Myeloid-secreted BDNF maintains innervation of inguinal adipose in male mice

53 Innervation of adipose tissue is essential for the proper function of this critical metabolic 54 organ. Numerous surgical and chemical denervation studies have demonstrated how 55 maintenance of brain-adipose communication through both sympathetic efferent and 56 sensory afferent nerves helps regulate adipocyte size, cell number, lipolysis, and 57 ‘browning’ of white adipose tissue. Neurotrophic factors are growth factors that promote 58 neuron survival, regeneration and outgrowth of neurites from adult nerves. Peripheral 59 blood immune cells have been shown to be a source of neurotrophic factors in humans 60 and mice. Although a number of these immune cells also reside in the adipose stromal 61 vascular fraction (SVF), it has remained unclear what roles they play in adipose 62 innervation. We have demonstrated that adipose resident myeloid lineage immune cells 63 secrete brain derived neurotrophic factor (BDNF) and that deletion of this neurotrophic 64 factor from the myeloid lineage led to a ‘genetic denervation’ of inguinal subcutaneous 65 adipose tissue, thereby causing decreased energy expenditure and increased adipose 66 mass. AAV-BDNF intra-adipose injections resulted in an increase in markers of innervation in two models with demonstrated adipose neuropathy. We and others have previously shown that noradrenergic stimulation via cold exposure increases adipose innervation in the inguinal depot. Here we have identified a subset of myeloid cells that are Cx3cr1+ monocytes/macrophages expressing adrenergic receptors. The quantity of 71 these mobile immune cells increased in adipose tissue upon cold stimulation, fitting with their function to hone to sites of tissue injury and repair, and these cells also expressed 73 BDNF. We propose that these cold induced neuroimmune cells (CINCs) are key players in maintaining adipose innervation as well as promoting adipose nerve remodeling under adrenergic stimuli such as cold exposure. an increase to neuroinflammation Microglial-derived BDNF in the promotes hippocampal synaptic plasticity [40] and neurogenesis [42]. We hypothesized that myeloid lineage cells may play a similar role in adipose tissue, and generated a myeloid specific BDNF knock-out mouse model, by crossing LysM cre/- and BDNF fl/fl mice. Here we report a role for myeloid derived BDNF in the specific maintenance of adipose innervation, and identify a subpopulation of monocyte/macrophages (Cx3cr1+) that infiltrate adipose tissue in response to cold stimulation and express BDNF.

Introduction 78 In order for the central nervous system (CNS) to regulate functions of distal 79 organs, peripheral innervation of tissues needs to be maintained and properly 80 coordinated. It has been demonstrated numerous times that loss of innervation of the 81 adipose organ (by surgical or chemical means) leads to dysfunction of the tissue and 82 disrupts whole body energy homeostasis. Denervation of brown adipose tissue (BAT) 83 greatly impairs the energy expending process of adaptive thermogenesis [1][2][3][4], while 84 denervation of white adipose tissue (WAT) results in fat mass accumulation via 85 hyperplasia and impaired lipolysis [5][6][7]. Furthermore, we have demonstrated that 86 nerves in subcutaneous adipose tissue (scWAT) undergo remodeling in response to 87 pathophysiological and environment stimuli, and exhibit signs of neuropathy in 88 obesity/diabetes as well as with aging [8]. Cold exposure served to enhance neurite 89 density within the scWAT depot [8]. 90 through an influx of monocytes (undifferentiated macrophages), neutrophils, T cells, B 126 cells and mast cells [31][32][33][34][35][36]. Macrophages and their monocyte precursors are myeloid 127 lineage immune cells and comprise the highest fraction of immune cells present in 128 adipose tissue [37]. They are highly heterogeneous cells that are polarized by 129 environmental stimuli to evoke differential responses within a tissue, including secretion 130 of cytokines. 