Astrocyte-to-neuron transportation of enhanced green fluorescent protein in cerebral cortex requires F-actin dependent tunneling nanotubes

Tunneling nanotube (TNT), a dynamic cell–cell contact, is dependent on actin polymerization. TNTs are efficient in transporting ions, proteins and organelles intercellularly, which are important mechanisms in physiological and pathological processes. Reported studies on the existence and function of TNTs among neural cells focus on cultured cell for the convenience in detecting TNTs’ ultrastructure. In this study, the adeno-associated virus (AAV-GFAP-EGFP-p2A-cre) was injected into the cerebral cortex of knock-in mice ROSA26 GNZ. GFAP promoter initiated the expression of enhanced green fluorescent protein (EGFP) in infected astrocytes. At 10 days post injection (10 DPI), EGFP transferred from astrocytes in layer I–III to neurons in layer V. The dissemination of EGFP was not through endocytosis or exosome. Applying microscopes, we found that the intercellular transportation of EGFP through contact connection was F-actin dependent. Therefore, we concluded that EGFP transported from astrocytes to neurons in cortex via F-actin dependent TNTs. This study first proved that proteins transported intercellularly via TNTs in brain.


Introduction
Tunneling nanotubes (TNT), a F-actin dependent principle of cell-cell contact communication, was first visualized on mammal cells by Rustom et al in 2004 1 . It was then discovered to play important roles in different physiological and pathological processes in various cells in vitro 2 . In previous studies, TNT was found to transport mitochondria from progenitor/stem cells to rescue endothelial cell line -HUVEC 3,4 and cardiomyoblast -H9c2 5 from chemical induced injury and ischemia respectively. Moreover, TNT was discovered to transfer calcium flux 6,7 , organelles 8 and cytosol 8,9 to synchronize neighbor cells and induce cell differentiation. In pathological studies, TNT was proven to diffuse HIV among T cells 10 and disseminate multi-drug resistance protein P-gp among breast cancer cells to acquire non-genetic resistance of chemotherapy 11 . In researches of neuron degeneration disease, TNT was demonstrated to spread prion 12,13 , mutant huntingtin 14,15 , amyloid β 16 and Tau 17 among neuronal cells, suggesting potential mechanisms in the deterioration of spongiform encephalopathies, Huntingtin's disease and Alzheimer's disease. Astrocytes, an important regulator of neurite in the central nervous system, communicate intimately with neurons 18, 19 . In vitro studies discovered that astrocytes can transmit calcium flux to neurons via TNT and induce depolarization of neurons 20 . Wang et al. demonstrated that developing neurons stretched filopodia to constructed TNT contacting astrocytes with a long distance 20 . After examining the ultrastructure of TNT among mouse cathecholaminergic CAD cells and human neuroblastoma SH-SY5Y cells by cryo-transmission electron microscope, Sartori-Rupp et al. illustrated that most TNTs comprised of a bundle of individual TNTs (iTNTs). Each iTNT is filled by regularly organized bundles of F-actin 21 . However no in vivo study reported whether there were contact interation between astrocytes and neurons yet. In this study, mediated by Adeno-associate virus (AAV), exogeneous Enhanced Green Fluorescent Protein (EGFP) was expressed specifically in astrocytes in mouse cortex. Suprisingly EGFP was tansfered to neurons in layer Ⅴ. Further we found that the dissemination of EGFP was inhibited by the F-actin inhibitor and was not affected by the exosome inhibitor. We concluded that EGFP transportation from astrocytes to neurons through TNT in cortex. Our study provide first evidence on the transportation of proteins via contact communication between astrocytes and neurons in vivo.

