Systemic Regulation of RAS/MAPK Signaling by the Serotonin Metabolite 5-HIAA

Human cancer is caused by the interplay of mutations in oncogenes and tumor suppressor genes and inherited variations in cancer susceptibility genes. While many of the tumor initiating mutations are well characterized, the effect of genetic background variation on disease onset and progression is less understood. We have used C. elegans genetics to identify genetic modifiers of the oncogenic RAS/MAPK signaling pathway. Quantitative trait locus analysis of two highly diverged C. elegans isolates combined with allele swapping experiments identified the polymorphic monoamine oxidase A (MAOA) gene amx-2 as a negative regulator of RAS/MAPK signaling. We further show that the serotonin metabolite 5-hydroxyindoleacetic acid (5-HIAA), which is a product of MAOA catalysis, systemically inhibits RAS/MAPK signaling in different organs of C. elegans. Thus, MAOA activity sets a global threshold for MAPK activation by controlling 5-HIAA levels. To our knowledge, 5-HIAA is the first endogenous small molecule that acts as a systemic inhibitor of RAS/MAPK signaling.


Introduction
Human cancer is a complex polygenic disease caused by somatic mutations in oncogenes and tumor suppressor genes together with inherited polymorphisms in cancer susceptibility genes. Many of the oncogenes and tumor suppressor genes that are mutated in different cancer types have been investigated in detail. However, relatively little is known about the effect of the genetic background on disease onset and progression. It thus remains a challenge to identify functional links between oncogenic traits and associated natural variants [1,2].
The components of the RAS/MAPK signaling pathway are mutated in a large fraction of human tumors. In particular, activating ("gain-of-function") mutations in HRAS and KRAS are among the most prevalent tumor initiating mutations found in human cancer cells [3]. Thanks to the strong conservation of this pathway in metazoans, genetic studies in model organisms, such as the nematode Caenorhabditis elegans, have provided important insights into various factors modulating RAS/MAPK signaling [4]. Moreover, C. elegans has become a platform species for quantitative genetic analyses of various phenotypes and pathways in order to identify and characterize polymorphic genes [5,6] In this study, we have used quantitative genetics to explore how the genetic background affects the phenotypes caused by the activating G13E (n1046) mutation in the C. elegans ras gene let-60 [7]. The n1046 mutation is homologous to the HRAS and KRAS mutations that are frequently found in human cancer cells [3]. For the purpose of this study, we compared RAS/ MAPK signaling in two highly diverse genetic backgrounds, C. elegans varieties Bristol (N2) and Hawaii (CB4856) [8]. Compared to the reference strain N2, the Hawaiian CB4856 strain on average contains one polymorphism every 412 bp with around 75% of all genes carrying at least one coding polymorphism [9].
To measure the activity of the RAS/MAPK pathway in different genetic backgrounds, vulval induction can be used as a quantifiable and reproducible readout. During vulval development, the anchor cell in the somatic gonad secretes the EGF-like ligand that activates via an EGFR family receptor tyrosine kinase the RAS/MAPK signaling pathway in the adjacent vulval precursor cells (VPCs) [10]. In combination with a lateral NOTCH signal, RAS/MAPK signaling induces three of the six VPCs to adopt a 2°-1°-2°pattern of vulval cell fates ( Fig 1A). Mutations that hyperactivate RAS/MAPK signaling, such as the n1046 allele, cause the differentiation of more than three and up to six VPCs and a Multivulva phenotype, while mutations that reduce RAS/MAPK signaling result in the induction of fewer than three VPCs and a Vulvaless phenotype. Hence, the average number of induced VPCs per animal, the vulval induction (VI), is a quantitative measure of RAS/MAPK signaling output in the VPCs [10,11]. Besides the vulva, RAS/MAPK signaling is activated in a variety of other tissues in C. elegans at different developmental stages, such as the meiotic germ cells in the hermaphrodite gonads, the excretory duct cell precursor in the embryo or the chemosensory neurons during olfaction in adults [4]. Using a quantitative genetics approach, we aimed at identifying globally acting as well tissue-specific modifiers of RAS/MAPK signaling. Here, we describe the identification of the polymorphic monoamine oxidase amx-2 gene as a global negative regulator of the RAS/MAPK pathway. amx-2 encodes a mitochondrial monoamine oxidase type A (MAOA) that catalyzes the oxidative deamination of biogenic amines such as dopamine (DA) and serotonin (5-HT) [12]. We further show that AMX-2 activity in intestinal cells controls the levels of the serotonin metabolite 5-hydroxyindoleacetic acid (5-HIAA), which acts as systemic inhibitor of MAPK phosphorylation.

