Neuronal calmodulin levels are controlled by CAMTA transcription factors

The ubiquitous Ca2+ sensor calmodulin (CaM) binds and regulates many proteins, including ion channels, CaM kinases, and calcineurin, according to Ca2+-CaM levels. What regulates neuronal CaM levels, is, however, unclear. CaM-binding transcription activators (CAMTAs) are ancient proteins expressed broadly in nervous systems and whose loss confers pleiotropic behavioral defects in flies, mice, and humans. Using Caenorhabditis elegans and Drosophila, we show that CAMTAs control neuronal CaM levels. The behavioral and neuronal Ca2+ signaling defects in mutants lacking camt-1, the sole C. elegans CAMTA, can be rescued by supplementing neuronal CaM. CAMT-1 binds multiple sites in the CaM promoter and deleting these sites phenocopies camt-1. Our data suggest CAMTAs mediate a conserved and general mechanism that controls neuronal CaM levels, thereby regulating Ca2+ signaling, physiology, and behavior.


Introduction
Calmodulin-binding transcription activators (CAMTAs) are a highly conserved family of CaM-binding transcription activators (Finkler et al., 2007). In plants, CAMTAs mediate transcriptional changes in response to Ca 2+ signals evoked by biotic and abiotic stress (Yang and Poovaiah, 2002;Du et al., 2009;Doherty et al., 2009;Pandey et al., 2013;Shkolnik et al., 2019). Mammals encode two CAMTA proteins, CAMTA1 and CAMTA2, respectively enriched in the brain and heart (Song et al., 2006). Loss of CAMTA1 in the mouse nervous system leads to defects in hippocampal-dependent memory formation, degeneration of cerebellar Purkinje cells and ataxia (Long et al., 2014;Bas-Orth et al., 2016). Humans heterozygous for lesions in the CAMTA1 gene exhibit a range of neurological phenotypes, including intellectual disability, cerebellar ataxia, and reduced memory performance (Huentelman et al., 2007;Thevenon et al., 2012;Shinawi et al., 2015). Mechanistically, however, little is known about the origin of these neuro-behavioral phenotypes.
CaM is a ubiquitously expressed Ca 2+ binding protein that plays a key role in transducing responses to Ca 2+ changes (Faas et al., 2011;Baimbridge et al., 1992). Ca 2+ -CaM modifies a host of neuronal functions, including signal transduction, ion currents, vesicle fusion, learning and memory, metabolism, and apoptosis (Hoeflich and Ikura, 2002;Berchtold and Villalobo, 2014), by regulating dozens of binding targets including the CaM kinases, calcineurin, and diverse ion channels (Wayman et al., 2008;Saimi and Kung, 2002). CaM levels are thought to be limiting compared to the combined concentration of Ca 2+ -CaM binding proteins (Sanabria et al., 2008), and relatively small changes in CaM levels are predicted to impact Ca 2+ -CaM regulation of downstream targets (Pepke et al., 2010). What mechanisms regulate neuronal CaM levels is, however, unclear. We identify CAMTA as a key regulator of CaM expression in multiple neuron types, and in both Caenorhabditis elegans and Drosophila, and suggest that it is a general and conserved regulator of Ca 2+ /CaM signaling in nervous systems.

