Regulation of response properties and operating range of the AFD thermosensory neurons by cGMP signaling

Animals lacking the AFD neurons, or mutant for the tax-4 cyclic nucleotide-gated gene, fail to initiate and maintain isothermal tracks

Mutations in the tax-4 cyclic nucleotide-gated (CNG) channel have been shown to abolish AFD thermosensitivity and IT [6, 7, 13, 15]. We first determined whether the inability of tax-4 mutants to track isotherms arises from defects in initiation and/or maintenance of IT behavior. Animals were grown at a T_c of 15°C or 20°C and placed on a 0.9°C/cm linear thermal gradient that includes the temperatures at which IT behavior is expected. We quantified the numbers of tracks perpendicular to the direction of the gradient to obtain a measure of initiated isothermal tracks, as well as isothermal track lengths to determine maintenance of IT behavior. Under our experimental conditions, the average length of an isothermal track on a plate in the absence of an imposed thermal gradient is ~3.6 mm (see Supplemental Experimental Procedures). Quantification of isothermal track numbers that were longer than 3.6 mm on a thermal gradient showed that wild-type animals initiated more tracks in a temperature band near T_c (Figure 1D). Moreover, in the presence of a gradient, wild-type animals maintained movement along isotherms near T_c as evident from longer isothermal track lengths in this temperature band (Figure 1D).

In contrast, we found that the isothermal tracks initiated by tax-4(p678) mutants were evenly distributed across the gradient (Figure 1E). Moreover, tax-4 mutants initiated fewer of these tracks overall (Figure 1G), and exhibited only rare longer tracks (Figure 1E). Since IT behavior requires forward movement and suppression of abrupt reorientation events or turns [8, 9], we confirmed that the tracking defects of tax-4 mutants were not due to increased basal turning rates or changes in forward movement velocity (Figure S1). Consistent with a major role of the AFD neurons in regulating IT behavior, animals in which the AFD neurons had been genetically ablated (generous gift from M. Goodman, Stanford University) also initiated significantly fewer isothermal tracks (Figure 1F, 1G) that were not maintained (Figure 1F). These results indicate that the AFD neurons and TAX-4 CNG channel are required to both initiate and maintain isothermal tracks.

The GCY-8 rGC is required for isothermal tracking behavior

We next determined whether one or more of the AFD-expressed GCY-8, GCY-18 or GCY-23 rGCs regulate IT behavior. We first confirmed that all examined gcy mutants exhibited normal locomotion and turning rates (Figure S1). We found that animals carrying two independent alleles of gcy-8 (Figure S2), exhibited significant defects in IT behavior. gcy-8(oy44), and to a lesser extent, gcy-8(tm949) mutants grown at either 15°C or 20°C showed defects in isothermal track maintenance as compared to wild-type animals (Figures 2A–C, 3A–C). These mutants also exhibited significantly fewer overall isothermal track numbers at both temperatures, with gcy-8(tm949) mutants exhibiting weaker defects than gcy-8(oy44) animals particularly at 20°C (Figures 2I, 3I). The tracking defects of gcy-8(oy44) mutants were partially rescued upon expression of low copy numbers of a cosmid containing wild-type gcy-8 genomic sequences in one of two transgenic lines examined (Figure 2D, 2I, 3D, 3I). gcy-8(oy44)/gcy-8(tm949) heterozygous animals also exhibited significant IT defects (data not shown).

In contrast, gcy-18 mutants exhibited similar isothermal track numbers and lengths as wild-type animals when grown at either 15°C or 20°C (Figures 2E, 2I, 3E, 3I). gcy-23 mutants grown at 15°C but not at 20°C initiated a significantly larger number of isothermal tracks on a thermal gradient (Figures 2F, 2I, 3F, 3I), and also exhibited longer tracks (Figure 2F). However, the distribution of tracks exhibited by both gcy-18 and gcy-23 mutants on the gradient was distinct from that of wild-type animals (discussed further below). The phenotypes of gcy-23 gcy-18 double mutant animals resembled those of gcy-18 mutants alone at both examined temperatures (Figures 2G, 2I, 3G, 3I), whereas gcy-23 gcy-8 gcy-18 triple mutant animals exhibited strong defects in IT behavior at both temperatures resulting from defects in both initiation and maintenance of isothermal tracks (Figures 2H, 2I, 3H, 3I). These results indicate that GCY-8 regulates the ability of the AFD neurons to execute IT behavior.

