Increasing adipocyte number and reducing adipocyte size: the role of retinoids in adipose tissue development and metabolism

White adipose tissue expansion and metabolic dysfunction

Adipose tissues are highly dynamic in response to nutrient status or environmental changes [23]. In response to chronic excessive energy intake, adipose tissue expands by hypertrophy first until the size of the adipocyte reaches a critical threshold [24]. Excessive adipocyte hypertrophy leads to hypoxia, inflammation, adipocyte senescence, and death [25]. The increased production of proinflammatory adipokines by dysfunctional adipocytes, such as tumor necrosis factor-α (TNF-α), monocyte chemoattractant protein-1 (MCP-1), interleukin (IL-6), and resistin, triggers a series of inflammatory processes [26]. In obese subjects, macrophages accumulate in adipose tissue to initiate its remodeling [27–29]. Necrosis of adipocytes recruits macrophages to form crown-like structures [30, 31]. Macrophages phagocyte lipid droplets, forming lipid-laden multinucleated syncytia, which is a hallmark of chronic inflammation [30]. Macrophages are polarized into inflammatory M1 and anti-inflammatory M2 macrophages [32]. Obesity is accompanied by a transformation in the polarized states of macrophages from M2 to M1 [29], which correlates with chronic low-grade systemic inflammation.

To prevent metabolic dysfunction, generating new adipocytes through adipogenesis is critical. Indeed, deletion of the master regulator of adipogenesis, Pparg, reduced fat mass due to the blockage of adipogenesis; however, the mutated mice had enlarged pancreatic islets, dramatically elevated levels of blood glucose, extreme insulin resistance, and massive fatty livers [33]. Surgically removing subcutaneous or visceral fat tissues has little effect on insulin sensitivity [34, 35]. On the contrary, when challenged by overnutrition, fat tissue removal aggravates diet-induced lipid deposition in the liver and the residual fats and systemic insulin resistance [36]. Losing of adipocytes increases the pressure of remaining adipocytes and other tissues. Lipodystrophy syndrome is characterized by ectopic steatosis, insulin resistance, and severe dyslipidemia which is primarily due to the loss of functional adipocytes [37]. Thus, compared to adipocyte hypertrophy, adipose tissue expansion through hyperplasia is healthy [6, 7]. During adipocyte hyperplasia, adipocyte progenitor cells differentiate into small adipocytes, which increase adipose tissue energy storage capacity, preventing lipid accumulation in peripheral organs. Clinically, “Metabolically Healthy Obesity” (MHO) has smaller adipocyte size than “Metabolically Unhealthy Obesity” (MUO) [38], while “Metabolically Unhealthy Normal Weight” have increased adipocyte size [39] (Figure 1). Thus, increasing adipogenesis is a way to treat obesity-related metabolic diseases [7, 40]. If we look at animals in the wild, sheep with a larger number of adipocytes in tail fat can store more energy and have improved energy homeostasis than animals with fewer adipocytes, which is an evolutional adaptation to the inconsistent food supply in extreme environments [41]. An increased number of adipocytes provides a higher capacity for energy storage.

Most adipocytes are formed during the fetal and early postnatal stages, and adipocyte hyperplasia largely ceases in perirenal fat after birth [42]. Although about 8% of human subcutaneous adipocytes are refreshed every year, the total number of adipocytes is fixed during childhood and early adulthood [43]. The adipogenic capacity of adipose tissue declines with the development of obesity [44] and aging [45]. Thus, the fetal stage and early childhood are critical periods for increasing adipocyte number and improving the energy storage capacity of adipose tissue.

Besides the adipocyte number, the expandability of each adipocyte also contributes to the energy storage ability of adipose tissues. Although the factors that determine adipocyte expandability are complex and remain to be fully understood, the extracellular matrix (ECM) plays a critical role in determining how big the size of each adipocyte can reach. Adipocytes release ECM proteins as well as the enzymes (proteinases) for ECM remodeling in response to changes in nutrient supply [46]. The relaxation of ECM allows healthy expansion of the fat pad, losing ECM plasticity leads to adverse metabolic consequences [47]. The stiffness of ECM limits the enlargement of adipocytes, which triggers pathological responses including inflammation, activation of stress-related pathways, and ectopic lipid deposition in other tissues [2]. A variety of collagens accumulate in the epididymal WAT of db/db mice [48]. MT1-MMP, a neutral endopeptidase that cleaves collagenous proteins, is required to modulate pericellular collagen rigidity during adipogenesis. Collagen VI is the predominant collagen produced by adipocytes. Depletion of the Col6a1 gene in leptin-deficient ob mice leads to larger adipocytes but reduced inflammation and improved lipid and glucose homeostasis [49]. Increasing the plasticity of ECM is a promising way to raise the threshold of adipocyte size, but knowledge in this area is currently very limited. However, great progress has been made in preventing adipocyte hypertrophy by increasing energy expenditure.

