Ecient Targeted Transgene Delivery to Injured Lower Motor Neurons

Peripheral nerve injuries yield devastating consequences, and surgical repair outcomes remain suboptimal. Novel therapeutic strategies such as gene therapy could improve peripheral nerve regeneration. Though adeno-associated virus (AAV) vectors have delivered transgenes to intact peripheral neurons, transduction of transected neurons relevant to management of peripheral nerve injuries has not been reported. Herein, in vivo transduction eciency of axotomized murine facial neurons using four AAV capsids packaging a uorescent reporter transgene, tdTomato, is characterized. Proximal stumps of transected facial nerve branches in C57Bl/6J mice were immersed in AAV solutions. Four weeks later, facial motor nuclei were volume-imaged via whole-mount two-photon excitation microscopy, and machine learning-based image segmentation quantied the proportion of transgene expressing neurons. We observed remarkable retrograde transduction eciency with AAV-PHP.S and AAV-F, with expression levels sucient to detect intrinsic tdTomato uorescence. This study conrms successful in vivo retrograde transgene delivery to transected peripheral neurons, an approach that carries potential as a research tool and future therapeutic strategy.

Viral-mediated gene delivery to the peripheral nervous system was rst reported by Geller and Breake eld, wherein cultured peripheral neurons were transduced using a modi ed herpes simplex virus 1 (HSV-1) vector to express beta-galactosidase [29]. Retrograde in vivo gene delivery to murine peripheral neurons was later described by Dobson et al. via intramuscular (IM) injection of a recombinant HSV-1 vector [30].
Subsequently, lentivirus, adenovirus, and adeno-associated virus (AAV) vectors have been employed for in vivo transduction of intact peripheral nerves via intravenous, intrathecal, peritoneal, intramuscular, or intraneural injection [31] [32][33][34][35][36]. AAVs are optimal vectors owing to their relatively low immunogenicity, lack of pathogenicity, wide range of infectivity, and capacity to establish long-term transgene expression in non-dividing cells such as neurons [37]. AAV vectors have demonstrated potential for treatment of neurodegenerative diseases [38]. FDA-approval has already been granted for two AAV-based gene therapies, one for treatment of lower motor neuron disease (onasemnogene abeparvovec for spinal muscular atrophy) [39][40][41]. Animal models suggest gene therapy may improve peripheral nerve regeneration and functional outcomes [28, [42][43][44]. Clinical translation of gene therapy for nerve repair, though, requires preclinical evidence of targeted transduction of transected axons relevant to peripheral nerve injuries. Prior work has established intraneural or intramuscular injection of speci c AAV vector serotypes may yield targeted transduction of intact peripheral neurons via long-distance retrograde transport [45,46]. Targeted transduction of transected peripheral neurons using AAV vectors has not been previously reported. Nerve transection prevents internalization of AAV vectors at the neuromuscular junction (NMJ), alters axonal transport machinery, and changes nuclear transcription processes [47][48][49][50][51][52]. Previous research has demonstrated axotomy decreases alphaherpesvirus retrograde tra cking, possibly indicating competition for axonal transport machinery between virions and local damage signals [53].
Herein, a murine model is employed to characterize in vivo transduction e ciency of axotomized murine facial neurons using four AAV capsids. AAV6 and AAV9 were studied as these vectors have demonstrated capacity for peripheral neuron transduction [54][55][56]. However, of capsids studied, AAV9 capsid variants, AAV-F and AAV-PHP.S, yielded the highest transduction e ciencies, quanti ed by whole-mount twophoton excitation microscopy (2PEM) and machine-learning based image segmentation.
To secondarily con rm our data demonstrating successful vector retrograde transport and facial motoneuron transduction, confocal microscopy immuno uorescence (IF) using antibodies against tdT and motor neurons was performed on coronal sections of brainstem containing the facial nucleus. Each brainstem from facial nerves exposed to vector had neuronal cell bodies immunoreactive for anti-red uorescent protein (RFP) (Supplementary Fig. S7). Double-labeled sections expressed tdT in choline acetyltransferase (ChAT) immunoreactive motoneurons (Fig. 4), with robust expression only in the ipsilateral facial nucleus.

