Oligodendrocyte lineage cells driven by neuronal activity in selected brain regions are required for episodic memory formation

The formation of long-term episodic memories requires the activation of molecular mechanisms in several regions of the medial temporal lobe, including the hippocampus and anterior cingulate cortex (ACC). The extent to which these regions engage distinct mechanisms and cell types to support memory formation is not well understood. Recent studies reported that oligodendrogenesis is essential for learning and long-term memory; however, whether oligodendrocyte lineage cells are required only in selected brain regions is still unclear. Also still unknown are the temporal kinetics of oligodendrocyte lineage cells involvement in memory processes and whether these cells are engaged in response to neuronal activity. Here we show that in rats and mice, episodic learning rapidly increases the oligodendrogenesis and myelin biogenesis transcripts Olig2, Myrf, Mbp, and Plp1 as well as oligodendrocyte precursor cells (OPC) proliferation and differentiation in the ACC, but not in the dorsal hippocampus (dHC). Region-specific knockdown or knockout of Myrf, a regulator of oligodendrocyte differentiation, revealed that cells of the oligodendrocyte lineage are required for memory formation in the ACC but not the dHC. Chemogenetic neuronal silencing in the ACC showed that neuronal activity is critical for learning-induced OPC proliferation. Hence, activity-driven oligodendrocyte lineage cells in the ACC, but not dHC, are critical for the formation of episodic memories. Impact statement Oligodendrocyte lineage cells are required in the anterior cingulate cortex but not in the hippocampus for long-term memory formation.

Second, what is the temporal progression of oligodendrocyte lineage cells requirement in the various 57 processes of learning and memory? And, finally, do oligodendrocyte changes require neuronal activation? 58 Steadman et al. (2020) reported that the acquisition of spatial memory in mice is accompanied by 59 an increase in oligodendrocyte precursor cell (OPCs) proliferation and/or differentiation mechanisms in the 60 CC1-cells from ACC from UT and trained rats perfused 1D after training (n = 787 and 889 cell across 169 three rats in UT and 1D groups respectively; two-tailed t-test; P<0.001); two-tailed t-test; * indicates 170 p<0.05, ** indicates p<0.01*** indicates p<0.001. For detailed statistical information, see Table 1 The induction of Myrf and OLIG2 in the rat ACC but not the dHC following training (Fig. 1)  176 suggested that ACC undergoes a learning-dependent differential increase in OPC proliferation and 177 oligodendrocyte differentiation. To test this hypothesis, we quantified the rate of newly: i) dividing OPCs, 178 ii) differentiating OPCs and iii) maturing oligodendrocytes in the ACC and dHC of mice. To label newly 179 dividing cells, we injected intraperitoneally (i.p.) the thymidine analog 5-ethynyl-2'-deoxyuridine (EdU) 180 into mice one hour before IA training and euthanized the mice one day later ( Fig. 2A), a timepoint at which 181 a significant training-dependent increase in OLIG2 protein levels was found ( Fig. 1G and I). To determine 182 the number of proliferating OPCs, we counted the number of cells doubly stained with fluorescent EdU and 183 platelet-derived growth factor receptor alpha (PDGFRa), a marker of OPCs (Rivers et al., 2008). To 184 determine the rate of newly differentiating oligodendrocytes, we measured the number of cells labelled with 185 three markers, EdU, OLIG2 (a marker that labels all stages of oligodendrocyte lineage including OPCs) 186 and CC1. To measure changes in number of maturing oligodendrocytes, we quantified the number of cells 187 triply stained with EdU, OLIG2, and Glutathione-S-transferase (GST)-Pi, as GST-Pi is a marker of mature 188 oligodendrocytes (Tansey et al., 1991). All these analyses were normalized against the total number of 189 DAPI-positive cells. We found a significant increase in the number of EdU+ cells, EdU+&PDGFRa+ cells 190 To extend the temporal window of detecting proliferating and differentiating OPCs, we labeled 201 cells with EdU for a more extended period and also let them progress in their differentiation before testing 202 their changes as a result of learning. Specifically, we injected EdU 4 times, once every other day, and then 203 let the mice rest for 7 days before subjecting them to IA training. We then perfused the mice 1 day after 204 training and performed immunohistochemistry (Fig. 2E). Similar to what found with EdU injected 1 hour 205 before training, we detected a significant increase in EdU+&OLIG2+&CC1+ cells in the ACC of trained 206 mice relative to untrained controls (Fig. 2F), but no significant change in EdU+ &OLIG2+ &GST-Pi+ cells 207 Collectively, these data indicated that IA training leads to a significant increase in proliferating 213 OPCs and differentiating oligodendrocytes in the ACC but not dHC, which is detectable starting one day 214 after training. However, newly proliferating OPCs did not undergo maturation in response to training during 215 this temporal window.   Myrf knockout disrupts memory formation 239 Next, we asked whether oligodendrogenesis is required for IA memory formation. We employed a 240 conditional knockout mouse model in which Myrf is deleted in OPCs. Because MYRF is a transcription 241 factor required for oligodendrocyte differentiation, its deletion in OPCs impairs oligodendrogenesis, and Myrf +/+ (P-Myrf +/+ ) wild-type littermates as controls. 248 To confirm the effect of Myrf deletion on OPCs differentiation, TAM-treated P-Myrf floxed/floxed and 249 P-Myrf +/+ mice received an injection of EdU one hour before IA training and were perfused one day later. 250 OPC differentiation was significantly inhibited in the ACC in P-Myrf floxed/floxed mice, as demonstrated by the 251 significant reduction in the number of cells positive for EdU, OLIG2, and CC1 in P-Myrf floxed/floxed mice 252 compared to P-Myrf +/+ littermate controls (Fig. 4A). 253 To test the effect of Myrf knockout on memory formation, TAM was administered to Myrf floxed/floxed 254 and P-Myrf +/+ mice seven days before IA training, and the mice were tested at 1, 7, and 28 days after training. 255 P-Myrf floxed/floxed mice exhibited significant memory reduction at all time points compared to P-Myrf +/+ 256 controls (Fig. 4B). To exclude the potential effects of multiple testing, a second experiment was conducted 257 in which P-Myrf floxed/floxed and P-Myrf +/+ littermates were tested only at 28 days after training, and we again 258 observed significant impairment in memory retention (Fig. 4C). We concluded that brain-wide 259 oligodendrogenesis is required for the formation of recent and remote long-term memories. 260 To determine whether oligodendrogenesis contributes to the persistence or storage of memory, we 261 administered TAM to P-Myrf floxed/floxed and P-Myrf +/+ mice 14 days after training, when the consolidation 262 process has significantly advanced (Bambah-Mukku et al., 2014; Squire et al., 2015). Memory retention 263 was tested 14 days after knockout, corresponding to 28 days after training, as well as at 36 days and 56 264 days after training. No difference was detected between groups (Fig. 4D), indicating that 265 oligodendrogenesis is not required for the persistence, retrieval, or storage of long-term memory. 266 Finally, to determine whether mechanisms involving oligodendrogenesis play a role in the 267 formation of non-aversive episodic memories, P-Myrf floxed/floxed and P-Myrf +/+ littermates were injected with 268 TAM seven days before being trained in novel object location (nOL), a hippocampus-dependent learning  Collectively, these results indicate that oligodendrogenesis is required for the formation of long-281 term hippocampus-dependent memories. 282 once every other day. Seven days after the last injection the mice underwent novel object location training 300 and were tested 4 hours later (two-way ANOVA followed by Bonferroni post hoc test). (F) Open field test 301 expressed as mean ±s.e.m. of (i) percent time spent in the center of the arena, (ii) total distance, and (iii) 302 mean velocity exploring the arena. (n = 9,12 mice per P-Myrf +\+ and P-Myrf flox\flox groups respectively, two-303 tailed t-test; * indicates p<0.05, ** indicates p<0.01, *** indicates p<0.001). For detailed statistical 304 information, see Table 4-Source Data1. 305 306 Myrf knockdown in the ACC but not the dHC of rats impairs memory consolidation but not learning 307 In order to investigate whether learning-dependent Myrf increase is differentially implicated in 308 distinct brain regions and memory processes, we employed a Myrf knockdown strategy. We achieved 309 region-specific and temporally restricted Myrf knockdown by using stereotactic injections to deliver an 310 antisense oligodeoxynucleotide (ASO-ODN) specific against Myrf (Myrf-ASO), and, as a control, a related 311 scrambled sequence (Myrf-SCR). We injected the ODNs bilaterally into the brain region of interest at 312 various times before and after training. 313 The rapid and temporally limited effect of the ODN-mediated knockdown approach offers the 314 opportunity to dissect, with highly defined temporal approaches, the dynamics of the requirement of with Myrf-ASO exhibited no significant differences in MBP protein expression in the ACC one day after 324 training, suggesting that Myrf-ASO treatment did not lead to demyelination (Fig. 5B). 325 In order to test whether MYRF increase is required for learning, we bilaterally injected Myrf-ASO 326 or Myrf-SCR into the ACC 15 minutes before training and tested the effect 1 hour after training. We 327 detected no differences in memory between the two groups ( Fig. 5C), indicating that MYRF is dispensable 328 in the ACC for learning and short-term IA memory. To test whether MYRF is required for memory 329 consolidation, bilateral injections of Myrf-ASO or Myrf-SCR were administered in the ACC 15 minutes 330 before and six hours after training, then memory was tested one day after training. Rats injected with Myrf-331 ASO exhibited significant memory impairment one day after training compared to rats that had received 332 Myrf-SCR injections (Fig. 5D), and the impairment persisted at 28 days after training (Fig. 5D). A reminder 333 shock given one day after the remote memory test was unable to reinstate memory, indicating that the 334 memory impairment was not due to a suppressed memory response but likely resulted from disrupted 335 memory consolidation. Furthermore, re-training one day later of rats who had been injected with Myrf-336 ASO resulted in a long-lasting memory, thereby excluding the possibility that they had experienced memory 337 loss because of ACC damage caused by surgery or injections. 338 By contrast, when Myrf-ASO was injected bilaterally into the dHC 15 minutes before and six hours 339 after IA training, we observed no effect on memory retention; memories of the two treatment groups were 340 similar at 1 day and 28 days after training (Fig. 5E). The lower level of retention in the dHC relative to the 341 ACC with stereotactic injections is typically observed. Thus, we concluded that Myrf in the ACC is critical 342 for the consolidation, but not the acquisition of IA, and is not required in the dHC. training and rats were tested at one hour post-training (n = 6 per group) or (D) injections were given 15 354 minutes before and 6 hours after training and rats were tested one day, and 28 days after training (n = 7 per 355 group). Rats received a reminder shock (RS) after the last testing, followed by another retention test a day 356 later (RS test). One day later rats underwent re-training (RT), and memory retention was tested a day later 357 (RT test). (E) Injections were given 15 minutes before and 6 hours after training and rats were tested one 358 day and 28 days after training (n = 6 per group). Data are presented as mean latency ± s.e.m. to enter the 359 dark chamber (in seconds, s; two-way ANOVA followed by Bonferroni post hoc test; * indicates P < 0.05, 360 ** indicates P < 0.01, *** indicates P < 0.001). For detailed statistical information, see table 5-source data1. differentiation (Dugas et al., 2006). Hence, we specifically tested whether MYRF-expressing differentiating 376 and mature oligodendrocytes are required in the ACC and dHC for learning and/or memory formation. 377 After two weeks to allow for viral expression, intraperitoneal injections of TAM were administered four 378 times, once every other day, then mice underwent IA training (Fig. 6A). As the virus was not engineered 379 with fluorophore sequences, we used diffusion of Chicago blue dye to confirm the injection diffusion area 380 into the ACC (Fig. 6B). Compared to Myrf +\+  Myrf +\+ littermates ( Supplementary Fig. 1E) suggesting that the effect of Myrf knockout at this timepoint is 385 detected only in EdU-labeled cells. As EdU+ cells labeled with OLIG2 and CC1 constitute a small fraction 386 of cells relative to their EdU-counterparts, we cannot exclude that changes also occur in the EdU-387 population but were undetectable due to insufficient sensitivity. Nevertheless, these results suggest that the 388 AAV-Mbp-CreER T2 injection decreased differentiation mechanisms in Mbp-expressing oligodendrocytes 389 validating the MYRF knockout. 390 To test whether ACC-specific Myrf knockout is required for learning and short-term memory, 391 another cohort of Myrf +\+ mice and Myrf flox\flox littermates underwent similar viral injections and TAM 392 protocol but were tested at one hour after IA training. No differences between groups were observed (Fig.  393   6D). To determine the role of oligodendrocytes in the ACC on memory processes, Myrf +\+ and Myrf flox\flox 394 littermates were treated with the same viral and TAM injection protocol as above but tested for memory 395 retention at one day and seven days after training. Compared to Myrf +\+ littermates, Myrf flox\flox mice showed 396 a significant memory impairment at both time points after training (Fig. 6E). These results suggest that 397 oligodendrocytes in the ACC are necessary for memory consolidation but dispensable for memory 398 acquisition and short-term memory in mice, just as in rats. 399 To test whether oligodendrocytes is required for memory formation in the hippocampus, AAV-400 Mbp-CreER T2 was bilaterally injected into the dHC of Myrf +\+ and Myrf flox\flox littermates using the protocol 401 described above. No differences in memory retention were observed at one day or seven days post-training 402 compared to control groups (Fig. 6F), leading us to conclude that oligodendrocytes are required in the ACC 403 for memory consolidation but not for learning or short-term memory. By contrast, oligodendrocytes are 404 dispensable in the dHC for the formation of hippocampus-dependent memories. 405 by confocal microscopy two weeks after viral infection and found to be mostly confined to the ACC (Fig.  442   6B). The mice were tested one day after training. Treatment with C21 significantly impaired memory 443 retention compared to vehicle injection (Fig. 7C), suggesting that neuronal activity in the ACC is required 444 for memory formation. 445 To determine whether blocking neuronal activity in the ACC affected learning-dependent 446 oligodendrocyte lineage cells we started by investigating OPC proliferation. Toward this end, AAV-hSyn-447 hM4D (Gi)-mCherry was bilaterally injected into the ACC and fourteen days later the mice were injected 448 with either C21 or vehicle in combination with EdU one hour before receiving IA training. The mice were 449 perfused one day after training at which point immunohistochemistry was used to quantify the number of 450 cells that were positive for both EdU and PDGFRa. Trained mice injected with C21 had significantly fewer 451 cells with EdU+&PDGFRA+ cells compared to mice injected with vehicle control, indicating that OPC 452 proliferation was significantly impaired (Fig. 7D). Thus, we concluded that neuronal activity in the ACC is 453 required for learning-induced changes in OPC proliferation. 454  Our western blot analyses confirmed that levels of OLIG2 significantly increased in the ACC 484 following learning, supporting the idea that oligodendrogenesis is rapidly upregulated in this brain region 485 in response to experience. Interestingly, MBP protein levels did not change, despite a significant increase 486 in Mbp mRNA levels. This dichotomy might be due to the large pool of MBP in the brain, and it is possible 487 that relatively small changes of MBP induced by a learning may not be easily detected with western blot. 488 However, more sensitive techniques should be able to address this question. differentiation, but rather suggests that the differentiation occurs either in only newly proliferating OPCs 504 that undergo differentiation or in small cell populations that require more sensitive detection approaches. 505 In fact, the EdU+ cell population accounts for a small number (< 0.5%) relative to the DAPI+ cell number 506 of the ACC to which the quantifications were normalized. Even the analysis of the number of 507 EdU+&OLIG2+&CC1+ cells relative to the total number of OLIG2+ cells in the ACC confirmed that the 508 learning-dependent increase in OPC differentiation is detectable in the EdU+ but not in EdU-cell 509 population. The EdU+ population is however less than 8% of OLIG2+ cells, again raising the question of 510 whether learning-dependent changes occur either mostly in the proliferating OPCs, which rapidly proceed 511 to differentiation, or also in other cell populations that however are too small to be detected with abundant 512 markers of oligodendrogenesis stages. 513 Our results also extended previous findings on motor, spatial, and contextual fear memories by 514 showing that brain-wide disruption of oligodendrogenesis impairs novel object location memories, 515 strengthening the conclusion that oligodendrocyte lineage cells are required for long-term memory 516 formation. Furthermore, by using multiple genetic and molecular approaches in rats and mice targeting 517 specific brain regions of interest, we provided evidence that Myrf expression is required in the ACC but not 518 the dHC, confirming the data across species. Thus, only certain brain regions in a given memory system 519 recruit Myrf for memory consolidation. OPCs prevents the formation of spatial and contextual memories. These studies showed that water maze 522 and contextual fear conditioning learning in mice rapidly induce oligodendrocyte precursor cell (OPC) 523 proliferation and differentiation into myelinating oligodendrocytes in cortical regions such as the ACC and 524 medial prefrontal cortex (mPFC), but not the hippocampus. They suggested that myelin remodeling 525 following training might be restricted to brain regions associated with long-term consolidation of 526 hippocampus-dependent memories. However, because these studies utilized a brain-wide OPC knockout 527 approach, they could not determine whether oligodendrogenesis in specific brain regions is required for 528 memory formation. Identification of region-and circuitry-specific requirements for oligodendrogenesis 529 and/or myelination in different types of learning and behavioral stimuli is important because it offers critical 530 knowledge for better understanding the role of myelin in healthy brain functions as well as in diseases. Such 531 knowledge should also expand our understanding of the circuitry that supports responses to learning. 532

Why oligodendrocyte lineage cells are required in the ACC but not the hippocampus is an open 533
question; one possible explanation is that they may subserve long-term changes required for memory 534 storage. In cortical regions including the ACC, but not in the hippocampus, episodic and spatial memories 535 are stored for the very long term via systems consolidation, a process that requires time (Dudai et al., 2015). oligodendrocytes may also be involved in long-term memory formation via their abilities to regulate 566 synaptic plasticity (Zemmar et al., 2018), neuronal excitability through extracellular potassium clearance 567 (Larson et al., 2018), and trophic signals to neurons (Du & Dreyfus, 2002). 568 Our findings showing that changes of differentiation markers colocalized with markers of 569 proliferation one day following IA learning are in line with previous studies showing that OPC 570 differentiation is observed four hours after motor skill learning (Xiao et al., 2016) and required one day 571 after water maze training (Steadman et al., 2020). Collectively these data suggest that the activation and 572 recruitment of oligodendrocyte lineage cell mechanisms following learning are relatively rapid. Within a 573 24-hour temporal window we also detected a significant learning-dependent increase in Mbp mRNA in the 574 ACC, which prompted us to test whether oligodendrogenesis and/or more mature oligodendrocytes played 575 a critical role in memory. While the majority of Mbp expression is found in mature oligodendrocytes, Mbp 576 is also expressed in differentiating/pre-myelinating oligodendrocytes, therefore, with the Mbp-driven 577 knockout of MYRF, we could not fully dissect which stage of Mbp-expressing oligodendrocytes are 578 critically recruited for memory formation. Nevertheless, we found that Mbp-regulated MYRF-expressing 579 cells are required in the ACC for memory consolidation, further supporting the hypothesis that not only 580 OPC proliferation but also oligodendrocyte differentiation in the ACC plays an important role in memory 581

consolidation. 582
In the present study, we also dissected the requirement for Myrf in various memory processes. We 583 found that training-induced Myrf expression in the ACC is necessary for the consolidation process but not 584 for the initial acquisition of memory (learning) or remote storage. In fact, knocking down or knocking out 585 Myrf before training did not affect short-term memory or acquisition, nor was there an effect on memory 586 when Myrf was knocked down at a remote time point. However, disruption of oligodendrogenesis after 587 training impaired long-term memory tested one day later, and the impairment persisted when the memory 588 was tested at remote time points, such as four weeks after training. The lack of an effect on memory when showed that brain-wide OPC knockout of Myrf at 25 days after water maze training did not impair memory 591 retention. From these results, we can conclude that MYRF-dependent changes in cortical regions are 592 necessary for the rapid and initial phase of consolidation, but not for learning, retrieval, or memory storage. 593 Our results also shed light on the kinetics of oligodendrogenesis requirement in recent memory 594 recall. Steadman et al. (2019) found that brain-wide Myrf conditional knockout in OPCs disrupts one-day-595 old spatial memory, and the disruption was still observed at a remote time point 28 days after training. By 596 contrast, Pan et al. (2020) reported that Myrf knockout mice trained in contextual fear conditioning (CFC) 597 had intact recent memory recall 1 day after training but impaired remote memories 28 days after training. 598 We found that brain-wide and ACC-targeted knockout of Myrf in mice as well as ACC-specific ODN-599 mediated knockdown of MYRF in rats impaired recent memories tested one day after IA training. The 600 impairments persisted in both rats and mice tested up to 28 days after training, leading us to conclude that 601 MYRF-dependent oligodendrogenesis is rapidly upregulated and engaged following learning to selectively 602 support a rapid phase of memory consolidation. It is possible that task-related differences in the kinetics of 603 MYRF requirements exist, and that CFC has a slower cortical recruitment of oligodendrogenesis relative 604 to water maze and IA tasks. It is also possible that Myrf knockout and knockdown under different conditions 605 (ASO, vs. OPCs, vs. mature oligodendrocytes) and promoters may uncover distinct contributions of 606 oligodendrocyte lineage cells to different memory phases. Knowing the role of oligodendrogenesis in 607 specific memory processes and temporal phases of memory provides valuable information for the future 608 development of temporally targeted treatments for cognitive symptoms of demyelinating diseases. 609 Finally, using a chemogenetic approach, we showed that the inhibition of neuronal activity in the 610 ACC prevents learning-induced OPC proliferation. These findings extend previous data indicating that In sum, our data support the view that activity-regulated Myrf-dependent changes in 616 oligodendrocyte lineage cells in select brain regions underlie long-term memory consolidation. We suggest 617 that these changes support the stabilization process required to store information long-term. 618  Plexiglas walls and floor measured at 12.5 (±2.5) lux in the center of a dim room. Visual cues were provided 775 within the box and on the walls of the room. Behavior was recorded with a video camera positioned above 776 the arena. Mice were first habituated to the arena for 10 minutes for 3 consecutive days before the training. 777

Materials and Methods
Twenty-four hours after the last habituation session, each animal was returned to the arena for its training 778 session. Training consisted of exposing the mice to two identical objects constructed from Mega Bloks 779 (Montreal, Canada) secured to the floor of the arena. Object sizes were no taller than twice the size of the 780 mice. Mice were initially placed facing a corner, away from the objects, and were allowed to explore the 781 arena and objects for 10 min. 4 hours after training; each animal was tested in the arena. During testing, one 782 object remained in the same location as during training, whereas the second object had been moved to a 783 novel location. Animals were placed in the arena facing the same direction as during training and were 784 allowed to explore for 10 min. The placement of the object in the novel location was counterbalanced 785 between subjects. The arena and objects were cleaned between sessions. Video files were coded and 786 scrambled. The experimenter was blind to treatment and scored the total time the mice spent actively 787 exploring each object in each session. Active exploration was defined as the mice pawing at, sniffing, or 788 whisking with their snout directed at the object from a distance of less than ∼1 cm. Sitting on or next to an 789 object was not counted as active exploration. Mice with less than 10s total exploration time were excluded. 790 If mice explored more than 15s, the exploration percentage was taken at 15s of total exploration time.

Mouse viral injections and C21 administration 802
Mice were anesthetized with isoflurane. The skull was exposed, and holes were drilled in the skull 803 bilaterally above the ACC or dHC. A Hamilton (Reno, NV) syringe with a 33 gauge needle, mounted onto 804 a nanopump (K.D. Scientific, Holliston, MA), 0.2ul microliters of the virus was injected per mouse 805 bilaterally into the ACC (+ 0.5mm anterior to bregma, ± 0.3 lateral of bregma, -2 dorsal of skull surface) 806 or 1ul per mouse bilaterally into the dorsal hippocampus (+1.7mm anterior to bregma ±1.5 lateral of bregma 807 -1.75 dorsal of skull surface) at a rate of 0.2μL/min. The injection needle was left in place for 5 min 808 following injection to allow complete dispersion of the solution and then the scalp was sutured. Meloxicam 809 (3 mg/kg) was used as an analgesic treatment after surgeries, and mice were allowed to recover for 14 days 810 before training. 811 The pAAV-Mbp-CreER T2 virus (titer: 10x13 GC/μl) was packaged into AAV-PHP.B capsid and 812 purchased from Vector Biolabs (Malvern, PA, cat# VB1545). The AAV-hSyn-hM4D(Gi)-mCherry (cat# 813 VB1545) was purchased from add gene (titer: 7×10¹² vg/mL; cat# 50475-AAV8). C21 (HB6124, Hello 814 Bio, Princeton, NJ) was dissolved in PBS pH7.4 and injected at 1mg/kg 60 min before training. After 815 behavioral experiments, mice were anesthetized with an i.p. injection of 750 mg/kg chloral hydrate and 816 transcardially perfused with 4% paraformaldehyde in PBS pH 7.4. Their brains were post-fixed in this 817 solution overnight at 4°C, followed by PBS pH7.4 with 30% sucrose for 72 h. 30 µm brain sections were 818 collected by cryosection for free-floating immunofluorescent staining. 819 820

Statistical analyses 821
Data were statistically analyzed using Prism software. The student's t-test was used to compare 822 statistical differences between two experimental groups. When more than two groups were compared, data 823 were analyzed with one-or two-way repeated-measure ANOVA followed by Bonferroni post hoc test. All Laboratory Animals and were approved by the New York University Animal Care Committees. All 844 surgeries were performed under isoflurane anesthesia and every effort was made to minimize suffering. 845