Dynamic serotonin biosynthesis is important for serotonin function; however, the mechanisms that underlie experience-dependent transcriptional regulation of the rate-limiting serotonin biosynthetic enzyme tryptophan hydroxylase (TPH) are poorly understood. Here, we characterize the molecular and cellular mechanisms that regulate increased transcription of Caenorhabditis elegans tph-1 in a pair of serotonergic neurons ADF during an aversive experience with pathogenic bacteria, a common environmental peril for worms. Training with pathogenic bacteria induces a learned aversion to the smell of the pathogen, a behavioral plasticity that depends on the serotonin signal from ADF neurons. We demonstrate that pathogen training increases ADF neuronal activity. While activating ADF increases tph-1 transcription, inhibiting ADF activity abolishes the training effect on tph-1, demonstrating the dependence of tph-1 transcriptional regulation on ADF neural activity. At the molecular level, the C. elegans homolog of CaMKII, UNC-43, functions cell-autonomously in ADF neurons to generate training-dependent enhancement in neuronal activity and tph-1 transcription, and this cell-autonomous function of UNC-43 is required for learning. Furthermore, selective expression of an activated form of UNC-43 in ADF neurons is sufficient to increase ADF activity and tph-1 transcription, mimicking the training effect. Upstream of ADF, the Gqα protein EGL-30 facilitates training-dependent induction of tph-1 by functional regulation of olfactory sensory neurons, which underscores the importance of sensory experience. Together, our work elucidates the molecular and cellular mechanisms whereby experience modulates tph-1 transcription.
Serotonin plays important roles in a wide range of neural functions (Azmitia, 2007). Tryptophan hydroxylase (TPH) is the rate-limiting enzyme for serotonin biosynthesis, and transcription of TPH is highly regulated. For example, in the rat raphe nuclei, TPH transcription varies with circadian rhythm (Malek et al., 2005), and the TPH mRNA level is elevated by stress, such as repeated immobilization stress (Chamas et al., 1999) and repeated forced swimming (Shishkina et al., 2008). In addition, the TPH mRNA level in the trigeminal ganglia of female mice oscillates with the natural estrous cycle (Berman et al., 2006). These characterizations underscore the importance of TPH transcriptional regulation; however, the underlying mechanisms are mostly unknown. The mammalian serotonergic neurons are highly heterogeneous in anatomy and function (Jensen et al., 2008), posing a challenge to the molecular and cellular characterization of TPH regulation.
The Caenorhabditis elegans tryptophan hydroxylase, TPH-1, is highly homologous to mammalian TPH and is essential for serotonin biosynthesis. An expression reporter of C. elegans tph-1 is expressed in three pairs of serotonin-producing neurons, NSM, ADF, and HSN (Sze et al., 2000). The anatomical positions and circuit connections of these neurons are well defined (White et al., 1986). Therefore, C. elegans is an accessible and suitable model system to study TPH regulation.
The transcription of C. elegans tph-1 in ADF serotonergic neurons is dynamically regulated by conditions including starvation, heat stress, and food quality (Estevez et al., 2004, 2006; Zhang et al., 2005; Moussaif and Sze, 2009), consistent with the modulatory role of serotonin in behaviors that are affected by these conditions (Chase and Koelle, 2007). Our previous work on a form of aversive olfactory learning identified an experience-dependent change in tph-1 transcription in ADF. We found that, as a bacteria-feeding nematode, C. elegans learns to avoid the smell of pathogenic bacteria after ingestion. This learning process requires the serotonergic signal from ADF neurons. Intriguingly, aversive training with pathogenic bacteria increases tph-1 transcription and the serotonin content in ADF, and the ADF serotonin signal promotes learning through several interneurons (Zhang et al., 2005). These findings further demonstrated the importance of tph-1 regulation and provided a system to examine its underlying mechanisms.
Here, we found that the activity of ADF serotonergic neurons regulates tph-1 transcription. Aversive training with pathogenic bacteria enhances the sensory-evoked intracellular calcium response of ADF. While activating ADF increases tph-1 transcription, inhibiting ADF activity suppresses the training-dependent upregulation of tph-1. The C. elegans homolog of calcium/calmodulin-dependent protein kinase II (CaMKII), UNC-43, acts within ADF to regulate neuronal activity, tph-1 transcription, and learning. Selective expression of an activated form of UNC-43 in ADF is sufficient to mimic the training effect to generate increased ADF activity and tph-1 transcription. The Gqα EGL-30 acts in the upstream olfactory sensory neurons to mediate the training effect on tph-1 transcription by regulating the sensory response of these neurons. Together, these results support a model in which aversive training activates ADF serotonergic neurons via the cell-autonomous function of UNC-43, which acts together with the non-cell autonomous function of Gqα to regulate an experience-dependent increase in tph-1 transcription (see Fig. 5E,F).