CAMT-1 functions in neurons to regulate multiple behaviors
Most natural isolates of C. elegans feed in groups. By contrast, the standard C. elegans lab strain, N2, feeds alone, due to a gain-of-function mutation in a neuropeptide receptor called NPR-1 (de Bono and Bargmann, 1998). Using npr-1(ad609) null mutants of the N2 strain (denoted as npr-1 throughout this manuscript), which aggregate strongly (Figure 1-figure supplement 1A), we performed a forward genetic screen for genes required for aggregation (Chen et al., 2017). The screen identified multiple aggregation-defective strains with mutations in camt-1, the sole C. elegans CAMTA (Figure 1-figure supplement 1A).
Aggregation is closely linked to escape from normoxia (21% O 2 ) (Busch et al., 2012;Rogers et al., 2006;Gray et al., 2005). Normoxia elicits rapid movement in npr-1 animals, which is rapidly suppressed when O 2 levels drop ( Figure 1A). Since aggregating animals create a local low O 2 environment, due to aerobic respiration, an animal encountering a group from normoxia switches from fast to slow movement, thereby staying in the group. camt-1 mutants showed defective responses to O 2 stimuli. Compared to npr-1 controls, animals from a mutant strain isolated in the screen, camt-1(db973); npr-1, which harbors a premature stop codon in CAMT-1 (Q222*), were hyperactive in 7% O 2 , and showed reduced arousal when switched from 7% to 21% O 2 ( Figure 1A-B). A deletion (The C. elegans Deletion Mutant Consortium, 2012) that removed 451 residues of CAMT-1, camt-1(ok515), conferred similar defects ( Figure 1A-B). A fosmid transgene containing a wild-type (WT) copy of the camt-1 genomic locus rescued camt-1 mutant phenotypes, restoring fast movement at 21% O 2 , and slow movement at 7% O 2 ( Figure 1C). These results indicate that CAMT-1 is required for C. elegans to respond appropriately to different O 2 levels.
In mice, humans, and flies, CAMTA transcription factors are expressed in many brain regions (Huentelman et al., 2007;Bas-Orth et al., 2016;Sato et al., 2019;Long et al., 2014). We generated a fosmid-based reporter to map the expression pattern of the longest isoform of C. elegans CAMTA, CAMT-1a. This fluorescent reporter was functional, as it rescued the behavioral defects of camt-1 mutants ( Figure 1C), and revealed that CAMT-1 was expressed broadly and specifically in the nervous system ( Figure 1D). We observed CAMT-1 expression in sensory neurons with exposed ciliated endings, motor neurons of the ventral cord, the URX O 2 -sensing neuron, and URX's post-synaptic partner, the RMG hub interneurons (Figure 1-figure supplement 2). camt-1's broad expression prompted us to ask if camt-1 mutants display pleiotropic behavioral phenotypes. We asked whether CAMT-1 is required for other aversive behaviors, such as avoidance of CO 2 , or for chemoattraction to odors and salts. In response to a rise in CO 2 , WT control (N2) worms transiently perform omega turns, W-shaped body bends that re-orient the animal away from the stimulus (Bretscher et al., 2008). camt-1 mutants exhibited abnormally high levels of W-turns without a CO 2 stimulus and a prolonged increase in W-turns in response to a rise in CO 2 ( Figure 1E). C. elegans avoids CO 2 but is attracted toward salt and a range of volatile compounds (Ward, 1973;Bargmann et al., 1993). Chemotaxis toward NaCl and odorant attractants such as benzaldehyde and diacetyl was reduced in camt-1 mutants, and these defects were rescued by a fosmid transgene containing WT CAMT-1 ( Figure 1F). Taken together, these data show that CAMT-1 function is important for multiple C. elegans behaviors.
Many deleterious human alleles of CAMTA1 alter the CG-1 DNA-binding domain (Thevenon et al., 2012). To assess the importance of the putative DNA-binding domain of CAMT-1, we used CRISPR-Cas9 to engineer mutations in conserved residues of the CG-1 domain (Figure 1figure supplement 1B). These mutants showed defects in aggregation and in their response to O 2 , recapitulating phenotypes of the camt-1 deletion mutants described above ( Figure 1G, Figure 1figure supplement 1B). These results suggest that CAMT-1 binding to DNA is essential for its function, at least for O 2 escape behavior.
We targeted CAMT-1 cDNA expression to different subsets of neurons in the neuronal circuit regulating the response to O 2 , to find out where CAMT-1 is required to promote aerotaxis. O 2 is sensed mainly by the sensory neurons URX, AQR, and PQR, and tonic signaling from URX to RMG drives high locomotory activity at 21% O 2 (Busch et al., 2012;Zimmer et al., 2009). Selectively expressing CAMT-1 to the RMG hub interneurons, but not O 2 sensing neurons, rescued the fast movement at 21% O 2 of camt-1 mutants (Figure 1-figure supplement 1D-E). The defective response of camt-1 mutants to 7% O 2 was not rescued by expressing CAMT-1 in RMG, or by simultaneous expression in RMG and O 2 -sensing neurons (Figure 1-figure supplement 1D  The domain organization of CAMT-1, highlighting camt-1 loss of function mutations used in this study. (C) A WT copy of the camt-1 genomic locus rescues the O 2 -response defects of camt-1(db973) mutants. (D) CAMT-1a::GFP driven from its endogenous regulatory sequences in a recombineered fosmid is expressed widely in the nervous system. (E) camt-1(db973) mutants exhibit an increased turning frequency both in the presence and absence of a CO 2 stimulus. Assays were performed in 7% O 2 . (F) camt-1(ok515) mutants show defects in chemotaxis to NaCl, benzaldehyde (Benz), and diacetyl (DI), which can be rescued by expressing a WT copy of CAMT-1. Colored bars indicate the mean and error bars indicate the SEM. (G) The O 2 -response defects of mutants harboring amino acid substitutions in the CG-1 DNAbinding domain (db1258, db1259, and db1260 alleles; see also  a model in which CAMT-1 acts in multiple neurons. As expected, pan-neuronal expression rescued camt-1 mutant phenotypes, and expression of the isoform a alone (CAMT-1a) was sufficient for rescue ( Figure 2A).
CAMTA transcription factors bind and can be regulated by CaM (Yang and Poovaiah, 2002;Du et al., 2009;Doherty et al., 2009;Pandey et al., 2013;Shkolnik et al., 2019). Ca 2+ -CaM dependent changes in gene expression are known to be important for both the development and function of the nervous system (West et al., 2002;Chin and Means, 2000). To test whether CAMT-1 activity is essential during development, we expressed CAMT-1 cDNA from a heat-shock-inducible promoter. Without heat-shock, this transgene did not rescue the hyperactivity phenotype of camt-1 mutants ( Figure 2B). By contrast, inducing CAMT-1 expression in the last larval stage/young adults rescued the aggregation (data not shown) and speed response defects, albeit not completely ( Figure 2C), suggesting that CAMT-1 can function in adults post-developmentally to regulate behavioral responses to ambient O 2 .