Discussion
Though targeted delivery of AAVs to adult rodent intact peripheral motoneurons has been described, transduction e ciencies are only 1-7% [34,35,55,57]. Clinically-relevant local delivery of AAV vectors to transected peripheral nerves has not been reported. Herein, we demonstrate successful in vivo retrograde transduction of transected murine facial nerves using AAV vectors. Exposure of nerve to a relatively low dose of AAV, 2.95x10 9 VGs (5.9x10 11 VGs/mL), of AAV-F and AAV-PHP.S transduced 18.7% (SD 6.95%) and 33.6% (SD 5.8%) of buccal facial motoneurons, respectively. This surprising, high e ciency tdTomato expression was limited to buccal branch facial motor subnuclei, indicating trans-synaptic spread of AAV is unlikely ( Supplementary Fig. S8). Transduction e ciencies of the two naturallyoccurring serotypes tested, AAV6 and AAV9, were far lower than the two engineered capsid variants.  [55]. This present report con rms poor tropism of AAV6 for murine motor neurons [55]. However, they injected intact nerves in close proximity to the facial nucleus, minimizing distance of retrograde transport [58-60]. Stern et al. also determined total facial motoneurons using retrograde labeling three weeks after nerve transection; this technique may underestimate total motoneurons exposed to virus as proximal stump axons degenerate after injury, decreasing axon count exposed to retrograde tracer [61-63]. In the present study, FG labeled 310 (SD 70) cell bodies when delivered three weeks after buccal branch transection, while 786.5 (SD 97) cell bodies were labeled when buccal motoneurons were exposed to FG immediately after transection. Control mice in which FG was delivered three weeks following nerve transection without virus delivery had 249 (SD 132) cell bodies labeled (Supplementary Fig S5). In contrast to prior work, the present study delivered AAV vectors to transected buccal branch motoneurons more than 2 cm from the facial nucleus, ensuring all transgene expression was due to long-distance retrograde vector transport [64]. While promising, systemic and intrathecal vector administration do not allow for restricted gene delivery to the facial nucleus, and requires higher doses risking toxicity [67-69]. Though intramuscular injection allows for targeted gene delivery, this strategy is not translatable to peripheral nerve transection injuries where the NMJ is not intact. In contrast, the present study reports high retrograde transduction e ciency when transected proximal nerve stumps were exposed to AAV-PHP.S and AAV-F, providing critical proof-ofconcept that targeted gene delivery to transected peripheral nerves is possible.
We employed a novel high-throughput murine model to quantify transduction e ciency. Prior research relies on delivering vector to intact nerve or NMJ, performing a subsequent secondary labeling study with a retrograde tracer, and then sectioning tissue for immunohistochemistry (IHC) or IF [35,42]. Alternatively, groups have co-stained for a viral transgene and neuronal marker such as ChAT to quantify transduction using IHC [34,55]. Herein, we delivered pAAV-CAG-tdTomato and, later, FG for co-labeling of facial motoneurons. These uorophores are readily separable by 2PEM excitation wavelength; tdTomato has 97% maximal excitation at 1040 nm and minimal emission at 830 nm excitation, whereas FG has 20% maximal tested excitation at 830 nm with minimal excitation at 1040 nm (Fig. 3) [70,71] [72]. Performing 2PEM using a dual-output laser at 830 nm and 1045 nm minimized spectral overlap between tdTomato and FG. Further, 1045 nm excitation of tdTomato minimized lipofuscin excitation (peak 2PEM excitation 770 nm, peak emission 650 nm), a primary cause of CNS auto uorescence [73,74]. Simple immersion of brainstems in refractive-index matching solution permitted deep imaging of specimens to depths over 350 nm, avoiding the need for tissue sectioning while enabling volumetric analysis of facial motor nuclei.
Machine-learning based image segmentation was employed to quantify the proportion of transgeneexpressing neurons. The software employed permits automated cell body counts, as well as volume and surface area measurements of segmented cell bodies ( Supplementary Fig. S6). Neuronal cell bodies transduced by AAV-F and AAV-PHP.S had greater surface areas and volumes compared to those transduced by AAV-6 and AAV-9. The reason for this difference is currently unclear. Neurons exposed to AAV-F and AAV-PHP.S had preserved cell body architecture and dendrites, while soma of neurons transduced by AAV6 and AAV9 were frequently rounded and vacuolated with loss of dendrite processes ( Supplementary Fig. S9). These morphological changes have been associated with neuronal toxicity and degeneration [75][76][77][78][79]. Though neurons exposed to FG have decreased long-term survival, this study's interval of 6 days between FG delivery and tissue harvest prevented neuronal loss secondary to FG toxicity [80].