Materials and Methods
The strains were cultivated and maintained under standard conditions at 20°C (Brenner, 1974). C. elegans hermaphrodites are used in this study. The expression of tph-1 was measured using GR1333 mgIs71 [tph-1::gfp, rol-6(su1006)] V (Sze et al., 2000). To measure tph-1 expression in different mutant backgrounds, all of the following strains were crossed with GR1333: MT2605 unc-43(n498n1186) IV, MT1092 unc-43(n498) IV, DA823 egl-30(ad805) I, MT1434 egl-30(n686) I, CG21 egl-30(tg26) I; him-5(e1490) V, KG1278 unc-73(ce362) I, RM2221 egl-8(md1971) V, and JT734 goa-1(sa734) I. The transgenic strains used in this work include CX10281 kyEx2373 [Pstr-2::GCaMP2; Punc-122::gfp], ZC405 yxEx274 [Psrh-142::YC3.60; Punc-122::dsred], ZC28 unc-43(n498n1186) IV; mgIs71 V; yxEx12 [Psrh-142::unc-43::mcherry; Punc-122::gfp], ZC35 unc-43(n498n1186) IV; yxEx19 [Psrh-142::unc-43::mcherry; Punc-122::gfp], ZC929 unc-43(n498n1186) IV; yxEx334 [Psrh-142::unc-43::gfp; Punc-122::gfp], ZC949 unc-43(n498n1186) IV; mgIs71 V; yxEx249 [Psrh-142::unc-43(E108K)::mcherry; Punc-122::gfp], ZC974 egl-30(ad805) I; mgIs71V; yxEx369 [Pstr-1::egl-30::mcherry; Podr-1::egl-30::mcherry; Punc-122::gfp], ZC1240 egl-30(n686) I; mgIs71 V; yxEx543[Pstr-1::egl-30::mcherry; Podr-1::egl-30::mcherry; Punc-122::gfp], ZC1145 yxEx476 [Pstr-1::GCaMP; Punc-122::gfp], ZC1493 egl-30(ad805) I; kyEx2373[Pstr-2::GCaMP2; Punc-122::gfp], ZC1512 egl-30(ad805) I; yxEx476 [Pstr-1::GCaMP; Punc-122::gfp], ZC1584 egl-30(ad805) I; yxEx476 [Pstr-1::GCaMP; Punc-122::gfp]; yxEx772 [Pstr-1::egl-30::mcherry; Punc-122::dsred], ZC1587 yxEx775 [Pstr-1::mcherry; Punc-122::dsred; KP326], ZC1588 yxEx776 [Podr-1::mcherry; Punc-122::dsred; KP326], ZC1863 yxEx944 [Psrh-142::mcherry; Punc-122::dsred; KP326], ZC947 egl-30(ad805) I; mgIs71 V; yxEx350 [Pstr-1::egl-30::mcherry; Punc-122::gfp], ZC371 egl-30(ad805) I; mgIs71 V; yxEx246 [Psrh-142::egl-30::mcherry; Punc-122::gfp], ZC1391 mgIs71 V; yxEx725[Psrh-142::NaChBac::mcherry; Punc-122::gfp], ZC1890 mgIs71V; yxEx960 [Psrh-142::twk-18(cn110)::mcherry; Punc-122::dsred], ZC2119 egl-30(ad805) I; mgIs71 V; yxEx249 [Psrh-142::unc-43(E108K)::mcherry; Punc-122::gfp], ZC2122 unc-43(n498n1186) IV; yxEx249 [Psrh-142::unc-43(E108K)::mcherry; Punc-122::gfp]; yxEx274 [Psrh-142::YC3.60; Punc-122::dsred], ZC2153 itr-1(sa73) IV; yxEx274 [Psrh-142::YC3.60; Punc-122::dsred], ZC2169 unc-2(e55) X; yxEx274 [Psrh-142::YC3.60; Punc-122::dsred], ZC2170 unc-68(e540) V; yxEx274 [Psrh-142::YC3.60; Punc-122::dsred], ZC 2175 unc-43(n498n1186) IV; yxEx12 [Psrh-142::unc-43::mcherry; Punc-122::gfp]; yxEx274 [Psrh-142::YC3.60; Punc-122::dsred]. KP326 is a plasmid that contains a Pegl-30::gfp transcriptional reporter (Lackner et al., 1999).
Generation of DNA transgenes and transgenic animals.
To generate the Psrh-142::YC3.60 construct, a 4 kb PCR product of the srh-142 promoter was cloned into Invitrogen pENTR to make an entry vector, which was used in an LR recombination reaction with a pCVG11 destination vector containing Cameleon YC3.60 and unc-54 3′UTR to generate the final construct. A PCR product containing the srh-142 promoter, YC3.60, and unc-54 3′UTR was generated from the plasmid and injected at 100 ng/μl. To generate the Psrh-142::unc-43::mcherry construct, the mcherry cDNA was used to replace the gfp sequence in pPD95.77 (a gift from Dr. A. Fire, Stanford University School of Medicine, Stanford, CA), and a 4 kb srh-142 promoter was inserted into the SphI site of the same plasmid to generate Psrh-142::mcherry. unc-43 cDNA, which was isolated from the clone yk213a11 (a gift from Dr. Y. Kohara, National Institute of Genetics, Mishima, Shizuoka, Japan), was inserted into Psrh-142::mcherry. Psrh-142::unc-43(E108K)::mcherry was constructed using a QuickChange Site-Directed Mutagenesis kit (Stratagene) from Psrh-142::unc-43::mcherry. To generate the Psrh-142::twk-18(cn110)::mcherry and Psrh-142::NaChBac::mcherry constructs, the coding region of twk-18(cn110) (a gift from Dr. E. Jorgensen, The University of Utah, HHMI, Salt Lake City, UT) or NaChBac (a gift from Drs. D. Clapham and D. Ren, Harvard Medical School, HHMI, Boston, MA) was amplified with primers engineered with SalI sites and was used to replace the unc-43 cDNA in the Psrh-142::unc-43::mcherry construct. An egl-30 cDNA was isolated from the clone yk587a10 (a gift from Dr. Y. Kohara). To generate Psrh-142::egl-30::mcherry, a srh-142 promoter fragment and a PCR product of egl-30 cDNA were inserted into a mcherry vector, which was generated by replacing the gfp sequence of pPD95.77 with a mcherry coding sequence. The transgenic animals were generated by germline transformation (Mello et al., 1991). The transgenes were injected at 25 ng/μl together with a coinjection marker, Punc-122::gfp or Punc-122::dsred (Miyabayashi et al., 1999), at 25 ng/μl, unless otherwise noted.
