|Home | About | Journals | Submit | Contact Us | Français|
Many neurons release classical transmitters together with neuropeptide cotransmitters whose functions are incompletely understood. Here we define the relationship between two transmitters in the olfactory system of Caenorhabditis elegans, showing that a neuropeptide-to-neuropeptide feedback loop alters sensory dynamics in primary olfactory neurons. The AWC olfactory neuron is glutamatergic and also expresses the peptide NLP-1. nlp-1 mutants have increased AWC-dependent behaviors, suggesting that NLP-1 limits the normal response. The receptor for NLP-1 is the G protein-coupled receptor NPR-11, which acts in postsynaptic AIA interneurons. Feedback from AIA interneurons modulates odor-evoked calcium dynamics in AWC olfactory neurons and requires INS-1, a neuropeptide released from AIA. The neuropeptide feedback loop dampens behavioral responses to odors on short and long timescales. Our results point to neuronal dynamics as a site of behavioral regulation and reveal the ability of neuropeptide feedback to remodel sensory networks on multiple timescales.
Embedded in an animal’s neuroanatomy are the pathways that drive its behavior. Over twenty years ago, a wiring diagram of the C. elegans nervous system was constructed from serial-section electron micrographs1. Despite this unique resource, the relationships between sensory inputs, neuronal activity, and specific behavioral programs are unclear2, 3. We are exploring these issues within a circuit that generates undirected search when animals are removed from food, and directed chemotaxis in odor gradients4–6. Both of these behaviors are initiated by AWC olfactory neurons and are based upon temporally-regulated turning: when animals are removed from food, a transient bout of turning produces undirected local search, and in the presence of an odor gradient, temporally-regulated turning produces a biased random walk for gradient climbing7, 8. Through cell ablation and quantitative behavioral analysis, neurons in this circuit have been traced from sensory input to motor output5, 6. Sensory neurons detect changes in odor levels or feeding state, and synapse onto a layer of interneurons that control turning rates in a coordinated fashion; a second layer of interneurons and downstream motor neurons regulate specific classes of turns and features of turns. The turning circuit has interesting temporal properties. First, even in a constant odor environment, an intrinsic process generates apparently stochastic turns. Second, to move up a spatial odor gradient using a biased random walk, animals calculate the time derivative of the odor signal over seconds. Third, odor history and food regulate the overall turning rate on a longer timescale of minutes. A meaningful understanding of this circuit must include neuronal properties that represent these varied behavioral timescales.
Previous studies demonstrated that AWC releases the neurotransmitter glutamate, which activates AIB interneurons via the glutamate-gated cation channel GLR-1 and inhibits AIY interneurons via the glutamate-gated chloride channel GLC-39 (Fig. 1a). AWC also expresses genes that encode predicted neuropeptides, including the buccalin-related peptide NLP-110. We characterized the function of NLP-1 by examining local search behavior, in which animals increase their turning rates during the initial 15 minutes after they are removed from food. Local search depends on AWC activity, and the rate of turning provides a quantitative measurement of AWC signaling6. A null mutant for nlp-1, generously provided by the C. elegans knockout consortium, had a higher turning rate than wild-type animals during local search (Fig. 1b). Like wild-type animals, nlp-1 mutants suppressed turning after 15 minutes off food (Supplementary Table 1). This behavioral profile suggests that nlp-1 inhibits AWC-induced turning behavior.
NLP-1 reporter genes are expressed in AWC, ASI, PHB, and BDU neurons and in the intestine10. The enhanced turning defect in nlp-1 mutants was rescued by transgenic expression of nlp-1 in AWC neurons, but not by expression in ASI neurons (Fig. 1b, Supplementary Table 1). Overexpression of nlp-1 in AWC neurons of wild-type animals reduced the turning rate during local search (Fig. 1b). The opposite effects of nlp-1 null mutants and gain-of-function transgenes suggest that this neuropeptide can be an instructive determinant of turning rates.