Identification of RAS/MAPK modifiers by quantitative C. elegans genetics
To identify polymorphic modifiers of the RAS/MAPK pathway, we generated a set of 228 "mutation included recombinant inbred lines" (miRILs) between the Bristol strain MT2124 that carries the activating let-60 ras(n1046gf) mutation [7] and the Hawaiian CB4856 strain ( Fig  1B). Since small genetic variations are efficiently buffered in a wild-type genome [13,14], the .p receives most of the inductive EGF signal from the anchor cell and activates the EGFR/RAS/MAPK pathway inducing the 1°cell fate (green arrows). Lateral signaling via the Notch pathway induces the 2°cell fate in the neighboring VPCs P5.p and P7.p (red arrows). The remaining VPCs (blue) adopt the non-vulval 3°cell fate. (B) Crossing scheme to generate the let-60 (n1046gf) miRILs. Hawaii males (red) were crossed with Bristol let-60(n1046gf) mutants (blue). For each example animal, the two chromosomes IV carrying the n1046 mutation and another arbitrary chromosome pair are shown. Random segregation of the two parental genomes was allowed except for the let-60 (gf) mutation that was kept homozygous from F2 generation onwards. After ten generations of self-fertilization to drive all regions to homozygosity, 228 independent miRILs were obtained. (C) Genotypes and phenotypes of the let-60(gf) miRILs sorted by increasing VI. Genotypes determined by FLP mapping [15] are plotted on the y-axis versus the miRIL numbers on the x-axis. Hawaii genotypes are indicated with red, Bristol genotypes with blue and missing genotypes with gray colors. The VIs for each miRIL are shown below the genotypes. Error bars indicate the standard error of the mean. (D) QTL mapping identified three regions (QTL1 through QTL3) above the threshold LOD score of 3 (dotted red line). In each of the panels showing chromosomes I through X, the locations of the FLP markers used for genotyping are indicated on the x-axis with vertical lines. For the exact locations of the FLPs used, see Materials and Methods and [15]. inclusion of the let-60(gf) allele created a sensitized genetic background, allowing us to identify genetic modifiers that increase or decrease RAS/MAPK signaling. After 10 generations of inbreeding and genotyping using fragment length polymorphisms (FLPs) [15], 173 of the miRILs homozygous for the n1046 allele were used for further analysis (Fig 1C, top) (see Materials and Methods for details on genotyping and the selection of informative miRILs). In addition, we quantified RAS/MAPK signaling output in each of these miRILs by measuring the VI of at least 20 animals. While the let-60(gf) allele in the Bristol background exhibits a VI of 3.7±0.06 (n = 100), the VIs of the miRILs varied between 3.0 and 5.7 (Fig 1C, bottom). Quantitative trait loci (QTL) mapping [14] identified at least three loci on chromosomes I (QTL1), II (QTL 2) and V (QTL 3) above the threshold LOD score of 3 that are associated with variation in the VI (Fig 1D). For QTL1, the Bristol genotype is associated with a decreased VI, while for QTL2 and QTL3 the Bristol genotype is associated with an increased VI (S1 Fig). To estimate the effect size of each QTL and explore how the QTLs affect the VI when combined, we used two mapping models, one where the QTLs have additive effects and another one where they show an interaction (S1 Table). This analysis did not detect any significant interactions between the QTLs. Since the let-60(gf) mutation maps to chromosome IV, our approach did not permit us to identify QTLs on this chromosome. Moreover, the genetic incompatibility between the Bristol and Hawaii genomes caused by the zeel-1 and peel-1 loci on the left arm of chromosome I may have prevented the detection of QTLs in this region [16]. To confirm and refine the mapping of the detected QTLs, introgression lines (ILs) carrying defined segments of the Hawaii genome in the QTL regions of interest were crossed to