CAMT-1 dampens Ca 2+ responses in sensory neurons
To test whether disrupting camt-1 altered physiological responses to sensory cues we used Yellow Cameleon (YC) Ca 2+ sensors to record stimulus-evoked Ca 2+ changes in the URX O 2 -sensor, and in the BAG and AFD neurons, which respond to CO 2 . BAG drives omega turns when CO 2 levels rise (Bretscher et al., 2011;Hallem and Sternberg, 2008). Expressing YC sensors in these neurons did not alter the response of animals to O 2 or CO 2 (Figure 3-figure supplement 1A-C). We found that baseline Ca 2+ and stimulus-evoked Ca 2+ responses in URX, BAG, and AFD neurons were significantly elevated in camt-1 mutants across all the O 2 /CO 2 conditions we tested ( Figure 3A . These data suggest that CAMT-1 activity somehow dampens the Ca 2+ responses of these sensory neurons. We obtained similar results for Ca 2+ measurements in BAG using a Ca 2+ reporter, TN-XL (Bazopoulou et al., 2017;Mank et al., 2006), which uses chicken troponin C instead of CaM to bind Ca 2+ (Figure 3-figure supplement 1E-F). We observed the converse phenotype, reduced Ca 2+ baselines and responses, when we overexpressed CAMT-1 cDNA specifically in O 2 sensors or in BAG neurons of control animals ( Figure 3C-D). Overexpressing CAMT-1 slightly reduced expression from the gcy-37 promoter we used to express YC in O 2 sensors, as measured using a gcy-37p::gfp reporter (Figure 3-figure supplement 1G). Although we cannot completely exclude that this contributes to the reduced baseline YFP/CFP ratio, we note that  cameleon is a ratiometric sensor. Taken together, our results suggest that camt-1 regulates the excitability of sensory neurons.
Calmodulin is one of only two genes whose expression is regulated by CAMT-1 across all neuronal types profiled To identify downstream targets of CAMT-1, we compared the transcriptional profiles of multiple neural types in camt-1; npr-1 and npr-1 control animals (Kaletsky et al., 2018). We separately profiled the O 2 -sensors URX/AQR/PQR, the RMG interneurons, the AFD thermosensors, and the BAG O 2 /CO 2 sensors. We collected the neurons using FACS from strains in which they were labeled with GFP, and performed 4-10 biological replicates for robust statistical power. Analysis of the data revealed altered expression of many genes, with most changes being neural-type specific ( Figure 4A, Supplementary files 1 and 2). A striking exception was cmd-1 (calmodulin-1), encoding C. elegans CaM. cmd-1 was one of only two genes whose expression was reduced in all four neural profiles relative to WT controls. The other gene, Y41C4A.17, has no known homolog in mammals.   Most camt-1-dependent gene expression changes in O 2 sensing neurons are associated with altered neural activity Altered Ca 2+ signaling can drive changes in neuronal gene expression (Yap and Greenberg, 2018). This prompted us to investigate if the altered Ca 2+ signaling we observed in camt-1 mutants contributed to the altered gene expression. To address this, we focused on the URX/AQR/PQR O 2 sensors, which showed the altered expression of 2370 genes in camt-1 mutants. Our profiling experiments were carried out in normoxia, when these neurons exhibit tonic high Ca 2+ levels due to sustained cGMP signaling mediated by a heterodimeric soluble guanylate cyclase composed of GCY-35 and GCY-36 subunits, which binds and is activated by O 2 (Zimmer et al., 2009;Couto et al., 2013). Disrupting GCY-35 or GCY-36 abolishes the O 2 response and causes these neurons to have a constitutive low baseline Ca 2+ (Zimmer et al., 2009). We therefore compared the number of genes differentially regulated in URX/AQR/PQR neurons that we isolated and sorted from gcy-35; gcy-36; npr-1 and gcy-35; gcy-36; npr-1; camt-1 mutant animals. We only observed 108 differentially regulated genes between these genotypes, a dramatic decrease from the 2370 genes we observed when we compared the same neurons between npr-1 and npr-1; camt-1. Out of the 108 genes, 33 genes are common across the two sets of comparisons ( Figure 4B). Sorting these 33 genes in decreasing order of expression (Table in Figure 4B), we found that they included cmd-1 and Y41C4A.17, the two genes regulated by camt-1 in all neuronal types we profiled. These results support the hypothesis that most of the genes expression changes we observe in O 2 sensing neurons in camt-1 mutants are due to altered Ca 2+ signaling rather than direct control by CAMT-1, but that cmd-1, encoding CaM, is an exception.
Supplementing CMD-1 in the nervous system using the lowest expressing rab-3p::cmd-1::SL2:: mCherry line also restored normal chemotaxis toward salt, benzaldehyde, and diacetyl in camt-1 mutants ( Figure 4E), and rescued the hyperexcitability defects in URX and BAG neurons of camt-1 mutants ( Figure 4F-G). By contrast, deleting the entire coding region of Y41C4A.17 did not affect aggregation of npr-1 animals (data not shown). Our data suggest that reduced CMD-1 expression accounts for camt-1 Ca 2+ signaling and behavioral defects (see also below).