AWC-dependent turning behavior is reduced in mutants in the vesicular glutamate transporter gene eat-4 or in the glutamate receptors glr-1 or glc-3, defects opposite to those of nlp-1 mutants9. Double mutants between nlp-1 and eat-4, glr-1, or glc-3 had reduced turning rates, resembling the mutants that affected glutamate signaling (Fig. 1b). This result suggests that NLP-1 functions as a cotransmitter whose effects are only apparent when the classical transmitter glutamate is also released: glutamate from AWC stimulates turning, and NLP-1 from AWC decreases the magnitude of this effect.
The C. elegans genome encodes about 100 G protein-coupled receptors (GPCRs) related to characterized neuropeptide receptors (www.wormbase.org). To identify the receptor for NLP-1, we examined GPCRs that lack known ligands, focusing on those for which expression has been observed in neurons connected to AWC11, 12. The orphan GPCR NPR-11, whose closest characterized homolog is the Drosophila neuropeptide F receptor 1, was one candidate. We found that npr-11 null mutants, like nlp-1 mutants, had increased turning during AWC-dependent local search behavior (Fig. 1c, Supplementary Table 1).
nlp-1 npr-11 double mutants resembled the single mutants in turning behavior, the expected result for a ligand-receptor pair that acts together and not additively (Fig. 1c). In addition, the npr-11 mutation fully suppressed the effects of nlp-1 overexpression on turning behavior (Fig. 1c). This result indicates that npr-11 is necessary for the biological activity of nlp-1, as predicted if npr-11 encodes an nlp-1 receptor.
npr-11 reporter genes are expressed in two postsynaptic targets of AWC, the AIA and AIY interneurons, and in other neurons11 (data not shown). The npr-11 behavioral defect in local search behavior resembled the defect after AIA interneurons are killed with a laser9, and the npr-11 defect was fully rescued by transgenic npr-11 expression in AIA interneurons under the gcy-28.d promoter (Fig. 1c). These results suggest that NLP-1 release from AWC activates the NPR-11 GPCR on AIA. In agreement with this hypothesis, the turning behavior of nlp-1 npr-11 double mutants was rescued only when nlp-1 was expressed in AWC and npr-11 was expressed in AIA together in the same strain (Fig. 1c, Supplementary Table 1).
A biochemical confirmation of the genetically inferred ligand-receptor relationship for NLP-1 and NPR-11 was obtained by expressing NPR-11 in HEK 293 cells together with the promiscuous G-protein α16Z13, and exposing the cells to synthesized peptide corresponding to one of four predicted MSFamide peptides encoded by nlp-1, MDANAFRMSFamide. NPR-11-expressing cells responded to micromolar concentrations of the peptide with calcium transients typical for GPCR activation (Fig. 1d). No responses were observed in response to scrambled NLP-1 peptide, in untransfected cells, or in cells expressing receptor or G-protein alone (Fig. 1d,e).
The relationship between AWC and AIA interneurons was examined directly using genetically-encoded calcium indicators of the G-CaMP family, which increase their fluorescence emission upon calcium binding14,15. C. elegans neurons are thought to lack sodium-based action potentials, but express voltage-gated calcium channels16, and calcium signals in these neurons correlate with neuronal depolarization17–22. Previous experiments suggest that AWC neurons have basal activity at rest, are hyperpolarized by odor addition, and are strongly activated by odor removal9. We found that AIA interneurons had an opposite response, showing large transient calcium increases upon odor addition (Fig. 2a,b). Odor responses in AIA required AWC sensory input, as they were strongly attenuated when AWC neurons were killed using a laser microbeam (Fig. 2c,d). These results suggest that AWC forms inhibitory synapses onto AIA, as previously suggested for the synapses between AWC and AIY (Fig. 1a)9. In the simplest model, basal AWC activity at rest tonically inhibits AIA; when odor is added, AWC is hyperpolarized, tonic inhibition of AIA is reduced, and AIA becomes active.
