Transcriptional Control Of Calmodulin By CAMTA Regulates Neural Excitability

Calmodulin (CaM) is the major calcium ion (Ca2+) sensor in many biological processes, regulating for example the CaM kinases, calcineurin, and many ion channels. CaM levels are limiting in cells compared to its myriad targets, but how CaM levels are controlled is poorly understood. We find that CaM abundance in the C. elegans and Drosophila nervous systems is controlled by the CaM-binding transcription activator, CAMTA. C. elegans CAMTA (CAMT-1), like its fly and mammalian orthologues, is expressed widely in the nervous system. camt-1 mutants display pleiotropic behavioural defects and altered Ca2+ signaling in neurons. Using FACS-RNA Seq we profile multiple neural types in camt-1 mutants and find all exhibit reduced CaM mRNA compared to controls. Supplementing CaM levels using a transgene rescues camt-1 mutant phenotypes. Chromatin immunoprecipitation sequencing (ChIP-Seq) data show that CAMT-1 binds several sites in the calmodulin promoter. CRISPR-mediated deletion of these sites shows they redundantly regulate calmodulin expression. We also find that CaM can feed back to inhibit its own expression by a mechanism that depends on CaM binding sites on CAMT-1. This work uncovers a mechanism that can both activate and inhibit CaM expression in the nervous system, and controls Ca2+ homeostasis, neuronal excitability and behavior.


Introduction
Calmodulin-binding transcription activators (CAMTAs) are a highly conserved family of Ca 2+ -regulated transcription factors 1 . In plants, CAMTAs are transcriptional effectors of Ca 2+ /CaM signaling in response to biotic/abiotic stress [2][3][4][5][6] . Mammals encode two CAMTA proteins that are enriched in heart and brain tissue 7 . Loss of CAMTA1 in the nervous system induces degeneration of cerebellar Purkinje cells, ataxia and defects in hippocampal-dependent memory formation 8,9 . A variety of neurological disorders, including intellectual disability, attention deficit hyperactivity disorder (ADHD), cerebellar ataxia, and reduced memory performance have been reported in individuals with lesions in the human CAMTA1 gene [10][11][12] . Mechanistically however, little is known about the origin of these neuro-behavioural phenotypes. We show here that the C. elegans ortholog of CAMTA, CAMT-1, regulates neuronal Ca 2+ signaling by controlling CaM expression. A variety of behaviours are dependent on CAMT-1, and Ca 2+ imaging in multiple neurons reveals that neural activity is abnormal in camt-1 mutants. By combining cell-type specific transcriptional profiling of these neurons with chromatin immunoprecipitation (ChIP) analysis of CAMT-1's DNA-binding sites, we find that CAMT-1 upregulates the expression of CaM to control neural activity and behaviours. CaM is a ubiquitously expressed Ca 2+ sensor that plays a key role in buffering intracellular Ca 2+ and in directing a cellular response to Ca 2+ changes 13,14 . It is pivotal to diverse processes, including metabolic homeostasis, protein folding, apoptosis, vesicular fusion, and control of neuronal excitability 15,16 , for example through regulation of CaM kinase II activity.
Importantly, cellular levels of CaM are limiting, compared to the concentration of CaM binding proteins 17 , and changes in CaM levels are likely to impact Ca 2+ /CaM regulation of downstream targets 18 . How CaM expression is regulated in specific cell-types and contexts is, however, poorly characterized. Our results identify CAMTA as a CaM regulator in neurons, and suggest CAMT-1 can both promote CaM expression and repress it, in a feedback loop by which CaM can negatively control its own expression by binding CAMT-1.

