The development of synapses at DD GABAergic spines

We next investigated the relative timing of developmental events during the formation of synapses onto spines. C. elegans progress through four larval stages of development (L1-L4) prior to adulthood. DD neurons undergo a well-characterized program of synaptic remodeling such that the mature circuit organization is established during the transition from L1 to L2 stage [34,35]. The formation of new synaptic connections between postembryonic born cholinergic neurons and the ventral dendrites of DD GABAergic neurons offers a window to investigate de novo formation of neuron-neuron synapses. We focused on understanding the time course of spine formation relative to 2 key events in synaptogenesis: 1) the formation of presynaptic release sites, and 2) the clustering of postsynaptic receptors and F-actin.

The appearance of newly born ventral cholinergic motor neurons progressed anteriorly to posteriorly starting roughly 16 hours after hatch, consistent with the birth and integration of cholinergic neurons during this time frame (S4 Fig) [36]. We analyzed the distribution of pre- and postsynaptic markers at timepoints before (12 hours after hatch), during (16, 18, 20 hours after hatch) and after (24, 32 hours after hatch, L4 stage/42-50 hours) DD synaptic remodeling (Figs 2, S4 and S5). We first assessed the formation of the active zone and presynaptic specialization by analyzing clustering of the Piccolo-like active zone protein, CLA-1, (GFP::CLA-1e) [37] and the synaptic vesicle-associated protein synaptobrevin/SNB-1 (SNB-1::GFP) (Figs 2A, 2D–2E and S5). Overall, we found CLA-1 is more discretely localized to putative synapses than SNB-1, both at earlier timepoints and throughout development, consistent with its specific localization to active zones. Immediately prior to remodeling (~12 hours after hatch), we did not observe significant localization of either presynaptic CLA-1 or SNB-1 adjacent to the DD process. Shortly after (16 hours after hatch), we noted the initial appearance of SNB-1- and CLA-1-associated fluorescence in presynaptic cholinergic processes. Over the next 4 hours (20 hours after hatch) individual CLA-1 clusters were more clearly distinguishable, suggesting CLA-1 association with developing active zone structures (Figs 2A, 2D, 2L and S5A). Synaptic vesicle fluorescence was initially diffuse at 16 hours after hatch and became more clearly organized into discrete puncta over a similar time course to CLA-1 (Figs 2A, 2E, 2L and S5B). We did not observe the emergence of DD dendritic spines until 24–32 hours after hatch, well after initial active zone formation and recruitment of synaptic vesicles (Figs 2A, 2C–2E and S5A and S5B). Spines continued to mature through L4 stage, increasing in both length and number (S4D–S4F Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Temporal order of molecular events during synapse and spine development.

(A) Fluorescent images of GABAergic dendrites at 12, 16, 20, 24, and 32 hours after hatch and at L4 stage (~42–50 hours post-hatch). Animals express a GABAergic dendrite marker (Pflp-13::mCherry) with either presynaptic active zone (Punc-17β::GFPnovo2::CLA-1e) (top) or synaptic vesicle marker (Pacr-5::SNB-1::GFP) (bottom). Images are pseudo colored to indicate spines (magenta) and presynaptic CLA-1 or vesicle clusters (cyan). (B) Fluorescent images of GABAergic dendrites at 12, 16, 20, 24, and 32 hours after hatch and L4 animals (~42–50 hours post-hatch). Animals express a GABAergic dendrite marker (Pflp-13::mCherry) with either postsynaptic AChR (Punc-47::ACR-12::GFP, top) or postsynaptic actin (Pflp-13::GFP::UtrCH, bottom) markers. Images are pseudo colored to indicate spines (magenta) and postsynaptic AChR or F-actin (cyan). White asterisks (*) indicate images used in Fig 2C. (C) Fluorescent images (inverted LUT) of GABAergic dendrites at 20 hours after hatch and L4 stage (~42–50 hours post-hatch) as indicated. Animals express a GABAergic dendrite marker (Pflp-13::mCherry). Dashed red boxes indicate regions used in Fig 2B (asterisks). (D-G) Quantification of spine number (purple) relative to clusters of active zone protein/CLA-1 (D), synaptic vesicle/SNB-1 (E), AChR/ACR-12 (F), and F-actin/UtrCH (G) (black) through development. Clusters of each of these markers are visible well prior to spine outgrowth. Blue shading indicates the approximate period of DD remodeling. For this and subsequent figures, numbering indicates animals quantified at each timepoint for each genotype. Data points and bars indicate mean ± SEM. (H) Fluorescent confocal images of DD1 AChR/ACR-12 (Pflp-13::ACR-12::GFP, magenta) and F-actin clusters (Pflp-13::mCherry::UtrCH, cyan) 24 hours after hatch. Dotted rectangle indicates high magnification image shown in 2I. (I) Expanded view of DD1 dendrites 24 hours after hatch from Fig 2H. (J) Fluorescent confocal images of DD1 AChR/ACR-12 (Pflp-13::ACR-12::GFP, magenta) and F-actin clusters (Pflp-13::mCherry::UtrCH, cyan) at L4 stage. Dotted rectangle indicates high magnification image shown in 2K. (K) Expanded view of DD1 dendrites 24 hours after hatch from Fig 2J. (L) Cartoon summary of the order of events during development of synapses onto DD neurons. In early development, presynaptic vesicles (SNB-1) and active zone protein (CLA-1) localize along presynaptic terminals while postsynaptic F-actin and postsynaptic receptors cluster along the postsynaptic process. Following their formation, dendritic spines begin to extend. Spines continue to elongate throughout development and into the mature circuit (L4).

