- Top of page
- Materials and methods
- Supporting Information
The biogenic amine serotonin (5-HT, 5-hydroxytryptamine) exerts powerful, modulatory control over multiple physiological functions in the brain and periphery, ranging from mood and appetite to vasoconstriction and gastrointestinal motility. In order to gain insight into shared and distinct molecular and phenotypic networks linked to variations in 5-HT homeostasis, we capitalized on the stable genetic variation present in recombinant inbred mouse strains. This family of strains, all derived from crosses between C57BL/6J and DBA/2J (BXD) parents, represents a unique, community resource with approximately 40 years of assembled phenotype data that can be exploited to explore and test causal relationships in silico. We determined levels of 5-HT and 5-hydroxyindoleacetic acid from whole blood, midbrain and thalamus/hypothalamus (diencephalon) of 38 BXD lines and both sexes. All 5-HT measures proved highly heritable in each region, although both gender and region significantly impacted between-strain correlations. Our studies identified both expected and novel biochemical, anatomical and behavioral phenotypes linked to 5-HT traits, as well as distinct quantitative trait loci. Analyses of these loci nominate a group of genes likely to contribute to gender- and region-specific capacities for 5-HT signaling. Analysis of midbrain mRNA variations across strains revealed overlapping gene expression networks linked to 5-HT synthesis and metabolism. Altogether, our studies provide a rich profile of genomic, molecular and phenotypic networks that can be queried for novel relationships contributing risk for disorders linked to perturbed 5-HT signaling.
Serotonin (5-hydroxytryptamine, 5-HT) provides for powerful modulation of physiology and behavior throughout the body and across the life span. Approximately 90% of 5-HT stores in the human body derive from gut enterochromaffin cell 5-HT synthesis and release (Gershon 2013), with a large stable pool found in platelets (Rand & Reid 1951) that acquire the indoleamine during their passage through the enteric circulation. Although blood and brain pools of 5-HT in the adult are separated by the blood–brain barrier (Mann et al. 1992), Bonnin et al. recently reminded us that another peripheral site, the placenta, can produce and release 5-HT that can travel to the central nervous system (CNS) and modulate axon guidance (Bonnin et al. 2011).
Central nervous system regions that regulate mood, emotion, feeding and reward are prominent sites of 5-HT synthesis and release (Steinbusch 1981). As such, a number of disorders have been reported to display alterations in CNS 5-HT homeostasis, including anxiety, depression, obsessive-compulsive disorder and addiction (Barondes 1994). Although a subject of significant debate (Risch et al. 2009), biochemical and genetic evidence continues to drive consideration that risk for depression in some individuals may be linked to a limited capacity for normal brain 5-HT signaling. For example, elevated brain 5-HT turnover has been repeatedly observed in unmedicated patients with depression (Gjerris et al. 1987; Sullivan et al. 2006), with levels returning to normal levels upon treatments with selective serotonin reuptake inhibitors (SSRIs) (Barton et al. 2008). Removal of tryptophan, the essential precursor for 5-HT synthesis, induces relapse in subjects who previously remitted with SSRI therapy (Barr et al. 1994; Delgado et al. 1990). With respect to blood 5-HT, hyperserotonemia (elevated whole-blood 5-HT) has been observed in a sizeable fraction (25–30%) of autistic children for over 50 years (Cook & Leventhal 1996; Cross et al. 2008; Schain & Freedman 1961). Genetic variation in the 5-HT transporter (SERT) promoter (5HTTLPR) has been associated with neuroticism (Lesch et al. 1996), the impact of early life stress on mood disorder risk (Caspi et al. 2003; Karg et al. 2011) and in SSRI response (Arias et al. 2003; Pollock et al. 2000; Serretti et al. 2007). With respect to 5-HT metabolism, deletion in humans of a region encompassing one or both of the 5-HT-metabolizing genes, MAOA and MAOB, has been found in subjects who display a spectrum of developmental and behavioral disorders, including mental retardation, autism-like features and psychosis (Lenders et al. 1996; Saito et al. 2013; Whibley et al. 2010). Finally, functional variation in MAOA has been reported to represent a point of convergence for gene × environment interactions through the moderation of the long-lasting effects of childhood maltreatment (Caspi et al. 2002).