Brown adipose tissue and energy expenditure

The best way to prevent excessive expansion of adipocytes is to increase energy expenditure, in which BAT has a prominent role. BAT only exists in eutherian mammals [50] and the thermogenic function of BAT is an evolutionary advantage for animals to survive during cold [5]. Initially, BAT is only found to be abundant in infants [51], which prevents hypothermia in neonates [52]. However, PET-CT discovered the existence of metabolically active BAT in the supraclavicular and spinal regions of adult humans [53]. Subsequent studies showed that cold exposure activates BAT thermogenesis and increases BAT mass in both rodents [54, 55] and humans [56, 57], which accounts for substantial amounts of basal energy consumption [58]. As estimated, BAT dissipates 25 to 100 kcal/day energy [24]. The volume and function of BAT vary in different human populations, individuals of south Asian origin have a lower BAT volume than that of white Caucasians and they have a very high risk of developing type 2 diabetes compared with the latter [59].

BAT thermogenesis is inducible by cold exposure or dietary bioactive compounds, which are respectively called cold-induced thermogenesis (CIT) and diet-induced thermogenesis (DIT). BAT and skeletal muscle-based thermogenesis are equally recruited during acute cold adaptation [60] but 4 weeks of cold acclimation in adult humans shifts uncoupling thermogenesis from skeletal muscles to BAT [61]. As for DIT, depending on the composition and amount of the diet, it comprises 5–15% of the daily energy expenditure [62, 63]. BAT is highly innervated [64], and its thermogenesis is controlled by the sympathetic nervous system (SNS). Cold exposure increases norepinephrine release from SNS to promote brown adipocyte hyperplasia or beige adipocyte generation in WAT [65–69]. Norepinephrine activates brown or beige adipogenesis through the β-adrenergic receptors (βARs) in adipocytes [4, 65, 70], which activates p38 MAPK signaling [71], a major stress signaling pathway [72]. P38 pathway induces expression of PRDM16 and brown adipogenesis [73], Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PPARGC1A or commonly called PGC1α) [74] increases the transcriptional activity of PPARγ and stimulates the expression of Ucp1[75–77]. PRDM16 drives both beige [78] and brown [79] adipocyte formation. As a large zinc finger-containing transcriptional factor, PRDM16 binds to and modulates the activity of other adipogenic transcriptional factors, including PPARγ [79], c/EBPβ [80], PPARα and PGC1α [81].

Cold therapy has been clinically approved to be an effective strategy to increase energy expenditure. Wearing a cold water-conditioned tube suit (18 °, 4.5 hours a day) for about 20 days, for example, can effectively increase the energy expenditure of human beings [57, 82]. The major disadvantage that restricts its wide application is the inconvenience being suffering from the prolonged cold [83]. βAR agonists provide an alternative to cold exposure, which directly activates βAR and induces thermogenesis. However, nonspecific sympathomimetic drugs increase cardiovascular activity [84, 85]. Thus, strategies bypassing βAR activation to activate brown/beige adipogenesis are desired. A huge amount of data show that polyphenolic compounds, which are abundant in fruits and vegetables, strongly enhance beige adipogenesis and thermogenesis of rodents [86–88]. Polyphenols act as activators of AMP-activated protein kinase, a key regulator of energy metabolism [88, 89]. Polyphenols are abundant in our diet and some of them like capsaicin have been proved to increase fatty acid oxidation and energy expenditure of human beings [90].