CAMTA promotes CaM expression in Drosophila melanogaster
Fly mutants of CAMTA show slow termination of photoresponses compared to WT controls (Han et al., 2006), and also exhibit defects in male courtship song (Sato et al., 2019). An allele of the Drosophila CaM gene that deletes part of the promoter and reduces CaM expression also shows slow termination of photoresponses (Scott et al., 1997). This phenotypic similarity, and our findings in C. elegans, prompted us to ask if CAMTA promotes CaM expression in flies too. We obtained two characterized alleles of Drosophila CAMTA (dCAMTA), tes 2 and cro, which respectively contain an L1420Stop mutation and a transposon insertion (Han et al., 2006;Sato et al., 2019). tes 2 mutants showed a modest decrease in dCAMTA mRNA level, suggesting that the premature stop late in the protein does not induce mRNA degradation ( Figure 5-figure supplement 1). The level of dCAMTA mRNA was strongly reduced in cro mutants as reported previously (Sato et al., 2019; Figure 5-figure supplement 1). We assessed the levels of CaM mRNA and CaM in the heads of dCamta mutant flies using quantitative RT-PCR and Western blots. Each method reported significant decreases in CaM expression compared to controls in both tes 2 and cro mutant flies ( Figure 5A-C). Moreover, immunostaining dissected retinas from cro mutants showed reduced CaM expression in rhabdomeres ( Figure 5D-E). These results suggest that the transcriptional upregulation of neuronal CaM by CAMTA is conserved from worms to flies.
CAMT-1 directly regulates CMD-1/CaM transcription through multiple binding sites at the cmd-1/CaM promoter To test whether CAMT-1 directly regulates C. elegans CaM expression by binding the cmd-1 promoter, we performed chromatin immunoprecipitation sequencing (ChIP-seq) using a CRISPR-knockin CAMT-1a::GFP strain. Our analysis revealed about 200 loci that were significantly enriched in CAMT-1a::GFP pulldowns compared to input, and to a mock pulldown (Supplementary file 3). At the top of the list was cmd-1: we observed three peaks at~6.3 kb, 4.8 kb, and 2.2 kb upstream of the CMD-1 translation start site in the CAMT-1a::GFP pulldown experiments ( Figure 6A, Figure 6figure supplement 1A). We called these peaks A, B, and C, respectively. Thus, CAMT-1 is recruited to multiple sites upstream of cmd-1. A CAMT-1 binding peak was also found in the promoter region of Y41C4A.17, the only other gene whose expression was reduced in all the neurons profiled from camt-1 mutants ( Figure 6-figure supplement 1B).
To test whether the CAMT-1 ChIP-seq peaks in the cmd-1 promoter region regulated CMD-1 transcription, we generated CRISPR strains that deleted one or more of these peaks. A strain harboring 110 bp and 136 bp deletions at peaks B and C, respectively ( Figure 6A-B, db1275), and a strain harboring a 200 bp deletion at peak A ( Figure 6A-B, db1280) exhibited aggregation and O 2 escape responses similar to npr-1 mutant controls ( Figure 6B). However, a strain harboring all three deletions ( Figure 6A-B, db1278) exhibited strong aggregation defects (Figure 1-figure supplement  1A) and defects in the locomotory responses to O 2 that mirrored those of camt-1 loss-of-function mutants ( Figure 6B, Figure 1A). Notably, the hyperactivity at 7% O 2 of db1278 mutants could be rescued by expressing additional CMD-1 in the nervous system. Like camt-1(ok515) mutants, cmd-1 (db1278) mutants also showed chemotaxis defects toward salt, benzaldehyde, and diacetyl that could be rescued by supplementing neuronal expression of CMD-1 (compare Figures 4D and  6C). These results suggest that CAMT-1 binds multiple sites in the CMD-1 promoter and acts redundantly at these sites to promote neuronal CaM expression.

