Lipocalin-2 promotes adipose–macrophage interactions to shape peripheral and central inflammatory responses in experimental autoimmune encephalomyelitis

Objective Accumulating evidence suggests that dysfunctional adipose tissue (AT) plays a major role in the risk of developing multiple sclerosis (MS), the most common immune-mediated and demyelinating disease of the central nervous system. However, the contribution of adipose tissue to the etiology and progression of MS is still obscure. This study aimed at deciphering the responses of AT in experimental autoimmune encephalomyelitis (EAE), the best characterized animal model of MS. Results and Methods We observed a significant AT loss in EAE mice at the onset of disease, with a significant infiltration of M1-like macrophages and fibrosis in the AT, resembling a cachectic phenotype. Through an integrative and multilayered approach, we identified lipocalin2 (LCN2) as the key molecule released by dysfunctional adipocytes through redox-dependent mechanism. Adipose-derived LCN2 shapes the pro-inflammatory macrophage phenotype, and the genetic deficiency of LCN2 specifically in AT reduced weight loss as well as inflammatory macrophage infiltration in spinal cord in EAE mice. Mature adipocytes downregulating LCN2 reduced lipolytic response to inflammatory stimuli (e.g. TNFα) through an ATGL-mediated mechanism. Conclusions Overall data highlighted a role LCN2 in exacerbating inflammatory phenotype in EAE model, suggesting a pathogenic role of dysfunctional AT in MS.


INTRODUCTION
Multiple Sclerosis (MS) is an immune-mediated process in which an abnormal response of the body's immune system is directed against myelin sheath [1]. Although the etiology of MS is still unknown, a tight relationship between dysfunctional adipose tissue and MS has been observed [2]. Adipose tissue (AT) is a fat-rich tissue and is characterized by a complex cell-to-cell network involving immune cells that respond to environmental stimuli [3]. Consistent with the role of AT dysfunction in MS, several authors observed that experimental autoimmune encephalomyelitis (EAE), the most commonly used experimental murine model for the study of MS, induces w15e20% body weight loss (mainly due to the loss of adipose mass), at the disease onset [4e7]. Similar results were obtained in cuprizone-treated mice, 1 a well-established model to study demyelination in rodents [8e11]. Remarkably, a massive weight loss was also described in EAE mice fed with high fat diet [12]. Even if many efforts have been made in the last years to identify a pathophysiological role of AT in MS, to date scarce evidence is available about the role of AT in the pathogenesis of MS. Dissecting the immunogenic potential of AT in murine models of MS could shed light on new pathogenetic mechanisms of this disease. Corroborating studies have also revealed that adipocytes release a plethora of secretory factors perturbing the immuno-metabolic homeostasis and could have a role in autoimmune diseases, including MS [13e16]. Among these, a role for lipocalin 2 (LCN2) in the pathogenesis of MS is emerging [17e20]. LCN2 is a siderophore-binding protein mainly released from AT (also referred as adipokine) [16,17,21]. LCN2 is increased in several pathological settings including MS and cancer-associated cachexia [21,22]. LCN2 directly promotes muscle atrophy, activation of immune cells and sarcopenia in obese mice [23,24]. Mice overexpressing LNC2 specifically in the adipose tissue (Tg-Adipose-LCN2) showed reduced body weight, which was accompanied by lower AT mass if compared with WT mice. Furthermore, Tg-Adipose-LCN2 also increased the energy dissipating phenotype in the adipose depots [25]. Additionally, mice with global LCN2 genetic ablation (Lcn2KO) led to accelerated weight gain and visceral fat deposition with age, when compared to wild type (WT) mice [26]. In MS patients, the increased levels of LCN2 seems to contribute to neurodegeneration through myelination-dependent pathway [17,20]. In murine models of MS, total ablation of Lcn2 (Lcn2 KO) ameliorates EAE-related symptoms, suggesting that LCN2 expression in spinal cord and peripheral immune organs contribute to EAE development [18]. In EAE models, recent findings described LCN2 upregulation in MS lesions at the single-cell level {Fournier, 2023 #53). Although AT results as the greater sink of circulating LCN2 in the body, a direct role of AT-derived LCN2 in modulating inflammation in EAE mice are unexplored. Herein we aimed at exploring the AT responses at the onset of EAE and we identified LCN2 as main protein released from dysfunctional adipocytes able to exacerbates the inflammatory setting in a mouse model of MS.