Results
Astrocytes transfered EGFP to neurons in the cortex. AAV-GFAP-EGFP-P2A-Cre has been injected in the prefrontal cortex of homozygous ROSA26 GNZ knock-in mice with the age of P30 for 5-10 days (Fig. 1a). The expression of cre and EGFP are initiated by promoter GFAP in GFAP+ cells. Recombinase cre recognizes and cuts the sequence of loxp-stop codon-loxp in transgenic mouse ROSA26 GNZ, and then induce the expression of GFP-β -galatosidase fusion protein with a nuclear positioning sequence (Fig. 1a) 22 . After 5 days (5 DPI), EGFP existed in both astroctytes and neurons in the cortex with propotions of 84.3±4% and 15.7±4% respectively (Fig. 1b,c,d,e). After 10 days (10 DPI), EGFP was also detected in both of astrocytes and neurons, and in the population of EGFP+ cells, 61.5±8.6% were astrocytes and 38.5±8.6% were neurons (Fig. 1b,c,d,e). The proportion of EGFP+ neurons at 10 DPI increased significantly than that at 5 DPI (Fig. 1d,e). EGFP+ astrocytes were mostly in layer Ⅰ-Ⅲ and EGFP+ neurons were predominantly in layer Ⅴ (Fig. 2a). To confirm the specificity of promotor -GFAP, we detected the expression of nuclear localized β -galatosidase induced by Cre. Results indicated that β -galatosidase expressed specifically in astrocytes (Fig. 2b). Since the expression of Cre is initiated by promoter -GFAP, the result demonstrated that the promoter -GFAP is an astrocyte specific promotor. Therefore, EGFP in neurons was not expressed in situ but transported from astrocytes.

EGFP was not transferred intercellularly via vesicles.
Most EGFP+ neurons existed in layer Ⅴ and are not close to EGFP expressing astrocytes in layer Ⅰ-Ⅲ of the cortex (Fig. 2a). One possible way of transferring EGFP from astrocytes to neurons is secretion. If EGFP was secreted as a free molecule to the extracellular environment, it might be endocytosed by the traget cell and then fused with lysosome. To detect whether EGFP was transferred from astrocytes to neurons through exocytosis and endocytosis, we examined whether lysosome marker -lamp1 23 was co-localized with EGFP. The result demonstrated that EGFP particles were not localized with lamp1 ( Fig. 3), indicating free EGFP was not secreted from astrocytes to extracellular environment and enter neurons through endocytosis. However, EGFP might be secreted in nanovesicles with membrane -exosomes. The exosome is an efficient vector to cargo biomolecules and transmit signals intercellularly in the neuron-astrocytes network 24 . GW4869 -an inhibitor of neutral sphingomyelinase can block the biogenesis and release of exosomes efficiently 25 . Therefore, the extent of GW4869 prohibiting intercellular spreading of EGFP manifests the reliance that EGFP transportaion placing on exosomes. To confirm whether EGFP was transferred from astrocytes to neurons via exosomes, we inject GW4869 in the same site injected with AAV-GFAP-Cre-EGFP and examin the dissemination of EGFP. Previous studies demonstrated that 10-20 μM GW4869 could decrease more than 70% exosomes released from neurons 26 and astrocytes 27 . Here we injected 1 μl of 80 μM GW4869 in the cortex to inhibit exosome formation thoroughly. To avoid the influence of GW4869 on the process of AAVs infecting cells, GW4869 was injected in the prefrontal cortex 30 minutes after injection of AAV-GFAP-EGFP-P2A-Cre, and the control group was injected with 1/200 (v/v) DMSO 30 minutes after the injection of AAV-GFAP-EGFP-P2A-Cre. After 10 days, we examined the existence of EGFP in the brain, and found that GW4869 did not affect the distribution of EGFP in the cortex (Fig. 4a). The results showed that EGFP existed in astrocytes and neurons with the proportion of 65.1±8.4% and 34.9±8.4% respectively (Fig. 4b,c,d,e), which is not significantly different from the control with 66.1±4.3% astrocytes and 33.9±4.3% neurons in EGFP+ population (Fig. 4b,c,d,e), indicating that EGFP was not transported intercellularly through exosomes.
