From injury to full repair: nerve regeneration and functional recovery in the common octopus, Octopus vulgaris

More than 160 years of studies support evidence of outstanding regenerative abilities of cephalopod molluscs, showing that cuttlefish, squid and octopus are all capable of recovering the structure and function of a variety of damaged/lost tissues and parts, including the nervous system (reviewed in Imperadore and Fiorito, 2018). As in other organisms, the regenerative capabilities of cephalopods appear to rely on the plasticity expressed at the cellular level, and on the involvement of pluripotent elements such as stem cells or dedifferentiated cells, cell proliferation and migration (Imperadore et al., 2017, 2018; Zullo et al., 2017), and other regulatory and/or trophic factors including the expression of specific genes (for a general review, see Birnbaum and Alvarado, 2008; see also Imperadore, 2017).

One of the most remarkable cases of regeneration in cephalopods is represented by the pallial nerve, a paired neural structure connecting the central nervous system (i.e. the sub-oesophageal mass) to the periphery via the stellate ganglia (for description, see Fredericq, 1878; Young, 1929, 1971, 1972). These two long nerves, which extend around 40 mm in length in an adult octopus of about 250 g body mass having a mantle length of 95 mm, innervate mantle muscles and chromatophores in the skin (Fig. 1). Axons of neurons travelling in this nerve are involved in the neural control of respiratory muscles and skin patterning (Imperadore et al., 2017, 2018; Sanders and Young, 1974; Sereni, 1929; Sereni and Young, 1932).

Unilateral traumatic injury of this nerve leads to paralysis of mantle muscles and immediate paling of the skin on the ipsilateral side of the mantle (e.g. Imperadore et al., 2017; Sanders and Young, 1974). The ability of the pallial nerve to regenerate in octopus was discovered by Sereni and Young (1932), who described the phenomenon from a structural and morphological point of view. Subsequently, recovery of the chromatic function of the skin was also reported. In the course of the above-mentioned observations, it was shown that degeneration of both the central and peripheral stump of the pallial nerve occurred within the first 10 days after lesion, followed by an intense regeneration originating from the central stump first and later from the peripheral one (Imperadore et al., 2017; Sereni and Young, 1932; Young, 1972). Haemocyte infiltration was reported to occur for several days after nerve transection at the level of the cut stumps (Sereni and Young, 1932), together with active phagocytosis, scar formation and proliferation (Imperadore et al., 2017). Proliferation and, possibly, re-differentiation of connective tissue cells within the injured nerve were first observed by Imperadore and co-workers (2017, 2018).

The first signs of functional recovery of body patterning (i.e. chromatic and textural appearance of the skin; Borrelli et al., 2006; Messenger, 2001; for review, see Packard and Hochberg, 1977) have been reported to appear from 30 days post-lesion (Sanders and Young, 1974), but recent studies found that this may occur earlier, i.e. 7 days after injury (Imperadore et al., 2017).

Here, we integrated available knowledge on the morphological and physiological events occurring after severing the pallial nerve in Octopus vulgaris, by extending our previous observations beyond the first 14 days after injury (Imperadore et al., 2017), and report a new set of data describing the complete recovery of the nervous components and corresponding phenotypic appearance. Backfilling of the pallial nerve allowed detection of regenerating fibres in the nerve as early as 5 days to 5 months post-lesion. We assessed physiological recovery through examination of the conductivity of the axons in the pallial nerve at 5 months post-lesion. Furthermore, we provide a phenomenological analysis of mantle muscle contraction for breathing and chromatic patterning of the skin from loss of function (after lesion) to full recovery (∼1–5 months post-lesion).

A total of 14 Octopus vulgaris Cuvier 1797 of both sexes (8 females, 6 males, body mass 200–450 g) were utilized in this study. Animals were caught in the Bay of Naples (Italy) and following the practice available in G.F.’s laboratory at the Stazione Zoologica, Naples, Italy (Amodio et al., 2014; Fiorito et al., 1990), each animal was identified, sexed, weighed and housed in an experimental tank with running seawater. Octopuses were selected for being intact (i.e. no sign of lesion or regenerating parts).

Experiments were conducted during spring and autumn (seawater temperature ranging between 18 and 22°C). Following Fiorito and co-workers, the experimental setting was designed in order to simulate natural conditions at 3–4 m depth (Amodio et al., 2014; Borrelli, 2007; Fiorito and Scotto, 1992; Fiorito et al., 1990). Animals were individually housed in tanks located in a holding room with dark walls to which access was possible for experimenters and carers only. The tanks (60×100×50 cm) are made of dark grey PVC except for the front side, which consisted of a transparent glass panel (45×35 cm) to allow remote observation and video recording. A yellowish-brown layer of sand (355–500 μm granular size, 1 cm layer thickness), obtained directly from the coast off La Gaiola-Posillipo (Bay of Naples, Tyrrhenian Sea), was added to the bottom of each tank and a pair of bricks, set in a corner, served as a den. At the top of the tanks, a series of lamps (Neodymlite, Oy Airam AB, Finland) were positioned and programmed to switch on and off automatically according to the seasonal and daily rhythm at the latitude of the Bay of Naples. A dark blue curtain, from ceiling to the floor and running the entire length of the tanks, was positioned at a distance of 1.5 m from the front glass panel of each tank to hide both video equipment (video cameras, tripods, etc.) and the experimenter from the animals' view. A second curtain was positioned from the ceiling to the surface of the water at the level of the front glass panel. These curtains had a similar brightness to the tank walls, and also helped to hide the tester during the experiments.