Although the basic elements controlling 5-HT synthesis, storage, response and inactivation have been known for many decades, novel features of the molecular and cellular landscape controlling or responding to 5-HT homeostasis continue to emerge. For example, only within the past decade did Walther et al. (2003a) establish that distinct isoforms of tryptophan hydroxylase (TPH) differentially dictate 5-HT synthesis in the periphery (TPH1) and CNS (TPH2). Serotonylation, a process whereby 5-HT is covalently attached to small G-proteins via the actions of transglutaminase II, has also only recently been recognized as a mechanism by which intracellular 5-HT can impact secretion in platelets (Walther et al. 2003b) and pancreas (Paulmann et al. 2009). To gain a better understanding of functionally relevant networks linked to 5-HT homeostasis and signaling, we exploited recombinant inbred (RI) mice, a community resource of clonal lines derived from C57BL/6J × DBA/2J parents (BXD lines) (Peirce et al. 2004; Taylor et al. 1977, 1999). Thousands of anatomical, biochemical, physiological and behavioral phenotypes as well as single nucleotide polymorphisms (SNPs) have been collected on these lines and are publically accessible via a curated database (http://www.genenetwork.org). This study represents the first systematic analysis of heritability, trait correlations and quantitative trait loci (QTL) based on BXD brain and blood 5-HT traits, efforts that reveal multiple, and in some cases gender-specific, associations. We identify a suite of phenotypes likely to influence, or be influenced by, 5-HT signaling capacity. Finally, we elucidate transcriptional networks linked to variation in 5-HT traits in the BXD lines and nominate candidate genes as determinants of 5-HT homeostasis and signaling.
- Top of page
- Materials and methods
- Supporting Information
BXD RI strains constitute a powerful reference population with which to assess both genotype–phenotype and phenotype–phenotype relationships, as well as to nominate genomic loci that influence measured traits (Houtkooper et al. 2013; Lu et al. 2008). Surprisingly, neurochemical measures are comparatively limited in the GeneNetwork BXD database, including CNS and peripheral levels of 5-HT and 5-HIAA. It should be recognized at the outset that the conclusions we draw from our analyses are limited by the number of lines and animals per line utilized as well as the extent of variation in 5-HT traits exhibited by the parental C57BL/6J and DBA/2J parental lines. For genetic influences that impact 5-HT trait variation more broadly (e.g. after stress or in the context of other genetic variants), no associations can be inferred. Thus, with respect to biochemical, molecular and behavioral correlations pursued with the BXD resource, ‘absence of evidence does not mean evidence of absence’.
Our efforts revealed an approximately twofold variation in the levels of 5-HT, 5-HIAA and 5-HT turnover in both brain and whole-blood extracts across BXD strains, with the across vs. within-strain variation establishing a significant heritability of these measures. Consistent with these findings, human blood 5-HT levels have been found to exhibit high heritability (Weiss et al. 2005). Significant differences in blood 5-HT levels have also been reported in C57BL/6 and BALB/c mouse strains (Flood et al. 2012). 5-Hydroxytryptamine measures in the midbrain, where raphe cell bodies and their intrinsic projections reside, were significantly correlated with the same measures in the diencephalon, a region that receives ascending raphe projections. In contrast, BXD strains did not demonstrate a significant correlation between brain and blood 5-HT measures. These findings are consistent with observations obtained in ICR mice (Pietraszek et al. 1992) and suggest that basal CNS 5-HT homeostasis is greatly impacted by locally derived factors as opposed to determinants shared with blood 5-HT. Enterochromaffin cells utilize TPH 1 (TPH1) to synthesize 5-HT, whereas TPH2 supports 5-HT synthesis neurons (Walther et al. 2003a) and as such, 5-HT synthesis is likely under the control of distinct regulatory networks, not to mention those received in a tissue-specific manner. Nonetheless, certain features of 5-HT homeostasis are shared between blood and brain. For example, the transporter responsible for packaging 5-HT in secretory vesicles for release in neurons (VMAT2, Slc18a2) is expressed by platelets, as is SERT, the chief mechanism of 5-HT clearance in the CNS. Additionally, the MAO isoform MAOB is present in both platelets and raphe neurons (Chen et al. 1993; Jahng et al. 1997). Despite the absence of evidence for molecular relationships dictating blood and brain 5-HT traits in our studies, we must consider that factors shared between these two measures may not vary across BXD strains or may be engaged in a context-dependent manner. Nonetheless, several CNS measures, including behavioral phenotypes, correlated with both brain and blood 5-HT traits. One conspicuous correlation with 5-HT turnover was found for brain weight, hippocampal weight and hippocampal stem cell number. Studies that manipulate 5-HT turnover independent of CNS 5-HT turnover are needed to determine whether these correlations reflect a result of 5-HT homeostasis per se vs. a shared determinant that may in early life impact the trajectory of brain development (Bonnin et al. 2007; Janusonis et al. 2004; Mazer et al. 1997).