131 In a simplistic paradigm, classically activated macrophages (M1) act in a pro-132 inflammatory manner, while alternatively activated macrophages (M2) produce an anti-133 inflammatory response and initiate tissue remodeling after injury. Both M1 and M2 cells 134 retain phagocytic behavior. During obesity, M1 adipose tissue macrophages (ATMs) 135 greatly increase in number, thus contributing to a chronic state of tissue inflammation 136 [38]; and inflamed, insulin-resistant adipose tissue histology is characterized by 137 macrophage crown-like structures surrounding hypertrophic, hypoxic and dying 138 adipocytes. On the other hand, it has been suggested that cold-induced browning of 139 adipose promotes an M2 phenotype in ATMs [38], possibly promoting tissue remodeling 140 and potentially serving as a source of NFs in adipose tissue. Importantly, the 141 immunology field has uncovered that M1 and M2 designations are an oversimplification, 142 where many immune cells express markers of both subtypes and may be inter-143 converting between the two polarities, but this has never been clearly demonstrated in 144 adipose tissues. 145 Although myeloid cells, including monocytes and macrophages, from peripheral 146 blood have been shown to store and release NFs [39], it is still unclear what role these 147 immune cells play in peripheral innervation, and it remains unknown how and if adipose-148 resident immune cells are stimulated to release NFs that act locally. Microglia, or the 149 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint CNS resident myeloid cells that are most similar to macrophages, are an accepted 150 source of BDNF in the brain [40] and increase secretion of BDNF in response to 151 neuroinflammation [41]. Microglial-derived BDNF in the CNS promotes hippocampal 152 synaptic plasticity [40] and neurogenesis [42]. We hypothesized that myeloid lineage 153 cells may play a similar role in adipose tissue, and generated a myeloid specific BDNF 154 knock-out mouse model, by crossing LysM cre/and BDNF fl/fl mice. Here we report a role 155 for myeloid derived BDNF in the specific maintenance of adipose innervation, and 156 identify a subpopulation of monocyte/macrophages (Cx3cr1+) that infiltrate adipose 157 tissue in response to cold stimulation and express BDNF. 3-5 to a cage providing for socialization, in a monitored temperature and humidity-173 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. SVF from bilateral whole inguinal adipose depots was isolated as described above and 266 resuspended in degassed buffer (1XPBS pH7.2, 0.5% BSA and 2mM EDTA). Single-267 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint cell suspensions were sorted on the MidiMACS Quadro magnetic-activated cell Adult (13-15 week old) male and female control mice (BDNF fl/fl ) were cold exposed for 287 10 days in a diurnal incubator as described above. Following cold exposed SVF from 288 bilateral whole inguinal adipose depots was isolated as described above. Single cell 289 suspensions were treated with unlabeled FC receptor blocking antibody cocktail 290 (CD16/CD32) in 50ul of FACS buffer (HBSS + 5mM EDTA + 2% FCS). Cells were then 291 1 3 incubated in an "staining antibody cocktail" against 20 cell surface markers for 60min at 292 4°C in 100ul. Cells were then washed in 1ml of FACS buffer and centrifuged at 400g for 293 7min, 2 times. Cells were then resuspended and analyzed on a five-laser 30-parameter 294 FACSymphony A5 cytometer (BD Biocsiences, San Jose, USA) using DAPI exclusion 295 for cell viability. For compensation of fluorescence spectral overlap, UltraComp eBeads 296 (eBioscience, Inc.) were used following the manufacturer's protocols. FCS 3.0 files 297 generated by flow cytometry were initially processed using FlowJo Software (Tree Star, 298 Ashland, USA) for automated compensation. Standard manual hierarchical gating was 299 performed to remove debris, cell doublets, (DAPI+) dead cells from analysis before 300 gating on CD45+ leukocyte populations. In preparation for performing the 20 marker 301 tSNE leukocyte population analysis, the CD45+ population from each sample was 302 down-sampled to 3000 events to normalize cellular input between samples. Using 303