Calmodulin can inhibit its own expression via CAMT-1
CaM is a key regulator of neural function. We speculated that CMD-1/CaM might homeostatically regulate its own expression via a negative feedback loop. To investigate this hypothesis, we built a transcriptional reporter for cmd-1 by fusing the 8.9 kb DNA fragment immediately upstream of the CMD-1 translational start site to sequences encoding GFP. This reporter showed strong fluorescence expression in neurons and muscle, including pharyngeal muscle (Figure 7-figure supplement 1A). We next introduced this reporter (cmd-1p::gfp) into a C. elegans line that overexpressed CMD-1/ CaM in neurons, using the rab-3 promoter (rab-3p::cmd-1), and measured neuronal GFP fluorescence in single (cmd-1p::gfp) and double (cmd-1p::gfp+rab-3p::cmd-1) transgenic animals. We normalized expression using pharyngeal GFP levels. Animals expressing rab-3p::cmd-1 reduced neuronal expression of GFP from the cmd-1p::gfp reporter. These data suggest that the high levels of CMD-1 can repress expression from the cmd-1 promoter (Figure 7-figure supplement 1B). To examine if this repression is achieved via CaM binding to CAMT-1, we introduced into the double transgenic background a camt-1 allele that disrupts the 4 IQ domains, noted as camt-1(4IQ*) ( Figure 1-figure supplement 1B). In this allele, codons encoding the conserved isoleucine residues in the four putative IQ domains of CAMT-1 were mutated to codons that encode asparagines. The camt-1(4IQ*) allele did not disrupt the O 2 -avoidance behaviors of npr-1 mutant animals (Figure 7figure supplement 1C), suggesting that CaM binding to CAMT-1 via the IQ motifs is not essential for O 2 escape behavior. By contrast, we found that camt-1(4IQ*) animals expressing cmd-1p::gfp  and rab-3p::cmd-1 showed neuronal GFP levels similar to those found in control animals lacking the rab-3p::cmd-1 transgene (Figure 7-figure supplement 1B). These data suggest that CMD-1/CaM can negatively regulate its own expression by binding the IQ domains of CAMT-1. Thus, CAMT-1 may not only activate cmd-1 expression, but also repress it when available CMD-1/CaM levels are high (Figure 7).

Discussion
We find that neuronal levels of CaM, a key mediator of Ca 2+ signaling, are controlled by the CaMbinding transcriptional activator CAMTA in both C. elegans and Drosophila. Reduced CaM levels appear to explain the pleiotropic phenotypes of C. elegans camt-1 mutants. First, camt-1  Profiling four different C. elegans neurons from camt-1 mutants and WT controls using FAC sorting and RNA Seq shows that CAMT-1 stimulates CaM expression in each of the four neurons. These results, together with the observation that CAMT-1 is expressed in most or all C. elegans neurons, suggest that CAMT-1 is part of a general mechanism that regulates CaM levels throughout the nervous system.
The RNA Seq experiments reveal a 2.5Â-4Â reduction in CaM mRNA levels in camt-1 mutants, depending on neuron type. These relatively small decreases in CaM mRNA are, however, associated with striking alterations in the stimulus-evoked Ca 2+ responses of each neuron. These findings suggest neural function is sensitive to quite small changes in CaM transcription. CaM levels may therefore provide a sensitive point of regulation of neural physiology. The increase in neuronal Ca 2+ levels we observe in the sensory neurons of camt-1 mutants could simply reflect a decrease in Ca 2+ buffering by CaM. An alternative explanation for the Ca 2+ imaging phenotypes is that reducing CaM levels disrupts the regulation of Ca 2+ /CaM's myriad binding partners. Previous work has identified multiple Ca 2+ -CaM feedback loops regulating C. elegans sensory responses, mediated for example by calcineurin/TAX-6 ( Kuhara et al., 2002), CaM kinase I/CMK-1 (Satterlee et al., 2004), and PDE1/PDE-1 (Couto et al., 2013). In addition, work in vertebrates (Saimi and Kung, 2002) has shown that CaM regulates the activity of cyclic nucleotide-gated ion channels and the L-type-Ca 2+ channel, which contribute to the Ca 2+ responses of these C. elegans sensory neurons. Further experiments are required to understand in mechanistic terms how altered CaM levels alter Ca 2+ signaling in camt-1 mutants.