Only animals fully capable of a normal predatory response (sensu Fiorito et al., 2015; Maldonado, 1963b) following acclimatization (about 1 week from capture; e.g. Fiorito et al., 1990; Maldonado, 1963a,b) were considered in this study.

Animals were presented with a live crab (Carcinus maenas) once a day to measure motivation to attack live prey – the so-called ‘readiness to attack’ (Amodio et al., 2014; see ‘Analysis of animal performance’, below, for details) – and as a food item.

Experiments with live octopuses included in this study were carried out before transposition of Directive 2010/63/EU in Italy (i.e. March 2014). Although no authorization was required, all procedures were performed in order to minimize the pain and distress of the animals involved (Andrews et al., 2013; Fiorito et al., 2014, 2015; Smith et al., 2013).

Octopus vulgaris were anaesthetized by immersion in 3.5% magnesium chloride hexahydrate (MgCl₂·6H₂O) in seawater for 15 min (Grimaldi et al., 2007). Surgery was performed as described in Imperadore et al. (2017). Briefly, octopuses were turned on their ventral side and the right pallial nerve exposed using a skin hooklet and completely transected using fine scissors. To standardize the site of injury along the nerve, the diameter of the stellate ganglion was measured for each animal. The same length was computed as the distance along the nerve to set the site of the lesion (3.5 mm on average). The completeness of the transection was verified under a stereo-microscope, allowing us to identify the two nerve stumps, which appeared with a gap in between of about 0.2 cm. The lesion did not require sutures. The left-side pallial nerve remained intact, serving as an internal control.

Animals were returned to their tanks soon after surgery and allowed to recover. Experiments were carried out only on animals that succeeded in recovering their predatory behaviour and that exhibited a stable and prompt attack behaviour in response to the live prey (Amodio et al., 2014; see ‘Analysis of animal performance’, below, for details).

After recovery, octopuses were maintained in experimental tanks and kept alive for 5 days (N=3), 30 days (N=3), 45 days (N=3) and 5 months (N=5). Animals were observed daily to identify functional recovery for skin patterning, papillae and mantle muscle contraction. Animal care was provided on the basis of the established practice in G.F.’s laboratory (Stazione Zoologica, Naples, Italy) and in compliance with principles stated in Directive 2010/63/EU and guidelines available for the care and welfare of live cephalopods (Fiorito et al., 2014, 2015). All experiments were video recorded by remote-controlled cameras (Panasonic HDC-SD80 or Canon EOS 80D), which were hidden from the animals' view (Borrelli, 2007; Fiorito and Scotto, 1992; see above and Fiorito et al., 1990; Imperadore et al., 2017).

A live prey was presented to the octopus for a maximum of 5 min every day to assess the predatory response (i.e. readiness to attack). A failure to attack within this time was classified as ‘no attack’ (Amodio et al., 2014; Fiorito et al., 1990; Maldonado, 1963a,b). All behavioural observations were carried out at the same time of day, in the afternoon.

Video recordings of all animals were examined by independent observers, blind to the experimental design. For the purpose of this work we considered the predatory performance, behavioural responses, breathing rate and welfare status of the octopuses, as described below.

The predatory performance of the octopuses was assessed through qualitative and semi-quantitative variables (Amodio et al., 2014; Borrelli, 2007; following Borrelli et al., 2006) by measuring readiness to attack and asymmetries during locomotory–predatory responses (i.e. type of attack/approach). As a measure of readiness to attack, we considered latency of attack (LA) measured as the time (in seconds) elapsed from the first appearance of a crab at the water surface to just before the octopus pounced on the prey.

The type of attack that O. vulgaris perform is reported to depend on the initial position and orientation that the animal assumes when the prey/stimulus comes into its visual field. In most cases, animals use one eye only (monocular vision; Packard, 1963), i.e. the ‘positive’ eye. We assessed the orientation towards the prey (i.e. asymmetries, side of attack/approach, SA: left, right) during attack/approach responses of the octopus to the prey.

The above parameters were measured every day before and after lesion.