FlowJo, a concatemer of all samples was performed. An unbiased T-Distributed 304
Stochastic Neighbor Embedding (tSNE) plugin algorithm was then run using defaults 305 with 22 parameters, on the whole sample pool to obtain a multi-sample population 306 reference map. Gating of each sample and experimental group was performed to 307 generate tSNE maps for each condition. Differential cell clusters were gated and 20 308 marker histograms plots were used to predict cluster identities. 309 Antibodies used for adipose flow cytometric panel shown here: Anti-Mouse CD45-310 The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint NK1.1 BV785 (PK136), Anti-Mouse CD11c_FITC (HL3), Anti-Mouse CD80 PE (16-316 SVF from bilateral whole inguinal adipose depots was isolated as described above. For 322 cell sorting, the following 5 marker panel was used with DAPI exclusion for viability: 323 (475301). Sorting was performed on a BD™ FACSAria II™ cell sorter with SVF gated 326 on CD45 and CD11b; CD45+CD11b-represented the non-myeloid population; 327 CD45+CD11b+ myeloid fraction was gated on Ly6c, followed by CCR2 and Cx3cr1. 328 329 Gene Expression (qPCR) 330 RNA was isolated from whole tissue depots using Trizol reagent, and total RNA 331 extracted using a Zymo (Irvine, CA) kit. RNA yield was determined on a Nanodrop; 332 cDNA was synthesized using a High Capacity Synthesis Kit (Applied Biosystems, 333 Foster City, CA). Real-time quantitative (q)PCR was performed with SYBR Green (Bio-334 Rad, Hercules, CA) on a CFX96 instrument (Bio-Rad, Hercules, CA). 335 336 Histology 337 Adipose 338 All rights reserved. No reuse allowed without permission.

Neuromuscular junction immunofluorescence, imaging, and analysis 348
Soleus and medial gastrocnemius muscles were removed and fixed in a 2% PFA at 4°C 349 for 2 hours. Tissues were rinsed with 1XPBS and incubated in blocking buffer 350 (1XPBS/2.5%BSA/0.5-1%Triton) at 4°C for at least 24 hours and up to 7 days. 351 Following blocking muscles were teased, tendons and fat removed, and tissue was 352 flattened by being placed between two tightly-bound glass slides for at least 30 minutes 353 at 4°C. Tissues were transferred to fresh blocking buffer at 4°C for at least 12 hours. 354 Immunostaining of innervation with primary antibodies was performed overnight at 4°C, The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint microscope slides using Millipore mounting fluid (Burlington, MA USA; Cat. # 5013) and 363 1 1/5 coverslips then sealed and allowed to set overnight. Stained sections were 364 imaged on a Nikon Eclipse E400 epiflourescent microscope equipped with Nikon DS-fi2 365 camera. Brightness, contrast, and sharpness were adjusted in Microsoft PowerPoint. 366 Up to 100 NMJs were counted for each tissue, statistics were conducted in GraphPad 367 PRISM software (La Jolla, CA, USA) using the multiple t-tests (one-per row) function. to the treatment. After 2 weeks animals were sacrificed, inguinal scWAT depots were 384 harvested and processed for western blot analysis. In a separate study, adult (11-13 385 week old) male C57BL/6 mice were placed on a 58% HFD for 17 weeks to induce 386 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint neuropathy, as previously described [45]. At 17 weeks AAV-BDNF and vehicle were 387 administered as described above. After 2 weeks animals were sacrificed, inguinal 388 scWAT depots were harvested and processed for western blot analysis. adipose BDNF, adult male C57BL/6 mice were cold exposed and whole SVF was 420 isolated from mature adipocytes of inguinal scWAT. Gene expression analysis revealed 421 that Bdnf was almost exclusively expressed in the SVF, while other common NFs (nerve 422 growth factor; NGF, and vascular endothelial growth factor, VEGFa) showed no 423 difference of expression between SVF and mature adipocyte fractions (Fig. 1A). In 424 FACS-sorted CD11b/F4/80 double-positive macrophages from scWAT SVF, BDNF was 425 expressed and had no change with cold stimulation (Suppl. Fig. S1A). These data also 426 indicated that non-F4/80 myeloid cells could express BDNF since those were contained 427 in the negative sort that was also BDNF positive. 