A major question for understanding neuronal maturation is: when do 5-HT neurons acquire adult physiological characteristics? There have been few studies to date to investigate the functional maturation of Pet-1+/5-HT neurons in the dorsal and median raphe (Rood et al., 2014; Morton et al., 2015). Rood et al. (2014) compared electrophysiological properties of 5-HT neurons at postnatal days 4, 12, 21, and 60 in mice (Table 1). First, the authors compared membrane characteristics within the dorsal and median raphe. They found that dorsal raphe neurons had a more depolarized resting membrane potential at P4 compared to P12 through P60. Membrane resistance was significantly higher at P4, then decreased at P12 and remained steady through P60. Morton et al. (2015) measured electrophysiological properties of putative 5-HT neurons located in the ventromedial dorsal raphe at P5 to P7 and P15 to P17. Morton et al. (2015) did not find the same changes in membrane properties as Rood et al. (2014), but the changes in action potential properties of threshold, peak amplitude, duration, and afterhyperpolarization amplitude were similar. The differences could be due to the mouse strain background, earlier time-point in Rood et al. (2014), or use of a direct or indirect approach of identifying 5-HT neurons (ePet::EYFP in Rood et al., 2014 and vGat::Venus- in Morton et al., 2015).

A hallmark feature of 5-HT neurons is autoinhibition of firing through 5-HT1AR. When Rood et al. (2014) applied a 5-HT1AR and 5-HT7R agonist (5-carboxamidotryptamine, 5-CT), no significant outward current was observed until P60, although P21 was trending toward significantly greater response than at P4. The 5-HT1AR response not only depends on expression of 5-HT1AR, but also on expression of the downstream G-protein signaling components. The authors used a non-hydrolyzable form of GTP that did not elicit an outward flow of current until P21 and P60. These results indicate that the genes for 5-HT1AR and the relevant G-protein signaling components are not fully activated until roughly the third week of life.

Rood et al. (2014) also evaluated 5-HT neuron dendritic morphology from P4 to P60. Initially at P4, dendrites in the dorsal raphe were shorter and more branched, then the dendrites grew longer until P21, then shortened dramatically by P60. Thus, DRN 5-HT neuron dendrites start out highly branched and seemingly undergo pruning late in maturation to become simpler in branching than P4 neurons with roughly equivalent length. Interestingly, MRN 5-HT neuron dendrites do not change dramatically from P4 to P60. In conclusion, the results found by Rood, et al., demonstrate remarkable alterations in electrophysiological properties and dendritic morphology of dorsal raphe neurons from the first week of life to adulthood representing a maturation process required for normal function in adulthood.

5-HT neurons establish expansive axonal connectivity patterns throughout the CNS. Just after 5-HT neurons are born, they begin sending axons into the brain and spinal cord and reach their target region between late embryonic through the third week of life (Lidov and Molliver, 1982). Surprisingly little is known about the mechanisms regulating 5-HT neuron axon outgrowth, axon guidance, terminal field recognition, and axon ramification. The growth-associated protein GAP-43 is highly expressed in 5-HT neurons and has a strong ascending gene expression trajectory (Wyler et al., 2016). In GAP-43-null mice, there are severe reductions of 5-HT fibers in terminal fields of the frontal cortex and hippocampus, with milder reductions in the piriform cortex and septum, while the ventrobasal thalamus is hyperinnverated (Donovan et al., 2002). The mechanism for how GAP-43 is involved in proper 5-HT neuron innervation patterns remains unknown, although it is possible that GAP-43-null 5-HT neurons are not able to sense NCAM, L1, or N-cadherin signals that could guide 5-HT axons (Meiri et al., 1998).

Another interesting study demonstrated that the Wnt planar cell polarity pathway is required for proper anterior-posterior (A-P) axon guidance of 5-HT neurons (Fenstermaker et al., 2010). The Wnt receptors Fzd3, Celsr3, and Vangl2 are expressed in 5-HT neurons and when deleted have various axon guidance defects. In each receptor mutant, ascending axons from B4-B9 5-HT neurons are partially misrouted laterally or posteriorly. Likewise, descending axons from B1-B3 5-HT neurons are misrouted anteriorly or have fasciculation defects. The Wnt5a ligand is expressed in opposing gradients with high expression at the isthmus decreasing to rhombomere 4, then increasing from r4 posteriorly. Thus, complex expression gradients of Wnt5a direct the correct A-P guidance of 5-HT axons early in 5-HT neuron maturation.

