Neuromodulation in invertebrate sensory systems: from biophysics to behavior

Animals respond to a wide variety of sensory stimuli on a daily basis. Moreover, an identical stimulus may elicit distinct reactions under different environmental or behavioral conditions. Since organisms have a finite number of neural circuits, neurons and networks must be multifunctional. Neuromodulation provides one powerful means to dramatically but reversibly reconfigure the function of a sensory circuit without changing the`hard-wiring'.

A neuromodulatory effect typically begins with the binding of a peptide or other small molecule to a metabotropic receptor. This triggers a cascade of biochemical reactions that ultimately changes the physiology of the cell and can elicit much more complex effects than the simple excitation or inhibition of classical neurotransmission. These may include modification of a neuron's membrane resistance, firing rate or bursting properties(Combes et al., 1997), the dynamics of adaptation (Zhang et al.,1992), the strength of its synaptic outputs, or even the shape of the action potential itself (Dunlap and Fischbach, 1978). Excellent general reviews of neuromodulation can be found in the literature (Kupfermann,1979; Kaczmarek and Levitan,1987; Lopez and Brown,1992; Katz and Frost,1996).

An understanding of the functional significance of neuromodulation may emerge from correlations of the cellular effects of a neuromodulator with the behavioral or physiological changes it induces in an animal. To such an end,research in invertebrate sensory systems is particularly advantageous. Obviously, invertebrate nervous systems are simpler than those of vertebrates. They have fewer cells that are generally larger and easier to maintain in vitro. In many species neurons are readily identifiable, and their synaptic connections have been thoroughly mapped. Individual invertebrate neurons can have significant or even unique roles in determining behavior, and typically only a few layers of neural processing separate sensory input from motor output. Finally, the simplicity inherent in invertebrate systems facilitates detailed mathematical modeling at both the cellular and network level.

One common effect of neuromodulation is to increase sensitivity of receptor neurons to their particular stimulus. In other words, neuromodulation can decrease the threshold stimulus required for generating both action potentials and behavioral responses. For example, in the male silkworm moth(Antheraea polyphemus) octopamine enhances the sensitivity of neurons that detect pheromones (Pophof,2000). This change in receptor sensitivity correlates with the increased behavioral sensitivity seen in males of several other moth species after octopamine injection (Linn and Roelofs, 1986; Linn et al., 1992, 1996). Male cabbage looper moths (Trichoplusia ni), for example, detect pheromones at a concentration two orders of magnitude lower than before injection of the modulator (Linn and Roelofs,1986). Octopamine presumably improves the ability of males to follow the odor plumes of sex pheromones emitted by females.

Understanding the function of a modulator is not always so easily interpretable. In addition to identification of the biophysical mechanisms involved, knowledge of the role neuromodulation plays in network activity may be required (Mercer, 1999). Below we discuss several examples of neuromodulation in invertebrate sensory systems that have been studied on a number of levels, from the effects on ion channels all the way up to behavior. The role of neuromodulation in several other well-studied invertebrate sensory systems has previously been reviewed(Pasztor, 1989).

Probably the best-known example of invertebrate sensory modulation involves the gill-withdrawal reflex of the sea slug Aplysia californica, a system that has been studied extensively at the behavioral, network, cellular and biophysical levels (for reviews and specific references, see Byrne and Kandel, 1996; Kandel, 2001). It is also one of the very few invertebrate sensory systems in which neuromodulation has been modeled mathematically (Baxter et al.,1999). In a classic series of experiments, Kandel and coworkers discovered that a weak tactile stimulus elicits a reflexive withdrawal of the siphon and gill. Although the response quickly habituates to repeated stimuli,a single noxious stimulus sensitizes the animal, temporarily reversing the habituation. Serotonin released by interneurons(Marinesco and Carew, 2002)mediates sensitization by modulating transmitter release at the sensory-motor synapse. Binding of the modulator to its receptor leads to the phosphorylation and subsequent inhibition of three K⁺ currents: a voltage-independent K⁺ current(Klein and Kandel, 1980), a voltage-dependent K⁺ current similar to the delayed rectifier(Baxter and Byrne, 1989, 1990) and a calcium-activated K⁺ current (Walsh and Byrne,1989). The resulting increase in membrane resistance broadens presynaptic action potentials, increasing Ca²⁺ influx. The subsequent enhanced release of neurotransmitter reverses the habituation of the reflex. Moreover, a series of noxious stimuli induces longer lasting sensitization that persists for days or weeks. In this case, serotonin promotes gene activation and the synthesis of new proteins that induce the growth of new synapses (Liu et al.,1997; Zhang et al.,1997; Müller and Carew,1998; reviewed in Kandel,2001).

The firing rate of a sensory neuron typically increases with stimulus intensity. Although `gain modulation' refers to modifications of this relationship, there is no clear consensus on the term `gain'. It is reasonable to define the gain of the response as the slope of the function relating firing rate to stimulus intensity (see Fig. 1A). Higher gain results in a larger differential response for a given change in stimulus. However, it also tends to limit the range over which a neuron can encode stimuli, since firing rates themselves are limited by biophysical properties of membranes (see Fig. 1B).