CAMT-1 functions in neurons to regulate multiple behaviours
We identified mutations in camt-1, the sole C. elegans CAMTA, in a forward genetic screen for mutations that disrupt C. elegans aggregation behavior 19 . This behavior, where animals form groups (Extended Data Fig. 1a), is closely linked to escape from ambient oxygen levels (21% O2) 20,21 . The screen used an N2 strain defective in the neuropeptide receptor NPR-1, npr-1(ad609), that aggregates similarly to most natural isolates of C. elegans; the N2 lab strain does not aggregate due to an npr-1 gain-of-function mutation 22 .
In mouse, humans and flies, CAMTA transcription factors are expressed in many brain regions [8][9][10]26 . We generated a fosmid-based reporter to map the expression pattern of the CAMT-1a, the longest isoform of C. elegans CAMTA. This reporter was functional, as it rescued the behavioural defects of camt-1 mutants (Fig. 1c), and revealed that CAMT-1 is expressed broadly and specifically in the nervous system (Fig. 1d). We observed CAMT-1 expression in sensory neurons with exposed ciliated endings, motor neurons of the ventral cord, the URX O2-sensing neuron, and the RMG hub interneurons (Extended Data Fig. 2).
This broad expression prompted us to ask if camt-1 mutants displayed pleiotropic behavioural phenotypes. We asked whether CAMT-1 is required for behavioural responses to chemical cues other than O2, and for other aversive behaviours, such as avoidance of CO2. In response to a rise in CO2, WT N2 worms transiently perform omega turns, Ω-shaped body bends that re-orient the animal away from the stimulus 27 . camt-1 mutants exhibited abnormally high levels of omega-turns without a CO2 stimulus and a prolonged increase in omega turns in response to a rise in CO2 (Fig. 1e). C. elegans avoids CO2 but is attracted towards salt and a range of volatile compounds 28,29 . Chemotaxis towards NaCl, benzaldehyde and diacetyl was reduced in camt-1 mutants, and these defects were rescued by a fosmid transgene containing WT CAMT-1 (Fig. 1f). Together these data show that CAMT-1 function is important for many C. elegans behaviours .
Many deleterious human alleles of CAMTA1 alter the CG-1 DNA binding domain 11 . We assessed the importance of the putative DNA-binding domain of CAMT-1 using mutants. Strains engineered using CRISPR to harbor mutations in conserved residues of the CG-1 domain showed defects in aggregation and in their response to O2, recapitulating phenotypes of the camt-1 deletion mutants described above (Extended Data Fig.1b, Fig. 1g). By contrast, substituting conserved isoleucine residues in all four putative IQ domains to asparagines did not disrupt the O2-avoidance behaviors of npr-1 mutant animals (Extended Data Fig. 1b and d). Moreover, expressing the a isoform of CAMT-1 with conserved residues in the CaMBD mutated rescued camt-1 mutant phenotypes (Extended Data Fig. 1c and e). These results suggest that CAMT-1 binding to DNA is essential for its function but binding to CaM is not, at least for oxygen escape behaviour.
We targeted CAMT-1 cDNA expression to different subsets of neurons to find out where it is required to promote aerotaxis. Restricting expression of CAMT-1 to RMG but not O2 sensing neurons rescued the fast movement at 21% O2 in camt-1 mutants (Extended Data Fig. 1f-g). Defective responses of camt-1 mutants to 7% O2 was neither rescued by the expression of CAMT-1 in RMG nor simultaneous expression in RMG and O2-sensing neurons (Extended Data Fig.   1f-g). This data was consistent with a model in which CAMT-1 acts in multiple neurons. As expected, pan-neuronal expression rescued mutant phenotypes; in particular, expression of the a isoform alone (CAMT-1a) was sufficient for rescue ( Fig. 1h). Interestingly, pan-neuronal overexpression of CAMT-1 reduced locomotory activity at 7% O2 to below levels found in the npr-1 mutant background, suggesting that animal speed in 7% O2 is anti-correlated with CAMT-1 levels (Fig. 1h).
CaM/Ca 2+ -dependent changes in gene expression are known to be important for both development and function of the nervous system 30,31 . To test whether CAMT-1 activity is required during development, we expressed CAMT-1 cDNA from a heatshock-inducible promoter. Without heat-shock, this transgene did not rescue the hyperactivity phenotype of camt-1 mutants (Fig. 1i). However, inducing CAMT-1 expression in late L4s/ young adults was sufficient to rescue the adult mutant defects (Fig. 1j), suggesting that CAMT-1 does not act developmentally to regulate behavioural responses to ambient O2.

