Molecular regulation of axon termination in mechanosensory neurons Free

Nervous system wiring requires precise execution of a genetically programmed developmental plan. During development, axons grow and navigate to their target site where they terminate growth in a spatially and temporally accurate manner, a process referred to as axon termination. Accurate, efficient axon termination is required to generate laminated termination zones that occur throughout the nervous system (Feldheim and O'Leary, 2010; Grueber and Sagasti, 2010; Lopez-Bendito and Molnar, 2003). In all organisms, the nervous system must wire efficiently in a physically restricted body plan, which represents a complex biological problem highlighted by recent studies mapping axon termination zones in the rodent brain (Gao et al., 2022; Harris et al., 2019). Although not our focus, recent studies found that multiple molecules that regulate termination are associated with neurodevelopmental disorders, and regulate axon regeneration and degeneration (AlAbdi et al., 2023; Chen et al., 2006b, 2015; Drozd and Quinn, 2023; Opperman et al., 2017; Xiong et al., 2012). Thus, understanding the molecular genetic programs for termination has implications for biomedical research. Despite its importance, axon termination remains relatively understudied compared with other neurodevelopmental processes, such as axon guidance (Chedotal, 2019) or synapse formation (Shen and Scheiffele, 2010).

Here, we discuss axon termination in mechanosensory neurons across several model organisms. We focus on mechanosensory neurons because their atypical, bifurcated axon anatomy requires termination programs in central and peripheral zones, and our understanding of axon termination is particularly well developed in mechanosensory neurons. Our discussion centers on low-threshold mechanoreceptors (LTMRs), also called gentle touch receptor neurons (Chalfie, 2009; Zimmerman et al., 2014), but principles likely pertain to pain-sensing nociceptors as well (Basbaum et al., 2009; Dubin and Patapoutian, 2010).

We first discuss the nematode Caenorhabditis elegans, for which we have the most knowledge about mechanisms regulating axon termination. C. elegans mechanosensory neurons share several similarities with mammalian counterparts in dorsal root ganglia (DRG). This includes anatomically precise axon termination, unipolar axon anatomy, bifurcated multi-functional axons that sense and relay touch information, and glutamatergic chemical transmission. Thus, it is possible that regulators of axon termination in C. elegans are relevant for vertebrates.

Axon termination studies in C. elegans principally utilize anterior lateral microtubule (ALM) or posterior lateral microtubule (PLM) neurons. These neurons are ideal for studying termination for several reasons. First, their single unipolar axons display spatially and temporally precise termination (Fig. 1A) (Borgen et al., 2017; Chalfie et al., 1985; Gallegos and Bargmann, 2004), and termination sites form tiled patterns reminiscent of zebrafish and mammals (discussed later). Termination studies are not complicated by glutamatergic chemical synapses, which are located on a collateral synaptic branch and mediate mechanosensory habituation (Fig. 1A) (Crawley et al., 2017; Giles et al., 2015; Rankin and Wicks, 2000). Finally, termination is not affected by electrical gap junction synapses, which mediate initial touch responses and are located primarily on the axon base (Chalfie et al., 1985; Meng et al., 2016; Wicks and Rankin, 1995). Although a small number of electrical connections occur at PLM termination sites, results suggest that formation of electrical connections is largely distinct from termination (Borgen et al., 2017; Zhang et al., 2013).

Mechanosensory axon development in C. elegans proceeds through three phases (Gallegos and Bargmann, 2004): (1) immediately after hatching, growth cones (GCs) are present and axon guidance occurs during a period of rapid outgrowth; (2) axon growth slows; and (3) axons stretch without GCs as the animal increases in size. GCs are present during the first developmental phase (Mohamed et al., 2012; Zhang et al., 2013). In the second phase, GC collapse during termination is protracted, requiring several hours for GCs to transition from dynamic to static states (Borgen et al., 2017). This protracted collapse process with a static state differs notably from rapid GC collapse and re-emergence, which occurs in minutes when axons change trajectory during guidance but do not terminate growth. Although protracted GC collapse is a cellular distinction of axon termination versus guidance, very few regulators of axon termination have been pursued at this depth in C. elegans or other models.

Early forward genetic studies with mechanosensory neurons identified mutants with growth defects, which included ectopic branching or premature termination (Chalfie and Sulston, 1981; Hedgecock et al., 1987; Siddiqui, 1990). Below, we discuss how complementary genetic and proteomic approaches have delved extensively into regulatory mechanisms of axon termination. Although we aim to provide a comprehensive overview of the axon termination process, we suggest readers return to the primary literature for complete details on whether a given player specifically affects termination in PLM and/or ALM neurons.

Ubiquitin ligases are required for axon termination of both PLM and ALM mechanosensory neurons. Findings to date indicate that a complex ubiquitin ligase network likely orchestrates the termination of axon growth in C. elegans.

The most extensively studied E3 ubiquitin ligase is regulator of presynaptic morphology 1 (RPM-1), which is orthologous to human MYCBP2 (PAM), mouse Phr1 (Mycbp2), zebrafish Mycbp2, and Drosophila Highwire (Grill et al., 2016; Virdee, 2022). These Pam/Highwire/RPM-1 (PHR) proteins form atypical RING ubiquitin ligase complexes with the F-box protein FSN-1/FBXO45 and SKR-1/SKP1 (Desbois et al., 2018; Pao et al., 2018; Saiga et al., 2009).

RPM-1 is an excellent example of a key player in axon termination, as rpm-1 loss-of-function (lf) mutants have severe failed termination defects (i.e. axon overgrowth beyond the normal termination site) for both ALM and PLM neurons (Fig. 1B) (Schaefer et al., 2000). Overexpressing RPM-1 yields the opposing phenotype: premature termination defects (Borgen et al., 2017; Li et al., 2008). RPM-1 localizes to terminated axon tips (Grill et al., 2007; Opperman and Grill, 2014), and physically interacts with FSN-1 in this subcellular compartment (Desbois et al., 2018). This interaction is important because termination is affected by a peptide that impairs RPM-1 binding to FSN-1 (Sharma et al., 2014). fsn-1 (lf) mutants also display termination defects, although defects are milder and less frequent than those in rpm-1 mutants (Grill et al., 2007; Liao et al., 2004). Importantly, RPM-1 functions as a ubiquitin ligase signaling hub, a concept that becomes clear below.