Aversive olfactory training.
The animals were trained based on a long-term training procedure as described previously (Zhang et al., 2005). Briefly, the embryos were extracted and placed on nematode growth medium (NGM) plates. For the naive condition, the NGM plate contained a lawn of Escherichia coli OP50. For the standard training, the NGM plate contained a lawn of Pseudomonas aeruginosa PA14 and a small lawn of OP50. For mutants with locomotory defects, L4 animals on the standard training plate were moved to another training plate that was completely covered by PA14 to ensure sufficient training.
Microscopy to measure the fluorescent intensity of the tph-1::gfp reporter.
The gfp images were collected using an Eclipse TE2000U fluorescence inverted microscope (Nikon) with a 40× NA1.3 oil-immersion objective and a CoolSNAP EZ CCD camera; the quantification was performed using NIS-element software and NIH ImageJ. The mean fluorescence intensity was generated for each neuron by subtracting the background intensity from the average fluorescence intensity of the neuron. Multiple worms were measured for each genotype.
The neuronal calcium response of ADF neurons was measured by quantifying the changes in the YFP/CFP ratio of the ratiometric calcium indicator Cameleon YC3.60 (Nagai et al., 2004). The neuronal responses of AWB and AWC were measured by quantifying the fluorescence intensity change of the calcium indicator GCaMP (Nakai et al., 2001) or GCaMP2.0 (Tallini et al., 2006). A microfluidic device (Chalasani et al., 2007; Chronis et al., 2007) was used to trap the worm and to deliver fluidic stimuli. The switch between the buffer and the bacteria-conditioned medium was controlled by a Valvebank II Perfusion System (Automate Scientific). To make the bacteria-conditioned medium, E. coli OP50 was cultured in NGM overnight until OD 600 reached 0.25–0.30. The bacterial culture was then filtered with a 0.22 μm polyethersulfone membrane filter. ND2 files of fluorescence images were collected using a CoolSNAP EZ CCD camera installed on an Eclipse Ti-U inverted microscope (Nikon) with a 40× NA 1.3 oil-immersion objective. The sampling frequency was 10 Hz. The fluorescence intensity was quantified by analyzing the ND2 files with a customized Matlab script that tracked and measured the average intensity of a designated number of brightest pixels in the selected region of interest in each image throughout the image series, which was then subtracted by a background signal of a nearby region. To correct the photo bleaching in the Cameleon signal, the signal was fitted into a bi-exponential decay function, and the corrected signal was the ratio of the recorded signal over the decay function. To calculate the change in signal, the average signal in a 4 s window before the switch of the stimuli was set as the baseline F0. The percentage change relative to F0 was calculated and plotted.
Slow-killing assays were performed using a procedure similar to the one described previously (Kawli et al., 2010). Briefly, the worms were made sterile by feeding on the pos-1 RNAi to minimize the effect of progeny hatching inside hermaphrodite bodies during the assays. The killing assays were performed on a 3.5 cm NGM plate covered with PA14 at 25°C. The worm corpses were picked out and scored every 8–10 h, and the worms that had died on the wall of the Petri dish were censored.
Aversive training enhances the sensory-evoked calcium response of ADF neurons
Previously, we have shown that training with pathogenic bacteria, such as P. aeruginosa PA14, induces olfactory aversion to the pathogen (Zhang et al., 2005). Naive animals that are never exposed to the pathogen often prefer the smell of the pathogen, whereas trained animals that have ingested the pathogenic bacteria avoid the bacterial smell (Fig. 1A,B). This form of behavioral plasticity requires the activity of the C. elegans TPH-1 in ADF serotonergic neurons, because selective expression of tph-1 in the ADF neurons of the tph-1 null mutants rescues the learning defect of the mutants. Interestingly, aversive training increases tph-1 transcription in ADF, as evidenced by the increased signal of a tph-1::gfp reporter (Sze et al., 2000) in trained animals (Fig. 1C) (Zhang et al., 2005).
To understand how tph-1 transcription in ADF neurons is regulated by training experience, we first examined the ADF neuronal activity using intracellular calcium imaging. We selectively expressed a genetically encoded calcium sensor, Cameleon YC3.60 (Nagai et al., 2004), in ADF using the srh-142 promoter and recorded the calcium response of transgenic animals in a microfluidic device (Chronis et al., 2007). We stimulated the transgenic animals with alternating fluid streams of buffer and the medium that was conditioned with the culture of E. coli OP50, the common laboratory bacterial food for C. elegans. We found that the ratio of YFP-to-CFP intensity increased when the stimulus switched from buffer to the conditioned medium, indicating increased intracellular calcium levels in response to the conditioned medium (Fig. 1D,E).