CAMT-1 dampens Ca 2+ responses in sensory neurons
To test whether disrupting camt-1 altered physiological responses to sensory cues we recorded stimulus-evoked Ca 2+ changes in O2-and CO2-sensing neurons with Yellow Cameleon (YC) sensors. URX activity tracks environmental O2 levels, and tonic signaling from URX to RMG drives high locomotory activity at 21% O2 20 . BAG and AFD neurons are CO2 sensors, and BAG drives omega turns when CO2 levels rise 32 . We found that Ca 2+ responses in URX, BAG and AFD neurons were significantly elevated in camt-1 mutants across all the O2/CO2 conditions we tested ( Fig. 2a-b, Extended Data Fig. 3a). These data suggest that CAMT-1 activity somehow dampens the Ca 2+ responses of these sensory neurons. We observed the converse phenotype, dramatically reduced Ca 2+ levels, when we overexpressed CAMT-1 cDNA specifically in O2 sensors or in BAG neurons of control animals (Extended Data Fig. 3b-c). Although Yellow Cameleon is a ratiometric sensor, we cannot exclude the possibility that its reduced expression in animals overexpressing CAMT-1 (Extended Data Fig. 3d) contributes to the reduced baseline YFP/CFP ratio.
RMG hub interneurons likely integrate inputs from multiple sensory stimuli, including food, pheromones and O2, and drive the change in locomotory state associated with a switch from 7% to 21% O2 33,34 . O2-evoked responses in RMG neurons depend on URX: URX ablation abolishes these responses.
Whether the RMG phenotype reflects a homeostatic response to hyperactivation by URX, or altered input into RMG from other neurons is unclear. However, these data suggest that the reduced locomotory activity of camt-1 mutants at 21% O2 is due to defective RMG responsiveness.

CAMT-1 phenotypes are associated with reduced expression of CMD-1/CaM
To identify downstream targets of CAMT-1 we compared the transcriptional profiles of multiple neural types in camt-1 and control animals 35 . We separately profiled the O2-sensors URX/AQR/PQR, the RMG interneurons, the AFD thermosensors, and the BAG O2/CO2 sensors. We used FACS to collect the neurons from strains in which they were labeled with GFP, and performed 4 -10 biological replicates for robust statistical power. Our analysis of the data revealed altered expression of many genes; most of these changes were neural-type specific (Fig. 3a, Extended Data Table 1 and 2). A striking exception was the C.
elegans CaM, cmd-1. cmd-1 was one of only two genes whose expression was reduced in all four neural profiles relative to WT controls, with mRNA levels 2.5 -4 fold lower than controls, depending on neural type ( Fig. 3a-b). The other gene, Y41C4A.17, is a short transmembrane protein with no known homolog in mammals. CaM plays a key role in regulating many functions in the nervous system 36,37 . We therefore speculated that reduced CMD-1/CaM expression could account for many camt-1 phenotypes.
We also found that supplementing CMD-1 levels rescued the hyperexcitability defects in URX and BAG neurons of camt-1 mutants (Fig. 3e, Extended Data Fig.   4c), and partially rescued Ca 2+ responses in RMG (Fig. 3f). Together these data suggest that reduced CMD-1 expression accounts for camt-1 Ca 2+ signaling and behavioural defects.

CAMTA promotes CaM expression in D. melanogaster
Fly mutants of CAMTA show slow termination of photoresponses compared to wild type controls 38 , and also exhibit defects in male courtship song 26 . An allele of the Drosophila calmodulin gene that deletes part of the promoter and reduces CaM expression also shows slow termination of photoresponses 39 . 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 26,38 . There is a modest decrease in dCAMTA mRNA level in tes 2 mutants, suggesting that the premature stop late in the protein may not induce mRNA degradation (Extended Data Fig.   4d). The level of dCAMTA mRNA decreases strongly in cro mutants as previously reported 26 (Extended Data Fig. 4d). We assessed the level of CaM in the heads of dCamta mutant flies using quantitative RT-PCR and Western blots.
Each method reported significant decreases in CaM expression in tes 2 and cro mutants compared to controls ( Fig. 3g-i). These results suggest that the transcriptional upregulation of CaM by CAMTA is conserved from worms to flies.