Three other ubiquitin ligases regulate termination: the F-box proteins LIN-23 and MEC-15, and the HECT ubiquitin ligase EEL-1 (Fig. 1B,D). LIN-23 has pleiotropic effects on termination, as lin-23 hypomorphic mutants show failed termination whereas null mutants show premature termination (Mehta et al., 2004). mec-15 (lf) mutants display premature termination, indicating an inhibitory role in termination (Zheng et al., 2020). Failed termination is observed in eel-1 mutants with genetic enhancer effects occurring in eel-1;fsn-1 double mutants (Opperman et al., 2017). LIN-23, MEC-15 and EEL-1 are conserved in humans and homologous to FBXW11, FBXW9 and HUWE1, respectively. LIN-23 and MEC-15 F-box proteins are likely to form multi-subunit ubiquitin ligase complexes with RING proteins, but the nature of these ubiquitin ligase complexes remains unknown.

Ubiquitination substrates for EEL-1, LIN-23 and MEC-15 that mediate effects on termination remain unclear. However, RPM-1 ubiquitinates and inhibits several kinases to influence termination, thus we move to a discussion of kinases next.

Several kinases and kinase cascades negatively and positively regulate axon termination. For example, two serine/threonine kinases, UNC-51/ULK and SAX-1/STK38, which function in termination, display opposing regulatory roles (Chen et al., 2011; Du and Chalfie, 2001; Gallegos and Bargmann, 2004). Premature PLM neuron termination occurs in unc-51 (lf) mutants, whereas sax-1 (lf) mutants display failed termination and overgrowth (Fig. 1C,D). RPM-1 ubiquitinates UNC-51, targeting it for proteasome-mediated degradation (Fig. 2A) (Crawley et al., 2019). SAX-1 functions in a linear genetic pathway with the microtubule-binding protein SAX-2/Furry to positively regulate termination (Gallegos and Bargmann, 2004). Genetic evidence that SAX-1 functions upstream of SAX-2 is absent, but it is possible SAX-1 could phosphorylate SAX-2. Another positive regulatory kinase similar to SAX-1 is NEKL-3/NEK7, which is regulated by an UNC-116 Kinesin/UNC-16 JIP3 (MAPK8IP3) complex (Fig. 1C) (Drozd and Quinn, 2023). CDK-5 is a further kinase that negatively regulates termination in ALM neurons (Fig. 2A) (Desbois et al., 2022). CDK-5 displays substrate-like interactions with the RPM-1/FSN-1 complex, CDK-5 is inhibited by RPM-1 ubiquitin ligase activity, and CDK-5 overexpression triggers failed termination and overgrowth. Finally, impairing the LRK-1/LRRK2 kinase results in failed ALM termination (Fig. 1B) (Kuwahara et al., 2016). Thus, the involvement of these five kinases indicates that accurate axon termination requires complex layers of kinase signaling.

Kinase cascades (composed of multiple kinases that sequentially activate one another) also regulate termination with the DLK cascade being the most studied (Fig. 2A). Full-length DLK-1 (DLK-1L) homodimerizes to initiate activation of its cascade, which influences axonal mRNA stability for the transcription factor cebp-1 (Yan and Jin, 2012; Yan et al., 2009). Whether CEBP-1 also influences transcriptional changes required for termination remains uncertain. Interestingly, complex layers of regulatory mechanisms converge on the DLK pathway. Two mechanisms negatively regulate DLK-1: RPM-1/FSN-1 which ubiquitinates and inhibits DLK-1; and a short inhibitory DLK-1S isoform which heterodimerizes with DLK-1L forming inactive heterodimers (Grill et al., 2007; Liao et al., 2004; Nakata et al., 2005; Yan and Jin, 2012). MEC-15 also affects DLK-1 protein levels in a similar manner to RPM-1 (Zheng et al., 2020). However, mec-15 mutants display the opposing termination phenotype to rpm-1 mutants. Why RPM-1 and MEC-15 both affect DLK-1 levels but yield opposing mutant phenotypes remains unclear. The ubiquitin E2 conjugating variant UEV-3 acts downstream in the DLK kinase cascade to inhibit PMK-3/p38 MAPK (Trujillo et al., 2010). Although little is known about how DLK-1 is activated in C. elegans, genetic loss- and gain-of-function studies suggest that MIG-2 Rac/RhoG impairs termination by activating DLK-1 (Fig. 2B) (Borgen et al., 2017).

The MLK-1 cascade, which includes the Jnk isoform KGB-1, is also inhibited by the RPM-1/FSN-1 complex (Fig. 2A) (Baker et al., 2015; Nix et al., 2011). Consistent with a role in preventing axon termination, MLK-1 or KGB-1/JNK overexpression causes failed PLM termination (Baker et al., 2015). Weaker phenotypes and weaker genetic suppression of RPM-1 by MLK-1 compared with DLK-1 indicate that the MLK cascade plays a more minor role compared with the DLK cascade.

Together, these observations highlight that axon termination is influenced by numerous kinases and their downstream cascades, which form a kinase signaling network (Fig. 2A). For kinases not in cascades, the downstream phosphorylation targets that influence termination remain unclear. Furthermore, the RPM-1 ubiquitin ligase signaling hub represents a key inhibitory mechanism that restrains numerous kinases in this network (Fig. 2A).

Phosphatases represent a further inhibitory mechanism that regulates the DLK and MLK cascades (Fig. 2A). Two phosphatases regulate termination: protein phosphatase Mg2⁺/Mn2⁺ dependent-1 (PPM-1) and PPM-2, which are single-subunit PP2C serine/threonine phosphatases.

PPM-1 functions in parallel with FSN-1 to restrict DLK and MLK-1 signaling with PPM-1 likely acting on MKK-4 or PMK-3/p38 MAPK in the DLK pathway, and MEK-1 or KGB-1/JNK in the MLK pathway (Baker et al., 2015; Tulgren et al., 2011). PPM-2 is an RPM-1-binding protein that functions in the RPM-1 signaling network to restrict DLK-1 directly via dephosphorylation at a residue required for DLK-1S binding (Baker et al., 2014). Thus, RPM-1 restrains DLK-1 via PPM-2 recruitment and ubiquitin ligase activity. Whereas PPM-1 broadly inhibits both the MLK and DLK pathways, PPM-2 is a more specific inhibitory mechanism that restrains only DLK-1.