428 To determine the contribution of myeloid-lineage BDNF to adipose innervation, 429 we created a knock-out (KO) mouse model using Cre-Lox technology. LysM cre/mice 430 were bred to BDNF fl/fl mice to generate LysM cre/-::BDNF fl/fl (KO) animals, which lacked 431 BDNF in myeloid lineage cells. Compared to littermate controls, KO animals exhibited a 432 significant, although not complete, decrease of Bdnf in adipose SVF as measured by 433 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint gene expression (Fig. 1B). Likely additional SVF cell types also contribute to tissue 434 BDNF. Since myeloid lineage cells are expressed in the brain, and BDNF has been 435 shown to play a role in energy balance via CNS control of satiety, we also investigated 436 whether our KO model affected BDNF expression in the hypothalamus. Gene week old) male mice were cold challenged at 5°C for 4 days, KO and control animals 447 maintained similar body weight ( Fig. 2A, left panel). However, KO animals had 448 significantly increased adiposity ( Fig. 2A, right panel). As cold exposure stimulates 449 catecholamine-induced lipolysis mediated through sympathetic nerves, we next 450 investigated innervation of the inguinal scWAT depot following a 7-day cold challenge. 451 Protein expression of the pan-neuronal marker PGP9.5 was markedly reduced in 452 inguinal scWAT of male KO mice compared to littermate controls (Fig. 2B, left panel). 453 Protein expression of tyrosine hydroxylase (TH), a marker of sympathetic innervation 454 and activation, was also drastically reduced in inguinal scWAT of KO animals compared 455 to controls (Fig. 2B, right panel). Cold exposure induces expression of uncoupling 456 protein 1 (UCP1) as required for adaptive thermogenesis, due to its ability to uncouple 457 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint 0 the mitochondrial respiratory chain resulting in a proton leak and heat production. UCP1 458 is therefore a unique marker of BAT and browning in scWAT. Gene expression of Ucp1 459 in inguinal scWAT of 7-day cold exposed mice was reduced in KO animals compared to 460 littermate controls (Fig. 2C). Thyroid hormone potentiates SNS activation of 461 thermogenesis in BAT but is produced via a distinct neuronal pathway from adipose 462 sympathetic drive, and no changes in circulating thyroid hormones were observed 463 between KO and CON animals (Fig. 2D) We next sought to determine whether genetic denervation of scWAT in KO 475 animals was restricted to adipose tissue or extended to other organs. BDNF is a known 476 myokine, and muscle is an energy expending tissue. We assessed innervation of fast 477 twitch (gastrocnemius) and slow twitch (soleus) muscle in CON and KO, by 478 investigating occupancy of neuromuscular junctions (NMJs) at basal conditions in adult 479 male mice. Immunostaining of the presynaptic nerve and vesicles (neurofilament and 480 SV2, respectively), and postsynaptic acetylcholine receptors was performed to allow 481 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint 1 visualization of NMJ (Suppl. Fig. S1C, left panel). Following counts of occupied, 482 partially occupied, and unoccupied NMJs it was determined that there was no evidence 483 of neurodegeneration in the NMJ of KO animals (Suppl. Fig. S1C, right panel). In the 484 same animals we also assessed axon numbers of spinal (L5 ventral root), motor, and 485 sensory nerves through cross-section imaging (Suppl. Fig. S1D). A lower axon count 486 could reflect neuronal death, but no difference was observed between CON and KO 487

animals. 488
Myeloid cells are also present in BAT tissue. We therefore wanted to evaluate 489 whether a lack of BDNF in BAT myeloid cells would have an effect on this tissue's 490 function. BAT of 7-day cold exposed 12-25 week old male CON and KO mice was 491 evaluated. Protein expression of UCP1, TH, and PGP9.5 did not differ between KO 492 mice and littermate controls (Suppl. Fig. S2A), which suggested BAT function was 493 neither impaired nor was the tissue denervated. Histological assessment of BAT 494 revealed no difference in cellular morphology, or UCP1 expression (Suppl. Fig. S2B). 495 These results suggested that BDNF may not play a modulatory role in BAT innervation. 496 Indeed, when adult male C57BL/6 mice were cold exposed or treated with the 497  The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint Loss of sympathetic innervation to inguinal scWAT has been shown to increase depot 505 mass [46], however, in our genetic model of scWAT denervation, no difference in 506 adiposity was observed under basal conditions (Suppl. Fig. S1A), despite the observed 507 decrease in energy expenditure (Fig. 1D). We next metabolically challenged CON and 508 KO mice with a 45% high fat diet (HFD). Adult (25 week old) male CON and KO mice 509 were placed on a HFD for 3-11 weeks, to assess adipose integrity and energy balance. 510 At 3 weeks of HFD feeding animals were characterized in metabolic cages. HFD 511 resulted in only a slight decrease in energy expenditure in KO mice compared to 512 littermate controls (Fig. 3A, top panel). However, KO mice showed a higher RER than 513 CON animals during the light cycle, indicative of preferential metabolism of 514 carbohydrates over lipids for fuel (Fig. 3A, bottom panel). These data fits with studies 515 demonstrating that adipose nerves are important for lipolysis [47] and that denervation 516 would shift fuel preference to carbohydrates. These physiological differences between 517 CON and KO animals were observed despite no difference in food intake or change in 518 body weight (Fig. 3B, top and bottom panels). After 6 weeks of HFD, KO mice 519 displayed altered glucose control compared to CON animals (Fig. 3C), perhaps due to a 520 shift in fuel utilization. By 11 weeks of HFD feeding, KO animals displayed greater 521 adiposity than littermate controls (Fig. 3D). Taken together, the challenge of HFD 522 feeding exacerbated fat accumulation in inguinal scWAT of KO animals and altered 523 glucose utlization. 524 525

AAV-BDNF delivery increases innervation of scWAT 526
To test whether BDNF is capable of promoting scWAT innervation, we used adeno-527 associated virus (AAV) to deliver BDNF directly to the inguinal scWAT of two mouse 528 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint models that we have demonstrated to display adipose neuropathy [8]. The AS/Rec2 529 dual cassette vector used to deliver the BDNF transgene, transduces adipose tissue 530 efficiently while restricting off-target transduction in liver [43]. Animals received a bolus 531 injection of 1x10 10 vg of AAV-BDNF in one unilateral inguinal scWAT depot and a 532 vehicle injection in the contralateral inguinal scWAT. In BTBR MUT male mice at 16 533 weeks of age, when adipose neuropathy is prominent, protein expression of PGP9.5 534 was significantly increased after two weeks in the fat pads that received virus, 535 compared to the vehicle treated depots (Fig. 4A, left panel). No difference was seen in 536 protein expression of TH or PSD95 following AAV-BDNF treatment (Fig. 4A, right 537 panels) for this dose and duration, in this neuropathic model. Using a HFD-induced 538 model of neuropathy, AAV-BDNF delivery to unilateral scWAT similarly increased 539 markers of innervation, but displayed a different trend than the ob/ob mice. A strong 540 increase in TH and an increase in PSD95 were observed, with no significant increase in 541 PGP9.5. Taken together, these data suggest a role for BDNF in promoting total 542 innervation on inguinal scWAT. 543 544

Cold-induced neuroimmune cells (CINCs) are recruited to scWAT and express BDNF 545
