The Vertical Lobe of Cephalopods: A Brain Structure Ideal for Exploring the Mechanisms of Complex Forms of Learning and Memory
Abstract and Keywords
We show that the cephalopod vertical lobe (VL) is a promising system for assessing the function and organization of the neuronal circuitry mediating complex learning and memory behavior. Studies in octopus and cuttlefish VL networks suggest an independent evolutionary convergence into a matrix organization of a divergence-convergence (“fan-out fan-in”) network with activity-dependent long-term plasticity mechanisms. These studies also show, however, that the properties of the neurons, neurotransmitters, neuromodulators, and mechanisms of induction and maintenance of long-term potentiation are different from those evolved in vertebrates and other invertebrates, and even highly variable among these two cephalopod species. This suggests that complex networks may have evolved independently multiple times and that, even though memory and learning networks share similar organization and cellular processes, there are many molecular ways of constructing them.
Today, because of the Internet, many people know that modern cephalopods (coleoids), octopuses, cuttlefishes, and squids, are exotic animals, which are believed to be the most intelligent invertebrates, rivaling the level of intelligence of many vertebrates. Not many are aware of the tremendous scientific work that the group of scientists, led by J. Z. Young, M. J. Wells, A. Packard, B. Boycott, H. Maldonado, and J. B. Messenger, invested in the subject in the middle of the previous century. They were interested in these animals because they believed that the simplicity of the invertebrate nervous system, together with the complex behavior that coleoids demonstrate, would provide a good opportunity for discovering the neuroethological bases of complex behaviors. Together with intensive studies of cephalopod behavior, they thoroughly investigated cephalopod neuroanatomy and, with the help of lesion and stimulation experiments, were able to assign a more or less specific function to each of the about 40 distinct, interconnected lobes of the central nervous system of the octopus (Fig. 24.1) (Young, 1971).
Surprisingly, they have found that the nervous system of cephalopods is extremely large, not only for an invertebrate, but even with respect to lower vertebrates (Packard, 1972). For example, the nervous system of the octopus contains half a billion nerve cells, the majority being very small. This contrasts sharply with the nervous system of the gastropod Aplysia californica, which contains relatively few large and identifiable neurons, allowing researchers to reveal the principles of simple forms of short- and long-term memory. Some of these mechanisms, like the involvement of cAMP-dependent gene expression (Dash et al., 1990), have (p. 560) been found to be universal for long-term memory acquisition (Kandel, 2001).
In the 1950s, Boycott and Young (1955) and their colleagues revealed the function of the vertical lobe as the memory acquisition system of the cephalopod brain. These studies were very inspiring because they took place in parallel with probably the most influential work revealing the role of hippocampus in human learning memory, done by Brenda Milner on the patient H. M. (Henry Molaison; Stellar, 1957). Unfortunately, the attempts to study the neural bases of learning and memory in the cephalopod nervous systems reached a dead end in the mid-20th century due to the lack of suitable techniques (Rose, 2003), dashing hopes that cephalopods would become the animal of choice to study mechanism of complex behavior.
The technological advances of recent years have allowed Octopus vulgaris and other cephalopods to now become effective preparations for investigating the computational and neurophysiological bases of short- and long-term synaptic modulation that are important for the control of the complex behaviors of the coleoids. In contrast to the defensive reflex of Aplysia, in which learning and memory take place at the level of the reflex circuit itself (Kandel, 2001), cephalopods use a very large dedicated neuronal network functioning in parallel to the circuitry controlling the learned behavior (Fig. 24.2 and Fig. 24.7). This is very similar to the organization of complex learning and memory systems in vertebrates and especially resembles that of insects, where the projection neurons from the antennal lobe innervate in parallel the mushroom body and the lateral horn (Heisenberg, 2003).
Here we summarize these new findings. Our main premise is that octopuses evolved a learning and memory network with connectivity of the most generic association networks (Helmstaedter, 2015). This network possesses universal cellular mechanisms that mediate learning and memory, such as activity-dependent LTP together with robust neuromodulation systems. However, the molecular mechanisms mediating these rather universal cellular processes have converged to completely different and novel mechanisms. They are based on radical adaptations of common invertebrate processes, especially those that are found in other molluscs.
In the Conclusion, the neurophysiology and neuroanatomy of the octopus VL findings are recapitulated in a descriptive model that offers an explanation of how the VL system controls the stereotypic attack behavior of the octopus. This model (p. 561) provides a convenient and promising system to continue the analyses of the neuroethological bases of the learning and memory network in an animal in which this network plays an important part in its complex behavior.
A Short Account of the Cephalopod Nervous System
The nervous system of modern cephalopods (coleoids) is divided into three parts: the central brain (50 million cells in Octopus vulgaris); the two optic lobes (each with 80 million cells in octopus); and the peripheral nervous system, which contains the majority of neurons (320 million). The central brain is composed of around 40 interconnected lobes (ganglia) in which the cell bodies of the monopolar neurons lie in the outer region of each lobe and their processes form the central neuropil, as usual in invertebrate ganglia (Fig. 24.1; Bullock & Horridge, 1965; Young, 1971). The central brain is divided into suboesophageal and supraoesophageal parts by the esophagus that runs through its center, as in all invertebrates (Fig. 24.1A). The suboesophageal part functions similarly to the brain stem and spinal cord of vertebrates, with lobes responsible for vegetative functions like breathing (Boycott, 1961; Bullock & Horridge, 1965) and relaying activation commands to various effectors like the arm and mantle. The supraoesophageal lobes are responsible for the higher cognitive and executive functions of the central nervous system and learning and memory. The most prominent lobes (Fig. 24.1B) are the basal lobes, responsible for higher motor control (Boycott, 1961; Zullo et al., 2009; see chapter on Motor Control in the Octopus); the median superior frontal lobes (mSFLs) with their connection to the vertical lobes (the VL system), responsible mainly for visual learning; and the inferior frontal lobe (IFL) with its connection to the subfrontal lobes (SubFLs), forming a system that deals with chemotactile learning (Wells, 1978).
The Vertical Lobe Is Involved in Learning and Memory
The VL system is best known for its involvement in associative learning, particularly in the framework of short- and long-term associative and operant learning to inhibit the octopus’s innate attack behavior (Maldonado, 1963, 1964). This robust behavior led Aristotle to pronounce octopuses “stupid,” because their natural predatory drive (curiosity) makes them attack any reasonably sized moving object in their vicinity; they can be easily caught by simply waving a hand underwater (Aristotle, 1910). When the attack is contingent with an aversive outcome, octopuses in the laboratory, and likely also in nature, learn very quickly and remember for a long time to restrain this behavior. This very robust and quick training paradigm has served numerous behavioral experiments, such as assessing the features important for visual discrimination (Wells, 1978).