The tetrapeptide FMRFamide affects the gain of neurons that maintain osmotic balance in the medicinal leech Hirudo medicinalis(Wenning and Calabrese, 1995). Ordinarily, fresh water leeches minimize solute loss by reabsorbing ions from the urine. However, after a blood meal the concentration of extracellular chloride can triple (Zerbst-Boroffka et al., 1997). Under these conditions salt reabsorption across the epithelia of urine-forming nephridia decreases, allowing the animal to excrete the excess solute. The nephridial nerve cells (NNCs) mediate this physiological change. They monitor extracellular chloride concentration and secrete FMRFamide to regulate the activity of both the urine-forming cells as well as the NNCs themselves (Wenning et al., 1993, 2001).

By modulating its own release, FMRFamide adjusts the chloride conductance and hence the gain of the NNC. Between meals chloride efflux depolarizes the cells, causing them to fire periodic bursts of action potentials that convey information about chloride concentration to the central nervous system(Wenning, 1989). This bursting also releases both a neurotransmitter and FMRFamide from neurosecretory terminals innervating the nephridia(Wenning et al., 1993). In addition to its effects on urine formation, the peptide inhibits its own release and partially inactivates the resting outward chloride conductance,thereby reducing the NNC chloride receptor gain(Wenning and Calabrese, 1995). In contrast, the high extracellular chloride concentration present after a blood meal inhibits the same chloride conductance as FMRFamide. As the NNCs hyperpolarize toward the potassium equilibrium potential, peptide release stops altogether (Wenning and Calabrese,1995). Under these conditions, the chloride conductance depends solely on the concentration of extracellular chloride. This makes the NNCs more sensitive to chloride and primes them to monitor the return to normal extracellular levels of the ion. As the chloride concentration decreases to the normal steady state levels, chloride channels in the NNCs re-open,allowing chloride efflux to resume. The resulting membrane depolarization initiates the normal periodic bursting that releases FMRFamide.

GABA and serotonin modulate phototaxis in the marine mollusc Hermissenda crassicornis. When the animal encounters a shadow in an otherwise uniformly lit field, it turns toward the light. However, repeatedly pairing light with turbulence suppresses the phototactic response(Crow and Alkon, 1978; Farley and Han, 1997). The underlying neural circuits can be activated behaviorally or in vitro. The cellular effects of classical conditioning can be observed even in individual cells isolated from a conditioned animal(Frysztak and Crow, 1997; Gandhi and Matzel, 2000). The neural wiring mediating phototaxis begins with type A photoreceptors. Acting through interneurons, these sensory cells excite motor neurons in the pedal ganglia, causing the animal to turn into the light(Goh and Alkon, 1984; Crow and Tian, 2000). Type B photoreceptors suppress the behavior by inhibiting type A photoreceptors(Goh and Alkon, 1984). Furthermore, vestibular hair cells excite type B photoreceptors, thereby facilitating their inhibitory effect on type A photoreceptors(Schultz and Clark, 1997).

Sensory information from photoreceptors and mechanoreceptors converges on type B cells to elicit phototactic behavior. Both GABA and serotonin act through the same biochemical pathway in these neurons(Schuman and Clark, 1994; Schultz and Clark, 1997). GABA, released by hair cells, binds to metabotropic receptors on the type B photoreceptors, stimulating phospholipase A₂ to release arachidonic acid and thereby activate protein kinase C (PKC)(Muzzio et al., 2001). Serotonin also modulates the function of type B photoreceptors via a PKC-dependent mechanism (Schuman and Clark, 1994; Frysztak and Crow, 1997). The injection of PKC into B cells and the activation of endogenous PKC both increase membrane excitability by phosphorylating potassium channels (Farley and Auerbach,1986; Alkon et al.,1988). The subsequent reduction in K⁺ conductance lengthens action potentials while decreasing the amplitude of the after-hyperpolarization (Farley and Auerbach, 1986; Matzel et al.,1992; Gandhi and Matzel,2000). In addition, serotonin augments a hyperpolarization-activated inward current that is active at the resting potential in these cells (Acosta-Urquidi and Crow, 1993). All of these mechanisms prolong the presynaptic depolarization of B cells, promoting increased transmitter release and facilitation of postsynaptic inhibition of type A photoreceptors.

Octopamine is often considered a `stress' hormone in insects, most closely analogous to norepinephrine in vertebrates(Adamo et al., 1995; Roeder, 1999). The hemolymph concentration of octopamine in the locust Schistocera gregaria, for example, can increase up to eightfold during exposure to noxious stimuli(Orchard et al., 1981; Davenport and Evans, 1984). In the locust, octopamine modulates the responses of sensory neurons of the femoral chordotonal organ (feCO). This sensory organ is a multineuronal proprioceptor at the femur-tibia joint that mediates reflexes important for posture and movement (Usherwood et al.,1968; Field and Rind,1981; Field and Matheson,1998). Neurons in the feCO preferentially respond to joint position, to the direction, velocity or acceleration of movement, or to a combination of these stimuli (Matheson,1990). The cells make excitatory synaptic connections with flexor and extensor motor neurons as well as with local and intersegmental interneurons (Burrows, 1987). Each feCO sensory-motor terminal is presynaptically inhibited by interneurons activated by other feCO neurons (Burrows and Laurent, 1993; Burrows and Matheson, 1994). This network may provide a gain control mechanism that allows the reflex properties of particular proprioceptive neurons to be modified independently of each other(Burrows and Matheson, 1994; Matheson, 1997).

One of the most common types of feCO neurons generates spike trains with a tonic component that reflects joint position, as well as a phasic component that describes movement of the tibia. Application of physiologically relevant concentrations of octopamine expands the range of tonic firing frequencies without affecting phasic responses(Matheson, 1997). This would suggest that octopamine might strengthen the reflex generated by these cells. However, octopamine also indirectly increases tonic inhibition at the sensory-motor terminal due to its enhancement of the firing rates of other neurons (Matheson, 1997). Since it augments both the excitability and the presynaptic inhibition of feCO neurons, the net effect of octopamine on the reflex is unclear. The consequence for any particular feCO input to the motor system would depend on the balance between the two neuromodulatory effects and could be stimulus-dependent. Determining the physiological importance of octopamine in this system will require a detailed understanding of the interactions between particular feCO neurons in the intact neural circuit.

Serotonin and octopamine have opposite effects on a visually evoked reflex in the honeybee Apis mellifera. Antagonistic effects of octopamine and serotonin on neural and behavioral activity have also been reported in other invertebrate systems (Livingstone et al., 1980; Claassen and Kammer, 1986; Linn and Roelofs, 1986; Bicker and Menzel, 1989). During flight, antennae of the honeybee are held in an outstretched position (Heran,1959) until just before landing when they are reflexively pushed forward, presumably to facilitate olfaction(Erber, 1984). Presentation of vertically moving black and white stripe patterns elicits similar antennal movements from restrained bees (Erber and Schildberger, 1980; Erber et al., 1993). Injecting serotonin into the region of the optic lobe containing dendritic terminals of motion-sensitive neurons(Hertel and Maronde, 1987; Hertel et al., 1987) inhibits the response (Erber and Kloppenburg,1995); in contrast, octopamine injection slightly enhances it(Erber and Kloppenburg, 1995). Analogously, serotonin decreases the amplitudes of field potentials recorded in the optic lobe while octopamine increases them(Kloppenburg and Erber, 1995). Related modulatory effects are also observed on the evoked responses of individual optic lobe neurons that process visual information(Kloppenburg and Erber,1995).

Octopaminergic and serotonergic processes branch throughout the bee's visual system and brain, suggesting that the amines modulate various functions of the central nervous system(Schürmann and Klemm,1984; Kreissl et al.,1994; Bicker, 1999)including olfaction and memory (Mercer and Menzel, 1982; Bicker and Menzel, 1989; Bicker,1999). Octopamine and serotonin are released in the CNS under different circumstances. Octopamine levels appear to increase during food arousal (Braun and Bicker,1992; Bicker,1999). This effect of food is supported by the observation that the injection of octopamine, but not dopamine or serotonin, into the antennal lobe of the bee (Hildebrandt and Müller, 1995a) increases protein kinase A (PKA) levels in the same manner as sucrose stimulation(Hildebrandt and Müller,1995b). On the other hand, serotonergic modulation of various insect sensory systems is subject to circadian control. For example, serotonin application increases photoreceptor sensitivity by modifying potassium channel kinetics, mimicking the increased sensitivity seen at night(Cuttle et al., 1995; Hevers and Hardie, 1995.)Moreover, light suppresses the activity of serotonin-immunoreactive neurons in the optic lobe of the butterfly Papilla xuthus(Ichikawa, 1994), and serotonin levels in the sphinx moth Manduca sexta peak around dawn and dusk (Kloppenburg et al.,1999). The multiple effects of amines in the bee CNS are probably mediated by a number of different receptors. The pharmacological properties of octopamine receptors in the bee's mushroom bodies differ from those in other parts of the brain (Erber et al.,1993), for example. A more complete identification and understanding of neuromodulatory effects in the sensory CNS will require continued progress in the application of molecular biological techniques(Maleszka, 2000; Blenau and Baumann, 2001).

Neuromodulators may endow neural networks with such a rich functional repertoire that unraveling the complexity can be a daunting task, even in a relatively simple invertebrate nervous system. Studies of neuromodulation have typically focused either on behavior or on the function of individual components of the nervous system - neurons, synapses, channels, etc. However,we cannot hope to understand fully the physiological significance of neuromodulation without exploring the possibility of multiple, interacting modulatory effects. The success of such investigations will be contingent on the refinement of biophysical, biochemical and molecular biological techniques. Moreover, the application of underutilized computational tools will suggest new lines of research and promote a deeper understanding of neuromodulation (Fellous and Linster,1998). With their relatively simple neural circuits and large,identifiable cells, invertebrate sensory systems are ideally suited for such investigations.

We thank Larry Abbott for insightful comments regarding gain modulation. This work was supported by an award from Research Corporation (to J.T.B.).