Consistent with our earlier report focused on phenotypes linked to SERT coding variation (Carneiro et al. 2009; Ye & Blakely 2011), we found many iron-related traits that significantly correlate with midbrain and diencephalon 5-HT levels. The associations of several phenotypic correlates with brain 5-HT pathways have been extensively studied and add credence to our analysis. For example, adult neurogenesis in the hippocampus significantly associated with midbrain 5-HT and 5-HIAA levels and is believed to play an important role in the chronic actions of SSRIs (Santarelli et al. 2003). Additionally, CNS 5-HT has been shown to be a strong regulator of bone formation and resorption (Yadav et al. 2009). More importantly, we identified several novel or underappreciated phenotypes linked to CNS or blood 5-HT levels, including attributes of the visual, auditory and immune systems. Combined with the candidate gene approaches discussed above, these findings encourage increased attention to these systems and their control by central and peripheral 5-HT signaling. One caveat to these associations is the large number of phenotypes (˜3800) that we accessed for our 5-HT trait correlations. A Bonferroni correction for multiple comparisons requires that we achieve a P value of <1.5 × 10−5. At this stringency level, none of the trait correlations tabulated would reach statistical significance. Thus, we offer the uncorrected tabulation as a hypothesis generator whose true significance must be assessed through other experiments.
In our studies, we analyzed 5-HT trait correlations for both males and females separately as well as in pooled samples where these measures exhibited significant correlations. Although the between-gender comparisons required that we use fewer lines (24 of 38 total) that had both male and female samples, we did find that male and female 5-HT traits from midbrain and blood were significantly correlated, whereas diencephalon traits lacked such a correlation. The latter observation suggests that gender-specific determinants of 5-HT homeostasis act more strongly in distal, forebrain projection areas. It will be interesting to re-examine these relationships using females at specific phases of the estrous cycle. Both Maoa and Maob genes, as well as the Htr2c gene that encodes the 5-HT2C receptor, are located on the X chromosome in mice and humans and this localization may contribute to the gender effects we observed, particularly the lack of gender correlations in MAO-controlled 5-HT turnover measures. Alternatively, one or more of the gender-specific QTLs for CNS 5-HT levels (see below) may be under strong gonadal hormone control. Such an idea is consistent with gender-specific modulation of human brain metabolism by fenfluramine (Anderson et al. 2004) and SSRI medications (Munro et al. 2012) as well as gender-dependent modulation of brain glucose consumption by SERT expression level (Dawson et al. 2009).
Both tissue and gender specificity were further evident in our studies identifying QTLs. For example, while a suggestive QTL on chromosome 10 (mb5htq3) for midbrain 5-HT levels as well as 5-HT turnover ratios also demonstrated suggestive correlation with diencephalon 5-HT levels (females) and 5-HT turnover (males), this locus was not identified as a QTL for blood 5-HT traits (either gender). Similarly, neither suggestive QTLs for blood 5-HT levels (bl5htq1 and bl5htq2) was identified as a QTL for the equivalent traits in midbrain or diencephalon.
The strongest correlation with midbrain 5-HT levels was observed on SNP rs13480650, located within Pcdh15 gene coding for protocadherin 15 (PCDH15). PCDH15 is a cell adhesion molecule belonging to the non-clustered protocadherin subfamily in the protocadherin superfamily, playing important roles in the development of cochlea hair cells and has been associated with human Usher syndrome (Alagramam et al. 2001a,b). Expression of a distinct cadherin gene cluster, protocadherin alpha, is required by serotonergic neuron to form proper projections (Katori et al. 2009), suggesting that multiple adhesive proteins may be critical for 5-HT neuron differentiation and signaling. Indeed, we recently validated that disruption of PCDH15 functions resulted in alterations in CNS 5-HT and SERT levels (Ye & Blakely, submitted). It is also worth noting that another cadherin gene, Cdh23, encoding cadherin 23 (CDH23 or octocadherin), locates within locus mb5htq2. CDH23 physically interacts with PCDH15 and this interaction is required for normal inner ear function (Lelli et al. 2010). Future studies that explore the interaction of CDH23 and PCDH15 in determining 5-HT homeostasis and signaling appear warranted.
In the genes identified from interval mapping of blood 5-HT levels, Nod2 (nucleotide-binding oligomerization domain containing 2) codes for Card15 (caspase recruitment domain-containing protein 15) that is highly expressed in peripheral blood leukocytes (Strober & Watanabe 2011). Card15 is also known as inflammatory bowel disease (IBD) protein 1 owing to its association with IBD/Crohn's disease (Hugot et al. 2001; Ogura et al. 2001) and Blau syndrome (Miceli-Richard et al. 2001). Card15's potential regulation of 5-HT levels is also interesting given the role the protein plays in inducing immune responses to bacterial antigens, including lipopolysaccharide, and the presence of SERT (Marazziti et al. 1998) and 5-HT signaling (Prasad et al. 2005) in the immune system.