CAMT-1 directly regulates CMD-1/CaM transcription through multiple binding sites at cmd-1/CaM promoter
To test whether CAMT-1 directly binds the cmd-1 promoter, we performed chromatin immunoprecipitation sequencing (ChIP-seq) using a CRISPR-knockin CAMT-1a::GFP strain. Our analysis revealed > 200 loci that were significantly enriched in CAMT-1a::GFP pulldowns compared to input, and to a mock pulldown (Extended Data Table 3). Importantly, we observed three peaks ~ 6.3kb, 4.8 kb and 2.2 kb upstream of the CMD-1 translation start site in CAMT-1a::GFP pulldown (Fig. 4a, we called these peaks A, B and C respectively, Extended Data Fig. 5a). Thus, CAMT-1 appears to be recruited to multiple sites upstream of cmd-1. A CAMT-1 binding peak was also found in the promotor region of Y41C4A.17 (Extended Data Fig. 5b).
To test whether the CAMT-1 ChIP-seq peaks in the cmd-1 promoter region regulate CMD-1 transcription, we generated CRISPR strains that delete one or more of these peaks. A strain harbouring 110 bp and 136bp deletion at peaks B and C respectively ( Fig. 4a-b, db1275) and a strain harbouring a 200 bp deletion at peak A ( Fig. 4a-b, db1280) exhibited aggregation and O2 escape responses similar to npr-1 mutant controls (Fig. 4b). However, a strain harbouring all three deletions ( Fig. 4b-c, db1278) exhibited strong aggregation defects (Extended Data Fig. 1a) and locomotory responses to O2 that mirrored the defects of camt-1(ok515) loss-of-function mutants (Fig. 4b, Fig. 1a). Notably, the hyperactivity at 7% O2 of db1278 mutants could be rescued by expressing additional CMD-1 in the nervous system. Like camt-1(ok515) mutants, camt-1(db1278) mutants also showed chemotaxis defects towards salt, benzaldehyde and diacetyl that could be rescued by neuronal expression of CMD-1 (compare Fig. 3d and 4c). These results suggest that multiple sites in the CMD-1 promoter act redundantly to recruit CAMT-1.
Together our data shows that CAMT-1 regulates CaM expression in a redundant manner through binding to multiple sites in the CaM promoter.

CMD-1/CaM can inhibit its own expression via CAMT-1
CMD-1 levels are important for proper neural function. We speculated that CMD-1 might homeostatically regulate its own expression by means of a negative feedback loop. To investigate this hypothesis, we built a transcriptional reporter of CMD-1 by fusing an 8.9 kb fragment immediately upstream of the CMD-1 translational start site with GFP. This reporter shows strong fluorescence expression in neurons and muscle, including pharyngeal muscle (Fig. 4d). We crossed this line with another C. elegans line that over-expresses CMD-1 in neurons using the rab-3 promoter (rab-3p::cmd-1) and measured neuronal GFP fluorescence in single and double transgenic animals. We normalized expression using pharyngeal GFP levels. Animals bearing both transgenes exhibited a decrease in neuronal GFP fluorescence, suggesting that high levels of CMD-1 can repress expression from the cmd-1 promoter (Fig. 4e). To examine if this repression is achieved via CaM binding to CAMT-1, we introduced into the double transgenic background the camt-1 allele that disrupts the 4 IQ domains, syb1919. We found that camt-1(syb1919) animals expressing cmd-1p::GFP and rab-3p::cmd-1 showed neuronal GFP levels similar to that found in control animals lacking the rab-3p::cmd-1 transgene (Fig. 4e). These data suggest that CMD-1 can negatively regulate its own expression by binding the IQ domains of CAMT-1. Thus CAMT-1 can not only activate cmd-1 expression but also repress it when available CMD-1 levels are high.