EGFP was transferred between astrocytes and neurons via TNT. TNTs are membrane processes connecting two cells and open on both ends, which makes it a highway to transport ions, molecules and organelles efficiently and specifically 2,28 . The TNT is a potential way of the intercellular transportation of EGFP. In vitro studies demonstrated that each TNT comprises of 2 to 11 iTNTs. The average diameter of iTNT is around 100 nm. Adding the space between iTNTs, the diameter of TNTs ranges from 145 to 700 nm with an average of 305 nm 1,21 . Furthermore, TNTs are transient intercellular connecting structure which lasts for 10-15 min 20 . As a fragile, tiny and dynamic ultrastrucutre, the TNT is susceptible to chemicals and the changing environment 29 in which we prepare experimental tissue samples. Since markers used to detect TNTs, such as membrane markerwheat germ agglutinin and F-actin binding molecule -phalloidin 1 are not TNTs specific but ubiquitous in membrane processes of various cells, the background is too noisy to discriminate TNTs in solid tissues. All in all, TNTs are hardly detected in solid tissues under microscopes. F-actin is a crucial skeleton to sustain the structure of TNTs 20,21 . To examin whether intercellular transferring EGFP was TNTs dependent, an inhibitor of F-actin assembly -cytochalasin B 30 was applied in the in vivo study. According to the kinetics study of F-actin, around 1μg/ml cytochalasin B could reduced 90% polymerization 30 . Studies on mammal cells found that less than 0.2 μg/ml cytochalasin B can inhibit 90% intercellular TNTs 31 . Here we used 1 μl of 50 μg/ml cytochalasin B to prohibit TNTs in cortex thoroughly. To avoid the influence of cytochalasin on the process of AAVs infecting cells, cytochalasin B was injected to the prefrontal cortex 30 minutes after the injection of AAV-GFAP-EGFP-P2A-Cre, and the control group was injected with 1 μl of 1/100 (v/v) DMSO 30 minutes after the injection of AAV-GFAP-EGFP-P2A-Cre. 10 days later, the existence of EGFP in the brain was examined. With the treatment of cytochalasin B, the distribution of EGFP in the cortex was obviously differently from the control (Fig. 5a). The proportion of astrocytes and neurons in EGFP+ cells were 91±8.8% and 9±8.8% respectively in cytochalasin B treated group (Fig. 5b,c,d,e), and the proportion of astrocytes and neurons in EGFP+ cells were 60.6±9.2% and 39.4±9.2% respectively in the control group (Fig. 5b,c,d,e). According to the statistic analysis, the population of EGFP+ neurons decreased significantly after the treatment of cytochalasin B (Fig. 5d,e). By capturing the z-stack movie of cerebral cortex, we found a tiny link between a neuron's apical dendrite and an astroctye's process (Fig. 6a, Movie 1), which is probably the TNT. To visualize the ultrastructure of TNTs, immunoelectron microscopy (IEM) was applied to detect the distribution of EGFP in the cortex. In IEM photos, EGFP was not found to be in membrane vesicles or exist in the extracellular environment (Fig. 6b). A structure with condensed EGFP was suspected to be the cross section of a TNT with a diameter around 150 nm (Fig. 6b). Results demonstrated that the transportation of EGFP from astrocytes to neurons was F-actin dependent TNT (Fig. 6c). Exosome secretion is also regulated by F-actin 32 , but exosome specific inhibitor GW4869 did not block the transferring of EGFP (Fig. 4a,b,c,d,e), which indicating that exosome is not critical in astrocyte-to-neuron transportating EGFP. Considering cytochalasin B distroyed EGFP transportation drastically, our study illustrated that EGFP transportaion from astrocytes to neurons in the cortext is mainly rely on F-actin dependent TNTs (Fig. 6).