Assessment of midbrain gene expression patterns permitted an identification of transcriptional signatures linked to 5-HT traits. In the GeneNetwork BXD database, transcriptome profiles of multiple brain regions have been deposited, including hippocampus, striatum, amygdala and the ventral tegmental area. However, midbrain mRNA expression profiles are lacking, precluding trait correlation analyses with expression variation derived from serotonin neurons. To remedy this situation, we pursued a microarray analysis of male midbrain mRNAs. Surprisingly, little correlation was found between 5-HT traits and the levels of expression of many genes known to control 5-HT homeostasis and signaling. Although at first surprising, this lack of correlation may simply reflect the significant control of these gene products by posttranslational mechanisms (e.g. kinase-mediated regulation of Tph2 activity) that would not be detected in our mRNA analyses. We did uncover significant correlations with Gch1, Slc18a2 and Maoa genes expressed by 5-HT neurons and that are involved in the biosynthesis, packaging and inactivation of the neurotransmitter. The network of genes associated with these molecules provides a window on other biological pathways that may rely on 5-HT signaling. One interesting gene that emerged from these networks is Hcrtr1, which encodes the hypocretin (orexin) receptor type 1. Orexin is a neuropeptide that is involved in the control of feeding behavior (de Lecea et al. 1998; Sakurai et al. 1998) and our NeuroBlast analysis suggests expression by 5-HT neurons. Agents that elicit 5-HT release diminish appetite (Curzon 1990) and Hcrtr1 null mice display behavior similar to that seen with antidepressant treatments (Scott et al. 2011). These connections, along with recent findings of variation in the HCRTR1 gene that associate with unipolar depression, warrant further studies of the control of 5-HT signaling by orexin (Rainero et al. 2011). Finally, we identified multiple 5-HT receptors whose expression significantly correlated with midbrain and diencephalon 5-HT traits. Among the receptors, excitatory 5-HT2A and inhibitory 5-HT1B receptors showed positive correlations with both midbrain 5-HT and 5-HIAA levels, suggesting their potential roles in upregulating 5-HT synthesis. 5-HT1B and 5-HT7 receptors correlated with serotonergic traits measured in both brain regions, indicating their involvement in dictating global 5-HT tone in the brain. These findings, should they relate to relationships present in the human brain, may be of use in the development of better medications for the spectrum of 5-HT-associated neuropsychiatric disorders.
Having established a midbrain 5-HT-associated transcriptome, we used this list to filter the genes identified in our QTL analysis. Genes in this combined list should derive from an impact of cells in the midbrain on 5-HT homeostasis. A caveat to this filtering is the possibility that genes driving variations in midbrain 5-HT could act through long-range connections (e.g. expression in frontal cortex neurons that project to the raphe) but we eliminated owing to the use of a midbrain mRNA data set. Nonetheless, this analysis resulted in 43 genes (Appendix S10). Interestingly, Pcdh15, which we noted above as containing the SNP that marks the peak of the most significant QTL in our study (mb5htq3), is among this list. Additionally, a bootstrap resampling analysis of mb5htq3 yields a location that solely includes the Pcdh15 gene (data not shown). Finally, in a separate study (Ye & Blakely, submitted) examining BXD determinants of SERT protein variation, we again identified Pcdh15 as a regulatory gene dictating CNS SERT protein levels. Together, these efforts point to an unexpected relationship between the cell adhesion protein and the development and/or function of 5-HT neurons, and further support the power of the BXD paradigm.
In summary, this study reports an interrogation of BXD RI mouse strains at multiple levels (neurochemical trait, mRNA and phenotypes) to generate rational hypothesis that could explain the genetic control of 5-HT homeostasis and signaling in both central and peripheral tissues. Despite the caveats noted above, we observed strong genetic control of both 5-HT traits and gene expression profiles. In examining these determinants, we (1) found blood and brain 5-HT traits were controlled by different sets of genetic loci, (2) identified a group of phenotypic traits that correlate with 5-HT blood and/or brain traits and (3) revealed both known and novel genes that associate with 5-HT traits. Given the success of our studies, it is evident that an even more clear picture should emerge through the use of a larger family of recombinant lines, such as those present in the Collaborative Cross, an effort based on RI strains derived from eight inbred and outbred lines, and where genetic variation similar to that of humans (Welsh et al. 2012). Finally, as more phenotypic data (such as protein expression data for 5-HT-related genes) are deposited into the GeneNetwork database, our data sets will offer additional opportunities to dissect the complex actions of 5-HT in the brain and periphery, and hopefully provide clues to disorders associated with 5-HT dysfunction.