Myeloid-resident neuropilin-1 influences brown adipose tissue in obesity

The beneficial effects of brown adipose tissue (BAT) on obesity and associated metabolic diseases are mediated through its capacity to dissipate energy as heat. While immune cells, such as tissue-resident macrophages, are known to influence adipose tissue homeostasis, relatively little is known about their contribution to BAT function. Here we report that neuropilin-1 (NRP1), a multiligand single-pass transmembrane receptor, is highly expressed in BAT-resident macrophages. During diet-induced obesity (DIO), myeloid-resident NRP1 influences interscapular BAT mass, and consequently vascular morphology, innervation density and ultimately core body temperature during cold exposure. Thus, NRP1-expressing myeloid cells contribute to the BAT homeostasis and potentially its thermogenic function in DIO.

Obesity has reached pandemic proportions in the Western world and represents a major risk factor for the development of type 2 diabetes, cardiovascular diseases and cancer 1,2 . It results from an imbalance between energy intake and energy expenditure causing an increment of fat mass in adipose tissue or liver. Adipose tissue can be classified as white adipose tissue (WAT), brown adipose tissue (BAT) or beige adipose tissue (BgAT) [3][4][5][6] . WAT participates mainly in energy storage and release 3 whereas BAT and BgAT can metabolize stored lipids to produce heat by a process known as nonshivering thermogenesis or adaptive thermogenesis 4,5 . This type of thermogenesis occurs in the mitochondria and is mediated by uncoupling protein 1 (UCP1), which facilitates the uncoupling of electrons from the synthesis of adenosine triphosphate (ATP) towards heat production 4 . Obesity is correlated with an increase in WAT mass and impaired BAT activity 7,8 . It has been proposed that potentiating thermogenesis in BAT could influence weight gain in humans 9,10 .
Adipose tissues are rich in effectors of both innate and adaptive immunity 11,12 whose numbers are altered with obesity 13,14 . The contribution of the innate immune system and specifically adipose tissue macrophages to low-grade inflammation and homeostasis in adipose tissue has been well documented [15][16][17] . Macrophage accretion in adipose tissue can increase inflammatory cytokines and lead to chronic low-grade inflammation [17][18][19] . With regards to BAT, alternatively activated macrophages have been suggested to influence BAT homeostasis and thermogenesis 20 . However, these findings were recently challenged 21 raising questions about the involvement of macrophages in BAT activation.
While investigating the role of innate immunity in obesity, we previously identified myeloid-resident neuropilin-1 (NRP1) as necessary for healthy weight gain and maintaining glucose tolerance in obesity 22 . NRP1 is a single pass transmembrane receptor that binds multiple ligands (e.g. semaphorins, VEGFs) and receptors (e.g. VEGFR2, integrins, plexins), influences intracellular signaling 23 , axonal guidance 24 and neuronal and vascular development 25 . Adipose tissue macrophages (ATMs) devoid of NRP1 are less efficient at internalizing lipids and shift their metabolism towards a more pro-inflammatory carbohydrate-based glycolytic metabolism 22 . Here we investigated the role of NRP1 expressing myeloid cells on BAT biology and the control of thermogenesis.

Results
Interscapular brown adipose tissue macrophages express high levels of neuropilin-1. In order to investigate the transcriptomic signatures of interscapular brown adipose tissue (iBAT) macrophages compared to resident macrophage populations of other tissues, we analyzed the transcriptomes of fluorescence-activated cell sorted (FACS) native macrophages across various organs 26 . We ran Gene Set Variation Analysis (GSVA) to identify enriched gene sets in iBAT-resident macrophages (iBAT-MPs) compared to macrophages from other tissues. We found enrichment in distinct gene sets across all macrophage populations (Fig. 1A, Supplemental Table S1). Six differential gene expression analyses by DESeq2 were performed between iBAT-MPs and all the other tissue macrophage populations; we identified 624 significantly upregulated and 375 significantly downregulated genes in a minimum of 5 out of 6 comparisons that were associated with iBAT-MPs (Fig. 1B, Supplemental Table S2-S3). Interestingly, by analyzing Gene Ontology (GO) biological processes, we found transcript enrichment in iBAT-MPs for cell migration involved in vasculogenesis, semaphorin-plexin signalling pathways and positive regulation of cell migration involved in sprouting angiogenesis (Fig. 1C, Supplemental Table S4). In order to identify specific genes of interest associated with the iBAT-MP function, we determined the most recurrent genes within the top 15 Gene Ontology (GO) biological processes. Of interest, the second-most frequently associated gene with the iBAT-MP enrichment signature was neuropilin-1 (Nrp1) (Fig. 1C-inset, Supplemental Table S5), a gene which we had previously demonstrated to play a critical role in white adipose tissue macrophage function 22 . Therefore, we assessed Nrp1 expression across screened macrophages and found iBAT-MP to express the highest levels of Nrp1 (Fig. 1D). These data raise the possibility that NRP1 may play a role in the function of iBAT-resident macrophages.