https://doi.org/10.1371/journal.pgen.1010016.g002

We next examined the clustering of postsynaptic acetylcholine receptors in DD dendrites (ACR-12::GFP). Prior to remodeling (12 hours after hatch), we did not observe detectable levels of cholinergic receptors in ventral DD processes. Surprisingly, we noted immature receptor clusters faintly visible in the dendritic shaft by 16 hours after hatch, well prior to the emergence of dendritic spines (Figs 2B, 2F, 2L and S5C). These receptor clusters increased in number and redistributed towards the tips of growing spines by 32 hours (Figs 2B, 2F, 2L and S5C). We found a similar developmental trend using DD neuron-specific labeling of the LEV-10 transmembrane auxiliary protein [38], previously shown to concentrate in spines [39] (S6A and S6B Fig). By L4 stage, receptor and LEV-10 clusters are clearly visible at the tips of mature spines. Our analysis indicates that the initial stages of development of postsynaptic (AChR and LEV-10 clusters) structures occur prior to spine outgrowth, raising questions about how these initial processes may be regulated.

To begin to address this question, we analyzed the distribution of F-actin in dendrites of the DD neurons during synapse formation (Figs 2B, 2G–2L and S5D). We noted that clusters of F-actin were evident in the ventral DD processes prior to the completion of synaptic remodeling (12 hours after hatch), prior to presynaptic CLA-1 and synaptic vesicle accumulation, and before the clustering of postsynaptic receptors. Dendritic F-actin-based structures became more abundant coincident with increases in the number of presynaptic CLA-1 and postsynaptic receptor clusters. To investigate this process in real-time, we used live imaging to examine the dynamics of postsynaptic F-actin (GFP::UtrCH) in the developing DD dendrite. We found that F-actin is highly dynamic during early developmental stages (16–20 hours after hatch) compared to L4 stage where the circuit has completed maturation (S6C–S6H Fig and S1 and S2 Videos). In young animals, GFP::UtrCH clusters often shuttled out of the cell body to the main dendritic process, perhaps indicating redistribution of F-actin to sites of postsynaptic assembly. By 24 hours after hatch, we observed clear co-localization of F-actin with newly formed AChR clusters in the dendritic shaft. By L4 stage, AChR clusters are stably sequestered at the tips of spines, while F-actin occupies the spine head and neck (Fig 2H–2K).

Dendritic spine formation does not strongly require synaptic activity

Our above analysis indicated that the localization of synaptic vesicles and active zone proteins, such as CLA-1, occur prior to spine outgrowth. In the rodent brain, spine morphogenesis is clearly regulated by presynaptic activity [40–43], but recent evidence suggests initial spine outgrowth may proceed independently of synaptic activity [44–46]. We therefore next asked whether presynaptic cholinergic activity is important for the formation of DD dendritic spines. To address this question, we analyzed the number of dendritic spines (Pflp-13::mCherry or Pflp-13::myrGFP) and cholinergic receptors (Pflp-13::ACR-12::GFP) at L4 stage in strains carrying mutations that affect various aspects of synaptic function and neuronal excitability (Table 1 and S7 Fig). Surprisingly, the abundance of spines and receptor clusters at L4 stage were not significantly affected by disruption of ACh synthesis in cholinergic motor neurons (mutation of the vesicular acetylcholine transporter unc-17) or by a strong reduction in synaptic vesicle exocytosis (mutation of the syntaxin binding protein unc-18 or the fusion regulator unc-13). Similarly, loss-of-function mutations in genes required for dense core vesicle release (CAPS/unc-31) and Ca²⁺ signaling (VGCC/unc-2) did not significantly reduce the number of spines or receptor clusters in the mature circuit (Table 1 and S7 Fig). Our data indicate that severe alterations in synaptic activity have very limited effects on the abundance of L4 stage dendritic spines. However, we cannot exclude the possibility of developmental delays in spine formation [23]. Notably, many of the mutations we examined significantly altered spine length and the size of receptor clusters (Table 1), suggesting that activity may influence other aspects of spine morphology and postsynaptic structure. Together, our findings point toward a model where spine formation and postsynaptic development proceed without a strong requirement for synaptic activity, while subsequent spine morphogenesis is more clearly affected.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Analysis of dendritic spines and cholinergic receptors in mutant strains with altered synaptic activity.

Dendritic spine number, spine length, ACR-12/AChR cluster number, and AChR cluster size at L4 stage normalized to the same measurements from wild type are shown (mean ± SEM) for each genotype (ACR-2/nAChR UNC-2/CaV2α, UNC-13/MUNC-13, UNC-17/VAChT, UNC-18/MUNC-18, UNC-31/CAPS). Mutations that affected synaptic activity had little effect on the density of spines or AChR clusters, but had variable effects on spine length and AChR cluster size. One-way AVOVA, Dunnett’s multiple comparisons test, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