Mouse models and treatments
Experimental Autoimmune Encephalomyelitis (EAE) was induced in female C57BL/6 mice (6e8 weeks old) purchased from Charles River as previously described. Briefly, mice were injected subcutaneously with 200 mL of emulsion containing 200 mg of myelin oligodendrocyte glycoprotein 35e55 (MOG 35e55 ) in Complete Freund's Adjuvant (CFA) containing 5 mg/mL of Mycobacterium tuberculosis (H37Ra strain, Difco), followed by intravenous administration of pertussis toxin (500 ng) twice (at days 0 and 2) as previously described ( [27]{Leuti, 2021 #55). Control animals received the same treatment as EAE mice without the immunogen MOG peptide (CFA). The animals were scored daily for clinical symptoms of EAE, according to the following scale: 0 ¼ healthy; 1 ¼ flaccid tail; 2 ¼ ataxia and/or paresis of hindlimbs; 3 ¼ paralysis of hindlimbs and/or paresis of forelimbs; 4 ¼ tetraparalysis; 5 ¼ moribund or death due to EAE. Intermediate clinical signs were scored adding 0.5 value. In EAE mice, first clinical symptoms (disease onset) appeared about 12e16 days post immunization (dpi) with a peak of severity at about 19e22 dpi. This stage is referred to as symptomatic or acute phase of the disease. Spleens and cells were extracted during this stage. C57BL/6 mice (Charles-River, Sulzfeld, Italy) were randomly assigned to standard cages (4e5 animals per cage) and kept at standard housing conditions with a light/ dark cycle of 12 h and free access to food and water. Beginning one week before the immunization, all animals were kindly handled every day to reduce the stress induced by operator manipulation during behavioral experiments. Cancer-associated cachexia was induced in female C57BL/6 mice (6e8 weeks old) purchased from Charles River. C57BL/6 male mice (3 m. o.) were subcutaneously inoculated in the right flank with 5 Â 10 6 Lewis Lung Carcinoma (LLC) cells in a total volume of 100 ml of sterile PBS.
Mice were maintained for 21 days under standard housing conditions and the total body weight, and the volume of tumor mass was monitored every other day. The tumor mass was measured with a caliber and the volume was calculated through formula V ¼ a*(b^2/2) where a is the major diameter and b is the minor diameter, according to [28]. The cachectic index, expressed as percentage of body weight loss, was calculated through the formula C.I. ¼ [(iWt e fWt þ fTW þ cWI)*100]/ (iWt þ cWI), where C.I. ¼ cachectic index; iWt ¼ initial weight of treated animal; fWt ¼ final weight of treated animal; fTW ¼ final tumor weight; cWI ¼ mean control animals weight increase, according to [29]. Mice were euthanized through cervical dislocation at 21 d. p.i. or when C.I. reached 15%, to reduce animal sufferance. Animal experiments were performed according to the Internal Institutional Review Committee, the European Directive 2010/63/EU and the European Recommendations 526/2007, and the Italian D. Lgs 26/2014. Global Lcn2 KO C57BL/6 mice were provided by Dr. Shizuo Akira from Osaka University, Japan. These mice were then backcrossed for 8e10 generations with a C57BL/6 strain to create homozygous animals [30]. The genomic DNA was extracted and PCR was performed to verify the absence of Lcn2 gene in these animals. EAE was induced in global Lcn2 KO mice as described above in mouse model section. Adipo Lcn2 KO mice were generated was previously described [31]. For all studies, the Lcn2-floxed (Flox) littermates served as controls in experiments on AdipoLKO.

Lipolysis detection assay
Media were collected after the treatments, centrifuged at 400 g for 5 min to remove dead cell and debris, and used for glycerol quantification. Glycerol assay kit (ThermoFisher Waltarm, MA, USA) was used according with manufacturer's instructions.

Nucleus and cytoplasm fractionation
For nucleus and cytoplasm extraction, 3T3-L1 cells were lysed in 1 mL of extraction buffer (EB; Hepes 1 mM, 0.5% NP-40, KCl 10 mM, MgCl 1.5 mM, Sucrose 250 mM and protease and phosphatase inhibitors from Sigma-Aldrich) and incubated in ice for 15 min, then centrifuged at 1500 g for 5 min at 4 C. Supernatant (cytosolic faction) was collected and pellet (nuclei) were washed twice with 1 mL of EB and centrifuged at 1500 g for 5 min at 4 C. Nuclei were resuspended with 200 ml of TBS buffer (NaCl 137 mM, KCl 2.7 mM, Tris-base 24,7 mM, 0.1% SDS). After protein dosage with Lowry protocol, protein where denaturized and loaded on polyacrylamide gels.