Neuropilin-1-expressing myeloid cells influence interscapular brown adipose tissue composition in diet-induced obesity. BAT mass and activity are reported to significantly decrease with obesity, age and in diabetic patients [27][28][29] . Given that loss of BAT function can impact the accumulation of body fat, we investigated the effect of a myeloid-specific knockdown of Nrp1 in an experimental model of obesity. Eightweek-old LysM-Cre:Nrp1 fl/fl mice or LysM-Cre:Nrp1 wt/wt control littermates were fed either a high-fat diet (HFD; 60% fat calories) or a matched regular diet (RD; 10% fat calories) for 12 weeks ( Fig. 2A). Consistent with our previous work 22 , HFD-fed LysM-Cre:Nrp1 fl/fl mice gained significantly more weight when compared to control LysM-Cre:Nrp1 wt/wt mice (Fig. 2B, C) despite not increasing food intake (Supplemental Figure 1A-B). Following diet-induced obesity, we analyzed distinct adipose tissues: (1) eWAT, which consists of a bilateral intraabdominal visceral depot attached to the epididymis, (2) interscapular brown adipose tissue (iBAT), which is localized between the scapulae and (3) subcutaneous adipose tissue (iWAT), located between the skin and the muscle fascia anterior to the lower segment of the hind limbs (Supplemental Figure 1C). To determine the degree of adiposity, we measured the total weight of eWAT, iBAT and iWAT fat pads ( Fig. 2D-F), and in order to determine if weight gain has an effect on adiposity, we normalized these values to whole-body weight (Supplemental Fig. 1D-G). At the beginning of the diet paradigms (0 weeks), LysM-Cre:Nrp1 fl/fl and control mice showed similar weights of total iBAT (Fig. 2D) while a slight increase in baseline eWAT and iWAT weight was noted in LysM-Cre:Nrp1 fl/fl mice (Fig. 2E,F). When the data were normalized to whole-body weight, only eWAT weight was increased (Supplemental Figure 1D-F). Surprisingly, after 12 weeks of diet, LysM-Cre:Nrp1 fl/fl mice on HFD showed significantly higher total or normalized weights of both iBAT and iWAT when compared to LysM-Cre:Nrp1 wt/wt mice, whereas eWAT showed a trend, but was not significant, despite eWAT weight being significantly higher after 4 weeks of diet in LysM-Cre:Nrp1 fl/fl mice ( Fig. 2D-F, Supplemental Figure 1D-M). Fat pad weights did not vary between genotypes on RD at this later time point (Fig. 2D-F). Consistent with this, histological assessment of iBAT with H&E staining revealed that HFD-fed LysM-Cre:Nrp1 fl/fl mice showed signs of hypertrophy in iBAT when compared to LysM-Cre:Nrp1 wt/wt mice (Fig. 2G).