While the mechanism by which AAV-PHP.S and AAV-F mediate e cient retrograde transduction of transected facial motoneurons remains to be determined, it is evident from our study they have higher e ciency than the parental capsid from which they were derived, AAV9. Both AAV-PHP.S and AAV-F have a unique 7-mer sequence, QAVRTSL and FVVGQSY, respectively, inserted after amino acid 588 in AAV9 VP1 which leads to surface display of ~ 50 copies of the peptide on the VRVIII loop at the three-fold axis of symmetry [32,33]. Interestingly, both capsids were identi ed after an in vivo selection with AAV9 peptide display libraries after systemic injection in mice. While the mechanism for AAV-mediated transduction is a myriad of steps, the observation that both capsids facilitate higher transduction of transected motor neurons compared to AAV9 gives a place to start investigations. One can compare entry and tra cking along axons of uorescently-labeled capsids (engineered capsids vs AAV9), to gain insight into the improved e ciency.
In the current study, axotomized facial motoneurons were exposed to virus immediately after nerve transection. Future studies will evaluate delayed AAV vector delivery to transected motoneurons, as delays of only a few days between nerve crush injury and exposure to adenovirus vector yields decreased transduction e ciency [81]. Lastly, while we demonstrate e cient and robust transduction of facial motor neurons using this approach, it will be important to test therapeutic transgene candidates such as glial derived neurotrophic factor in order to advance this technology towards clinical application [28,43].
In conclusion, this study reports e cient in vivo retrograde transgene delivery to transected lower motor neurons using AAV-F and AAV-PHP.S capsids. This model can be used to test therapeutic strategies to improve peripheral nerve regeneration, as well as a tool to speci cally probe the biology of facial motor neuron responses to axonal injury.
Conduit reservoir delivery of AAV-CAG-tdTomato and Fluoro-Gold to Transected Facial Nerves. Twentyfour adult C57 mice (7-9 weeks, 12 female) were used for vector and Fluoro-Gold™ (FG) (Fluorochrome LLC, Denver, Colorado) delivery. All animal surgeries were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and with approval by the Massachusetts Eye and Ear Animal Care Committee (ACC Protocol # 16 − 006, IRBNet ID 884247). This study was carried out in compliance with the ARRIVE guidelines.
Buprenorphine (0.05 mg/kg subcutaneously) and meloxicam (1.0 mg/kg subcutaneously) were administered prior to skin incision. A 1 cm infra-auricular incision was made on the left, and skin-muscle aps were elevated. The exorbital lacrimal gland was retracted to expose the buccal branch of the facial nerve. This branch was meticulously dissected circumferentially using the operating microscope at 25x magni cation. Care was taken to avoid crush injury. The nerve was transected just proximal to the distal pes and the proximal stump then immersed in a pipette tip containing 5 µL of titer-matched vector solution (n = 4 mice/vector group, 2 males and 2 females per vector group) (5.9x10 11 VG/mL) for 10 minutes (Supplementary Figs. S1 and S2a). In four control mice, the proximal nerve stump was dipped in 5 uL of 2% FG (Fluorochrome, Denver, CO, w/v in distilled water) for 10 minutes. In a second control group of three mice, the nerve was transected but no virus or dye was delivered. In one experiment, high-dose AAV-PHP.S (1.72x10 13 GC/mL) carrying the pAAV-CAG-tdTomato transgene expression cassette was delivered to one mouse nerve. After nerve transection, the proximal stump was placed overlying the masseteric fascia, separated from its distal stump. The wound bed was irrigated with saline and closed in a single layer using 4 − 0 absorbable suture (Polysyn, Sharpoint, Westwood, MA). Animals recovered from general anesthesia and returned to their cages. Postoperative Meloxicam was given for 72 hours post-procedure.
Three weeks after AAV delivery, the 16 vector-treated mice (n = 4 mice/vector x 4 vectors) and three mice who had undergone prior nerve transection without delivery of virus, underwent iso urane anesthesia at the above dosing with the same analgesics. A 1 cm infra-auricular incision was made on the left and the previous buccal branch transection was identi ed. The nerve was transected just proximal to the neuroma and the proximal stump immersed in 5 µL of 2% FG (Supplementary Fig. S9b). The high-dose AAV-PHP.S mouse underwent facial nerve main trunk transection with proximal stump immersion in 5 µL of 2% FG. The wound bed was irrigated with saline and closed in a single layer using 4 − 0 absorbable suture (Polysyn). Animals were recovered from general anesthesia. Postoperative Meloxicam was given for 72 hours post-procedure.