To identify the source of ADF calcium transients, we examined the effects of several mutations that disrupt either intracellular calcium release from endoplasmic reticulum or calcium influx from extracellular space. First, we imaged ADF neurons in itr-1(sa73) and unc-68(e540) mutant animals, which carried a reduction-of-function mutation in the inositol-1,4,5-trisphosphate receptor ITR-1 (Dal Santo et al., 1999) and a splicing-acceptor site mutation in the ryanodine receptor UNC-68 (Sakube et al., 1997), respectively. We quantified the sensory-evoked ADF calcium transients by calculating the time integral of the YFP/CFP ratio change during the 30 s exposure to the conditioned medium. We did not observe any obvious defect in ADF calcium responses to OP50-conditioned medium in these two mutants compared with wild-type animals (p > 0.05 for time integrals over 30 s stimulus exposure, Student's t test) (Fig. 1H,I). Because previous studies have indicated important roles of ITR-1 and UNC-68 in regulating intracellular calcium release (Maryon et al., 1996; Sakube et al., 1997; Clandinin et al., 1998; Baylis et al., 1999; Dal Santo et al., 1999; Busch et al., 2012), these results suggest that intracellular calcium stores are not likely to be the main source for the sensory-evoked calcium transients in ADF. Next, we tested ADF calcium response in unc-2(e55) animals, which carry a nonsense mutation in the C. elegans non-L-type voltage-gated calcium channel UNC-2 (Schafer and Kenyon, 1995; Estevez et al., 2004). We found that compared with wild-type animals, unc-2(e55) mutants displayed weaker ADF calcium transients in response to OP50-conditioned medium (Fig. 1J), and the integral of the calcium response during 30 s exposure of the stimulus is smaller in unc-2(e55) than in wild-type animals (p < 0.05, Student's t test). The partial defect of unc-2(e55) could be attributable to potential redundant roles played by other calcium channels. For example, the L-type voltage-gated calcium channel EGL-19 (Lee et al., 1997) has been shown to regulate sensory-evoked calcium transients in several types of sensory neurons (Suzuki et al., 2003; Hilliard et al., 2005; Frøkjaer-Jensen et al., 2006). However, we could not assess the effect of egl-19(ad1006) mutants because of weak expression of Cameleon YC3.60 in the ADF neurons of the egl-19 mutants. Both intracellular and extracellular sources of calcium contribute to the calcium responses of C. elegans neurons (Suzuki et al., 2003; Hilliard et al., 2005; Frøkjaer-Jensen et al., 2006; Busch et al., 2012; Hendricks et al., 2012). Our results together suggest that influx of extracellular calcium regulates the sensory-evoked calcium transients in ADF.
Next, we examined whether the aversive training alters ADF activity by comparing the calcium responses of ADF neurons in the naive and trained animals. We found that trained animals exhibited calcium transients for an extended amount of time, ranging from 10 to 30 s (Fig. 1D,E), in response to the conditioned medium. To quantify this change, we calculated the time integral of the YFP/CFP ratio change during the last 20 s exposure to the conditioned medium and found that PA14-trained animals displayed a significant increase compared with naive animals (Fig. 1F). In contrast, training with PAK, a nonpathogenic strain of P. aeruginosa, did not induce similar changes (p > 0.05 for time integrals of the calcium transients over the last 20 s stimulus presentation, Student's t test) (Fig. 1G), suggesting that the training-dependent enhancement in ADF activity is contingent on the aversive experience with the pathogenic bacterium PA14. Together, these results indicate that the sensory-evoked calcium response in ADF is enhanced in the PA14-trained animals, suggesting that aversive training activates ADF neurons.
Activating ADF neurons increases tph-1 transcription, whereas inhibiting ADF abolishes training-dependent induction of tph-1
Because aversive training increases both ADF neural activity and tph-1 transcription in these cells, we examined whether increased ADF activity can lead to upregulation of tph-1 transcription. First, we activated ADF neurons by selectively expressing a bacterially derived voltage-gated sodium channel, NaChBac (Ren et al., 2001), in ADF. Ectopic expression of NaChBac depolarizes and increases the excitability of several types of neurons in flies and mice (Nitabach et al., 2006; Kelsch et al., 2009). We found that compared with nontransgenic siblings (1.0 ± 0.06, intensity normalized by the signal in nontransgenic animals; n ≥ 49 animals), the transgenic animals that expressed NaChBac only in ADF neurons displayed significantly enhanced tph-1 expression in these neurons even in the absence of training (1.34 ± 0.08, intensity normalized by the signal in non-transgenic animals; p < 0.01, Student's t test; n ≥ 50 animals; mean ± SEM), thus supporting the hypothesis that increased ADF neuronal activity during aversive training induces tph-1 expression.
Next, we examined whether the training-dependent increase in tph-1 transcription depends on ADF neuronal activity. We inhibited ADF neurons by selectively expressing an activated form of potassium channel twk-18(cn110) in ADF neurons. The expression of twk-18(cn110) in Xenopus oocytes generates a potassium channel that conducts significantly larger outward currents than the wild-type channels (Kunkel et al., 2000). We found that training-induced tph-1 transcription was abolished in the transgenic animals, whereas nontransgenic wild-type siblings displayed a normal level of induction (Fig. 1K). Together, these results support our hypothesis that activation of ADF neurons during the aversive experience leads to increased tph-1 transcription.