Profiling of O 2 sensors revealed that many genes showed altered expression in camt-1 mutants compared to controls. Our analysis of mutants that abolish O 2 -evoked Ca 2+ responses in these neurons shows that most of these expression changes are linked to increased Ca 2+ levels in camt-1 mutants, rather than loss of CAMT-1 per se. This is consistent with the known role of Ca 2+ in regulating neuronal transcription (Yap and Greenberg, 2018). Our ChIP-seq studies identified CMD-1 as one of the major direct targets of CAMT-1. While binding motif analysis of the ChIP-seq data using prediction tool MEME did not find hits that coincide with CAMT-1 binding sites at the cmd-1 promoter(data not shown), we note that there are four mouse CAMTA1 binding motifs (Long et al., 2014;Long et al., 2009) overlapping with the CAMT-1 binding peaks of the cmd-1 promoter.
CAMTA regulates CaM expression not only in C. elegans but also in Drosophila. Mutations in the sole Drosophila CAMTA, dCAMTA, cause an approximately two fold reduction in CaM mRNA and protein in the Drosophila head. These results suggest that the regulation of CaM expression by CAMTA proteins is conserved across phylogeny. Conservation may extend beyond metazoa, as in Arabidopsis, CAMTA3/AtSR1 binds in vitro to the promoter of CaM2, although whether this regulates CaM2 expression in vivo is unknown (Yang and Poovaiah, 2002).
Like CAMT-1, dCAMTA is expressed broadly in the nervous system (Sato et al., 2019). Previous work found that dCAMTA mutants have defective termination of photoresponses (Han et al., 2006). A separate study showed that a promoter mutation in the fly CaM gene that reduces CaM expression also disrupts photoresponse termination in Drosophila photoreceptors (Scott et al., 1997). Since dCAMTA mutants show reduced levels of CaM in photoreceptors (although not to the same extent as the promoter mutation), part of the photoresponse termination defect in these animals may reflect reduced levels of CaM. More generally, it would be interesting to ask if supplementing neuronal CaM levels can rescue the dCAMTA behavioral phenotypes.
Mammals encode two CAMTA genes, CAMTA1 and CAMTA2. CAMTA1 is expressed broadly in both the mouse and human nervous systems. Homozygous mice and heterozygous human patients bearing mutations in CAMTA1 exhibit pleiotropic behavioral phenotypes, including memory defects and neurodegeneration (Han et al., 2006;Sato et al., 2019;Long et al., 2014;Bas-Orth et al., 2016;Thevenon et al., 2012;Huentelman et al., 2007). Our work raises the possibility that these defects are functionally associated with a reduction in CaM expression (Zalcman et al., 2018;Wayman et al., 2008). CAMTA2 is expressed in cardiomyocytes, and is implicated in promoting cardiac growth: overexpressing CAMTA2 in the mouse heart leads to cardiac hypertrophy (Song et al., 2006). Selectively overexpressing CaM in the mouse heart also induces cardiac hypertrophy, by a calcineurin-dependent mechanism (Obata et al., 2005). It would be interesting to ask if the cardiac hypertrophy in CAMTA2 overexpressing mice reflects increased CaM levels.
While CAMTAs were initially characterized as transcriptional activators, they have also been shown to mediate transcriptional repression (Du et al., 2009;Kim et al., 2017;Sun et al., 2020). Our data suggest CAMT-1 not only promotes CaM expression in C. elegans neurons, but can also inhibit it when available CaM levels are high, by a feedback loop in which CaM regulates its own transcription by binding to IQ domains of CAMT-1. These data suggest CAMT-1 can play a homeostatic role in regulating CaM levels (Figure 7). Mutant analyses in plants and flies have already suggested that CaM binding regulates CAMTA activity (Du et al., 2009;Nie et al., 2012;Kim et al., 2017;Choi et al., 2005). Our data suggest that binding to CaM converts CAMT-1 from an activator to a repressor. However, more data are required to establish if this feedback is relevant under physiological conditions. The absence of an obvious behavioral phenotype in mutant animals in which CAMT-1's four IQ motifs have been disrupted suggests that native CaM levels may simply not be high enough in the circuits we have studied to evoke negative feedback regulation of CaM expression.
In summary, our data suggest that we have discovered a general and conserved mechanism by which neurons control levels of CaM using CAMTA, a transcription factor that is expressed broadly in the nervous system across Metazoa. Toggling CAMT-1, the C. elegans CAMTA, up and down, can change neural excitability, circuit function, and behavior. We speculate that the activity of CAMTA transcription factors is regulated in response to upstream signals, and provides a mechanism to alter CaM levels and thereby modulate neural excitability and behavior.