Both glutamate and NLP-1 neurotransmitters contributed to odor-evoked calcium responses in AIA. AIA responses to odor were diminished in eat-4 mutants that lack the vesicular glutamate transporter, and in glc-3 mutants that lack a glutamate-gated chloride channel (Fig. 2e,f). The involvement of glc-3 is consistent with its known expression in AIA, its molecular identity as an inhibitory glutamate-gated channel, and the prediction that the AWC-to-AIA synapse is inhibitory11, 23. nlp-1 mutants also had a diminished response in AIA interneurons, which was partly rescued by expression of nlp-1 in AWC neurons (Fig. 2g). The partial rescue of nlp-1 may be due to variable expression of transgenes or to a combined action of nlp-1 in AWC and additional neurons.
Although a dual action of glutamate and NLP-1 on AIA calcium signals was consistent with the behavioral analysis, the relationship between the transmitters was unexpected. In turning behavior, glutamate and NLP-1 had opposite effects, but in AIA calcium imaging experiments, the effects of nlp-1 and glc-3 were congruent. This distinction suggests that the behavioral effect of NLP-1 cannot be entirely explained by its observed effect on AIA activity, and prompted further examination of neurons in the AWC circuit.
The activity of sensory neurons such as AWC is thought to be defined primarily by sensory input, and not by network interactions. Unexpectedly, calcium imaging of the odor-evoked response in AWC neurons revealed alterations in nlp-1 mutants. In both wild-type and nlp-1 animals, initial calcium responses to odor removal peaked within ten seconds of odor removal (Fig. 3a–c, Supplementary Figs. S1, S2). However, in nlp-1 mutants, the first peak was frequently followed by large secondary calcium transients that continued for at least two minutes after odor removal, the longest duration that was practical for calcium imaging (Fig. 3b,c). Thus NLP-1 neuropeptide signaling suppressed AWC calcium transients at long times after odor removal.
To provide a quantitative description of the altered activity in nlp-1 mutants, a discrete Fourier transform was used to analyze the temporal character of the AWC calcium response to odor removal. Fourier analysis is well suited to reveal oscillatory or repeating signals, which appeared by inspection to be present in nlp-1 traces. Indeed, spectral analysis suggested the existence of increased power in a mid-frequency domain in nlp-1 mutants, with a significantly greater contribution than wild type from periodic components between 0.033 Hz - 1 Hz (period 1 – 30s) (Fig. 3d). The effects were statistically detectable at all time points, and suggest the presence of irregular slow oscillations with a preferred period near 20s. Cell-selective transgenic expression of nlp-1 in AWC resulted in partial but significant rescue of the defect (Fig. 3c,d, Supplementary Fig. S1). The irregular oscillating signal was not observed in the absence of odor, excluding trivial optical or mechanical artifacts (Supplementary Fig. S1), but it was observed to some degree in wild-type traces as well as nlp-1 traces, suggesting that it is a component of the normal AWC response.
Enhanced secondary AWC calcium signals were also observed in npr-11 mutants after odor removal (Fig. 3e–g Supplementary Figs. S1,S2). Normal AWC calcium signals were restored by transgenic expression of npr-11 in AIA neurons (Fig. 3g,h, Supplementary Fig. S1). The functional effect of a receptor in AIA on odor responses in AWC strongly suggests the existence of feedback from AIA to AWC. In summary, these results suggest that NLP-1 release from AWC, sensed by NPR-11 in AIA, results in feedback onto AWC that suppresses secondary calcium transients after odor removal.
If modulatory feedback from the circuit contributes to normal AWC activity, other synaptic mutants could also affect AWC odor responses. Indeed, AWC responses to odor removal were reduced in eat-4 mutants, which lack glutamatergic transmission from AWC and other neurons, and prolonged in unc-31 mutants, which have reduced neuropeptide release from all neurons (Supplementary Fig. S2c,d)24, 25. These results suggest that the long-lasting AWC responses to odor removal are affected by synaptic inputs, with positive inputs from glutamatergic neurons and negative inputs from peptidergic neurons.