CaMKII/UNC-43 functions in ADF neurons to regulate ADF neuronal activity, tph-1 transcription, and aversive learning
Next, we sought the molecular mechanisms whereby tph-1 transcription in ADF neurons is upregulated by training experience. Because UNC-43, the only C. elegans homolog of CaMKII (Reiner et al., 1999; Rongo and Kaplan, 1999), regulates ADF tph-1 expression under normal conditions (Zhang et al., 2004), we examined whether UNC-43 also regulates training-dependent tph-1 expression in ADF.
First, we found that the training-dependent upregulation of tph-1 transcription in ADF neurons was completely abolished in the unc-43(n498n1186) null mutants, measured by the expression of the transcriptional reporter tph-1::gfp in both naive and trained animals (Fig. 2A). In addition, the unc-43(n498n1186) mutants failed to learn to avoid the smell of PA14 after training (Fig. 2C). UNC-43 is ubiquitously expressed in the nervous system (Reiner et al., 1999; Rongo and Kaplan, 1999). We found that selective expression of a wild-type unc-43 cDNA in the ADF neurons of unc-43(n498n1186) mutants fully restored the training-dependent upregulation of tph-1 expression in ADF and partially rescued the learning defects in the mutants (Fig. 2B,C). One potential reason for the partial rescue of learning is that UNC-43 also regulates the function of motor neurons and muscles (Reiner et al., 1999; Rongo and Kaplan, 1999), and the transgenic animals with UNC-43 activity restored only in ADF were still defective in locomotion. Thus, UNC-43 functions cell-autonomously in ADF neurons to regulate the training-dependent enhancement of tph-1 transcription and aversive olfactory learning.
The mammalian CaMKII isoform αB is localized in the nucleus (Brocke et al., 1995). However, we did not observe any clear signal in the nuclei of transgenic animals that expressed a translational reporter unc-43::mcherry only in ADF neurons under either the naive or the trained condition (data not shown). The mammalian αB-CaMKII possesses a nuclear localization signal (Brocke et al., 1995) that is absent in UNC-43. This notion is consistent with the imaging results. Therefore, UNC-43 likely functions through further downstream signaling events to regulate tph-1 transcription in ADF neurons.
Next, we investigated how UNC-43 functions cell-autonomously to regulate training-dependent tph-1 transcription in ADF neurons. First, we found that tph-1 transcription in the ADF neurons of a gain-of-function mutant, unc-43(n498) (Reiner et al., 1999; Rongo and Kaplan, 1999), was significantly elevated under the naive condition (Fig. 2D), consistent with previous studies (Zhang et al., 2004). In addition, this elevated tph-1 level was not further increased by training (Fig. 2D). The unc-43(n498) mutant exhibits opposing phenotypes to the unc-43 loss-of-function mutant in both locomotion and egg laying (Reiner et al., 1999; Rongo and Kaplan, 1999; Robatzek and Thomas, 2000), and the amino acid substitution E108K in unc-43(n498) results in constitutive activation of UNC-43 that is calcium independent (Umemura et al., 2005). These results suggest the possibility that activation of UNC-43 in ADF neurons during aversive training leads to increased tph-1 expression. To further test this possibility, we selectively expressed the constitutively active form UNC-43(E108K) in the ADF neurons of the unc-43(n498n1186) null mutants and found that this ectopic expression enhanced tph-1 transcription [1.84 ± 0.14, intensity normalized by the signals in unc-43(n498n1186) mutants that expressed wild-type UNC-43 only in ADF; n ≥ 37 animals; mean ± SEM] compared with transgenic animals that expressed the wild-type UNC-43 in the ADF neurons of the same mutants (1.0 ± 0.03, normalized intensity; n ≥ 37 animals; p < 0.001, Student's t test; mean ± SEM). This enhancement strongly mimicked the training effect. The transgenic animals that expressed ADF-specific UNC-43(E108K) appeared to exhibit higher expression of tph-1 in ADF neurons than the unc-43(n498) gain-of-function mutants (Fig. 2D). This difference may result from a higher expression level of UNC-43(E108K) in the ADF neurons of the transgenic animals and/or some unidentified defects in the unc-43(n498) mutant animals. Together, these findings support that activated UNC-43 in ADF neurons increases tph-1 transcription during aversive training.
UNC-43 regulates training-dependent enhancement of ADF calcium responses
Having characterized the roles of UNC-43 and ADF neuronal activity in regulating tph-1 transcription in ADF, we examined whether UNC-43 regulates ADF activity. We performed calcium imaging on the ADF neurons of unc-43(n498n1186) null mutants (Fig. 2E,F). We found that loss of UNC-43 activity abolished training-induced increase in ADF calcium transients and that this defect was fully rescued by selective expression of wild-type UNC-43 in the ADF neurons (Fig. 2E,F). Conversely, ADF-specific expression of the activated UNC-43(E108K) isoform in the unc-43(n498n1186) null mutants is sufficient to increase ADF calcium transients, mimicking the training effect (Fig. 2G,H). Moreover, training did not further enhance ADF calcium response in these transgenic animals. Together, our results indicate that aversive training experience activates ADF serotonergic neurons via activated function of UNC-43 in ADF, which leads to upregulation of tph-1 transcription (Fig. 5E,F).