Bulk RNA-sequencing and data integration
The adipose tissue samples were subject to RNA-sequencing using an Illumina NextSeq550 and the indexed libraries were prepared from 1 mg-purified RNA with TruSeq-stranded mRNA (Illumina) Library Prep Kit according to the manufacturer's instructions. The quality of the single-end reads was evaluated with FastQC v.0.11.5 (https://www. bioinformatics.babraham.ac.uk/projects/fastqc/). All the fastqc files were filtered to remove low-quality reads and adapters with Trimmomatic v.0.3671. The resulting reads were mapped to the Mus musculus genome (GRCm38) with HISAT2 v.2.1.072 using default parameters, while Stringtie v1.3.4d73 was applied to the BAM files obtained with HISAT2 to generate expression estimates and to quantify the transcript abundance as transcripts per kilobase per million of mapped reads (TPM). The count matrices generated by Stringtie were imported in R, where differential expression analysis was performed using the Deseq2 package to compare the two different conditions. The functional annotation was performed through the AnnotationDbi R library (http://bioconductor.org/packages/release/bioc/html/Annotation Dbi.html). Functional enrichment analysis was performed by Funrich v3.0 tool (http://funrich.org/index.html) and David 6.8. Data from gene expression omnibus (GEO) were analyzed by GEO2R and integrated by Venn diagram.

Real-time PCR
Total RNA was extracted using TRI ReagentÒ (Sigma-Aldrich). RNA (3 mg) was retro-transcripted by using M-MLV (Promega, Madison, WI). qPCR was performed in triplicate by using validated qPCR primers (BLAST), Applied BiosystemsÔ PowerÔ SYBRÔ Green Master Mix, and the QuantStudio3 Real-Time PCR System (Thermo Fisher, Whaltam, MA, USA). mRNA levels were normalized to Rpl8 mRNA, and the relative mRNA levels were determined through the 2ÀDDCt method. The nucleotide sequences of the primers employed for mouse samples in the qPCR were as follows: 2.8. Analysis of mitochondrial ROS production and lipid peroxidation Differentiated 3T3-L1 cells were treated with TNFa and DMF and after 24 h were stained using MitoSOX probe (ThermoFisher Waltarm, MA, USA) for 30 0 at 37 C. For Lipid peroxidation analysis, differentiated 3T3-L1 cells downregulated for LCN2 and treated with TNF for 24 h were stained with BODIPY C11 (ThermoFisher, Waltarm, MA, USA) for 30' at 37 C. Flow cytometry analysis was performed using AmnisÒ CellStreamÒ flow cytometry (Luminex company). All Flow cytometry analyses were performed with AmnisÒ CellStreamÒ software program (Luminex Company).

Fluoromyelin staining
The spinal cord sections were rehydrated in PBS before being incubated with fluoromyelin green fluorescent myelin stain (1:300; Invitrogen, Eugene, OR) for 20 min at 25 C. The tissues were then washed three times in PBS for 10 min each. Images were obtained under fluorescent microscope and Image J was used to highlight the demyelination area and total white matter area in the spinal cord sections. The pixel area of each sample was calculated, and the percentage of demyelination was measured by dividing the total white matter area by the total demyelinated area.
2.10. Immunophenotyping and flow cytometry Cellular phenotypes were assessed using multiparametric flow cytometry panels containing markers to identify cell types and markers to assess activation states. The use of these markers allowed us to exclude all cells of no interest based on physical parameters (side and forward scatter) and to gate on specific cells of interest. For the immunophenotyping of sWAT, single suspension cells isolated from sWAT were stained with different panels of cell surface markers (see Table 1). For the identification of the main infiltrated leukocyte populations, total leukocytes were identified gating on CD45þ cells. Inside this gate, neutrophils were identified as Ly6G þ Ly6Clow cells and monocytes as Ly6C þ Ly6Glow cells. Ly6C-Ly6G-cells were further gated to identify CD3þ T-lymphocytes and NK.1þ NK cells. Further gating on CD3-NK.1-cells, allowed us to identify CD19þ B-lymphocytes. In another panel, cells were first gated on CD45þ cells and then on CD11bhighF4/80high to identify total antigen presenting cells, that could be further subdivided into dendritic cells (CD11c þ F4/80-) and macrophages (F4/80highCD11c-). Macrophages were further stained for the expression of M1 (anti-CD86, anti-CD40 and anti-MHC-II) or M2 (anti-CD206 and anti-CD200R) markers. Samples were acquired on a 13-color Cytoflex (Beckman Coulter) and for each analysis, at least 0.5 Â 10 6 live cells were acquired by gating on aqua Live/Dead negative cells [35,36]. For the immunophenotyping of spinal cord in adipo-Lcn2 EAE mice, after identifying as CD45 þ CD3-population, Blymphocytes were identified as CD19þ cells and neutrophils as CD11b þ Ly6G þ cells. CD11b þ Ly6G-cells were further analysed as monocytes as CD19-CD11c-cells and macrophages as F4/ 80highCD11c-cells. In another panel, CD3þ T-lymphocytes were divided into CD4þ and CD8þ subpopulations. The same gating strategy was applied by staining the single cell suspension with the panel of antibodies shown in Table 2 (Yu Wang). Samples were then acquired on an 18-color LSR Fortessa (Beckton Dickinson). All Flow cytometry analyses were performed with FlowJo software program (Treestar, Ashland, OR, USA).