Adipose tissue vasculature plays an essential role in nutrient and oxygen supply to iBAT as well as in heat dissipation [30][31][32] . We therefore assessed iBAT vascularization by CD31 immunofluorescence staining on sections of iBAT from HFD-fed mice. We observed a decrease in vascularized area per analyzed section ( Fig. 2H-J). Importantly, when we accounted for iBAT hypertrophy (iBAT weight), we observed a significant increase in total vessel area and total vessel length in iBAT from LysM-Cre:Nrp1 fl/fl mice as would be expected from an increased organ size (Supplemental Figure 1N-O). We also assessed blood vessel lacunarity, a morphological measure pertaining to gaps and heterogeneity, and detected a significant increase in vascular lacunarity per section, in iBAT from LysM-Cre:Nrp1 fl/fl mice (Fig. 2K). Brown adipose tissue is also highly innervated by sympathetic nerve fibers, which regulate thermogenesis, and changes in innervation can influence lipid storage in iBAT 26,[33][34][35] . We therefore used the iDISCO method adapted for adipose tissue 36 to clear iBAT from HFD-fed mice and stained for tyrosine hydroxylase (TH) to visualize sympathetic axons and evaluate the impact of myeloid specific Nrp1knockdown on sympathetic innervation. 3D reconstructions of 4-5 randomly selected regions in iBAT were performed to evaluate sympathetic fiber length (Fig. 2L). While we detected a decrease in fiber length in LysM-Cre:Nrp1 fl/fl mice (Fig. 2M), when we normalized fiber length to total iBAT weight to account for the effect of iBAT expansion on innervation, we observed a significant increase in the amount of fibers as expected for a larger tissue (Supplemental Figure 1P). We also evaluated fiber density and we detected a significant decreased fiber density in LysM-Cre:Nrp1 fl/fl mice (Fig. 2N), due to adipocyte hypertrophy. Therefore, the observed decrease in vascularization and sympathetic innervation length in iBAT of LysM-Cre:Nrp1 fl/fl mice is not attributed to a decrease in total vasculature or innervation, but rather to tissue expansion, given the increased size of the adipocytes. These data indicate that upon HFD, NRP1-expressing myeloid cells influence iBAT mass by inducing hypertrophy of the tissue, resulting in an expansion of vascular networks and sympathetic innervation likely due to tissue stretch. Compared to controls, the resulting iBAT in LysM-Cre:Nrp1 fl/fl mice exhibit irregular vascular morphology and innervation of reduced density.    (Fig. 3A). Upon return to room temperature, the core temperature of LysM-Cre:Nrp1 fl/fl mice took slightly longer to normalize relative to controls (Fig. 3A). These data suggest that LysM-Cre:Nrp1 fl/fl mice do not adapt to cold as well as control LysM-Cre:Nrp1 wt/wt do. Analysis of energy expenditure between HFD-fed LysM-Cre:Nrp1 fl/fl and LysM-Cre:Nrp1 wt/wt mice did not reveal significant differences between groups (Fig. 3B, Supplemental Figure 2A-C) nor did we observe discrepancies in lower locomotor activity (beam breaks) (Fig. 3C). www.nature.com/scientificreports/ Since LysM-Cre:Nrp1 fl/fl and LysM-Cre:Nrp1 wt/wt mice had a lower core temperature, yet similar food intake and levels of activity, we next sought to investigate consumption of lipids to fuel systemic metabolism. We thus measured respiratory exchange ratio (RER) during cold exposure by indirect calorimetry to determine the respiratory quotient (ratio between the volume of CO 2 produced and the volume of O 2 consumed) that indicates the substrate (e.g. lipid-derived carbohydrate) that is metabolized to supply the body with energy 37 . Consistent with our previous work 22 , RER did not vary between HFD-fed LysM-Cre:Nrp1 fl/fl and LysM-Cre:Nrp1 wt/wt mice both at room temperature and during cold exposure (Fig. 3D). We found that during HFD, lipids become the predominant source of energy or substrate in both control and LysM-Cre:Nrp1 fl/fl mice. Therefore, it is unlikely that the differences observed in core temperature during cold exposure are due to discrepancies in reduced lipid utilization between strains. Furthermore, we evaluated body composition and consistent with our previous work 22 , analysis of animals by Echo MRI showed an increase in fat mass and in percentage of total body fat mass in LysM-Cre:Nrp1 fl/fl while lean mass did not vary (Supplemental Figure 2D-E). The discrepancy in fat mass was not attributed to difference in eating patterns as average daily or total food intake and energy intake from food of LysM-Cre:Nrp1 fl/fl or LysM-Cre:Nrp1 wt/wt mice were similar after 10-12 weeks of HFD whilst housed at room temperature (Fig. 3E,F, Supplemental Figure 1A-B). Altogether, our data suggest that NRP1-expressing myeloid cells impact thermogenic capacity following cold exposure.  Figure 3D-F). We verified this by means of H&E staining and showed that treatment with CL316,243 was able to reverse the hypertrophy observed in HFD-fed LysM-Cre:Nrp1 fl/fl mice (Fig. 4F). These data suggest that iBAT from LysM-Cre:Nrp1 fl/fl mice remains responsive to β3-adrenergic stimulation, suggesting it retained the ability to induce nonshivering thermogenesis.