At present, relatively little is known about how players that influence termination affect the protracted GC collapse process that is required for axon termination. One player that has been studied at this depth is RPM-1 (Borgen et al., 2017). RPM-1 restricts GC size and regulates GC collapse by affecting the second static portion of the collapse process. Indeed, RPM-1 is initially localized at low levels in GCs during the developmental window for collapse and later concentrates at terminated axon tips (Borgen et al., 2017; Opperman and Grill, 2014). Thus, RPM-1 likely influences local signaling in subcellular axonal compartments to regulate GC collapse and termination.

Several groups have examined β- and α-tubulins. Initially, there were differing reports on termination defects in mec-7/β-tubulin and mec-12/α-tubulin mutants with reports of premature termination (Du and Chalfie, 2001; Lockhead et al., 2016) or failed termination with overgrowth defects (Chen et al., 2014). To solidify the field, the Zheng and Chalfie labs evaluated a staggering number of mec-7 and mec-12 alleles (Lee et al., 2021; Zheng et al., 2017). Mild premature termination defects occur in mec-7 (lf) mutants, whereas mec-12 (lf) mutants do not display defects. Dominant-negative mutants for mec-7 or mec-12 yield greater reductions in microtubule stability and display severe premature termination (Lee et al., 2021; Zheng et al., 2017). Multi-tubulin knockouts do not increase premature termination defects, but reveal that two α-tubulin isoforms, MEC-12 and TBA-1, function in parallel (Lockhead et al., 2016). Studies with the microtubule-destabilizing drug colchicine provides further evidence that reducing microtubule stability results in premature termination (Borgen et al., 2017; Lockhead et al., 2016).

EFA-6, a guanyl-nucleotide exchange factor, is known to be required for microtubule catastrophe, and efa-6 (lf) mutants display failed termination defects (Chen et al., 2015, 2011). In contrast, impairing microtubule stabilizers, such as the tubulin acetyltransferases atat-2 and mec-17, the minus-end binding protein ptrn-1 or the microtubule-binding protein ptl-1/Tau (Mapt) causes premature termination (Figs 1D and 2B) (Borgen et al., 2017). The RNA-binding protein MBL-1 is required for the stability and expression of mec-12 and mec-7 RNAs with mbl-1 mutants displaying premature termination (Fig. 1D) (Puri et al., 2023).

RPM-1 signaling displays several further links to microtubules (Fig. 2B). First, RPM-1 binds and functions in part via the microtubule-binding protein RAE-1 (Baker and Grill, 2017; Grill et al., 2012). Second, results with colchicine and taxol, drugs that destabilize and stabilize microtubules respectively, indicate that RPM-1 is a microtubule destabilizer (Borgen et al., 2017). Third, RPM-1 functions coordinately with KIN-18/TAOK1 while opposing MEC-17 and ATAT-2 (Borgen et al., 2017). Fourth, the DLK-1 cascade, which RPM-1 inhibits, influences microtubules (Borgen et al., 2017; Ghosh-Roy et al., 2012). Finally, the microtubule-binding protein PTL-1/Tau potentially inhibits RPM-1, an intriguing observation as this suggests the RPM-1 signaling hub is potentially regulated by a prominent neurodegenerative molecule. Thus, a consensus emerges that reduced microtubule stability causes premature termination and increased microtubule stability generates failed termination and overgrowth.

Regulators of the actin cytoskeleton also facilitate precise termination based on two observations. First, PLM axons display premature termination when the actin elongation factor UNC-34/Ena, is impaired (Mohamed et al., 2012). Second, opposing failed termination defects occur with WSP-1/Wasp (lf), which activates actin branching. Indeed, F-actin is enriched in mechanosensory GCs (Borgen et al., 2017). This indicates that more complex, branched actin networks facilitate termination and actin elongation inhibits termination and promotes growth.

Further evidence for actin modulation has emerged from studies on Rac GEFs (guanine nucleotide exchange factors) and Rac GTPases. Rac GEF unc-73/Trio (lf) mutants display severe premature termination similar to other Trio GEF components such as UNC-53/Nav1 and SEM-5/Grb2 (Fig. 1D) (Zheng et al., 2016). The Rac isoforms CED-10 and MIG-2 function redundantly with double mutants displaying premature termination defects. Studies on GC collapse have examined other regulators of actin, including UNC-33/CRMP, CDC-42 and RHO-1 (Fig. 2B) (Borgen et al., 2017). Inhibiting these players causes mild failed termination defects and enhances rpm-1 (lf). In contrast, the Rac isoforms CED-10 and MIG-2 suppress failed termination caused by rpm-1 (lf), whereas mig-2 gain-of-function (gf) mutants show severe failed termination. Thus, MIG-2 and CED-10 Rac isoforms and the TRIO Rac GEF have opposing effects to CDC-42 and RHO during termination.

The importance of proteostasis was initially supported by studies with ubiquitin ligases and proteasome-mediated protein degradation. Findings on autophagy and late endosome/lysosome biogenesis have further solidified the importance of proteostasis in axon termination.

The role of late endosomes was first revealed by identification of the Rab GEF GLO-4 as an RPM-1-binding protein (Grill et al., 2007). RPM-1 positively regulates the GLO pathway, which consists of GLO-4, the Rab GTPase GLO-1 (orthologous to Rab32/38) and AP-3/APM-3 (Fig. 2C). Impairing GLO pathway components yields mild failed termination defects with enhanced phenotypes occurring in double mutants of GLO components with FSN-1. The LRK-1/LRRK2 kinase functions downstream of GLO-1 (Kuwahara et al., 2016). Recently, the dynein-binding function of UNC-16/JIP3 and effects on LRK-1 kinase have been found to promote elimination of axonal late endosomes during termination (Drozd et al., 2024). Thus, efficient biogenesis and trafficking of axonal late endosomes is required for termination of axon growth.