Spinal cord dissociation and flow cytometry
After isolation of the lumbar spinal canal, the spinal cord sample was cut into small pieces and transferred to a shaking bath at 37 C in Dulbecco's Modified Eagle Medium (DMEM) containing 1 mg/mL collagenase Type I (Gibico TM, Waltham, MA, USA). The suspension was then filtered (100 mM) and aspired the supernatant after centrifuging at 500Âg. The pellet was resuspended in iced phosphatebuffered saline (PBS) and labelled with fluorophore-conjugated antibodies (Table 2). After removing the antibodies, the cell pellet was resuspended in PBS (within 5 nM EDTA) for subsequent analyses.
2.12. TNFa ELISA kit Whole blood samples were collected 4 h after fasting to normalized biochemical parameters. Serum from whole blood was collected and the concentrations of TNFa (R&D SYSTEM, Minneapolis, MN, US), was measured by enzyme linked immunosorbent assay (ELISA) kit according to the Manufacturer's protocols.  Original Article 2.13. Statistical analysis Data are reported as mean AE SD. Statistical analyses were performed using GraphPad Prism software version 9.5.0. Differences between groups which involved a single variable or factor were analyzed using Student T test or a one-way analysis of variance (ANOVA). In groups involving two or more factors, a two-way ANOVA test was performed. All analyses were subjected to Tukey's correction for multiple comparisons and statistical significance was accepted at p < 0.05.

EAE induces at loss and resident macrophage activation
Inflammatory status is a hallmark of multiple sclerosis (MS), a frequent autoimmune demyelinating disease of the central nervous system [37]. Although experimental and clinical data postulated a pathogenic link between dysfunctional AT and MS, the responses of AT during demyelinating conditions have never been explored. In the experimental autoimmune encephalomyelitis (EAE), the most commonly used experimental murine model for the study of MS, w15e20% body weight loss (mainly due to the loss of adipose mass) was observed about 15-days post immunization: (15dpi) [4e7]. Here we aimed at investigating the molecular and metabolic features of AT in EAE mice at the onset of disease (15dpi) and in line with other data, a significant loss of the total body weight was recorded ( Figure 1A). Next, we asked whether the reduction in the total body weight was caused by reduced food intake. To this end, food consumption was measured every 3 days up to 21-days post immunization and no differences between EAE and CTRL mice were observed (Suppl. Figure 1A). This result led us to postulate that other factors mediated the acute loss in the body weight independently to calorie consumption. In order to investigate the tissuespecific importance underlying observed body weight loss in EAE mice, we analyzed the responses of skeletal muscle and white adipose tissues. Although, skeletal muscle mass (tiabialis anterior) was reduced (Suppl. Figure 1B), any modulation in the atrophy-related genes such Atrogin1 and Murf1 was observed in EAE mice (Suppl. Figure 1C). Differently, epididymal white adipose tissue (eWAT) was totally lost in EAE mice at 15 dpi, whereas subcutaneous white adipose tissue (sWAT) mass was strongly reduced ( Figure 1B), suggesting that the significant reduction in body weight was mainly attributable to the reduction in the white adipose tissue. Based on these results, we aimed to study the metabolic and molecular signatures of subcutaneous WAT (sWAT). The histological features of sWAT showed massive fibrosis in EAE mice ( Figure 1C,D), which was associated with an increased level of fibrosis markers such as Col1A, Col3A1 mRNAs and a-SMA protein ( Figure 1E,F). In line with this, diminished levels of anti-fibrotic and proadipogenic marker PPARg were observed in sWAT of EAE mice ( Figure 1F). Fibrosis is often associated with unresolved inflammation and persistence of inflammatory stress [38]. Accordingly, increased expression of inflammatory genes such as Tnfa and Il1b was observed in sWAT of EAE mice ( Figure 1G), which was consistent with the increased circulatory levels of TNFa (Suppl. Figure 1D). Next, to better clarify the immune cells in the sWAT of EAE mice, we performed a immunophenotyping through high dimensional flow cytometry (Suppl. Figure. 1E and Figure 1H). A first analysis of the total CD45 þ cells revealed that leukocytes were significantly increased in sWAT of EAE (Suppl. Figure 1E). When dissecting the individual contribution of the different leukocyte populations, we found that CD19 þ B-lymphocytes, CD3 þ T-lymphocytes, CD11c þ CD11b þ Ly6C À dendritic cells and Ly6G þ Ly6C À neutrophils remained unchanged (Suppl. Figure 1F), whereas NK1.1 þ NK cells, Ly6C þ monocytes and F4/ 80 þ CD11b þ CD11c À total macrophages were increased in sWAT of EAE mice (Suppl. Figure 1G and Figure 1H). To further characterize macrophage inflammatory activity and polarization, we assessed the expression of key M1-like and M2-like markers and we found that sWAT of EAE mice displayed an accumulation of M1-like macrophages ( Figure 1I), characterized by higher levels of MHC-II, CD86 and CD40, and a concomitant reduction of CD206-and CD200R-expressing M2like macrophages ( Figure 1J) was observed, suggesting that innate immunity, and in particular monocytes/macrophages, promotes an inflammatory milieu surrounding adipocytes.