https://doi.org/10.1371/journal.pgen.1010016.t001

Presynaptic NRX-1 stabilizes postsynaptic components

In previous work, we found that mature spines are severely reduced in L4 stage nrx-1(wy778) mutants [20]. We found a similarly severe reduction in nrx-1(wy1155) null mutants where the entire nrx-1 coding region is deleted [47], indicating that both alleles are nulls and highlighting the potential importance of activity-independent mechanisms for spine formation (S7D and S7E Fig). Mutation of nrx-1 also disrupted dendritic calcium responses evoked by presynaptic stimulation of cholinergic motor neurons (S8 Fig). To assess involvement of presynaptic NRX-1 in the establishment of spines, we asked when NRX-1 first localizes to presynaptic terminals of cholinergic motor neurons. Axonal clusters of NRX-1::GFP are present by 16 hours after hatch (Fig 3A and 3B), approximately the same time at which assemblies of the presynaptic scaffold CLA-1 become visible (Fig 2). Diffuse synaptic vesicle material is also visible at this time (Fig 2) but has yet to be organized into discrete clusters. Notably, presynaptic NRX-1::GFP clusters and spines are similar in number by L4 stage, while presynaptic CLA-1 and SNB-1 clusters appear more abundant than spines (Fig 2D and 2E). This may suggest that a subset of presynaptic release sites lacking NRX-1 are not associated with dendritic spines; however some of these presynaptic puncta may represent trafficking vesicles rather than mature release sites, complicating interpretation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Dendritic spines and postsynaptic components form initially but subsequently collapse without the synaptic adhesion protein, NRX-1.

(A) Fluorescent confocal images of dendritic spines (Pflp-13::mCherry, magenta) and cholinergic expression of NRX-1::GFP (Punc-17β::NRX-1::GFP, cyan) at 16, 20, and 24 hours after hatch and L4 stage (~42–50 hours post-hatch). Presynaptic clusters of NRX-1::GFP are visible by 16 hours after hatch and increase in size until roughly 24 hours after hatch. (B) Quantification of the average number of spines (purple) and NRX-1 clusters (black) at the time points indicated. Blue shading indicates the approximate period of DD remodeling. Numbers indicate the number of animals quantified at each timepoint. Data points indicate mean ± SEM. (C) Fluorescent images (inverted LUT) of DD dendritic spines (Pflp-13::myrGFP) in wild type (left) and nrx-1(wy778) (right) animals at 16, 18, 20, and 24 hours after hatch and at L4 stage. Immature dendritic spines clusters form initially but then collapse in nrx-1 mutants. (D) Quantification of the percentage of wild type (black) and nrx-1(wy778) (purple) animals with 2 or more DD dendritic spines at the timepoints indicated. YA, young adult. Numbers indicate the number of animals quantified at each timepoint. (E) Fluorescent images of ACR-12/AChR clusters (Pflp-13::ACR-12::GFP) in DD dendrites of wild type (left) and nrx-1(wy778) mutants (right) 16, 20, and 24 hours after hatch and at L4 stage (~42–50 hours post-hatch). Immature AChR clusters initially form but then disperse in nrx-1 mutants. (F) Quantification of the number of ACR-12::GFP clusters in DD dendrites of wild type (black) and nrx-1(wy778) mutants (purple). Two-way ANOVA, Sidak’s multiple comparisons, *p<0.05, **p<0.01, ***p<0.001. (G) Fluorescent confocal images of dendritic spines (Pflp-13::mCherry) and F-actin (Pflp-13::GFP::UtrCH) in DD dendrites of wild type and nrx-1(wy778) animals at 16 hours after hatch and L4 stage.

https://doi.org/10.1371/journal.pgen.1010016.g003

The early arrival of NRX-1 to synaptic terminals could indicate a role in either initial synapse formation or subsequent stabilization and maturation. To distinguish between these possibilities, we analyzed the development of dendritic spines and cholinergic receptor clusters in nrx-1(wy778) null mutants. We quantified spine and AChR cluster number over a similar time course as previously completed for wild type. Surprisingly, we noted that immature spines and receptor clusters were evident in nrx-1 mutants during early development, albeit at slightly reduced density relative to wild type (Figs 3C–3F and S9). The density of spines in nrx-1 mutants increased significantly over the next several hours until 24 hours after hatch (Figs 3C, 3D and S9A). However, after this time, spine density decreased dramatically such that DD dendrites of nrx-1 mutants were almost completely devoid of spines by L4 stage [20]. Measurements of spine formation and disassembly in live imaging studies of wild type and nrx-1 mutants offered further support for this conclusion. Mature wild type spines (L4) were remarkably stable over 1–2 hours of recording, but were more dynamic in the developing circuit (16–20 hrs after hatch) (S9C–S9I Fig and S3–S5 Videos). Developing spines in nrx-1 mutants were also highly dynamic. Notably, nrx-1 mutant spines exhibited heightened dynamics into later stages of development compared with wild type spines. For example, at 21–24 hours after hatch, almost 90% of wild type spines were stable over the recording period, while less than 50% of nrx-1 mutant spines remained stable.