Changes in the appearance of chromatic and textural patterns between the lesioned and control sides were assessed. These were recorded daily when the animal was at rest (at least 10 min before) and after the presentation of the live prey in the octopus' tank to assess its readiness to attack. Through analysis of the video recordings, two independent observers scored the body pattering and the approximate areas of blanching of the skin (mantle and any other body part) according to descriptions provided elsewhere (Packard, 1991, 1995; Sanders and Young, 1974). Body patterning observed during the analysis of video recordings was also coded following Borrelli et al. (2006).

We assessed breathing rate through analysis of the video recordings. We also scored any side-bias in the use of the mantle musculature in providing efficient breathing movements. Snapshots of selected frames, where the animal appeared motionless, showing both sides of the mantle opening to the camera, were saved. At least 40 frames (∼5 snapshots per second considered in the analysis) taken from video recordings of each trial were subsequently analysed through ImageJ (https://imagej.nih.gov/ij/) to measure, from each frame, octopus head width (in pixels, as a reference) and the extension of the maximum mantle opening (for each side, calculated in pixels). For the sake of the following analysis, measures of mantle openings were considered as the relative difference in the maximum extension observed during each breathing act of the mantle opening on both sides.

Behavioural observations also served as an assessment of animals according to the principles stated in Directive 2010/63/EU. Signs based on appearance, behaviour and physiology were monitored following those included in table 5 of Fiorito et al. (2015).

At the selected time points (5, 30, 45 days and 5 months), octopuses were anaesthetized by immersion in a 3.5% solution of MgCl₂ in seawater. Animal behavioural responses were also observed and video recorded during anaesthesia; skin reactions to pinch on both sides of the mantle were assessed using grooved forceps (as described in Butler-Struben et al., 2018) in lesioned and control animals. When terminal anaesthesia was achieved (>30 min; Grimaldi et al., 2007), death was confirmed by transection of dorsal aorta, in compliance with recommendations included in Annex IV of Directive 2010/63/EU (see Andrews et al., 2013; Fiorito et al., 2015). After death, both nerves were harvested and used for backfilling experiments to evaluate nerve regeneration patterns and target re-innervation.

Only nerves from animals kept alive for 5 months post-lesion, prior to undergo backfilling, were used for electrophysiology experiments. Electrical stimulation of axons in the nerve was applied before the injury site with a stimulation electrode and recording of the induced activity was measured close to the stellate ganglion with a recording electrode. Analogous areas were selected for the contralateral control nerves.

The pallial nerve and the stellate ganglion of recovered and control sides of experimental animals were harvested and placed in a saline-filled dish. Bipolar hook electrodes (e.g. Büschges et al., 1992; Sauer et al., 1995) were placed at both ends of the pallial nerve and the electrodes were isolated from the surrounding saline by silicon paste (Baysilone, Fa. GE Bayer Silicones). For stimulation of axons at one electrode, rectangular current pulses of 0.2 ms duration and variable amplitude were used (generated by a Grass S48 stimulator via a Grass SIU5 isolation unit). The resulting compound spike of axons in the pallial nerve was recorded extracellularly with the second electrode. Signals were pre-amplified by an isolated low-noise preamplifier (model MA101, Electronics workshop, Institute for Zoology, University of Cologne). The signal was further amplified and high- and low-pass filtered (high-pass: 300 Hz; low-pass: 3 kHz) using a 4-channel amplifier/signal conditioner (model MA102, Electronics workshop, Institute for Zoology, University of Cologne). Signals from the stimulation and recording electrodes were digitized and recorded at a sampling rate of 12 kHz using a Micro-1401-3 analog-to-digital converter (Cambridge Electronic Design Ltd, Cambridge, UK) and Spike2 software (Cambridge Electronic Design). The threshold amplitude of the stimulus for eliciting a compound spike in the pallial nerve was set to 1.0× threshold (T) (Büschges et al., 1992; Hooper and Schmidt, 2017). The amplitudes for the stimuli were normalized to this threshold.

For evaluation of amplitudes of the compound action potentials induced by electrical stimulation of the pallial nerve, for each stimulation strength, the mean and standard deviation were calculated (considering five original compound action potentials per stimulus strength). Mean values of compound action potentials (lesioned and contralateral intact nerve) were also plotted versus stimulus strength.

Pallial nerves together with stellate ganglia and surrounding tissues were harvested from animals after death for neural tracing experiments. The backfill protocol was performed following Imperadore et al. (2019). In brief, the far end of the pallial nerve was placed in a Vaseline pool filled with tracer solution (Neurobiotin, 5% in distilled water, Vector Laboratories SP-1120, ID RRID:AB_2313575) while the remaining tissue was immersed in seawater. The tracer was left to diffuse for around 24 h at 4°C.