The results described above suggest effects of neuropeptide signaling and feedback on long-lasting dynamics in AWC neurons. The 15 minute local search behavior is one sustained AWC-dependent behavior; a second sustained AWC-dependent behavior is olfactory adaptation, a behavioral change in which prolonged exposure to an odor leads to reduced chemotaxis to that odor (Fig. 4a,b)26. nlp-1 and npr-11 mutants failed to adapt to odors after a 60 min exposure (Fig. 4b), although they did adapt after a 90 minute exposure (Supplementary Fig. S3a). Like the defects in local search behavior, the olfactory adaptation defects were rescued by transgenic expression of nlp-1 in AWC and transgenic expression of npr-11 in AIA, respectively (Fig. 4b).
Previous studies have identified signaling components required cell-autonomously for olfactory adaptation in AWC, including GPCR regulators, TRPV channels, and a cGMP-dependent protein kinase27–32. However, the effects of adaptation on AWC activity have not been described. We found that wild-type animals adapted to odor for 60 minutes failed to respond to odor removal with AWC calcium transients, suggesting that adaptation blocked an early step of AWC signaling (Fig. 4c). By contrast, both nlp-1 and npr-11 mutants that had been exposed to odor for 60 minutes responded with AWC calcium transients after odor removal (Fig. 4c,d). Transgenic expression of nlp-1 in AWC and npr-11 in AIA neurons restored the wild-type AWC calcium response (Fig. 4c,d). These results indicate that NLP-1 release from AWC, acting on the NPR-11 receptor in AIA, generates feedback that reduces the primary AWC response to odor removal after adaptation.
The results described above suggest that AIA releases a signal that suppresses odor-evoked calcium responses in AWC. The insulin-related neuropeptide INS-1, which is expressed in AIA, modulates the function of the ASE gustatory neurons in a salt conditioning assay33, suggesting INS-1 as a possible signal from AIA to sensory neurons. Therefore, ins-1 mutants were examined for their effects on AWC-regulated behaviors. Like nlp-1 and npr-11 mutants, ins-1 mutants had increased turning in AWC-dependent local search behavior (Fig. 5a, Supplementary Table 1). Double mutants between ins-1 and nlp-1 resembled each single mutant, suggesting that these genes act in a common process (Fig. 5a). An ins-1 mutation fully suppressed the effects of NLP-1 overexpression on turning behavior, suggesting that ins-1 activity is necessary for the biological effects of nlp-1 (Fig. 5a).
ins-1 is expressed in multiple neurons including AIA, RIC, ASI, and AWC33. Transgenic expression of ins-1 from an AIA-selective promoter rescued turning behavior in the ins-1 mutant, but expression from AWC, ASI, or RIC promoters did not (Fig. 5a, Supplementary Table 1, and data not shown). These results suggest a specific requirement for ins-1 in AIA neurons. In double mutant studies, nlp-1 ins-1 double mutants were restored to normal turning behavior only when nlp-1 was expressed in AWC and ins-1 was also expressed in AIA (Fig. 5a). nlp-1(overexpressor) ins-1 double mutants were restored to the behavior of nlp-1(overexpressor) strains when ins-1 was expressed in AIA (Fig. 5a). These experiments support a site of ins-1 action in AIA neurons.
In calcium imaging experiments, ins-1 mutants had secondary AWC transients after odor removal, with temporal properties similar to nlp-1 and npr-11 mutants (Fig. 5b, Supplementary Fig. S4). AIA specific expression of ins-1 significantly rescued the altered odor response in AWC neurons, suggesting that AIA neuropeptides influence AWC activity (Fig. 5b, Supplementary Fig. S4).
ins-1 was also required for AWC-dependent olfactory adaptation. Like nlp-1 and npr-11 mutants, ins-1 mutants did not adapt after a 60 min exposure to odor, but did adapt partially after 90 minutes (Fig. 5c, Supplementary Fig. S3). Adaptation in nlp-1 ins-1 double mutants was similar to adaptation in each single mutant, suggesting that these two genes affect a common pathway (Supplementary Fig. S3c). Moreover, AWC neurons in ins-1 animals exposed to odors for 60 minutes showed a calcium response after odor removal, unlike wild-type animals in the same conditions (Fig. 5d). Both the behavioral adaptation and the neuronal correlate of adaptation observed in calcium imaging were rescued by transgenic expression of ins-1 in AIA (Fig. 5c,d Supplementary Fig. S3). These results suggest that INS-1 release from AIA acts directly or indirectly on AWC sensory neurons to limit their activity.
