ABSTRACT
Dendritic arbor development is a complex, highly regulated process. Post-transcriptional regulation mediated by RNA-binding proteins plays an important role in neuronal dendrite morphogenesis by delivering on-site, on-demand protein synthesis. Here, we show how the Drosophila fragile X mental retardation protein (FMRP), a conserved RNA-binding protein, limits dendrite branching to ensure proper neuronal function during larval sensory neuron development. FMRP knockdown causes increased dendritic terminal branch growth and a resulting overelaboration defect due, in part, to altered microtubule stability and dynamics. FMRP also controls dendrite outgrowth by regulating the Drosophila profilin homolog chickadee (chic). FMRP colocalizes with chic mRNA in dendritic granules and regulates its dendritic localization and protein expression. Whereas RNA-binding domains KH1 and KH2 are both crucial for FMRP-mediated dendritic regulation, KH2 specifically is required for FMRP granule formation and chic mRNA association, suggesting a link between dendritic FMRP granules and FMRP function in dendrite elaboration. Our studies implicate FMRP-mediated modulation of both the neuronal microtubule and actin cytoskeletons in multidendritic neuronal architecture, and provide molecular insight into FMRP granule formation and its relevance to FMRP function in dendritic patterning.
INTRODUCTION
Neurons display complex and diverse dendritic arbor morphologies that determine their ability to receive and process different signals. Numerous molecular and cellular regulatory mechanisms are required during dendrite morphogenesis to establish and maintain proper dendritic patterning and field coverage (Jan and Jan, 2010). Defects in dendritic arbor development cause a variety of neurological and neurodegenerative diseases (Lukong et al., 2008; Kulkarni and Firestein, 2012; Koleske, 2013). However, how dendritic patterning is precisely controlled to ensure neuronal function and how failure of this regulation results in neuropathology are incompletely understood.
RNA binding proteins (RBPs) play key roles in dendritic arbor development and function by allowing dynamic spatial and temporal control of gene expression, especially on-site, on-demand protein synthesis (Lukong et al., 2008). In neurons, numerous mRNAs are incorporated into ribonucleoprotein particles (RNPs) and translationally silenced for transport from the soma to the dendrites, where they can be locally translated. RBPs have been shown to modulate dendritic mRNA localization and/or local translation for proper spine maturation and synaptic plasticity (Martin and Zukin, 2006; Bramham and Wells, 2007; Besse and Ephrussi, 2008; Jung et al., 2014; Glock et al., 2017), but less is known about the role of RBPs in dendrite morphogenesis, in particular what processes require them and what RNAs they regulate for these processes to occur properly.
Fragile X mental retardation protein (FMRP), a highly conserved RBP, was previously identified as a negative regulator of dendrite growth. Loss of functional FMRP results in Fragile X syndrome (FXS), the most common inherited intellectual disability (Bagni and Greenough, 2005; Bassell and Warren, 2008; Santoro et al., 2012). In addition to mammalian systems, Drosophila is a widely used FXS model as mutations in dfmr1 (Fmr1 – FlyBase), the gene encoding the Drosophila FMRP homolog, cause phenotypes resembling those in individuals with FXS (Drozd et al., 2018). Similar to mouse (Comery et al., 1997; Nimchinsky et al., 2001) and human (Irwin et al., 2001) Fmr1 knockout neurons, overly complex and disordered dendritic patterns (including longer and denser branches) were observed in Drosophila dfmr1 mutant neurons (Lee et al., 2003; Pan et al., 2004).
FMRP has a broad spectrum of mRNA targets and, in neurons, it has been shown to direct microtubule-dependent dendritic RNA transport (Antar et al., 2005; Estes et al., 2008; Dictenberg et al., 2008) and to repress mRNA translation by multiple mechanisms through interactions with ribosomes (Darnell et al., 2011; Chen et al., 2014), miRNAs (Ishizuka et al., 2002; Xu et al., 2008; Yang et al., 2009) and cytoplasmic FMRP interacting protein 1 (CYFIP) (Schenck et al., 2003; Napoli et al., 2008; Abekhoukh et al., 2017). Consistent with a role for FMRP in regulation of dendrite architecture, many FMRP target transcripts identified by high-throughput methods encode proteins involved in dendrite development and cytoskeletal remodeling (Brown et al., 2001; Darnell et al., 2011; Ascano et al., 2012; McMahon et al., 2016; Maurin et al., 2018). Functional studies showed that FMRP regulates microtubule and actin modulators, including map1b/futsch (Zhang et al., 2001), profilin/chickadee (Reeve et al., 2005; Tessier and Broadie, 2008) and arc/arg3.1 (Park et al., 2008), for proper axonal branching, synaptic structures and neuronal activity. However, the role of FMRP in generating the complex patterns of dendritic arbors through its regulation of cytoskeletal targets is not well understood.
Here, we have addressed this gap by investigating FMRP function in Drosophila larval dendritic arborization (da) neurons, an ideal model of dendrite morphogenesis. These sensory neurons are subdivided into four classes (classes I-IV), each with its own distinct dendritic arbor morphology and function (Grueber et al., 2002; Jan and Jan, 2010). Class IV da (C4da) neurons develop the most elaborate arbors, which completely and non-redundantly cover the entire larval body wall to detect noxious stimuli (Tracey et al., 2003; Hwang et al., 2007; Xiang et al., 2010; Zhong et al., 2010; Terada et al., 2016). These arbors are largely confined to a two-dimensional space beneath the epidermis (Han et al., 2012; Kim et al., 2012) and can be easily visualized throughout larval development. The complex C4da dendritic pattern is generated by a highly dynamic process during which the main branch network is established initially while higher-order branches elaborate and are refined dynamically throughout larval development to provide proper field coverage. This process is controlled by both intrinsic factors, including transcriptional and cytoskeletal regulators, and extrinsic factors, such as the ECM and epidermal signals (Jan and Jan, 2010).
We show that FMRP is required in C4da neurons for proper dendrite morphogenesis and neuronal function, and acts by modulating the neuronal cytoskeleton. FMRP controls microtubule stability in main branches and microtubule dynamics in distal, higher-order branches, which prevents over-elaboration of dendritic arbors. In addition, FMRP controls terminal branch extension in part through its regulation of chickadee (chic), which encodes the Drosophila profilin homolog. FMRP colocalizes with chic mRNA in dendritic granules, and regulates its dendritic localization and protein expression. Finally, we show that FMRP function in dendritic arborization requires two of its three RNA-binding domains (RBDs), KH1 and KH2, but not its RGG domain. FMRP granule formation and chic mRNA association is primarily mediated by the KH2 domain and correlates with FMRP activity in C4da neurons. Together, our results support a model in which FMRP controls dendrite stability and outgrowth via both microtubule and actin remodeling to generate the appropriate multidendritic architecture for sensory neuron function. Our findings also suggest that FMRP granule formation is linked to its function in dendritic arborization.
RESULTS
FMRP negatively regulates branching of C4da dendrites
Previous analysis of dfmr1-null mutant larvae showed that C4da neuronal arbors have excess terminal branches when compared with those of wild-type larvae (Lee et al., 2003), but it is unclear how dfmr1 loss of function leads to abnormal dendrite morphogenesis. To address this issue, we first verified FMRP expression in C4da neurons using anti-FMRP immunofluorescence (IF). FMRP was detected in the soma, mainly in the cytoplasm (Fig. S1A-A″). Furthermore, FMRP particles were visible along C4da dendrites in neurons expressing GFP-tagged FMRP (Fig. S1D-D″), consistent with a RNA regulatory role in dendritic arborization.
To better assess the requirement for FMRP in C4da dendrite morphogenesis, we examined the branch morphology of FMRP loss-of-function mutant (dfmr1Δ50M) neurons in more detail, through three larval developmental stages (L1, L2 and L3). C4da dendrite morphology was visualized at the end of each larval stage using the membrane marker CD4::tdGFP driven by ppk-GAL4, a C4da neuron-specific driver. Primary branches of wild-type C4da neurons are established early, and the size and complexity of arbors increase to maintain the dendritic space filling pattern as the larva grows dramatically (Fig. 1A-A″). In dfmr1Δ50M neurons, an increase in branch number was first observed at the end of L1 (Fig. 1B,E), and this defect was exacerbated as development proceeded through L3. During this period, dfmr1Δ50M neurons exhibited dramatic branch outgrowth, resulting in an increase in the number of branches (branch points), total dendritic length and field coverage, and an overall dendritic pattern that was denser than that of wild type (Fig. 1B′,B″,D-G).