LCN2 is a hallmark of dysfunctional AT
To identify molecular mediators linking AT remodeling and immunological perturbations in MS, we initially applied a text-data mining strategy by analyzing the top 10 up-regulated genes (p adj < 0.05) in the spinal cord (GSE44989), dura mater (GSE37191) and choroid plexus (GSE35363) isolated from EAE mice. Through Venn diagram we identified Lcn2 as the unique overlapping gene (Figure 2A), suggesting its predictive role in the MS pathogenesis. Interestingly, a high expression level of Lcn2 was observed in sWAT at the onset of EAE ( Figure 2B,C), leading us to suppose that its induction accompanies the pathological loss of AT. To give more insight on this aspect, we generated an in vivo model of cancer cachexia by inoculating Lewis Lung Carcinoma cells (LCC) in wild type mice. Body weight and tumor mass were monitored every other day up to 21-days post injection. Tumor mass weight increased progressively with a doubling time of 3.8 g/day ( Figure 2D) starting from day 11 dpi. In parallel we observed a cachectic index (CI), expressed as percentage of body weight loss, which reached approximately 11% of increase 21 dpi ( Figure 2E). The decrease in the body weight was not associated with reduced food intake (Suppl. Figure 2A), whereas a significant reduction of sWAT ( Figure 2F) and a substantial increase of LCN2 protein level was observed ( Figure 2G). Similar results were observed in eWAT, in which LCN2 protein levels were increased concomitantly to eWAT mass loss (Suppl. Figure 2B). To corroborate the occurrence of cachectic phenotype, skeletal muscle mass and markers of muscle atrophy such as Atrogin 1 and Murf1 were measured. As reported in Suppl. Figure 2C, although the tibialis anterior weight was unaffected in LLC mice, an increased expression level of Atrogin1 was observed.
To explore the mechanisms governing Lcn2 up-regulation in cachectic AT, we performed a bulk RNA-sequencing in both EAE and LCC mice. The up-regulated genes (Log2FC > 0.6; p < 0.05) were analyzed for biological processes ( Figure 2H) and among the main representative processes, we observed inflammatory and hypoxia responses in both conditions ( Figure 2H). In order to broaden our evidence, we aimed at analyzing the cellular landscapes of AT following sepsis-induced cachexia (SIC), which causes AT loss similarly to EAE and LCC mice [39]. Consistent with cancer cachexia and EAE mouse models, an increase in macrophages as well as neutrophils and T cells was observed in AT of SIC mice ( Figure 2I). Through MacSpectrum tool we analyzed macrophage polarization index (MPI) and the activationinduced macrophage differentiation index (AMDI) [40], and we identified that AT-resident macrophages in SIC mice developed M1-like phenotype (Suppl. Figure 2D). Consistently, Lcn2 expression was abundant in neutrophils and in a subpopulation of fibroblast (Suppl. Figure 2E). To explore if Lcn2 was induced in adipocyte precursors following sepsis condition, we subclusterized fibroblasts ( Figure 2J) and a significant increase of Lcn2 expression was detected in fibroblast adipocyte precursors (FAP) ( Figure 2K). Based on these data, we asked if Lcn2 induction in FAP was part of an adaptive inflammatory response. To solve this question, we performed a gene set enrichment analysis and an increased inflammatory pathway was observed in sepsis (Suppl. Figure 2F). Overall collected data suggest that adipocyte-related Lcn2 participates as stress responsive protein during inflammatory conditions associated with cachexia-like states.
3.3. TNFa-treated adipocytes release LCN2 through redoxdependent mechanism Based on transcriptomics data of AT in EAE and cachectic mice, herein we postulated that LCN2 induction was consequential to inflammatory or hypoxic conditions in adipocytes. To test this, mature adipocytes were treated TNFa or cobalt chloride (CoCl 2 ), a chemical model to induce hypoxia in vitro [41,42]. Interestingly, both TNFa and CoCl 2 caused LCN2 release from adipose cells ( Figure 3A,B) as well as mitochondrial reactive oxygen species (ROS) production ( Figure 3C), which is consistent with the effectiveness of TNFa to promote mitochondrial damage in adipocytes ( [43]{Chen, 2010 #59)}. To give more insight about the redox control of LCN2 induction in dysfunctional adipocytes, we aimed at using dimethylfumarate (DMF), a fumaric acid ester approved for the treatment of relapsing-remitting MS, which promote Nrf2-mediated antioxidant defense [44,45]. As expected, DMF reduced the expression and release of LCN2 in TNFaand CoCl 2treated 3T3-L1 white adipocytes ( Figure 3A,B) and this response was accompanied by nuclear accumulation of Nrf2 (Suppl. Figure 2G). In accordance with the antioxidant role of Nrf2 [46], we found that DMF increased the expression levels of antioxidant genes (Suppl. Figure 2H) and diminished the production of mitochondrial ROS in TNFa-treated adipocytes ( Figure 3C). Remarkably, treatment with the antioxidant Nacetylcysteine (NAC) prevented LCN2 induction and release from TNFa and CoCl2-treated adipocytes ( Figure 3D and Suppl. Figure 2I). Next, we selectively induced mitochondrial stress by using an inhibitor of electron transport chain such as Antimycin A (AA). As expected, a significant increase of LCN2 was observed in mature adipocytes, which was effectively prevented by NAC (Suppl. Figure 2J). The collected results suggest that LCN2 take part to mitochondrial stress response in adipose cells.

Adipose Lcn2 controls TNFa-mediated body weight loss and macrophages activation
LCN2 is a protein that is particularly represented in the adipose tissue (also called as adipokine) and its global depletion protects from demyelination (Suppl. Figure 3A) and neuroinflammatory setting in the spinal cord of EAE mice (Suppl. Figure 3B). Accordingly, Lcn2 treatment activated microglial cells to levels comparable to TNFa (Suppl. Figure 3C), suggesting a key role of LCN2 in driving inflammation. Remarkably, the transcriptome of female mice overexpressing LCN2 specifically in the adipose depots (Adipose LCN2up ) showed an increased expression of Hif1a as well as of several fibrosis and inflammatory markers including Col6a1, Col5a2, Col6a3, Ccl2, S100a8, Saa3 [47]. Overall, these data suggest that adipose-derived LCN2 could be part of an autoregulatory loop linking adipocyte dysfunction to macrophage activation. To explore the role of adiposederived LCN2 in macrophage activation, we tested whether LCN2 was peculiar to mature adipocytes. To this end, we compared the levels of LCN2 in differentiated (day 8) versus undifferentiated (day 0) adipocytes and higher release of LCN2 was detected in mature adipocytes ( Figure 4A). Next, we carried out TNFa treatment on adipo-Lcn2 KD and the culture media were collected and used to treat macrophages. The analysis of inflammatory cytokines expression showed that macrophages cultured with adipo-Lcn2 KD and TNFa-treated conditioned medium showed reduced expression of pro-inflammatory M1-like markers such as TNFa, Nos2 and IL1b and concomitant higher levels of anti-inflammatory M2-like marker Arg1 with respect to macrophages cultured with conditioned medium from TNFatreated adipocytes with normal levels of LCN2 ( Figure 4B). These data corroborate the inflammatory role of adipose-derived LCN2 and draw a direct link between this adipokine, AT loss and M1-like macrophage activation. Consistent with this, was recently demonstrated that Lcn2 downregulation, limited AT loss and inflammatory cell infiltration in a pancreatic cancer cachexia mouse model [21]. Based on these data, we aimed at exploring the role of adipose-derived LCN2 in EAE mice.
To this end a mouse model where LCN2 is specifically downregulated in the adipose tissue (adipo-Lcn2 KO) was generated and the inflammatory phenotype of EAE mice was analyzed. Although we did not observe any improvement in the clinical score (data not shown), adipo-Lcn2 KO were resistant to body weight loss following EAE induction compared to control mice (CFA) ( Figure 4C). Interestingly, the spinal cord of adipo-Lcn2 KO revealed a diminished mRNA expression of monocytes/macrophages markers such as Adgre1 (F4/80), Itgam and Nos2 ( Figure 4D), as well as inflammatory cytokines and chemokines such as Tnfa, Il6 and Cxcl2 ( Figure 4D) compared to adipo-Lcn2 Flox. In line with these data, immunophenotyping of leukocytes within the spinal cord through high dimensional flow cytometry revealed a diminished number of total infiltrated leukocytes (CD45 high CD11b þ/À ) ( Figure 4E) and macrophages (CD45 high F4/ 80 þ CD11b þ CD11c À ) ( Figure 4F). Consistent with recent findings (das Neves et al., 2022), our results suggest that adipose-LCN2 participates in the AT loss and exacerbates the innate immune response during inflammatory conditions. To explore the role of LCN2 in nonhormonally-stimulated adipocytes, we downregulated LCN2 specifically in mature adipocytes (adipo-Lcn2 KD) and lipolysis markers were analyzed following TNFa treatment. Interestingly, adipo-Lcn2 KD showed a limited glycerol release from adipo-Lcn2 KD ( Figure 5A). Similar finding was observed in AT of cachectic mice [21], suggesting  Original Article a key role of LCN2 in mediating AT loss during cachectic-like conditions. Remarkably, DMF efficiently reduced glycerol release from TNFa and CoCl 2 -treated adipocytes (Suppl. Figure 3D). The significant reduction in the glycerol release was accompanied with the reduced levels of the rate limiting enzyme of lipolysis ATGL (adipose triglyceride lipase) and the phospho-active form of HSL (hormone sensitive lipase) ( Figure 5B). Recent findings revealed that ATGL inhibition in AT, rather than other lipases, protects from tumor-associated AT cachexia [48]. To demonstrate if ATGL also controls LCN2 induction during cachectic stimuli, we selectively inhibited ATGL by Atglistatin and we observed a limited induction and release of LCN2 from TNFatreated adipocytes ( Figure 5C,D). Next, we asked whether the mechanism promoting LCN2 release by TNFa belongs to the canonical lipolysis signaling pathways. To test this, mature 3T3-L1 adipocytes were treated with 3-isobutil-1-metilxantina (IBMX), a cellpermeable inhibitor of cAMP phosphodiesterases, which activates cAMP-dependent protein kinase (cAMP/PKA). Although IBMX massively increased glycerol release from mature adipocytes ( Figure 5E), LCN2 expression and release were unffected ( Figure 5F).
Remarkably, lipolysis activation by b3-receptor agonist isoproterenol increased glycerol release (Suppl. Figure 4A), whereas LCN2 induction and release LCN2 were unaffected (Suppl. Figure 4B). The collected results suggest that LCN2 released from adipose cells is part of an adaptive stress response to inflammatory mileua. Consistent with this, similar to TNFa, also IL6 strongly increased LCN2 induction and release from adipocytes (Suppl. Figure 4C). The reciprocal relationship between LCN2 manipulation and lipolysis makes it difficult to interpret the release of LCN2 as the driver of macrophage inflammation. Accordingly, several authors demonstrated that free fatty acids released from dysfunctional adipocytes elicit macrophage inflammation [49]. To demonstrate a direct role of adipose-derived LCN2 to promote the inflammatory responses in macrophages, we co-cultured TNFa-treated adipocytes with macrophages downregulating LCN2 receptor (LCN2r KD). As reported in Figure 5G, macrophages LCN2r KD showed a limited inflammatory activation when co-cultured with dysfunctional adipocytes. These data demonstrate that the inflammatory inputs are causative of LCN2 release from adipose cells, which directly activates macrophages through LCN2 receptor.

DISCUSSION
Adipose tissue (AT), a large lymphoid tissue characterized in heterogenous cell population including immune cells [50]. Factors secreted by downregulating LCN2 receptor (LCN2r KD) co-cultured with TNFa-treated adipocytes. TNFa (10 ng/mL: 6 h in serum-free media) was added to co-culture system. Data was reported as mean percentages of positive cells AE SD. Student's T test or ANOVA test *p < 0.05; **p < 0.01.
adipose tissue have an impact on immuno-metabolic homeostasis, contributing to several inflammatory-associated diseases. Interestingly, the treatment with metformin and pioglitazone, two compounds used for the management of metabolic syndrome, has been shown to increase the adiponectin/leptin ratio and improve the inflammatory profile, coinciding with decreased brain lesions in MS patients [51] and reduced cancer risk [52]. Adipocytes also release extracellular vesicles (EVs) that play a role in several pathophysiological conditions [53]. It is worth noting that EVs released by adipocytes stimulate fatty acid oxidation and migration in melanoma cells [54] and are involved in macrophage activation [55]. These findings emphasize the importance of monitoring the AT secretome to enhance disease-modifying therapies for conditions such as multiple sclerosis (MS) [56,57] and cancer cachexia [21,58]. Emerging findings revealed that adipose-derived LCN2 participates in several pathophysiological conditions including inflammation [21,25,26,31,47]. LCN2 is a prognostic marker of cancer-associated cachexia (CAC) and its downregulation improve CAC symptoms. High expression levels of LCN2 were observed in several pre-clinical models of MS as well as in human. In vitro studies demonstrated the capability of LCN2 to promote inflammatory cells activation such as microglial and macrophages [59,60]. For the first time, our study explored the responses of AT in EAE mice revealing a cachectic-like condition. We identified LCN2 as stress responsive protein released by dysfunctional adipocytes, which activates innate immunity. LCN2 downregulation improved demyelination and inflammatory status in EAE mice. Of note, when LCN2 was downregulated specifically in AT, EAE mice were protected against body weight loss and inflammatory macrophage infiltration in the spinal cord. However, despite thoroughly characterizing the effects of LCN2 on macrophage inflammation, it appears that animals lacking LCN2 in adipocytes do not show improvement in their clinical score in the EAE model. The rationale behind our findings stems from the profound impact of macrophage-driven innate immunity on the clinical scores of experimental autoimmune encephalomyelitis (EAE), particularly during the chronic phase that occurs at 25e30 dpi. The spatio-temporal distributions and contributions to disease development have revealed that monocytes/macrophages play a role in the early (presymptomatic) phase of EAE, promoting the activation of T cell-mediated autoreactivity [61] and the onset of MS-related symptoms. The significant reduction of macrophages in the spinal cord of Adipo-Lcn2 KO mice in the early phase of EAE strongly suggests that their pathogenic activity will subsequently be reflected in an attenuated chronic phase of the disease. Consequently, we anticipate that the influence of adipocyte-derived Lcn2 on the clinical score will become evident during the later phase of the disease. Similar findings reported that LCN2 downregulation, limited AT loss and tissue macrophage infiltration in a murine model of cancer cachexia [21]. In murine model of renal injury, LCN2 downregulation specifically in adipose depots, protects protected disease progression [31]. In line with these findings, female mice overexpressing LCN2 specifically in AT showed increased fibrosis and tissue inflammation [47]. Emerging findings highlighted a role of LCN2 in lipolysis and lipid peroxides production, which are markers of ferroptosis [62,63]. This is consistent with data reporting that Lcn2 expression is induced by reactive oxygen species and its downregulation protect against from oxidative stress, inflammation and ferroptosis [63e66]. Herein, we demonstrate that the release of LCN2 from adipocytes is strictly dependent on inflammatory milieu. In fact, the canonical stimulation of lipolysis by hormones was ineffective in promoting LCN2 release from adipocytes, thereby highlighting a pathognomonic function of LCN2 in dysfunctional adipocytes. In adipose cells, treatment with the Nrf2-agonist dimethyl fumarate (DMF) effectively reduced LCN2 induction following inflammatory or hypoxic stimuli. Of note DMF prevents weight loss in murine models of MS [67] and sepsis [68]. Although DMF is a first-line-treatment for relapsing-remitting multiple sclerosis (RRMS), the mechanisms of action remain still unclear. Herein we provided evidence that DMF might limit MS-related symptoms through the reduction of LCN2 production from dysfunctional adipocytes.
In conclusion, we identified adipose-derived LCN2 as a stress responsive protein, which participates in several pathological conditions exacerbating inflammatory responses and disease progression. In this scenario, obesity take part to EAE disease progression by releasing higher amount of LCN2 from adipose depots.

DECLARATION OF COMPETING INTEREST
The other authors declare no competing interests.

DATA AVAILABILITY
Data will be made available on request.