Many neurons release both classical neurotransmitters and neuropeptides34–36. Our results suggest that AWC releases both the classical neurotransmitter glutamate and the neuropeptide NLP-1 to modulate behavior (Fig S4e). The glutamate signaling pathway promotes local search and odor chemotaxis, whereas the neuropeptide pathway limits local search and promotes odor adaptation. Glutamate signals are interpreted by multiple glutamate receptors on interneurons9, and NLP-1 is sensed by NPR-11 on AIA interneurons, which release INS-1 that directly or indirectly limits AWC activity and behavior. In one straightforward model, INS-1 is released from active AIA neuron when odor is present; the INS-1 released during a short odor exposure inhibits repetitive calcium transients from AWC when odor is removed, and the prolonged INS-1 release during a long odor exposure has a stronger effect, completely suppressing AWC calcium transients upon subsequent odor removal. This feedback loop may act as a gain control circuit to dampen AWC output to strong stimuli, a role that negative feedback plays in other neuronal systems37. In this interpretation, the increased oscillations in nlp-1 and other mutants result from ungoverned high-gain signaling in AWC.
A number of inhibitory pathways involving a single neuropeptide are known; this pathway is unusual in its apparent requirement for two neuropeptides from two neurons, working in a feedback loop. Several properties of this feedback loop remain to be determined. For example, we do not know when NLP-1 and INS-1 are released with respect to food or odor cues, and with respect to AWC and AIA activity. If the peptides are released when the neurons are active, NLP-1 would be released in alternation with INS-1 and not simultaneously; the slow biochemical timescale of G protein signaling may permit temporal integration across asynchronous activity of AWC and AIA38. Another missing component of the feedback model is the receptor for INS-1, as mutation of the one characterized C. elegans insulin receptor gene, daf-2, did not mimic or suppress ins-1 mutations as predicted for an ins-1 receptor (data not shown). ins-1 antagonizes daf-2 in the developmental dauer larva pathway39, and can act either as an agonist or an antagonist of daf-2 in food-regulated thermal learning and salt learning paradigms33, 40, but C. elegans has over 30 insulin-regulated peptides, and daf-2 may not be the only receptor for this peptide family. Precedent for alternative insulin receptors exists in mammals, where GPCRs are receptors for the insulin-related relaxin peptides41.
The behavioral functions of ins-1 in AWC olfactory adaptation are related to its previously described functions during ASE salt chemotaxis learning, where ins-1 from AIA suppresses chemotaxis after salt is paired with starvation33. It will be interesting to ask whether NLP-1 or another sensory peptide initiates ASE salt chemotaxis learning, and whether ins-1 signaling alters ASE sensory dynamics. Although we specifically examined the relationship between AWC and AIA, the AIA interneurons also receive synaptic input from neurons that sense food, tastants, pheromones, and repellents1, 33, 42. This connectivity might enable AIA to act as a local integrator of sensory information.
Increasing evidence in many animals indicates that peripheral olfactory signaling is modulated by top-down signals, internal states, and neuromodulators. State-dependent inputs affect olfactory signaling in rodents, acting as early as the synapses of the olfactory sensory neurons43,44. In Drosophila, tachykinin peptides expressed by local interneurons mediate presynaptic inhibition of olfactory receptor neurons45. The modulation of primary sensory neurons is a prominent feature of pain-sensing pathways46; we suggest that it will also be prominent in olfactory systems. In C. elegans, the use of specific mutants allows these feedback mechanisms to be directly linked to olfactory responses and behaviors, and also reveals their functional diversity. The neuropeptide feedback loop described here acts to dampen AWC output, while a different modulatory pathway involving the receptor guanylate cyclase GCY-28 can regulate an AWC switch from behavioral attraction to repulsion8. The C. elegans genome contains about 113 genes that encode more than 250 predicted neuropeptides, a rich potential source of behavioral variability10, 47, 48.