FMRP negatively regulates branching of C4da dendrites. (A-C) Representative images of C4da neurons labeled with CD4-tdGFP in wild-type (A-A″), dfmr1Δ50M (B-B″) and dfmr1OE (C-C″) larvae at the end of L1 (A,B,C), L2 (A′,B′,C′) and L3 (A″,B″,C″) stages. (D-G) Quantification of average terminal length (D), terminal branch number (E), total dendritic length (F) and the dendritic field coverage ratio (G) from C4da neurons with the indicated genotypes. Terminal branch number and the total dendritic length were normalized to the area of the image field at each time point. All images are confocal z-series projections. Scale bars: 50 μm. Each dot represents a value for one neuron. (H) Quantification of time elapsed (latency) before the larva exhibits each of five sequential, stereotyped behaviors (head thrash, roll, whip, seizure and paralysis); wild-type (n=53), dfmr1RNAi (n=48), dfmr1OE (n=55) and dfmr1RNAi, dfmr1OE (rescue, n=42) larvae. dfmr1OE likely partially attenuates dfmr1 RNAi, and vice versa, leading to a physiological level of dfmr1 expression. We noticed that in some cases the rolling and whipping behaviors occurred in the reverse order. Each dot represents a value for one larva. Data are mean±s.d.; ns, not significant; *P<0.05, **P<0.01, ****P<0.0001; one-way ANOVA with Dunnett's multiple comparisons test.
FMRP negatively regulates branching of C4da dendrites. (A-C) Representative images of C4da neurons labeled with CD4-tdGFP in wild-type (A-A″), dfmr1Δ50M (B-B″) and dfmr1OE (C-C″) larvae at the end of L1 (A,B,C), L2 (A′,B′,C′) and L3 (A″,B″,C″) stages. (D-G) Quantification of average terminal length (D), terminal branch number (E), total dendritic length (F) and the dendritic field coverage ratio (G) from C4da neurons with the indicated genotypes. Terminal branch number and the total dendritic length were normalized to the area of the image field at each time point. All images are confocal z-series projections. Scale bars: 50 μm. Each dot represents a value for one neuron. (H) Quantification of time elapsed (latency) before the larva exhibits each of five sequential, stereotyped behaviors (head thrash, roll, whip, seizure and paralysis); wild-type (n=53), dfmr1RNAi (n=48), dfmr1OE (n=55) and dfmr1RNAi, dfmr1OE (rescue, n=42) larvae. dfmr1OE likely partially attenuates dfmr1 RNAi, and vice versa, leading to a physiological level of dfmr1 expression. We noticed that in some cases the rolling and whipping behaviors occurred in the reverse order. Each dot represents a value for one larva. Data are mean±s.d.; ns, not significant; *P<0.05, **P<0.01, ****P<0.0001; one-way ANOVA with Dunnett's multiple comparisons test.
In contrast to loss of dfmr1, overexpression of full-length FMRP (dfmr1OE) in C4da neurons caused a reduction in terminal branch length and dendritic field coverage when compared with wild type at the end of L1 (Fig. 1C,D,G). These defects became more severe as development progressed to the end of L3 where higher-order branches were missing, leading to a sparse distribution of dendritic arbors (Fig. 1C′,C″,D-G). Severing and fragmentation of branches was observed peripherally in dfmr1OE neurons at the end of L2 (Fig. S2). This process is characteristic of the dendrite pruning that occurs during pupariation (Williams and Truman, 2005), suggesting that the decrease in arbor complexity of dfmr1OE neurons is due in part to the elimination of higher-order branches through premature pruning. Together, the dfmr1 mutant and overexpression phenotypes indicate that FMRP acts as a negative regulator to limit C4da dendrite branching during larval development.
FMRP regulates dendrite development cell autonomously
To test whether FMRP modulates dendrite morphogenesis in a cell-autonomous manner, we knocked down FMRP in C4da neurons by expressing UAS-dfmr1RNAi specifically in these neurons using ppk-GAL4. FMRP levels were greatly decreased in dfmr1RNAi neurons and dramatically increased in dfmr1OE neurons (Fig. S1B-B″,C-C″), as determined by anti-FMRP immunofluorescence. Similar to dfmr1Δ50M neurons, dfmr1RNAi neurons displayed an overelaboration defect, including an increase in branch number, total branch length and dendritic field coverage throughout larval development (Fig. S3A-A‴,B-B‴,D-G). Thus, FMRP acts cell autonomously to regulate C4da dendrite branching. Given the high knockdown efficiency of dfmr1 RNAi and the similar over-elaboration defect exhibited by both dfmr1Δ50M and dfmr1RNAi neurons, we performed subsequent experiments with dfmr1 RNAi to simplify genetic manipulations.
Loss of FMRP impairs neuronal functions
C4da neurons are nociceptors that perceive noxious thermal and mechanical stimuli (Tracey et al., 2003; Hwang et al., 2007; Zhong et al., 2010; Terada et al., 2016). The morphological defects induced by FMRP knockdown in C4da neurons suggested that sensory function might also be impaired. To test this, we performed a heat-plate assay in which larvae are subjected to a noxious heat stimulus and monitored for a stereotypic series of locomotor responses (see Materials and Methods; Chattopadhyay et al., 2012). Compared with wild-type larvae, dfmr1RNAi larvae exhibited a significantly faster response to high temperature, as evidenced by the shorter time required to elicit the first locomotory behavior (head thrash), and they displayed longer response periods for subsequent behaviors prior to end-stage paralysis (Fig. 1H). Conversely, we observed a significant delay of the head thrash response in dfmr1OE larvae. Overexpression of FMRP in dfmr1RNAi C4da neurons rescued the nocifensive behavioral defects (Fig. 1H). Together, these results indicate that FMRP is required in C4da neurons for their function in thermal nociception.
Loss of FMRP leads to increased terminal branch growth activity
The increased branch number, terminal branch length and field coverage defects in FMRP mutant and RNAi neurons suggest that FMRP might control terminal branch growth. To test this idea, we performed a time-lapse analysis and compared branch morphology of the same neurons at two time points (0 min and 30 min) during L2. In wild-type larvae, about 50% of dendrites were dynamic, exhibiting a balance of both growth (new outgrowth and/or extension of existing dendrites) and retraction events. We found that 38% of terminal branches in dfmr1RNAi neurons extended during the period when compared with 26% in wild type, while the frequency of retraction was comparable (24% of terminal branches in dfmr1RNAi and 26% in wild type) (Fig. 2A,A′,B,B′,D). This 1.7-fold change in the dendrite growth/retraction ratio between dfmr1RNAi and wild-type neurons (Fig. 2E) indicates a net increase in terminal branch growth activity in the absence of FMRP. dfmr1OE neurons exhibited a wild-type frequency of growing dendrites (28%) and a slight increase in the frequency of retracting dendrites (32%, P=0.0597) (Fig. 2A,A′,C,C′,D), suggesting that, along with inappropriate pruning, increased retraction activity of terminal branches could contribute to branch loss in dfmr1OE neurons. Together, these data show that FMRP controls branch growth activity and that FMRP-mediated dendritic arbor defects may be the result of increased terminal branch outgrowth.
Loss of FMRP leads to increased terminal branch growth activity. (A-C) Representative images of dendritic terminal dynamics at 0 and 30 min in wild-type (A,A′), dfmr1RNAi (B,B′) and dfmr1OE (C,C′) 2nd instar larvae. C4da neurons are labeled with CD4-tdGFP. Arrowheads indicate growing (magenta), retracting (cyan) or stable (gray) branches. (D) Proportions of growing (magenta), retracting (cyan) and stable (gray) termini. (E) Quantification of the growth to retraction ratios. Each dot represents a value for one neuron. All images are confocal z series projections. Scale bar: 50 μm. Data are mean±s.d.; ns, not significant; **P<0.01; one-way ANOVA with Dunnett's multiple comparisons test.
Loss of FMRP leads to increased terminal branch growth activity. (A-C) Representative images of dendritic terminal dynamics at 0 and 30 min in wild-type (A,A′), dfmr1RNAi (B,B′) and dfmr1OE (C,C′) 2nd instar larvae. C4da neurons are labeled with CD4-tdGFP. Arrowheads indicate growing (magenta), retracting (cyan) or stable (gray) branches. (D) Proportions of growing (magenta), retracting (cyan) and stable (gray) termini. (E) Quantification of the growth to retraction ratios. Each dot represents a value for one neuron. All images are confocal z series projections. Scale bar: 50 μm. Data are mean±s.d.; ns, not significant; **P<0.01; one-way ANOVA with Dunnett's multiple comparisons test.
FMRP regulates organization of stable microtubules along C4da dendrites
To elucidate the mechanisms by which FMRP restrains branching, we first focused on microtubules, which play a key role in dendritic arbor architecture. Stable microtubules are found throughout the main branches of C4da neurons (Jenkins et al., 2017) and are thought to provide structural support for arbor architecture, while microtubule nucleation and polymerization in nascent branches contribute to dendrite branching (Delandre et al., 2016).
Given the importance of microtubules in dendrite morphogenesis, we hypothesized that FMRP controls branching in part by regulating microtubule stability. To address this hypothesis, we analyzed the effect of dfmr1 RNAi on stable microtubule levels by immunostaining for Futsch, a neuron-specific stable microtubule marker (Hummel et al., 2000; Roos et al., 2000). In dfmr1RNAi neurons, Futsch levels were significantly elevated and the Futsch signal extended further from the soma when compared with wild-type neurons (Fig. 3B-B″,D,E). On the contrary, dfmr1 overexpression resulted in a dramatic downregulation of Futsch in both the soma and dendrites (Fig. 3C-C″,D,E). Similar results were obtained when we visualized acetylated α-tubulin (Fig. S4), another widely used stable microtubule marker (Jenkins et al., 2017). To better characterize the effects of dfmr1 on stable microtubules, we quantified the proportion of branches that were Futsch positive. dfmr1 RNAi caused a reduction in the percentage of Futsch-positive branches but this was most likely due to the dramatic increase in the number of higher-order branches without Futsch (Fig. 3F). From the above results, we conclude that FMRP destabilizes microtubules, primarily in main branches.
FMRP regulates the organization of stable microtubules along C4da dendrites. (A-C) Immunofluorescence detection of Futsch, a neuronal stable microtubule marker, in wild-type (A-A″), dfmr1RNAi (B-B″) and dfmr1OE (C-C″) larvae. C4da neurons are labeled with CD4-tdGFP (A-C, red). Futsch (A′-C′, green) was detected with anti-22C10 antibody and the non-ddaC signals were masked to better visualize Futsch in C4da neurons. A″-C″ show merged images. (D) Quantification of relative Futsch intensity (normalized to the intensity of the neuronal marker) in C4da dendrites from the indicated genotypes. (E) Plot of relative Futsch intensity along the primary dendrites of wild-type (green), dfmr1RNAi (red) and dfmr1OE (purple) C4da neurons. Lines indicate the mean values; lighter shading denotes s.d. (F) Quantification of proportion of branches that were Futsch positive. Percentages are the number of Futsch-positive branches divided by the number of total branches. All images are confocal z series projections. In D,F, each dot represents a value for one neuron. Scale bar: 50 μm. Data are mean±s.d.; *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; one-way ANOVA with Dunnett's multiple comparisons test.
FMRP regulates the organization of stable microtubules along C4da dendrites. (A-C) Immunofluorescence detection of Futsch, a neuronal stable microtubule marker, in wild-type (A-A″), dfmr1RNAi (B-B″) and dfmr1OE (C-C″) larvae. C4da neurons are labeled with CD4-tdGFP (A-C, red). Futsch (A′-C′, green) was detected with anti-22C10 antibody and the non-ddaC signals were masked to better visualize Futsch in C4da neurons. A″-C″ show merged images. (D) Quantification of relative Futsch intensity (normalized to the intensity of the neuronal marker) in C4da dendrites from the indicated genotypes. (E) Plot of relative Futsch intensity along the primary dendrites of wild-type (green), dfmr1RNAi (red) and dfmr1OE (purple) C4da neurons. Lines indicate the mean values; lighter shading denotes s.d. (F) Quantification of proportion of branches that were Futsch positive. Percentages are the number of Futsch-positive branches divided by the number of total branches. All images are confocal z series projections. In D,F, each dot represents a value for one neuron. Scale bar: 50 μm. Data are mean±s.d.; *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; one-way ANOVA with Dunnett's multiple comparisons test.
Loss of FMRP increases microtubule dynamics in distal dendrites
In addition to microtubule stability in main branches, microtubule dynamics in higher-order branches plays an important role in dendrite patterning (Delandre et al., 2016). Therefore, we next sought to test whether FMRP controls microtubule dynamics by visualizing growing microtubules in vivo through live imaging of GFP-tagged EB1, a microtubule plus-end-binding protein (Stone et al., 2008; Stewart et al., 2012). EB1-GFP appears as comets at microtubule plus-ends, and the number and the speed of motile EB1-GFP comets represent the number of growing microtubules and the growth rate, respectively. We performed time-lapse live imaging of C4da neurons expressing EB1-GFP at 72 h and quantitatively analyzed EB1-GFP comets in more distal, higher-order dendrites where the comets are easily visible (Fig. 4A). Consistent with the predominantly minus-end out orientation of microtubules in C4da neuron arbors (Stone et al., 2008; Ori-McKenney et al., 2012), we found that most EB1-GFP comets moved along dendrites toward the soma in wild-type neurons, and that microtubule polarity was unaltered in both dfmr1RNAi and dfmr1OE neurons (Fig. 4B,C, Movies 1-3). However, dfmr1 RNAi led to more and faster motile EB1-GFP comets in distal dendrites (Fig. 4D-G), indicating increased microtubule dynamics. Although we were unable to quantify EB1-GFP motility in the terminal branches, our data support the idea that loss of FMRP affects the microtubule remodeling by increasing both stable microtubule levels along main branches and microtubule dynamics in higher order dendrites, thereby stabilizing dendritic arbors and promoting branch growth, respectively.
Loss of FMRP increases microtubule dynamics in distal dendrites. (A) Representative image of C4da neurons with GAL4-477-driven UAS-EB1-GFP. The region of the arbor used to quantify EB1 dynamics is indicated by the orange box. (B) Representative kymographs showing the movements of EB1-GFP comets (red line) in wild-type, dfmr1RNAi and dfmr1OE C4da neurons. (C) Quantification of percentages of EB1-GFP comets moving in anterograde or retrograde direction. (D-G) Quantification of the number (D,E) and the speed (F,G) of motile EB1-GFP comets in distal, higher-order dendrites of wild-type, dfmr1RNAi and dfmr1OE C4da neurons. Overall: EB1-GFP comets moving in either an anterograde or a retrograde direction. Retrograde: EB1-GFP comets moving in a retrograde direction. Scale bars: 5 μm and 100 s. In D-G, each dot represents a value for one neuron. Data are mean±s.d.; ns, not significant; *P<0.05, **P<0.01; one-way ANOVA with Dunnett's multiple comparisons test.
Loss of FMRP increases microtubule dynamics in distal dendrites. (A) Representative image of C4da neurons with GAL4-477-driven UAS-EB1-GFP. The region of the arbor used to quantify EB1 dynamics is indicated by the orange box. (B) Representative kymographs showing the movements of EB1-GFP comets (red line) in wild-type, dfmr1RNAi and dfmr1OE C4da neurons. (C) Quantification of percentages of EB1-GFP comets moving in anterograde or retrograde direction. (D-G) Quantification of the number (D,E) and the speed (F,G) of motile EB1-GFP comets in distal, higher-order dendrites of wild-type, dfmr1RNAi and dfmr1OE C4da neurons. Overall: EB1-GFP comets moving in either an anterograde or a retrograde direction. Retrograde: EB1-GFP comets moving in a retrograde direction. Scale bars: 5 μm and 100 s. In D-G, each dot represents a value for one neuron. Data are mean±s.d.; ns, not significant; *P<0.05, **P<0.01; one-way ANOVA with Dunnett's multiple comparisons test.
FMRP negatively regulates Chic expression in C4da neurons to control dendrite morphology
In addition to microtubules, dendrites contain actin assemblies that direct dendrite branching. In C4da neurons, actin filaments are distributed intermittently throughout the arbor and dynamic F-actin assemblies pre-localize at future branch sites to promote terminal branch outgrowth (Nithianandam and Chien, 2018). As FMRP modulates terminal branch dynamics and actin remodeling is implicated in dendrite outgrowth, we reasoned that FMRP might regulate actin polymerization to control C4da branch growth. Notably, chickadee (chic) mRNA, which encodes the Drosophila profilin homolog that promotes actin turnover by catalyzing the exchange of actin-bound ADP to ATP (Karlsson and Dráber, 2021), was identified as a FMRP-associated transcript (Reeve et al., 2005). Moreover, previous work has shown that chic is required for normal dendrite morphology (Nithianandam and Chien, 2018). We therefore tested whether FMRP regulates chic to limit dendrite outgrowth.
Immunoblotting of extracts from adult Drosophila heads revealed that Chic protein levels were significantly higher in dfmr1Δ50M homozygotes than in wild-type and heterozygous siblings (Reeve et al., 2005; Tessier and Broadie, 2008), consistent with downregulation of chic by FMRP. To determine whether FMRP regulates chic in C4da neurons, we performed anti-Chic immunofluorescence. Chic protein was detected in the soma of wild-type C4da neurons but not in dendrites (Fig. 5A,A′). In dfmr1RNAi neurons, Chic levels were elevated in the soma and Chic also became readily detectable along C4da dendrites (Fig. 5B,B′,D,E). In contrast, dfmr1 overexpression led to reduction of Chic in the soma and no detectable protein in dendrites, similar to wild-type neurons (Fig. 5C-E). These data suggest that FMRP represses Chic protein expression in C4da neurons.
FMRP negatively regulates Chic expression in C4da neurons to control dendrite morphology. (A-C) Immunofluorescence detection of Chic (A-C, green) in wild-type (A,A′), dfmr1RNAi (B,B′) and dfmr1OE (C,C′) larvae. The somas of C4da (ddaC) and C3da (ddaF) neurons are labeled with orange and blue arrowheads, respectively; C4da (ddaC) and C3da (ddaF) dendrites are labeled with orange and blue arrows, respectively. Merged images (A′-C′) of CD4-tdGFP-labeled C4da neurons (red) and Chic (green). Scale bar: 30 μm. (D,E) Quantification of relative Chic intensity in the soma and dendrites of C4da neurons from the indicated genotypes. Each dot represents a value for one neuron. (F-M) Representative images of wild-type (F), chicRNAi (G), chicOE (H), dfmr1OE (I), chicOE, dfmr1OE (J), dfmr1RNAi (K), chic221/+ (L) and chic221/+, dfmr1RNAi (M) C4da neurons labeled with CD4-tdGFP. F′-M′ are the magnified views from regions outlined by a purple dashed line. Scale bars: 50 μm. (N-P) Quantification of average terminal branch length (N), terminal branch number (O) and the dendritic field coverage ratio (P) in the image field from C4da neurons with the indicated genotypes. Each dot represents a value for one neuron. Data are mean±s.d.; ns, not significant; *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; one-way ANOVA with Tukey's multiple comparisons test. All images are confocal z series projections.
FMRP negatively regulates Chic expression in C4da neurons to control dendrite morphology. (A-C) Immunofluorescence detection of Chic (A-C, green) in wild-type (A,A′), dfmr1RNAi (B,B′) and dfmr1OE (C,C′) larvae. The somas of C4da (ddaC) and C3da (ddaF) neurons are labeled with orange and blue arrowheads, respectively; C4da (ddaC) and C3da (ddaF) dendrites are labeled with orange and blue arrows, respectively. Merged images (A′-C′) of CD4-tdGFP-labeled C4da neurons (red) and Chic (green). Scale bar: 30 μm. (D,E) Quantification of relative Chic intensity in the soma and dendrites of C4da neurons from the indicated genotypes. Each dot represents a value for one neuron. (F-M) Representative images of wild-type (F), chicRNAi (G), chicOE (H), dfmr1OE (I), chicOE, dfmr1OE (J), dfmr1RNAi (K), chic221/+ (L) and chic221/+, dfmr1RNAi (M) C4da neurons labeled with CD4-tdGFP. F′-M′ are the magnified views from regions outlined by a purple dashed line. Scale bars: 50 μm. (N-P) Quantification of average terminal branch length (N), terminal branch number (O) and the dendritic field coverage ratio (P) in the image field from C4da neurons with the indicated genotypes. Each dot represents a value for one neuron. Data are mean±s.d.; ns, not significant; *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; one-way ANOVA with Tukey's multiple comparisons test. All images are confocal z series projections.
To determine whether FMRP controls dendrite growth by regulating chic, we first evaluated how Chic levels affect C4da neuron morphology. Effective knockdown and overexpression of chic in C4da neurons were confirmed by anti-Chic immunofluorescence (Fig. S5A,B). chic RNAi did not affect the number of terminal branches (Fig. S6G), suggesting that chic is not required for branch formation per se. By contrast, loss of chic caused a significant reduction of terminal branch length and dendritic field coverage (Fig. 5F-G′ and Fig. S6E,F), whereas its overexpression in C4da neurons led to increased terminal branch length (Fig. 5F,F′,H,H′,N). Thus, the Chic level must be maintained within a certain range for proper dendrite outgrowth. Furthermore, chic RNAi resembled dfmr1 overexpression in causing a significant decrease in terminal branch length (Fig. 5G,G′,I,I′,N and Fig. S6B,E), consistent with negative regulation of chic by FMRP.
We further tested their genetic interaction by co-overexpressing FMRP and Chic in C4da neurons. Overexpression of Chic rescued the terminal branch length and dendritic field coverage defects in dfmr1OE neurons (Fig. 5F,H-J,N-P), suggesting that FMRP regulates terminal branch growth through chic. Reducing Chic levels in dfmr1 RNAi neurons using a heterozygous chic loss-of-function mutation decreased terminal branch length (Fig. 5F,K-N) and mitigated the dendritic field coverage defects (Fig. 5P) without significantly affecting terminal branch number (Fig. 5O). Similarly, knockdown of both dfmr1 and chic together, as confirmed by anti-Chic and anti-FMRP immunofluorescence (Fig. S5C,D), caused a reduction of terminal branch length resembling chicRNAi neurons (Fig. S6B,D,E) and resulted in significant phenotypic rescue of dfmr1RNAi neurons (Fig. S6A,C-G). Together, these observations provide evidence that FMRP negatively regulates chic to control dendrite morphology, particularly terminal branch growth.
KH1 and KH2 domains are crucial for FMRP-mediated dendritic regulation
FMRP contains three RNA-binding domains (RBDs): two hnRNP K homolog domains (KH1 and KH2) and one RGG box enriched in arginine and glycine (Fig. 6A). The RGG box has been known to interact with G-quadruplex-containing mRNAs for RNA localization (Darnell et al., 2001; Goering et al., 2020). Two missense mutations, I244N and I307N, within KH1 and KH2 domains, respectively, were found to abolish translational regulation and ribosomal interaction (Feng et al., 1997; Chen et al., 2014). To gain molecular insight into FMRP function in C4da neurons, we generated a set of UAS-dfmr1 transgenes with the missense mutation I244N or I307N, or with different FMRP RBDs deleted (Fig. 6A). Expression of dfmr1ΔRGG successfully rescued the dfmr1RNAi overelaboration defect, similarly to the intact dfmr1 (dfmr1FL) (Fig. 6B-E,P-S), indicating that the RGG domain is dispensable for FMRP function in C4da dendrites. Importantly, expression of dfmr1ΔKH1, dfmr1I244N, dfmr1ΔKH2 or dfmr1I307N failed to compensate for dfmr1 knockdown (Fig. 6F-I,P-S), revealing the requirement of KH1 and KH2 domains, as well as the contributions of I244 and I307, in FMRP-mediated dendrite arborization.
KH1 and KH2 domains are crucial for FMRP-mediated dendritic regulation. (A) Schematic illustration of domain structures of different FMRP variants. (B-I) Representative images of CD4-tdGFP labeled wild-type C4da neurons (B), and neurons co-expressing dfmr1RNAi in wild-type (C), dfmr1FL (D), dfmr1ΔRGG (E), dfmr1ΔKH1 (F) dfmr1I244N (G), dfmr1ΔKH2 (H) and dfmr1I307N neurons (I). Although the transgenic dfmr1 transcripts are targeted by the RNAi, their overexpression likely partially attenuates the RNAi, and vice versa, leading to a physiological level of dfmr1 expression. (J-O) Representative images of CD4-tdGFP labeled C4da neurons overexpressing dfmr1FL (J), dfmr1ΔRGG (K), dfmr1ΔKH1 (L), dfmr1I244N (M), dfmr1ΔKH2 (N) and dfmr1I307N (O). The UAS-dfmr1 transgenes used in this figure were generated by PhiC31-based integration, whereas the UAS-dfmr1OE line used in previous figures was generated by P element-mediated insertion. Scale bar: 50 μm. (P-S) Quantification of terminal branch number (P), average terminal length (Q), total dendritic length (R) and the dendritic field coverage ratio (S) from C4da neurons with the indicated genotypes. All images are confocal z series projections. Each dot represents a value for one neuron. Data are mean±s.d.; ns, not significant; *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; one-way ANOVA with Dunnett's multiple comparisons test.
KH1 and KH2 domains are crucial for FMRP-mediated dendritic regulation. (A) Schematic illustration of domain structures of different FMRP variants. (B-I) Representative images of CD4-tdGFP labeled wild-type C4da neurons (B), and neurons co-expressing dfmr1RNAi in wild-type (C), dfmr1FL (D), dfmr1ΔRGG (E), dfmr1ΔKH1 (F) dfmr1I244N (G), dfmr1ΔKH2 (H) and dfmr1I307N neurons (I). Although the transgenic dfmr1 transcripts are targeted by the RNAi, their overexpression likely partially attenuates the RNAi, and vice versa, leading to a physiological level of dfmr1 expression. (J-O) Representative images of CD4-tdGFP labeled C4da neurons overexpressing dfmr1FL (J), dfmr1ΔRGG (K), dfmr1ΔKH1 (L), dfmr1I244N (M), dfmr1ΔKH2 (N) and dfmr1I307N (O). The UAS-dfmr1 transgenes used in this figure were generated by PhiC31-based integration, whereas the UAS-dfmr1OE line used in previous figures was generated by P element-mediated insertion. Scale bar: 50 μm. (P-S) Quantification of terminal branch number (P), average terminal length (Q), total dendritic length (R) and the dendritic field coverage ratio (S) from C4da neurons with the indicated genotypes. All images are confocal z series projections. Each dot represents a value for one neuron. Data are mean±s.d.; ns, not significant; *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; one-way ANOVA with Dunnett's multiple comparisons test.
We also tested the effect of the dfmr1 mutations on FMRP function by overexpression. Consistent with the rescue analysis (Fig. 6D,E), dfmr1ΔRGG neurons resembled dfmr1FL neurons in having less dendritic coverage, shorter terminal branches and reduced total dendritic length when compared with wild-type neurons (Fig. 6J,K,P-S), while dfmr1ΔKH1, dfmr1ΔKH2 and dfmr1I244N neurons displayed a more wild-type morphology (Fig. 6L-N,P-S). Despite the inability of I307N to rescue the dfmr1RNAi phenotype (Fig. 6I,P-S), surprisingly, dfmr1I307N neurons still had overexpression defects, although less severe than those of dfmr1FL neurons (Fig. 6O,P-S). Together, these results indicate that KH1 and KH2 are each required for FMRP-mediated regulation in C4da neurons, whereas the RGG domain is dispensable.
KH2 domain is critical for FMRP granule formation
FMRP is found in neuronal transport granules and regulates granule motility and localization of mRNAs to dendrites (Antar et al., 2005; Estes et al., 2008; Dictenberg et al., 2008). Dendritic transport of FMRP was shown to impact the growth of dendritic spine protrusions (Dictenberg et al., 2008), but whether FMRP granule formation is related to its dendritic functions is not clear. As FMRP granules accumulate in C4da dendrites (Fig. S1D), we further investigated whether FMRP-mediated dendritic growth is related to its granule-forming ability. Previous studies have shown that FMRP granule formation can occur by phase separation mediated by its RGG-containing C-terminal low-complexity region (Fig. 6A) (Dictenberg et al., 2008; Tsang et al., 2019) and by RNA interactions through KH2-dependent mechanisms (Ascano et al., 2012; Siomi et al., 1994; Gingrich et al., 2018 preprint; Starke et al., 2022). To identify the RBD(s) essential for FMRP granule formation in C4da neurons, we visualized FMRP in neurons expressing different variants using anti-FMRP immunofluorescence. All FMRP variant-expressing neurons had significantly higher FMRP levels in the soma than levels in wild-type neurons (Fig. 7A-G). Under these conditions, FMRP granules were detected along dendrites of dfmr1FL, dfmr1ΔRGG and dfmr1ΔKH1 neurons (Fig. 7B-B″,C-C″,D-D″). The number of FMRP granules was unaltered in dfmr1ΔRGG neurons but was slightly decreased in dfmr1ΔKH1 neurons compared with that of dfmr1FL neurons (Fig. 7H). On the contrary, FMRP signals were largely diffuse in the dendrites of dfmr1ΔKH2 neurons (Fig. 7F-F″), with few FMRP granules detected (Fig. 7H). Although we cannot rule out the possibility that FMRP overexpression may influence granule formation, these data indicate that the KH2 domain is crucial. The finding that deletion of the KH2 domain eliminated both granule formation (Fig. 7F-F″,H) and the ability to rescue dfmr1 knockdown (Fig. 6H) and prevented the severe branching defect caused by dfmr1 overexpression (Fig. 6J,N) is consistent with the idea that granule formation is required for FMRP function. Alternatively, the KH2 domain may contribute to FMRP function in C4da neurons by means other than granule formation, such as FMRP dendritic transport.
The KH2 domain is crucial for FMRP granule formation. (A-G) Immunofluorescence detection of FMRP using an anti-FMRP antibody in wild-type C4da neurons (A-A″) and in neurons overexpressing dfmr1FL (B-B″), dfmr1ΔRGG (C-C″), dfmr1ΔKH1 (D-D″), dfmr1I244N (E-E″), dfmr1ΔKH2 (F-F″) and dfmr1I307N (G-G″). Green arrowheads indicate FMRP particles (A′-G′, green) along dendrites of C4da neurons labeled by CD4-tdGFP (A-G, red). Merged images for two channels (A″-G″). All images are confocal z series projections. Scale bar: 30 μm. (H) Quantification of the number of FMRP particles in C4da neurons from the indicated genotypes. Each dot represents a value for one neuron. Data are mean±s.d.; ns, not significant; ****P<0.0001; one-way ANOVA with Dunnett's multiple comparisons test.
The KH2 domain is crucial for FMRP granule formation. (A-G) Immunofluorescence detection of FMRP using an anti-FMRP antibody in wild-type C4da neurons (A-A″) and in neurons overexpressing dfmr1FL (B-B″), dfmr1ΔRGG (C-C″), dfmr1ΔKH1 (D-D″), dfmr1I244N (E-E″), dfmr1ΔKH2 (F-F″) and dfmr1I307N (G-G″). Green arrowheads indicate FMRP particles (A′-G′, green) along dendrites of C4da neurons labeled by CD4-tdGFP (A-G, red). Merged images for two channels (A″-G″). All images are confocal z series projections. Scale bar: 30 μm. (H) Quantification of the number of FMRP particles in C4da neurons from the indicated genotypes. Each dot represents a value for one neuron. Data are mean±s.d.; ns, not significant; ****P<0.0001; one-way ANOVA with Dunnett's multiple comparisons test.
Despite forming granules (Fig. 7D-D″), dfmr1ΔKH1 neurons did not display morphological defects (Fig. 6J,L), suggesting that granule formation alone is insufficient for FMRP function in dendrite arborization. Consistent with the weaker overexpression phenotype (Fig. 6O), dfmr1I307N neurons were still capable of forming dendritically localized FMRP granules, but the number of granules was dramatically reduced (Fig. 7G-G″). Interestingly, dfmr1I244N neurons also had few dendritic FMRP particles (Fig. 7E-E″), raising the possibility that I244N within the KH1 domain is a dominant-negative mutation that prevents KH2-mediated granule formation. Together, these results provide evidence that granule formation correlates with FMRP-mediated dendritic patterning and is mediated primarily by the KH2 domain.
FMRP is colocalized with chic mRNA and promotes chic dendritic localization
chic mRNA and FMRP were previously shown to colocalize in Drosophila cultured neurons and to coimmunoprecipitate (Reeve et al., 2005; Estes et al., 2008), suggesting that FMRP physically interacts with chic mRNA to regulate its expression. To further test this, we employed the MS2-MCP fluorescent RNA labeling system (Misra et al., 2016) together with FMRP-GFP to visualize chic mRNA and FMRP, respectively. Live imaging of C4da neurons expressing MS2-tagged chic RNA along with MCP-mCherry and FMRP-GFP showed chic mRNA (Fig. 8A,B) and FMRP (Fig. 8A′,B′) particles distributed along the dendrites. Among these, 54±8% of chic particles colocalized with FMRP and 36±8% of FMRP particles colocalized with chic (Fig. 8A-A″,B-B″). These values may underestimate the degree of colocalization due to the presence of unlabeled endogenous chic and FMRP in these neurons. Furthermore, FMRP may contribute to multiple distinct RNPs.
FMRP is colocalized with chic mRNA in promoting chic dendritic localization. (A) Representative images of C4da neurons expressing chic-MS2 mRNA, MCP-mCherry and FMRP-GFP. C4da neurons are labeled with CD4-tdGFP. chic mRNA (A, magenta arrowheads) and FMRP-GFP (A′, green arrowheads) are colocalized in dendritic granules (A″, white arrowheads). (B) Magnified view of merged signals (B″, white arrowheads) for chic mRNA (B, magenta arrowheads) and FMRP-GFP (B′, green arrowheads) granules from the indicated region (orange box in A″). Scale bar: 10 μm. (C-E) Representative images of MCP-mCherry (C-E) in CD4-tdGFP-labeled C4da neurons (C′-E′) expressing MCP-mCherry alone (C,C′, n=16 neurons), chic-MS2 together with MCP-mCherry (D,D′, n=17 neurons), and chic-MS2, MCP-mCherry and dfmr1RNAi (E,E′, n=11 neurons). MCP particles are indicated by white arrows. Scale bar: 30 μm. All images are confocal z series projections. (F) Quantification of the number of MCP particles (normalized to the field area) from C4da neurons with the indicated genotypes. Values are median; boxes range from lower to upper quartile; extended lines indicate the minimum and maximum. ns, not significant; ****P<0.0001; one-way ANOVA with Tukey's multiple comparisons test. (G) Relative chic mRNA levels normalized to rp49 in wild-type and nSyb-GAL4-driven dfmr1RNAi adult heads. Each dot represents one biological replicate consisting of three technical replicates. Data are mean±s.d.; ns, not significant; unpaired Student's t-test. (H) Immunoblotting of lysates from S2 cells transfected with the indicated 3x-Flag-tagged dfmr1 constructs with (+) or without (−) induction with CuSO4. Proteins were detected with anti-DYKDDDDK antibody. (I) RT-PCR of chic mRNA in FMRPFL immunoprecipitates. rp49 was used as a negative control. (J) Real-time PCR analysis showing the chic mRNA levels normalized to rp49 in indicated input samples compared with the control input samples. (K) Real-time PCR analysis showing the fold-enrichment of chic mRNA levels from indicated samples after immunoprecipitation with DYKDDDDK magnetic agarose. Data are mean±s.d.; ns, not significant; ****P<0.0001; one-way ANOVA with Dunnett's multiple comparisons test.
FMRP is colocalized with chic mRNA in promoting chic dendritic localization. (A) Representative images of C4da neurons expressing chic-MS2 mRNA, MCP-mCherry and FMRP-GFP. C4da neurons are labeled with CD4-tdGFP. chic mRNA (A, magenta arrowheads) and FMRP-GFP (A′, green arrowheads) are colocalized in dendritic granules (A″, white arrowheads). (B) Magnified view of merged signals (B″, white arrowheads) for chic mRNA (B, magenta arrowheads) and FMRP-GFP (B′, green arrowheads) granules from the indicated region (orange box in A″). Scale bar: 10 μm. (C-E) Representative images of MCP-mCherry (C-E) in CD4-tdGFP-labeled C4da neurons (C′-E′) expressing MCP-mCherry alone (C,C′, n=16 neurons), chic-MS2 together with MCP-mCherry (D,D′, n=17 neurons), and chic-MS2, MCP-mCherry and dfmr1RNAi (E,E′, n=11 neurons). MCP particles are indicated by white arrows. Scale bar: 30 μm. All images are confocal z series projections. (F) Quantification of the number of MCP particles (normalized to the field area) from C4da neurons with the indicated genotypes. Values are median; boxes range from lower to upper quartile; extended lines indicate the minimum and maximum. ns, not significant; ****P<0.0001; one-way ANOVA with Tukey's multiple comparisons test. (G) Relative chic mRNA levels normalized to rp49 in wild-type and nSyb-GAL4-driven dfmr1RNAi adult heads. Each dot represents one biological replicate consisting of three technical replicates. Data are mean±s.d.; ns, not significant; unpaired Student's t-test. (H) Immunoblotting of lysates from S2 cells transfected with the indicated 3x-Flag-tagged dfmr1 constructs with (+) or without (−) induction with CuSO4. Proteins were detected with anti-DYKDDDDK antibody. (I) RT-PCR of chic mRNA in FMRPFL immunoprecipitates. rp49 was used as a negative control. (J) Real-time PCR analysis showing the chic mRNA levels normalized to rp49 in indicated input samples compared with the control input samples. (K) Real-time PCR analysis showing the fold-enrichment of chic mRNA levels from indicated samples after immunoprecipitation with DYKDDDDK magnetic agarose. Data are mean±s.d.; ns, not significant; ****P<0.0001; one-way ANOVA with Dunnett's multiple comparisons test.
To determine whether FMRP is required for dendritic chic mRNA localization, we knocked down dfmr1 by RNAi in C4da neurons expressing chic-MS2 RNA and MCP-mCherry. The number of chic mRNA dendritic particles in dfmr1RNAi neurons was reduced to a level comparable with that in neurons expressing MCP-mCherry alone (Fig. 8C-F), suggesting that FMRP is crucial for dendritic localization of chic mRNA. To exclude an effect of FMRP on chic mRNA stability, we analyzed chic mRNA levels in wild type and dfmr1RNAi adult heads by qPCR. No significant difference in chic mRNA levels was observed (Fig. 8G), indicating that FMRP does not generally affect chic mRNA stability. These data, together with the analysis of Chic protein levels above, lead us to conclude that FMRP post-transcriptionally regulates chic by promoting dendritic localization of chic mRNA but inhibiting Chic protein expression.
The KH2 domain mediates association of FMRP with chic mRNA and chic repression
Our findings that FMRP is colocalized with chic mRNA in dendritic granules and that the KH2 domain is required for FMRP granule formation raise the possibility that this domain could mediate the interaction of FMRP with chic mRNA. Although it was not genetically feasible to test which FMRP domain mediates chic localization, we interrogated the set of FMRP variants for their ability to associate with chic mRNA by RNA-coimmunoprecipitation (RNA-IP). We expressed 3xFLAG-tagged FMRP or FMRP RBD deletion mutants in Drosophila S2 cells. After anti-FLAG IP, proteins were detected by anti-Flag immunoblotting (Fig. 8H) and chic mRNA was assayed by RT-PCR and RT-qPCR (Fig. 8I-K). The relative abundance of chic mRNA was unchanged in S2 cell lysates after induction of FMRP-3xFlag proteins (Fig. 8J), which is consistent with our results that chic mRNA levels were not altered by FMRP in Drosophila adult heads (Fig. 8G). In line with the RT-PCR gel analysis (Fig. 8I), chic mRNA levels were fivefold enriched in full-length FMRP immunoprecipitants over input (Fig. 8K), further confirming their association. More importantly, the enrichment was completely abolished by the loss of the KH2 domain but not significantly affected by the loss of RGG and KH1 domains (Fig. 8K), suggesting that the KH2 domain is indispensable for the interaction between FMRP and chic mRNA. The requirement for KH2 in chic binding by FMRP predicts that, in contrast to wild-type FMRP, expression of FMRP lacking KH2 in C4da neurons should not affect Chic levels. We therefore compared Chic levels in wild-type, dfmr1FL and dfmr1ΔKH2 neurons using anti-Chic immunofluorescence. Whereas Chic expression was decreased relative to wild type in dfmr1FL neurons, it was not significantly different from wild-type levels in dfmr1ΔKH2 neurons (Fig. S7), consistent with loss of FMRP binding to and regulating chic.
DISCUSSION
Previous studies have focused primarily on the function of FMRP in dendritic spine maturation and its relevance to synaptic function (Bagni and Greenough, 2005; Bassell and Warren, 2008). Here, we have investigated the role of FMRP in the development of foundational dendritic arbor architecture. Loss of FMRP in C4da neurons resulted in an overelaboration defect and elevated dendrite dynamics compared with wild-type neurons, suggesting that FMRP prevents excessive branching of dendritic arbors. As C4da neurons do not receive synaptic connections, our results reveal a primary role for FMRP in dendrite branching morphogenesis that is independent of any effect on synaptic activity. This regulation of multidendritic architecture by FMRP may also apply to neurons with complex dendritic patterns in the central nervous system. We also found that loss or gain of FMRP specifically in C4da neurons affects the larval nocifensive response to noxious thermal stimuli. Given the influence of dendritic architecture on neuronal function, the morphological defects in C4da neurons lacking FMRP may compromise their sensory function. Individuals with FXS, and Fmr1 knockout mice and rats exhibit sensory hypersensitivity and abnormal sensory processing (Rais et al., 2018). Our results raise the possibility that defects in peripheral sensory neuron development and/or function could also contribute to FXS phenotypes.
Cytoskeletal remodeling plays a key role in dendrite arborization (Hummel et al., 2000; Roos et al., 2000; Stone et al., 2008; Delandre et al., 2016; Jenkins et al., 2017; Nithianandam and Chien, 2018). Our findings support a model in which FMRP regulates dendritic arbor morphology through effects on both microtubule and actin polymerization. Results from both dfmr1 knockdown and overexpression experiments suggest that FMRP modulates microtubule stability in main branches and microtubule dynamics in distal, higher-order branches, generating the requisite balance for proper arbor complexity. Although futsch mRNA was identified as a target of FMRP in synaptic boutons (Zhang et al., 2001), knockdown of futsch in C4da neurons results in excess rather than fewer dendrites relative to the wild type (Jenkins et al., 2017). Furthermore, removing one copy of futsch the dfmr1 RNAi context did not rescue the dfmr1 phenotype (data not shown). Thus, it seems likely that the elevated Futsch levels in dfmr1RNAi neurons is a consequence rather than a cause of the overelaboration defect. Given that microtubule remodeling is controlled by a variety of factors (Kapitein and Hoogenraad, 2015; Delandre et al., 2016), the impact of FMRP on this process may be an integrative effect of its regulation of multiple targets involved in microtubule nucleation and/or stabilization, such as tau (Kobayashi et al., 2017), spastin (Yao et al., 2011) and tppp (ringer) (Darnell et al., 2011; Ascano et al., 2012; Maurin et al., 2018).
Additionally, our results suggest that FMRP regulates actin polymerization through its interaction with and post-transcriptional regulation of chic mRNA. The appropriate Chic level maintained by FMRP could be crucial for actin assembly-based terminal branch growth (Nithianandam and Chien, 2018). Loss of chic affects terminal branch length but not branch number, suggesting that chic is required for branch elongation rather than formation. This is consistent with evidence that profilin enhances fluctuations in the length of actin filaments to promote polarized growth of actin filaments (Pernier et al., 2016). Branch number is affected by loss or overexpression of FMRP, however, indicating that FMRP must control additional targets, such as those encoding regulators of actin nucleation, localization or dynamics. Notably, mRNA encoding the GTPase Rac1, a regulator of actin dynamics, has been suggested to be a potential target of FMRP in C4da neurons (Lee et al., 2003).
Previous analysis of FMRP target RNAs found little overlap between those whose localization to neurites depends on FMRP and those whose translation is regulated by FMRP (Goering et al., 2020), suggesting that FMRP has two distinct and independent regulatory modes. By contrast, we find that FMRP forms dendritic granules that contain chic mRNA and downregulates Chic expression, suggesting that, in this case, localization and translational control are coupled by FMRP. RNA granules are often associated with translational repression, and studies in cultured neurons have shown that dendritically localized RNA granules store translationally silenced RNAs that, upon synaptic stimulation, can be shuttled to spines, and released and translated (Martin and Ephrussi, 2009; Glock et al., 2017). Such local translation plays an important role in synaptic plasticity and remodeling. Similarly, FMRP granules may modulate the spatiotemporal dynamics of Chic protein production, such that it is translated only at specific dendritic sites where and when Chic is needed for branch outgrowth. In this scenario, transient local signals coming from the overlying epidermal cells or extracellular matrix could release chic from granules and FMRP-mediated repression. Alternatively, FMRP may tune chic translation to a basal level throughout the dendrite that is commensurate with needs during arbor development. These potential functions of FMRP in chic mRNA granule distribution and/or translational repression may be modulated by its post-translational modifications, as phosphorylation status of FMRP is crucial for its regulation of condensate assembly (Tsang et al., 2019; Kim et al., 2019) and translation (Ceman et al., 2003; Coffee et al., 2012).
FMRP has three RBDs, different combinations of which may confer distinct functionalities. The RGG domain was previously shown to mediate RNA localization through its interactions with transcripts containing G-quadruplex structures, whereas the KH domains regulate translation (Darnell et al., 2001; Chen et al., 2014; Goering et al., 2020). We find, however, that the RGG is dispensable for FMRP-mediated dendritic patterning and granule formation in C4da neurons. Instead, our findings that FMRP is colocalized with chic mRNA in dendritic granules, that FMRP granule formation is primarily mediated by the KH2 domain, that chic mRNA coimmunoprecipitates with FMRP in S2 cells in a KH2-dependent manner, and that KH2 is required for downregulation of Chic together indicate that KH2 domain is crucial for RNA interaction, transport and repression of chic in C4da dendrite morphogenesis. Consistent with this, the I307N mutation in KH2, which impacts FMRP RNA binding through ACUK RNA recognition elements (Siomi et al., 1994; Ascano et al., 2012), compromised FMRP granule formation.
Unlike the KH2 domain, the KH1 domain is crucial for FMRP-mediated dendritic regulation but is dispensable for FMRP-mediated granule formation. Previous studies have shown that the I244N mutation within KH1 domain abolishes ribosomal interaction and, in turn, translational repression by FMRP (Feng et al., 1997; Chen et al., 2014). Thus, the KH1 domain may similarly confer the translational repression function of FMRP in C4da neurons. Whether this repression occurs through the interaction of FMRP with ribosomes or by other mechanisms associated with FMRP, including recruitment of the miRNA-induced silencing complex (Ishizuka et al., 2002; Xu et al., 2008; Yang et al., 2009) and/or other post-transcriptional regulators (Napoli et al., 2008; Bienkowski et al., 2017; Kim et al., 2019), is an interesting question for future studies.
MATERIALS AND METHODS
Fly strains
The following transgenic stocks were used: FRT82B dfmr1Δ50M (Tessier and Broadie, 2008); ppk-GAL4, UAS-CD4::tdGFP (Bhogal et al., 2016); UAS-dFmr1-GFP (Lee et al., 2003); UAS-dfmr1-RNAi (TRiP HMS00248; Bloomington stock 34944); UAS-Fmr1.Z (Bloomington stock 6931); UAS-chic-RNAi (TRiP HMS00550; Bloomington stock 34523); nSyb-gal4 (Bloomington stock 51635); UAS-fmr1-RNAi (VDRC 110800 KK); and Gal4-477, UAS-EB1-GFP (Stewart et al., 2012). ppk-GAL4, UAS-CD4::tdTom was generated by recombination between ppk-GAL4 (Bloomington stock 32079) and UAS-CD4::tdTom (Bloomington stock 35837). ppk-GAL4 was used to drive expression of UAS transgenes specifically in C4da neurons with the exception of UAS-EB1-GFP, which was expressed using GAL4-477. To enhance GAL4/UAS efficiency, the experiments using UAS-RNAi lines were performed at 29°C. All other crosses were performed at 25°C.
Plasmid construction
UAS-dfmr1 transgenes
Full-length dfmr1 cDNA was amplified from EST clone LD09557 (DRGC #1297175) using Fwd_EcoRI_LD09557 (5′-GATCCTGAATTCCACGAGGCGAAAGTGTG-3′) and Rev_XbaI_LD09557 (5′-GGCTATTCTAGATTTTTTTTTTTTTTTTTTGCAATTATATCATTCTCG-3′) primers containing EcoRI and XbaI sites, respectively. PCR products were digested with EcoRI and XbaI, and then inserted into pattB-UASt digested with EcoRI and XbaI to generate pUASt-dfmr1-FL. The deletions removing different FMRP RBDs and the missense mutation I244N were generated by ligating PCR products amplified from pUASt-dfmr1-FL with the following pairs of primers using Q5 site-directed mutagenesis kit (NEB E0554S): Fwd_KH1_del (5′-GAATACGCCGAGGAGTTC-3′) and Rev_KH1_del (5′-TCCACGGCTCATCAGTTT-3′); Fwd_KH2_del (5′-GTGCTGTTGGAGTATCATCTGTCGC-3′) and Rev_KH2_del (5′-TTCGAGCATCGCGCGGGC-3′); Fwd_RGG_del (5′-CGGCCGCCACGCAACGAT-3′) and Rev_RGG_del (5′-ATCGTTGTAGCCACGCTGCTG-3′); and Fwd_I244N (5′-GGCTCGAATAATCAAGCGGCAC-3′) and Rev_I244N (5′-ATGCGAGCCAATCGCCAA-3′). The missense mutation I307N was generated by amplifying PCR products from pUASt-dfmr1-FL with Fwd_I307N (5′-GGGCGCATTAACCAGGAGATT-3′) and Rev_I307N (5′-AATCTCCTGGTTAATGCGCCC-3′) using PfuTurbo DNA polymerase (Agilent, 600250).
UAS-chicCDS-MS2-chic3′UTR transgene
Chic coding and 3′UTR sequences were amplified from pUASt-Flag-chicCDS-chic3′UTR (Medioni et al., 2014) with Fwd_XhoI_chicCDS (5′-GATCCTCTCGAGATGAGCTGGCAAGATTATG-3′) and Rev_Esp3I_chicCDS (5′-CGTACGTCTCAGGTTCTAGTACCCGCAAGTAATC-3′), and Fwd_Esp3I_chic3utr (5′-CGTACGTCTCATCATGAGAATAGATCAACAC-3′) and Rev_XbaI_chic3utr (5′-GGCTATTCTAGACGTGTGGATTTATGTACG-3′), respectively. The 24xMS2v5 fragment was amplified from pUC57_24xMS2V5 (Wu et al., 2015) with Fwd_Esp3I_24xMS2v5 (5′-CGTACGTCTCTAACCTACAAACGGGTGGAG-3′) and Rev_Esp3I_24xMS2v5 (5′-CGTACGTCTCTATGAGATCTGAGGTGTTTG-3′). The three PCR products were digested, respectively, with the indicated enzymes and ligated together with XhoI- and XbaI-digested pattB-UASt to produce pUASt-chicCDS-24xMS2-chic3′UTR.
UAS-MCP-mCherry transgene
The UAS-MCP-mCherry transgene is identical to UAS-MCP-RFP (Brechbiel and Gavis, 2008), except that a NotI fragment containing the EGFP-coding sequence was replaced by a PCR-generated NotI fragment containing the mCherry-coding sequences (Shaner et al., 2004). The transgene was introduced into y w67c23 embryos by P element-mediated germline transformation.
pMT-dfmr1-3xFlag constructs
The full-length dfmr1-coding sequence was amplified from EST clone LD09557 (Drosophila Genomics Resource Center 1297175) using Fwd_KpnI_dfmr1-CDS (5′-GATCCTGGTACCGCCGCCATGGAAGATCTCCTCGTGGAAG-3′) and Rev_BamHI_dfmr1-CDS (5′-GGCTATGGATCCGGACGTGCCATTGACCAGGCC-3′), and was digested with KpnI and BamHI. The pMT backbone was amplified from pMT-Hrp48-ADARcd-V5 (Addgene 81172) using Fwd_AgeI_pMT (5′-GGCTATACCGGTCTGATCAGCCTCGACTGTG-3′) and Rev_KpnI_pMT (5′-CTTCGAGGTACCCCGATCTAGATC-3′), and was digested with AgeI and KpnI. A Ty1-3xFlag gBlock Gene Fragment (5′-GGTCAATGGCACGTCCGGATCCGAAGTGCATACCAATCAGGACCCGCTGGACGACTACAAAGACCATGACGGTGATTATAAAGATCATGACATCGATTACAAGGATGACGATGACAAGTAGACCGGTCTGATCAGCCTCGACTGTG-3′) synthesized by Integrated DNA Technologies was digested with BamHI and AgeI. The three fragments were ligated together to generate pMT-dfmr1-FL-3xFlag. DNA fragments with deletions of FMRP RBDs were amplified respectively from pUASt-dfmr1-RGG/KH1/KH2 using Fwd_KpnI_dfmr1-CDS (5′-GATCCTGGTACCGCCGCCATGGAAGATCTCCTCGTGGAAG-3′) and Rev_BamHI_dfmr1-CDS (5′-GGCTATGGATCCGGACGTGCCATTGACCAGGCC-3′) primers, digested with KpnI and BamHI, and inserted into pMT-dfmr1-FL-3xFlag digested with KpnI and BamHI.
S2 cell culture and transfection
Drosophila S2 cells were cultured at 25°C in Schneider's Drosophila Medium (Gibco, 21720024) with 10% fetal bovine serum (Gibco, 10437010) and 1% penicillin/streptomycin (Sigma, P4333). S2 cells were transfected with pMT-dfmr1-3xFlag constructs using Effectene Transfection Reagent (Qiagen, 301425) in six-well plates. 24 h post transfection, expression of different FMRP variants was induced by 0.5 mM CuSO4 (Sigma, 209198-100G). Uninduced S2 cells were used as controls. Cells were harvested 24 h post induction, lysed in buffer containing 10 mM Tris HCl (pH 7.5), 150 mM NaCl, 0.5 mM EDTA (pH 8.0), 0.5% IGEPAL CA-630 (Sigma, I8896), 1× protease inhibitor cocktail (Roche, 11873580001) and RNase inhibitor (NEB #M0307L), and cleared by centrifugation at 16,100 g for 25 min at 4°C.
RNA immunoprecipitation and RT-qPCR
Uninduced and induced cleared S2 cell lysates were incubated with equilibrated anti-DYKDDDDK magnetic agarose beads (Thermo Scientific, A36797) on a nutator for 20 min at room temperature. 25% of each immunoprecipitation sample was removed for immunoblotting and the remainder was used for RNA extraction and RT-qPCR analysis. FMRP-3xFlag proteins were detected using 1:2000 anti-DYKDDDDK tag monoclonal antibody (Invitrogen, MA1-91878) and 1:2000 HRP sheep anti-mouse antibody (VWR, 95017-332). Total RNA from S2 cell lysates or IP samples was extracted using TRIzol (Invitrogen) and reverse transcribed with SuperScript III First-Strand Synthesis System (Invitrogen, 18080051) and oligo(dT)20 primer (Invitrogen, 18418020). For the −RT control, reverse transcriptase was replaced by nuclease-free water. qPCR was performed using SYBR Green PCR Master Mix (Thermo Fisher, 4309155) and StepOnePlus Real-Time PCR System (Applied Biosystems), according to the manufacturer's protocol. The following primers were used: chic_Fwd (5′-TGCACTGCATGAAGACAACA-3′) and chic_Rev (5′-GTTTCTCTACCACGGAAGCG-3′); and rp49_Fwd (5′-CGGATCGATATGCTAAGCTGT-3′) and rp49_Rev (5′-GCGCTTGTTCGATCCGTA-3′). To prepare Drosophila adult heads for qPCR, heads expressing wild-type or UAS-dfmr1-RNAi by nSyb-GAL4 were frozen in liquid nitrogen and total RNA was extracted as described above.
Immunofluorescence
Late 3rd instar larva preparation and staining were performed as previously described (Bhogal et al., 2016). For better immunostaining efficiency, the larval body wall muscles were removed as described previously (Tenenbaum and Gavis, 2016). FMRP expression was detected using anti-FMRP monoclonal antibody (1:100, Abcam, ab10299). Stable microtubules were visualized using anti-Futsch monoclonal antibody (1:20, DSHB, 22C10) and anti-acetylated α-tubulin antibody (1:500, Sigma-Aldrich, T6793). The surfaces function in Imaris (Oxford Instruments) was used to mask non-ddaC signals to better visualize Futsch and acetylated tubulin labeling within ddaC neurons in Fig. 3 and Fig. S4, respectively. Chic levels were detected using anti-Chic monoclonal antibody (1:2, DSHB chic 1J). AlexaFluor 568 goat anti-mouse (1:500, Life Technologies, A-11004) was used as the secondary antibody. All antibodies were incubated in blocking buffer containing PBS/0.3% TritonX-100 with 5% normal goat serum (NGS) either overnight at 4°C (primary antibodies) or for 1.5 h at room temperature (secondary antibodies).
Microscopy
Confocal imaging of fixed and live samples was performed on a Leica SP5 laser scanning confocal microscope (40×/1.25 NA oil objective or 63×/1.3 NA glycerol objective). For consistency, class IV ddaC neurons from abdominal segments A2-A5 were imaged. For confocal imaging of live samples, larvae were mounted individually in 80% glycerol between a slide and a coverslip. To perform branch dynamic analysis, larvae at 72 h AEL were imaged using a Leica SP5 40×/1.25 NA oil objective and returned to juice agar plates. The same neurons were re-imaged after 30 min for further quantification.
Time-lapse EB1 imaging
For real time analysis of EB1 dynamics, EB1-GFP comets were captured using a Leica SP5 63×/1.3 NA GLY objective. One ddaC neuron per larva at 72 h AEL was live imaged every 8 s for 160 s. Videos were analyzed with ImageJ software (https://imagej.nih.gov/ij/). Movement of EB1-GFP comets were tracked by drawing lines (width=3) along dendrites. Kymographs were generated with ImageJ built-in Multi Kymograph function. The number of EB1 comets and proportions of EB1 polarity were quantified manually from videos and further confirmed by kymographs. The speed of EB1 comets was analyzed by measuring angles of EB1 tracks from kymographs and calculating cotangent (angle) for distance over time.
Analysis of dendrite morphology
C4da (ddaC) neurons from live larvae were labeled using ppk-GAL4, UAS-CD4::tdGFP and were imaged by a Leica SP5 40×/1.25 NA GLY objective. All images are confocal z-series projections. At least seven neurons from four or more larvae were imaged and analyzed for each genotype. Quantitative analysis of dendrite morphology was performed with ImageJ software. Total dendritic length and terminal branch length were analyzed by tracing neurons using the NeuronJ plug-in. The field coverage of dendritic arbors was quantified by overlaying a grid of 20×20 squares on the image of interest and counting the number of empty squares. Sholl analysis was performed using ImageJ built-in Sholl function.
Quantification of fluorescence intensity
Images with maximum projections were quantified in ImageJ. For quantification of Futsch and acetylated tubulin intensity, intensity values were measured by ROI tracing of dendrites. The normalized intensity was calculated by dividing background subtracted Futsch or acetylated tubulin intensity to background subtracted membrane intensity. The plot of Futsch intensity along dendrites was generated with the ImageJ built-in Plot Profile function. For quantification of Chic levels, intensity values were measured by ROI tracing and background was subtracted.
Quantification of dendritic granules
For quantification of MS2-tagged transcripts, at least ten neurons from five or more larvae were imaged using a Leica SP5 63×/1.3 NA glycerol objective. The MCP channel was thresholded automatically. For quantification of dendrite-localized RNA granules, MCP signals from the axon were removed before particle detection. chic mRNA particles were counted with ImageJ built-in Analyze Particles function. Particle identification parameters were set as follows: particle size, 0-20 μm2; particle circularity, 0.00-1.00. Quantification of FMRP granules was carried out manually.
Global heat-plate assay
The global heat-plate assay was performed on mid-3rd instar larvae as previously described (Chattopadhyay et al., 2012). Briefly, an individual larva was deposited in an 80 μl water droplet on a 60×15 mm Petri dish, which was then placed on a heat block pre-equilibrated to 95°C. The complete sequence of locomotory behavioral responses (head thrash, roll, whip, seizure and paralysis) was video-recorded. The time elapsed from the start of the treatment to onset of each behavior (latency) was measured.
Statistical analysis and plotting
All data were analyzed and plotted using GraphPad Prism 9 (https://www.graphpad.com/). Comparisons between two groups and three or more groups were performed with an unpaired Student's t-test and one-way ANOVA with Dunnett's or Tukey's multiple comparisons test, respectively. Values are mean±s.d.; ns, not significant; *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.
Acknowledgements
We thank the Bloomington Drosophila Stock Center, the Vienna Drosophila Resource Center, K. Broadie, F.-B. Gao and M. Rolls for fly stocks, and the Drosophila Genomics Resource Center, F. Besse and H. Garcia for plasmids. We thank the Princeton Confocal Imaging Facility, a Nikon Center of Excellence in the Department of Molecular Biology, for assistance with confocal imaging, and J. Yan, A. Hakes and Y. Peng for comments on the manuscript.
Footnotes
Author contributions
Conceptualization: H.L., E.R.G.; Methodology: H.L.; Validation: H.L.; Formal analysis: H.L.; Investigation: H.L.; Data curation: H.L.; Writing - original draft: H.L.; Writing - review & editing: H.L., E.R.G.; Supervision: E.R.G.; Project administration: E.R.G.; Funding acquisition: E.R.G.
Funding
H.L. is supported by a pre-doctoral fellowship jointly funded by the China Scholarship Council and Princeton University.
Peer review history
The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200379.
References
Competing interests
The authors declare no competing or financial interests.