After demonstrating the significance of myeloid derived BDNF to scWAT we sought to 546 determine which myeloid cells were the source of BDNF. Given their multifaceted role in 547 adipose tissue, being a source of BDNF in the brain (microglia), and the phenotype 548 observed in a myeloid-lineage KO, we hypothesized that monocytes/macrophages were 549 the leading source of BDNF in scWAT. Since previous studies indicated that BDNF is 550 increased in scWAT with noradrenergic stimulation [8], CD11b+ F4/80+ macrophages 551 were isolated from SVF of inguinal scWAT of room temperature and 5-day cold exposed 552 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint C57BL/6 adult (12 week old) male mice. Bdnf gene expression did not differ between 553 room temperature and cold exposed CD11b+ F4/80+ macrophages (Fig. 4A). Although 554 F4/80+ is considered a pan-macrophage marker, it was too broad to reveal 555 subpopulation changes, and does not effectively mark monocytes (macrophage 556 precursors), that could be infiltrating the tissue. Based on these findings we applied a 557 different approach to determining which myeloid cells are the source of scWAT BDNF. 558 Adult (12 week old) female control animals were maintained at room temperature or 559 cold exposed for 10 days. Inguinal scWAT SVF was isolated and flow cytometrically 560 analyzed using a custom antibody cocktail against immune cells. Surprisingly, cold 561 exposure did not have an effect on either M1 or M2 ATMs ( Fig 4B). Instead, the 562 greatest increase was in Ly6c+ monocytes. Both Ly6C+ CCR2+ and Ly6C+ CCR2+ 563 Cx3cr1+ populations increased in inguinal scWAT following cold exposure (Fig. 4C). 564
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint for increased expression of Bdnf compared to cold exposed non-myeloid cells (Fig. 4F). 577 Infiltration of more of these CINC cells to adipose upon cold exposure, and not an 578 increase in per-cell BDNF levels, likely underlie the increased BDNF in adipose that was 579

observed. 580
Adrβ3 gene expression in Cx3cr1+ cells confirmed the presence of 581 norepinephrine (NE) receptor on these cells indicating the potential to be responsive to 582 SNS stimulation (Suppl. Fig. S3), a likely mode for promoting BDNF release to the 583 tissue. Taken together, these data indicated that Cx3cr1+ cells are cold-induced 584 neuroimmune cells (CINCs) that have the potential to be stimulated by sympathetic 585 nerves, and express BDNF. 586 587

Discussion 588
Here we present evidence of the necessity for myeloid-derived BDNF in 589 maintaining inguinal scWAT innervation. Loss of BDNF from LysM+ myeloid cells 590 resulted in a severe decrease in total innervation of the inguinal adipose tissue as 591 measured by the pan-neuronal marker PGP9.5, and a near complete denervation of 592 sympathetic nerves (as measured by TH protein expression. We attribute this 'genetic 593 denervation' as a cause for the observed decrease in energy expenditure exhibited 594 under basal conditions, impaired thermogenic potential evidenced by lack of UCP1 595 induction and impaired lipolysis in WAT during cold stimulation, and worsened response 596 to HFD. These phenotypes fit with the known roles of adipose nerves in regulating 597 proper adipogenesis, lipolysis, thermogenesis and overall metabolic health in adipose 598

tissues. 599
All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint Diet-induced obesity results in chromic low grade adipose inflammation due to 600 infiltration of pro-inflammatory immune cells to the tissue, and ATM content increase 601 from 10-15% (in lean mice) to ~50% [31,48]. In addition, phenotypic switching of CD4+ 602

T cells and recruitment of T cell and B cells precedes macrophage infiltration, and 603
macrophage polarity is affected to exhibit a more inflammatory phenotype. 604 We initially hypothesized that anti-inflammatory (M2) ATMs were the source of 605 scWAT BDNF. However, we observed no changes in M1 or M2 populations in scWAT 606 after cold exposure. Instead we saw an increase in Ly6c+CCR2+Cx3cr1+ monocytes in 607 both male and female mice. We have named these cells CINCs since they are recruited 608 to scWAT by cold, and express BDNF. Surprisingly, it would appear that CINCs are a 609 pro-inflammatory cell type as Ly6C+ is a marker of inflammatory monocytes. Although SAMs do exhibit pro-inflammatory markers. SAMs hone to WAT in the obese state but 623 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint conversely we see CINCs hone to WAT with cold/noradrenergic-stimulation, indicating 624 the two immune populations may have opposing roles, including potential for 625 phagocytosing nerve debris or stimulating nerve regeneration respectively. Fitting with 626 this, SAMs sequester and degrade NE, while CINCs express BDNF. One could 627 speculate then, that SAMs serve to decrease SNS input to the tissue while CINCs serve 628 to increase innervation of WAT. On the other hand, Cx3cr1+ macrophages have been 629 shown to play diverse and even opposing roles in the intestines [50], a phenotypic 630 plasticity which may also be present in adipose tissue. One thing that is clear is that the 631 M1/M2 paradigm of ATM classification is an oversimplification of the functionally distinct 632 populations of macrophages active in adipose tissue. However, further markers beyond 633 Cx3cr1+ are necessary to fully understand and phenotype the rich variety of 634 macrophages in adipose. Compounding the difficulty in clearly delineating macrophage 635 populations is that macrophage activation is a dynamic process. Phenotypic switching 636 appears to be sequential, responding to microenvironment stimuli [51] and may be 637 dependent on spatiotemporal differences in tissue resident immune cell subtypes [52]. 638 We postulate that CINCs contribute to nerve remodeling under noradrenergic 639 stimulation, leading to BDNF release. Cao, et al have recently suggested that 640 sympathetic nerve plastically is dependent on cold-induced, adipose-derived NGF [53]. 641 Although they provide evidence that NGF is involved to some degree in promoting cold 642 induced sympathetic nerve density and browning of scWAT, there is insufficient 643 evidence as to the cellular source of NGF (our data indicate it is produced more by 644 mature adipocytes than SVF cells, and thus may be promoting innervation of the 645 adipocytes and not the SVF). After determining that NGF gene expression increased in 646 the first couple of days of cold exposure, Cao, et al used an NGF neutralizing antibody 647 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint to prevent NGF activity and observed decreased sympathetic nerve density in response 648 to cold. Several caveats to this study exist since cold-induced sympathetic nerve 649 density was not assessed in the whole adipose depot, and we know that regional 650 innervation patterns exist from our previous works as well as others [54]. Furthermore, 651 NGF neutralizing antibody was administered systemically, and there was no control for 652 off-target effects, including any that may have affected hypothalamic control of 653 sympathetic drive to scWAT. However, their approach in preventing TrkA receptor The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint 9 and wrote the manuscript. KLT is the guarantor of this work and, as such, had full 692 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https: //doi.org/10.1101/787234 doi: bioRxiv preprint access to all the data in the study and takes responsibility of the data and the accuracy 693 of the data analysis. 694 695

Conflict of Interest Statement/Duality 696
The authors do not have any conflicts of interest to disclose. 697 698 699 700 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint 1 The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint :BDNF fl/fl (CON) and LysM cre/-:BDNF fl/fl (KO) mice were cold exposed at 5°C for 4 days; 720 body weight and adiposity were compared between CON and KO groups (A). Body and 721 tissue weight data were analyzed by two-tailed Student's T-Test, N=5 CON, N=8 KO. 722 Protein expression of PGP9.5 and tyrosine hydroxylase (TH) in inguinal scWAT was 723 measured by western blotting from adult (12-25 week old) 7-day cold (5°C) exposed 724 WT/CON and KO male animals (B). β-tubulin was used as a loading control for 725 normalization. Data were analyzed by two-tailed Student's T-Test, N=3 WT/CON, N=4 