Early experiments showed that even after the removal or lesions of the VL, octopuses continued to attack crabs despite receiving electrical shocks, unless the intertrial interval was less than approximately 8 min (Boycott & Young, 1955). A lesion study also demonstrated that the VL is important for learning by observation (Fiorito & Chichery, 1995), a highly advanced form of social learning reported by Fiorito and Scotto in 1992. Lesions in the ventral part of the VL of the cuttlefish (Sepia officinalis) led to marked impairment in the acquisition of spatial learning, whereas lesions in the dorsal part of the VL impaired its long-term retention (Graindorge et al., 2006). The study in cuttlefish suggests possible significant differences between the behavioral roles of the VL in different coleoids. The cuttlefish VL may be important for navigation: in a very early experiment, Sanders and Young (1940) found that the cuttlefish VL enabled the animals to follow a shrimp to its hiding place. These experiments suggest that the VL, especially in cuttlefish, may be the site of “working memory” as it keeps information for only a short time. The hypothesis that cephalopod’s VL is involved in navigation is intriguing because it promotes the interesting possibility of existing “place cells” in this lobe, storing navigation information like the mammalian hippocampus (McNaughton et al., 2006).
The Anatomy of the Octopus Vertical Lobe System
The VL and the IFL (Fig. 24.1A) are morphologically unique structures in the coleoids brains, and both are associated with learning and memory. The VL is highly developed among all modern cephalopods, yet only octopuses (octopodes) are noted for their large IFL (Fig. 24.1A), which can be poorly developed or absent in decapods (cuttlefishes and squids; Hanlon & Messenger, 1996). These differences in IFL development are correlated with the reliance of the different animal on their arms for their interaction with the environment (see Motor Control chapter). Indeed, the octopus IFL is involved in chemotactile learning (Wells, 1978). (p. 562)
The large numbers of neurons in the VL and IFL are organized in layers with their processes aligned more or less in parallel, resembling a vertebrate brain organization. The insect mushroom body is similarly organized (Strausfeld et al., 2009): relatively few projecting-neurons originating from the antennal lobe innervate a large number of small Kenyon interneurons en passant in the mushroom body neuropil (Heisenberg, 2003). This organization pattern, typical for fan-out, fan-in networks (Fig. 24.2; Shomrat et al., 2011), is quite unusual in the invertebrate nervous system. It is not surprising, therefore, that such structure is highly suggestive of an invertebrates’ brain area involved in cognitive functions associated with learning and memory.
The octopus VL contains only two types of neuron, amacrine cells (AM, ~ 5 μm diameter) and large efferent neurons (LN, ~15 μm diameter), both of which are morphologically typical invertebrate monopolar neurons (Fig. 24.2B; Gray, 1970; Young, 1971). Twenty-five million AMs converge onto only approximately 65,000 LN. The AMs are intrinsic interneurons, as their neurites remain within the VL (Fig. 24.2B). The LN are efferent neurons whose axons form the only output of the VL, leaving from its ventral side in organized axon bundles, or roots, that are easy to identify and record from in brain slice preparations (Fig. 24.1B and Fig. 24.3).
The octopus VL receives inputs from the mSFL, which is thought to receive visual and other sensory information (Young, 1971), although this has not yet been physiologically investigated. The octopus mSFL contains 1.8 million neurons of only one morphological type, whose axons project to the VL in distinct tracts running between the VL neuropil and its outer cell body layer (Fig. 24.1). This arrangement allows each mSFL axon to make en passant synapses with many AM neurites along the VL neuropil, creating an organization like the fan-out layer of an associative matrix (Fig. 24.2). There is still no quantitative histological information on the pattern and density of connections between the mSFL axons and the AMs.
The fan-in connection of the AMs onto LNs is mediated by special serial synapses, thus named because the neurites of the AMs serve both as the postsynaptic site for the mSFL axon terminals and are presynaptic to the spines of the LNs dendrites (Fig. 24.2B; Gray, 1970). The two-layer synaptic connections of the VL fan-out, fan-in connectivity are thus confined to the AM neurites. The matrix-like organization of the mSFL–VL complex (Fig. 24.2A) is shared by the inferior frontal-subfrontal (p. 563) lobes complex (Fig. 24.1), which is involved in chemotactile learning and memory (Wells, 1978).
At least some of the outputs of the VL terminate in the subvertical lobe (SubVL), but little is known about other possible targets of the LN. Young (1971 and 1995) suggested that some LNs send their axons back into the lateral SFL, creating a recurrent loop between the VL and the SFL. Modern tracing techniques have clearly revealed such connections in cuttlefish (Graindorge, 2008). Such recurrent excitatory connections are computationally attractive, as they may create a reverberatory circuit that may subserve working memory by maintaining ongoing electrical activity, as suggested by Young (1991, 1995).
Neurophysiology of the Superior Frontal Lobe Input to the Octopus Vertical Lobe
The SFL input to the VL has been investigated using the extracellular local field potential (LFP) recordings (Fig. 24.3). Neuron activity, such as an action potential, usually generates a tiny amount of current flowing in the extracellular space. As the extracellular resistance is very low (relatively to the neurons’ membrane resistance), only summed field potentials generated by many neurons become large enough to be reliably detected against the background electrical noise. Such summation occurs only if the neurons are synchronously active and their currents flow in the same direction; otherwise, they would cancel each other out. This condition is achieved only when the neurons have long processes running in parallel and close to each other. Such an arrangement, common in the vertebrate brain, occurs in the mSFL–VL system, as the mSFL axonal input to the VL is organized as a tract. Similarly, the millions of AM cell bodies lie in the outer zone of the VL, whereas their neurites, which are organized in bundles, project into the lobe neuropil, perpendicular to the mSFL axonal tract (Fig. 24.2B). This architecture would be expected to generate a significant LFP like that generated, for example, close to the Schaffer collaterals in the hippocampus (Kandel et al., 2012).