the let-60(gf) Bristol strain [17]. Lines homozygous for the introgressions and the let-60(gf) mutation were compared to sibling lines without introgressions to identify those introgressions that cause significant differences in the VI (see Materials and Methods). The results for the fine mapping of QTL1 are shown in Fig 2A and for all QTLs in S2 Fig IL mapping revealed that QTL1 is composed of two adjacent QTLs, termed 1a and 1b, and that QTL1b maps to an interval of 1.43 Mbp containing 142 polymorphic genes (Fig 2A). Through this approach, we have identified several regions in the C. elegans genome that contain modifiers of the RAS/MAPK pathway. Notably, for QTL1a and QTL1b the Bristol genotype caused reduced RAS/MAPK activity, while for QTL2 and QTL3 the Bristol background increased RAS/MAPK activity.
The polymorphic amx-2 gene negatively regulates RAS/MAPK signaling Since the QTL1b region does not contain any known regulators of RAS/MAPK signaling, we performed RNAi knockdown of 107 of the 142 genes in this region in let-60(gf) single mutants as well as in let-60(gf) mutants carrying the ewIR17 introgression, which spans QTL1b. We envisioned two possible scenarios that are not mutually exclusive: (1) The QTL1b region in the Bristol strain may contain a negative regulator of RAS/MAPK signaling that is inactive or weakly active in the Hawaii background. (2) The Hawaii background may contain a positive regulator of RAS/MAPK signaling that is inactive or weakly active in the Bristol background. We thus screened for candidates exhibiting allele-specific RNAi effects (S2 Table). Note that when grown on the E.coli strain HT115 that is commonly used in RNAi feeding experiments [18], the let-60(n1046) allele exhibits an increased VI compared to animals grown on standard OP50 bacteria [19]. Knockdown of five genes significantly increased the VI in the let-60(gf) but not in the ewIR17; let-60(gf) background, defining potential negative regulators of RAS/MAPK signaling that are active in the Bristol background (highlighted in green in S2 Table), whereas knockdown of ten genes reduced the VI in the ewIR17; let-60(gf) but not in the let-60(gf) background, defining potential positive regulators active in the Hawaii background (highlighted in blue in S2 Table). These data suggested that the QTL1b region is oligogenic, containing several polymorphic modifiers of RAS/MAPK signaling. Of particular interest was the amx-2 gene because it fulfilled the criteria of a polymorphic negative regulator of RAS/MAPK signaling acting in the Bristol strain, but being inactive in the Hawaii strain. amx-2 RNAi had no significant effect on the ewIR17; let-60(gf) background, but amx-2 RNAi caused a robust increase in the VI of let-60(gf) mutants ( Fig 2B). Furthermore, the amx-2(ok1235) deletion mutant, which most likely represents a null allele (www.wormbase.org), increased the VI of let-60(gf) mutants in the Bristol background ( Fig 2C). To individually assess the activities of the Bristol and Hawaii amx-2 variants, we generated single-copy insertions on chromosome II [20] of a 7.8 kb genomic fragment spanning the amx-2 locus that was isolated either from the Bristol or the Hawaii genome. These single-copy transgenes were then introduced (homozygously) into the amx-2 (lf); let-60(gf) background. Insertion of the Bristol but not the Hawaii amx-2 variant reduced the VI of amx-2(lf); let-60(gf) double mutants to the value observed in let-60(gf) single mutants ( Fig 2C). These results confirmed the different physiological activities of the two amx-2 variants. In addition, amx-2(lf) partially suppressed the Vulvaless phenotype caused by reductionof-function mutations in let-60 ras [7] or the EGFR homolog let-23 [10] and enhanced the Multivulva phenotype of the let-23 gain-of-function mutation sa62 [21] (Fig 2D). We thus conclude that the Bristol variant of the amx-2 gene inhibits RAS/MAPK signaling in the VPCs.