Discussion
Astrocytes are crucial elements in the nervous system to provide strucutral and metabolic support for the activity and homeostasis of neurons 33 . As far as we know, astrocytes and neurons exchange neurotransmitters and metabolic molecules through particular membrane channels 34,35 . Astrocytes were also reported to transport proteins, such as apolipoprotein D to neurons through exosomes 36 . In vitro studies demonstrated that astrocytes transferred prions 12 and calcium ions 20 to neurons via TNT, unveiling a highway for specific intercellular transportation between astrocytes and neurons. However, no in vivo study reports TNT between astrocytes and neurons yet. Here we discovered that EGFP produced in astrocytes of cortex layer Ⅰ-Ⅲ were transferred specificly and efficiently to neurons in layer Ⅴ (Fig. 2a). The process of the intercellular transportation is F-actin dependent (Fig. 5) and endocytosis/exosome independent (Fig. 3&4), suggesting that EGFP transferred from astrocytes to nerons through TNT. Our study firstly provides in vivo evidence of intercellular transportation of proteins via TNTs in the nervous system, and experimental mehods to study TNTs in solid tissues. In this study, there is a P2A sequence between recombinant gene EGFP and cre vected by adeno-associated virus (Fig. 1a), which can depart productes of two genes efficiently. According to Le et al.'s study, cre recombinance is a nuclear localizing protein both in free state and in cre-EGFP fusion state 37 . The result in this study that neurons did not expressβ -galatosidase (Fig.  2b), demonstrated that cre recombinanse expressed in astrocytes could not be transferred to neurons, which re-proved the nuclear localization property of cre. TNTs are contact passages between nearby cells. In the cortext, GFP was trasferred from astrocytes in layer Ⅰ-Ⅲ to neurons in layer Ⅴ. Although somas of both cells are distant, membrane processes, such as apical dendrites of neurons, can extend near astrocytes in layer Ⅰ -Ⅳ 38 (Fig. 6). According to Wang et al.'s study, in vitro developing neurons streched out processes to distant astrocytes to form TNTs, which is microtubules and F-actin dependent 20 . Since directed dendrite growth requires dynamic F-actin population stalling at the branching sites 39,40 , we hypothesize that TNTs in the cortex were formed between apical dendrites and astrocyte somas/processes with a close distance (Fig. 6 & Movie 1). Wang et al. discovered that connexin 43 is an important inducer of TNT-like ultrastructure to connect unmature neurons and astrocytes which transfer Ca2+ signals between them 20 , suggesting that TNTs form in the base of the gap junction. Given that each gap junction protein can form homotypic or hetergenic channels with specific gap junction proteins 41 , TNTs might form between specific astrocytes and neurons for their expression of particular gap junction proteins, which might explain the targeted dissemination of EGFP from astrocytes in layer Ⅰ-Ⅲ to neurons in layer Ⅴ. β -Amyloid(1-42) (Aβ42), a major component of amyloid plaques, accumulates within pyramidal neurons in the brains of individuals with Alzheimer's disease (AD) and Down syndrome. Nagele et al. found that besides pyramidal neurons, Aβ42 accumulated in astrocytes of cortical molecular layer (layer Ⅰ), which showed moderate to advanced AD pathology 42 . How Aβ42 transferred from neurons to astrocytes in vivo is unclear yet. In this study, we found that fluorecent proteins were transported between neurons in layer Ⅴ, which looked like pyramidal neurons according to their location and shape (Fig. 1c&2b), and astrocytes in layer Ⅰ-Ⅲ via TNTs. Considering Wang et al.'s study reported that Aβ42 was transferred among astrocytes via TNTs in vitro 16 , the conclusion of our study that TNTs formed between pyramidal neurons and astrocytes might explain the mechanism of Aβ42 dissemination from pyramidal neurons to astrocytes and the exacerbation of AD pathology. Prion, an infectious protein, can spread from intercellularly in the central nervous system. The aggregation of prion in the neuron induce neuron degeneration causing spongiform encephalopathies. In that case, the mechanism of intercellular transmission of prion is a potential target to cure spongiform encephalopathies. In vitro studies discovered that prion can transfer from astrocytes to neurons and among neurons via TNTs 12 . This study provide an in vivo model to study the intercellular transmission of prion. Furthermore, in vitro studies, prion like proteins such as mutant huntingtin 15 and pathological Tau 17 were found to disseminate among neurons or astrocytes through TNTs, suggesting they might be transported between astrocyte and neuron in the cortex via TNTs like EGFP. In conclusion, results of our study provide experimental evidences for pathological studies of neuron degeneration diseases.

