Rosenthal and Bezanilla, 2002; Rosenthal and Gilly, 2003). Such channels do occur in gastropod molluscs (Gilly et al., 1997), however, and cannot be entirely ruled out.

We found no evidence of any residual neural plexus containing L-Glu or 5-HT in the skin of denervated D. opalescens. We have not successfully carried out immunohistochemical studies that specifically label all neurons, so other neuronal processes could have remained after denervation and escaped detection. However, no other endogenous neurotransmitters have been identified in the chromatophore network in loliginid squid (Messenger et al., 1997; Messenger, 2001) and no such nerve net has been found in either classical histological studies (Florey, 1966) or in conjunction with tubulin-labeling of the chromatophore layer in D. opalescens hatchlings (Mackie, 2008). Therefore, the existence of a peripheral neural plexus is highly unlikely, and we propose that TTX-resistant wave activity is most likely myogenic.

Our results suggest the existence of a physical connection between the branching distal ends of radial muscle fibers from different chromatophores in both species of squid studied. Previous descriptions of direct connections between chromatophore muscle fibers in other cephalopods (Froesch-Gaetzi and Froesch, 1977; Packard, 1995a) were largely dismissed after intracellular injections of Lucifer Yellow failed to reveal any such connections in L. vulgaris (Reed, 1995b). Although these experiments confirmed intercellular coupling between the nucleated basal regions of muscle fibers (Florey and Kreibel, 1969), they failed to reveal the distal branching of radial muscle fibers that is clearly evident in both D. opalescens (Florey and Kreibel, 1969; present study) and D. pealei (Bell et al., 2013), species that are closely related to L. vulgaris.

Although visualization using confocal microscopy alone cannot prove the existence of a physical connection between muscle fiber processes (rather than a simple region of apposition) or elucidate the structural nature of the connection, we found evidence of putative morphological connections in 8% of radial muscle fibers in D. gigas but in only 0.3% of muscle fibers in D. opalescens. The much larger fraction in D. gigas is consistent with the propagation of waves in fresh skin preparations and potentially in the living animal in conjunction with flickering behavior (Rosen et al., 2015). Waves occur in D. opalescens only after denervation of chromatophores, and this will be discussed below in regard to the loss of serotonergic inhibition.

Transmission across a distal connection between radial muscle fibers might involve intercellular junctions through which ions or small molecules can move (e.g. gap junctions), a physical connection through which mechanical activity of neighboring muscle fibers could be sensed by stretch-activated channels, or some other mechanism. Wave activity described in this study propagates at a velocity of ∼1 cm s⁻¹, at least an order of magnitude faster than intercellular Ca²⁺ waves in a large variety of tissues (Jaffe, 1991; Haas et al., 2006), and this discrepancy strongly suggests that Ca²⁺ waves alone cannot account for propagation of waves through the chromatophore network of squid.

Mechanisms responsible for the initiation of chromatophore waves are also unknown, but excitability properties of radial muscle fibers would be relevant. These cells do not show evidence of voltage-gated sodium channels in D. opalescens (Florey and Kreibel, 1969), but they do have voltage-gated calcium channels that provide a graded type of excitability that can support spikes that arise from generator potentials during spontaneous pulsations of single chromatophores (Florey, 1966). Such pulsations do not propagate, but it seems likely that a similar form of electrogenesis would be involved in initiating waves.

Vertebrate skeletal muscle fibers can become spontaneously active after denervation due to changes in density and properties of several types of ion channels (Harris and Thesleff, 1971; Miledi et al., 1971; Pappone, 1980; Caldwell and Milton, 1988; Neelands et al., 2001; Midrio, 2006). Comparable studies do not appear to have been reported for invertebrate muscle, but excitability of radial muscle fibers in D. opalescens could be enhanced post-denervation, thereby facilitating initiation and propagation of myogenic activity. Preliminary recordings using whole-cell patch-clamp methods have not revealed any qualitative change in electrical properties of chromatophore muscle fibers following chronic denervation in D. opalescens (W.F.G., unpublished data). Although changes in voltage-dependent Ca²⁺ and/or K⁺ currents cannot be ruled out at this time, Na²⁺ currents are not observed. Electrical properties of radial muscle fibers in D. gigas are unknown, but they might show enhanced excitability under normal conditions, and such a feature would be consistent with the relative paucity of glutamatergic neuromuscular contacts in this species.

Electrically stimulated responses and spontaneous chromatophore waves in D. gigas are both inhibited by 5-HT, but only wave activity in denervated D. opalescens was inhibited by this agent. We propose that the extensive serotonergic innervation of the chromatophore network in intact D. opalescens is responsible for the lack of spontaneous activity but does not interfere with electrically stimulated excitation via motor axons. This suggests that denervated skin in D. opalescens may be similar to the normal situation in D. gigas, where limited contacts between serotonergic axons and radial muscle fibers suggest that serotonergic inhibition is relatively weak, thereby facilitating a higher level of spontaneous activity.