amx-2 in intestinal cells inhibits RAS/MAPK signaling cell nonautonomously
To determine the site of amx-2 action, we generated transcriptional P amx-2 ::gfp reporters. amx-2 was expressed in head neurons, the intestine and in a subset of cells of the rectum and in the adult vulva (Fig 2E-2H). However, we did not observe any amx-2 expression in the VPCs during vulval induction, though amx-2 reporter levels could be below the detection limit. Since neurons have a low sensitivity to RNAi [22], yet amx-2i efficiently phenocopied the amx-2(lf) phenotype, we suspected that amx-2 might act in intestinal cells, where we detected strongest expression. Intestine-specific amx-2 RNAi using an rde-1(lf); let-60(gf); P elt-2 ::rde-1(+) strain [23] increased the VI to a similar degree as systemic RNAi, while vulva-specific RNAi using the P lin-31 ::rde-1(+) transgene [24] had no detectable effect, which is consistent with lack of detectable amx-2 reporter expression in the VPCs (Fig 2I, note that the overall lower VI in the vulvaspecific RNAi strain is due to the genetic background [24]). Taken together, AMX-2 most likely acts in the intestinal cells to negatively regulate RAS/MAPK signaling in the VPCs.

Expression of amx-1 is increased in amx-2(lf) mutants
To investigate a possible redundancy between the MAOA amx-2 and the MAOB gene amx-1, we measured the transcript levels of amx-2 and its paralog amx-1 by quantitative real-time PCR. The abundance of amx-2 and amx-1 transcripts was not significantly different between the Bristol and Hawaii backgrounds ( Fig 2J). However, amx-2 transcript levels were around 60% decreased and amx-1 levels around 40% increased in amx-2(lf) mutants. Possibly, the elevated amx-1 expression can partially compensate for a loss of amx-2 expression.  (Fig 3A) [12]. The products of the DA and 5-HT deamination reactions, 3,4-dihydroxyphenylacetaldehyde and 5-hydroxyindole-acetaldehyde respectively, are further oxidized by aldehyde dehydrogenases into 3,4-dihydroxyphenylacetic acid and 5-hydroxyindoleacetic acid (5-HIAA), which in humans are secreted through the kidneys (Fig 3A) [26]. Consistent with the predicted function of AMX-2 in degrading 5-HT, total extracts of amx-2(lf) worms contained elevated levels of 5-HT when compared to wild-type extracts (Fig 3B).
We thus investigated if AMX-2 inhibits RAS/MAPK signaling by controlling the levels of DA, 5-HT or their metabolites. The addition of 10mM DA to the growth medium had no significant effect on the VI of let-60(gf) single or amx-2(lf); let-60(gf) double mutants (Fig 3C).
However, 1mM 5-HT as well as 1mM of the 5-HT metabolite 5-HIAA caused a strong reduction of the VI in both backgrounds and a suppression of the Multivulva phenotype (Fig 3C  and 3D). Addition of 0.6mM melatonin (MT), another 5-HT metabolite (Fig 3A), had a slightly weaker yet significant effect on the VI (Fig 3C). We conclude that the 5-HT metabolites, in particular 5-HIAA, inhibit RAS/MAPK signaling. To test the sensitivity of the RAS/MAPK pathway to 5-HT and 5-HIAA, we performed dose-response experiments in the presence and absence of amx-2. For both compounds, the maximum reduction of the VI was observed at concentrations above 1mM (Fig 3E and 3F). However, let-60(gf) single mutants displayed a higher sensitivity to low concentrations (1μM) of 5-HT than amx-2(lf); let-60(gf) double mutants, while the effects of 5-HIAA were independent of the amx-2 genotype. Overall, 5-HT exerted a slightly stronger effect than 5-HIAA, suggesting that additional 5-HT metabolites besides 5-HIAA may inhibit RAS/MAPK signaling.