After dye diffusion, the tissues were fixed in 4% paraformaldehyde (PFA) in seawater for 1 h 30 min at 4°C and then washed in 0.1 mol l⁻¹ phosphate-buffered saline (PBS, pH 7.4). Dehydration was performed with an ascending ethanol series (20 min each), followed by a mix of ethanol 100% and methyl salicylate (1:1, 20 min) and by two changes of methyl salicylate (5 min each). Rehydration was achieved with a descending ethanol series (20 min each). The samples were treated with collagenase/dispase and hyaluronidase in 0.1 mol l⁻¹ PBS for 30 min at 37°C on a shaker (final concentration of the enzymes was 1 mg ml⁻¹).

After washing in 0.1 mol l⁻¹ PBS+1% Triton X-100 (PBS-TX), samples were placed for pre-incubation in 10% normal goat serum (NGS) in PBS-TX overnight at 4°C and the following day incubated with streptavidin conjugated to Cy3 (Jackson ImmunoResearch JI016-160-084, 1:200) in PBS-TX+10% NGS overnight at 4°C. Nuclei counterstain was performed with DAPI (14.3 μmol l⁻¹) in PBS-TX at room temperature (2 h).

Samples were washed in 0.1 mol l⁻¹ PBS and dehydrated with an ascending ethanol series (20 min each). For clearing, samples were immersed in 100% ethanol and methyl salicylate (1:1) for 20 min and then in pure methyl salicylate until the preparation was fully transparent and clear (usually 15 min). For visualization under the microscope, the whole nerve and ganglion were embedded in methyl salicylate in a metal slide (2 or 3 mm thick) closed on both sides by cover slips.

Samples were imaged on a Leica TCS SP8X laser scanning confocal microscope or a Leica DMi8 microscope using a 10× (zoom factor 0.40) or a 20× air objective (zoom factor 0.75). Tile Z-stack experiments of whole-mount tissues were performed on backfilled nerve and stellate ganglion, for both lesioned and control samples. The laser was set at 550 nm for neurobiotin detection, while DAPI was excited at 405 nm. Z-step size was set at 5 µm. Pixel format size was set at 1024×1024.

To visualize connective tissues in the samples, an upright multiphoton microscope was used (Leica SP8 MP). Laser Chameleon Vision II (Coherent) was tuned to 1050 nm for simultaneous excitation of SHG and neurobiotin; and 405 nm for excitation of DAPI. Two non-descanned HyD detectors were utilized to detect SHG signal (filter BP 525/50) and red signal (BP 585/40); one internal HyD was utilized for the DAPI objective. In this case, a 10× (zoom factor 0.40) air objective with a numerical aperture of 0.3 was utilized. Z-stacks were set at a step size of 10 µm.

Images were processed using Leica software LAS X or Fiji software.

Analyses of data on lateralization and latency of attack were carried out on a total of 68 O. vulgaris: N=34 (20 females, 14 males) before and after surgery (‘injured’), and N=34 (21 females, 13 males) naive octopuses (before and after anaesthesia) that served as the control.

Of the 34 animals belonging to the injured group, only 14 individuals were the live animals utilized in this study. The behavioural data for the remaining 20 animals were deduced from laboratory records and video recordings from previous experiments (Imperadore et al., 2017, 2018). In addition, and in application of the 3Rs principle (www.nc3rs.org.uk/the-3rs), behavioural data for the control group (N=34) were deduced from previous data available for O. vulgaris held in G.F.’s lab (at the Stazione Zoologica, Naples, Italy, between 2010 and 2013). Based on log notes for the G.F. lab animals, octopuses were selected for the following characteristics: (i) absence of lesion of any kind; (ii) body mass (ranging between 250 and 450 g); (iii) water temperature of the holding tanks during experimentation (18–22°C); and (iv) because they underwent anaesthesia according to the same protocol as the injured group of animals included in this study.

The number of octopuses used in this study was chosen in an attempt to both obtain a statistically representative group and implement the 3Rs strategy, reducing the number to a minimum. To further reduce the number of animals used, we obtained behavioural data from previous experiments, conducted in the same laboratory (at Stazione Zoologica, Naples, Italy) in comparable conditions (see ‘Animals utilized for statistical analysis’, above). Experiments, data and subsequent analyses were performed following recommendations on experimental design and analysis in laboratory animals (Festing and Altman, 2002) and, more specifically, guidelines for the care and welfare of cephalopods in research (Fiorito et al., 2015). Behavioural observations from video recordings and data analysis were performed by investigators blind to the experimental conditions. The experimental design was assessed through the NC3Rs Experimental Design Assistant (https://www.nc3rs.org.uk/experimental-design-assistant-eda), achieving power >0.90.

Statistical analyses were performed using SPSS software (SPSS Inc., released 2009, PASW Statistics for Windows, version 18.0, Chicago) to analyse (i) lateralization, (ii) latency of attack and (iii) recovery of the breathing function. Asymmetries (i.e. lateralization) and latencies of attack for the 68 octopuses taken into account were noted before anaesthesia/surgery and on the day following the procedures. Following Zar (1999) and Siegel and Castellan (1988), we applied McNemar's test to examine differences in the side animals' utilized during attack/approach (SA; within groups) for the two conditions (injured and control); comparisons between groups were tested using the chi-square test. Yates' correction for continuity was applied when necessary. Within- and between-groups comparisons on latency of attack data were tested through the Wilcoxon test and Mann–Whitney U test, respectively.

