Feedforward and feedback mechanisms cooperatively regulate rapid experience-dependent response adaptation in a single thermosensory neuron type

Experience-dependent adaptation of the response threshold in AFD has been largely assessed via determining the temperature above which these neurons exhibit changes in intracellular calcium dynamics (T*_Calcium) in response to a rising temperature ramp (9, 17–19) (Fig. 1B). We considered two mutually nonexclusive mechanisms by which the threshold of temperature-evoked calcium influx via CNG channels in AFD is rapidly altered by the animal’s temperature history. In the first, the threshold of activation or inhibition of one or more of the AFD-expressed thermosensor rGCs or PDEs (Fig. 1A), respectively, may be reset by temperature history resulting in a shift in the gating threshold of the CNG channels. In the second, only the threshold of CNG channel opening may be reset. The first but not the second model predicts that similar to the adaptation of T*_Calcium, the cGMP response threshold (T*_cGMP) will also adapt rapidly upon a temperature change.

To distinguish between these models, we shifted adult animals from 15 to 25 °C for different periods of time, and compared T*_cGMP and T*_Calcium in AFD neurons expressing the genetically encoded FlincG3 (24) or GCaMP6s cGMP and calcium sensors, respectively (Fig. 1B). Both T*_cGMP and T*_Calcium adapted rapidly within minutes to a significantly higher temperature (Fig. 1 C and D and SI Appendix, Fig. S1 A and B), and adapted to their final values upon overnight growth at 25 °C (Fig. 1 C and D and SI Appendix, Fig. S1 A and B) (25). T*_Calcium adaptation was similar at the AFD sensory endings and soma indicating that rapid adaptation occurs at the sensory endings themselves (SI Appendix, Fig. S1C). These observations indicate that similar to adaptation of T*_Calcium, temperature experience also rapidly modulates T*_cGMP in AFD. T*_Calcium adaptation may be a consequence of T*_cGMP adaptation alone, and/or due to additional modulation of CNG channels.

Temperature-regulated cGMP levels in AFD are determined by the opposing actions of cGMP production and hydrolysis by the GCY-8, GCY-18, and GCY-23 thermosensor rGCs, and multiple PDEs, respectively (Fig. 1A). The roles of individual rGCs in regulating rapid thermosensory adaptation in AFD are unknown. Since misexpression of GCY-18 or GCY-23, but not GCY-8, is sufficient to confer temperature responses onto other cell types (26), we focused on the contributions of GCY-18 and GCY-23 to rapid thermosensory adaptation.

We examined cGMP and calcium responses in AFD neurons in gcy-8 gcy-18 double mutants expressing GCY-23 alone, (Fig. 2 A–F), or in gcy-23 gcy-8 double mutants expressing GCY-18 alone (Fig. 2 G–L). cGMP response amplitudes were significantly dampened in both double mutants as compared to responses in wild-type animals (Fig. 2 A and G). While expression of either rGC alone was sufficient for rapid T*_cGMP adaptation to warmer temperatures (Fig. 2 B and H), T*_cGMP in AFD expressing GCY-18 alone adapted to warmer temperatures more quickly following a temperature upshift (Fig. 2H). T*_cGMP was significantly lower in GCY-23- but not GCY-18-expressing neurons upon overnight growth at 25 °C (Fig. 2 C and I) (25). Consistent with lower levels of intracellular cGMP leading to the opening of fewer calcium channels, the amplitude of temperature-evoked calcium responses was also lower in neurons expressing either rGC (Fig. 2 D and J). While T*_Calcium also adapted rapidly upon a temperature upshift in neurons expressing either rGC alone, T*_Calcium values were consistently lower and higher in GCY-23- and GCY-18-expressing neurons, respectively (Fig. 2 E and K). T*_Calcium was also lower in GCY-23- but not GCY-18-expressing animals upon overnight growth at 25 °C (Fig. 2 F and L). T*_Calcium in ASE chemosensory neurons misexpressing GCY-23 did not exhibit rapid response adaptation upon a temperature upshift (SI Appendix, Fig. S2), suggesting that the adaptation mechanisms are likely to be regulated by AFD-specific pathways. We infer that expression of either rGC alone is sufficient to mediate rapid adaptation of AFD responses, but that each rGC drives adaptation at distinct rates.

How might a temperature shift reset the response threshold of a thermosensory rGC? rGCs such as the guanylyl cyclase A (GC-A) natriuretic peptide receptor and the sea urchin sperm guanylyl cyclase chemoreceptor are phosphorylated at residues in their intracellular juxtamembrane kinase homology domain in the unliganded state (27, 28). Ligand binding results in dephosphorylation of these residues leading to response desensitization (29–31). We hypothesized that the response threshold of thermosensory rGCs could be modulated via similar phosphorylation/dephosphorylation cycles upon a temperature upshift.

Five of six residues targeted for phosphorylation in the intracellular domains of GC-A and the sea urchin sperm guanylyl cyclase are conserved in GCY-18 (Fig. 3A). We generated a gcy-18 allele in which all five predicted phosphorylation sites were mutated to alanine (gcy-18(5A); Fig. 3A) via gene editing at the endogenous locus in the gcy-23 gcy-8 double mutant background. We found that expression of GCY-18(5A) resulted in a significantly lower T*_Calcium both upon short- and long-term temperature upshift (Fig. 3B). Examination of the response dynamics in individual neurons showed that following activation at the respective T*_Calcium, intracellular calcium levels appeared to oscillate in a subset of GCY-18(5A)-expressing AFD neurons (Fig. 3C). These oscillations increased significantly upon a rapid temperature upshift but were observed less frequently upon overnight growth at 25 °C or in neurons expressing wild-type GCY-18 (Fig. 3C). We suggest that phosphorylation of one or more of the targeted residues in GCY-18 may be necessary for both short- and long-term adaptation of the response threshold of this rGC. Moreover, in the absence of this phosphorylation, GCY-18 responses may desensitize rapidly after activation and are reactivated as temperatures rise, resulting in the observed calcium oscillations in response to a rising temperature ramp.

cGMP and calcium signaling pathways are intricately interconnected in vertebrate phototransduction. Low intracellular calcium levels upon light-induced hyperpolarization activate retinal rGCs via EF-hand containing guanylyl cyclase activating calcium sensor proteins (GCAPs) to regenerate cGMP (32–34). Since buffering of calcium has previously been shown to modulate sensory adaptation dynamics in AFD (16), we asked whether calcium influx feeds back to modulate cGMP in AFD.

Mutations in the tax-4 CNG channel α subunit largely abrogate temperature-evoked calcium dynamics in AFD (17). We were unable to generate a strain expressing the FlincG3 cGMP sensor in AFD in tax-4 mutants. tax-2 CNG channel β subunit mutants retained partial but highly variable temperature-evoked calcium and cGMP responses in AFD (SI Appendix, Fig. S3 A and B), possibly due to decreased expression of the imaging reporters in this mutant background (35). Although the response variability and reduced response amplitude precluded accurate measurement of T*_cGMP, these observations suggest that calcium influx may modulate cGMP levels and/or dynamics in AFD.

To determine whether calcium acts via one or more GCAP-like proteins to regulate rGCs in AFD, we examined the contributions of C. elegans neuronal calcium sensor homologs in modulating temperature-evoked cGMP levels in AFD. Of the three GCAP-related neuronal calcium sensor proteins (NCS-1 to -3) predicted to be encoded by the C. elegans genome, NCS-1 and NCS-2, but not NCS-3, are expressed in AFD (36). Although NCS-1 has previously been shown to regulate AFD temperature responses and AFD-driven thermosensory behaviors (37–39), we observed no defects in calcium or cGMP responses in AFD in ncs-1 mutants (SI Appendix, Fig. S3C).

We noted that basal FlincG3 fluorescence levels were significantly higher in ncs-2 mutants as compared to wild-type animals upon growth at either 15 °C or 25 °C overnight (Fig. 4A). Expression of gfp driven under the same promoter driving FlincG3 expression did not result in a similar increase in fluorescence indicating that this increased expression is unlikely to arise from altered expression levels (SI Appendix, Fig. S3D). While T*_cGMP was not altered in animals under any examined condition (Fig. 4B), the amplitude of the cGMP response in AFD in ncs-2 mutants was increased in animals upon rapid temperature upshift as well as following overnight growth at 25 °C (Fig. 4C). Increased cGMP response amplitudes and increased basal cGMP levels are predicted to increase the amplitude of the calcium response due to the opening of additional CNG channels, and decrease the threshold of T*_Calcium by enabling the channels to open at a lower temperature following a temperature upshift. T*_Calcium adapted to a lower value in two independent ncs-2 mutants (Fig. 4D), indicating that within minutes after a temperature upshift, CNG channels open at a lower temperature than in wild-type animals in ncs-2 mutants. In addition, the amplitude of the calcium response was increased in ncs-2(ju836) mutants after 3 to 10 min, and in both ncs-2 alleles upon overnight growth, at 25 °C (Fig. 4E). We infer that NCS-2 may inhibit basal cGMP levels in both overnight and rapid temperature shift conditions to alter T*_Calcium but not T*_cGMP adaptation.

Although multiple sites of direct interaction between GCAPs and the intracellular domains of retinal rGCs have been postulated (40–42), mutating an Arg residue to Pro in the dimerization domain between the kinase homology and catalytic domains of RetGC1 was shown to abrogate GCAP binding in cultured cells (Fig. 4F) (43). We mutated the homologous conserved residues in GCY-18 [GCY-18(K860P)] and GCY-23 [GCY-23(R820P)] (Fig. 4F) via gene editing in gcy-8 gcy-23 and gcy-8 gcy-18 double mutant backgrounds, respectively, and measured AFD responses. While T*_Calcium in neurons expressing GCY-23(R820P) adapted similarly to neurons expressing wild-type GCY-23 alone (Fig. 4G), unlike in ncs-2 mutants, T*_Calcium adapted to a warmer temperature upon expression of GCY-18(K860P) (Fig. 4H). These observations suggest that NCS-2 is unlikely to interact solely via the targeted residues on either GCY-23 or GCY-18 to modulate their functions, and may act via other as yet unidentified mechanisms to regulate basal cGMP levels in AFD.

cGMP-mediated opening of CNG channels results in calcium influx; intracellular calcium feeds back via calmodulin to modulate CNG channel properties and sensory adaptation (11, 44–46). The TAX-2 and TAX-4 CNG thermotransduction channel subunits have been reported to not contain calcium-calmodulin binding sites (47). However, phosphorylation of TAX-2 by the cGMP-dependent protein kinase (PKG) EGL-4 is critical for short-term odorant adaptation in the AWC olfactory neurons in C. elegans (15). Phosphorylation also modulates CNG channel functions in other sensory systems (48–50). We tested whether modulation of the response threshold of the CNG channels via PKG-mediated phosphorylation contributes to T*_Calcium adaptation in AFD.

We found that the tax-2(S727A) mutation that affects olfactory adaptation only minimally affected T*_Calcium in AFD following a temperature upshift on either a rapid or slow timescale (SI Appendix, Fig. S4A), suggesting that this residue and/or the TAX-2 subunit are not the major targets of modulation in AFD. In addition to tax-2 and tax-4, the cng-3 α subunit of CNG channels is expressed in multiple sensory neuron types including in AFD (51, 52) and has been shown to be necessary for robust thermotaxis behavior (53, 54). This channel subunit has also been implicated in the regulation of thermotolerance (52), as well as in short-term olfactory adaptation in the AWC olfactory neurons (47). Temperature-evoked calcium responses in cng-3(jh113) null mutants were similar to those in wild-type animals (Fig. 5 A and B), indicating that unlike TAX-2 and TAX-4, this channel subunit is not essential for primary thermotransduction. However, T*_Calcium was consistently lower than in wild-type animals in cng-3 mutants upon a rapid temperature upshift but unaltered upon overnight growth at 25 °C (Fig. 5C). No effects were observed on T*_cGMP adaptation (SI Appendix, Fig. S4B). The S20 residue in CNG-3 (Fig. 5D) has been predicted to be a PKG target mediating short-term olfactory plasticity in the AWC neurons (47). Similar to observations in cng-3(jh113) mutants, rapid T*_Calcium adaptation was reduced in animals carrying an S20A mutation in the endogenous cng-3 locus (Fig. 5D and Movie S1), but not upon overnight growth at 25 °C (Fig. 5D). cng-3(S20A) mutants also exhibited defects in thermotaxis navigation behaviors such that while wild-type animals grown at 15 °C navigated to colder temperatures on a thermal gradient, cng-3 mutants failed to do so (Fig. 5E) (53, 54). These results suggest that phosphorylation of CNG-3 plays a critical role in modulating T*_Calcium, but that this CNG channel subunit is dispensable for primary thermosensory responses.

The EGL-4 PKG has previously been implicated in mediating multiple forms of olfactory adaptation in AWC in C. elegans (4, 15). In particular, EGL-4-mediated phosphorylation of the S727 and S20 residues in TAX-2 and CNG-3, respectively has been suggested to be critical for short-term olfactory adaptation in AWC (15, 47). Given the role of CNG-3 phosphorylation in rapid thermosensory adaptation, we tested whether EGL-4 also regulates T*_Calcium adaptation in AFD.

We found that neither egl-4(n479 lof) nor egl-4(ad450 gof) mutants exhibited defects in T*_Calcium adaptation upon either a short- or long-term temperature upshift (Fig. 6A). In addition to egl-4, the C. elegans genome encodes a less well-characterized PKG encoded by pkg-2 (55). pkg-2(tm5814 lof) animals exhibited relatively minor although significant defects at the 10 min timepoint in T*_Calcium adaptation but not upon overnight growth at any examined temperature (Fig. 6B). However, egl-4; pkg-2 double mutants exhibited strong defects in short- but not long-term T*_Calcium adaptation similar to the phenotypes of cng-3(jh113) and cng-3(S20A) mutants (Fig. 6C). As in the case of cng-3 mutants, calcium response amplitudes were unaffected in egl-4; pkg-2 double mutants (SI Appendix, Fig. S5A). T*_cGMP adaptation on either rapid or slow timescales was unaffected in egl-4; pkg-2 double mutants suggesting that these enzymes do not target the rGCs (Fig. 6D). EGL-4 has previously been reported to undergo nuclear translocation upon prolonged odorant exposure in the AWC olfactory neurons (56); the subcellular localization of a functional GFP-tagged EGL-4 protein was unaltered upon a 15 min temperature upshift in AFD (SI Appendix, Fig. S5B). These results indicate that the EGL-4 and PKG-2 cGMP-dependent kinases act redundantly to mediate short-term T*_Calcium adaptation possibly via phosphorylation of CNG-3.

Conclusions regarding AFD neuronal activity are limited by using calcium as a proxy for neuronal activity. We are unable to determine absolute calcium or cGMP concentrations in AFD using fluorescent sensors. Responses to only a 10 °C step change from 15 °C to 25 °C at a single rate of change were examined in this work; distinct pathways may contribute differentially to adaptation in response to other temperature change paradigms. We have not directly measured the cGMP affinities of the channel subunits, and whether this affinity changes upon phosphorylation. Whether functional changes in phosphodiesterases also contribute to rapid T*_cGMP adaptation was not examined.

C. elegans were maintained using standard conditions on Escherichia coli OP50-seeded nematode growth agar plates. PCR-based sequencing was used to confirm the presence of molecular lesions in strains. See SI Appendix, Table S1 for the list of all strains used in this work.

Calcium imaging was performed essentially as described previously (23, 26). T*_Calcium and T*_cGMP were defined as the temperature at which ΔF/F increased by a minimum of 2% over at least 8 consecutive seconds with an average slope of >0.3% per second. Calcium imaging data reported in all figures were acquired from AFD soma except in Fig. 1 which reports measurements from the AFD sensory endings, and SI Appendix, Fig. S1C which shows calcium responses from both AFD soma and sensory endings. All cGMP measurements were performed from the AFD sensory endings. Additional details are provided in SI Appendix, Supplemental Methods.

Thermotaxis behavior was performed essentially as described (26, 69). A detailed protocol is described in SI Appendix, Supplemental Methods.

All cRNAs, tracrRNAs, and Cas9 protein were obtained from Integrated DNA Technologies (IDT). Point mutations in gcy-18, gcy-23, tax-2, and cng-3 were generated using ssODN repair templates which included the mutated site along with 35 bp 5′ and 3′ homology arms. Injection mixes for all manipulations included the ssODN repair template (110 ng/μL), Cas9 protein (250 ng/μL), tracerRNA(100 ng/μL), crRNA (28 ng/μL), and a coinjection marker [myo-3p::mCherry (50 ng/μL)]. F1 animals expressing the coinjection marker were analyzed for the presence of the desired mutations via PCR and sequencing of the relevant genes. F2 progeny were subsequently screened for homozygous edits via PCR and sequencing. Sequences of all oligonucleotides are listed in SI Appendix, Table S2.

All statistical analyses were performed using GraphPad Prism version 10.0.0 (www.graphpad.com) which was also used to generate all box plots. Statistical test details and the number of analyzed samples are reported in each figure legend.

T.J.H. and P.S. designed research; T.J.H. performed research; T.J.H. contributed new reagents/analytic tools; T.J.H. analyzed data; P.S. acquired funding; and T.J.H. and P.S. wrote the paper.

The authors declare no competing interest.