The experiments described here reveal neuropeptide-regulated dynamic properties of C. elegans neurons that correlate with the dynamics of behavior. The prolonged AWC calcium signals in nlp-1, npr-11, and ins-1 mutants correlate with their increased turning during local search behavior. The reduced AWC calcium signals after prolonged odor exposure correlate with olfactory adaptation, which is reduced in nlp-1, npr-11, and ins-1 mutants. Further studies of the AWC circuit should generate a better understanding of the relationship between behavior and the time-varying acute responses of AIA, sustained responses of AIB, and repetitive responses of AWC and AIY9. Together with previous results in C. elegans mechanosensory neurons49, these results suggest that circuit input tunes sensory responses to external stimuli based on sensory history and internal states, generating a rudimentary form of decision making.
To generate an AIA imaging line, GCaMP2.2b15 was expressed under the gcy-28.d promoter8, which is expressed strongly in AIA and weakly in RIA, ASK, AVJ, and other cells. The AWC GCaMP1 imaging line was described previously9. The lines were scored quantitatively for AWC local search behaviors, and those with normal responses were used for calcium imaging. Animals were trapped in a custom-designed microfluidic device made of the transparent polymer PDMS in which animals are restrained in a small chamber matching their dimensions, exposed to odors in liquid streams under laminar flow, and monitored using wide-field fluorescence microscopy9,22. Fluorescence from the cell of interest was captured after the presentation of isoamyl alcohol (10−4 dilution) and after odor removal five minutes later.
Metamorph and a Coolsnap HQ (Photometrics) camera were used to capture stacks of TIFF images at 10 frames/sec during the addition and removal of odor stimulus. A region of interest encompassing the cell was identified in all frames and the average fluorescence intensity recorded. A Matlab script used the data generated by MetaMorph to plot the imaging responses. The average fluorescence of the region of interest was generated by subtracting the recorded value from the average intensity of the background region of a similar area. The average fluorescence in a 3 second window (t=1–4s) was set as Fo. The percent change in the fluorescence intensity for the region of interest relative to Fo was plotted for all stacks, and these data were used for further analysis. Raw traces were corrected for fluorescence bleaching by subtracting a fixed correction function obtained by fitting an exponential curve to wild-type control traces using the equation: y(t) = 100% (exp(−t/τ) – 1) where t is time and τ is time constant. Since bleaching rates were found to vary among neurons, separate time constants were obtained for AWC (τ = 330.7 s; 95% confidence interval: 327.1 – 334.4 s; R2 = 0.91) and AIA (τ = 615.5 s; 95% confidence interval: 607.2 – 623.8 s; R2 = 0.86).
Because the irregular secondary calcium transients in AWC were not visible in averaged traces, the ratio of change in fluorescence to total fluorescence for individual traces were plotted in heat maps in Fig. 3 and Supplementary Figs. S1 and S4. The wild-type controls for each figure were interleaved with mutants tested over the same time period.
Individual animals were scored for exploratory behavior while in the presence of food (5 minutes), immediately after removal from food (1–12 minutes), and at long times off food (35–40 minutes)6, 9. All turns and reversals were scored by eye, by an investigator blind to the genotype of the animal. Reversals and turns were identified as described6. Results in Figs. 1 and and55 show turning rates scored 7–12 minutes after removal from food, and are reported as RevOmega values, which represent reversals coupled to omegas. Qualitatively similar results were obtained when large reversals or omega turns were scored individually (Supplementary Table 1). Data were analyzed using Perl scripts to calculate reversal and omega frequencies.
Adaptation assays were performed as previously described26. Animals were washed and plated on 3% assay agar plates. 16 µl of isoamyl alcohol was placed on agar plugs on the plate lid and the plates were sealed with parafilm (conditioning plate). After 60 min or 90 min, animals were washed and tested for chemotaxis on fresh plates. Controls were treated identically except that isoamyl alcohol was omitted from the conditioning plate. For imaging experiments, animals were conditioned for 60 min with or without isoamyl alcohol and then loaded into the device for calcium imaging.