Indeed, stimulating the SFL tract with short current pulses evokes a typical LFP: a triphasic (positive–negative–positive) tract potential (see TP in Fig. 24.3B1) generated by the volley of action potentials propagating along the stimulated axons in the SFL tract. The delay after the stimulus artifact depends on the distance between stimulus and recording electrodes. A mainly negative LFP follows immediately after the second positive wave of the TP. This potential is a glutamatergic postsynaptic field potential (see fPSP in Fig. 24.3B1), as it disappears in zero-calcium high-magnesium physiological solutions and is blocked by AMPA-like antagonists such as CNQX, DNQX, or kynurenate (Hochner et al., 2003). Because the AM cell bodies are inexcitable (i.e., they do not generate regenerative action potentials; Hochner et al., 2006), the fPSP exhibits an amplitude-independent waveform with no population spikes (Hochner et al., 2003). This differs, for example, from the fPSP recorded in the hippocampal CA1 pyramidal neurons, in which the synaptic input into the neurons can elicit action potentials as in all excitable neurons. The short latency between the peak negativity of the TP and the fPSP onset suggests a monosynaptic delay (~3 ms, Fig. 24.3B1). The physiological results thus agree well with Gray’s anatomical scheme (Fig. 24.2B; Gray, 1970), in which the terminals of the mSFL axons synapse directly on the AM neurites, suggesting that the first synaptic layer of the VL appears to be a glutamatergic synaptic input from the SFL neurons onto the AMs (synapse labeled green in Fig. 24.2B).
Recording LFPs from slices prepared from the cuttlefish VL showed similar LFP characteristics to those found in octopus (Shomrat et al., 2011). Thus, cuttlefish and octopus VL appear to have similar membrane properties and connectivity patterns despite significant differences in the overall anatomy of the two VL systems (Nixon & Young, 2003).
To investigate the nature of the AMs input to the LNs, two techniques were employed (see recording configurations at Fig. 24.3). Infrared differential interference contrast (DIC) microscopy aided intracellular recordings from LN cell bodies (Fig. 24.3B2), and extracellular recordings of spiking activity in their axon bundles (Fig. 24.3B3). Figure 24.3B2 shows excitatory postsynaptic potentials (EPSPs) in a cuttlefish LN evoked by stimulating the mSFL tract, as well as spontaneous EPSPs. Although the fine anatomical details in the cuttlefish are not as well known as in the octopus, intracellular recording from the LNs in cuttlefish VL revealed similar neurophysiological properties. Like most invertebrate neurons, the cell bodies of the LN are inexcitable, as demonstrated by the small nonovershooting spikelets (arrowheads in Fig. 24.3B2 and Fig. 24.4C). These low-amplitude spikelets result from the decrease in the amplitude of the action potential as it propagates passively along the neurite from a distant spike initiation zone (usually at the junction between the dendritic tree and the axon) to the cell body. (p. 564)
In both cuttlefish and octopus, the synaptic input to the LN is cholinergic, as both the evoked and spontaneous EPSPs are blocked by hexamethonium (Shomrat et al., 2011), a muscarinic receptor antagonist that also blocks the synaptic potential at the neuromuscular junctions of the octopus arm (Matzner et al., 2000). Hexamethonium also blocked the evoked spiking activity recorded from the LN axonal bundles. As expected, the glutamatergic fPSP of the first synaptic layer in octopus and cuttlefish was unaffected by the cholinergic antagonists. Thus, the cholinergic synapse must be the AM onto LN synapse (synapse labeled orange in Fig. 24.2B). Both cholinergic and glutamatergic antagonists blocked the LNs output as measured by recording from their axon bundles (method 3, Fig. 24.3). These findings indicate that there is no strong, direct connection from mSFL axons to the LNs and that the main connections within the VL are the mSFL inputs onto the AMs, which, in turn, innervate the LNs as shown diagrammatically in Figure 24.2A (Shomrat et al., 2011), again supporting the connectivity suggested by Gray (1970) and depicted in Figure 24.2B.
The VL systems of cuttlefish and octopus thus appear to be organized as a simple feed-forward, fan-out fan-in type of network (Fig. 24.2B and Fig. 24.5). The first fan-out synaptic layer may create high-dimensionality neural representations of the incoming sensory information in a suitable form for further processing at the fan-in layer (Shomrat et al., 2011). The two layers of fan-out fan-in network architecture, like in the VL and the mushroom body, are likely the simplest in terms of connectivity among biological association neuronal networks. Yet interestingly, it was the first connectivity architecture to be implemented in artificial networks that learned to classify inputs (Rosenblatt, 1958; Helmstaedter, 2015).
The Neuronal Output From Octopus and Cuttlefish Vertical Lobe Demonstrates Activity-Dependent Long-Term Potentiation
The input/output relationship of the VL can be relatively simply determined by stimulating the mSFL tract and measuring the VL output, either by recording intracellularly from the LN cell bodies or recording their spiking activity in the axonal (p. 565) bundles leaving the VL (Fig. 24.3). Applying four high-frequency (HF) trains to the mSFL tract (20 pulses at 50 Hz with 10 s intertrain interval) induces a robust activity-dependent LTP in the output of the VL in both octopus and cuttlefish (Fig. 24.4A1, A2 and B1, B2, respectively). Thus, the region in the coleoid brain associated with learning and memory is endowed with activity-dependent LTP, a property universally believed to be essential for networks mediating behavioral learning and memory. What are the mechanisms of this neural plasticity? Are they important for cephalopod behavior?
Synaptic Plasticity in the Octopus and Cuttlefish Vertical Lobe
Surprisingly, the synaptic plasticity discovered within the VL is located at different sites in octopus and cuttlefish (summarized in Fig. 24.5). The LFP recorded at the mSFL tract of the octopus demonstrates a very robust activity-dependent plasticity. Tetanization can lead to an LTP of an average of about fourfold increase in fPSP amplitude (as far as we know the largest activity-dependent LTP that has been described) without affecting the amplitude of the presynaptic TP (Fig. 24.3B; Fig. 24.4A1 and 24.A2). The potentiation of the glutamatergic (p. 566) synaptic input to the AMs is long term. Tetanizing a second time showed that the LTP was saturated and no further long-term enhancement was obtained. Notably, the same experiments in the cuttlefish VL revealed no activity-dependent plasticity at this synapse (Fig. 24.4B1 and 24.B2; Fig. 24.5).
Does activity-dependent plasticity occur only at the fan-out glutamatergic connections of the octopus VL? In the octopus, HF tetanization inducing LTP of the fPSP also caused a long-term increase in the extracellular spiking activity of the LN axon bundles (Fig. 24.4A1 and 24.A2). This result was not surprising: the facilitation of the synaptic input to the AMs (shown by the amplitude increase of the fPSP) should increase their cholinergic input to the LNs, thus enhancing LNs output. Therefore, this result does not reveal whether there is also synaptic plasticity at the input to the LNs. However, because the same fPSP amplitude gave rise to the same level of bundle activity irrespective of the LTP, LTP in the octopus VL must occur only at the first fan-out synaptic layer (see Shomrat et al., 2011, for details).