5-HIAA acts via the 5-HT receptor SER-1 and the EGL-30 Gqα pathway
We further characterized the inhibitory effect of 5-HIAA on the RAS/MAPK pathway by testing mutants in the 5-HT pathway for their response to 5-HIAA treatment. A mutation in the tryptophan hydroxylase gene tph-1, which is essential for 5-HT biosynthesis [29], slightly reduced the VI in let-60(n1046gf) animals in the absence of 5-HIAA ( Fig 3H). However, treatment of tph-1(lf); let-60(n1046gf) double mutants with 4mM 5-HIAA further reduced the VI, indicating that 5-HIAA acts in the absence of endogenous 5-HT and hence does not compete with 5-HT. By contrast, the VI of let-60(n1046gf) animals carrying a mutation in the 5-HT receptor gene ser-1 [30] was not reduced by 5-HIAA treatment. Surprisingly, the VI of let-60(gf); ser-1(lf) double mutants was even increased after 5-HIAA treatment. Moreover, a gain-of-function mutation in egl-30, which encodes a Gqα protein acting in the 5-HT pathway [31], rendered let-60(n1046gf) mutants resistant to 5-HIAA and caused a slight increase of the VI in untreated animals (Fig 3H). Since the SER-1/EGL-30 pathway plays an essential role in 5-HT stimulated egg laying [30], we tested the effects of 5-HIAA on the egg laying rate with and without 5-HT stimulation. While 5-HIAA treatment alone caused a slight reduction in the egg laying rate, 5-HIAA did not significantly compete with the 5-HT stimulated increase in egg laying (S5 Fig). We conclude that 5-HIAA acts via the SER-1 receptor and the downstream EGL-30 Gqα signaling pathway to repress RAS/MAPK activity. However, the inhibitory effect of 5-HIAA is independent of 5-HT activity.

5-HIAA attenuates RAS/MAPK signaling in multiple organs of C. elegans
Besides the VPCs, RAS/MAPK signaling is required in several other organs of C. elegans [4]. Hence, let-60(gf) mutants exhibit multiple defects besides a Muv phenotype. For example, the temperature-sensitive let-60(ga89gf) allele causes accelerated exit of meiotic germ cells from the pachytene stage, resulting in the accumulation of many immature oocytes in the proximal gonad arm at the restrictive temperature [32,33] (Fig 4A). Moreover, let-60(n1046gf) mutants frequently contain two duct cells expressing the lin-48::gfp marker [34] (Fig 4B). Treatment with 4mM 5-HIAA partially suppressed the let-60(gf) phenotypes both in the germ line and the duct cell (Fig 4A and 4B). To measure the global effect of 5-HT and 5-HIAA treatment on MAPK activation, we quantified the levels of activated, phosphorylated MPK-1 in total extracts of L4 larvae [11]. Treatment with 5-HT and 5-HIAA caused a similar reduction in phospho-MPK-1 levels in let-60(gf) mutants. However, in the amx-2(lf); let-60(gf) background 5-HIAA exerted a stronger effect than 5-HT ( Fig 4C). Thus, 5-HIAA supplemented into the culture medium exerts a systemic effect to inhibit RAS/MAPK signaling in different organs of C. elegans.