The plots in Figs. 3d,h, ,5b,5b, and Supplementary Figure S1 and S4d were generated by transforming data in the time domain (the raw fluorescence trace y(t)) into the frequency domain (Y(f)), using the Matlab function fft (discrete Fourier transform) which computes
at 1200 equally-spaced frequencies from f0 =0 to f1199 =10 Hz, the sampling rate. Aliasing, or contribution from oscillatory components at higher than the Nyquist frequency (5Hz), is negligible based on the rapid drop-off of the Fourier transform with increasing frequency because the dynamics of calcium fluorescent indicators is slower than the sampling rate. The normalized energy spectral density, was then calculated for each trace and averaged across each group of traces at each fk sample point to give versus fk (Figs. 3d, 3h and and5b).5b). The value |Y (f0)| represents the mean of the entire time-domain vector y(t) and was excluded from the calculation of energy spectral density normalization and band power ratio (described below), although including the value in the calculations did not significantly affect results. Values of f above fN/2 were also excluded from calculations due to the redundancy of the magnitude of the second half of the discrete Fourier spectrum when performed on real-valued signals. For simplicity, no smoothing or windowing was applied to the energy spectral density. Applying various windowing schemes did not significantly affect quantifications (data not shown).
This representation of calcium signal data was chosen for two reasons: First, because calcium events were not registered in time, transients after the initial peak response were typically lost when averaging traces from multiple animals (Supplementary Figs. S2a, S2b). In contrast, the frequency-domain representation separates magnitude and phase information and secondary transients are preserved when averaging over many experiments. Second, this representation allows the identification and separation of signal components in various frequency bands. Low frequencies capture the magnitude of the primary response, high frequencies capture signal measurement noise, and middle frequencies capture secondary transients.
As an estimate of the portion of the total signal contributed by oscillatory components in a band bounded by two frequencies fmin and fmax, we compute the band power ratio as
The complete time domain trace y(t) can be exactly reconstructed from its complex Fourier spectrum via the inverse discrete Fourier transform (Matlab function ifft) which computes:
For visualizing signal contributions from each frequency band, Fourier spectra were divided into 3 frequency bands, and time-domain signals represented by the contributions from a single frequency band were reconstructed via the inverse discrete Fourier transform (Eq. 3) of the complex Fourier spectrum clipped to zero everywhere outside of the particular frequency band. Middle band reconstructions are shown in Supplementary Figures S1 and S4d. By mathematical identity, the sum of the low, middle, and high-frequency trace reconstructions equals the full time-domain fluorescence trace. Since reconstructions based on sharply clipped samples can potentially result in ringing effects (continued oscillations during a time period when the original trace is not oscillating), we also performed reconstructions using sloped frequency windows. No major changes in the appearance of trace reconstructions during the 30–120 sec period were observed (not shown).
The choice of middle frequency band range was made by separating the Fourier spectra into many bands (10 cutoffs at 0.001, 0.01, 0.02, 0.033, 0.1, 0.2, 0.5, 1, 2, 5 Hz) and selecting the bands that significantly distinguished wild type from nlp-1 responses (data not shown). We found that most nlp-1 secondary transients are captured in a narrower f = 0.033 – 0.2 Hz band, but nlp-1 traces also show increased oscillation in the f = 0.2 – 1 Hz range; hence, the combination of these bands (f = 0.033 – 1 Hz) was selected as the region of interest for Supplementary Figs. S1 and S4d. This frequency range corresponds to periodic oscillations with periods between 1 and 30 s. The different genotypes do not show significant differences in the frequency domain less than 0.033Hz or greater than 1 Hz, suggesting little effect on the primary response or signal noise.