Discussion
As CaM is a central regulator of Ca 2+ signaling, its dysregulation is likely to have profound effects on cellular and organismal processes, and drive of disease 16,36,37 . We discovered that CAMTA is a major regulator of CaM in the nervous system of C. elegans and Drosophila. The increase in neural excitability and pleiotropic behavioural defects of camt-1 mutants are consistent with the many known roles of CaM in the nervous system, including controlling activation properties of ion channels 40 and regulating CaM-dependent kinases 36,37,41 .
Likewise, CaM has been shown to regulate the termination of visual response in   b camt-1 mutants exhibit altered locomotory responses to 21% O2 and hyperactive movement at 7% O2.
c A WT copy of the camt-1 genomic locus rescues the O2-response defects of camt-1 mutants.
d CAMT-1a::GFP driven from its endogenous regulatory sequences in a recombineered fosmid is expressed widely in the nervous system. VNC, ventral nerve cord.
e camt-1 mutants exhibit an increased turning frequency both in the presence and absence of a CO2 stimulus. Assays were performed in 7% O2.
c The db1278 allele confers chemotaxis defects to NaCl, benzaldehyde and diacetyl that can be rescued by supplementing CMD-1 expression in the nervous system.  f-g Defective responses of camt-1(ok515) mutants to 21% O2 are rescued by expression of CAMT-1a in RMG (using the flp-5 promoter, f) but not in O2-sensing neurons (using the gcy-32 promoter, g). The defective response camt-1(ok515) mutants to 7% O2 are neither rescued by expression in RMG nor by expression in both RMG and O2-sensing neurons (gcy-32p+flp-5p::camt-1). # and $ marked the interval used for time points used for statistical test at 7% and 21% O2 respectively.   Numbers on the right indicate the scale (normalized read counts).

Extended Data Tables
Extended Data Table 1: The 100 most highly expressed genes (in order of decreasing read counts, in TPM) from neuron-specific RNA profiling. Table 2: Genes differentially-expressed in camt-1 and WT in the profiled neural types. Table 3: Genomic locations differentially bound by CAMT-1 identified using the DiffBind algorithm for ChIP-seq data with a False Discovery Rate (FDR) threshold of 0.05. Genes overlapping or within 10kb 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 non-coding RNA linc-128. Table 4: List of C. elegans strains used in this study.

Methods
No statistical methods were used to predetermine sample size. The experiments were not randomized.

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 32

Behaviour assays
O2-and CO2-response assays were performed as described previously 46  Chemotaxis assays were performed as previously described (Bargmann et al., 1993) with minor modifications. 9 cm assay plates were made with 2% Bacto Agar, 1mM CaCl2, 1mM MgSO4 and 25mM K2HPO4 pH 6. Test and control circles of 3cm 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, 1μl odorant (Benzaldehyde 1/400 or Diacetyl 1/1000 dilution in ethanol) or 1μl ethanol, and 1μl 1M NaN3, was added to each circle.
For gustatory assays, an agar plug containing 100 mM NaCl was added the night before and removed prior to assay. Assays were allowed to proceed for 30-60min, after which point plates were moved to 4°C. Chemotaxis index was calculated as (number of animals in test circle -number of animals in control circle) / total number of animals that have left starting area.

Heat-shock
Animals were raised at 15 °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 room temperature for 6 h.

Ca 2+ imaging
Neural imaging was performed as previously described 46 , with a ×2 AZ-Plan 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 cells sorted as described previously 35 . Briefly, C. elegans with GFP-labelled neurons were synchronized using the standard bleaching protocol and eggs placed 3 days before cell sorting on 90 mm NGM plates seeded with OP50. For each sample, we used >50 000 worms. The worms were washed 3 times with M9, prewashed and then incubated for 6.5 min with 750 μl lysis buffer

Confocal microscopy and image analysis
Young adult worms were mounted for microscopy on a 2% agar pad in 1M sodium azide. Image analysis and fluorescence quantification was carried out using Fiji (ImageJ, Wayne Rasband, NIH).
water-immersion objective. Colocalization of CAMT-1(fosmid)-GFP with neuronal markers was imaged on a Andor Ixon EMCCD camera coupled to a spinning disk confocal unit, with a 60x lens and 100 ms exposure time..

ChIP-seq
The ChIP-seq protocol used was similar to that describe in Wormbook
ChIP-seq data were analyzed using a nucleome processing and analysis toolkit which 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. Comparisons between different ChIP-seq conditions were carried out using the DiffBind package 47 . ChIP-seq processed data was visualized using IGV 48,49 .

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). 3 replicates for male and 3 replicates for female flies were done for each genotype. RNA was reverse transcribed into cDNA using a 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 with the one used in Sato et al. 26 . CaM was amplified using the primer pair 5'-TGCAGGACATGATCAACGAG-3' (forward) and 5'-ATCGGTGTCCTTCATTTTGC-3' (reverse). Data processing was performed using LightCycler Software (Roche).

Statistical tests
Statistical tests are two-tailed and were performed using Matlab ( Figure 4