Discussion
In this study, we demonstrate that macrophages residing in interscapular brown adipose tissue express high levels of Nrp1 compared to other tissue-resident macrophages. Given the metabolic role of BAT during obesity, we investigated the impact of macrophage-resident NRP1 deficiency on the thermogenic function of iBAT in a HFD model of weight gain. Our findings suggest that upon HFD, myeloid-resident NRP1 influences iBAT hypertrophy, impacting vasculature morphology, sympathetic innervation density and ultimately core temperature when exposed to a cold challenge. Mouse activity as evaluated by beam breaks as well as RER showed similar rates in both strains suggesting that the reduced ability of LysM-Cre:Nrp1 fl/fl mice to maintain body temperature in a cold challenge was not due to differences in mouse activity or substrate consumption. Although no differences were detected in energy expenditure between LysM-Cre:Nrp1 fl/fl and LysM-Cre:Nrp1 wt/wt mice, we observed that iBAT from LysM-Cre:Nrp1 fl/fl mice is responsive to β3-adrenergic stimulation and hence these mice are able to induce nonshivering thermogenesis. Adipose tissue is subjected to tight immune regulation with studies having predominantly focused on immune regulation of WAT in obesity and diabetes. We previously showed that NRP1-expressing myeloid cells in WAT contribute to weight gain, insulin sensitivity and modulate metabolic homeostasis 22 . Furthermore, NRP1 deficiency in white adipose tissue macrophages as well as in peritoneal macrophages led to an increase in a polarization towards a classic pro-inflammatory phenotype 22 . With regards to iBAT, alternatively activated macrophages have been shown to play a role in the modulation of BAT thermogenesis through catecholamine synthesis 20 . However, these findings have been challenged with the finding that there are insufficient levels of tyrosine hydroxylase to synthesize relevant amounts of catecholamines in alternatively activated macrophages 21 , raising questions about the underlying signaling pathways in macrophages that could mediate BAT activation. Nonetheless, other studies showed that mutations associated with BAT resident macrophages influenced BAT innervation as well as thermogenesis 26,33,40 . The importance of innervation in iBAT has been demonstrated by denervation studies in animals subjected to cold exposure or HFD where it leads to a reduction in UCP1 expression levels, mitochondrial activity, blood flow and glucose uptake 41,42 . Moreover, in obesity, alterations in BAT vascular density can lead to hypoxia and mitochondrial dysfunction resulting in lipid droplet accumulation 43 . Given the critical role of NRP1 in myeloid-driven angiogenesis 23,44 , lipid uptake and mitochondrial lipid utilization 45 , it is possible that NRP1-expressing myeloid cells are influencing BAT homeostasis through vascular networks and local lipid metabolism. Future investigation on the discrepancies in the expression of thermogenic markers between LysM-Cre:Nrp1 wt/wt and LysM-Cre:Nrp1 fl/fl will provide insight on the role NRP1-expressing myeloid cells in iBAT function. Furthermore, while we focused on iBAT, iWAT is also reported to produce heat following browning of the tissue, hence future work could explore the contribution of myeloid cells expressing NRP1 to iWAT-mediated thermogenesis and browning.
In sum, our study suggests an indirect role for myeloid cells expressing NRP1 in BAT homeostasis during diet-induced obesity. Our data further supports the importance of vasculature 43   Bioinformatics analysis with DESeq2, GSEA, GSVA and GO enrichment data analysis. Differential expression analysis was performed with DESeq2 and used for pre-ranked gene set enrichment analysis (GSEA 4.0.3). Genes that were exclusively upregulated or downregulated in a minimum of 6 out of 7 tissues were considered to determine the iBAT macrophages-specific signature. GSVA analysis was performed in R (v3.6.0.). Gene ontology (GO) analysis was performed with PANTHER classification system 49-51 using the GO aspect: biological processes. Biological processes with a with FDR-adjusted P value (< 0.05) were considered as significant.