Materials and methods
No statistical methods were used to predetermine sample sizes. The sample size and replicate number were similar to or greater than that used in previously published papers (behavior assays, Ca 2+ imaging) or in the scientific literature (RNA-seq, ChIP-seq, Western blot, and qPCR). The experiments were not randomized. This work used only biological replicates (biologically distinct samples that capture random biological variation) but not technical replicates (repeated measurements from the same sample).

Molecular biology
We obtained a clone containing the camt-1 locus from the C. elegans fosmid library (Source BioScience). To insert GFP immediately prior to the termination codon of camt-1 we followed established protocols (Tursun et al., 2009). The primers used to amplify the recombineering cassette from pBALU1 were: ATCATCCATGGGACCAATTGAAACCGCCGTATGGTTGCGGAACACTTGCAATGAG TAAAGGAGAAGAACTTTTCAC and aaaccaataaaaaaaatcggcatcttctaaaagtgacaccggggcaaTTATTTG TATAGTTCATCCATGCCATG. To generate transgenic lines, we injected a mix of 50 ng/ml fosmid DNA and 50 ng/ml co-injection marker (unc-122p::dsRED).

Behavioral assays
O 2 -and CO 2 -response assays were performed as described previously (Flynn et al., 2020), using young adults raised at RT. 15-30 young adults were assayed in a microfluidic PDMS chamber on an NGM plate seeded with 20-50 ml OP50. The indicated O 2 /CO 2 mixtures (in nitrogen) were bubbled through H 2 O and pumped into the PDMS chamber using a PHD 2000 Infusion Syringe Pump (Harvard Apparatus). Videos were recorded at two fps using FlyCapture software (FLIR Systems), and a Point Gray Grasshopper camera mounted on a Leica MZ6 microscope. Custom MATLAB software (Zentracker: https://github.com/wormtracker/zentracker, Laurent et al., 2015) was used to measure speed and omega turns.
Chemotaxis assays were performed as previously described (Bargmann et al., 1993) with minor modifications. 9 cm assay plates were made with 2% Bacto Agar, 1 mM CaCl 2 , 1 mM MgSO 4 , and 25 mM K 2 HPO 4 pH 6. Test and control circles of 3 cm diameter were marked on opposite sides of the assay plate, equidistant from a starting point where >50 animals were placed to begin the assay. For olfactory assays, 1 ml odorant (Benzaldehyde 1/400 or Diacetyl 1/1000 dilution in ethanol) or 1 ml ethanol, and 1 ml 1M NaN 3 , were added to each circle. For gustatory assays, an agar plug containing 100 mM NaCl was added the night before to the assay plates and removed prior to assay. Assays were allowed to proceed for 30-60 min, after which point plates were moved to 4˚C, to be counted later. The chemotaxis index was calculated as (number of animals in test circleÀnumber of animals in control circle)/total number of animals that have left the starting area.

Heat-shock
Animals were raised at 20˚C to reduce leaky expression from the hsp-16.41 heat-shock promoter. To induce heat-shock, parafilm-wrapped plates were submerged in a 34˚C water bath for 30 min, and then recovered at 20˚C for 10 hr.

Ca 2+ imaging
Neural imaging was performed as previously described (Flynn et al., 2020), with a 2Â AZ-Plan Fluor objective (Nikon) on a Nikon AZ100 microscope fitted with ORCA-Flash4.0 digital cameras (Hamamatsu). Excitation light was provided from an Intensilight C-HGFI (Nikon), through a 438/24 nm filter and an FF458DiO2 dichroic (Semrock). Emission light was split using a TwinCam dual-camera adapter (Cairn Research) bearing a filter cube containing a DC/T510LPXRXTUf2 dichroic and CFP (483/32 nm) and YFP (542/27) filters. We acquired movies using NIS-Elements (Nikon), with 100 ms or 500 ms exposure time. YFP/CFP ratios in URX were reported by YC2.60 driven from the gcy-37 promoter, in BAG by YC3.60 and TN-XL driven from the flp-17 promoter, in AFD by YC3.60 driven from the gcy-8 promoter.
Single-neuron-type cell sorting and RNA sequencing We used C. elegans lines in which neuronal types were labelled by expressing GFP under specific promoters: oxygen sensing neurons (gcy-37p), BAG (flp-17p), RMG (combination of ncs-1p::CRE and flp-21::loxP::STOP::loxP::GFP; Macosko et al., 2009), and AFD (gcy-8p). These markers were crossed into either npr-1(ad609) or npr-1(ad609); camt-1(ok515) backgrounds. C. elegans cells were dissociated and GFP-labelled neurons were sorted as described previously (Kaletsky et al., 2018). Briefly, C. elegans with GFP-labelled neurons were synchronized using the standard bleaching protocol 3 days before the cell sorting and the eggs were placed on 90 mm rich NGM plates (7.5 g peptone/liter) seeded with OP50. For each sample, we used >50,000 worms. Worms were washed three times with M9, prewashed, and then incubated for 6.5 min with 750 ml lysis buffer (0.25% SDS, 200 mM DTT, 20 mM HEPES pH 8.0, and 3% sucrose). The worms were then rapidly washed five times with M9. We dissociated the cells by adding 500 ml of Pronase (Roche) 20 mg/ml and by either pipetting up-and-down or stirring continuously for 12 min using a small magnetic stirrer. The pronase was inactivated by adding 500 ml of phosphate-buffered saline (PBS)+2% fetal bovine serum (FBO) (Gibco). The solutions were passed through a 5 mm pore size syringe filter (Millipore), and filtered cells were further diluted in PBS+2% FBS for sorting using a Sony Biotechnology Synergy High Speed Cell Sorter. Gates for detection were determined using cells prepared in parallel from nonfluorescent animals using the same protocol. An average of 3000 cells was collected for each library, and sorted directly into lysis buffer containing RNAse inhibitor (NEB E6420). cDNA libraries were made from RNA using NEB's Next Single Cell/Low Input RNA Library Prep Kit for Illumina (NEB E6420). Libraries were sequenced on an Illumina HiSeq 4000 with single-end reads of 50 bases.

Confocal microscopy and image analysis
Young adult worms were mounted for microscopy on a 2% agar pad in 1 M sodium azide. Image analysis and fluorescence quantification were carried out using Fiji (ImageJ, Wayne Rasband, NIH). The expression pattern of CAMT-1(fosmid)-GFP was imaged as previously described (Flynn et al., 2020) on an Inverted Leica SP8 confocal microscope using a 63Â/1.20 N.A. water-immersion objective. Lines expressing a cmd-1 transcriptional reporter (cmd-1p::gfp) and a red neuronal marker (either rab-3p::mCherry or rab-3p::cmd-1::SL2::mCherry) were imaged on an LSM800 inverted microscope (Zeiss) using a 63x/1.40 N.A. oil-immersion objective. The region between the two pharyngeal bulbs (Figure 7-figure supplement 1A) was imaged using stacks with a step size of 0.3 mm. A 3 mm section (10 images) around the middle of the pharynx was projected using the maximum projection method. Neurons were identified by thresholding the intensity of the red marker (mCherry). The neuronal regions overlapping with the pharynx or body wall muscles were excluded. The relative fluorescence in (Figure 7-figure supplement 1B) was defined as the GFP level in neurons minus background fluorescence divided by the level of fluorescence in the pharynx (metacorpus+isthmus +terminal bulb) minus background fluorescence.
Images of fly retinae were acquired using a Zeiss LSM800 microscope with a 20Â objective. Only retinae oriented so that the long axis of the rhabdomeres was visible were selected for quantitative analysis. A representative region of the image, as shown in Figure 5D1' and D2', was thresholded to segment the rhabdomeres, and the mean fluorescence intensity was measured, corrected to the background fluorescence, and plotted.

Chromatin immunoprecipitation sequencing
The ChIP-seq protocol used is described in Wormbook (http://www.wormbook.org/chapters/www_ chromatinanalysis/chromatinanalysis.html). Briefly, mixed-stage worms were grown in liquid culture, harvested, washed three times in PBS, and resuspended in PBS+Protease Inhibitor (PI, Sigma-Aldrich). Worm 'popcorn' was prepared by dripping worm solution into liquid nitrogen, and then hand ground to a fine powder. For each ChIP replicate we used 2.5 g of packed worms. Crosslinking was carried out by incubating samples in 1.5 mM EGS in PBS for 10 min, then adding 1.1% formaldehyde and incubating for a further 10 min. The reaction was quenched using 0.125 M glycine. The pellet was washed once in PBS+PMSF 1 mM and once in FA buffer (50 mM HEPES/KOH pH 7.5, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, and 150 mM NaCl)+PI. The pellet was resuspended in 4 ml of FA buffer+PI+0.1% sarkosyl and sonicated using a Diagenode Bioruptor Plus with 40 cycles, 30 s on, 30 s off. The sample was then spun in a tabletop microcentrifuge at top speed (15,000 rpm) for 15 min. The supernatant was incubated with 1 ml of anti-GFP antibody from Abcam (Abcam Cat# ab290, RRID:AB_303395) overnight at 4˚C. 60 ml of Protein A conjugated Dynabeads was added and the resulting solution incubated for 3 hr at 4˚C. Pulldown, washing, and de-crosslinking steps were as described in http://www.wormbook.org/chapters/www_chromatinanalysis/chromatinanalysis.html. For preparing ChIP libraries, we used NEBNext Ultra II DNA Library Prep Kit for Illumina with half of the pulldown and 30 ng of input. DNA libraries were then sequenced on an Illumina HiSeq 4000 platform with single read of 50 bases.
ChIP-seq data were analyzed using a nucleome processing and analysis toolkit that contains an automated ChIP-seq processing pipeline using Bowtie2 mapping and MACS2 peak calling. The software is available on Github at https://github.com/tjs23/nuc_tools (Stevens, 2021). Comparisons between different ChIP-seq conditions were carried out using the DiffBind package (Stark and Brown, 2011). ChIP-seq processed data were visualized using IGV (Robinson et al., 2011;Thorvaldsdó ttir et al., 2013).

Quantitative PCR
qPCR was performed using the Janus Liquid Handler (PerkinElmer) and a LightCycler 480 system (Roche). Total RNA was extracted from the heads of 20 male adults or 17 female adults using a Monarch Total RNA Miniprep Kit (NEB). Three replicates for male and three replicates for female flies were done for each genotype. RNA was reverse transcribed into cDNA using an ImProm-II Reverse Transcription System (Promega). cDNA was mixed with Luna Universal qPCR Master Mix (NEB). RpL32 (rp49) was amplified as an internal control. Primer sequences for Rpl32 and CAMTA were identical to those used in Sato et al., 2019. CaM was amplified using the primer pair 5 0 -TGCAG-GACATGATCAACGAG-3 0 (forward) and 5 0 -ATCGGTGTCCTTCATTTTGC-3 0 (reverse). Data processing was performed using LightCycler Software (Roche).

Supplementary files
. Supplementary file 1. The 100 most highly expressed genes (in order of decreasing read counts, in TPM) from neuron-specific RNA profiling.
. Supplementary file 2. Genes showing differential expression between camt-1 and WT in the profiled neural types.
. Supplementary file 3. Genomic locations differentially bound by CAMT-1 identified using the Diff-Bind algorithm for ChIP-seq data with a false discovery rate (FDR) threshold of 0.05. Genes overlapping or within 10 kb downstream of these sites are reported. The table is sorted in the order of increasing FDR. Note that the CAMT-1 binding site at the cmd-1 promoter was annotated with the overlapping long intervening noncoding RNA linc-128.
. Supplementary file 4. List of C. elegans strains used in this study.
. Supplementary file 5. Exact p-values and n numbers for experiments reported in this study.
. Transparent reporting form Data availability Sequencing data have been deposited in GEO under accession codes GSE164671.
The following dataset was generated: