PTEN knockout using retrogradely transported AAVs restores locomotor abilities in both acute and chronic spinal cord injury

Restoring function in chronic stages of spinal cord injury (SCI) has often been met with failure or reduced efficacy when regenerative strategies are delayed past the acute or sub-acute stages of injury. Restoring function in the chronically injured spinal cord remains a critical challenge. We found that a single injection of retrogradely transported adeno-associated viruses (AAVrg) to knockout the phosphatase and tensin homolog protein (PTEN) in chronic SCI can effectively target both damaged and spared axons and restore locomotor functions in near-complete injury models. AAVrg’s were injected to deliver cre recombinase and/or a red fluorescent protein (RFP) under the human Synapsin 1 promoter (hSyn1) into the spinal cords of C57BL/6 PTENFloxΔ/Δ mice to knockout PTEN (PTEN-KO) in a severe thoracic SCI crush model at both acute and chronic time points. PTEN-KO improved locomotor abilities in both acute and chronic SCI conditions over a 9-week period. Regardless of whether treatment was initiated at the time of injury (acute), or three months after SCI (chronic), mice with limited hindlimb joint movement gained hindlimb weight support after treatment. Interestingly, functional improvements were not sustained beyond 9 weeks coincident with a loss of RFP reporter-gene expression and a near-complete loss of treatment-associated functional recovery by 6 months post-treatment. Treatment effects were also specific to severely injured mice; animals with weight support at the time of treatment lost function over a 6-month period. Retrograde tracing with Fluorogold revealed viable neurons throughout the motor cortex despite a loss of RFP expression at 9 weeks post-PTEN-KO. However, few Fluorogold labeled neurons were detected within the motor cortex at 6 months post-treatment. BDA labeling from the motor cortex revealed a dense corticospinal tract (CST) bundle in all groups except chronically treated PTEN-KO mice indicating a potential long-term toxic effect of PTEN-KO to neurons in the motor cortex. PTEN-KO mice had significantly more β-tubulin III labeled axons within the lesion when treatment was delivered acutely, but not chronically post-SCI. In conclusion, we have found that using AAVrg’s to knockout PTEN is an effective manipulation capable of restoring motor functions in chronic SCI and can enhance axon growth of currently unidentified axon populations when delivered acutely after injury. However, the long-term consequences of PTEN-KO may exert neurotoxic effects.

inducing axon growth when applied in both acute and chronic conditions of SCI (9,14). Whether deleting 88 PTEN from other spinal-projecting neurons can improve function after SCI remains unknown. 89 To date, most genetic manipulations to PTEN signaling have occurred via direct injection of adeno-90 associated viruses (AAVs) into the motor cortex or red nucleus to target single spinal tracts (e.g. the 91 corticospinal or rubrospinal tracts) (9)(10)(11)14). While targeting defined axonal populations is sufficient to study 92 the effects on regeneration, the effects are limited to only a single spinal tract and may not be sufficient to elicit 93 locomotor improvements. Use of AAVs that are pseudotyped for retrograde transport (AAVrg) can be injected 94 at the lesion site and taken up by near-all neurons projecting to or below the level of the lesion, with a few 95 notable exceptions including neurons of the caudal Raphe nucleus and Locus Coeruleus (16)(17)(18). The use of 96 AAVs for retrograde targeting of spinal axons was first performed by Klaw and colleagues (2013) prior to the 97 development of a designer AAVrg envelope protein capable of more efficient retrograde transport (17,19). The 98 advent of the AAVrg pseudotype has gathered much interest as a prospective gene-delivery method for SCI 99 due to its ability to affect both damaged and spared fibers of most spinal-projecting neurons (18). To the best 100 of our knowledge, most AAVrg applications have only targeted acutely injured axons. Whether chronically 101 injured axons can be targeted to facilitate functional improvements is not well understood. 102 To determine the answers to these outstanding questions regarding the chronically injured spinal cord, 103 we applied AAVrg to delete PTEN acutely and chronically after a complete T8 crush SCI. Through a series of 104 AAvrg and tract-tracing experiments, we demonstrate that a single spinal injection of AAVrg's can target 105 neuronal populations that are otherwise challenging to reach via conventional stereotactic methods and that a 106 single AAVrg injection is sufficient to restore motor functions when applied in both acute and chronic SCI. 107 Despite observing functional improvements in both conditions, detectable effects on axon growth were only 108 observed when applied immediately post-injury. We further identify that the human Synapsin1 promoter 109 (hSyn1) is potentially affected by PTEN deletion suggesting a role of sustained growth in regulating synaptic 110 proteins; coincidently we observed that long-term PTEN deletion may also exert progressive neurotoxic effects 111 on neurons in the motor cortex. Together, our findings indicate that chronically injured axons can be targeted 112 with AAVrg to facilitate functional recovery after severe SCI. Additional work is needed to determine whether 113 AAVrg delivery can facilitate sustained functional recovery without neurotoxicity in chronic SCI. 114 2.0 Results 116

Experimental Design 117
To study the effects of PTEN-KO on axon growth and regeneration we utilized a complete crush model 118 of SCI in PTEN FloxΔ / Δ mice as described in prior literature (8,9,14,20). Complete crush models leave limited 119 spared axons through the lesion and are hallmarked by astrocyte infiltration into the lesion in chronic SCI,120 which is essential to support axon regeneration (14,21). To target spinal-projecting neurons throughout the 121 neuroaxis we delivered an AAvrg carrying cre recombinase and/or a red fluorescent protein (RFP), which were 122 produced under the hSyn1 promoter. 123 AAVrg's were intraspinally injected rostral to the lesion either immediately after SCI, or 3 months post-124 SCI. After SCI, locomotor abilities were analyzed every other week using the Basso Beattie and Bresnahan 125 scale of locomotor recovery (BBB)(22). The BBB is typically used to assess locomotor recovery in rats; 126 however, the more common Basso Mouse Scale (23) has a limited sensitivity at the lower ranges of functions, 127 e.g. for animals without weight support which is anticipated after complete spinal crush (see methods section 128 5.3). Functional outcomes were obtained in mice treated either acutely or chronically after SCI. A separate 129 group of acutely treated mice was used for both anterograde tracing of the CST using biotinylated dextran 130 amines (BDA) or retrograde tracing of spinal-projecting neurons using Fluorogold. See Fig. 1 for a graphical 131 summary of the experimental design. Groups, sample sizes, attrition rates, and outcomes are available in 132  and attrition are reported for each outcome in each experiment. High attrition was observed in acute 136 experiments which were caused in large part by bladder rupture within a few days post-SCI during daily 137 bladder care in male mice. 138

Distribution and Fluorescent Expression of AAVrg's Reveal a Consistent and Stable Labeling 139
Throughout the Neuroaxis in Control but not PTEN-KO Mice. 140