Three of the genotypes (nlp-1, npr-11 and ins-1) showed strong secondary calcium transients. The nlp-1 energy spectral density shows a broad peak in frequency-domain amplitude at f = 0.0583 Hz (Fig. 3d), corresponding to a period of 17s; this period is similar to the time period between large calcium transients determined by a manual measurement of nlp-1 traces (19.5 ± 2.2 s (mean ± SD); Fig. 3b). Similarly, npr-11 and ins-1 animals show increased amplitude at frequencies corresponding to a period between 10 and 20 s (Fig 3f,h, Supplementary Figs. S4b,d). These transients were not observed during imaging of wild-type or nlp-1 animals during buffer exchange protocols (Supplementary Fig. S1). All calculations were performed with Matlab (7.0R14, The MathWorks, Natick, MA).
Two peptides, one corresponding to the NLP-1 sequence (MDANAFRMSFamide) and the second with a scrambled sequence (MSMRFANADFamide), were synthesized by the Proteomics resource center at The Rockefeller University. Human embryonic kidney 293 cells (HEK293) were cultured in DMEM supplemented with 10% FBS at 37°C in a humidified atmosphere containing 5% CO2. 50–60% confluent cells were transfected with a 1:1 ratio of pME18s-npr-11 and pcDNA3-α16Z (a promiscuous G-protein13, a kind gift from Prof. Y.H. Wang) using Lipofectamine 2000 and incubated for at least 24 hours; the transient transfection efficiency was ~70%, estimated by cotransfecting β2-adrenergic receptor and Gα15 plasmids and counting the fraction of cells responsive to isoproterenol. In control experiments, the receptor or G protein was transfected alone. Experiments were conducted on three plates for each condition on three different days. The transfected cells were loaded with 2.5uM fura-2/AM for 20 min at 37°C, peptide solution was applied sequentially to the cells for 15s with a peristaltic pump, and fluorescence at 510 nm by excitation at 340/380 nm was monitored by MetaFluor calcium imaging system. The calcium response trace was calculated using all cells in randomly chosen fields. For receptor and G protein transfections, ten fields were analyzed; for G protein alone, seven fields, and for receptor alone, eight fields.
In transgenic animals expressing GCaMP2.2b under an AIA-specific promoter, AWC neurons were identified based on their position and morphology using Nomarski optics, and killed with a Micropoint laser system. Operated animals were tested in parallel with control animals from the same strain on the same day.
cDNA or genomic regions corresponding to the entire coding sequences of nlp-1, npr-11, and ins-1 were amplified by PCR from mRNA or genomic DNA and expressed under cell-specific promoters as indicated. For behavioral experiments in transgenic lines, a splice leader (SL2) fused to a gfp transgene was used to confirm cell-specific expression of the gene of interest, and only animals expressing GFP were scored. Selective expression in AWC, ASI, AIA or AIY was achieved using the promoters odr-3 (AWC>AWB), str-3 (ASI alone), gcy-28.d (AIA) and ttx-3 (AIY alone). Germline transformations were carried out my microinjection of plasmids at concentrations between 10 ng/µl and 50 ng/µl. Strains were grown and maintained under standard conditions50.
We thank the C. elegans knockout consortium and the Caenorhabditis Genetic Center (CGC) for strains, Y.H. Wang for α16Z chimera, L. Looger for GCaMP2.2b, Y. Iino for discussions about ins-1, and L. Vosshall, G. Lee, E. Feinberg, M. Tsunozaki, J. Gray, J. Garrison, M. McGrath, and members of the Bargmann laboratory for critical help, advice and insights. Peptide synthesis was performed by the Proteomics Resource Center of the Rockefeller University. This work was funded by the G. Harold and Leila Y. Mathers Charitable Foundation and by the Howard Hughes Medical Institute. D.R.A. holds a Career Award at the Scientific Interface from the Burroughs Wellcome Fund. C.I.B. is an Investigator of the Howard Hughes Medical Institute.
Author contributionsS.H.C. conceived, conducted, and interpreted experiments and co-wrote the paper; S.K., D.R.A., and L.F.A. performed and interpreted data analysis; T.N. performed HEK expression experiments; C.I.B. conceived and interpreted experiments and co-wrote the paper.