RAB39B-mediated trafficking of the GluA2-AMPAR subunit controls dendritic spine maturation and intellectual disability-related behaviour

Abstract

Mutations in the RAB39B gene cause X-linked intellectual disability (XLID), comorbid with autism spectrum disorders or early Parkinson's disease. One of the functions of the neuronal small GTPase RAB39B is to drive GluA2/GluA3 α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) maturation and trafficking, determining AMPAR subunit composition at glutamatergic postsynaptic neuronal terminals. Taking advantage of the Rab39b knockout murine model, we show that a lack of RAB39B affects neuronal dendritic spine refinement, prompting a more Ca²⁺-permeable and excitable synaptic network, which correlates with an immature spine arrangement and behavioural and cognitive alterations in adult mice. The persistence of immature circuits is triggered by increased hypermobility of the spine, which is restored by the Ca²⁺-permeable AMPAR antagonist NASPM. Together, these data confirm that RAB39B controls AMPAR trafficking, which in turn plays a pivotal role in neuronal dendritic spine remodelling and that targeting Ca²⁺-permeable AMPARs may highlight future pharmaceutical interventions for RAB39B-associated disease conditions.

Fig. 1. Generation of a Rab39b KO mouse model.

a Scheme of the Rab39b locus: exons are black boxes spaced from one intron; 5′- and 3′-UTRs are white boxes. Enlargement of the Rab39b sequence to highlight the target sequence of the MM0000573713 guide (upper), 14 bp deletion (middle; −14 bp) and 10 bp insertion (bottom; +10 bp). b Expression profile of Rab39b transcripts in Rab39b WT, KO −14 bp and KO +10 bp brain lysates (n = 3). Data are expressed as Rab39b expression normalised to histone H3 (2^{-ΔCt(Rab39b–H3)}). c Representative western blots showing RAB39B protein expression in Rab39b WT, KO −14 bp and KO +10 bp brain lysates (n = 3). Calnexin is the loading reference. d Body weight in grams (g) of Rab39b WT (n = 10) and KO (n = 11) littermate mice. e Representative images of 20- and 90-day-old Rab39b WT and KO littermate mice. The scale bar is 2 cm. f Representative MRI scan figures showing abdominal axial “pure fat” obtained by image subtraction (T1 weighted—T1 weighted with fat saturation); after subtraction, the signal of non-fat tissue becomes null. The following main anatomic landmarks were reported: 1-bowel, 2-kidneys, 3-spine and vertebral muscles. The scale bar is 0.5 cm. Graph indicates pure fat in mm³. g Quantification of the ratio between mature (M) and immature (IM) forms of AMPAR subunits after EndoHf digestion. Lower panels show representative western blots for GluA1 (WT = 6, KO = 6, KO + CherryRab39b = 3 independent lysates), GluA2 (WT = 7, KO = 6, KO + CherryRab39b = 3 independent lysates) and GluA3 (WT = 6, KO = 6, KO + CherryRab39b = 3 independent lysates) after PNGasef (P) or EndoHf (E) digestion; ND: nondigested neurons. h Quantification of the neuronal surface expression density of GluA1 (WT = 8, KO = 12 images), GluA2 (WT = 17, KO = 12 images) and GluA3 (WT = 15, KO = 16 images), expressed in the number of positive puncta/µm² and representative images. The number of images belongs to a minimum of three experimental replicates. All the data are expressed as the mean ± SEM. The scale bar is 50 µm. *p < 0.05; **p < 0.01; ***p < 0.001.

Fig. 2. RAB39B-driven trafficking controls spine refinement.

a Timeline of hippocampal dendritic spine development: spinogenesis at 20 days (P), synaptic pruning at P30 and spine maintenance at P90. Representative images of Golgi-impregnated CA1 hippocampal apical dendrites of Rab39b WT and KO mice (n = 3). The scale bar is 10 µm. b Spine density in number of spine heads/µm. c Partitioning of spine morphology along a dendrite in % of type of spine/dendrite. d Representative fluorescence images of Rab39b WT and KO primary hippocampal neurons fixed at 14 DIV and/or 7 DIV and Rab39b KO neurons transduced with CherryRab39b or transfected with CherryGluA2 (WT 7 DIV = 14; WT 14 DIV = 18; KO 14 DIV = 18; KO 14 DIV + CherryRab39b = 18; KO 14 DIV + CherryGluA2 = 11 dendrites from three independent preparations). The scale bar is 5 µm. e Spine density as the number of spine heads/µm. f Spine morphology as the type of spine/µm. g Representative images of primary hippocampal neuronal dendrites and quantification of the neuronal surface expression density of GluA2 as the number of positive puncta/µm² of cultured GFP-transduced Rab39b WT (n = 9 images), KO (n = 9 images), KO transduced with CherryRab39b (n = 13 images) or transfected with CherryGluA2 (n = 8 images) and WT transfected with CherryGluA2 (n = 13 images). Images are from at least three independent experiments. The scale bar is 50 µm. h Representative live imaging frames of primary hippocampal neuronal dendrites of GFP-transduced Rab39b WT 7 DIV (n = 7) and 14 DIV (n = 8), KO 14 DIV (n = 7), KO transduced with CherryRab39b (n = 9) and KO after NASPM treatment (n = 9). n dendrites are from at least three independent experiments. Examples of spine behaviour: stable spines in cyan and transient spines in magenta arrows. The scale bar is 10 µm. i Spine density expressed as spine-head/30 µm dendritic length. j % of stable or transient spines/dendrite. k Dynamic rate as the pixel size variance within the stable spines. Data are the mean ± SEM except for (k), where the box-central mark is the median, the 25th and 75th percentiles are bottom and top edges, and ‘+’ are outliers. *p < 0.05, **p < 0.01, ***p < 0.001.

Fig. 3. The lack of RAB39B impairs neuronal function.

a Diagram of a patched region. b Example traces of pyramidal neuron excitability. c Summary data of the number of APs in response to 50 pA current steps. d Scatter plot of the resting membrane potential. e Scatter plot of the input resistance. f Diagram of a patched and stimulation region. g Example traces of mEPSCs. h Scatter plot of mEPSC amplitude. i Scatter plot of mEPSC frequency. j Scatter plot of mEPSC decay time. k Example traces of AMPA current decay time. l AMPAR-EPSCs example traces (−60, 0, and 40 mV). m Scatter plot of the rectification index recorded from cortical pyramidal neurons. n AMPAR/NMDA example traces at +40 mV. o Scatter plot of the AMPA/NMDA ratio recorded from cortical pyramidal neurons. p Scatter plot of AMPA currents sensitive to NASPM. q AMPA EPSC example traces before and after NASPM bath application. r Time course of NMDA EPSCs in the presence of ifenprodil (WT mice in grey, Rab39b KO mice in red). s NMDA EPSC example trace before and after ifenprodil bath application. t Scatter plot of ifenprodil inhibition calculated as % of baseline between 30 and 40 min after ifenprodil bath application. u Scatter plot of NMDA EPSC decay time before ifenprodil bath application. All the data are expressed as the mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001. For each experiment 3 KO and 3 WT mice were used.

Fig. 4. Increase activity and curiosity towards novelty in Rab39b KO.

Comparison of (a, c, e, g, and i) emergence (WT = 20, KO = 21 mice) and (b, d, f, h, and j) novelty (WT = 12, KO = 14 mice) tests for several variables. a, b Distance travelled in metres. c, d Velocity in metres/second. e, f % of time spent in the three motion states: resting, scanning and progressing. g, h % of time in exploration (E), transition (T) and home (H) zones, highlighted in black in the pictograms below the graphs. i % of time spent in the home box (WT = 8, KO = 7). j Distance to the object while in the exploration zone in centimetres. Data are presented as the mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001.

Fig. 5. Rab39b KO mice are impaired in working and associative memory.

a Pictogram of the 8-arm radial maze apparatus. Graphs indicate (b) the number or errors and (c) the number of correct arm choices before the first error during the 8-arm radial maze test (WT = 18, KO = 25 mice). The dashed line is the chance level performance of 5.5 correct successive arm visits. d Pictogram of the spontaneous alternation apparatus. Graphs show (e) the number of visits and (f) the % of correct alternations in the spontaneous alternation test (WT = 14, KO = 15 mice). g–i Standard and (j–l) trace fear conditioning protocols (standard: WT = 10, KO = 12; trace: WT = 19, KO = 24 mice). g, j % of freezing elicited by repeated exposure to the conditioned stimulus (CS) during the training session. BL indicates the baseline. Pictograms on the graphs represent (g) the standard fear conditioning protocol where the 15 s tone (CS, black box) is superimposed with a foot shock (lightning) for the last 2 s and (j) the trace fear conditioning protocol where the CS and the foot shock are separated by a 15 s trace (white box). (h, i, k, l) % of freezing during the memory test, 24 h after training, (h, k) for context, and (i, l) for tone. Data are presented as the mean ± SEM. *p < 0.05, ***p < 0.001.