With regard to autophagy, proteomic substrate ‘trapping’ identified the autophagy-initiating kinase UNC-51/ULK as an RPM-1 substrate (Fig. 2A) (Crawley et al., 2019). Using CRISPR-tagged UNC-51 and autophagosome markers, RPM-1 was found to restrict both UNC-51 levels and autophagosome formation at axon termination sites. Results with other autophagy components (ATG-9, EPG-8, BEC-1 and EPG-6) indicate that RPM-1 inhibits UNC-51 to restrict autophagy. Studies with the voltage-gated calcium channel EGL-19 have revealed further links with autophagy (Buddell et al., 2019). egl-19 (gf) mutants, which mimic human genetic variants associated with autism, exhibit mild failed termination (Fig. 1C). Overgrowth is suppressed by impairing the autophagy components WDFY-3/Alfy and EPG-7, or altering the lysosome biogenesis regulator CUP-5/mucolipin. A final link to autophagy is potentially UNC-69/SCOC and UNC-76/FEZ1, which form a complex with both mutants displaying premature termination (Chen et al., 2011; Su et al., 2006). Whether this complex affects autophagy in the nervous system is unclear, but studies in non-neuronal cells have shown effects on autophagy (McKnight et al., 2012; Wirth et al., 2021). Collectively, these studies indicate that axonal autophagy inhibits termination with genetic perturbations that increase autophagy having a pro-growth effect during development.

The first evidence that signaling in the soma affects distal termination events in the axon stems from findings on the RPM-1 network, which has components that function at the nuclear envelope (Tulgren et al., 2014). Consistent with this, RPM-1 localizes to the PLM soma during the developmental window of termination (Borgen et al., 2017). RPM-1 binds and functions via ANC-1/nesprin (also called SYNE1), which interacts with UNC-84/SUN in a nuclear anchoring complex (Fig. 2C) (Tulgren et al., 2014). RPM-1 positively regulates ANC-1, which functions in parallel with GLO-4 and FSN-1. ANC-1 also inhibits the emerin homolog EMR-1/LEMD1 to prevent nuclear export of BAR-1/β-catenin, thereby facilitating POP-1/TCF transcription factor activity.

RPM-1 signaling in the soma also inhibits CDK-5 (Fig. 2A) (Desbois et al., 2022). CRISPR-engineered tags on RPM-1 and CDK-5 indicate that CDK-5 shows substrate-like interactions with RPM-1 in the soma, but CDK-5 has not been detected at the termination site in the axon. This suggests that RPM-1 restricts CDK-5 in the soma, but we note that low sensitivity of CRISPR reagents could prevent visualization of axonal CDK-5.

In C. elegans, extracellular cues and receptors regulating axon termination remain poorly understood. Receptors and ligands identified have relatively small roles in termination, which suggests that a complex receptor code potentially halts axon outgrowth.

The first receptor implicated in termination was the VAB-1 Eph receptor (Mohamed and Chin-Sang, 2006). vab-1 (lf) mutants display mild, low-penetrance failed termination defects in PLM neurons similar to triple knockouts for ephrins (EFN-2, EFN-3 and EFN-4). VAB-1 regulates WSP-1/Wasp and the Arp2/3 actin regulator via the NCK-1 adaptor protein (Mohamed et al., 2012; Mohamed and Chin-Sang, 2011). Live imaging in PLM neurons has demonstrated that VAB-1 (gf) reduces dynamic filopodial extensions in GCs, precipitating premature termination (Mohamed et al., 2012). VAB-1 functions in parallel to the FSN-1 component of the RPM-1 network (Chang et al., 2024). If and how RPM-1 is regulated by receptors remains unknown.

Because Wnts affect axon polarity, effects on axon termination must be carefully examined and are often evaluated in individual neurons not displaying substantial polarity defects. PLM neurons of lin-44 Wnt (lf) mutants have mild failed termination defects (Tulgren et al., 2014). lin-44; fsn-1 double mutants display increased termination defects, and findings with fsn-1 double mutants implicate a second Wnt, egl-20, in termination. A second study focused on posterior PLM process development has argued that LIN-44 functions as a repulsive cue to facilitate axon outgrowth rather than controlling axon polarity (Zheng et al., 2015). Although still debated, severely truncated axons in lin-44 mutants, often considered polarity defects, might represent severe premature termination defects.

Extracellular cues also influence axons when termination fails, such as in rpm-1 (lf) mutants. Under this failed termination scenario, two guidance systems have been suggested to influence PLM axons after they overgrow: (1) UNC-6/netrin and its receptor UNC-5, and (2) SLT-1/Slit and its receptor SAX-3/Robo (Li et al., 2008). Subsequent independent studies showed that unc-5 or unc-40 netrin receptor mutants do not suppress failed termination in rpm-1 mutants (Crawley et al., 2019). Rather, unc-5 and unc-40 mutants display mild failed termination defects in ALM neurons, which are enhanced by rpm-1 (lf) (Crawley et al., 2019). To date, no follow-up studies have examined how Slit and Robo affect termination defects in rpm-1 mutants.

Collectively, these findings indicate there is not a single receptor ‘stop’ cue that triggers termination. Rather, there are potentially multiple positive and negative cues forming a termination code. Although this is an intriguing possibility, solidifying the role of Ephrin, Wnt, Netrin and Robo receptors in termination awaits further testing. This is also not necessarily the full complement of receptors required for axon termination in C. elegans mechanosensory neurons.

Drosophila melanogaster contains multiple touch-sensing organs and several types of mechanosensory neurons that differ between larva and adult (Figs 3 and 4). For the genetics and circuitry of touch sensation, we recommend other reviews (Karkali and Martin-Blanco, 2017; Tuthill and Wilson, 2016). We focus on axon termination in embryonic and larval mechanosensory neurons before pivoting to adults. Given the limited understanding of how termination is regulated in Drosophila mechanosensory neurons, we also comment on progress with nociceptive and proprioceptive neurons. We discuss only central axon termination sites because flies have bipolar touch neurons with elaborate peripheral dendritic arbors, and mechanisms of dendrite tiling have been reviewed previously (Grueber and Sagasti, 2010).

Embryonic and larval flies have three groups of mechanosensory neurons. Two are multidendritic (md) neurons: dendritic arborization (da) neurons sense gentle touch and pain, and dorsal bipolar dendritic (dbd) neurons sense proprioception. Chordotonal (ch) neurons are a third group of neurons that sense gentle touch in specialized chordotonal organs. da (da I-IV), dbd and ch neurons send single axons into the central nervous system where they terminate in precise locations in the neuropil (Fig. 3) (Grueber et al., 2007; Merritt and Whitington, 1995; Schrader and Merritt, 2000). Different mechanosensory neurons terminate in anatomically distinct sites that form tiled laminar patterns (Grueber et al., 2007; Merritt and Whitington, 1995; Schrader and Merritt, 2000). Recently, drivers for class cIVda and cIIIda neurons showed that touch-sensing cIII and nociceptive cIV neurons terminate in adjacent laminated central zones (Galindo et al., 2023). Laminated central termination is an evolutionarily conserved principle because this also occurs with different classes of mammalian LTMRs and nociceptors (discussed later). Recent live imaging with fly somatosensory neurons has begun to establish the developmental windows for central axon development, which could prove valuable for future studies on termination (Galindo et al., 2023). ch and dbd neurons terminate at distinct locations in the neuropil stopping at ventral nerve cord (VNC) intermediate and medial zones, respectively (Fig. 3B) (Zlatic et al., 2003). Similar failed termination defects occur in both robo and slit (lf) mutants, in which ch and dbd axons overextend into or beyond the midline (Fig. 3B). Thus, the Robo receptor and its Slit ligand form a repulsive cue to trigger termination of ch and dbd axon growth in the medial-lateral direction.