In contrast, in the cuttlefish, the site of synaptic plasticity mediating the activity-dependent increase in the VL output is the cholinergic synaptic input to the LNs (Fig. 24.4B1 and 24.B2; Fig. 24.5). Intracellular recordings from LN cell bodies of cuttlefish clearly show that HF stimulation of the SFL tract leads to a robust enhancement of the EPSP (Fig. 24.4C and 24.D). Whereas in the octopus the increase in VL output involves LTP of the glutamatergic connection onto the AMs (the fan-out connections), in the cuttlefish the LTP occurs at the converging or fan-in cholinergic connections of the AMs into the LNs.
What Do the Vertical Lobes of Octopus and Cuttlefish Compute?
It is possible that these differences between octopuses and cuttlefish (summarized in Fig. 24.5) are related to different behaviors yet to be systematically examined (see “The VL Is Involved in Learning and Memory”). However, computational considerations suggest a quite different possibility. The VL of both octopus and cuttlefish show a nearly linear relation between the fPSP amplitude and the integrated level of spiking activity in the LN axonal bundles; that is, the VL has a linear input/output relationship (Shomrat et al., 2011). This is most unusual in neuronal networks, since the transformation of postsynaptic potential into regenerative action potentials is by nature a nonlinear process due to the transformation of an analog (EPSP) to binary signal (action potentials). Interestingly, the inexcitability of the AMs interneurons may contribute to this uncommon property.
The linear input/output relationship has important computational consequences. In a linearly operating fan-out fan-in network, similar computation capacity is obtained regardless of whether the plasticity is localized at the fan-out or the fan-in layer (Shomrat et al., 2011). In fact, a computationally linear input/output relation at the intermediate layer (here the input/output of the AMs), which indeed occurs in both octopus and cuttlefish, is sufficient to indicate similarity in computation. If the two networks have indeed evolved to the same (p. 567) computational capacity, either via evolutionary selection or via “self-organizational” mechanisms, then computational constraints, rather than specific properties of the neurons, may determine the network properties (Shomrat et al., 2011).
Mechanisms of Long-Term Potentiation Induction in the Octopus Vertical Lobe
The discovery of activity-dependent LTP in the VL raised the question of whether the synaptic plasticity in the VL fulfills Hebb’s rule and whether this rule is universal for all associative plasticity. Hebb’s rule states that the synaptic connection between pre- and postsynaptic cells is strengthened only when both are simultaneously and sufficiently active (Hebb, 1949). The NMDA channel, whose discovery was one of the most exciting breakthroughs in modern neuroscience, shows such a coincidence-detecting gating property. Direct tests for the involvement of an NMDA-like receptor involved in the postsynaptic current (fPSP) or LTP induction gave negative results: neither APV, which blocks NMDA-like current in cephalopod chromatophore muscles (Lima et al., 2003) nor MK-801 blocked these phenomena (Hochner et al., 2003).
The crucial test for whether the octopus VL has evolved an NMDA-independent Hebbian plasticity is to completely block the postsynaptic response pharmacologically and to check whether tetanization of the presynaptic axons (mSFL tract) blocks LTP induction. Such experiments showed that the LTP in the octopus VL use both associative and nonassociative induction mechanisms (Hochner et al., 2003). In slightly fewer than half the experiments, LTP induction was largely blocked by glutamatergic antagonists; in the remaining experiments, there was hardly any blocking effect, and LTP appears to have occurred in the absence of a postsynaptic response, thus refuting the Hebbian mechanism. If the two revealed processes were thus evenly distributed among the different synaptic connections in the VL, a normal distribution of the results would have been expected. The bimodal distribution can be explained by the two types of plasticity being segregated in anatomically different regions, as in the hippocampus (CA3 vs. CA1; Kandel et al., 2012). Such differentiation has not yet been demonstrated morphologically in the octopus VL.
Characterizing associative/Hebbian-type activity-dependent synaptic plasticity requires determining whether LTP results from presynaptic increase in the amount of transmitter released or an increase in postsynaptic response or both—an issue that is still not completely resolved in the various hippocampal synapses neither in the insects mushroom body. Detailed analysis of the changes in the properties of synaptic transmission accompanying LTP induction in the VL suggests that the expression of LTP is mainly, if not exclusively, presynaptic. That is, it involves an increase in the probability of transmitter release occurring in the mSFL neuron synapses with the AMs (Hochner et al., 2003). It is therefore conceivable that this synapse, with its Hebbian type of LTP induction mechanism, involves some sort of retrograde messenger from postsynaptic neurons (e.g., the AMs).
Involvement of Nitric Oxide in Octopus Long-Term Potentiation—A Novel Molecular Mechanism
The classic example of a retrograde messenger in both invertebrate and vertebrate neuronal plasticity is nitric oxide (NO; Garthwaite, 2008; Hardingham et al., 2013). Because NO diffuses readily through lipid membrane, it is an ideal molecule to transfer the association signal from the postsynaptic cell back to the presynaptic terminals to induce LTP.
NO is most likely important in the presynaptic expression of the octopus LTP, as inhibitors of the NO synthase (NOS), like L-NNA or L-NAME (but not its inactive enantiomer D-NAME), reversed the potentiated fPSP to the amplitude before the induction of LTP (Turchetti-Maia et al., 2014). The LTP was either fully expressed again following the washout of the drug, or a tetanization was required to resume a full LTP. The reversal of LTP by NOS inhibitors also reversed the twin-pulse facilitation ratio of the fPSP (i.e., the ratio between the second and the first fPSPs in twin-pulse stimulation). This finding confirms that, indeed, the blocking effect involves reversal of a presynaptic mechanism that increases the probability of transmitter release from the presynaptic terminal of the mSFL neurons during LTP. A strong support for the presence of the NOS system specifically in the octopus VL (and SubFL) is the intense, homogeneous labeling of the VL neuropil for NADPH-diapharose, an indication of NOS activity (Hope et al., 1991; Moroz et al., 1999).
These results suggest a novel mechanism for mediating long-term plasticity: persistent activation of NOS in the postsynaptic cells (AMs?) leads to the persistent production of NO that diffuses back to the presynaptic mSFL terminals to induce the expression of LTP by facilitating transmitter (p. 568) release from the presynaptic terminals. Previous behavioral results support the involvement of NO in long-term memory, as drugs blocking NO production interfered with tactile and visual learning in the octopus (Robertson et al., 1995; Robertson et al., 1996). NO has been shown to be an important neuromodulator in other invertebrates, a well-studied example being the digestive system of gastropods (Susswein & Chiel, 2012).