Postsynaptic receptor clusters were also present early in the development of nrx-1 mutants. The number and fluorescence intensity of ACR-12 clusters at 16 hours after hatch in nrx-1(wy778) animals (6.8 ± 0.8 clusters/25 μm) were similar to wild type (5.6 ± 0.7 clusters/25 μm) (Figs 3E, 3F and S9B). Within a few hours however, ACR-12 cluster number and intensity decreased significantly in nrx-1 mutant dendrites, and remained low throughout the remainder of development (Figs 3E, 3F and S9B). Remarkably, the organization of dendritic F-actin was similarly affected by nrx-1 deletion. The distribution of dendritic GFP::UtrCH was comparable across nrx-1 mutants and wild type in early development (16 hrs after hatch), positioned in discrete clusters along the main dendritic process (Fig 3G). By L4 stage, the F-actin marker became almost exclusively associated with wild type dendritic spines, but was diffusely localized along the length of L4 stage nrx-1 mutant dendrites (Fig 3G). Taken together, our results indicate that presynaptic NRX-1 is dispensable for the earliest stages of postsynaptic assembly and spine formation, but is critical for stabilizing dendritic spines, F-actin assemblies and AChR clusters, and promoting their maturation.

Presynaptic neurexin localizes to cholinergic terminals and stabilizes postsynaptic components in a Kinesin-3/UNC-104 dependent manner

To investigate this model further, we sought to determine how presynaptic neurexin is transported to active zones in vivo. Prior work showed that both synaptic vesicles and CLA-1 depend on the Kinesin-3 motor UNC-104 for their delivery to synapses [37,48]. In particular, disruption of unc-104 causes an accumulation of synaptic vesicles within neuronal somas and a corresponding loss of synaptic vesicles within axons [48] (S10A and S10D Fig). unc-104(e1265) carries a D1497N mutation in the PH domain of UNC-104 that impairs cargo binding and leads to a severe reduction in axonal UNC-104 abundance [49]. We found that NRX-1 endogenously tagged with GFP localized to puncta in neuronal processes of the nerve ring and nerve cords of wild type animals (S10E Fig), but was strikingly decreased in the nerve cords of unc-104(e1265) mutants (Figs 4A, 4C and S11A–S11D). A striking deficit in axonal NRX-1::GFP localization was also evident with specific cholinergic expression of NRX-1::GFP in unc-104 mutants (Figs 4B, 4C and S11E–S11H). By contrast, NRX-1::GFP fluorescence was significantly increased (5.4-fold) in unc-104 mutant cholinergic somas compared to wild type (S11C, S11D, S11G, S11H Fig). Consistent with UNC-104 involvement in cholinergic NRX-1 trafficking, we found that UNC-104 and NRX-1 are partially colocalized in cholinergic axons (S11I and S11J Fig). Accumulation of NRX-1 in cholinergic cell bodies, coupled with a decrease in axons, indicates a failure of NRX-1 transport in the absence of functional UNC-104. Cholinergic-specific expression of wild type unc-104 in unc-104 mutants restored axonal NRX-1::GFP localization, indicating a cell autonomous requirement (Fig 4B and 4C). In contrast, mutation of the Kinesin-1 motor unc-116 did not produce significant accumulation of NRX-1::GFP in cholinergic somas and caused comparatively less severe decreases in axonal NRX-1::GFP, demonstrating preferential involvement of Kinesin-3 for NRX-1 transport (S12A–S12D Fig). Live imaging studies offered additional evidence that synaptic vesicle and NRX-1 trafficking may share a common dependence on UNC-104. We found that NRX-1::GFP trafficking events occurred with similar anterograde and retrograde velocities to SNB-1::GFP (labeling synaptic vesicles) trafficking events, though less frequently (S12E–S12I Fig and S6, S7 Videos). Prior studies have noted that active zone protein trafficking events occur at reduced frequency compared with synaptic vesicle trafficking [9]. Together, our analyses demonstrate that NRX-1 localization to synapses is strongly UNC-104-dependent while UNC-116 may contribute more indirectly, perhaps through effects on other synapse-associated proteins [50,51].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Presynaptic NRX-1 localizes to cholinergic terminals and stabilizes spines in an UNC-104/KIF1A-dependent manner.

(A) Representative confocal images of NRX-1::GFP (endogenous nrx-1 knock-in) from dorsal cords (inverted LUT) of wild type and unc-104(e1265) mutants. Images on each line are from different young adult animals (5 shown for each genotype). Mutation of unc-104 impairs axonal NRX-1:GFP localization. (B) Representative confocal images of cholinergic NRX-1::GFP (Punc-129::NRX-1::GFP, inverted LUT) from the dorsal nerve cord of wild type, unc-104(e1265) and unc-104(e1265) mutants rescued by cholinergic expression of wildtype unc-104 (Punc-17β). Images on each line are from different young adult animals (5 shown for each genotype). (C) Scatterplot of the average number of NRX-1::GFP (endogenous or transgenic), CLA-1::GFP, UNC-10::GFP, or ELKS-1::mCherry clusters per 50 μm of dorsal nerve cord in unc-104(e1265) mutants or rescue as indicated. Values are normalized to their respective controls. Red dotted line indicates 50% of wild type values. Bars indicate mean ± SEM. Student’s t-test, ****p<0.0001. n ≥ 11 animals. Raw values with respective controls are shown in S13 Fig. (D) Fluorescent images (inverted LUT) of DD dendritic spines (Pflp-13::myrGFP) from L4 stage wild type, unc-104(e1265), and unc-104(e1265) mutants rescued by cholinergic expression of wild type unc-104. (E) Fluorescent images of AChR/ACR-12 clusters in DD dendrites (Pflp-13::ACR-12::GFP) of L4 stage wild type, unc-104(e1265), and unc-104(e1265) mutants rescued by cholinergic expression of wild type unc-104. (F) Scatterplot showing quantification of DD dendritic spines per 15 μm (Pflp-13::myrGFP) (purple) and DD cholinergic receptor clusters (Pflp-13::ACR-12::GFP) (green) in wild type, unc-104(e1265), and indicated cell-specific rescue lines. Inset, number of rescuing lines/total transgenic lines tested for each rescue construct. Bars, mean ± SEM. One-way ANOVA, Dunnett’s multiple comparisons test, ****p<0.0001. n ≥ 10 animals. (G) Fluorescent images of DD dendritic spines (Pflp-13::myrGFP or Pflp-13::mCherry) from L4 stage wild type, nrx-1(wy778), cla-1(ok937), unc-10(md1117), or elks-1(ok2762) mutants. Only mutation of nrx-1 affects dendritic spines. (H) Representative confocal images of cholinergic CLA-1 (Punc-17β::CLA-1::GFP, inverted LUT) from the dorsal nerve cord of wild type or unc-104(e1265) mutants. Images on each line are from different young adult animals (5 shown for each genotype). (I) Representative confocal images of cholinergic UNC-10 (Punc-129::UNC-10::GFP, inverted LUT) from dorsal nerve cords of wild type and unc-104(e1265) mutants. Images on each line are from different young adult animals (5 shown for each genotype). (J) Representative confocal images of cholinergic ELKS-1 (Punc-129::ELKS-1::mCherry, inverted LUT) from dorsal nerve cords of wild type and unc-104(e1265) mutants. Images on each line are from different young adult animals (5 shown for each genotype).

https://doi.org/10.1371/journal.pgen.1010016.g004

The density of spines and AChR clusters were also severely reduced in L4 stage unc-104(e1265) mutants (Fig 4D–4F), similar to nrx-1 mutants (Fig 4G). These effects were rescued by either native or cholinergic expression of wild type unc-104 in unc-104(e1265) mutants, but not by GABA- or muscle-specific unc-104 expression (Fig 4D–4F), demonstrating that UNC-104-mediated transport is critical in presynaptic cholinergic axons for postsynaptic spine development and receptor localization.

We next asked how a failure in delivery of AZ components may contribute to the severe reductions in spines and receptor clusters we observed in unc-104 mutants. In addition to CLA-1 [37], transport of both ELKS-1/ELKS/CAST and UNC-10/RIM were severely disrupted by mutation of unc-104, indicating that Kinesin-3 mediated transport is key for the synaptic delivery and assembly of several active zone constituents in these neurons (Figs 4H–4J and S13). Remarkably, we found that only mutation of nrx-1 produced significant decreases in dendritic spines (Fig 4G), demonstrating that presynaptic ELKS, UNC-10, and CLA-1 are dispensable for postsynaptic development. Our results demonstrate a specific requirement for NRX-1 in the stabilization and maturation of postsynaptic structures and provide evidence that a failure in synaptic delivery of NRX-1 is a primary causal factor in the postsynaptic structural defects of unc-104 mutants.

The stabilization of mature dendritic spines requires ongoing synaptic delivery of NRX-1

Our results suggest UNC-104 mediated transport positions NRX-1 at the presynaptic terminal in the early stages of synapse formation where it acts to stabilize growing postsynaptic structures, including spines and receptor clusters. We next sought to address whether there is a requirement for presynaptic UNC-104-dependent transport in the stabilization of mature spines. To address this question, we used the previously characterized temperature-sensitive allele, unc-104(ce782). unc-104(ce782) animals carry a G105E missense mutation in the motor domain of UNC-104 [52]. When grown at the permissive temperature (13.5°C), unc-104(ce782) axonal synaptic vesicle abundance and animal motility are modestly reduced compared to wild type. In contrast, synaptic vesicles are completely absent from unc-104(ce782) axons following growth at restrictive temperature (20–25°C), and animal motility is severely compromised within 12 hours of a shift to 23°C [52]. We raised unc-104(ce782) animals to L4 stage at the permissive temperature (13.5°C). We then shifted L4 stage animals to the restrictive temperature (25°C) for 16–20 hrs, and quantified NRX-1::GFP localization and spine density immediately following this shift (Fig 5). Axonal NRX-1::GFP clusters were strikingly decreased in unc-104(ce782) mutants subjected to the temperature shift compared with control animals subjected to the same shift or unc-104(ce782) mutants raised continuously at the permissive temperature (Fig 5B, 5C). unc-104(ce782) mutants grown at the permissive temperature had a slightly reduced number of spines overall compared with wild type (Fig 5E). A shift to the restrictive temperature at L3 or L4 stage produced a striking reduction in spine density for unc-104(ce782) mutants (Figs 5D, 5E and S14), but not for wild type. Taken together, our findings indicate an ongoing requirement for UNC-104 transport that extends well beyond the period of initial synapse formation and spine outgrowth. Among the UNC-104 cargoes we investigated, only deletion of nrx-1 produces a significant reduction in spine density at L4 stage. We therefore propose a model where ongoing UNC-104 delivery of presynaptic NRX-1 is critical for postsynaptic maturation and maintenance of mature spines.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. UNC-104/KIF1A transport of NRX-1 is required in the mature circuit to maintain postsynaptic structure.

(A) Cartoon depiction of experimental timeline. Animals were grown at 13.5°C until L4 stage before shifting animals to the restrictive temperature of 25°C for 16–20 hours and imaging. (B) Line scan fluorescent images of NRX-1::GFP (knock-in to endogenous locus, inverted LUT) from dorsal cords of wild type and unc-104(ce782) mutants grown continuously at the permissive temperature (13.5°C) (middle panel) or shifted to the restrictive temperature (25°C) (bottom panel) at L4 stage for 16–20 hours before imaging (bottom panel). Each line scan indicates an individual animal. (C) Quantification of NRX-1::GFP clusters along a 50 μm region of the dorsal nerve cord in wild type and unc-104(ce782)ts animals. Two-way ANOVA, Tukey’s multiple comparisons test, ns, not significant, *p<0.05, n ≥ 11 animals. Data points indicate mean ± SEM. (D) Fluorescent images of DD spines in wild type and unc-104(ce782)ts animals grown continuously at the permissive temperature (13.5°C) (middle panel) or shifted to the restrictive temperature (25°C) (bottom panel) at L4 stage for 16–20 hours before imaging. (E) Quantification of the number of spines per 15 μm in wild type and unc-104(ce782)ts animals. Two-Way ANOVA, Tukey’s multiple comparisons test, ns, not significant, **p<0.001. n ≥ 13 animals. Data points indicate mean ± SEM.

https://doi.org/10.1371/journal.pgen.1010016.g005

C. elegans GABAergic motor neurons are decorated with functional dendritic spines

Dendritic spines are known to act as specialized sites for compartmentalizing neurotransmission and are widely observed across various classes of neurons in both vertebrates and invertebrates [53]. Though suggested by prior EM studies, the presence of functional dendritic spines on C. elegans neurons had not been fully appreciated until recently [20,22,23]. Here we examined the developmental trajectory of postsynaptic structures (dendritic spines and cholinergic receptor clusters) located on GABAergic motor neurons relative to presynaptic release sites and showed that the stabilization of growing dendritic spines requires presynaptic NRX-1/neurexin.

Neural circuit organization and performance is dictated by the locations of synaptic connections and the identities of the interconnected neurons. Elucidation of the processes governing the construction of synapses and their coordination across presynaptic axons and postsynaptic dendrites are therefore fundamental for understanding circuit assembly and function. However, relatively few studies have monitored in real-time the coordinated assembly of pre- and postsynaptic specializations in vivo. Our work tackled this important question, taking advantage of the GABAergic DD spine model and the cellular precision offered by the C. elegans motor circuitry. We find that C. elegans DD dendritic spines share many of the hallmark features of vertebrate dendritic spines, but also have some key differences. Most notably, mature C. elegans spines display limited overt activity-dependent dynamics in wild type animals and appear comparatively stable relative to spines in rodents. This raises the important question of what is the functional role of spines in GABAergic neurons? Prior studies have provided evidence for compartmentalized calcium signals in C. elegans neurons [54]. Therefore, one intriguing possibility is that, similar to the situation in mammals, spines serve to compartmentalize calcium and perhaps other biochemical signals in GABAergic neurons. Spines may also be required in order to achieve the unusual dyadic arrangement of synapses in the ventral nerve cord, where presynaptic specializations of cholinergic motor neurons are positioned for transmission onto both GABA motor neuron and muscle postsynaptic partners [55]. In this case, GABAergic spines may have developed to intercept cholinergic release sites onto muscles, as suggested previously [56]. Although not yet characterized by light microscopy, prior electron microscopy studies indicated the presence of spine-like structures in other C. elegans neurons such as RMD, RME, SMD, and RIP [21,24,56]. Additional investigation of these structures should help to elucidate their precise functional roles.

We found that F-actin is compartmentalized to discrete regions within the dendritic shafts of DD neurons very early in the development of the circuit (prior to the appearance of spines) and later becomes exclusively localized to dendritic spines. F-actin labeling of spines is apparent as soon as they are detectable, suggesting that F-actin assemblies may participate in the earliest stages of spine development. Prior to remodeling of DD neurons, sites of GABA release are located on the ventral processes of DD neurons. As F-actin based structures are often associated with release sites [57,58], our finding that F-actin is clustered in ventral DD processes prior to remodeling may suggest that F-actin initially associated with release sites persists in the DD neurite through remodeling and is then repurposed for postsynaptic development, perhaps acting as a landmark or actin-based mechanism for stimulating postsynaptic maturation.

Notably, we also observed punctate cholinergic receptor fluorescence in GABAergic dendrites several hours prior to the formation of dendritic spines. These immature receptor clusters colocalize with dendritic F-actin assemblies, suggesting that receptors are trafficked into dendrites and positioned with F-actin to rapidly populate growing spines. Indeed, we observed that receptor clusters are visible in growing spines as soon as spines can be clearly resolved. This is consistent with recent findings from time-lapse 2-photon imaging of hippocampal organotypic slice cultures showing that receptor accumulation in spines occurs concurrently with spine outgrowth [12]. Similarly, we found that synaptic material accumulates in presynaptic axons prior to the emergence of spines from dendritic processes. In particular, initial accumulations of the presynaptic scaffold CLA-1 and the synaptic vesicle marker synaptobrevin/SNB-1 were visible in the axon with a similar time-course to the appearance of dendritic cholinergic receptor clusters. Presynaptic material became more clearly localized to discrete puncta over the next 4–6 hours, occurring roughly coincident with receptor accumulation at the tips of growing spines. Notably, our findings do not support a strong requirement for presynaptic neurotransmitter release in spine formation.

Transsynaptic NRX-1/Neurexin signaling stabilizes dendritic F-actin assemblies to promote postsynaptic maturation

Numerous studies across a variety of model systems have examined roles for neurexins in synapse development and function. However, no single consensus view of neurexin function has emerged. Instead, functions for neurexins at specific synapses are thought to be dictated by cellular and molecular context [19,59]. Mammalian genomes encode three Nrxn genes that can give rise to α-, β-, and γ-Nrxn isoforms. These isoforms share common intracellular and transmembrane regions but differ in their extracellular domains. α-Nrxn has six extracellular LNS domains interleaved with three EGF-like repeats. β-Nrxn has only one LNS domain, while γ-Nrxn lacks all identifiable extracellular domains. The multi-faceted roles for neurexins at vertebrate synapses likely emerge as a consequence of the many neurexin isoforms that are generated by alternative splicing, the complexity of their cellular expression, and the potential for these isoforms to selectively interact with distinct postsynaptic partners.

The C. elegans genome encodes a single neurexin gene, nrx-1, which generates long α and short γ isoforms [60]. As is the case in other systems, C. elegans NRX-1 has roles in presynaptic organization, for example in calcium channel clustering at the active zone [47]. Interestingly, the short γ-NRX-1 isoform that lacks identifiable ectodomains performs these roles (worms do not encode β-NRX-1), but even in the absence of both isoforms, a level of presynaptic functionality is maintained, as indicated by the presence of evoked responses in muscle cells following presynaptic motor neuron stimulation [20]. Our prior studies pointed toward the importance of NRX-1 at synapses between cholinergic and GABAergic motor neurons [20]. Surprisingly, our time course studies here revealed that initial spine outgrowth and AChR clustering occur normally in nrx-1 mutants lacking both isoforms. However, spines and AChRs in nrx-1 mutants become destabilized within hours and disperse, such that GABAergic dendrites are almost completely devoid of these postsynaptic components by L4 stage. Dendritic F-actin assemblies, that normally show discrete localization to spines, are also disorganized in nrx-1 mutants, and are diffusely distributed throughout GABAergic dendritic processes in nrx-1 mutants. We propose a model where presynaptic NRX-1 is required for the stabilization of dendritic spines and other postsynaptic structures in GABAergic dendrites. We suggest that presynaptic NRX-1 directs the organization of dendritic F-actin to stimulate maturation and stabilization of spines and postsynaptic receptor clusters. A similar form of F-actin based reorganization is a key step in presynaptic differentiation [25,57]. Our findings contrast with previous culture studies that implicated neurexin in postsynaptic induction [61].

Our work has parallels with recent studies in Xenopus and Drosophila. In embryonic Xenopus brain, presynaptic β-neurexin stabilizes dendritic filopodia through an adhesive partnership with neuroligin to direct the development of dendritic arbors [62]. Similarly, during fly metamorphosis, neurexin/neuroligin-based adhesion promotes the growth of neurite branches independently of synaptic activity (Constance et al., 2018) [63]. In the present studies, we showed NRX-1 stabilization of dendritic spines occurs independently of synaptic activity. Previous studies of nlg-1 reporters [64,65] and more recent transcriptome profiling efforts [66] show that nlg-1 is not strongly expressed in GABAergic motor neurons. Consistent with these findings, we previously showed deletion of neuroligin/nlg-1 did not significantly alter either dendritic spines or receptor clusters in GABAergic neurons at L4 stage [20]. These results suggest NRX-1 interaction with alternate binding partners in GABAergic neurons that to date remain unidentified. Identifying such partners will be an interesting area for future studies.

UNC-104/Kinesin 3-dependent localization of NRX-1/Neurexin is an essential early step required for synapse maturation and stabilization

Efficient trafficking of synaptic and active zone cargoes to presynaptic terminals is essential for synapse formation and neuronal communication. Long-range axonal transport of synaptic vesicles and active zone proteins is carried out by anterograde kinesin and retrograde dynein motors. However, significant questions remain about which synaptic proteins are trafficked together and their dependence on specific motors. We showed NRX-1 localization to presynaptic terminals is dependent on kinesin-3 UNC-104/KIF1A. The anterograde transport of synaptic and dense core vesicles are also strongly dependent upon UNC-104 [48,67]. In contrast, several studies suggest that other AZ proteins exhibit weaker requirements for UNC-104 in their delivery. For example, ELKS-1 and other AZ proteins localize to C. elegans HSN synapses in an UNC-104-independent manner [6]). Similarly, UNC-10/RIM is properly concentrated in nerve ring processes of unc-104(e1265) mutants [68]. Here, we found that NRX-1, UNC-10/RIM, ELKS-1/ELKS and CLA-1/piccolo all show a dependence on UNC-104 for anterograde transport to cholinergic synapses. The unc-104(e1265) allele used in our study carries a point mutation in the PI(4,5)P2 binding pocket of the PH domain and reduces the ability of UNC-104 to bind cargo [49], perhaps implicating this domain in UNC-104-dependent localization of these proteins.

Prior studies of rodent cultured neurons also suggested a requirement for the Kinesin-3 motor in neurexin transport, indicating that mechanisms for NRX-1 delivery to synapses are conserved [69]. We found that dendritic spines collapse and AChR clusters disperse when UNC-104 delivery of synaptic cargoes is disrupted. Of the putative UNC-104 cargoes we analyzed, only mutation of nrx-1 produced significant disruption of dendritic spines. We therefore propose that NRX-1 is a key UNC-104 cargo required for spine stabilization. Our live imaging points to similar rates of transport for synaptic vesicles and NRX-1. However, it remains unclear whether NRX-1 is delivered as a component of synaptic vesicles or may be segregated into a distinct vesicular population, such as synaptic vesicle protein transport vesicles (STVs). Our findings that mutations which severely impair synaptic vesicle fusion do not affect spine density point toward the latter possibility.

The prior identification of a temperature-sensitive allele of unc-104 allowed us to explore the temporal requirements for UNC-104 delivery of NRX-1. Importantly, we found that UNC-104 delivery was not solely required during early stages of synapse maturation but was critical throughout the developmental progression of the circuit toward maturity. These findings argue that delivery of presynaptic NRX-1 is required to maintain postsynaptic structure long after circuit assembly is complete. This raises interesting questions about the relationship between NRX-1 transport and synapse stability, perhaps suggesting that alterations in the rate of axonal NRX-1 transport may directly impact synaptic connectivity in mature animals. More broadly, our studies provide a new view of adhesive mechanisms in circuit connectivity and highlight a novel role for neurexin in the stabilization of mature synapses and dendritic spines.

Strains

All strains are N2 Bristol strain derivatives (wild type) and were maintained at room temperature (20–24°C) on nematode growth media plates (NGM) seeded with E. coli strain OP50. Transgenic strains were obtained by microinjection to achieve transformation [70] and identified using co-injection markers. Integrated lines were produced with X-ray irradiation and outcrossed to wild type/N2 Bristol eight times. Only hermaphrodites (L1-L4, young adults) were used in this study. A complete list of all strains used in this study is found in S1 File. Worms used for time course studies were staged by transferring freshly hatched larvae to seeded OP50 plates and transferring to 25°C (time point 0).

Molecular biology

Plasmids were constructed using the two-slot Gateway Cloning system (Invitrogen) and confirmed by restriction digest and sequencing.

Hydrogel solution

Hydrogel was prepared as an aqueous solution of polyethylene glycol (20% w/v) and the photoinitiator Irgacure (0.5% stock, 0.1% working concentration), and stored at 4°C in the dark.

Confocal microscopy

All strains were immobilized with sodium azide (0.3 M) on a 2% or 5% agarose pad. Images were obtained using an Olympus BX51WI spinning disk confocal equipped with a 63x objective. For time course analyses, newly hatched larvae were transferred to a seeded OP50 plate and maintained at 25°C.

Confocal microscopy analysis

All image analysis was conducted using ImageJ software (open source) within defined ROIs using intensity threshold values determined from control experiments. ROIs were located 15–30 μm anterior to DD1, DD2 or DD3 somas.

Fluorescence intensity analysis

Live Imaging

An event is defined as an incidence of spine destabilization/pruning, elongation/formation, or a stable spine during an entire live imaging recording. Z-stacks of DD1 dendrites were acquired every 5 minutes for 60–160 minutes. Numbers indicates number of events for each genotype at each timepoint.

Calcium Imaging

Animals were grown on plates seeded with OP50 containing 2.7 mM All-Trans Retinal (ATR). Plates were stored at 4°C under dark conditions and used within one week. Imaging was carried out using L4 animals immobilized in hydrogel [72]. Animals were transferred to 7.5 μL of the hydrogel mix placed on a silanized glass slide and covered with a glass slide. Hydrogel was cured using a handheld UV Transilluminator (312 nm, 3 minutes). Post-curing, the covering slide was removed and replaced with a coverslip. Imaging was carried out using a Yokogawa CSU-X1-A1N spinning disk confocal system (Perkin Elmer) equipped with EM-CCD camera (Hamamatsu, C9100-50) and 63X oil immersion objective. Chrimson photoactivation (~30 mW/cm²) was achieved using a TTL-controlled 625 nm light guide coupled LED (Mightex Systems). A 556 nm BrightLine single-edge short-pass dichroic beam splitter was positioned in the light path (Semrock) (S3 Fig). Data were acquired at 10 Hz for 15 s using Volocity software and binned at 1×1 during acquisition. Analysis was performed using ImageJ. The DD neurite process in each time series was extracted using the straighten function, background subtracted and photobleaching correction was carried out by fitting an exponential function to the data (CorrectBleach plugin). A smoothing function was applied to the data to enhance signal-to-noise. Individual spine ROIs were identified using the mCherry fluorescence. Post imaging processing, pre-stimulus baseline fluorescence (F₀) was calculated as the average of the data points in the first 4 s of the recording. Data was normalized to prestimulus baseline as ΔF/F₀, where ΔF = F–F₀. Peak ΔF/F₀ was determined by fitting a Gaussian function to the ΔF/F₀ time sequence using Multi peak 2.0 (Igor Pro, WaveMetrics). All collected data were analyzed, including failures (no response to stimulation). Control recordings were carried out in the absence of Retinal.

Temperature shift experiments

Wild type and unc-104(ce782)ts animals eggs were raised at 13.5°C until L3 or L4 stage as indicated, then shifted to 25°C for 16–20 hrs prior to imaging. Animals were age-matched at the time of the temperature shift to account for developmental delays of unc-104(ce782)ts mutants, approximately 144 hours to L4 stage for unc-104(ce782)ts mutants, and 120 hours for wild type.