EGL-30 regulates increased tph-1 transcription during aversive training
Next, we investigated the molecular pathways in which UNC-43 acts to regulate tph-1 induction during training. Previously, it was shown that UNC-43 regulates locomotion and egg laying through a pathway of GOA-1 and EGL-30, the C. elegans Goα and Gqα, respectively (Miller et al., 1999; Robatzek and Thomas, 2000; Lee et al., 2004). To test whether UNC-43 signals in the same pathway to regulate tph-1 in ADF neurons during aversive learning, we analyzed the effect of mutations in the Goα/Gqα pathway on this process. Because mutants in this pathway are often deficient in locomotion, to ensure sufficient training we transferred trained animals from a standard training plate to a plate that was fully covered by a lawn of PA14 during the L4 larval stage and measured ADF tph-1 levels 12 h later (Fig. 1A).
First, we found that both egl-30(ad805) and egl-30(n686) (Brundage et al., 1996), two different mutations that disrupt EGL-30 function, abolished the training-dependent upregulation of tph-1 transcription in ADF (Fig. 3A,B). Surprisingly, the gain-of-function mutant egl-30(tg26) (Doi and Iwasaki, 2002; Bastiani et al., 2003) is similarly defective (Fig. 3C), suggesting that proper activity level of Gqα/EGL-30 is important in regulating tph-1 transcription. In contrast, a canonical mutant, goa-1(sa734), exhibited normal training-dependent induction of tph-1 despite an increased level of tph-1 expression under the naive condition (Fig. 3D). These results suggest that the functions of GOA-1 and EGL-30 in regulating training-dependent expression of tph-1 are different from their antagonistic roles in locomotion and egg laying and thus suggest a new mechanism of EGL-30 in regulating tph-1 transcription.
Previous studies have implicated the Rho guanine nucleotide exchange factor (GEF) domain of UNC-73 (Williams et al., 2007) and the phospholipase Cβ EGL-8 as downstream effectors of EGL-30 in regulating locomotion and egg laying (Lackner et al., 1999; Miller et al., 1999). We found that a nonsense mutation in egl-8(md1971), which produces strong defects in locomotion and egg laying (Miller et al., 1999), displayed normal training-dependent induction of tph-1 expression in ADF neurons (Fig. 3E), suggesting that EGL-30 functions mostly independently of EGL-8 to enhance tph-1 transcription during aversive training. This result is consistent with a previous finding that EGL-8 acts as a minor downstream effector or in a parallel pathway of EGL-30 to regulate egg-laying behavior (Bastiani et al., 2003). Similarly, the unc-73(ce362) mutant that contains a mutation in the Rho GEF domain was also not significantly defective in tph-1 induction (Fig. 3F), implicating an unidentified or novel effector(s) of Gqα/EGL-30 in regulating training-dependent tph-1 transcription. Together, our results indicate that UNC-43 functions together with EGL-30 to modulate tph-1 expression in ADF neurons during aversive training.
EGL-30 mediates sensory response of AWB and AWC olfactory neurons to regulate tph-1 induction during aversive training
Next, we examined the mechanisms whereby EGL-30 regulates tph-1 transcription in ADF neurons during aversive training. egl-30 is known to express widely in the nervous system (Lackner et al., 1999). First, we confirmed the expression of egl-30 in ADF neurons (Fig. 4A–C) and the major olfactory sensory neurons AWB and AWC (Fig. 4D–F) by examining the coexpression of a transcriptional reporter, KP326 (Pegl-30::gfp) (Lackner et al., 1999), with mCherry reporters expressed in these neurons. Next, we found that, different from the cell-autonomous role of UNC-43, selective expression of a wild-type egl-30 cDNA in ADF neurons did not rescue the defect of egl-30(ad805) mutants in regulating training-dependent induction of tph-1 expression in ADF (Fig. 4G). This defect and the similar defect in the other egl-30 mutant, n686, required the activity of egl-30 in both AWB and AWC sensory neurons to rescue (Fig. 4H,I). AWB synapse onto ADF with multiple chemical synapses (White et al., 1986), but expression of egl-30 in AWB alone did not rescue (Fig. 4J). Together, our results reveal a small neuronal network in which unc-43 functions cell-autonomously in ADF neurons and egl-30 functions non-cell-autonomously in the upstream sensory neurons to produce activity-dependent upregulation of tph-1 transcription in ADF (Fig. 5E,F). Consistently, ADF-specific expression of the activated UNC-43(E108K) in egl-30(ad805) mutants significantly increases ADF tph-1 expression [Student's t test, p < 0.001 between transgenic animals (3.26 ± 0.18) and nontransgenic siblings (1 ± 0.08); tph-1 signals were normalized by the signals in nontransgenic animals; n ≥ 25 animals each group; mean ± SEM], further supporting the possibility that UNC-43 acts downstream of EGL-30 in regulating tph-1 expression in ADF.