A recent study investigated the function of Ephrin/Eph signaling in 5-HT axon guidance and target region recognition (Teng et al., 2017). The authors showed strong expression of the EphA5 receptor in dorsal raphe 5-HT neurons. Then by knocking out EphrinA5 ligand globally, 5-HT neuron axons from the dorsal raphe aberrantly targeted the ventromedial hypothalamus and the external plexiform layer of the olfactory bulbs, which are normally targeted by median raphe neurons. Overexpression of the EphrinA3 ligand in the amygdala and piriform cortex resulted in reduced innervation of those regions, suggesting EphrinA signaling is sufficient to induce avoidance by 5-HT neurons. Thus, EphrinA/EphA signaling is a key repulsive cue for preventing dorsal raphe 5-HT neurons from innervating at least the ventromedial hypothalamus and external plexiform layer of the olfactory bulb that are normally only innervated by median raphe 5-HT neurons. However, much still remains unknown about the intrinsic molecular pathways controlling 5-HT axon guidance, terminal field recognition, axon branching, and 5-HT neurons special ability to regrow axons after injury (Jin et al., 2016). Furthermore, the transcriptional regulatory factors required to orchestrate these processes within the various raphe are unknown. Dissecting the roles of the transcription factors and their target pathways expressed during 5-HT neuron maturation will likely yield tremendous insight into the complex wiring pattern of 5-HT neurons in the CNS.

A key step in understanding the molecular and functional properties of a specific cell-type is developing a “parts list” comprised of expressed RNAs and proteins for that cell-type from birth of the cell throughout life. The first study to define 5-HT neuron cell-specific transcriptomes, used a combination of fluorescence-activated cell sorting (FACS) and microarray techniques to isolate 5-HT neurons and profile genome-wide gene expression patterns (Wylie et al., 2010). This study used a transgenic mouse line that expresses EYFP in 5-HT neurons under control of the Pet-1 enhancer region (ePet::EYFP), so the authors were able to microdissect the rostral and caudal portions of the hindbrain at E12.5, dissociate cells, then flow sort YFP+ neurons. Comparing expression data between the rostral and caudal samples revealed distinct markers for each group with Hmx2/3 marking rostral Pet-1+/5-HT neurons and a group of Hox genes marking caudal 5-HT neurons. In the rostral group, an additional subtype was defined by Engrailed gene expression resulting in Hmx⁺En⁺ and Hmx⁺En⁻ rostral subtypes.

Many studies have used transgenic mouse lines that express EYFP or Cre recombinase under the control of the Pet-1 enhancer region (ePet::EYFP, ePet::Cre, Scott and Deneris, 2005; Scott et al., 2005; Liu et al., 2010; Fox and Deneris, 2012). The ePet promoter/enhancer fragment in these constructs directs gene expression to the majority of 5-HT neurons in the various raphe nuclei (Scott et al., 2005; Zhao Z.-Q. et al., 2006). Some of these Pet-1+ cells do not present an overt 5-HT identity (Tph2+/5-HT+) when assayed with standard in situ hybridization or immunohistological protocols. The number of Pet-1+ cells without detectable 5-HT identity was reported to be about 27% in the MRN while in the B7 and B6 portions of the DRN this number was only 1% (Pelosi et al., 2014). These findings were interpreted as evidence that not all Pet-1+ cells are 5-HT neurons (Sos et al., 2016). However, recent single cell RNA seq studies have revealed that all of the single Pet-1+ cells sequenced in the DRN and MRN express Tph2, although some had a much lower number of Tph2 reads that other Pet-1+ cells (Okaty et al., 2015). The Pet-1+ cells with low Tph2 were all from the MRN and no Pet-1+ cell profiled from the DRN had a low number Tph2 reads. This was corroborated with single molecule FISH. These findings indicate that the vast majority and perhaps even all Pet-1+ cell in the DRN and MRN are 5-HT neurons, but some do not express strong 5-HT (Tph2) identity. Thus, recent optogenetic studies of the DRN using ePet::Cre transgenic line were likely manipulating nearly exclusively 5-HT neurons (Liu et al., 2014).