For analysis of the recovery of breathing function after nerve regeneration, we applied a Friedman ANOVA followed by Bonferroni correction. Wilcoxon matched-pairs signed-rank tests were used as post hoc procedure.

Whenever appropriate, Monte Carlo simulations were utilized to calculate exact probability (Monte Carlo exact module available in SPSS), providing asymptotic P, a Monte Carlo probability value (MCP) and a 95% confidence interval (CI) of the MCP.

Where box and whisker plots were utilized to plot data, we used the SPSS graphical output standards (i.e. boxes represent the interquartile range, 25th to 75th percentiles, bars within boxes indicate the median values, whiskers indicate the 10th and 90th percentiles, and outliers are marked). In all cases, statistical differences were considered significant at P<0.05.

After severing the right pallial nerve and following recovery from anaesthesia, all O. vulgaris returned to their dens, behaved normally and did not exhibit any sign of distress. In some instances and immediately after recovery from anaesthesia, octopuses exhibited grooming behaviour (see description in Borrelli et al., 2006) close to the injured area and/or inside the mantle cavity (Fig. 2E) on the side of lesion. In most instances, grooming disappeared after a few hours.

All octopuses were presented with a live crab 60 min after recovery and promptly attacked it in less than 30 s, performing their normal predatory behaviour; on most occasions, a full attack response was observed (see descriptions in Borrelli et al., 2006; Maldonado, 1963a,b; Packard, 1963). In order to evaluate any effect of the lesion, we considered the octopuses' predatory performance (median latency to attack before and after lesion, injured: LA_before=6.0 s, LA_after=5.9 s, P=0.966; control: LA_before=6.2 s, LA_after=6.9 s, P=0.993; after Wilcoxon signed-rank, N=34) and found no significant differences. We also compared predatory performance before and after anaesthesia or surgery in the two groups (injured LA_before versus control LA_before: Z=−0.343, P=0.731; injured LA_after versus control LA_after: Z=−0.527, P=0.598; both after Mann–Whitney U-test, N₁, N₂=34), confirming the two groups behaved similarly, independent of the treatment.

Furthermore, we considered the type of attack (i.e. SA: left or right) to assess whether the octopuses' behaviour was affected by treatment. Animals approached the prey consistently as they did before lesion (SA left versus right, injured: P=0.109; control: P=0.549; both after McNemar's test, N=34). Similarly, we tested the effect of SA between groups and found no significant difference (injured SA_before versus control SA_before: χ²=0.23, P=0.627; injured SA_after and control SA_after χ²=1.47, P=0.224; chi-squared test after Yates correction, N=68).

Before lesion, animals were fully able to exhibit their specific body pattern depending on the need, situation or stimulus they encountered in the tank and, in particular, to control bilaterally and unilaterally the chromatophores on each side of the mantle (Fig. 2A,B; see also Fig. S1A,B). However, immediately after recovery from surgery, half of the mantle (right side of the animal, i.e. the lesioned side) appeared white because of the absence of any chromatic and textural pattern (Fig. 2D). The contralateral side of the mantle retained the ability to perform the full range of body patterns, as in a normally behaving animal (Fig. 2D). We also observed paralysis of the mantle muscles on the lesioned side (compare Fig. 2C and Fig. 2F).

One to two days post-surgery, brown spots appeared on a pale background in the denervated area (see Fig. S1C) (see also Imperadore et al., 2017). At around the end of the first week/beginning of the second week post-surgery, injured octopuses exhibited some chromatic patterns on the lesioned side (see Fig. S1D), matching that on the contralateral uninjured side (at least in tone and grooves; see Fig. S1E). However, this capability was retained only when animals were at rest. Once moving around the tank and/or performing an attack/approach during the predatory response, immediate paling of the lesioned side was observed (see Fig. S1F). We did not notice further improvement at 2 weeks post-lesion. Full reappearance of the normal colour pattern of the skin occurred (injured O. vulgaris both at rest and during attack) after 45 days and in some cases 4 months after injury (see Fig. S1G,H; see also bottom right image in Fig. 3A). We observed recovery of the ability to control skin texture (linked to contraction and relaxation of papillae) from about 1 month after lesion (data not shown).

Pallial nerve transection also induced impairment of mantle contractions required by normal breathing (first week post-lesion; Fig. 3A, upper right image), when compared with both the contralateral mantle side and the same side before lesion (Fig. 3A, upper left image). We did not notice significant progress during the second week post-lesion, but improvement was observed between the third and fourth week post-lesion. Overall, we observed a significant departure from the baseline condition immediately after injury (Fig. 3B) and a recovery to the normal condition starting from the fifth week post-lesion. Efficient breathing on the lesioned side appeared restored only after about 1 month (between 37 and 45 days after surgery; Fig. 3A, lower left and right image, respectively).

We observed a trend in variation in the efficiency of breathing events after lesion (contraction/expansion of the mantle openings; χ²=45.02, N=7, P<0.001, after Friedman ANOVA; Fig. 3B). The recovery continued to progress with time, with the difference between day 30 and day 37 post-lesion not being significant (day 30 versus day 37: Z=−0.68, N=7, P=0.499; Monte Carlo P<0.575, 95% CI=0.565–0.585, after Wilcoxon signed-rank test; Fig. 3B).

We evaluated changes in appearance of the mantle skin at the end of the experiments once animals had been anaesthetized again for humane killing and tissue harvesting. At 5–8 min after immersion in the anaesthetic solution, we observed that the right side of the mantle (injured octopuses: 5, 14 and 30 days post-lesion) appeared yellowish, while the contralateral side and the remaining part of the body were pale (see Fig. S2), as normally occurs following immersion in the MgCl₂ solution (Butler-Struben et al., 2018; Gleadall, 2013). When the dorsal part of the mantle was taken out of the anaesthetic solution, but leaving the other parts of the octopus body immersed, we observed dark waves (i.e. wandering clouds) on a grey-white background at the level of the denervated mantle (see Fig. S2).

The denervated skin was also highly sensitive to mechanical stimulation, contracting following pinching; the same stimulus did not elicit any reaction on the contralateral side, as expected during anaesthesia (Butler-Struben et al., 2018; Gleadall, 2013).

This denervated skin area appeared to reduce in size with time. Indeed, using the four main papillae of the mantle as landmarks, we observed shrinkage of the area affected by these phenomena (i.e. wandering clouds, yellowing and mechanical excitability) a few days (around 5–7 days) post-lesion, as also described by Packard (1995), which continued to reduce with time. In particular, the skin along the midline, the mantle tip and close to the head on the injured side returned to its original appearance (see Fig. S2; see also Fig. 3A).

We observed the above-mentioned phenomena on the denervated skin during anaesthesia at all time points considered, until 30 days post-lesion (see Fig. S2). On occasion, when lesioned animals were anaesthetized 45 days post-lesion or later, the denervated side of the skin appeared as pale as the contralateral side (see Fig. S2), and wandering clouds were no longer observable. Under similar circumstances, skin contractions in response to light pinch or any mechanical stimulation were absent.

Backfilled contralateral control nerves did not show any differences in structure compared with naive control nerves (see Fig. S3 and Movie 1). Axons were traced from the nerve to the ganglion (see Fig. S3B,C); inside the ganglion, fibres grow around motoneurons and form the neuropil (see Fig. S3C left, and enlargement in the yellow rectangle). Some fibres were seen to exit through the stellar nerves, probably innervating chromatophores in the skin (see Fig. S3C). We also observed fibres originating from the centripetal cells of the stellate ganglion (see Fig. S3C right, Movie 1).

Lesion of the pallial nerve determined the formation of two stumps, a central one still connected to the brain (which was used for the tracing experiments) and a peripheral one, connected to the stellate ganglion in the periphery. Within the site of lesion and a few days post-surgery, a bridge of connective tissue appeared (see Fig. S4B,Bi). Fibres at 5 days post-lesion were not seen to cross the lesion site, but remained confined to the backfilled stump. When transected nerves were traced 30–45 days after surgery, axons were seen to cross the gap between the lesion and the ganglion (Fig. 4A,B; see also Movie 2). Even though axons appeared to regenerate in several directions, the majority of the tracked fibres found the right path, allowing them to enter the ganglion (Fig. 4A,B), where they formed a network around motoneurons (Fig. 4A,B,Biii). Furthermore, some of these axons were able to reach the stellate ganglion and also to leave it through the stellar nerves, probably to reach their final targets (i.e. chromatophores). At this time point (30–45 days post-lesion), no positive cells (i.e. centripetal cells) were ever observed inside the stellate ganglion, contrary to observations in naive and contralateral control ganglia (see Fig. S3C, Movie 1).

Regenerating fibres at the lesion site were moderately well organized in bundles, with each bundle able to direct toward the ganglion. Noteworthy, some of these axons were found close to the stellate ganglion, but did not make it through and were even seen to grow backwards and into the musculature (Fig. 4B,Bii).

The regenerated part of the growing axons did not appear to have a homogeneous calibre as in the case of intact fibres, but they showed multiple varicosities and some swelling. Branching and axon tip swelling were also observed for some of the axons that were not able to enter the ganglion (Fig. 4Bii).

At the level of the injury, regenerating fibres appeared to grow around non-fluorescent obstacles, possibly scar tissue or debris, which apparently diverted axons from a straight route (Fig. 4Bi; see also Movie 2).

Backfilling of pallial nerves at complete functional recovery showed a chaotic arrangement of fibres at the lesioned site, even though the majority of axons appeared to regenerate in the correct direction, i.e. towards the stellate ganglion (Fig. 5B, neurobiotin). We detected the presence of fibre thickenings (Fig. 5B), which were not observed at other time points or in the control nerves; varicosities and branching were still observable (Fig. 5D). Fibres were able to reach the ipsilateral stellate ganglion, form a network around motoneurons (Fig. 5C) and exit through the stellar nerves. Centripetal cells appeared, which traced back into the ganglion (Fig. 5C).

Tracing experiments also highlighted an increased number of vessels and small capillaries in lesioned nerves (Fig. 5B,D) again not visible in control nerves, suggesting the occurrence of angiogenesis during the regenerative events.

The site of lesion was recognizable as a tissue interruption (Fig. 5B, DAPI), being more evident on the dorsal side and less evident in the middle and ventral sides.

Regenerated fibres in the pallial nerve expressed axonal properties, as tested through two bipolar electrodes placed before the lesion site (central stump) and close to the stellate ganglion (peripheral stump) on the surgically isolated pallial nerve of regenerated animals 5 months post-lesion (Fig. 6A). Electrical stimulation via the stimulating electrode close to the central stump was found to systematically induce activity in axons crossing the lesion site in the recording electrodes close to the stellate ganglion in all experiments (Fig. 6B). With increasing stimulation amplitude, the extracellularly recorded compound action potential increased in two out of three preparations. Such an increase in compound action potential amplitude reveals that axons of different diameters are recruited upon stimulation of increasing strength (Fig. 6C). Interestingly, the activation of axons upon electrical stimulation in the contralateral control pallial nerve (Fig. 6A,C) showed a similar systematic correlation between stimulation strength and amplitude of the compound action potential.

Lesion of one of the two pallial nerves in O. vulgaris leads to the interruption of the neural circuits between the central brain and the stellate ganglion in the mantle (i.e. periphery). As each nerve and related ganglion controls only one side of the mantle (Fig. 1), loss of function can be immediately recognized on the ipsilateral side of the injury (see Fig. 2): this corresponds to skin paling (as originally observed by Fredericq, 1878) due to a lack of control of chromatophores in the skin and to the impairment of muscle function for the control of mantle breathing movements and papillae. The contralateral side preserves intact functionality (Imperadore et al., 2017). These dramatic changes appear reversible, because of the ability of the pallial nerve to functionally regenerate (Sanders and Young, 1974; Sereni and Young, 1932). We evaluated the timing and pathways of neural fibre regeneration and functional recovery for this nerve after complete transection.

All octopuses recovered the ability to control skin papillae at around 1 month following surgery. The first signs of mantle muscle contraction recovery, described as muscular tone improvement in Imperadore et al. (2017), were observed at 14 days post-lesion. Normal breathing was resumed between 30 and 37 days post-surgery for all individuals.

These findings are also consistent with the morphological data obtained from imaging of the backfilled pallial nerves. Axons from the central stump required a comparable amount of time to reach motoneurons in the stellate ganglion (controlling breathing muscles; Monsell, 1977; Young, 1972), where they form a net surrounding them and possibly build synapses.

The time required by fibres to reach the chromatophores in the skin was consistently longer, as fibres had to reach the ganglion and exit from it through the stellar nerves, thus to travel for several centimetres to innervate the millions of chromatophores spread throughout the skin of the mantle (estimated at around 200 mm⁻² in O. vulgaris: Messenger, 2001). All octopuses kept alive for 5 months after lesion were able to fully recover skin patterning. However, for one of them, 45 days was sufficient time, while two required around 2 months and the last two required 4 months for complete recovery. In all cases, animals anaesthetized for humane killing did not show wandering clouds on the skin of the lesioned side as observed at previous time points. The skin appeared as pale and as unreactive to pinch as the contralateral side of the mantle, comparable to what was observed in uninjured animals under anaesthesia.

If we consider regeneration and re-innervation in terms of fibre growth speed, we can speculate that axons from the central stump reach the stellate ganglion and completely re-establish synapsis with motoneurons at a speed of 10 µm h⁻¹, as they had to cover a distance of ∼3.5 mm to reach the ganglion and the same distance to pass through the ganglion. Our data confirm the rate of growth originally described by Sereni and Young (1932).

However, fibres directed from the nerve to the chromatophores apparently regenerate faster. Fibre speed in this case appears to be between approximatively 34 and 90 µm h⁻¹, considering 7 mm distance to exit the ganglion and around 90 mm to reach the farther chromatophores in the mantle, as is the case at complete recovery.

The large number of fibres found in the lesioned area together with their intricate pathways made it difficult to assess the true origin of the fibres. Even though we tracked axons from the central stump, we are not able to assert with certainty that these fibres originated from the central brain and not from the peripheral stump of the pallial nerve, which also start to regenerate a few days post-injury (Imperadore et al., 2017).

Fibre varicosities, swelling, branching, obstacle deviation and thickenings were visible in the regenerating nerves; the last of these were never observed in uninjured or contralateral nerves. Sereni and Young (1932) also report regenerating fibres deviating around obstacles; these may represent the residue of scars or debris of the degenerating materials, both identified a few days post-lesion (Imperadore et al., 2017, 2018) and also described to occur in mammals during peripheral nerve regeneration (Kang and Lichtman, 2013).

Swelling, branching and fibre aberration appear to persist for months; the regenerated nerves never regained their original appearance, even 5 months post-lesion, although all animals had perfectly recovered the normal function by this time.

We cannot exclude that alternative mechanisms might also be involved and aid in chromatophore re-innervation, such as intervention through collateral nerve sprouting of the stellar nerves from the opposite stellate ganglion (as suggested by Packard, 1995). This is further supported by the fact that the mantle midline and the areas close to it recover functionality much earlier than other areas (see Fig. S2). Further investigation is required to support this hypothesis.

Our electrophysiological experiments on samples taken 5 months post-lesion confirmed that lesioned nerves are able to re-establish functional connections between the brain and the periphery. The regenerated fibres express axonal properties, including voltage-dependent ion channels, and generate and propagate action potentials. Electrical stimulation via the stimulating electrode close to the central stump was found to systematically induce activity in axons crossing the lesion site in the recording electrodes close to the stellate ganglion. Increasing stimulation amplitude resulted in an increase in the extracellularly recorded compound action potential. Such an increase in compound action potential amplitude reveals that axons of different diameters are recruited upon stimulation of increasing strength (Büschges et al., 1992; Hooper and Schmidt, 2017).

Previous studies on regeneration of the pallial nerve in the octopus (Sanders and Young, 1974; Sereni and Young, 1932) considered 60–150 days for complete recovery of function of body patterning after injury (either through crush or cut). According to the authors, in some cases this never occurred and the animal continued to show a lack of full recovery (Sanders and Young, 1974; Sereni and Young, 1932). Complete recovery of body patterns after crushing the nerve required 9–10 weeks. However, no octopuses showed signs of colour pattern recovery until 50 days post-lesion, either in summer or in autumn when experiments were carried out (Sanders and Young, 1974; Sereni and Young, 1932). In the original study, six out of 10 animals recovered complete colour pattern (between 60 and 69 days), while only in two out of 10 octopuses was it possible to observe functional recovery of textural components (i.e. papillae; between 30 and 50 days). In contrast, only four animals (out of 10) recovered colour patterns following complete transection of the pallial nerve (signs of recovery at 30 days). In this case, the complete regain of function required 109 days. In addition, seven out of 10 octopuses recovered the ability to raise papillae (Sanders and Young, 1974; Sereni and Young, 1932).

Strangely enough, when regenerated nerves were inspected, Young and co-workers were not able to determine any clear correlation between histological appearance and functional recovery, also noting the inability in following fibre pathways and orientations (Sanders and Young, 1974; Sereni and Young, 1932).

In summary, our results show that during pallial nerve regeneration, neural connectivity between the brain and the stellate ganglion can be re-established by the formation of axons crossing the lesion site.

In our view, the octopus pallial nerve provides an exceptional example of regeneration in cephalopods, because: (i) regeneration is always relatively quick and efficient; (ii) it is immediately possible to evaluate loss and regain of function by evaluating respiration (i.e. breathing movements) and skin patterning on the mantle of the animal; (iii) each animal has a pair of nerves, one on each side of the mantle, allowing the same animal to serve as both experiment and control; (iv) the same nerve can be lesioned several times, providing almost complete regeneration at each instance (Packard, 1995); and (v) the majority of the cell soma of axons in the pallial nerve reside inside the brain and the others reside in the ganglia at the periphery – both can regenerate, allowing comparison of the regeneration ability of the two ‘systems’, one being spatially more centrally and the other more peripherally located; and (vi) octopuses represents a ‘simpler’ animal compared with vertebrates. However, regeneration/degeneration phenomena following lesion were found to be similar to those occurring in higher vertebrates (Wallerian degeneration; Imperadore et al., 2017; Sereni and Young, 1932).

The authors are thankful for the assistance and guidance given by Dr Astrid Schauss and Dr Christian Jüngst (CECAD Imaging facility, CECAD Research Center, Cologne, Germany) for accessing the imaging facility and for the use of Leica SP8 multiphoton microscope. The authors are also grateful to Leica Microsystems (P. Romano, K. Orellana) for assistance during various phases of this work. This work benefited from the networking initiative of the COST Action FA1301 – CephsInAction.