Next, we investigated the function of egl-30 in AWB and AWC sensory neurons. First, we found that expression of egl-30 in AWB and AWC neurons did not significantly alter locomotory speed (transgenic animals, 0.13 ± 0.04 pixels s−1; nontransgenic siblings, 0.27 ± 0.13 pixels s−1; Student's t test, p > 0.05; n = 5 animals each group; mean ± SEM), suggesting that egl-30 function in these neurons is not essential for locomotion. Using slow-killing assays (Kawli et al., 2010) to measure resistance to PA14 infection, we found that the activity of egl-30 in AWB and AWC also did not significantly change the susceptibility of the egl-30(n686) mutants (Fig. 5A), consistent with the previous finding on the role of intestinal EGL-30 in regulating pathogen resistance (Kawli et al., 2010).
Finally, we examined whether egl-30 mediates the sensory response of AWB and AWC, the major sensory neurons that detect repellents and attractants, respectively (Bargmann et al., 1993). The sensory ability of these two pairs of neurons is essential for worms to learn to avoid the smell of pathogenic bacteria (Ha et al., 2010), and both AWB and AWC cells are activated by the removal of medium conditioned by the worm food E. coli OP50 (Chalasani et al., 2007; Ha et al., 2010). Using intracellular calcium imaging and transgenic animals that selectively expressed the calcium-sensitive fluorescent protein GCaMP (Nakai et al., 2001) or GCaMP2.0 (Tallini et al., 2006), we found that compared with wild-type animals, egl-30(ad805) mutants displayed a significantly weaker calcium response in AWB neurons after the removal of OP50-conditioned medium (Fig. 5B). This defect was fully rescued by selective expression of egl-30 cDNA in AWB neurons (Fig. 5C). Similarly, egl-30(ad805) mutants showed a mild defect in the calcium response of AWC neurons (Fig. 5D). Together, these results indicate that EGL-30 functions in AWB and AWC neurons to regulate olfactory sensory response, suggesting an important role of sensory experience in the upregulation of tph-1 transcription during aversive training.
Regulated expression of TPH, the rate-limiting biosynthetic enzyme of serotonin, is essential for serotonergic signaling. Here, we characterize a neuronal signaling network that regulates the experience-dependent transcription of C. elegans tph-1 during aversive olfactory learning. We show that aversive training activates ADF neurons and that increased ADF neuronal activity results in enhanced tph-1 expression. This experience-dependent increase in ADF activity and tph-1 expression requires the function of UNC-43 in ADF. ADF-selective expression of an activated form of UNC-43 produces the training effect on ADF neuronal activity and tph-1 transcription in the absence of training. We also show that Gqα/EGL-30 regulates the sensory function of AWB and AWC olfactory neurons, the sensory neurons upstream of ADF, to mediate the training-dependent enhancement of tph-1 expression, implicating the importance of sensory experience. Thus, our work provides mechanistic insights into experience-dependent regulation of tph-1 expression.
L-type voltage-gated calcium channel EGL-19 regulates sensory-evoked calcium transients in several C. elegans mechanosensory neurons (Suzuki et al., 2003; Frøkjaer-Jensen et al., 2006); nociceptive neuron ASH (Hilliard et al., 2005); and oxygen-sensing neurons, such as URX (Busch et al., 2012). Although we could not assess the role of egl-19 in ADF calcium transients because of weak expression of the calcium indicator Cameleon YC3.60 in egl-19(ad1006) mutants, we have identified the role of UNC-2, a non-L-type voltage-gated calcium channel, in regulating sensory-evoked calcium response in ADF. Previous studies have shown that mutations in unc-2 decrease tph-1 expression in ADF (Estevez et al., 2004). Together, these results suggest that the UNC-2-mediated intracellular calcium level is important for tph-1 expression in ADF neurons.
Intracellular calcium levels are known to regulate serotonergic signaling. In mammals, an increase in the intracellular calcium concentration in brain tissues leads to an increase in serotonin synthesis (Elks et al., 1979; Hamon et al., 1979). Here, we found that the sensory-evoked calcium response of ADF serotonergic neurons is enhanced after aversive training and increased ADF neuronal activity generates elevated transcription of tph-1. Thus, we propose that activity-dependent regulation of tph-1 transcription is one of the mechanisms that underlie calcium-regulated serotonin production. In rats, repeated immobilization or repeated forced swimming stress elevated the mRNA level of TPH in the raphe nuclei (Chamas et al., 1999; Shishkina et al., 2008). We speculate that increased serotonergic neuronal activity may underlie the experience-dependent regulation of TPH in these cases.
Importantly, we have identified a cell-autonomous role of CaMKII/UNC-43 in regulating training-dependent activity increase in ADF neurons. We have also shown that increased ADF activity enhances tph-1 transcription. Thus, we propose that aversive training experience upregulates tph-1 transcription in ADF neurons via UNC-43-mediated neuronal activity enhancement (Fig. 5E,F). In the mammalian serotonergic system, TPH can be directly phosphorylated and activated by CaMKII (Ehret et al., 1989). Here, we show that CaMKII also enhances tph-1 transcription during aversive olfactory learning, providing a new mechanism whereby CaMKII regulates serotonin signaling. Previous studies have shown that UNC-43 is required to maintain the basal level of tph-1 transcription under normal conditions (Estevez et al., 2004, 2006; Zhang et al., 2004). Our findings reveal several novel functions of UNC-43 in regulating tph-1. First, we show that UNC-43 regulates tph-1 transcription in a way that is dependent on training experience. Second, we show that UNC-43 regulates the training-dependent change in ADF activity and tph-1 transcription in a cell-autonomous manner, linking the signal transduction of UNC-43 and tph-1 transcription within the same cells. Third, we show that ADF-specific expression of a constitutively active form of UNC-43 mimics the training effect on ADF neuronal activity and tph-1 transcription, suggesting that activation of UNC-43 in ADF is a functional consequence of training that leads to increased ADF neural activity and tph-1 expression. Meanwhile, we have also shown that increased ADF neuronal activity, measured by intracellular calcium levels, leads to increased tph-1 transcription, causally linking the UNC-43-mediated neuronal activity change with increased tph-1 transcription within ADF (UNC-43 → neuronal activity → tph-1 transcription). During hippocampal long-term potentiation, high-frequency stimulation activates postsynaptic CaMKII, which in turn alters neuronal activity by modulating the conductance and localization of ionotropic receptors (Lisman et al., 2012). We propose that training experience activates the worm CaMKII, UNC-43, in ADF neurons, which enhances ADF neuronal activity by increasing intracellular calcium levels. In the mammalian nervous system, elevated intracellular calcium levels activate kinase pathways, including those of PKA, CaMK, and RSK/MSK, which results in transcriptional changes via modulation of transcriptional regulators (Flavell and Greenberg, 2008). Our study suggests that the signal of UNC-43-mediated calcium increase in trained ADF neurons is conveyed into the nucleus to enhance tph-1 transcription (Fig. 5E,F).
CaMKII is a calcium-activated enzyme that plays an essential role in calcium signaling and regulates activity-dependent changes in a variety of neuronal functions (Lisman et al., 2002). Mutations in unc-43, which encodes the C. elegans CaMKII, generate pleiotropic defects, including abnormal locomotion, egg laying, and defecation (Reiner et al., 1999; Rongo and Kaplan, 1999; Robatzek and Thomas, 2000), and UNC-43 regulates these different behaviors through different effectors. Particularly, in locomotion and egg laying, UNC-43 acts with a Gοα (GOA-1) and a Gqα (EGL-30), which play antagonistic roles (Mendel et al., 1995; Segalat et al., 1995; Lackner et al., 1999; Miller et al., 1999; Nurrish et al., 1999; Robatzek and Thomas, 2000; Lee et al., 2004). Here, we show that GOA-1 is not essential for the training-induced tph-1 upregulation in ADF neurons, whereas both UNC-43 and EGL-30 are required for this regulation. Furthermore, before our study, it was not clear whether UNC-43 and EGL-30 act in the same cells for many of their signaling functions. Here, we show that UNC-43 plays a cell-autonomous role in ADF neurons, whereas EGL-30 plays a non-cell-autonomous role in the upstream olfactory sensory neurons, to induce tph-1 transcription during training. Thus, our results reveal another mechanism for the diverse function of the UNC-43 and EGL-30 signaling pathway.
Our study also identifies an unexpected role of Gqα/EGL-30 in mediating the sensory response of olfactory neurons. Similar to vertebrates, C. elegans olfaction also signals through GPCR and G-protein pathways. At least 20 Gα subunits are encoded in the C. elegans genome (Jansen et al., 1999). One Gα, ODR-3, plays an important role in mediating olfactory sensation of the AWB and AWC neurons (Roayaie et al., 1998), but a strong mutant of odr-3 is only partially defective in long-range avoidance of 2-nonanone, a repellent detected by AWB, suggesting potential additional G-proteins that mediate AWB sensation (Troemel et al., 1997). Because each C. elegans olfactory sensory neuron can express multiple GPCRs, more than one Gα protein may be used for signal transduction in individual neurons. In this study, we show that EGL-30 acts in AWB to mediate its sensory response and mutations in egl-30 also impair the sensory response of AWC. Therefore, it is possible that Gqα/EGL-30 is one of the Gα proteins involved in olfactory sensation of AWB and AWC. Intriguingly, the activity of EGL-30 in AWB and AWC is essential for aversive training-enhanced tph-1 expression in ADF neurons, which receive many synapses from AWB (White et al., 1986). These results implicate the importance of sensory experience during training in the upregulation of tph-1 expression. AWB and AWC may perceive either sensory inputs from a training pathogen or internal signals generated by infection; the nature of these sensory cues that elicit AWB and AWC responses and mediate tph-1 regulation will require further investigation.
This work was supported by The Esther A. and Joseph Klingenstein Fund, March of Dimes Foundation, The Alfred P. Sloan Foundation, The John Merck Fund, and NIH Grant R01 DC009852 to Y.Z. We thank the Caenorhabditis Genetics Center, which is funded by NIH Office of Research Infrastructure Programs (Grant P40 OD010440), for strains; Dr. A. Miyawaki for YC3.60 cDNA; Dr. Y. Kohara for unc-43 and egl-30 cDNAs; Dr. J. Nakai for GCaMP DNA; Dr. J. Kaplan for the egl-30::gfp construct; Drs. D. Clapham and D. Ren for the NaChBac cDNA; Dr. E. Jorgensen for the twk-18(cn110) DNA; Dr. C. Bargmann for the imaging strain CX10281; Dr. A. Samuel for an imaging Matlab script; and Sarah Zhang for making the ADF::NaChBac::mcherry construct.
The authors declare no competing financial interests.
- Correspondence should be addressed to Yun Zhang at the above address.