3D fluorescence-imaging. Collected adipose tissue was processed as previously reported 36  3D imaging acquisition and quantification. The optically cleared adipose tissues were imaged on a LaVision Light Sheet Ultramicroscope equipped with an Andor Neo sCMOS camera and a MVPLaPO 2x/0.50 objective lens and a 6 mm with medium length dipping cap and a 4 mm with long dipping cap. Version v328 of Imspector Microscope controller software was used. The whole view of the samples was imaged at a 0.63X effective magnification (1.26 × zoom) and scanned by two lightsheets by the left or right side with a step size of 3 µm. For imaging at 4X effective magnification (8 × zoom) and scanned by two lightsheets by the left or right side with a step size of 2 µm. The image stacks were acquired by continuous lighsheet scanning method without the contrast-blending algorithm. Imaris v64 (9.1.2), Imaris software available at: http:// www. bitpl ane. com/ imaris/ imaris was used to reconstruct the image stacks obtained from the volume imaging. The representative images of iBAT were acquired with the orthogonal perspective of the image stacks. Sympathetic fiber length and overall fiber density by total iBAT weight were obtained by multiplying the respective values by total iBAT wet weight. Histology analysis. Collected adipose tissue was fixed in Formalin 10% followed by standardized paraffinembedding. Paraffin-embedded tissues were cut into 12 µm thick sections. Samples were deparaffinized and rehydrated and stained with Harris hematoxylin and eosin (H&E), followed by dehydration and mounting with PERTEX (HistoLab Products AB). For each sample, representative DIC images were taken with a Zeiss Axio-Imager Z2 (Zeiss) with a coupled AxioCam ICc 1 (Zeiss). Zeiss Axio Vision software (Zeiss) was used for image processing and editing.
For immunofluorescence, tissues were fixed in 4% paraformaldehyde overnight at 4°C, incubated with 30% PBS-sucrose for at least 48 h at 4°C and flash-frozen in OCT with liquid nitrogen before cryostat sectioning.
OCT-embedded tissues were cut into 12 µm thick sections. iBAT sections were permeabilized in methanol and blocked in 3% BSA, containing 0.05% Tween, 0.2% Triton for 1 h at room temperature in a humidified chamber. Followed by an overnight incubation with anti-rat CD31 [#550274, BD Biosciences] primary antibody (diluted at 1:500) in 3% BSA, containing 0.05% Tween, 0.2% Triton at 4 °C. Samples were incubated with Alexa Fluor secondary antibody at RT for 2 h. For each sample, Z-stacks were taken at 10 X on Olympus FV1000 confocal microscope.
Vessel quantification. AngioTool 50 analysis was performed on Z-stacks compressed on single images taken at 10X on an Olympus FV1000 confocal microscope. CD31 staining was measured.
Vessel area, total vessel length and vessel lacunarity by total iBAT weight were calculated by multiplying the respective values per total iBAT wet weight.
Metabolic studies and comprehensive lab animal monitoring system (CLAMS). Mice were implanted intraperitoneally with sterile temperature probes (G2 HR E-mitter, Bio-Lynx) 1 week prior to measurement of core temperature. Mice were placed in the CLAMS animal monitoring system for 24 h of acclimatization period, followed by 96 h of data collection at different temperatures. After the acclimatization, mice were Scientific Reports | (2021) 11:15767 | https://doi.org/10.1038/s41598-021-95064-w www.nature.com/scientificreports/ exposed as follows: 24 h at room temperature, 48 h at 4°C and 24 h of recovery at room temperature. Indirect calorimetry, O 2 consumption, CO 2 production, RER, energy expenditure, food intake, water intake, and locomotor activity were measured during 96 h. Average food intake was determined by measuring the mean of food consumption (light and dark cycle) of LysM-Cre:Nrp1 fl/fl or LysM-Cre:Nrp1 wt/wt of mice at 10, 11 and 12 weeks of diet.

Body composition. Lean and fat mass was examined by Echo MRI (Echo Medical Systems).
β3-Adrenergic activation. Mice were injected intraperitoneally once daily for 4 consecutive days with CL316,243 at 1 mg/kg of weight. Saline was used as vehicle.
Statistical methods. All results are presented as mean ± SEM. Analysis and statistical significance was analyzed using GraphPad Prism 5.0 (GraphPad Software; www. graph pad. com) by two-way ANOVA, when comparing multiple groups, and two-tailed unpaired Student's t-test, when comparing only two groups. Statistical significance was considered when p < 0.05 and it is indicated as: *p < 0.05; **p < 0.01; ***p < 0.001. Biological experiment numbers were listed in figure legends.

Data availability
Data, material and reagent information regarding this work can be inquired upon reasonable request to the corresponding author.