Discussion
We have identified several genetic modifiers of the oncogenic RAS/MAPK signaling pathway by comparing miRILs derived from the backgrounds of two highly diverged C. elegans isolates. The two parental strains used in this study display a level of sequence divergence that is comparable to the genetic variation observed in the human population [35]. The genetic modifiers of RAS/MAPK signaling we identified through this quantitative approach could not have been found in conventional forward genetic screens, as each locus alone only exerts a minor effect. Interestingly, both genetic backgrounds analyzed contain QTLs that enhance (i.e. QTLs 2 and 3 for Bristol) as well as QTLs that reduce (i.e. QTL 1 for Bristol) the relative strength of RAS/ MAPK signaling. Thus, each isogenic background may represent a balanced state exhibiting intermediate RAS/MAPK pathway activity thanks to the opposing effects of the different modifiers. The interplay of these modifiers may be necessary to keep the activity of the RAS/MAPK pathway within a certain range and avoid the detrimental effects caused by increased or reduced RAS/MAPK signaling.
The molecular characterization of one particular region (QTL 1b) identified the monoamine oxidase gene amx-2 as a negative regulator of RAS/MAPK signaling in multiple organs of C. elegans. Though, the RNAi analysis of the QTL1b region indicated that this region contains possibly up to ten additional polymorphic modifiers of RAS/MAPK signaling besides amx-2. Single-copy gene insertion experiments [20] demonstrated that the Bristol variant can fully rescue an amx-2 deletion allele, while insertion of the Hawaii locus had no significant effect in this assay, indicating that amx-2 activity in the Hawaii background is severely reduced or even absent.
The identification of a monoamine oxidase as a negative regulator of RAS/MAPK signaling was initially a surprising result, since MAOA is primarily known for its role in degrading neurotransmitters in the nervous system [12]. However, we observed strong AMX-2 expression in non-neuronal tissues, especially in the intestinal cells. The 5-HT metabolites such as 5-HIAA that result from AMX-2 catalysis are likely to be released into the body cavity in order to modulate RAS/MAPK signaling in distant organs. Such a globally acting regulatory mechanism may be useful to rapidly adjust RAS/MAPK signaling in response to changing environmental conditions, after food intake and to adapt the speed of reproduction [36]. Epistasis analysis by applying exogenous 5-HIAA points at a step downstream of RAS and upstream of MAPK that is repressed by 5-HIAA. Hence, 5-HIAA may simultaneously repress the RAS/MAPK pathway activated by various receptor tyrosine kinases in different tissues [4]. The observation that 5-HT exerts an inhibitory effect even in amx-2(0) mutants may be explained by the presence of additional redundant MAOs, notably AMX-1, and by spontaneous oxidation of 5-HT. Our epistasis analysis further indicates that 5-HIAA acts via the SER-1 receptor, which activates the EGL-30 Gqα signaling pathway [31]. One possible scenario is that 5-HIAA and 5-HT exert opposing effects on SER-1, such that the balance between 5-HT and 5-HIAA levels determines the strength of EGL-30 activation, which in turn promotes RAS/MAPK signaling. In line with this model, Moghal et al. [37] have previously shown that egl-30 signaling in neuronal cells positively regulates vulval induction under different environmental conditions. The role of 5-HT as a neurotransmitter in the mammalian nervous system is well documented [12]. However, over 90% of the 5-HT in the human body is found outside of the nervous system, especially in enterochromaffin cells of the intestine [38]. Remarkably, Rybaczyk et al. [39] reported that the expression of the human 5-HT degrading enzyme MAOA, the closest AMX-2 homolog, is consistently down-regulated across many human tumor types. The functional implications and mechanisms of reduced MAOA expression in cancer cells are unclear. Our findings that systemic application of the 5-HT metabolite 5-HIAA globally inhibits RAS/MAPK signaling may explain the physiological consequences of MAOA down-regulation. Tumors expressing low levels of MAOA may generate less oncostatic 5-HIAA and at the same time contain higher levels of 5-HT, which can promote tumor growth and survival via crosstalk to the RAS/MAPK pathway [40,41]. Thus, MAOA levels may set a global threshold for the activation of the RAS/MAPK cascade by different extracellular signals. To our knowledge, 5-HIAA is the first endogenous small molecule that acts as a systemic inhibitor of the RAS/ MAPK pathway.

General methods and strains used
Strains were maintained on NGM agar seeded with OP50 bacteria at 20°C [42], unless otherwise stated. C. elegans Bristol refers to the wild-type N2 strain and Hawaii to CB4856 [8]. Transgenic lines were generated as described in [32,43].