The only study that profiled adult 5-HT neuron gene expression by microarray used the translating ribosome affinity purification method (TRAP, Dougherty et al., 2013). The TRAP method relies on expressing a GFP-tagged ribosomal protein subunit, L10a. The authors used a Slc6a4 BAC to drive expression of the GFP-L10a construct in 5-HT neurons. Since L10a is a component of the ribosome, mRNAs that are being actively translated can be crosslinked to the ribosome and affinity-purified. The adult microarray expression profile overlapped significantly with the results from Wylie et al. (2010) despite the age difference and several sample preparation differences. Interestingly, when the investigators intersected their gene expression profile with data from the Autism Genetic Research Exchange (AGRE, Geschwind et al., 2001) they found a 5-HT neuron enriched gene, Celf6, with an inherited deleterious mutation in an affected individual. By generating a genetic knockout of Celf6 in mice, the authors found several autism-related phenotypic behaviors including resistance to change and decreased ultrasonic vocalizations. Intriguingly, the Celf6 knockout mice also had mildly decreased whole-brain levels of 5-HT, suggesting a disruption in 5-HT metabolism. Although Celf6 is predicted to be an RNA binding protein, it is not yet clear what role it plays in 5-HT neurons. Overall, the adult 5-HT neuron expression profile provides a useful dataset for testing gene function in adulthood.

To define gene expression trajectories of maturing 5-HT neurons from early fetal to early postnatal stages, Wyler et al. (2016) isolated highly purified populations of Pet-1+ neurons from the rostral raphe nuclei (DRN and MRN combined) at E11.5, E15.5, and P1-P3 (PN) for RNA sequencing analysis. Hierarchical cluster analysis revealed genes that significantly changed in expression level through all three time points. The resulting clusters clearly delineated groups of genes with highly related ascending or descending temporal expression profiles. The clusters with descending trajectories contained genes with high expression at E11.5 and decreased through E15.5 to significantly lower levels by early postnatal stage. Gene ontology analysis showed that the genes in these clusters were enriched for transcriptional regulator, chromatin remodeler, and embryonic development genes that were expected for genes being highly expressed during the progenitor stage of 5-HT neurons. Concurrently, the gene clusters with ascending trajectories were off or lowly expressed at E11.5 and increased expression through the early postnatal stage. These clusters were enriched for genes involved in axonogenesis, ion channel activity, synaptic vesicles, and synaptic plasticity signifying the process of transitioning to a mature functional state of 5-HT neuron identity (Figure 3). While these results present a comprehensive transcriptional profile of 5-HT neuron maturation, they do not extend into the third week of life when Rood et al. (2014) showed 5-HT neurons are functionally mature. This can be resolved by considering the adult microarray profile (Dougherty et al., 2013), as well as, single-cell RNA-seq of 5-HT neurons at around 1 month of age (Okaty et al., 2015). The authors used manual sorting of neurons derived from r1, r2, r3, r5, and r6-r8 and performing RNA-seq on single cells or pools of cells, Okaty et al. (2015) provided a comprehensive transcriptional profile of multiple 5-HT neuron sub-types that are functionally mature. Indeed, many of the genes with ascending trajectories in Wyler et al. (2016), such as Adra1b, Gria2, Lpar1, and Htr1a are highly expressed in the data from Okaty et al. (2015). It is now clear that extensive temporal transcriptional changes underlie 5-HT neuron maturation, which raises the question of what regulatory strategies are employed to control these changes and promote the maturation process. Are terminal selector TFs involved in shutting off the early specification program and activating the later maturation program? Do other regulatory factors get activated early in maturation and drive higher gene expression in later ages?

As discussed previously, Pet-1 functions as a terminal selector in 5-HT neurons by activating the core 5-HT neuron gene battery, autoregulating its own expression, and maintaining expression throughout the life of the cell (Deneris and Hobert, 2014). It was not known whether Pet-1 is involved in the functional maturation of 5-HT neurons. Previous work has shown that Pet-1-deficient neurons have increased spontaneous firing frequency and lack 5-HT1AR autoregulation (Liu et al., 2010). To determine if Pet-1 knockout neurons are functionally immature, Wyler et al. (2016) tested several passive and active membrane properties. Interestingly, Pet-1 knockout neurons had a depolarized resting membrane potential, increased membrane resistance, and an increased membrane time constant (tau, Table 2). These properties correspond with the previously reported immature 5-HT neuron phenotype at early postnatal ages (P4–P12, Rood et al., 2014). Action potential (AP) characteristics were also disrupted in Pet-1 knockout neurons. The AP amplitude was decreased with a hyperpolarized AP firing threshold. The afterhyperpolarization amplitude was also decreased. These passive and active membrane properties contribute to neuronal excitability and measuring current-induced excitability of Pet-1^−/− neurons demonstrated a permanent increase in excitability through adulthood. It is important to note that Pet-1^−/− neurons are maintained in the same numbers as in wildtype or control animals and do not appear to be directed to an alternative cell-fate. Therefore, the changes in functional properties are not due to recording from dying cells or adoption of functional properties of another neuron-type. These findings together reveal that Pet-1 function is not limited to the induction of 5-HT transmitter identity, but also for the acquisition of mature electrophysiological properties.