GCY-18 and GCY-23 rGCs regulate the temperature range of tracking

The temperature range in which animals track isotherms reflects the animals’ memory of T_c [6, 7, 10]. Mutants that are able to respond to oscillating temperature stimuli and track isotherms may nevertheless do so in a different temperature range if their T_c memory is distinct from that of wild-type animals. Since the distribution of isothermal tracks exhibited by gcy-18 and gcy-23 mutants appeared to be different from that of wild-type animals, we quantified the temperature range in which these gcy mutants exhibited significant IT ability. We found that the distribution range of isothermal tracks in gcy-18 mutants was shifted towards colder temperatures at both T_c=15°C and T_c=20°C, with the shift being larger at T_c=20°C (Figures 2J, 3J). However, while the distribution of isothermal tracks in gcy-23 mutants was also shifted toward colder temperatures at T_c=20°C (Figure 3J), the range was shifted to warmer temperatures at T_c=15°C (Figure 2J). The range in gcy-23 gcy-18 double mutants was not significantly different from that of gcy-18 single mutants at T_c=15°C (Figure 2J), but showed an additive effect at T_c=20°C (Figure 3J). These results imply that while GCY-18 and GCY-23 are not essential for the execution of IT behavior, these rGCs play important roles in setting T_c memory.

Mutations in gcy-8 affect the ability of the AFD neurons to respond to oscillating thermal stimuli

IT behavior requires that animals be able to sense oscillating temperature stimuli around a reference temperature; the amplitude and frequency of the oscillating stimulus are dependent on the steepness of the gradient and the speed of the animals’ sinusoidal movement [8, 9, 19]. The thermosensory responses of the AFD neurons have previously been shown to be bidirectional and adaptive [12, 14, 15], allowing animals to respond to the oscillatory temperature changes that the animal experiences while navigating a linear thermal gradient. We hypothesized that the failure of gcy-8 mutants to track isotherms on steep gradients could arise from defects in the ability of the AFD neurons to respond to these rapidly oscillating temperature changes. To address this possibility we expressed the genetically encoded ratiometric Ca²⁺ sensor YC3.60 cameleon [20, 21] in the AFD neurons under the gcy-8 promoter, and examined intracellular Ca²⁺ dynamics in response to sinusoidal variations in temperature.

We examined Ca²⁺ dynamics in response to oscillatory stimuli of 0.04 Hz frequency and 0.2°C amplitude centered around ~16.1°C and ~18.6°C for animals grown at 15°C and 20°C, respectively (Figure 4A). These stimuli correspond to temporal temperature variations of approximately 0.008°C/sec, which is within the range of sensitivity of temporal sensory processing required for isothermal tracking [8]. The reference mean temperatures were chosen since they were above the corresponding T^*_AFD, and within the temperature range at which IT behavior is exhibited. Most, if not all examined wild-type AFD neurons (90–100%; n=10 each) responded robustly to this stimulus at both temperatures, and exhibited intracellular Ca²⁺ dynamics that were phase-locked onto the provided stimulus (Figure 4B, Table 1; Figure S3A). Since the ability of the AFD neurons to track oscillatory thermal stimuli is likely associated with their ability to direct IT behavior [12] we elected to quantify the number of AFD neurons exhibiting oscillatory responses in mutants regardless of response amplitude. In contrast to wild-type animals, only 10–20% (n=10 each) of the examined AFD neurons in gcy-8(oy44) and gcy-8(tm949) animals responded to the stimulus in animals grown at 15°C. Only 20% of the AFD neurons in gcy-8 mutants also responded to oscillating stimuli of a larger amplitude of 0.5°C with a similar stimulus frequency (Figure S3B; Table 1). Whereas 50% of the AFD neurons in gcy-8(oy44) mutant animals responded when grown at 20°C, all AFD neurons of gcy-8(tm949) mutant animals responded at 20°C (Figures 4C, 4D, Table 1) consistent with weaker isothermal tracking behavioral defects exhibited by gcy-8(tm949) mutants at this temperature (Figure 3). We verified that the observed defects were due to mutations in gcy-8 by confirming that the defects in AFD Ca²⁺ dynamics in gcy-8(tm949) at T_c=15°C mutants could be rescued upon expression of wild-type gcy-8 sequences (Figure 4E).

Both gcy-18 and gcy-23 mutants retained the ability to respond to the oscillating stimuli with 0.2°C amplitude at either T_c examined (Figures 4F, 4G, Table 1) consistent with their ability to track isotherms when grown at these temperatures. We noted that a subset of the neurons in gcy-23 mutants appeared to initiate responses later for reasons that are presently unclear (Figure 4G, Table 1). The majority of AFD neurons in gcy-23 gcy-18 mutants also exhibited Ca²⁺ responses, with a small subset exhibiting delays in response initiation (Figure 4H, Table 1). Together, these results indicate that mutations in gcy-8, but not in gcy-18 and gcy-23, affect the ability of the AFD neurons to respond to oscillating temperature stimuli centered around a mean temperature.

cGMP levels regulate setting of the cultivation temperature memory

As indicated above, while gcy-18 and gcy-23 mutants retained the ability to track isotherms, they did so in a distinct temperature range, suggesting that their T_c memory was altered. To determine whether altered T_c memory correlated with changes in their T^*_AFD, we provided animals with thermal oscillations superimposed on warming ramps and quantified the average temperature at which intracellular Ca²⁺ transients are first observed (Figures 1B, 1C). Wild-type AFD neurons showed onset of Ca²⁺ changes at an average of 13.8°C or 18°C in animals grown at 15°C or 20°C, respectively (Figures 5A, 5B, Figure S4) [10, 12, 13]. Consistent with a larger downward shift in T_c memory of gcy-18 mutants grown at 20°C, the T^*_AFD in their AFD neurons was also significantly lower, although no significant changes were observed at T_c=15°C (Figures 5A, 5B). However, T^*_AFD of gcy-23 mutants was not altered at either temperature (Figures 5A, 5B) suggesting that the altered T_c in these animals may reflect changes in the threshold of AFD synaptic output but not in its sensory response [10]. gcy-23 gcy-18 double mutants exhibited T^*_AFD that was significantly lower than that exhibited by gcy-18 or gcy-23 mutants alone at T_c=20°C, reflecting the markedly lower T_c memory at this temperature as measured behaviorally (Figure 5B).

Although defects in tracking behavior precluded accurate measurements of T_c memory via behavioral assays in gcy-8 mutants, the AFD neurons in these animals responded robustly to oscillating stimuli superimposed on an upward linear ramp (Figure S4) allowing quantification of T^*_AFD. Interestingly, T^*_AFD of gcy-8 mutants was also lower than wild-type T^*_AFD in animals grown at either temperature (Figures 5A, 5B), indicating that mutations in AFD-expressed rGC genes regulate the range of IT behavior in part via regulation of T^*_AFD.

Since mutations in gcy genes are expected to alter intracellular cGMP levels, we next quantified T^*_AFD in wild-type animals grown in the presence of the non-hydrolyzable cGMP analog 8-Br-cGMP. Addition of this analog to the cultivation plates raised T^*_AFD (Figure 5A). We could not measure their T_c memory behaviorally due to general deficits in locomotory behavior in the presence of this analog (data not shown). Previous observations have also suggested that increased levels of cGMP may shift T_c memory towards higher temperatures [16]. Together, these observations strongly suggest that alterations in intracellular cGMP signaling via mutations in rGC genes or by the addition of cGMP analogs alter T_c memory at least in part via altering the threshold of AFD responsiveness.

cGMP levels set the upper bound of the response range of the AFD neurons to oscillating temperature stimuli

The lower bound of the temperature range in which animals exhibit IT behavior is partly set by T^*_AFD [12, 13], which in turn appears to be regulated by intracellular cGMP signaling. However, how the upper bound is set is unclear. It is possible that the AFD neurons are able to respond to oscillating temperature stimuli only within the defined temperature range in which IT behavior is exhibited, although this possibility has not been tested experimentally. We hypothesized that the upper bound of the range in which IT behavior is exhibited at a given T_c is also set by intracellular cGMP signaling, thereby defining an operating range of the AFD neurons to this stimulus.

To test this notion, we examined the Ca²⁺ responses of wild-type AFD neurons to an oscillating stimulus of 0.04 Hz frequency and 0.2°C amplitude centered around ~13.9°C, ~14.5°C and ~18.6°C for animals grown at 15°C and compared these responses to their responses at 16°C (Figures 5C–E, Table 2). Only 20–30% of wild-type AFD neurons exhibited responses at either the low or the high temperature extremes, compared to a 100% response rate at 14.5°C or 16°C (Figures 5C–E, compare with Figure 4B, Table 2), providing a first indication of a defined operating response range of the AFD neurons to this stimulus. The T^*_AFD of gcy-8 mutants is lower than that of wild-type; we thus considered the possibility that the operating range of gcy-8 mutants may be shifted towards lower temperatures. However, gcy-8(oy44) mutants continued to exhibit significant defects in response to an oscillating stimulus centered around 13.9°C (Figure 5C). Since addition of 8-Br-cGMP raised T^*_AFD (Figure 5A), we next determined whether the upper bound of the response was also shifted concomitantly. We found that growth of wild-type animals in the presence of 8-Br-cGMP resulted in a more robust response at the higher temperature with 80% of neurons responding albeit with alterations in phase (Figure 5E, Table 2). These results imply that intracellular cGMP signaling sets both the upper and lower bound of responsiveness to oscillating stimuli, thereby defining the temperature range for IT behavior.

Individual rGC genes regulate different aspects of AFD neuron function

Loss-of-function mutations in gcy-8 result in marked IT behavioral defects due to defects in both isothermal track initiation and maintenance. Correspondingly, the AFD neurons of gcy-8 mutants exhibit defects in their ability to respond to oscillating temperature stimuli centered around T_c. In contrast, gcy-18 and gcy-23 mutants retain the ability to track isotherms, and the AFD neurons in these mutants continue to respond to these oscillating stimuli. However, gcy-18 and gcy-23 mutants track isotherms in distinct temperature zones, and both gcy-18 and gcy-8 mutants exhibit altered T^*_AFD. Thus, while all three rGC genes play roles in regulating the temperature range in which IT behavior can be exhibited, only gcy-8 regulates the ability of the AFD neurons to execute IT behavior.

Genes encoding these three rGCs are linked on LG IV and share extensive sequence homology [16]. Phylogenetic analyses indicate that these proteins belong to a larger nematode-specific clade and are likely to have arisen via gene duplication [22]. However, their functions in the AFD neurons appear to have diverged. GCY-18 and GCY-23 do not have completely redundant functions, since the majority of phenotypes of the gcy-23 gcy-18 double mutants are not significantly stronger than those of gcy-18 single mutants alone. Mutations in gcy-23 affect the temperature range of IT behavior without affecting T^*_AFD, suggesting that the changes in cGMP signaling in gcy-23 mutants may be sufficient to alter the threshold of AFD synaptic output but not that of sensory responses. It is possible that GCY-18 and GCY-23 are required under other conditions that require AFD function [16], and contribute to other phenotypes including regulation of lifespan [23, 24] and systemic heat shock responses [25]. The AFD neurons have been reported to express additional rGCs [16, 18] which may further contribute to AFD functional complexity. Similar complexity in rGC functions has been reported in the ASE chemosensory neurons in C. elegans which express >10 rGC genes [18, 26].

cGMP signaling likely sets the lower and upper bound of AFD response range

Our results indicate that the AFD neurons are able to respond to oscillating temperature stimuli only within a narrow temperature zone relative to T^*_AFD. The lower and upper bounds of this response range are altered in rGC mutants and upon growth in the presence of 8-Br-cGMP, suggesting that reduction or increase in cGMP signaling results in a lowering or raising of T^*_AFD, respectively, with corresponding shifts in the upper bounds.

A decrease or increase in intracellular cGMP signaling requires a lower or higher temperature, respectively, to initiate AFD neuronal activity at a given T_c. A simple possibility is that the enzymatic activity of the rGCs themselves is temperature-dependent, and that the TAX-2/4 CNG channels adapt to prevailing cGMP signaling at a given T_c. Indeed, C. elegans guanylyl cyclases have been reported to exhibit temperature-dependent activity [17, 27]. Thus, in AFD neurons with decreased cGMP signaling due to loss-of-function mutations in one or more rGCs, the CNG channels may have adapted to lower cGMP concentrations at T_c. Hence, the increase in intracellular cGMP upon a small rise in temperature may be sufficient to open the channels resulting in a lower T^*_AFD than in wild-type AFD neurons. Conversely, if the CNG channels are adapted to high cGMP levels at T_c, higher temperatures may be needed to generate sufficient cGMP in order to enable channel opening, leading to a higher T^*_AFD. CNG channel opening and sensitivity are known to be regulated by diverse mechanisms including phosphorylation, binding of Ca²⁺/calmodulin, and phospholipids [28–35]. One or more of these mechanisms may contribute to the adaptation of the channels at T_c. Additional mechanisms such as changes in expression of the rGCs may also play a role in setting the T^*_AFD [36].

Why do the AFD neurons respond to oscillating stimuli only at temperatures close to T^*_AFD? One possibility is that in this range, the concentration of cGMP and the number of open channels are optimal to permit a high degree of sensitivity to rapid temperature changes. At higher temperatures, more channels may be open decreasing the gain, which may in turn set the upper bound by decreasing neuronal sensitivity. Loss of GCY-8 affects the ability of the AFD neurons to respond to rapid temperature changes around T_c, suggesting that GCY-8, but not GCY-18 or GCY-23, plays an important role in rapid temperature-dependent cGMP production around T_c. However, gcy-8 mutants respond normally to upward oscillating ramps, perhaps due to involvement of other rGCs when temperatures begin to rise. It is important to note, however, that current experiments are unable to distinguish between temperature-dependent regulation of rGCs vs phosphodiesterases.

Regardless of the exact mechanism, adaptation to T_c and modulation of cGMP signaling allows the AFD neurons to both maintain a high degree of sensitivity in the required temperature range while retaining the ability to respond to a broad temperature range. This is reminiscent of mechanisms of peripheral and central gain control in sensory systems such as in vision and audition that permit both sensitivity and accuracy in sensory responses across a wide dynamic range [37, 38].

Different thermosensory neuron response properties evoke specific behaviors in defined thermal zones

We and others previously showed that the AWC polymodal sensory neurons exhibit different responses at temperatures near, above or below T_c [39, 40]. We now find that the AFD neurons are also differentially responsive to oscillating temperature stimuli in the temperature range at which IT behavior is exhibited. Thus, the temperature range-specific responses of the AFD neurons and AWC neurons permit IT behavior in a defined temperature zone. Integration of differential responses from multiple sensory neurons under specific conditions may increase the robustness and precision of the thermosensory behavioral response.

Thermotaxis in C. elegans is a remarkably complex experience-dependent behavior. Identification and characterization of the molecular mechanisms underlying sensory neuron response properties to distinct features of thermal stimuli may now permit analyses of how these patterns are integrated and translated through the circuit into precise changes in locomotor output.