726 KO, * denotes data that was removed from analysis due to lack of expression of loading 727 control. Gene expression of Ucp1 was measured in adult (12-25 week old) 7-day cold 728 (5°C) exposed WT/CON and KO males (C). Data were analyzed by two-tailed Student's 729 T-Test, N=5 WT/CON, N=5 KO. Circulating thyroid hormones, thyroxine (T4) and 730 triiodothyronine (T3), were measured by ELISA from serum of adult (22-23 week old) 4-731 day cold (5°C) exposed CON and KO male mice (D). Data were analyzed by One-way 732 ANOVA, with Tukey's multiple comparisons test, N=5 CON, N=8 KO. 733 Immunofluorescent staining for UCP1 was performed on inguinal scWAT sections of 734 adult (22-23 week old) male CON and KO mice following 4-day cold (5°C) exposure (E). 735 Typogen Black, used to quench lipid autofluorescence, provided staining cell 736 morphology which was visualized under brightfield microscopy. Images were acquired 737 with a 10X objective for a total magnification of 100X, and are representative of N=5 738 CON, N=8 KO. For all error bars are SEMs. 739 740 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/787234 doi: bioRxiv preprint were placed on a 45% HFD, daily food intake (represented as cumulative food intake) 752 (B, top panel) was measured for the first week of HFD feeding. Percent change in body 753 weight (B, bottom panel) was measured for the first 7 days of HFD feeding. Data were 754 analyzed by two-tailed Student's T-Test, N=6 CON, N=4 KO. Glucose tolerance testing 755 was performed at 6 weeks of HFD feeding (C). Data were analyzed by Two-way 756 ANOVA, with Tukey's multiple comparisons test, N=4 CON, N=4 KO. Adiposity was 757 measured for CON and KO animals after 11 weeks of HFD feeding as a percentage of 758 scWAT over body weight (D). Data were analyzed by two-tailed Student's T-Test, N=4 759 CON, N=4 KO. For all error bars are SEMs. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 760 0.0001. 761 762 All rights reserved. No reuse allowed without permission.

Figure 5. Cold induced neuroimmune cells (CINCs) hone in to inguinal scWAT 777
and express Bdnf. Adult (12 week old) male C57BL/6 were either maintained at room 778 temperature (RT) or cold exposed (5°C) for 5 days, ATMs from inguinal scWAT depots 779 were isolated using magnetic-activated cell sorting (MACS) by positive selection of 780 CD11b+ followed by F4/80+ cells. Bdnf gene expression in doubly labeled CD11b+ 781 F4/80+ macrophages was compared between RT and cold exposed animals (A). Data 782 were analyzed by two-tailed Student's T-Test, N=4 per group. Adult (12 week old) 783 female control animals were either maintained at room temperature (RT) or cold 784 exposed (5°C) for 10 days, SVF from bilateral inguinal scWAT was isolated and FACS 785 sorted using a 20 cell surface marker panel for myeloid lineage immune cells (B-C). 786 Changes in M1/M2 polarity (B) and Ly6C+ macrophage precursors/monocytes (C) were 787 measured between RT and cold exposed animals. Data were analyzed by two-tailed 788 Student's T-Test, N=3 per groups. Adult male (M) and female (F) control animals were 789 either maintained at room temperature (RT) or cold exposed (5°C) for 10 days, SVF 790 from bilateral inguinal scWAT was isolated and FACS sorted using a 20 cell surface 791 marker panel for myeloid lineage immune cells (N=5 per group). t-Distributed Stochastic 792 Neighbor Embedding (tSNE) analysis was performed to identify myeloid lineage cell 793 population changes in response to cold exposure; Ly6c+CCR2+ and 794 Ly6c+CCR2+Cx3cr1+ were identified as adipose cold induced neuroimmune cells 795 (CINCs) (D). Adult (12-13 week old) male C57BL/6 were either maintained at room 796 temperature (RT) or cold exposed (5°C) for 14 days; SVF from bilateral inguinal scWAT 797 was isolated and FACS sorted (E-F). Previously identified CINCs, Ly6c+CCR2+Cx3cr1-798 and Ly6c+CCR2+Cx3cr1+ cells were compared between RT and cold exposed animals 799 All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the author/funder.

AAV-BDNF Delivery to Adipose of HFD-Induced Neuropathy Mice
A. C.