AAVrg transduces neurons throughout the neuroaxis but with low efficacy in the Locus 141
Coeruleus and caudal Raphe nucleus. 142 Recently published reports which utilized spinal cord injections of AAVrg's in the absence of injury or 143 acutely after SCI have characterized the distribution of viral expression throughout the neuroaxis (16,18,24). 144 Our experiment, however, specifically evaluated the expression patterns in neurons as long as 9 months post-145 SCI in chronic-treated conditions, utilized the hSyn1 promoter to limit expression to neurons, and evaluated the 146 effects of sustaining PTEN-KO for up to 6 months. 147 When evaluating the labeling patterns in the brain we focused on a few main areas of interest identified 148 through prior published reports (16,18). Specifically, neuron labeling was evaluated in the motor cortex, red 149 nucleus, Locus Coeruleus, Barrington's Nucleus, as well as the caudal Raphe nucleus at 2 weeks post-150 injection in acutely injured spinal cords using AAVrg-hSyn1-eGFP. We validated that labeling could be 151 detected at 2 weeks post-injection throughout the motor cortex, red nucleus, and Barrington's nucleus (Fig. 2). 152 Next, we validated that similar labeling patterns were observed at both 9 weeks and 9 months post-SCI in 153 control AAVrg-treated mice (Figs. 3,4). At all time-points, co-labeling between DBH and eGFP within the Locus 154 Coeruleus was scarce. We also observed scarce co-labeling of eGFP + neurons with 5-HT within the Raphe 155 nucleus cluster. Our results confirm previous reports demonstrating that AAVrg targets neurons throughout the 156 neural axis with some limitations. Here we confirm these effects take place after SCI and are sustained for up epicenter. Labeling was found in axons within the white matter and neuronal cell bodies within the gray matter 163 rostral to the lesion. Further, we identified axons spanning the lesion site primarily in the ventral aspect of the 164 spinal cord. These axons may indicate that although the crush lesions were severe, many were not 165 anatomically complete, or are indicative of limited sprouting/regeneration after injury (White arrow; Fig. 3A). 166 Axons that spanned the lesion were of similar morphology to what would be expected of axon sparing, 167 rather than regeneration (25). Further, we observed the existence of cell bodies caudal to the lesion which 168 could have been labeled either due to having axons that were spared from injury, potential leakage of the virus 169 across the lesion site, or could represent axons that took up the virus immediately after injury and transported 170 the viral particles to the soma prior to becoming axotomized during secondary injury cascades (Yellow arrow; 171 reporter gene expression below the injury as a measure of axon regeneration and will refer to the lesion 173 severity as severe and near-complete. 174 In mice which received PTEN-KO, fluorescent labeling was drastically reduced. Scant neurons were 175 labeled throughout the analyzed locations by 9-weeks post-treatment and labeled neurons appeared 176 hypertrophic (Fig. 3B) as previously reported with PTEN inhibition (8,14,20). By 9-months post-PTEN-KO, 177 AAVrg-mediated fluorescent was extremely limited or non-existent in every location (Fig. 3). 178 2.2.3 PTEN-KO exerts signs of long-term toxicity to neurons in the motor cortex at 6-months, but not 9-179 weeks post-injection. 180 To address if PTEN-KO was leading to neuronal death, we performed retrograde labeling in a subset of 181 SCI animals using Fluorogold at 9 weeks post-injection. In both PTEN-KO and control mice treated acutely 182 post-SCI, we detected Fluorogold labeling of neurons throughout the neural axis in locations such as the motor 183 cortex (Fig. 4A), suggesting that neuronal viability remained at 9-weeks post-treatment. To determine the long-184 term effect of PTEN-KO on neuron viability we performed track tracing 6 months after injection following 185 treatment in mice with chronic SCI (3 months post-injury). In contrast to 9-weeks after injection, at 6 months 186 post-AAVrg injection (9 months post-SCI) in PTEN-KO mice, we failed to detect strong Fluorogold labeling in 187 4B). Interestingly, however, we observed a conserved density of Fluorogold-labeled neurons in the red nucleus 189 and throughout the brainstem in PTEN-KO mice at this 6 month timepoint (Fig. 4C,D). 190 Taken together we have validated that the use of AAVrg's does indeed expand the gene-therapy effect 191 to a wider range of neurons throughout the neuroaxis. However, we have identified two confounding effects 192 when performing PTEN-KO: 1) PTEN-KO may silence the hSyn1 promoter which is both consistent with a 193 need to reduce synaptic strength for regeneration (26-30); and 2) sustaining long-term growth via  may have adverse effects on neuronal health and viability of specific populations of neurons.

Acute PTEN-KO using AAVrg's restores weight supported stepping. 197
To determine if PTEN-KO using AAVrg's could improve functional recovery, we first applied AAVrg's 198 above the lesion immediately post-SCI and monitored for locomotor recovery bi-weekly for 9-weeks. Relative 199 to control-treated mice, PTEN-KO significantly improved hind-limb function with most mice displaying a robust 200 increase in range of motion, sweeping behaviors, or plantar placing (treatment effect, F(1,15)=16.47, p=0.001; 201 time by treatment interaction, F(4,60)=9.107, p<0.0001) (Fig. 5A). Several mice regained weight supporting 202 abilities sufficient to perform at least occasional overground stepping (Fig. 5B). In contrast, control-treated mice 203 had no recovery of hind limb function and displayed only minimal range of motion usually within the ankle joint. 204 When characterizing the behavioral return there was an apparent increase in function beginning after 5-weeks 205 post-treatment (Fig. 5B). 206

Chronic PTEN-KO using AAVrg's exerts injury-severity dependent effects. 207
To determine whether PTEN inhibition could improve function after chronic SCI, we delivered control or 208 PTEN-KO AAVrg in mice 3 months after SCI. Animals were randomly assigned to balanced treatment groups 209 based upon locomotor function prior to treatment. We then evaluated locomotor performance bi-weekly for up 210 to 6-months post-treatment ( Fig. 5C-H). 211 In contrast to the acutely treated mice, chronically treated mice displayed complicated behavioral 212 changes in response to PTEN-KO. At first glance of the data when all mice are analyzed in their respective 213 groups, we do not see a main effect of PTEN-KO on motor performance but did detect a significant time-by-214 treatment interaction (treatment effect, F(1,33)=0.006, p=0.937; time by treatment interaction, F(16,528)=1.845,

PTEN-KO augments axon growth into the lesion in acute but not chronic SCI 242
Our original intent was to characterize axon growth and axon regeneration by determining RFP labeling 243 within and beyond the lesion. As mentioned above this was compromised by a PTEN-KO-specific loss of RFP 244 signal. While we have quantified RFP labeling within the lesion, results should be interpreted according to 245 these major limitations. Instead of relying on viral RFP expression, we labeled sections against β-Tubulin III 246 (β3-Tub) to visualize all axons within the lesion. The lesion boundaries were traced, and the proportional area 247 covered in β3-Tub labeling was assessed as a total axon growth marker. At 9-weeks post-SCI in acutely 248 treated mice, β3-Tub + area was significantly increased within the lesions of PTEN-KO mice, displaying a 2-fold 249 increase relative to controls (control M=14.07, PTEN-KO M=30.27; T(16)=3.141, p=0.0067) (Fig. 6A,C). In 250 contrast, 6 months after treatment in chronically injured mice, β3-Tub + area was not significantly increased in 251 the lesions of PTEN-KO mice relative to controls (control M=10.37, PTEN-KO M=12.67; T(28)=1.037, 252 p=0.308) (Fig. 6B,D). 253 Our observation of increased axon growth into the lesions of PTEN-KO mice at acute, but not chronic, 254 time points are consistent with previous findings that assert treatment strategies aimed at inducing 255 regeneration have less efficacy when delayed into chronic settings (9). However, while we did not observe a 256 total increase in β3-Tub + area in PTEN-KO mice in chronic SCI, we did observe a difference in the labeling 257 intensity. The β3-Tub + area within the lesion had a brighter and more vivid labeling intensity in PTEN-KO mice 258 compared to controls in the chronic (T(28)=2.308, p=0.028) (Fig. 6D), but not acute experiment (T(16)=1. 355, 259 p=0.194) (Fig. 6C). This suggests that while total axon growth might not have been enhanced from PTEN-KO in 260 chronic SCI, there are potential neurobiological differences occurring within the axons. For RFP labeling within 261 the lesion, no differences were found at 9-weeks following acute treatment (T(16)=0.55, p=0.589) (Fig. 6C); 262 labeling was not quantified at 9-months after chronic treatment due to an almost complete absence of RFP 263 expression in PTEN-KO mice (Fig. 6D). Next, we evaluated the regenerative effects on 5-HT axons. While there was growth of 5-HT axons into 266 and beyond the lesion even in control mice, there was no difference in 5-HT axon labeling at 9-weeks post-267 treatment following acute treatment (main effect, F(1,10)=0.004, p=0.948). Further, while 5-HT axon growth was 268 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 observed in several animals from both groups, not every animal displayed such effects. Specifically, at 9-269 weeks post-SCI 5 of 8 control-treated, and 7 of 9 PTEN-KO treated mice had evidence of 5-HT growth caudal 270 to the lesion (χ 2 =0.476, p=0.49), and this growth was limited to 400 µm caudal to the lesion border at the 271 longest observed fiber (Fig. 7A,C). 272 In contrast to our acute experiments, all chronically treated had evidence of 5-HT growth caudal to the 273 lesion 9 months after SCI and 6 months after treatment. Importantly, some mice had evidence of spared 5-HT 274 fibers which were detectable via 5-HT axon sprouting throughout the grey matter caudal to the injury in the 275 lumbar cord (furthest caudal tissue evaluated) (Fig. 7B,D). However, similar to our acute experiment, there 276 were no differences in axon regeneration between groups (main effect, F(1,26)=0.203, p=0.656; Fig. 7,D). Not 277 observing significant differences in 5-HT axon regeneration between groups is not surprising considering the 278 low number of 5-HT neurons successfully transduced with AAVrg in the caudal Raphe-Nucleus ( Fig. 1C) (16, 279 18, 24). 280

Corticospinal tract axons did not regenerate after PTEN-KO in either acute or chronic conditions. 281
Due to a loss in RFP expression in PTEN-KO mice, as well as evidence of sparing and labeling of 282 neuronal soma caudal to the lesion, we needed a method to evaluate regeneration specific to supra-spinal 283 neurons. We performed BDA tracing from the motor cortex to determine the regenerative effects of PTEN-KO 284 on the CST. Contrary to prior reports, we did not observe CST regeneration across the lesion in any of the 285 animals at any time point, or in any group (Fig. 7E,F). Surprisingly, we did observe that most chronically 286 treated PTEN-KO mice had poor or absent labeling of the CST (Fig. 7F). 287 To quantify our observation that PTEN-KO mice had less CST axons labeled with BDA at 6-months 288 post-treatment, we rated CST labeling based on being visually completely filled, moderately filled, sparse 289 evidence of axons, or no evidence of axons, and clustered animals into two groups for comparison using Chi-290 Squared. Mice with evidence of the CST being completely filled or moderately filled comprised one cluster, and 291 mice with sparse evidence of axons or no evidence of axon labeling were clustered together. There were 292 significantly less PTEN-KO treated mice with evidence of complete or moderate filling of the CST (χ 2 =6.677, 293 p=0.0098) (Fig. 7G). Concerned that chronic PTEN-KO may be neurotoxic, we performed Fluorogold tracing in 294 our last cohort of mice. As indicated in Fig. 4, the reduction in Fluorogold-labeled neurons in the motor cortex 295 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; https://doi.org/10.1101/2023.04.17.537179 doi: bioRxiv preprint may explain the lower density and prevalence of CST labeling in mice with PTEN-KO at 6-months post-296 treatment (Fig. 4,7). 297

Long Term Health and Adverse Outcomes 298
Throughout both acute and chronic experiments we observed one unanticipated outcome that was 299 specific to the PTEN-KO group. In mice receiving PTEN-KO in either the acute or chronic study, most-to-all 300 mice receiving treatment developed an elevated chronic muscular tone of the abdominal muscles which we 301 were not able to appropriately quantify, and therefore are only reporting through an empirical observation (for a 302 photographic example see Supplementary Materials 1). An increased abdominal tone phenotype was apparent 303 to the touch during routine bladder care. While we have not obtained quantitative evidence to support this 304 claim, we believe that this hypercontraction of the abdomen was the cause of the higher functioning mice 305 treated in the chronic condition to lose weight-supporting abilities after treatment. It is possible that mice 306 maintaining a hyper-contracted abdomen experienced a disrupted ability to balance and maintain weight 307 support. While we did see mice with the most severe lesions regain weight-supporting abilities in both the 308 acute and chronic conditions, an increased abdominal tone could potentially support the ability of these 309 severely injured mice to be capable of bringing their limbs under their body to assist in the recovery of weight 310 supporting abilities. Importantly, when considering our AAVrg delivery approach, we injected the virus into the 311 T8-9 vertebral segments which undoubtably affects neurons that control the abdomen. It is therefore important 312 to consider that injection of viral vectors into the spinal cord does not just affect long-tract spinal projecting 313 neurons but also affects neurons in the vicinity of the injection. There are several important and novel findings accumulated from work in this manuscript. 1) First, we 317 have replicated that delivery of AAVrg's can exert a wide genetic influence throughout the neuroaxis when 318 applied rostral to an SCI lesion but that transduction efficacy within the caudal Raphe nucleus and Locus 319 Coeruleus remain limited. 2) Even in severe and near-complete crush lesions, PTEN-KO using AAVrg's 320 improves locomotor outcomes in both acute and chronic SCI conditions, however, the effects are severity 321 dependent. 3) PTEN-KO using AAVrg's does increase axon growth into the lesions of acutely injured mice, 322 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; https://doi.org/10.1101/2023.04.17.537179 doi: bioRxiv preprint demonstrating a growth effect on axons. 4) However, driving the RFP fluorescent reporter gene under the 323 hSyn1 promoter revealed an unexpected interaction between the viral gene expression and PTEN-KO that 324 diminished the RFP reporter. 5) Finally, we observed a potential toxic response to prolonged PTEN-KO in 325 motor neurons comprising the CST, which is corroborated by a progressive loss in motor abilities over time, a 326 loss of Fluorogold labeling within the motor cortex, as well as an absence of normal CST labeling within the 327 spinal cords of PTEN-KO mice. 328 Use of gene delivery approaches to induce axon regeneration in SCI is an emerging strategy that has 329 demonstrated preliminary success (30). To date, most gene-delivery approaches to manipulate PTEN have 330 been performed with the intent of understanding the biological capacity for specific spinal tracts to regenerate, 331 mainly the CST and rubrospinal tract. Use of retrogradely transported viral particles to target a larger breadth 332 of spinal tracts via injection directly into the spinal cord confers major advantages to both study-specific spinal 333 tracts of interest, but also to target neuronal populations that are otherwise difficult to affect. Prior work by both 334 Blackmore and colleagues (16) and Metcalfe and colleagues (18) have characterized the capability for the 335 AAVrg pseudotype to affect neurons and have found similar results, both concluding that glutamatergic 336 neurons are very efficiently transduced, while neurons arising from the Raphe nucleus and Locus Coeruleus 337 display limited transduction. We observed similar trends in the brains analyzed from our experiment, however, 338 we did observe that a few labeled neurons in the Raphe nucleus were co-labeled with 5-HT. Additionally, we 339 observed that use of the hSyn1 promoter is stable for up to 9-months post-transduction in every population of 340 neurons analyzed, but the stability of this promoter may be dependent upon the regenerative state of the 341 neuron. 342 PTEN-KO leads to increased activity of AKT and the mammalian target of rapamycin (mTOR) which 343 exert modulatory roles across neuronal physiology. PTEN-KO leads to hyperexcitability of neurons and can 344 induce epilepsy or seizures when occurring throughout the cortex (31). Effects of sustained hyper-activity of 345 AKT/mTOR has also been found to induce spinogenesis on affected neurons and induce the formation of more 346 synaptic contacts with pre-synaptic neurons (32, 33), as well as facilitate sprouting of axons to make more 347 synaptic connections with post-synaptic neurons (8,9,11,14). Taken together, both mechanisms are 348 independent of long-distance axon regeneration but may exert therapeutic properties by leveraging the 349 excitability and influence of axons that have been spared from injury. Due to the relatively fast improvements in 350 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; https://doi.org/10.1101/2023.04.17.537179 doi: bioRxiv preprint locomotor functions observed in our study (observed within 3-weeks after AAVrg injection), it is likely the 351 therapeutic mechanisms of action are explained by the effects of AAVrg on spared, rather than regenerating 352

axons. 353
One of the most peculiar findings in our work is the loss of RFP fluorescence observed in the PTEN-KO 354 mice. Prior reports using PTEN-KO have utilized ubiquitous promoters such as the CMV promoter or have 355 utilized transgenic mice which express a reporter gene when exposed to the virally delivered Cre recombinase. 356 While these models apparently exert a more stable reporter gene expression, detecting a down-regulation of 357 the reporter gene while under the hSyn1 promoter is of both extreme interest and concern. A loss of RFP 358 induced by silencing of the hSyn1 promoter may prove to be consistent with an emerging theory on axon 359 regeneration. Specifically, retrograde signals from developed synapses may suppress axon growth, which 360 suggests a need to induce synaptic instability in order to facilitate axon regeneration (26-30). There is little 361 argument that PTEN inhibition induces a regenerative state within affected neurons (2, 8-14, 18, 20, 21, 31, 362 34-38). Therefore, the suppression of mature synaptic proteins such as synapsin 1 after PTEN-KO would be 363 consistent with the theory of synaptic silencing during regeneration. While silencing of the hSyn1 promoter 364 after PTEN-KO remains speculative, the implications of such findings can be extrapolated to the larger 365 concerns with gene-therapy approaches. 366 As gene therapies emerge for use in regenerative medicine it will be vital to regulate the effects in a 367 controlled and predictable manner. Use of the hSyn1 promoter may therefore not be the best option for driving 368 genetically induced regeneration, particularly if that growth effect will inevitably become downregulated as the 369 neuron enters a regenerative state. Future work should identify if this same interaction exists using other 370 neuronal specific promoters. One element of gene therapy that also needs to be addressed, particularly after 371 potentially observing toxicity to neurons in the motor cortex, is the long-term repercussions of sustaining a 372 growth and regenerative state in neurons. Prior work has not established a causal link between PTEN-KO and 373 neuronal toxicity, therefore our findings raise concerns about the feasibility of maintaining neurons in a 374 prolonged regenerative state (9,20). was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; https://doi.org/10. 1101/2023 to 1 year post-PTEN-KO from the motor cortex without observing toxicity (9). For reasons unknown, our 379 findings deviate from these prior reports and may suggest that either 1) limiting the effects of PTEN-KO to 380 neurons exerts divergent long-term responses compared to using a ubiquitous promoter that may involve 381 astrocytes and microglia in the vicinity of the neuronal soma, or 2) there is something different about using 382 AAVrg's to knockout PTEN that results in reduced long-term viability. 383 It is not inconceivable to suggest that a sustained growth response could exert adverse effects at both 384 the level of circuitry but also at the level of neurobiology. One of the main downstream targets that is activated While using AAVrg's to treat SCI holds significant translational promise, more work is needed to refine 398 the approaches and tease out potential complications associated with this methodology. More importantly, 399 however, our findings have raised a potentially important consideration requiring immediate attention. 400 Specifically, more knowledge is needed to understand the long-term consequences of sustaining growth in 401 neurons after SCI to both determine if our findings of neuron toxicity are reproducible and to determine if 402 stimulating regeneration using different approaches causes similar complications. Regenerating damaged 403 axons in the human condition will likely require significantly more time compared to rodents due to the 404 difference in the spinal cord size, however, we need to know if neurobiology can sustain the kind of prolonged 405 growth required to induce regeneration over an extended period of time. 406 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; https://doi.org/10.1101/2023.04.17.537179 doi: bioRxiv preprint Finally, while not observing regeneration of the CST after PTEN-KO may at first seem inconstant with 407 prior literature, our findings are in line with what has been described previously. The lesions in our animals, 408 although remarkably small relative to a clinically relevant contusion lesion, are wider in thickness than prior 409 reports. Here we induced crush injury with forceps approximating 0.2 mm in diameter. Prior studies have 410 utilized forceps that have been filed to be extremely thin, and only when a crush has been performed using 411 extremely thin forceps does axon regeneration of the CST appear to occur (14,21). This has been suggested 412 to be causal to the ability of astroglia bridges to form and span the ultra-thin lesions, which does not sufficiently 413 occur when the lesions are too wide (21) as in the current study. Zukor and colleagues (2013) described the 414 failure of axon regeneration when astroglia bridges do not form due to experimental lesions induced using 0.5 415 mm in thickness forceps compared to 0.1 mm thick forceps (21). In our tissue, we did not often see astroglia 416 bridges span the lesion, and when astrocyte bridges were observed they were observed in far-lateral sections 417 or far ventral regions in the cord, both of which are relatively far distances away from the main CST bundle. 418 Not finding regeneration of the CST even in our acute experiment, while not as hypothesized, is consistent with 419 previous reports considering the nature of our lesions. Our results support a growing idea that astrocytes are 420 indeed essential to support axon regeneration through a lesion and that other inhibitory elements within the 421 lesion mediate regenerative failure (21). 422

Lay summary of findings. 423
A primary cause for the loss of function experienced after a spinal cord injury (SCI) is due to the 424 disruption of communication to and from the brain. In an uninjured spinal cord, the processes of neurons that 425 carry a signal to other neurons, called axons, project long distances that lead from the brain to the lowest 426 regions of the spinal cord, or from the periphery/lower spinal cord up into the brain. Singular axon processes 427 from neurons within the brain travel uninterrupted down the spinal cord to make connections to other neurons 428 within the spinal cord. Those axons become severed at the location of the injury, preventing communication 429 with the intended target. One of the goals of regeneration after SCI is to force those damaged axons to 430 elongate and grow through the lesion and ideally make connections back to the original and intended target. 431 To date, the ability to induce long-distance regeneration is limited. Regeneration is prevented by inhibitory 432 molecules within the spinal cord, a lack of growth permissive signals/environment, as well as inhibitory 433 mechanisms ongoing within the neurons themselves. One inhibitory molecule that exists inside neurons and 434 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; https://doi.org/10.1101/2023.04.17.537179 doi: bioRxiv preprint actively suppresses regeneration is a protein called the phosphatase and tensin homolog protein (PTEN)(14). 435 Prior work has knocked out PTEN from the genomes of mouse neurons and found that when PTEN is deleted, 436 some axons within the spinal cord are capable of limited regeneration (9,14). 437 It is important to note a few things regarding the regenerative observations made in prior studies. First,438 prior experiments that knocked out PTEN as a growth-promoting strategy have evaluated regeneration only in 439 a single spinal tract at a time within the spinal cord, including the corticospinal tract (CST) or rubrospinal tract 440 (8)(9)(10)(11)14). The CST is both essential to produce voluntary dexterous movements but also has a history of 441 being difficult to coerce to regenerate. Observing regeneration of the CST after PTEN knockout is a huge 442 success, however, relative to all descending CST axons, relatively few grow through the spinal cord lesion There is a need to expand regenerative efforts to include more spinal tracts to maximize the restoration 450 of functions. To date, methods to induce regeneration by inhibiting PTEN have been induced either by using 451 gene-therapy approaches which involve the injection of DNA-delivering viral vectors into the locations in the 452 brain that derive the CST (or other singular fiber tracts), or by delivering a drug systemically that blocks PTEN 453 from functioning (36, 45, 46); both approaches have strengths and limitations. Several of the locations in the 454 brain and brainstem that make up other spinal tracts besides the CST would either be impossible to reach by 455 way of injecting a viral vector, or not advisable due to the potential for off-target effects on surrounding neurons 456 that are not implicated in SCI pathology. Further, it is likely impractical to inject into every location where 457 neurons exist that project into the spinal cord due to the wide distribution throughout the nervous system. 458 Using systemically delivered drugs to induce regeneration may be capable of reaching a greater diversity of 459 spinal-projecting neurons but may also come with off-target effects, including affecting non-neural cells or 460 neural cells not involved in SCI pathology. Ultimately, we do not know how long it will take for a damaged axon 461 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; https://doi.org/10.1101/2023.04.17.537179 doi: bioRxiv preprint to grow and regenerate in a human spinal cord, and it will likely require sustaining growth for long periods of 462 time. The safety profile for chronically delivering pro-growth compounds systemically remains unknown. 463 Our project made use of a specific form of a viral vector that is commonly used for gene therapies in 464 both humans and animals. Specifically, a viral particle called an adeno-associated virus (AAV) has been 465 engineered to be taken up by axons at distances far away from the cell body and transported into the cell body 466 (a process called retrograde transport) where it can elicit the desired genetic effects (17). Importantly, AAVs 467 that are used for this experimental purpose cannot replicate so when injected into the spinal cord, only cells or 468 axon processes local to the injection site take up the virus. Our study took advantage of this retrograde 469 property of the novel AAV and through injection directly above (rostral) to the lesion site, we attempted to 470 knockout PTEN from most of the spinal projecting neurons throughout the neuroaxis without affecting many 471 off-target neurons. 472 We delivered our gene therapy in both acute and chronic SCI conditions and observed that several 473 mice with a near-complete loss of function regained weight supporting abilities regardless of when the AAV 474 was delivered. In mice that were treated immediately after injury, we saw a corresponding increase in axon 475 growth into the lesions. In mice that were treated in chronic SCI, we observed that the worse a mouse was 476 able to function at the time of treatment, the better that mouse performed after knocking out PTEN. However, 477 we also observed that some mice that could weight support before treatment, lost weight supporting abilities 478 after knocking out PTEN. Most importantly, in the mice that responded with a gain of function after knocking 479 out PTEN, the functional improvements were not retained for longer than 9-weeks. 480 We kept chronically treated mice alive for 6-months after treatment to monitor for motor abilities and 481 found that all mice receiving PTEN knockout lost deletion-mediated improvements from 9-to 33-weeks post-482 treatment. When evaluating tissue obtained from these mice, we detected a possible toxic response to neurons 483 in the motor cortex that give rise to the CST. We currently do not know the mechanisms underlying toxicity 484 caused by sustained PTEN knockout. At this time, our experimental findings cannot be directly translated into 485 humans for several reasons. First, our data raise concern for the long-term consequences of sustaining growth 486 through PTEN interference, therefore further work is needed to identify the safety of similar approaches and 487 optimize the delivery methods. Next, our study made use of mice which have been genetically engineered to 488 be capable of responding to the gene therapy used in this study. More sophisticated approaches will be 489 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; https://doi.org/10.1101/2023.04.17.537179 doi: bioRxiv preprint required to create a similar gene-therapy approach that will function in other animals including humans (13). 490 Finally, our observation of injury-severity dependent responses to PTEN knockout suggests that more work is 491

needed to determine what injury characteristics (complete vs incomplete, and what fiber tracts should or 492
should not be spared from injury) will benefit from retrogradely transported AAVs to elicit axon growth. While 493 observing any improvement in function in chronic SCI conditions is exciting, it remains important to interpret 494 our findings as yet another small step forward. There is still a significant amount of work ahead before an 495 analogous treatment similar to what was performed in this study could be considered for experimental 496 purposes in humans. 497 498

Conclusions 499
To conclude, use of AAVrg's to knockout PTEN exerts dynamic effects that are dependent upon the 500 time applied post-SCI, the injury severity, and the duration of recovery post-treatment. PTEN-KO using 501 AAVrg's improves locomotor abilities in near-complete SCI when used to treat both acute and chronic SCI, but 502 the effects on less-severe injuries appear detrimental to weight-supporting abilities. We observed that long-503 term PTEN deletion showed negative effects even though there was an initial improvement in the most severe 504 injuries. While the use of spinal-injections of AAVrg's provides a promising approach to target the breadth of 505 axons implicated in the pathophysiology of SCI, future work is required to optimize the methods and delivery 506 approaches, as well as to identify the injury conditions most likely to experience a therapeutic gain. 507 508 5.0 Methods 509

Animals and Spinal Cord Injury Modelling 510
All procedures were approved by the University of Kentucky Institutional Animal Care and use 511 Committee. C57BL/6J mice with loxp sites flanking the catalytic domain of PTEN (B6.129S4-PTEN tm1hwu /j; 512 strain #006440; The Jackson Laboratory) were obtained and bread as homozygous breading pairs. Male mice 513 were used for acute behavioral studies while female mice were used for chronic behavioral studies. Tract 514 tracing experiments for the acute-treatment time point utilized a mix of male and female mice. Due to the long-515 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; https://doi.org/10.1101/2023.04.17.537179 doi: bioRxiv preprint timepoint used in the chronic experiment, female mice were used to mitigate unexpected mortality and health 516 issues that arise in male mice (47). Experimental groups and group sizes are included in Table 1. A graphical 517 overview of experimental methods and timelines is available in Figure 1. 518 Mice were injured between 11-13 weeks of age through a complete spinal crush injury at the T9 519 vertebral spinal level. Following dissection of the skin and musculature above the T9 vertebral process, a T9 520 laminectomy was performed and then crush injuries were performed using fine-tipped forceps as described in 521 prior literature (9,14). Forceps were closed on the spinal cord using care to ensure the tips were scraping the 522 ventral aspect of the vertebral column to minimize the risk of sparing axons at the lesion. Significant force was 523 applied to the forceps and the forceps were held closed for 8-seconds. For acute experiments, mice were 524 suspended by vertebral clips immediately after crush SCI and 2-µL of AAVrg was injected at 2-mm rostral to 525 the lesion, centered on midline of the spinal cord, and approximately 0.7-mm below the dura surface. Injections 526 were performed on a stereotactic device using a Hamilton syringe containing a leur lock tip for interchangeable 527 tips (80001; Hamilton Company, Reeno, NV). Fine glass-pulled pipette tips were purchased at a calibrated 528 10.0 µm diameter (TIP10TW1-L; World Precision Instruments, Sarasto, FL). The entire Hamilton syringe with 529 fitted pipette tip was filled with mineral oil to create a hydraulic continuity between the plunger and aspirated 530 virus. Virus was injected at a rate of 0.3 µL/minute and the needle was left in place for 3 minutes after finishing 531 the injection. For chronic experiments, the spinal cord was re-exposed at 12 weeks post-SCI and 2 µL of 532 AAVrg was injected as just described. 533 For all surgical procedures, mice were anesthetized using intraperitoneal injections of ketamine (100.0 534 mg/kg) and xylazine (10.0 mg/kg). Incisions of mice were sutured using absorbable sutures and the skin was 535 closed using non-absorbable sutures. Mice were given antibiotic (enrofloxacin; 5.0 mg/kg), and saline (1.0-536 mL/day) support for 5 consecutive days after SCI. Analgesic support was provided using a 1x delivery of 537 buprenorphine slow-release formulation (Buprenex SR, 1.0 mg/kg). Bladders of mice were expressed 2x/day 538 for the duration of the study consisting of 9 weeks for acute experiments and 9 months for chronic 539 experiments. AAVrg's were purchased from Addgene and included either AAVrg-hSyn-Cre-p2A-dTomato 540 (1.5x10 13 vg/mL; 107738-AAVrg; Addgene, Watertown, MA), which was used to knockout PTEN, or AAVrg-541 hSyn-mCherry (2.0x10 13 vg/mL; 114472-AAVrg; Addgene, Watertown, MA) as a vector control. For our initial 542 co-labeling experiments to identify the distribution of transfected neurons throughout the brain, we used 543 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 AAVrg-hSyn-eGFP (1.7x10 13 vg/mL; 50465-AAVrg; Addgene, Watertown, MA). AAV's were used undiluted 544 after 2 freeze-thaw cycles, one to allocate and the other prior to injections, as recommended by the 545 manufacturer. pAAV-hSyn-Cre-P2A-dTomato was a gift from Rylan Larsen (Addgene viral prep # 107738-546 AAVrg; http://n2t.net/addgene:107738; RRID:Addgene_107738). pAAV-hSyn-mCherry was a gift from Karl 547 Deisseroth (Addgene viral prep # 114472-AAVrg; http://n2t.net/addgene:114472; RRID:Addgene_114472). 548 pAAV-hSyn-eGFP was a gift from Bryan Roth (Addgene viral prep # 50465-AAVrg; 549 http://n2t.net/addgene:50465; RRID:Addgene_50465). 550 Mice in acute experiments were allowed to survive for up to 9 weeks post-SCI. The 9-week time-point 551 was chosen due to prior work demonstrating a need to prolong experiments at least 9 weeks in order to detect 552 regeneration of the corticospinal tract (CST) in mice receiving PTEN-KO delivered to the sensory-motor cortex 553 (14). For chronic experiments, mice were allowed to survive for 9 months. The 9-month time-point was similarly 554 chosen due to the singular report of PTEN-KO performed in chronic SCI that identified limited axon 555 regeneration of the corticospinal tract at 4 months post-knockout with better regenerative effects detected at 7 556 months (9). For chronic experiments, mice survived for a total of 3 months before treatment, and 6 months 557 post-treatment, with the 6-month post-treatment time point being determined by the emergence of severe 558 autophagia in several mice, unrelated to group assignment, causing early termination of the study. 559

Tract Tracing Procedures 560
Experiments using tract tracing procedures utilized either Fluorogold (1.0% w/v; Fluorochrome; Denver, 561 CO) injections into the spinal cord rostral to the lesion as a retrograde tracer or injections of biotinylated 562 dextran amines (BDA, D1956; 10,000 kD, 10.0% w/v, ThermoFisher Scientific, Waltham, MA) into the motor 563 cortex as an anterograde tracer to visualize the CST. Tracer injections were performed two weeks prior to 564 euthanasia. Mice were anesthetized again with ketamine (100.0 mg/kg) and xylazine (10.0 mg/kg) and either 565 had their lesion sites re-exposed or received craniotomies to expose the motor cortex of mice. Craniotomies 566 were performed by gentle grinding of the skull using a rotary tool, and removal of the bone over the injection 567 coordinates at -0.3, -0.8, -1.3 mm caudal to bregma at all 1.5, 1.0, -1.0, and -1.5 mm lateral to bregma. BDA 568 injections (0.4 L/injection) were made into all 12 locations at a rate of 0.3 L/min using pre-pulled glass 569 pipettes as described above, at a depth of approximately 0.6 mm below the dura surface. The needle was left 570 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 in place for at least 1 minute after each injection. For Fluorogold injections, the spinal cord was revealed after 571 removal of developed scar tissue over the dura and a pulled glass pipette, prepared as described above, was 572 lowered approximately 0.7 mm below the dura surface. 1.0-L of Fluorogold was injected into the spinal cord 573 at least 2.0 mm rostral to the lesion to ensure effective labeling of axons that have died back away from the 574 site of injury. The needle was left in place for at least 3-minutes after injection. Skin lesions over the skull were 575 closed using reflex clips, while lesions over the spine were closed using non-absorbable sutures. All mice were 576

Behavioral Monitoring 582
Due to the severe nature of complete crush SCI models, we did not predict any mouse to recover 583 enough function to perform weight-supported stepping. Instead, if improvements were to be made, we 584 estimated kinematic improvements would be detectable through increased range of motion of legs. We chose 585 to utilize the Basso Beattie and Bresnahan scale of locomotor improvements (22) (BBB), which was developed 586 for rats with contusive SCI, rather than the Basso Mouse Scale (23) (BMS) that is developed for mice with 587 contusive SCI, due to more resolution within the BBB at lower ranges of function. Unfortunately, there are no 588 other standardized and objective measures within the SCI field to evaluate hind limb kinematic improvements 589 prior to weight-supported abilities, which underlies the strength of using the BBB scale to assess this range on 590 the continuum of recovery. 591 As described by Basso and colleagues (2006), detection of range of motion of all joints in mice is 592 challenging which is the rationale for only assessing ankle movement on the BMS as opposed to assessing the 593 ankle, knee, and hip in the BBB. However, we chose to accept this limitation and move forward with utilizing 594 joints of the hind limbs indicated as either no movement, slight movement, or extensive movement, and include 598 observations to characterize sweeping behaviors, all which emerge before plantar placement of the paws. 599 Evaluating the range of motion of all joints, characterizing sweeping behaviors, and separating pre-(BBB 600 scores of 8 and below) vs post-weight supporting abilities (BBB scores above an 8) marks an increase in 601 resolution on the BBB scale relative to the BMS scale for severe conditions of SCI. The BMS scale only 602 evaluates the ankle joint, doesn't characterize sweeping behaviors, and combines pre-weight supported 603 plantar placement, weight supported stance, and up to consistent weight-supported dorsal stepping as a single 604 score (score of 3 on the BMS). The clustering of kinematic behaviors that transition from pre-to post-weight 605 supporting abilities in the BMS scale provides significantly less resolution for discriminating differences 606 occurring in severe, near-complete SCI. Mice were assessed at 1-week post-SCI and bi-weekly thereafter until 607 euthanasia. 608

Tissue Prep 610
At 8-weeks post-SCI mice were again anesthetized using an overdose of ketamine (~200.0 mg/kg) and 611 xylazine (~20.0 mg/kg) and perfused using 0.1 M phosphate buffered saline (PBS) followed by perfusion with The brains were cut between the caudal end of the cerebellum and the rostral aspect of the motor cortex. 620 Brains were cut at 20-30 µm and captured as serial sections on slides. 621

Immunohistochemistry Sample Processing 622
Before performing immunohistochemistry (IHC), sections were dehydrated by immersion in graded 623 ethanol baths containing 70-, 95-, and 100% ethanol and cleared of lipids by immersion in xylene for 5-624 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; minutes. We have found that clearing sections using organic solvents can often improve antibody penetration 625 and labeling quality of thicker sections, particularly in the white matter, when IHC labeling is performed on 626 slides. After rehydration of sections, when performing chromogen labeling, slides were treated with quenching 627 buffer consisting of 60/40% PBS/methanol and 0.3% H2O2 for 30-minues then washed in PBS. When 628 performing fluorescent labeling, or after quenching during chromogen labeling, unspecific labeling was blocked 629 using 60-minute incubations in 5.0% normal goat serum solution made in PBS and 0.1% Triton-X 100. All 630 primary antibody labeling was performed at room temperature overnight. Secondary antibodies include 631 Alexafluor-conjugated Goat anti-IgG/IgY (1:500; A11008, A21449; ThermoFisher Scientific.), peroxidase-632 conjugated Goat anti-IgG (1:250; PI-1000, Vector Laboratories), or alkaline phosphatase-conjugated goat anti-633 IgG/IgY (1:1000; A16057; ThermoFisher Scientific) and were incubated for 1-hour in PBS and 0.1% Triton-X 634 100 at room temperature. Spinal cord sections labeled for fluorescent IHC were treated with True Black (0.5x; 635 23007; Biotium) after antibody labeling to reduce auto-fluorescence caused by lipofuscin accumulation in 636 macrophages. 637

Spinal Cord Labeling 638
Sections were labeled to visualize the astrocyte scar using antibodies that target glial fibrillary acidic 639 protein (GFAP; 1:3000; Aves Labs) and co-labeled with β-tubulin III (β3-Tub; 1:2000, 5568S; Cell Signaling 640 Technologies) to visualize axons within the lesion. The red fluorescence from both mCherry and dTomato, 641 when present (see discussion below), was bright enough to not require immunohistochemical labeling. 642 Serotonin (5-HT) positive axons were identified using chromogen labeling against 5- HT (1:4000;20080;643 ImmunoStar) and were either not co-labeled in the first acute experiment, or were co-labeled with GFAP 644 (1:5000) using alkaline phosphatase-conjugated secondaries to visualize the astroglia borders. The CST was 645 labeled using the Avidin Biotin Complex method (ABC) to detect the BDA by incubating in ABC (5 µL/mL A and 646 B; PK-6100; Vector Laboratories) overnight at room temperature and co-labeled against GFAP (1:5000) to 647 visualize astroglia borders. 648 Barrington's nucleus were made either through landmark identification, morphology, or histological verification. 652 The Locus Coeruleus was identified using histological identification of dopamine beta-hydroxylase (DBH; 653 1:2000; 22806; ImmunoStar) positive neuron clusters at the trunk of the cerebellar peduncle shaped as a half 654 moon, while the Barrington's nucleus was identified as DBHneurons clusters cradled by the half-moon of the 655 Locus Coeruleus (48). Serotonergic neurons within the caudal Raphe nucleus were histologically identified by 656 labeling against 5-HT (1:4000) and detected in the ventral-medial aspect of the brainstem. The red nucleus is 657 easily identifiable by a bi-lateral circular cluster of neurons within the mid-brain that label vividly using AAVrg 658 approaches, while neurons of the motor cortex are identifiable as the only neurons labeled within layer 5 of the 659 cortex. In animals labeled with Fluorogold, similar brain regions were assessed for the expression of both 660 Fluorogold and the RFP expressed by the viral genome to determine differences in viability of infected neurons 661 between groups. 662

Imaging and Quantification 663
Fluorescently labeled spinal cord sections were imaged using confocal microscopy to obtain the 664 necessary pixel resolution and signal-to-noise ratio required for evaluating fine axon fibers within the lesion. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 For BDA labeling of the CST there were no BDA labeled axons caudal to the lesion in any animal, 679 therefore quantification of regenerated CST was not performed. Instead, due to an unexpected but apparent 680 loss of CST labeling in chronically treated PTEN-KO mice, the quality of CST labeling was categorically coded 681 as either completely filled, moderately filled, sparse labeling, or no labeling. A reviewer who was blinded to 682 group conditions analyzed all labeled sections for the visibility of CST labeling. Completely filled CST labeling 683 was apparent through a dense central CST bundle usually apparent in multiple midline sections. Moderately 684 filled CST labeling was considered when a singular CST bundle could be identified but had less density or 685 appeared on fewer sections compared to completely filled examples. Sparse labeling was considered when 686 axon collaterals could be detected in the grey matter, but no sections contained a singular or identifiable CST 687 bundle. No labeling was considered when no axon collaterals were observed. Due to the technical complexity 688 of BDA cortical injections, any animal with surgical notes that conferred question about the labeling quality 689 were excluded from assessment. Mice with evidence of complete or moderately labeled CST bundles were 690 analyzed in one group, and mice with sparse or no labeling were delegated into a second group. The 691 frequency of complete/moderately labeled vs sparse/no labeling were assessed using Chi-Squared. 692 Brain sections were labeled using the same immunohistochemical strategy as for spinal cord sections. 693 Neuronal populations of interest were identified to qualitatively assess the distribution of affected brain regions 694 at 2-weeks, 9-weeks, or 9-months post-SCI. Prior work by Blackmore and colleagues (2018) had identified 695 minimal co-labeling within the Raphe nucleus and almost no labeling within the Locus Coeruleus at 2-weeks 696 post-treatment (24). Our study evaluated neurons at much longer time points post-treatment which gives 697 significantly more time to either stop expressing, or start expressing, in ways undescribed from prior literature. 698

Statistics 699
Repeated-measures Analysis of Variance (ANOVA) was used to assess BBB scores as well as 5-HT 700 axon regeneration beyond the lesion. Post-hoc comparisons were performed using Fisher's least significant 701 differences when evaluating between-group effects on the repeated measures model, and Dunnett's pairwise 702 comparisons when comparing within group effects on repeated measures designs. T-tests were used to 703 compare axon labeling within the lesions, and Chi-Squared was used to evaluate the frequency of occurrences 704 when data was categorical. Probability values obtained at p < 0.05 were considered significant. 705 706 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; into the spinal cord retrogradely transport from the injection site into neurons throughout the brain. RFP 856 reporter proteins are then anterogradely transported to fill axons above the lesion. Next, in a subset of mice, 857 the corticospinal tract (CST) was traced by injecting biotinylated dextran amines (BDA; 10,000 kD) into the 858 motor cortex to anterogradely label axons within the spinal cord. Finally, in another subset of mice, Fluorogold 859 was injected into the spinal cord as a retrograde dye to detect surviving neurons throughout the brain, enabling 860 visualization of neurons that lost RFP labeling. We utilized three major experiments to assess the efficacy of 861 AAVrg's to knockout PTEN as a treatment for SCI. First, male mice were treated at the time of SCI and were 862 evaluated bi-weekly for behavioral analyses for up to 9-weeks post-injury. Next, both male and female mice 863 were treated at the time of injury and were labeled with either BDA or Fluorogold at 7-weeks post-injury. 864 Finally, female mice were treated with AAVrg's at 12-weeks post-SCI, monitored for locomotor improvements 865 bi-weekly, and either received BDA or Fluorogold labeling 9-months post-SCI. This image was created with 866 BioRender.com. 867 868 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; Both the existence of spared fibers and transduced neurons limit the ability to use RFP caudal to the lesion as 889 a metric of axon regeneration. Blue labeling is Dapi. Dopamine beta hydroxylase (DBH). Scale bars = 100 m. 890 891 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  the spinal cords of mice were injected with Fluorogold rostral to the lesion, 2-weeks prior to euthanasia. The 896 motor cortex was evaluated at 9-weeks post-SCI following acute treatment and revealed a preservation of 897 neuron densities, suggesting that neurons remain viable after PTEN-KO and that mechanisms other than 898 neuronal loss underlie the loss of RFP expression (A). However, by 6-months post-PTEN-KO in mice treated at 899 12 weeks after SCI there was an apparent loss of neurons labeled with Fluorogold in the motor cortex relative 900 to controls (B). Fluorogold was still observed within the red nucleus (C) and throughout the brainstem (D) at 6-901 months post-treatment. A loss of Fluorogold in the motor cortex at 6-months, but not 9-weeks post-treatment, 902 points to a potential toxic effect of sustained PTEN-KO in specific neuronal populations including the motor 903 cortex. Note: Most of the red puncta in the chronically treated PTEN-KO images (B,C,D) is caused by an 904 accumulation of autofluorescence (i.e. lipofuscin), rather than viral expression. Some RFP expression is 905 detected in the corpus callosum in chronically treated tissue (B). Scale bars = 100 m. 906 907 908 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; https://doi.org/10.1101/2023.04.17.537179 doi: bioRxiv preprint was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 significantly by group (F(1,31)=11.55, p=0.0019). Several of the most severely injured mice regained weight 920 supporting abilities for the first 9-weeks post-treatment (E), where-as many mice with weight supporting 921 abilities at the time of treatment lost the ability to weight support (F). To account for pre-treatment abilities 922 being a significant co-variate that introduced a divergence in our analyses, for treatment initiated 12-weeks 923 after SCI, we separated mice into pre-and post-weight supporting abilities at time of treatment. After 924 separating mice into pre-and post-weight supporting groups, we observed a significant time-by-treatment 925 interaction (F(16,176)=1.845, p=0.0287) in pre-weight supported groups. PTEN-KO mice exhibited both within 926 and between group improvements within the first 9-weeks post-treatment relative to baseline and control 927 treated mice respectively. In contrast, most mice with weight support at the time of treatment lost weight 928 support within the first 9-weeks post-treatment. Regardless of how mice were split for analysis, after 9-weeks 929 post-PTEN-KO, mice began to lose function progressively until the end of the study at 33-weeks post-SCI. 930 Green dotted horizontal line = the boundary between weight support and no weight support on the BBB scale 931 ( comparisons = *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Within group comparisons for PTEN-KO 937 mice = † p < 0.05, † † p < 0.01. 938 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 17, 2023. ; https://doi.org/10.1101/2023.04.17.537179 doi: bioRxiv preprint Figure 6. Using AAVrg to knockout PTEN increases β3-Tub axon labeling within the lesions of acutely, 940 but not chronically, treated mice. Spinal cords were extracted from mice at either 9-weeks or 9-months post-941 SCI that were treated with either a control AAVrg-RFP, or AAVrg-Cre-RFP. Spinal cords were extracted at 9-942 weeks post-SCI in mice treated acutely post-injury, or at 6-months post-treatment in mice treated at 12-weeks 943 post-SCI. Spinal cords were labeled for GFAP to identify the glial scar, as well as β-tubulin III (β3-Tub) to 944 identify axons within the lesion (A,B). As previously described, RFP expression was diminished in PTEN-KO 945 mice at 9-weeks post-treatment and was almost completely absent at 9-months post-treatment. The 946 percentage β3-Tub labeling within the lesion was quantified as a measure of axon growth. Further, the labeling 947 intensity of positive β3-Tub pixels was determined. A similar analysis was performed for RFP fluorescence in 948 tissue obtained from acutely treated mice but was not performed for tissue obtained from chronically treated 949