Studies on sensory neurons in general suggest that Robo3, PlexA and PlexB could affect termination of mechanosensory axons in the VNC (Zlatic et al., 2009). Overexpression of Robo3 and PlexB drives premature sensory axon termination (including ch, dbd and da neurons) in the medial-lateral axis (Zlatic et al., 2009). ch axons in PlexB mutants have prominent axon terminal morphology defects, which could reflect defects in synapse formation (Wu et al., 2011). Nonetheless, results suggest that PlexB, Sema2b and Sema2a potentially affect termination as well. In Sema2b mutants, many ch axons terminate laterally compared with wild type, which is consistent with premature termination. In contrast, Sema2a mutants display failed termination medially. Finally, PlexB mutants and Sema2a; 2b double mutants display abnormal lateral and medial termination events (Fig. 3B). Thus, PlexB, Sema2b and Sema2a play a complex role in ch termination, providing both inhibitory and positive regulation. Overall, a combination of Robo and Plexin signaling is likely to be required for ch and dbd termination.

In larval flies, da neurons have tightly tiled axon termination sites making studies on termination challenging. Nonetheless, there has been valuable progress principally with class IV da (cIV) nociceptive neurons, which we highlight here.

Genetic loss- and gain-of-function approaches showed that PlexB and PlexA are required for cIV axon termination along the dorsal-ventral axis with PlexB having a more prominent role (Fig. 3B) (Zlatic et al., 2009). cIV termination is also defective in mutants for the Plex ligands Sema1a and Sema2a.

robo (robo1), robo2 and robo3 mutants have sensory neuron axon terminal patterning defects (Grueber et al., 2007). However, further experiments with single-cell labeling to visualize individual axon termination sites will help define how different Robo isoforms affect termination. Overexpressing Robo3 in all da neurons or specifically in cIV neurons causes cIV axon undergrowth and absence of midline crossing (Galindo et al., 2023; Grueber et al., 2007). In contrast, Robo3 overexpression in cIV neurons results in failed termination and overgrowth of cIII axons (Galindo et al., 2023). Thus, engineered Robo3 can promote premature or failed termination depending upon neuronal context. This dual function is reminiscent of mammalian Robo3, which has alternative splice forms with attractive and repulsive activities (Chen et al., 2008). Whether Robo3 splicing affects axon termination in flies remains uncertain. Ablation of Down-and-Back (DnB) postsynaptic neurons does not impair cIII or cIV termination, indicating that termination could be distinct from synapse formation (Galindo et al., 2023), similar to C. elegans (Borgen et al., 2017; Zhang et al., 2013). Genetic ablation of cIV neurons results in failed termination of cIII axons, but the opposite does not occur. Competitive presynaptic formation between cIII and cIV neurons results in ectopic synapse formation when cIII axons overgrow. Thus, different classes of mechanosensory neurons influence one another during termination.

The Down syndrome cell-adhesion molecule (Dscam; also known as Dscam1) is a third cue involved in cIV termination (Kim et al., 2013). Dscam mutants have undergrown cIV axon terminals suggesting that premature termination occurs, whereas Dscam overexpression causes failed termination with overgrowth in the longitudinal and lateral directions (Fig. 3C). Similar failed termination events occur when Highwire/RPM-1 is impaired or with overexpression of its substrate Wallenda/DLK (Fig. 3C) (Kim et al., 2013; Wang et al., 2013b). Interestingly, the Highwire/DLK pathway promotes termination via two mechanisms: by restricting Dscam RNA stability or expression (Kim et al., 2013), and via the Kay/Fos transcription factor (Wang et al., 2013b) (Fig. 3D). Whether these are independent signaling mechanisms or converge on Dscam remains unclear.

Dscam levels are further regulated by Amyloid precursor protein-like (Appl). Ectopic overexpression of Appl causes overgrowth, and opposing premature termination occurs in Appl (lf) mutants (Fig. 3C) (Pizzano et al., 2023). Thus, it is possible Dscam and Appl might bind and act as co-receptors, or Appl might regulate Dscam processing.

Adult flies contain numerous mechanosensory neurons including: (1) external sensilla that innervate bristles on the wings and legs; (2) chordotonal organs; (3) campaniform sensilla in the cuticle; (4) scutellar (Sc) and dorsal central (Dc) neurons on thorax; and (5) md nociceptors that innervate the abdomen. Touch neurons display precise termination with different neuron classes terminating in distinct laminar zones in the VNC (Fig. 4). For example, mechanosensory neurons on leg bristles and wing campaniform sensilla have anatomically precise termination with different types of somatosensory axons, including mechanosensory axons, terminating at laminar neuropils that mirror body position (Chen et al., 2006b; Dickinson and Palka, 1987; Mamiya et al., 2018; Murphey et al., 1989; Tsubouchi et al., 2017). Thus, mechanosensory neurons in adult Drosophila and mammals both display laminar central termination patterns. Precise central termination also occurs in locust, moth and honeybee (Ai et al., 2007; Grueber et al., 2001; Pfluger et al., 1994).

The most extensively studied adult touch-sensing neurons are on the thorax, which display precise central termination sites (Murphey et al., 1999; Shepherd and Smith, 1996). Elegant subsequent studies identified two players that regulate termination of Sc and Dc neurons: Dscam (Dscam1) and the PTP69D receptor tyrosine phosphatase (Chen et al., 2006b; Dascenco et al., 2015). At single-cell resolution, Sc and Dc axons display severe premature termination in Dscam mutants, consistent with Dscam mutant GCs having reduced size (Chen et al., 2006b; He et al., 2014). Core Dscam function is required for outgrowth to the primary termination zone, whereas alternative splicing of Dscam regulates branching and growth to secondary termination sites (Fig. 4). These findings suggest that Dscam has dual roles in axon termination and synapse formation with few synapses forming at the primary termination site and many synapses forming on collateral axon branches at secondary termination sites (Urwyler et al., 2015). Hypomorphic mutants and phosphatase substrate trap Ptp69D mutants phenocopy premature termination observed in Dscam mutants affecting both primary and secondary termination sites (Dascenco et al., 2015). PTP96D functions with Slt to regulate Dscam phosphorylation and termination at secondary sites but not primary termination sites. To date, cues that positively regulate termination (i.e. mutants with failed termination defects) have not been identified for any class of adult mechanosensory neuron.

In fish, there are three types of mechanosensory neurons that use glutamatergic transmission (Carmean and Ribera, 2010; Pietri et al., 2009; Wang et al., 2013a). Trigeminal neurons sense touch on the head, while Rohon–Beard neurons and DRG neurons sense touch to the body in larva and adults, respectively (Fig. 5). Single-cell studies indicate that Rohon–Beard neurons (Bernhardt et al., 1990; Kuwada et al., 1990; Liu and Halloran, 2005; Metcalfe et al., 1990) and trigeminal neurons (Metcalfe et al., 1990; Pan et al., 2012) have single pseudo-unipolar axons, similar to mammalian DRGs, that bifurcate and form both central and peripheral termination sites. For Rohon–Beard neurons, most axons terminate in the hindbrain away from Mauthner dendrites or within the spinal cord (SC), and a small number terminate near Mauthner neurons (Fig. 5A) (Palanca et al., 2013). Peripheral axons have variable, complex arborization patterns and terminate in tiled, non-overlapping locations between the epithelial layers of skin (O'Brien et al., 2012). Although little is known about central termination for adult DRG neurons, peripheral axons undergo anatomical remodeling during development and display complex arborization patterns in the skin similar to Rohon–Beard axons (Fig. 5B) (Rasmussen et al., 2018). Interestingly, skin is required for proper outgrowth of DRG axons and is likely to affect peripheral termination, but the molecular players involved remain unknown.

One mechanism affecting peripheral termination is axon–axon repulsion (Fig. 5A) (Grueber and Sagasti, 2010). Time-lapse imaging has shown that GCs from different peripheral axons of trigeminal and Rohon–Beard neurons actively repulse one another during growth (Liu and Halloran, 2005; Sagasti et al., 2005). Transplanting trigeminal or Rohon–Beard neurons into mutants lacking somatosensory neurons demonstrates that axons grow expansively in the absence of tiling cues from neighboring axons (Sagasti et al., 2005). Findings from mechanically denervated infant rats and from leech suggest that axon–axon repulsion is likely to be a conserved principle of peripheral termination (Jackson and Diamond, 1984; Kramer and Stent, 1985).

Sema3D regulates peripheral termination in Rohon–Beard neurons (Fig. 5A) (Liu and Halloran, 2005). Morpholino knockdown of Sema3D causes premature termination of peripheral Rohon–Beard axons with reduced outgrowth. Ectopic expression of Sema3D can also trigger peripheral axon GCs to undergo an extended retraction period for hours, but has negligible effects on central axon GCs. Thus, bifurcated pseudo-unipolar axons are differentially sensitive to termination cues. Interestingly, the extended retraction of peripheral axon GCs caused by Sema3D is reminiscent of the protracted GC collapse that occurs during termination in C. elegans (Borgen et al., 2017). These findings demonstrate that Sema3D acts as a repulsive cue that inhibits peripheral termination of Rohon–Beard axons. Identifying further molecular cues and signaling mechanisms that regulate peripheral termination remains a task for the future. At present, the regulatory players that govern central termination in fish remain unknown (Fig. 5B).

Axon termination has been studied in two major types of mechanosensory neurons in mammals: LTMRs that sense gentle mechanical stimuli (Jenkins and Lumpkin, 2017; Meltzer et al., 2021), and nociceptors that sense harsh touch (Basbaum et al., 2009; Dubin and Patapoutian, 2010). LTMRs and nociceptors are located in DRG or trigeminal ganglia and form pseudo-unipolar axons that bifurcate to yield morphologically complex termination sites in the periphery and laminated termination patterns in the SC (Fig. 6A,B). Mammalian mechanosensory neurons are principally glutamatergic (Zhang et al., 2018). Interestingly, multi-functional, bifurcated axons and glutamatergic transmission are ancient evolutionary principles at work in mechanosensory neurons from C. elegans to mammals. Our discussion principally focuses on LTMRs in mice, but we also touch on studies that examine somatosensory neurons more generally. We highlight regulatory cues that might generally affect LTMRs and cues that are specific for subsets of LTMRs.

Studies in mice have dramatically increased our understanding of the anatomy and genetic classification of LTMRs (Olson et al., 2016). Historically, LTMRs are categorized by physiology and include Aβ rapid-acting (RA)-LTMRs, Aβ slow-acting (SA)-LTMRs, Aδ-LTMRs, C-LTMRs and C-tactile fibers. As we discuss below, genetic labels for individual LTMR classes have dramatically increased our understanding of LTMR anatomy, including termination sites. Single-cell sequencing has identified even more somatosensory neuron classes in DRG (Haring et al., 2018; Usoskin et al., 2015), which suggests that peripheral and central termination sites could be even more complex than what we describe here.

Studies with cats first showed that mechanosensory neurons have precise termination sites in the SC (Brown et al., 1977). Subsequent histological and physiological work in rodents, cats and monkeys showed that different mechanosensory classes terminate growth in different parts of the dorsal column (Fig. 6) (Brown et al., 1980; Li et al., 2011; Light and Perl, 1979; Liu et al., 2007; Sugiura et al., 1986; Woodbury et al., 2001). Elegant work with genetic reporters confirmed that different classes of LTMRs have precise, laminated central termination in mice, but also display partial overlap in termination layers (Fig. 6B) (Olson et al., 2016; Zylka et al., 2005). Moreover, individual neurons of the same class that innervate adjacent areas of skin terminate in adjacent or overlapping patterns in the SC. The degree of overlap varies depending on LTMR type with Aβ-RA axons showing the highest overlap (Kuehn et al., 2019). Trigeminal LTMRs also display precise termination in SC lamina (Durham and Woolsey, 1984; Jacquin et al., 1984, 1986; Ma, 1991; Shigenaga et al., 1990).

Mammalian peripheral sensory axon endings show variable terminal anatomy allowing for functional specification (Fig. 6B) (Cauna, 1973; Kruger et al., 1981; Munger and Ide, 1988; Olson et al., 2016; Pare et al., 2002). For LTMRs, imaging studies using genetic cell-specific labels and electron microscopy demonstrate that complex morphologies correspond with touch sensitivity (Bai et al., 2015; Handler et al., 2023; Li et al., 2011; Zylka et al., 2005). Constructing these termination sites is likely to require a similarly complex genetic signaling program. Findings indicate that Aβ RA-LTMRs form hair-innervating lanceolate endings, as well as Meissner's and Pacinian corpuscles (Fig. 6B, right). Aβ SA-LTMRs form Merkel and Ruffini endings. Aδ-LTMRs form palisades of lanceolate endings around hair follicles. C-fiber mechanoreceptors can be divided into two sub-populations: C-LTMRs with lanceolate endings surrounding hair follicles, and C-tactile fibers that terminate with free nerve endings (Fig. 6B, right) (Chalfie, 2009; Zimmerman et al., 2014).

Owing to space limitations, we primarily discuss regulatory players that affect termination of LTMR axons from DRG. However, we do not assume that termination programs for DRG and trigeminal neurons are the same. Given the number of DRGs, it is plausible that LTMRs innervating different portions of SC might be regulated by different cues.

Secreted semaphorins (Sema) and their receptor neuropilin 1 (Npn1/Nrp1) affect central termination for a subset of sensory axons in the SC, which likely include LTMRs (Fig. 6B,C) (Gu et al., 2003; Messersmith et al., 1995). A clever genetic strategy whereby the Sema-binding motif was eliminated from Npn1 has demonstrated that failed central termination and overgrowth occur in these animals. These findings, coupled with results from Drosophila, indicate that semaphorins play an evolutionarily conserved role in central termination.

Studies with Aβ-RA axons have shown that the Ret growth factor receptor influences central axon outgrowth (Honma et al., 2010; Luo et al., 2009). Cell-specific removal of Ret in Aβ-RAs reduces outgrowth with loss of tertiary axons. Conversely, ectopic expression of the Ret ligand neurturin (Nrtn) causes failed termination (Honma et al., 2010), and Nrtn mutants display premature termination (Fleming et al., 2015). The complexity of ligands for Ret in central termination is substantial with premature termination and undergrowth also occurring in Gfra2/Grfa1 double mutants, which phenocopy Ret mutants (Fleming et al., 2015). Interestingly, Gfra1/2 function via trans and cis signaling to influence central projection growth. Thus, an Ntrn/Gfra1/Gfra2-Ret growth factor receptor system is required for Aβ-RA central termination (Fig. 6B,C).

Thus, Sema/Npn1 signaling is broadly implicated in general sensory axon termination, whereas Ntrn/Grfa1/Grfa2-Ret activity affects central termination of Aβ-RA axons. Whether Npn1, Ret and their ligands form a combination of cues with effects on Aβ-RAs or more broadly on LTRMs awaits further study. With genetic tools for LTMR subtypes established, the field is well positioned to understand how these cues affect central termination of individual LTMR classes.

Several transcription factors potentially influence central termination. Runx1, a runt domain transcription factor, regulates central termination of C-tactile nociceptors (Chen et al., 2006c). In Runx1 mutant mice, premature termination of C-fibers occurs in layer I of the SC rather than layer II. More general studies suggest that Runx1 might be required for proper termination of other LTMRs (Chen et al., 2006a). Isl1, Pou4f1 (Brn3a) and Pou4f2 (Brn3b) broadly affect mechanosensory termination (Fig. 6B,C). Interestingly, Isl1 knockouts display axon over- and undergrowth suggesting that balanced Isl1 activity mediates accurate termination (Sun et al., 2008). Pou4f1 and Pou4f2 work coordinately to promote growth and prevent termination (Zou et al., 2012). Central termination of Aβ-RA axons is affected by the basic leucine-zipper transcription factor Maf, and the homeobox transcription factor Shox (Hu et al., 2012; Scott et al., 2011). In Maf and Shox (lf) mutants, a decrease in glutamatergic terminals occurs in layer III/IV of the dorsal horn where termination normally occurs for Aβ-RA axons, which suggests that premature termination might occur. However, this could be due to fewer DRG neurons resulting in fewer axons in these mutants. Although several transcription factors potentially influence central termination, we still lack a clear understanding of how a putative transcription factor code contributes to laminated central termination patterns.

Precise peripheral termination is also required for proper development and function of mammalian mechanosensory neurons (Wang et al., 2013a). To date, most studies have focused generally on somatosensory neurons. Thus, the cues discussed below could influence peripheral termination of LTMRs and nociceptors, but this often remains unclear.

Secreted Sema3A influences peripheral termination via the Npn1 receptor, and a second set of PlexA and PlexB co-receptors affect peripheral termination with Npn1 and Npn2 (Nrp2) (Fig. 6B,C) (Gu et al., 2003; Kitsukawa et al., 1997; Taniguchi et al., 1997; Yaron et al., 2005). Excess axon branching occurs in Sema3A, Sema3F, Npn1, Npn2, PlexA3 and PlexA4 knockout animals, which is accompanied by substantial failed termination and overgrowth. Consistent with a role in termination, Sema3A-induced GC collapse is mediated by Npn1 in cultured DRG neurons (Kitsukawa et al., 1997). Thus, a putative model suggests that peripheral termination is regulated by a complex multi-receptor signaling network composed of Sema3A/Npn1, Sema3A/Npn1/PlexA4 (Plxna4) and Sema3F/Npn2/PlexA3 (Plxna3). Whether Sema/Npn/Plex signaling axes mediate the axon–axon repulsion mechanism required for tiling remains unknown, and we do not know how different Sema/Npn/Plex signaling axes affect individual LTMR classes.

A recent study examining Aβ-RA terminals has shown that knockout of netrin G1 (a GPI-anchored protein) or its ligand (the adhesion protein Lrrc4c) produces mixed outcomes on Aβ-RA axons with reduced terminal numbers of increased size (Meltzer et al., 2022). This indicates that balanced netrin G1/Lrrc4c signaling is required for termination (Fig. 6B,C).

Transgenic overexpression of growth factors in the skin, such as bone-derived neurotrophic factor (BDNF) and neurotrophin 4 (NT4; also known as Ntf5), generates failed termination at Aβ Meissner's corpuscles without affecting sensory neuron numbers (Fig. 6B,C) (Krimm et al., 2006; LeMaster et al., 1999). Furthermore, mice lacking BDNF or its receptor, TrkB (Ntrk2), show reduced Meissner's corpuscles (Gonzalez-Martinez et al., 2005, 2004; Perez-Pinera et al., 2008). BDNF is expressed near termination sites in the skin and eliminating BDNF production by epithelial cells reduces the complexity of Aδ-LTMR termination sites (Rutlin et al., 2014). Knockout studies with other growth factors and their receptors, including NT3 (Ntf3), NGF, TrkA (Ntrk1), TrkC (Ntrk3) and p75 (Ngfr), suggested potential roles in peripheral axon outgrowth (Airaksinen et al., 1996; Fundin et al., 1997). Ectopic overexpression of nerve growth factor (NGF) in the skin drives failed termination with hyper-innervation (Albers et al., 1994). For NT3, heterozygous knockouts show reduced axon terminals at Merkel endings and those present display prominent premature termination (Airaksinen et al., 1996). Nonetheless, a major caveat is that impairing these growth factors can severely reduce LTMR numbers, which complicates interpretations relating to axon growth and termination (Crowley et al., 1994; Farinas et al., 1994).

Interestingly, Ret affects peripheral termination of Aβ-RAs without affecting RA neuron numbers (Honma et al., 2010; Luo et al., 2009). In Ret knockout mice, premature termination and undergrowth are observed for Aβ termination sites at Meissner corpuscles, Pacinian corpuscles and lanceolate endings (Fig. 6B,C). The Ret ligands Nrtn and Gfra2 are also implicated in peripheral termination (Luo et al., 2009). Thus, numerous growth factors could influence termination with Ret, BDNF and NT3 being most clearly involved.

Two transcription factors, Isl1 and Pou4f1, broadly regulate outgrowth of peripheral mechanosensory axons (Fig. 6B,C). Conditional knockouts for Isl1 that circumvent lethality show substantially undergrown peripheral axons (Sun et al., 2008). Loss of Pou4f1 causes extensive overgrowth of sensory axons and concomitant premature termination, which suggests that balanced Pou1f1 activity facilitates proper growth and termination (Eng et al., 2001). Interestingly, Isl1 and Pou4f1 function together with severely reduced axon outgrowth occurring in double mutants (Dykes et al., 2011).

Runx1 and Runx3 affect termination by C-LTMRs and proprioceptive neurons, respectively (Lallemend et al., 2012; Lou et al., 2013). In both cases, knockouts display undergrown termination sites. Runx3 is necessary and sufficient for axonal growth in vitro with pharmacological inhibitors of Rock kinase rescuing growth defects in Runx3 knockout neurons (Lallemend et al., 2012).

Finally, Maf affects peripheral terminals of Aβ-RAs, which have undergrown Messner endings in knockouts (Fig. 6B,C) (Wende et al., 2012). Although fewer Pacinian corpuscles are also observed, this might reflect a reduction in axon number rather than an effect on growth. In sum, peripheral termination is regulated by a complex milieu of repulsive cues and growth factors that are likely to influence multiple transcription factors.

We currently know little about how intracellular signaling shapes central and peripheral termination in mammals beyond a handful of transcription factors. One player implicated in termination of sensory neurons is Phr1 (Mycbp2). As noted earlier, mounting evidence in C. elegans and Drosophila indicate Phr1 (RPM-1 or Highwire, respectively) is a conserved master regulator of axon termination in mechanosensory neurons.

Regarding peripheral termination, Phr1 (lf) mutant mice display heavy overgrowth of cutaneous sensory axons consistent with failed termination (Fig. 6B,C) (Lewcock et al., 2007). Conversely, peripheral axon growth is reduced in motor neurons, indicating that neuronal context influences Phr1 effects on outgrowth and termination. Mechanistically, Phr1 affects sensory GC dynamics via effects on microtubules. Excess microtubules in Phr1 mutant GCs likely contributes to failed termination, and suggests that Phr1 is a microtubule destabilizer in sensory neurons. Interestingly, Phr1 mutant GCs remain sensitive to Sema3A and NGF in vitro suggesting that other cues might regulate Phr1 during peripheral termination. Furthermore, Phr1 stabilizes the EphB2 receptor to influence axon development ex vivo (Chang et al., 2024). Whether Phr1 and EphB2 influence termination of LTMRs in vivo remains to be tested.

Phr1 also functions in sensory neurons to regulate central termination (Holland et al., 2011). Conditional knockout of Phr1 in nociceptive neurons causes failed termination and overgrowth of sensory axons beyond layer II in the SC. This suggests that Phr1 likely regulates termination in C-LTMRs. Future experiments with more-specific genetic tools will be needed to determine whether Phr1/MYCBP2 affects C-LTMRs, and influences central and peripheral termination for other classes of LTMRs.

From our cross-species survey, several key concepts governing axon growth and termination emerge. (1) Multiple cell surface cues are required to orchestrate termination with semaphorin/plexin and netrins playing roles across flies, fish and mice. In mammals, growth factors expand the cadre of cell surface receptors involved in growth and termination. (2) Adhesion molecules, such as Dscam, also mediate cellular interactions to promote central termination. (3) Laminar termination of different classes of mechanosensory neurons is an evolutionarily conserved feature of central termination from invertebrates to mammals. (4) Multiple transcription factors influence peripheral and central termination. (5) Finally, complex intracellular signaling networks containing multiple ubiquitin ligases and kinases affect termination by influencing GC development, proteostasis and microtubule stability. Despite many conserved principles, different molecular players have also emerged from different models. This could reflect unbiased versus gene-specific approaches, the molecular environment of neurons under study, or the class or morphological complexity of a given LTMR.

Studies to date indicate that axon termination is developmentally distinct from axon guidance. This is based on developmental timing, effects on axon length (termination) versus trajectory/direction (guidance), and evidence that termination is mediated by a protracted GC collapse process whereas guidance involves more dynamic, transient growth cone alterations. It remains less clear how termination is distinct from or coordinated with axon branching. Building a comprehensive mechanistic understanding of how receptors, intracellular signaling networks and transcriptional changes influence axon termination is likely to reveal points of molecular distinction and integration between termination and other neurodevelopmental events.

We appreciate input from Dr Wesley Grueber.