Materials and Methods
Animals. Homozygous ROSA26 GNZ knock-in mice were from the Jackson Laboratory (stock #008606).
Primers OIMR8038 (5′-TAAGCCTGCCCAGAAGACTC-3′), OIMR8545 (5′-AAAGTCGCTCTGAGTTGTTAT-3′) and OIMR9539 (5′-TCCAGTTCAACATCAGCCGCTACA-3′) with a 575 bp PCR product were used for genotyping of ROSA26 GNZ. Unless indicated, mice were housed in a room with a 12-h light/dark cycle with access to food and water ad libitum. Homozygous ROSA26 GNZ mice with either sex were used for experiments. Animal experiments were approved by the Institutional Animal Care and Use Committee of the Hangzhou Normal University.
Antibodies and reagents. Information on commercial mouse antibodies are as follows: GFAP (MAB360, 1:500 for IF) and NeuN (MAB377, 1:500 for IF) were from Millipore (Temecula, CA, USA). Information on commercial rabbit antibodies is as follows: GFP (TP401, 1:500 for IF) was from Torrey Pines Biolabs Inc. (Secaucus, NJ, USA). Information on commercial chicken antibodies is as follows: β-galatosidase (ab9361, 1:500 for IF) was from Abcam (Cambridge, MA, USA). Information on commercial rat antibodies is as follows : Lamp1 (ab25245, 1:500  Immunofluorescence staining. Mice were anesthetized by 1.2% Avertin (0.2 ml/10 g, i.p.) and were perfused with 4% paraformaldehyde in 0.9%(w/v) NaCl. Mouse brains were isolated and fixed in 4% PFA overnight, and then washed with 0.01 M PBS (KH 2 PO 4 2 mM, Na 2 HPO 4 8 mM, NaCl 136 mM, KCl 2.6 mM, pH 7.4) twice. The fixed brains were kept in 0.01 M PBS with 1% ProClin 200 in 4 °C until sectioned by vibrating microtome (Leica VT1000S). Soft agar-embedded mouse brains were cut into 50 μm sections and subjected to immunostaining. Briefly, brain slices were incubated with blocking buffer (10% fetal bovine serum and 0.1% TritonX-100 in 0.01 M PBS) for 1 h at room temperature, and then incubated at 4 °C overnight with primary antibodies diluted in blocking buffer. After being washed three times with PBS, samples were incubated at room temperature for 1 h with secondary antibodies, and then washed and mounted on adhesion microscope slides (CITOTEST) with fluorescent mounting medium (0.5% N-propyl gallate, 50% glyceral in 20 mM Tris, PH 8.0). Images were taken by a Zeiss LSM710 confocal microscope and Nikon ECLIPSE Ti with exactly same scanning conditions for paired experiments, and analyzed by Image J.
Immunoelectron microscopy. Mice were anesthetized by 1.2% Avertin (0.2 ml/10 g, i.p.) and were perfused with 4% paraformaldehyde and 0.5% glutaraldehyde in 0.1 M PB (NaH 2 PO 4 *H 2 O 0.023M, Na 2 HPO 4 0.077M, pH 7.4). Mouse brains were isolated and fixed in perfusion solution overnight. The fixed brains were cut into 50 μm sections by vibrating microtome and subjected to immunostaining. Brain slices were kept and re-fixed in perfusion solution for 2 hours. Statistical analysis. All data were expressed as the mean ± the standard deviation from at least three independent experiments and analyzed using unpaired two tailed Student's t-test by Excel 2007 (Microsoft, Redmond, Washington, USA). P<0.05 was considered to indicate a statistically significant difference.

Data avalability statement
The data that support the findings of this study are available from the corresponding author upon request.        Pyramidal neurons in layer Ⅴ strech out apical dendrites to layer Ⅰ-Ⅲ with branches close to EGFP expressing astrocytes. With the assistance of F-actin which is enriched in sprouting dendrite branches, pyramidal neurons develope tunneling nano-tubules to astrocytes. EGFP disseminates from astrocytes to neurons through the tunneling nano-tubule. Red arrows, TNT. TNT, tunneling nano-tubule.