Mechanisms of peripheral inhibition by 5-HT are not clear in either case. Intracellular recordings in D. opalescens did not reveal any effects of 5-HT on electrical properties of radial muscle fibers or on postsynaptic potentials (Florey and Kreibel, 1969), nor did 5-HT have significant effects on the rates of contraction and relaxation following nerve stimulation in fresh skin preparations. These observations led to the proposal that 5-HT acts intracellularly in radial muscle fibers to reduce the rise in intracellular Ca²⁺ necessary for muscle contraction (Florey and Kreibel, 1969; Lima et al., 1997, 1998). Serotonergic inhibition could occur at any of several levels between release of L-Glu from motor axons and the rise in intracellular Ca²⁺ in radial muscle fibers, i.e. the processes that define excitation–contraction coupling. In the case of myogenic activity, effects on excitability might also be relevant. Further discussion of mechanisms for serotonergic modulation of chromatophore activity in D. gigas or loliginid squid must be considered speculative at this time.

Despite uncertainty over mechanisms, serotonergic inhibition probably plays an important role in chromogenic behaviors of the living squid. Stimulated activity was unaffected by 5-HT in D. opalescens and there was some indication in our experiments that spontaneous activity in D. gigas is more readily suppressed by 5-HT than is stimulated activity. A significant difference in sensitivity would imply that endogenous 5-HT in the skin could provide inhibitory control over spontaneous chromogenic activity without interfering with neutrally driven activity. In D. opalescens, inhibition of spontaneous wave activity might be particularly important, because this species relies heavily on spatially well-defined displays for both camouflage and communication. Any significant degree of spontaneous chromatophore activity, particularly wave-type activity that characterizes the denervated condition, would seemingly disrupt these displays. Conversely, dynamic chromogenic displays are common to the repertoire of D. gigas (Rosen et al., 2015), and we propose that flickering behavior in the living squid represents a form of wave activity with weak serotonergic inhibition acting to keep flickering in a balanced state. Stronger and global inhibitory control of flickering is also possible in this species, as demonstrated by the squid's ability to rapidly halt this ongoing behavior, often in conjunction with the onset of other chromogenic displays, particularly flashing. It seems likely that 5-HT would also underlie this stronger inhibition via descending control through the serotonergic innervation.

Coordinated activity in the chromatophore network that is not directly driven by descending (vertical) motor control was previously identified as a distributed, horizontal control process (Packard, 2001, 2006). Results of the present study support this idea. Spontaneous waves of chromatophore activity are common in D. gigas, and this activity propagates via a TTX-resistant pathway within the skin. Transmission may involve the putative connections between the branching distal ends of radial muscle fibers, although the mechanism of coupling remains unknown. Nevertheless, such a horizontal control system might coordinate chromogenic activity in the living squid with minimal, or even no, descending vertical control.

Waves of chromatophore activity are not evident in living D. opalescens or in fresh skin preparations, but such activity is clearly present in denervated animals. Our comparative approach suggests that, if horizontal control is operative in intact D. opalescens, it must be much less well developed and/or more strongly inhibited than it is in D. gigas. Given the similarities in the chromatophore system between the two distantly related species of squid studied (i.e. different families), we propose that horizontal control is common to all species of coleoid cephalopods, but the extent to which it is involved in natural chromogenic behaviors will depend on the precise needs of a particular species in the context of its environment.

How horizontal and vertical processes coordinate to control chromogenic behaviors in real time remains to be elucidated for any specific case, but this interaction is evident from a different perspective. Position, resting size and color of individual chromatophores making up the network are the result of developmental history (Packard, 1985), itself a hierarchical, vertical process, and waves of chromatophore activity tend to involve chromatophores of the same size/color class (Packard, 2001; http://gillylab.stanford.edu/assets/packard/ap_vid1_320x240.ogv). This clearly provides a manifestation of vertical and horizontal control working in concert.

We thank Patrick Daniel for technical assistance and photography of squid on the R/V Puma at the invitation of Dr Carlos Robinson, UNAM, Mexico City; Diana Li, Elan Porter, Alex Norton and Joe Welsh (Monterey Bay Aquarium) for squid collection; Russel Williams (Hopkins Marine Station) for programming assistance; Leonel Orozco and staff at Minera Metallurgica Boleo, for general support; Karmina Arroyo Ramirez and Institutuo Tecnologico Superior de Mulege for assistance in the field; and Dr Christopher Lowe (Hopkins Marine Station) for guidance in immunohistochemistry procedures. We are particularly grateful for critical comments on the manuscript by Andrew Packard (research affiliate of Hopkins Marine Station).