Modulatory Signals in the Octopus Vertical Lobe
In addition to the activity-dependent modulation of synaptic connections in neural networks involved in learning and memory, all these networks also exhibit neural plasticity processes that are subject to up- and down-regulation by neuromodulators. This type of mechanism is important for enabling “supervised learning,” as it can facilitate or inhibit learning depending on error signals and reward contingencies. Known examples for such modulatory systems are the dopaminergic and adrenergic systems in mammals (Schultz, 2010) and octopaminergic and dopaminergic neurons in insects (Burke et al., 2012; Giurfa, 2006; Perry & Barron, 2013; Waddell, 2013).
In the octopus VL, serotonin (5-HT), a well-documented facilitatory neuromodulator in molluscs (Kandel, 2001), facilitates the glutamatergic synaptic connection between the mSFL neurons and the AMs—exactly the synaptic connection showing activity-dependent LTP (Shomrat et al., 2010). However, in comparison to the sensory-motor synapse of Aplysia, in the octopus a higher 5-HT concentration is required to achieve maximal level of presynaptic facilitation (~100 µM compared to ~10 µM). A further difference is that, in the octopus, unlike Aplysia, prolonged or repeated exposure to 5-HT does not lead to an intermediate-term facilitation of the synaptic connection between the mSFL and the AM (Antonov et al., 2010; Kandel, 2001; Shomrat et al., 2010).
Serotonergic input to the octopus VL enhances, however, the activity-dependent induction of LTP by facilitating the postsynaptic response due to 5-HT induced presynaptic facilitation (Shomrat et al., 2010). This, in turn, most likely facilitates the mechanism mediating LTP induction. In the presence of 5-HT, even a test pulse once every 10 s was sufficient to induce a slow developed LTP (Fig. 24.6A). Tetanization did not lead to further facilitation and, following 5-HT washout, tetanization was no longer effective in inducing additional LTP as the synapse was already long-term potentiated (Fig. 24.6A). Also, in the presence of 5-HT, a modest tetanic stimulus (four trains of 3 instead of 20 pulses), which would induce only partial or no LTP under control conditions, can lead to a fully saturated LTP (Shomrat et al., 2010). These findings, together with the immunohistochemical demonstration of 5-HT extensively present in processes in the VL neuropil but not in cell bodies (Shomrat et al., 2010; Shigeno & Ragsdale, 2015), suggest that 5-HT may convey modulatory signals from other lobes to the VL. This serotonergic input may thus serve to transmit reinforcement signals that facilitate, for example, the induction of a long-term association between stimuli that are temporally coupled with a punishment. This punishment (e.g., electric shock to the arms; Shomrat et al., 2008) may activate the serotonergic input to the VL (see Proposed Model). Summarizing, these results suggest that the serotonergic system in molluscs is versatile in its roles and may be adaptively implemented to achieve various modulatory functions.
Octopamine (OA), an excitatory neuromodulator in the feeding system of gastropods (Vehovszky et al., 2004; Wentzell et al., 2009), has conserved, like 5-HT, its molluscan short-term facilitatory effect in the octopus VL. Although OA (100 µM), like 5-HT, causes gradual facilitation of the fPSP responses to test pulses given every 10 s (Fig. 24.6B), this facilitation seems not to involve activation of long-term facilitation as the enhancement diminished on OA washout. Then, and in contrast to 5-HT (Fig. 24.6A), a robust LTP could be induced. This large residual LTP was induced, even though a high-frequency tetanization was applied in the presence of OA (Fig. 24.6B). The simplest conclusion is that in the presence of OA, induction of LTP is blocked. Thus, irrespective of its short-term facilitatory effect, octopamine function has adapted to a specific effect in the octopus VL, inhibiting LTP induction. 5-HT and OA thus appear to convey opposing reinforcement signals (Greenwood et al., 2009; Shomrat & Hochner, 2015).
Long-term depression can be induced in mammals by repeated low-frequency stimulation patterned (Collingridge et al., 2010). In Aplysia, the neuropeptide Phe-Met-Arg-Phe-amide (FMRFamide) induces long-term synaptic depression (Montarolo et al., 1988). Attempts to reveal similar long-term depression in octopus VL have failed (Shomrat et al., unpublished data). The effect of OA is therefore an important mechanistic addition, as it may veto the induction of LTP and (p. 569) thus may convey a signal with opposite reinforcement consequences to that of 5-HT (see Fig. 24.7 and Proposed Model).
Are the Octopus Vertical Lobe and Its Long-Term Potentiation Involved in Behavioral Learning and Memory?
The octopus VL may be the preparation of choice for investigating how neurons participate in storing memories and how memories are retrieved for the control of complex behaviors. Pioneering steps were made by Young, Boycott, Wells, and colleagues during the second half of the 20th century. Their behavioral, morphological, and lesion experiments brought them to the conclusion that the octopus VL is important for learning and memory (Wells, 1978). The discovery of activity-dependent, long-term plasticity and its neuromodulation in the VL, described earlier, provides physiological support for such a function. Now, direct experimental testing has confirmed the involvement of these physiological processes in behavioral learning and memory (Shomrat et al., 2008).
To directly test the involvement of the LTP in the VL in learning and memory, Shomrat et al. (2008) followed the experiments of Moser et al. (1998; Moser & Moser, 1999) in rat hippocampus. In these experiments, artificial saturation of LTP, induced by tetanization, impaired spatial learning when the tetanization was applied prior to learning. This technically challenging experiment in the rat is much simpler in the octopus, as the VL lies most dorsally in the brain and is relatively easily accessible to a large electrode for global tetanization of the entire VL system (the VL and the SFL lobes; Fig. 24.1). Such tetanization, when applied in an anaesthetized animal, induced on average 56% of the total available LTP (Shomrat et al., 2008).
(p. 570) Shomrat et al. (2008) then tested whether this reduction in the “available” synaptic plasticity affected a learning task given 75 min after recovery from anesthesia and tetanization. The learning task was a passive avoidance task in which a mild electric shock “taught” the octopus to stop attacking a red ball. As discussed earlier, this behavior is an innate stereotypical attack behavior. The results of LTP saturation by the tetanization were compared with the effects of transecting the mSFL tract to the VL, which disconnects the VL from its sensory input. Inducing strong LTP of the glutamatergic synaptic input into the AM may be viewed as “short-circuiting” the VL (see VL circuitry in Fig. 24.2). Neither tetanization nor transection eliminated completely the octopuses’ ability to learn the task during the training phase. Nevertheless, both affected the short-term learning, but in the opposite way. Transection slowed the learning rate relative to that of sham-operated animals (the transected animals stopped attacking the ball after around 15 trials, whereas the sham-operated animals stopped after the ninth trial). Saturating the LTP had the opposite effect; it enhanced the learning process relative to that of nontetanized animals (tetanized animals stopped attacking after the sixth trial on average).