Tissue harvest. Six days following FG delivery, animals underwent CO 2 euthanasia and cardiac perfusion using 2% phosphate-buffered paraformaldehyde xative (PFA) solution. Animal heads were placed in 2% PFA overnight, then underwent brainstem harvest at the level of the facial nucleus. The intracranial facial nerve was used as a landmark for facial motor nucleus identi cation. Brainstems were placed in PBS in a light-tight container and stored at 4 o C for seven days prior to whole mount imaging.
Immuno uorescence staining of brain stems.
One animal from each treatment group underwent CO 2 euthanasia followed by cardiac perfusion using 4% PFA solution. Animal heads were placed in 4% PFA for 48 hours prior to brainstem harvest and overnight cryoprotection in 30% sucrose solution. Brainstems were then embedded in O.C.T. media (Tissue-Tek) and cryosectioned in the coronal plane at 40 µm. Floating sections were permeabilized with 0.5% Triton X-100 in PBS for 30 minutes at room temperature and blocked with 5% normal chicken serum (NCS, Abcam) in PBS for 1 hour at room temperature. Primary antibodies were incubated overnight (1:200 dilution factor) at 4°C in 1.5% NCS PBS. Primary antibodies used for this study were: rabbit anti-red uorescent protein (RFP, Code: 600-401-379, Rockland Antibodies,.); goat anti-ChAT (cat#AB144P EMD Millipore, Burlington, USA).
After washing sections, uorophore-conjugated secondary antibodies were incubated in 1.5% NCS in PBS at 1:1000 dilution. The secondary antibody for anti-RFP primary was chicken anti-rabbit Alexa Fluor 594 (Thermo, Cat A-21442) and the secondary antibody for anti-ChAT primary was chicken anti-goat Alexa Fluor 488 (Thermo, Cat A-21467). After washing in PBS, nuclei were labeled with a 1:10,000 dilution of 4′,6-diamidino-2-phenylindole (DAPI). Sections were mounted onto slides with a ne brush, dried for 2 hours, and cover slipped with Dako uorescent mounting medium (Agilent).
Imaging of whole mount brainstems and immuno uorescent tissue sections.
For IF imaging, sections were mounted with immersion oil (Code 1261, Cargille Laboratories, Cedar Grove, NJ) and imaged at multiple focal planes using a 40x objective (0.24mm WD, HC PL APO 40x/1.3 Oil CS2, Leica) on a Leica DM 6000 CS confocal microscope. For Alexa Fluor 488 visualization, an Argon excitation laser was used and emitted uorescence between 500-550 nm was collected. For Alexa Fluor 594 visualization, a Diode-pumped solid state (DPSS) 561 laser was used for excitation and emitted uorescence between 600-650 nm was captured.
2PEM imaging was performed on a commercial multiphoton microscope (TrimScope II, LaVision Biotech) powered by a dual-output femtosecond laser (Insight X3, SpectraPhysics) at 830 nm and 1045 nm. Images were acquired with a set of galvanometer mirrors and piezo XYZ-stage for large-eld volumetric imaging. Commercial image-analysis software (Bitplane Imaris 9.2; Oxford Instruments, Zurich, Switzerland) was used for stitching of tile scan 2PEM images. Optical clearing was achieved using a glycerol-immersion objective lens (CLr Plan-Neo uar 20x, Carl Zeiss) and refractive index matching solution (EasyIndex, LifeCanvas Technologies, Cambridge, MA). Fluorescent signal was spectrally ltered, a combination of short-pass and long-pass lters were used for collecting below 495 nm in the FG channel and above 560 nm in the tdTomato channel. . This classi er was used to generate a signal channel in Aivia's 3D object mesh recipe console to highlight and segment neuronal cell bodies. The object meshes were generated in a region of interest and adjusted in an iterative manner, during which morphological smoothing was performed, minimum object radius and edge intensity de ned, and holes lled until a satisfactory result was obtained. The segmentation parameters were then applied to the entire volume to generate a channel highlighting the pixels comprising cell bodies and automated cell counts were performed. This process was performed independently for FGlabeled and tdT-expressing cell bodies. The segmentation parameters were kept consistent for all reconstructions and all analyses were blinded to vector.

Statistical analyses
All statistical analyses were performed using SPSS (IBM SPSS Statistics 27). Levene's test was used to verify homogeneity of variances. Normality of data was con rmed using the Shapiro-Wilk and Kolmogorov-Smirnov tests. One-way AVOVA followed by Bonferroni post hoc analysis was performed to compare cell bodies expressing tdT, as well as FG-labeling of neuronal soma. One-way ANOVA with Games-Howell post hoc analysis was used to compare surface area and volume of segmented cell bodies. Con dence level was assessed at 95% (P < 0.05).