Phenotyping
To measure the VI, vulval induction was scored in L4 larvae using Nomarski optics as described [48], and the average number of induced VPCs per animal was calculated. The duct cell duplication phenotype was scored using the lin-48::gfp marker to visualize the duct cells using fluorescence microscopy [34]. The oocyte maturation phenotype was scored in 2 day old adults under Nomarski optics microscopy.
QTL mapping and data storage QTL mapping was performed using a single marker model on the per miRIL averages. Significance threshold was estimated using 1000 permutations [14]. All QTL data, phenotypes, QTL profiles and genotypes are stored in www.WormQTL.org [49].
Generation of single-copy insertion lines A 7.8 kb genomic fragment spanning the entire amx-2 locus was amplified with the primers OTS123 (GATTTTGGAGAAGAAACGAGGG) and OTS124 (ACTTCACTATGTTCCTCT ACCG) using either Bristol or Hawaii genomic DNA as template and subcloned into the XhoI restriction site of pCFJ151 [20]. Single-copy insertions of the amx-2 Bristol and amx-2 Hawaii containing plasmids into the ttTi5605 region on chromosome II were generated using the protocol by [20] to yield zhSi73 and zhSi74, respectively. The insertions were verified by PCR amplification using primers flanking the insertion site before crossing them into the amx-2(lf); let-60(gf) background. For each genotype, at least three independent lines were scored.

Transcriptional amx-2 reporters
Primers OTS219 (AAA AGG ATC CTT AGG TTT ATT GCT GGA AAA AT) and OTS220 (AAA AGG ATC CCC TTA ACC AAA TTT CAT ACC C) were used to amplify 4kb of upstream promoter region. The PCR fragment was further cloned into the BamHI restriction site of pPD95.67 to generate a the P amx-2 ::gfp transcriptional reporter plasmid that was co-injected at 50ng/μl with 2.5ng/μl of the pharyngeal P myo-2 ::mcherry marker.

Measurement of 5-HT levels
Animals were grown in 100ml liquid cultures and harvested by flotation on 50% sucrose. Worm pellets were resuspended in 2ml PBS buffer and lysed using a swing-mill homogenizer followed by high-speed centrifugation to remove insoluble debris. Total protein concentrations were measured in each sample using the amidoblack staining assay [50]. 5-HT levels were determined with an ELISA kit according to the manufacturer's instructions (BA E-5900, Labor Diagnostika Nord) and normalized to the total protein concentrations in the extracts. The average 5-HT concentrations for each genotype were determined with two separate measurements, each done in triplicate using extracts obtained from two independently grown cultures.

Quantification of ERK phosphorylation
Phospho MPK-1 levels in total extracts of C. elegans L4 larvae were determined by Western blotting as described in [11]. As loading controls, total MPK-1 levels were quantified on parallel blots loaded with the same amounts of protein (20μg) from the identical samples. Protein bands were quantified using the integrated density function in ImageJ. The ratios of phospho-MPK-1 to total MPK-1 levels were calculated for each extract and normalized to the ratios in untreated controls. Antibodies used: anti-MAP Kinase (Sigma-Aldrich, M5670), anti-phospho-MAP Kinase, Activated (Diphosphorylated ERK-1&2, Sigma-Aldrich, M8159).  Fig 1D with ILs. ILs covering the predicted QTL regions were chosen from [17] and crossed with the let-60(n1046) Bristol mutant. Significant differences in the VI indices between sibling lines with and without introgression were used to verify and further refine the different QTL regions. Several overlapping introgression lines allowed us to further narrow down the genomic intervals for further studies. The sizes and approximate positions of the ewIR introgressions are depicted below. For the exact locations of the breakpoints in each IL, see [17]. Error bars indicate the standard error of the mean, and ÃÃÃ indicates p<0.001 in a Student's t-test. Egg laying rates of one day-old adults were determined in liquid as described in [30]. For each genotype and condition, in total 48 animals were assayed in four independent experiments and the average egg laying rates per animal are shown. Error bars indicate the standard error of the mean, ÃÃ indicates p<0.01, Ã p<0.05 and n.s. p>0.1 in a Student's t-test. (JPG) S1 Table. Estimated QTL effect sizes in additive and interactive mapping models. (PDF) S2 Table. Results of the RNAi screen in the QTL1b region. Rows highlighted in green indicate allele-specific effects in the Bristol background, rows highlighted in blue indicate allelespecific effects in the Hawaii background, and rows highlighted in beige allele-independent effects. 1 SE indicates the standard error of the mean. 2 p-values compared to empty vector controls were determined with a Student's t-test. (PDF)