Thus, LTP of specific synaptic connections in the VL does not appear to be involved in short-term learning itself. Instead, the VL output most likely controls short-term learning processes occurring elsewhere. These results are consistent with the LNs output having a general inhibitory effect on the circuit controlling attack behavior. In this case, cutting the input to the VL would reduce this inhibitory input to the attack motor circuit, and global LTP of the VL (i.e., “short-circuiting” the VL synapses) would enhance the VL inhibition output (see Proposed Model and Fig. 24.7).
Testing for long-term memory 24 h after training produced more straightforward results: both treatments severely impaired long-term memory (Shomrat et al., 2008). Sham-operated and control animals did not demonstrate perfect memory of the task, as approximately 70% of the animals attacked the ball on their first trial. However, they demonstrated a fast recollection as they stopped attacking the ball in the following test trials. The transected animals remembered almost nothing of the avoidance task they had learned the day before, whereas the tetanized animals showed severe impairment. These results suggest that memories acquired a few hours after tetanization or transection were not consolidated. In contrast, a memory acquired before tetanization or transection, such as to attack a crab or a white ball (associated with positive reward in pretraining), was not impaired (Shomrat et al., 2008). The results confirm earlier lesion experiments showing that previous memories were not affected by lesions of the VL (Sanders, 1975), similar to results obtained in mammals and even humans with a severed hippocampus (Stellar, 1957; Corkin, 2002).
These experiments demonstrate that the VL and its LTP are not important for the actual storage of long-term memory. Short- and long-term memory traces appear to be stored outside the VL, possibly in the circuitry mediating the attack behavior. Instead, the LTP of the specific synaptic connections in the VL enhance its output, controlling the consolidation of short-term memory into long-term memory occurring elsewhere.
A System Model for Octopus Learning and Memory
We conclude by proposing a model for the octopus learning and memory system incorporating the anatomical, physiological, and behavioral findings presented here for a passive avoidance task (Fig. 24.7). The building blocks of this descriptive yet comprehensive model are summarized in the following list.
1. Sensory inputs feed in parallel to the VL system and to the circuits controlling the attack behavior. (Experimental support: transection and tetanization did not affect the behavior. The octopuses still showed their stereotypical attack behavior.)
2. Long-term memory is stored outside the VL. (Experimental support: both tetanization and transection did not erase old memories.)
3. The output of the VL modulates the rate of short-term learning taking place outside the VL system, possibly by inhibiting the attack circuit. (Experimental support: tetanization accelerated learning [inhibited the tendency to attack] and transection slowed learning [increased the tendency to attack]. However, both treatments did not prevent the attack behavior.)
4. The LTP in the VL system is crucial for the consolidation of long-term memory outside the VL system. (Experimental support: both treatments prevented the consolidation of short-term into long-term memory.) (p. 571)
5. The probability of turning the associative activities into long-term changes depends on whether a concomitant LTP-reinforcer (5-HT) or LTP-suppressor (OA) signal is delivered to the VL. (Experimental support: physiological and immunohistochemical results.)
6. The VL output, via the LN axons, inhibits the attack behavior. (Experimental support: the LNs are GABAergic and tetanization of the VL decreases and transection increases the tendency to attack.)
7. Because the VL output inhibits the attack behavior, 5-HT is likely to signal punishment and OA to signal reward contingencies. (Experimental Support: inferred from points 5 and 6.)
Model-Based Working Hypotheses
8. Multimodal sensory information is integrated in the mSFL and likely significant features are transmitted to the VL. (Support: anatomical connections.)
9. The features of the various sensory modalities is represented sparsely in the higher dimensionality of the input layer of the VL due to the fan-out en passant innervation of the AM neurites. (Support: morphology and computational inference.)
10. Associative potentiation of certain connections to the AMs is likely to occur at the AMs that receive multimodal signals simultaneously (coincidence detection) leading to a local LTP. (Support: physiological results.)
11. The matrix of facilitated connections to the AMs converge sharply (fan-in) into a less variable matrix of LNs that are now reliably activated by either of the temporally associated multimodal sensory inputs. (Support: connectivity and computational considerations.)
How Does the Model Explain Passive Avoidance Learning?
In the passive avoidance task used by Shomrat et al. (2008), the octopus learns to refrain from attacking a red ball (actually dark, as octopuses are color-blind) by receiving an electric shock on its arms when it attacks. On sighting the target, the information on its shape and brightness is fed into the attack behavior circuitry, activating the natural attack behavior. This information is also fed, in parallel, to the mSFL. Each quality (brightness and shape) is then transferred by a different set of mSFL neurons to the VL, creating a sparse representation of each sensory quality in the matrix-like connections of the mSFL neurons with the AMs. Those AM cells receiving inputs from both qualities are more likely to undergo LTP due to their higher level of coincident activity. The “dark with round” association is reinforced if they are conjugated with the nociceptive signal conveyed to the VL by the serotonergic system. The strengthening of this association during training, in turn, creates a long-term enhancement of the AM input to the set of LNs driven by this mutual sensory representation. The VL output inhibits the tendency to attack and can be regarded as an inhibitory supervising signal. The short-term modifications in (p. 572) the attack circuitry, that are driven by the input from the VL, mediate the long-term acquisition by long-term facilitating the response of the attack circuitry to the conditioned sensory inputs.
We have shown that cephalopods are valuable animals for exploring the neural mechanisms subserving complex behaviors. Cephalopod behaviors are mediated by a nervous system uniquely organized into three separate, large components: the arm nervous system, the optic lobes, and a central brain. The cephalopod nervous system still maintains the basic, simpler features of invertebrate morphology and neurophysiology, and the division of the octopus brain into discrete interconnected lobes facilitates the analysis of connectivity. That is, as we have shown here, the complexity of cephalopod behavior is achieved by a simple organization of simple invertebrate neuronal elements. Most likely, because size and mass of the nervous system are less constraining than in terrestrial or flying animals, evolution shaped large but simpler neural networks to control complex behavior of cephalopods, and these simpler networks are more accessible for assessing how neural networks are embedded in the organization of complex behaviors. The large but simply organized fan-out, fan-in and likely feedforward network of the octopus is a vivid demonstration of this idea.
Our research is supported by the United States–Israel Binational Science Foundation, the Israel Science Foundation, the Smith Family Laboratory at The Hebrew University, the European Commission EP7 projects OCTOPUS and STIFFFLOP, and The National Institute for Psychobiology in Israel (to T.S.). We thank Prof. Jenny Kien for editorial assistance and suggestions.
Antonov, I., Kandel, E. R., & Hawkins, R. D. (2010). Presynaptic and postsynaptic mechanisms of synaptic plasticity and metaplasticity during intermediate-term memory formation in Aplysia. The Journal of Neuroscience, 30, 5781–5791.Find this resource:
Aristotle (1910). Historia animalium. In The works of Aristotle, translated into English under the editorship of J. A. Smith and W. D. Ross. Vol IV. Oxford, UK: Clarendon.Find this resource:
Boycott, B. B. (1961). The functional organization of the brain of the cuttlefish Sepia officinalis. Proceedings of the Royal Society of London B Biological Sciences, 153, 503–534.Find this resource:
Boycott, B. B., & Young, J. Z. (1955). A memory system in Octopus vulgaris Lamarck. Proceedings of the Royal Society of London B Biological Sciences, 143, 449–480.Find this resource:
Bullock, T. H., & Horridge, G. A. (1965). Structure and function in the nervous systems of invertebrates. San Francisco, CA: Freeman.Find this resource:
Burke, C. J., Huetteroth, W., Owald, D., Perisse, E., Krashes, M. J., Das, G., . . . Waddell, S. (2012). Layered reward signalling through octopamine and dopamine in Drosophila. Nature, 492, 433–437.Find this resource:
Collingridge, G. L., Peineau, S., Howland, J. G., & Wang, Y. T. (2010) Long-term depression in the CNS. Nature Reviews Neuroscience, 11, 459–473.Find this resource:
Corkin, S. (2002). What’s new with amnesic patient H.M.? Nature Reviews Neuroscience, 3, 153–160.Find this resource:
Dash, P. K., Hochner, B., & Kandel, E. R. (1990). Injection of the cAMP-responsive element into the nucleus of Aplysia sensory neurons blocks long-term facilitation. Nature, 345, 718–721.Find this resource:
Fiorito, G., & Chichery, R. (1995). Lesions of the vertical lobe impair visual discrimination learning by observation in Octopus vulgaris. Neuroscience Letters, 192, 117–120.Find this resource:
Fiorito, G., & Scotto, P. (1992). Observational learning in Octopus vulgaris. Science, 256, 545–547.Find this resource:
Garthwaite, J. (2008). Concepts of neural nitric oxide-mediated transmission. European Journal of Neuroscience, 27, 2783–802.Find this resource:
Giurfa, M. (2006). Associative learning: The instructive function of biogenic amines. Current Biology, 16, 892.Find this resource:
Graindorge, N. (2008). Vertical lobe and formation of the mnesic trace in the cuttlefish, Sepia officinalis (Doctoral dissertation). Université de Caen Normandie.Find this resource:
Graindorge, N., Alves, C., Darmaillacq, A. S., Chichery, R., Dickel, L., & Bellanger, C. (2006). Effects of dorsal and ventral vertical lobe electrolytic lesions on spatial learning and locomotor activity in Sepia officinalis. Behavioral Neuroscience, 120, 1151–1158.Find this resource:
Gray, E. G. (1970). The fine structure of the vertical lobe of octopus brain. Proceedings of the Royal Society of London B Biological Sciences, 258, 379–394.Find this resource:
Greenwood, S., Shomrat, T., & Hochner, B. (2009). The involvement octopamine in short- and long-term plasticity in the octopus vertical lobe. Program No. 890.12.2009, Neuroscience Meeting Planner. Chicago, IL: Society for Neuroscience. http://www.abstractsonline.com/Plan/ViewAbstract.aspx?sKey=c7e22b3d-3daa-4c68-8286-42aae5f8cc6a&cKey=27d53769-b018-49e8-bbab-3738e2a4a957&mKey=%7b081F7976-E4CD-4F3D-A0AF-E8387992A658%7d
Hanlon, R. T., & Messenger, J. B. (1996). Cephalopod behaviour. Cambridge, UK: Cambridge University Press.Find this resource:
Hardingham, N. R., Dachtler, J., & Fox, K. D. (2013). The role of nitric oxide in pre-synaptic plasticity and homeostasis. Frontiers in Cellular Neuroscience, 7, 190.Find this resource:
Hebb, D. O. (1949). The organization of behavior: A neuropsychological theory. New York, NY: Wiley.Find this resource:
Heisenberg, M. (2003). Mushroom body memoir: from maps to models. Nature Reviews Neuroscience, 4, 266–275.Find this resource:
Helmstaedter, M. (2015). The mutual inspirations of machine learning and neuroscience. Neuron, 86, 25–28.Find this resource:
Hochner, B., Brown, E. R., Langella, M., Shomrat, T., & Fiorito, G. (2003). A learning and memory area in the octopus brain manifests a vertebrate-like long-term potentiation. Journal of Neurophysiology, 90, 3547–3554.Find this resource:
Hochner, B., Shomrat, T., & Fiorito, G. (2006). The octopus: A model for a comparative analysis of the evolution of learning (p. 573) and memory mechanisms. The Biological Bulletin, 210, 308–317.Find this resource:
Hope, B. T., Michael, G. J., Kniegget, K. M., & Vincent, S. R. (1991). Neuronal NADPH diaphorase is a nitric oxide synthase. Proceedings of the National Academy of Sciences, 88, 2811–2814.Find this resource:
Kandel, E. R. (2001). The molecular biology of memory storage: A dialogue between genes and synapses. Science, 294, 1030–1038.Find this resource:
Kandel, E. R., Schwartz, J. H., Jessell, T. M., Siegelbaum, S., & Hudspeth, A. J. (2012). Principles of neural science (5th ed.). New York, NY: McGraw-Hill.Find this resource:
Lima, P. A., Nardi, G., & Brown, E. R. (2003). AMPA/kainate and NMDA-like glutamate receptors at the chromatophore neuromuscular junction of the squid: Role in synaptic transmission and skin patterning. European Journal of Neuroscience, 17, 507–516.Find this resource:
Maldonado, H. (1963). The visual attack learning system in Octopus vulgaris. Journal of Theoretical Biology, 5, 470–488.Find this resource:
Maldonado, H. (1964). The control of attack by Octopus. Zeitschrift fur vergleichende Physiologie, 47, 656–674.Find this resource:
Matzner, H., Gutfreund, Y., & Hochner, B. (2000). Neuromuscular system of the flexible arm of the octopus: Physiological characterization. Journal of Neurophysiology, 83, 1315–1328.Find this resource:
McNaughton, B. L., Battaglia, F. P., Jensen, O., Moser, E. I., & Moser, M. B. (2006). Path integration and the neural basis of the “cognitive map.” Nature Reviews Neuroscience, 7, 663–678.Find this resource:
Montarolo, P. G., Kandel, E. R., & Schacher, S. (1988). Long-term heterosynaptic inhibition in Aplysia. Nature, 333, 171–174.Find this resource:
Moroz, L. L., Gillette, R., & Sweedler, J. V. (1999). Single-cell analyses of nitrergic neurons in simple nervous systems. Journal of Experimental Biology, 202, 333–341.Find this resource:
Moser, E. I., Krobert, K. A., Moser, M. B., & Morris, R. G. (1998). Impaired spatial learning after saturation of long-term potentiation. Science, 281, 2038–2042.Find this resource:
Moser, E. I., & Moser, M. B. (1999). Is learning blocked by saturation of synaptic weights in the hippocampus? Neuroscience & Biobehavioral Reviews, 23, 661–672.Find this resource:
Nixon, M., & Young, J. Z. (2003). The brain and lives of cephalopods. Oxford, UK: Oxford University Press.Find this resource:
Packard, A. (1972). Cephalopods and fish: The limits of convergence. Biological Reviews, 47, 241–307.Find this resource:
Perry, C. J., & Barron, A. B. (2013). Neural mechanisms of reward in insects. Annual Review of Entomology, 58, 543–562.Find this resource:
Robertson, J. D., Bonaventura, J., & Kohm, A. (1995). Nitric-oxide synthase inhibition blocks octopus touch learning without producing sensory or motor dysfuntion. Proceedings of The Royal Society of London Series B-Biological Sciences, 261, 167–172.Find this resource:
Robertson, J. D., Bonaventura, J., Kohm, A., & Hiscat, M. (1996). Nitric oxide is necessary for visual learning in Octopus vulgaris. Proceedings of The Royal Society of London Series B-Biological Sciences, 263, 1739–1743.Find this resource:
Rose, S. P. R. (2003). The making of memory: From molecules to mind. New York, NY: Anchor Books.Find this resource:
Rosenblatt, F. (1958). The perceptron: A probabilistic model for information storage and organization in the brain. Psychological Review, 65, 386–408.Find this resource:
Sanders, F. K., & Young, J. Z. (1940). Learning and other functions of the higher nervous centres of Sepia. Journal of Neurophysiology, 3, 501–526.Find this resource:
Sanders, G. D. (1975). The cephalopods. In Invertebrate learning, vol. 3. New York, NY: Plenum Press.Find this resource:
Schultz, W. (2010). Dopamine signals for reward value and risk: Basic and recent data. Behavioral and Brain Functions, 6, 24.Find this resource:
Shigeno, S., & Ragsdale, C. W. (2015). The gyri of the octopus vertical lobe have distinct neurochemical identities. Journal of Comparative Neurology, 523, 1297–12317.Find this resource:
Shomrat, T., Feinstein, N., Klein, M., & Hochner, B. (2010). Serotonin is a facilitatory neuromodulator of synaptic transmission and “reinforces” long-term potentiation induction in the vertical lobe of Octopus vulgaris. Neuroscience, 169, 52–64.Find this resource:
Shomrat, T., Graindorge, N., Bellanger, C., Fiorito, G., Loewenstein, Y., & Hochner, B (2011). Alternative sites of synaptic plasticity in two homologous “fan-out fan-in” learning and memory networks. Current Biology, 21, 1773–1782.Find this resource:
Shomrat, T., & Hochner, B. (2015). Serotonin may convey positive and octopamine negative reinforcement signals to the learning network of Octopus vulgaris. Program No. 629.06. 2015 Neuroscience Meeting Planner. Chicago, IL: Society for Neuroscience 2015.Find this resource:
Shomrat, T., Zarella, I., Fiorito, G., & Hochner, B. (2008). The octopus vertical lobe modulates short-term learning rate and uses LTP to acquire long-term memory. Current Biology, 18, 337–342.Find this resource:
Stellar, E. (1957). Physiological psychology. Annual Review of Psychology, 8, 415–436.Find this resource:
Strausfeld, N. J., Sinakevitch, I., Brown, S. M., & Farris, S. M. (2009). Ground plan of the insect mushroom body: Functional and evolutionary implications. Journal of Comparative Neurology, 513, 265–291.Find this resource:
Susswein, A. J., & Chiel, H. J. (2012). Nitric oxide as a regulator of behavior: New ideas from Aplysia feeding. Progress in Neurobiology, 97, 304–317.Find this resource:
Turchetti-Maia, A. L., Shomrat, T., & Hochner, B. (2014). Nitric oxide synthase is involved in maintenance but not in induction of activity-dependent LTP in the vertical lobe of the octopus. Program No. 561.04. 2014 Neuroscience Meeting Planner. Washington, DC: Society for Neuroscience 2014.Find this resource:
Vehovszky, A., Szabó, H., & Elliott, C. J. (2004). Octopamine-containing (OC) interneurons enhance central pattern generator activity in sucrose-induced feeding in the snail Lymnaea. Journal of Comparative Physiology A, 190, 837–846.Find this resource:
Waddell, S. (2013) Reinforcement signalling in Drosophila; dopamine does it all after all. Current Opinion in Neurobiology, 23, 324–329.Find this resource:
Wells, M. J. (1978). Octopus. London: Chapman and Hall.Find this resource:
Wentzell, M. M., Martínez-Rubio, C., Miller, M. W., & Murphy, A. D. (2009). Comparative neurobiology of feeding in the opisthobranch sea slug, Aplysia, and the pulmonate Snail, Helisoma: Evolutionary considerations. Brain, Behavior and Evolution, 74, 219–230.Find this resource:
Young, J. Z. (1971). The anatomy of the nervous system of Octopus vulgaris. Oxford, UK: Clarendon Press.Find this resource:
Young, J. Z. (1991). Computation in the learning-system of cephalopods. Biological Bulletin, 180, 200–208.Find this resource:
Young, J. Z. (1995). Multiple matrices in the memory system of octopus. In Cephalopod neurobiology (pp. 431–443). Oxford, UK: Oxford University Press.Find this resource:
Zullo, L., Sumbre, G., Agnisola, C., Flash, T., & Hochner, B. (2009). Nonsomatotopic organization of the higher motor centers in octopus. Current Biology, 19, 1632–1636. (p. 574) Find this resource: