The HERC1 ubiquitin ligase regulates presynaptic membrane dynamics of central synapses

HERC1 is a ubiquitin ligase protein, which, when mutated, induces several malformations and intellectual disability in humans. The animal model of HERC1 mutation is the mouse tambaleante characterized by: (1) overproduction of the protein; (2) cerebellar Purkinje cells death by autophagy; (3) dysregulation of autophagy in spinal cord motor neurons, and CA3 and neocortical pyramidal neurons; (4) impairment of associative learning, linked to altered spinogenesis and absence of LTP in the lateral amygdala; and, (5) motor impairment due to delayed action potential transmission, decrease synaptic transmission efficiency and altered myelination in the peripheral nervous system. To investigate the putative role of HERC1 in the presynaptic dynamics we have performed a series of experiments in cultured tambaleante hippocampal neurons by using transmission electron microscopy, FM1-43 destaining and immunocytochemistry. Our results show: (1) a decrease in the number of synaptic vesicles; (2) reduced active zones; (3) less clathrin immunoreactivity and less presynaptic endings over the hippocampal main dendritic trees; which contrast with (4) a greater number of endosomes and autophagosomes in the presynaptic endings of the tambaleante neurons relative to control ones. Altogether these results show an important role of HERC1 in the regulation of presynaptic membrane dynamics.

www.nature.com/scientificreports/ These data together strongly suggest that HERC1 mutation of tbl mouse might alter the normal dynamic of excitatory presynaptic terminals. Furthermore, CLT mediated endocytosis (CME) is a key step for synaptic vesicle recycling 18 ; thus, alterations of HERC1-CLT interaction 1,17 might alter the normal CLT cycle interfering with the normal synaptic function. Therefore, to elucidate the putative role of HERC1 in the synaptic vesicle populations of central excitatory synapses and in their presynaptic dynamics, we have analyzed tbl hippocampal neuronal cultures in vitro by using transmission electron microscopy, immunocytochemical, GFP pull-down and FM1-43 destaining methods.

Results
In present experiments those synapses of control (Fig. 1A) and tbl (Fig. 1B) hippocampal cultures showing a clear synaptic cleft, and evident thickening of pre-and postsynaptic zones were only considered for vesicle counts and active zone length measurement. The number of round and clear synaptic vesicles counted and the active zone length were significantly fewer (Fig. 1C) and shorter (Fig. 1D) in tbl synapses relative to control ones. However, significant differences were neither found in the mean diameter of the synaptic vesicles (Fig. 1E,∅) nor in the intervesicular distance (Fig. 1E) between control and tbl synapses. Furthermore, there was no statistically significant differences in the number of tethered synaptic vesicles (Fig. 1G); and, although the numbers of vesicles located in the nearest (< 75 nm) and farthest (225-300 nm) compartments of tbl synapses were fewer than in control ones, their values were not significant (p = 0.0853 and p = 0.2318 respectively) (Fig. 1H). However, a clear significant statistical difference in the number of docked vesicles was found, whose number was ever higher in control synapses relative to in tbl ones -as much in absolute values (Fig. 1F, AZ) as in normalized values relative the mean value of tbl active zone values (Fig. 1F, 500 nm AZ).
To quantify synaptic vesicles pools we use the fluorescence lipophilic dye FM1-43, that label synaptic boutons in an activity-dependent manner that involves dye uptake by synaptic vesicles that are recycling 19 . Hippocampal cultures of control and tbl mutants were loaded with FM1-43 dye after the application of 600 pulses ( Fig. 2A 20 . The application of 700 consecutive pulses, sufficient to deplete the reserve pool, showed also a significant difference between the control and the tbl mutant hippocampal cultured neurons (238.9 ± 108 vs. 122 ± 51.3; n = 100 and n = 75 boutons, p = 0.007).
Cultured tbl hip 21 pocampal neurons seem not show evident qualitative differences in size relative to control ones. The use of microtubule associated protein 2 (MAP2) as neuronal dendritic marker allowed us to analyze the neuronal expression of CLT ( Fig. 3A-B). Control hippocampal neurons expressed strongly both markers relative to tbl ones; and the quantitative analyses demonstrated the existence of significant differences in the absolute ( Fig. 3C and E) and normalized values relative to the analyzed area ( Fig. 3D and F) of MAP2 and CLT expression. As CLT plays a key role in the synaptic vesicle endocytic recycling 22 , we have analyzed the immunocytochemical expression of CLT together to the presynaptic SV2B protein as general marker of synaptic vesicles ( Fig. 4A-B). The expression of both markers was significantly higher in control cultured neurons than in tbl ones ( Fig. 4C-F). Moreover, the counts of presynaptic boutons expressing SV2B alone or co-expressing with CLT ( Fig. 5A-B) clearly indicated that the number of presynaptic boutons lying over the main dendrites of control hippocampal neurons analyzed here was higher than those ending on the main dendrites of tbl hippocampal neurons (Fig. 5C, T); this significant difference in the number of synaptic boutons is consistent from both SV2B single (Fig. 5C, SV2B) and double labelled boutons (Fig. 5C). When these numbers were normalized regarding their density relative to each µm of dendritic shaft measured, difference clearly ratifies that the density of synaptic boutons ending on the main dendrite of tbl hippocampal neurons was the half in relation to the control (Fig. 5E). The reduction in the density of presynaptic boutons is also correlated with a significant decrease in the immunolabelling of synaptotagmin 1 (SYT1), a presynaptic ubiquitous protein 22 , in tbl cultured neurons in relation to control cultured neurons (Fig. 6D).
In our hands the commercial antibody used here against the HERC1 in hippocampal cultures (Figs. 6 and 7) shows that the expression of this protein is distributed through the cell nucleus and the cytoplasm (Figs. 6A-B and 7A-B). In all analyzed cultures, HERC1 expression was significantly more intense in tbl hippocampal neurons than in controls (Figs. 6C and 7C).
Apart the differences in the number of rounded vesicles, an also clear qualitative difference on the size of the endosomes was observed between tbl and control presynaptic endings (Fig. 8A to D). From seminal works of Heuser and Reese 23 the endosomes located in the presynaptic endings have been related with the synaptic vesicles recycling. In present observations, quantitative analysis of the main components of the CLT recycling pathway 24-25 demonstrated statistic significant differences on the number of CLT coated vesicles, endocytic pits, and endosomes. Thus, while control presynaptic terminals contained more coated vesicles, and endocytic pits than tbl ones (Fig. 8E); the number of endosomes was greater in tbl presynaptic boutons relative to control ones (Fig. 8E). Moreover, the qualitative appearance that tbl endosomes were bigger than control ones is clearly ratified by the measurement of their maximum diameters that were twice in tbl presynaptic endings relative to the www.nature.com/scientificreports/ control ones (Fig. 8F). This dysregulation of the recycling endocytosis pathway is reinforced by present immunocytochemical experiments. Early endosome antigen 1 (EEA1) distributed through the neuronal cytoplasm ( Fig. 7A-B); and the quantitative analysis shows significant differences of the immunocytochemical expression of EEA1 between tbl and control cultured hippocampal neurons, being almost twice in tbl neurons regarding the control ones (Fig. 7D). Another important finding of present experiments is related with the presence of autophagy vacuoles. Thus, from earliest reports, the ultrastructural features of these vacuoles as it is the presence of double membrane encircling damaged or ubiquitinated proteins 26 help their easy identification. Autophagosomes and mitochondrial debris (Fig. 9B) are often observed in tbl presynaptic boutons; and, in addition to the previously described  www.nature.com/scientificreports/ presynaptic terminals (n = 26) relative to control ones (n = 30) (Fig. 9C). From these observations it seems that the membrane dynamics of the presynaptic terminal might be altered in the tbl mutation. As the synaptic www.nature.com/scientificreports/ transmission takes place at the terminal active zone, which is smaller in tbl presynaptic terminal than in control ones ( Fig. 1C-D), we have measured the perimeter of the presynaptic boutons, the perimeter of the endosomes www.nature.com/scientificreports/ and the perimeter of the autophagosomes and compared these in absolute values and as their ratios in relation to the active zone length (Fig. 9). Thus, while the values of presynaptic boutons perimeters and of the sum of the perimeter of the plasma membrane surrounding the bouton and that of the endo-and autophagosomes did not shown significant differences between control and tbl terminals (Fig. 9D), the perimeter of endosomes and www.nature.com/scientificreports/ autophagosomes was significantly greater in tbl terminals than in control ones (Fig. 9D). When the size of active zone is analyzed according their ratio respective to the values of the perimeter of the rest of presynaptic ending www.nature.com/scientificreports/ membranes were measured, there were significant differences between control and tbl presynaptic terminals in which these ratios were ever highest in control presynaptic boutons (Fig. 9E).
Although previous experiments described that HERC1-CLT interaction takes place through RLD2 domain 17 , we have tried to analyze the possible role that mutated RLD1 domain could play in this interaction. However, the fusion of proteins with amino acid residues of HERC1 RLD1 domain in cultured control and tbl hippocampal neurons experiments was not yet attained accurate levels in our hands. Therefore, as indirect analysis we have transfected HEK-293 T cells transfected with pFG41 (RLD1 control), pFG44 (RLD1 tbl) or GFP fusion constructs. 72-h post-transfection, the immunoblot of protein retained in the resin of the cell lysates demonstrated that in contrast to pFG41, pFG44 did not interact with CLT (Fig. 10, and Supplementary Information). Thus, this result shows for the first time that tbl RLD1 domain alters HERC1-CLT interaction, and open the possibility that this lack of interaction might be also present in tbl central synapses.

Discussion
Biallelic Herc1 mutation causes in humans a polymorphic syndrome with varied signs and symptoms [6][7][8] , together with intellectual disability [see table 1 in ref. 8 ]. Recently a new Herc1 mutation characterized by the synthesis of a truncated Herc1 protein lacking the C-terminal of HECT domain has been described, in addition with signs of the autism spectrum 27 . The animal model of HERC 1 mutation, the tambaleante mouse, was first described as an example of adult cerebellar ataxia by autophagy Purkinje cell death 5,[9][10][11][12] . Later studies showed more neurodevelopmental damages to tbl mice phenotype as: (1) delay in synaptic development and maturation of NMJ 16 ; (2) anomalous myelination of peripheral nerves and delayed action potential transmission 15 ; (3) increase of autophagy signs in projection neurons of central nervous system as the hippocampal CA3 pyramidal neurons, the neocortical pyramidal neurons, and the motor neurons of the spinal cord 13 ; and, (4) impairment of associative learning linked to absence of LTP and anomalous spinogenesis of the neurons of the lateral amygdala 14 . Therefore, as part of the ubiquitin-proteasome system (UPS) whose alterations have been related with several neurodegenerative diseases of nervous system [28][29][30][31][32][33][34] , all these findings have been explained as the effect of the dysregulation of the autophagy 5,13-15 and/or of the mammalian target of rapamycin complex 1 (mTORC1) altered regulation 5,15 . Present results in hippocampal cultured neurons show that tbl presynaptic terminals possess a noticeable decrease in the number of presynaptic terminals relative to the control ones, which in addition display: (1) a lower number of the ready releasable and the reserve pools of synaptic vesicles; (2) a shorter zone active; (3) less docked synaptic vesicles; (4) less SYT1; (5) less CLT coated vesicles and CLT immunoreactivity; (6) increased number of big endosomes correlated with and also high expression of EEA1; and, (7) an important presence of autophagosomes. These results are in accordance with previous findings in tbl mice in vivo. The decrease in the number of clear and round synaptic vesicles, in SYT1 expression as in the SV2B expression in tbl cultured neurons correlates with the decrease of SV2B-VGLUT1 co-expressing vesicles found in the lateral amygdala of adult tbl mice 14 . Furthermore, a diminution of the ready releasable pool of synaptic vesicles (RRP) together with a decrease in the probability of release was also found at the tbl NMJ 16 . Hence, all these data together strongly suggest that HERC1 might play a pivotal role in the presynaptic activity and the synaptic vesicle dynamics.
Postmitotic neurons are highly specialized cells whose presynaptic terminals lye several dozens or hundreds of microns far from the cell soma, and therefore part of the protein turnover resulting from the high synaptic activity takes place locally. Thus, more and more experimental evidences demonstrate that autophagy homeostasis, as well as their dysregulation eliciting neurodegenerative diseases, takes place at the presynaptic terminals [35][36][37][38] . In fact, studies in another model of ataxia involving Purkinje cell degeneration, the Lurcher mutant mouse, demonstrated that Purkinje cell axons react earlier and more strongly to autophagy than the Purkinje cell bodies 39 . Furthermore, Lurcher mice lacking the Atg7 gene that encodes the autophagy-related protein 7 showed that axonal degeneration occurs prior to and independently of Purkinje cell death 40 . Indeed, it is now well established that axon terminals accumulate the greatest number of neuronal autophagosomes 41 and that autophagy is key to preventing axonal degeneration after injury 42 .
The synaptic vesicle recycling pathways are far to be completely understood [43][44][45] . The endosomes play a pivotal role at least in two of these pathways: the CLT mediated endocytosis (CME), and the ultrafast endocytosis (UFE) 23,44 provide: (1) the renewal of synaptic vesicles (CME and UFE); and/or (2) the elimination of damaged components of synaptic vesicles by lysosomal degradation (CME) 46 . Our results demonstrate an increase in the number of endosomes within the tbl presynaptic endings, which correlates to an also higher expression of EEA1 and with the presence of multivesicular bodies (Fig. 9). The clear deficit of CLT together to the reduction in synaptic vesicles number in tbl presynaptic endings could imply that less fused synaptic vesicles could integrate within the early endosomes, and could alter the complex process of endosomes maturation. In fact, alterations in the endosomes size and maturation have been reported in several neurological and neurodegenerative diseases [for a rev see 46 ]. Therefore, our present findings open an interesting work line to analyze what of the proteins (CME interactors 47 , GTP-Rab interactions, i.e., see below) involved in the complex process of endosome maturation are altered by the HERC1 mutation. In fact, the role of CLT in the maintenance of synaptic vesicles number and in the endosomes homeostasis suggested by our observations, have been elegantly demonstrated by Kononenko et al. 48 , whose experiments with CLT KO mice, clearly demonstrated the depletion in synaptic vesicles together to endosome accumulation; two morphological data which are similar to that present the tbl presynaptic endings.
Amongst their many several functions, the GTPases Rab family has been involved in synaptic vesicles autophagy through the canonical endosomal-lysosomal system 36,49 . Moreover, the work of Binotti et al. 50  www.nature.com/scientificreports/   1,51 . Therefore, it could be conceivable that mutated RLD1 domain might alters the regulatory action of HERC1 on the GTPases Rab family affecting the mechanism of synaptic vesicles recycling (Fig. 11A). CME is a key pathway for synaptic vesicles recycling at the presynaptic terminal 18,24,25 . CLT coated vesicles are uncoated by the action of the complex Endophilin A (Endo A)-Synaptojanin 1 (Synj 1) 25,51 . Synj 1 has been recently described as an autophagy regulator promoting the autophagosomes maturation 36,51,52 . Furthermore, a link between CLT coated vesicles and the autophagosome precursors has been described in basis to CLT-Atg 16 L1 (Autophagy related 16 like 1 protein) interaction 53,54 . HERC1 RLD2 interact with CLT and cytoplasmic vesicle trafficking 1,17 , and tbl mutation might alters CLT balance (Fig. 11B), as demonstrated by the decrease of CLT immunoreactivity and coated vesicles found in cultured hippocampal tbl neurons. These, correlate with the decrease in the number of synaptic vesicles and the increase in number and size of the endosomes in these neurons, indicative of a dysregulation of the synaptic vesicles recycling via CME. Whether the relationships between autophagy factors such Synj 1 or Atg 16 L1 should be further investigated to exclude that results found here are exclusively promoted by the CLT downregulation. Further, our GFP pull-down experiments in transfected HEK-293 T cells demonstrated that mutated HERC1 RLD1 domain was unable to interact with CLT (Fig. 11C), indicative that HERC1 mutation directly alters normal CLT dynamic 55 , irrespective of the role that such interaction will play through RLD2 domain, being suggestive that this disbalance could be also elicited in tbl central synapses.
The mTORC1 plays a key role in autophagy 56 , and its inhibition enhances autophagy, inducing a loss of dopaminergic synaptic vesicles 57 . In normal conditions, HERC1 could regulate mTORC1 activity through interaction with the tuberous sclerosis complex 2 (TSC2) 58 , helping thus to maintain the homeostatic autophagy; however, the activity of mTORC1 is reduced by HERC1 overproduction in the tbl mutation 5 , and this decrease in mTORC1 action could dysregulate autophagy and neuronal cell death found in tbl mice 5 . Therefore, the increase of autophagosomes and multivesicular bodies, and at least in part the decrease in the number of synaptic vesicles could be explained by a misfunction of the HERC1 protein in tbl mutation (Fig. 11D).
In conclusion, although the specific targets and pathways damaged by the HERC1 mutation are still unsolved, our results ratify and extend previous in vivo findings and leads us to hypothesize that HERC1 mutation alters central synapses dynamic by impairing the normal CLT balance, altering thus the homeostasis of synaptic vesicles recycling. Therefore, we postulate HERC1 as an important regulator in neurodevelopment, and particularly in the alterations of synaptic transmission homeostasis, one of the bases of the intellectual disability.   For synaptic vesicles count w ® e have followed the methodology proposed by Schikorski and Stevens 59 . Presynaptic endings (control n = 30; and tbl n = 26) were divided in four compartment (75 nm width each) parallels to the active zone. All counting was done with the Fiji ImageJ software (W. Rasband, National Institutes of Health, https ://image j-nih.gov/ij/). Parameters considered here were: (1) the total number of vesicles; (2)   Synaptic terminals were loaded by substitution of basal solution with 70 mM [K + ] Locke solution (generated by equimolar replacement of NaCl by KCl) containing 4 μM FM1-43 (Molecular Probes) for 2 min. For imaging experiments, we use an inverted microscope IX-71 (Olympus) located on an isolation table. The recording chamber was set up on a separate XY stage attached to an optical bench column. A 3-axis mechanical manipulator was also attached for micropipette positioning. The microscope camera port allows 100% light reflection to which a QE-Sensicam camera (12 bit, PCO) was attached. The light source was a monochromator polychrome IV (100 W Xenon lamp) (Till Photonics). Image acquisition and illumination were controlled with TILL Vision software. Images were collected through a plan-NeoFluor 40 × objective (0.75 NA). Excitation of the FM1-43 was done at 488 nm and the emission light was collected through a band-pass emission filter (510 ± 10 nm). The pixel size was 0.32 μm. Neurons were stimulated by placing a micropipette near the soma (≈ 5-10 μm). Micropipettes were pulled in a P97 (Sutter Instruments) puller, having a resistance between 1-2 MΩ, when filled with the recording solution. Neuronal stimulation was done with an electrical stimulator (model 2,100, AM-Systems) by passing biphasic 1 ms pulses of 45-50 μA. A TTL signal generated with the TILL Vision software triggered stimulation and data acquisition. We collect 250 images per experiment. Exposure time was 800 ms, the image sampling rate was 1 frame/s. We used Igor Pro 6.22 (Wavemetrics Inc) custom made macros for image analysis. Synaptic responses were identified as regions the destaining of fluorescence. The ROIs were a square of 5 × 5 pixels (1.6 μm by 1.6 μm for a 40 × objective). For analysis, we use a Kolmogorov-Smirnov to check the normality of the samples and apply a t test to compare mean values.
The images were acquired on an upright Olympus FluoView 1,000 confocal laser scanning microscope. The figures were prepared using the Photoshop 7.0 (Adobe ® ) software without any additional correction.
Immunoreactivity was quantified as indicated previously 14 . Briefly, an alternating sequence of laser pulses was used to activate the different fluorescent probes during image acquisition. Images were acquired with a 60 × oilimmersion objective (N.A. 1.42). Images from the hippocampal cultures of control and tbl mice were obtained in the same session under similar conditions (laser intensities and photomultiplier voltages). Quantification of the fluorescent labelling intensity was performed offline with ImageJ and the size of the areas measured was determined automatically by defining outline masks based on the brightness thresholds from maximal projected confocal images. The control and tbl images of CLT, EEA1, HERC1, MAP2, SV2B and SYT were captured as follows and expressed in arbitrary units: (1) CLT: Rhodamine laser intensity 5% with photomultiplier settings HV 410, Gain 2, Offset 25; (2) EEA1: FITC laser intensity 4% with photomultiplier setting HV 530, Gain 3., Offset 15; (3) HERC1: Rhodamine laser intensity 1% with photomultiplier setting HV 600, Gain 3, Offset 13; (4) MAP2: FITC laser intensity 10% with photomultiplier setting HV 413, Gain 3, Offset 10; (5) SV2B: FITC laser intensity 5%, with photomultiplier settings HV 401, Gain 3, Offset 15; and (6) SYT1: FITC laser intensity 4% with photomultiplier setting HV 530, Gain 3., Offset 15.