The establishment of Pet-1 as a TF required for functional maturation raises the question of what gene categories are under control of Pet-1 during maturation. It is clear from the gene expression trajectories from E11.5 to the early postnatal period, 5-HT neurons undergo tremendous remodeling of their transcriptomes, while maintaining their core 5-HT transmitter identity (Figure 3). Wyler et al. (2016) tested whether Pet-1 controls maturation of gene expression trajectories and how any disruptions relate to the functional immaturity of Pet-1 deficient 5-HT neurons. Multiple broad categories of genes encoding GPCRs, ion channels, adhesion molecules, and transcription factors were found to be regulated by Pet-1, both positively and negatively. A significant proportion of these genes were also found to be directly bound by Pet-1 using a chromatin-immunoprecipitation sequencing (ChIP-seq) strategy. The disrupted expression of genes involved in many different 5-HT neuron relevant pathways emphasizes the broad regulatory role of Pet-1 in establishing the mature 5-HT neuron identity.

Cell-types within the nervous system are categorized based on a variety of characteristics including location, morphology, neurotransmitter usage, activity, and gene expression. Investigation of how these specific cell-types arise is often focused on identifying a transcription factor that is required to activate the genes required for that cell-type. Each cell-type, however, may be broad in definition with many distinct sub-types. These sub-types may have different morphologies, alternate release or co-release of other transmitters, and functional characteristics. 5-HT neurons are such a class of neurons with a variety of sub-types (Wylie et al., 2010; Spaethling et al., 2014; Okaty et al., 2015; Fernandez et al., 2016). 5-HT neurons can be divided by rostral vs. caudal projections, brain region specific innervation patterns, afferent inputs, vesicular glutamate transporter expression (vGlut3+), neuropeptide usage, and electrophysiological properties. The regulatory network strategies for creating such diversity are not well elucidated. A recent study by Kerk et al. (2017), investigated the regulatory strategy controlling cholinergic neuron diversity in C. elegans. The cholinergic motor neuron terminal selector UNC-3 regulates ACh pathway genes, ion channels, signaling proteins, and neurotransmitter receptors in these neurons, yet different sub-types are defined by anatomical and functional traits. The terminal selector TF, UNC-3, is expressed in all of these neurons, therefore it is not able to selectively activate these different features alone. The expression of co-activators or repressors are two possible strategies for generating the various sub-types. Kerk et al. (2017) use genetic screens to test whether either or both strategies are utilized in C. elegans cholinergic motor neurons. Their study revealed a surprisingly consistent use of repressors to inhibit specific genes from being activated by UNC-3 in the different sub-types. Thus, UNC-3 has the role of broadly activating cholinergic neuron genes and then other repressor proteins override the UNC-3 program to diversify cholinergic neuron sub-types. In contrast, Pet-1 functions as a terminal selector for mouse 5-HT neurons and has shown to act as both an activator and repressor (Fyodorov et al., 1998; Maurer et al., 2003; Wyler et al., 2016). Pet-1 represses Nkx2-2, Hcrtr1, and Penk as mentioned above. Penk encodes the enkephalin peptide and is highly expressed in caudal 5-HT neurons (Okaty et al., 2015). This suggests Pet-1 is repressing a caudal 5-HT neuron gene to separate the rostral and caudal sub-types. Pet-1 expression is maintained in caudal 5-HT neurons, so it is unclear how Penk expression is increased in caudal neurons relative to rostral neurons. Other transcription factors that are expressed and/or regulated by Pet-1 in 5-HT neurons likely contribute to 5-HT neuron heterogeneity including Lmx1b, Gata2, Gata3, En1, and others described above (Ding et al., 2003; Zhao Z.-Q. et al., 2006; Zhao et al., 2007; Song et al., 2011; Yan et al., 2013; Haugas et al., 2016).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.