Besides environmental stimulus, food intake itself also increases energy expenditure, which is called DIT. The same as CIT, DIT depends on norepinephrine stimulated β3-AR activation and the downstream signaling [91, 92] but DIT is mediated through uroguanylin (UGN) which is released from the intestine after a meal [93, 94]. UGN binds to its receptor guanylate cyclase C (GC-C) in the hypothalamus and activates β3-AR for BAT activation [93, 94]. DIT contributes 5%~15% of total daily energy expenditure [63]. Thus, DIT spontaneously reduces the burden of white adipocytes in no need of extra stimulation. Variation in DIT could be a primary determinant of energy balance for individuals living a sedentary lifestyle because most people are normally stay in a thermoneutral condition when CIT is not activated. We recently found the pesticide chlorpyrifos, which is a common chemical residue on vegetables and fruits, suppresses DIT and promotes obesity related metabolic dysfunctions [95]. A decline in DIT could be an important contributor to the epidemic of obesity, more efforts are needed to understand the regulatory signaling pathways of DIT and find out nutritional and environmental factors that change the level of DIT in human beings.

Adipogenesis and its transcriptional regulations

Adipogenesis is a process by which multipotent mesenchymal progenitors differentiate into adipocytes. Adipogenesis can be roughly separated into two stages: the first step is the commitment of multipotent, fibroblast-like mesenchymal progenitors into preadipocyte, which is restricted to adipocyte lineage; the second step is the differentiation of preadipocyte into mature adipocyte [96]. In vitro, studies showed that bone morphogenetic protein (BMP)2 and BMP4 induce the commitment of C3H10T1/2 pluripotent stem cells into white adipocytes [97, 98]. BMP2/4 activate SMAD4 signaling which in turn activates PPARγ to initiate adipogenesis [97]. Zinc finger protein 423 (ZFP423) is an important transcription factor that promotes preadipocyte formation. Depending on its SMAD-binding activity, ZFP423 amplifies the BMP signaling pathway and upregulates Pparγ [99, 100]. ZFP467 also plays a role in white preadipocyte commitment as it alters osteoblast into adipocyte lineage [101].

Compared to the commitment stage, the terminal differentiation stage has been extensively studied in the 3T3-L1 cell line [102]. These studies identified PPARγ and C/EBPs as critical regulators in the terminal differentiation of white adipocytes [103]. PPARγ is necessary and sufficient for terminal adipogenesis [104]. C/EBPβ/δ is expressed in the very early stage of adipogenesis and triggers the expression of PPARγ [105], an essential transcription factor for adipogenic differentiation [104, 106]. Sterol-responsive element-binding protein-1c (SREBP-1c) is also an important regulator of adipogenesis, especially during terminal differentiation and lipid accumulation [107, 108]. It enhances adipose conversion by stimulating the generation of PPARγ ligands that in turn activate the transcriptional activity of PPARγ [109]. Consistently, SREBP-1c induces the expression of adipocyte signature genes including fatty acid synthase and fatty acid binding protein (aP2).

The two types of thermogenic adipocytes, brown and beige adipocytes, are developed from different lineages[110, 111], but share some regulators. The RNA-binding protein associates with the transcription factor Y-box binding protein 1 (YBX1) to activate BMP7 expression, BMP7, in turn, promotes brown/beige adipogenesis [64, 73, 112]. BMP4[113, 114] also induces while myostatin [115] prevents brown adipocyte lineage determination. Serum response factor (SRF)/Myocardin-related transcription factors A and B (MRTF-A and MRTF-B) signaling, which is essential for myocyte lineage determination, inhibits the brown preadipocyte commitment by modulating TGFβ/BMP pathway [116]. PRDM16 in collaboration with C/EBPβ determines the commitment of myf5⁺ cells to the brown adipocytes, loss of PRDM16 in brown adipose progenitors promotes muscle differentiation [79]. PRDM3 and ZFP516 are partners of PRDM16 which are also required for brown adipogenesis [117, 118]. EBF2 is selectively expressed in embryonic brown adipose progenitor cells, and ectopically expressing EBF2 reprograms myoblasts and fibroblasts into brown adipocytes [119]. Similar to the terminal differentiation of white adipocytes, PPARγ, C/EBPβ, and FABP4 are also involved in brown/beige adipocyte maturation [64]. Beige adipocytes can be generated by de novo adipogenesis [120] or converted from existing adipocytes [121]. ZFP423 maintains white adipocyte identity through suppression of beige adipogenesis, inactivation of Zfp423 converts mature white adipocytes into beige adipocytes [122].

White, brown, and beige adipocytes belong to different cell lineages. BMPs/ TGFβ, PRDM16, and ZFPs, are key regulators determining the fate of the stem/progenitor cells, controlling the expression of these genes could change the number and type of adipocytes and would greatly affect whole-body energy homeostasis. However, the regulatory network of the differentiation of these cells is complex and remains fully understood.

The vascular system in adipose tissue contributes to healthy adipose tissue expansion.

Adipose tissue is highly vascularized and the process of adipogenesis is spatially and temporally associated with vascular development [123]. The angiogenic activity is associated with the metabolic status of adipose tissue. In diabetic mice, adipose-derived stem cells have lower angiogenic potential than that of healthy mice [124]. Insufficient angiogenesis due to decreased angiogenic capacity was also observed in human subcutaneous adipose tissue from obese subjects [125]. Hypertrophic obesity is commonly associated with hypoxia in AT when the vasculature is unable to keep pace with AT growth [126]. Hypoxia triggers the expression of hypoxia regulators like hypoxia-inducible factor 1 alpha (HIF1α), causing inflammation, insulin resistance, and other metabolic dysfunctions [126, 127].

There exists close communication between the blood vessels and adipocytes. Adipokines leptin and adiponectin play important roles in cardiovascular functions. Adiponectin stimulates the production of nitric oxide in vascular endothelial cells [128], leading to vasodilation and increased blood flow. A proper level of circulating adiponectin and leptin protects cardiovascular functions [129, 130]. Adipocyte-derived vascular endothelial growth factor-A (VEGF-A) stimulates AT vascularization and improves metabolic health [131, 132]. It is recently found that endothelial cells communicate with adipocytes via releasing or exchanging extracellular vesicles (EVs), which contain proteins and lipids that capable of modulating cellular signaling pathways [133]. The transfer of EVs is enhanced in fasted state due to glucagon stimulation of EVs production in EC and is attenuated in obese tissue [133]. In vitro, TNFa treated 3T3L1 adipocytes secrete EVs that promotes leukocyte attachment to vascular endothelial cells [134].

The function of blood vessels in adipose tissue is not limited to providing oxygen and nutrients, or releasing regulatory molecules, but also serves as a cellular reservoir to provide adipose progenitors for adipogenesis [135]. Lineage tracing studies showed that adipose progenitors reside in the mural cell compartment of the adipose vasculature [136]. More directly, pericytes and endothelial origin cells were found to differentiate into adipocytes [137, 138]. Adipose progenitors such as PDGFRα⁺ cells can differentiate into both white and brown adipocytes [139] and PDGFRβ⁺ mural preadipocytes differentiate into white adipocytes in response to high-fat diet (HFD) [140]. A later study showed that PDGFRα/PDGFRβ signaling balance determines progenitor commitment to beige (PDGFRα) or white (PDGFRβ) adipogenesis [141]. In the subcutaneous WAT, PDGFRα⁺ lineage give rise to PDGFRβ^- preadipocytes that differentiate into beige adipocytes or PDGFRβ⁺ preadipocytes that can differentiate into white adipocytes [141]., It is still controversial whether the perivascular adipose progenitors are mural cells or adventitial fibroblasts [142–144], lineage tracing using single cell marker can’t unequivocally discriminate different cells types, for example, PDGFRβ is commonly used as a mural cell marker but it is also expressed in some adventitial fibroblasts [142]. In spite of that, blood vessels undoubtedly act as a stem cell/progenitor pool for adipocyte replenishment and expansion in mature adipose tissue [135]. Differentiation of perivascular preadipocyte into adipocyte contributes to healthy adipose tissue expansion in obesity. Researchers overexpressed PPARγ in PDGFRβ⁺ perivascular preadipocytes which leads to de novo adipocyte differentiation from these cells and protected against pathologic visceral adipose expansion in obesity [145]. Figure 2 summarises the interactions between blood vessels and adipocytes and present the adipogenic cells on blood vessels.

Given the important roles that vasculatures play in adipose tissue homeostasis, it has been considered a potential therapeutic target for obesity and metabolic diseases[146]. TNP-470, an inhibitor of angiogenesis, reduces adipose vascular density and prevents diet-induced obesity [147, 148]. However, later studies showed that TNP-470 reduces body weight loss by inhibiting food intake [149]. Despite a significant reduction in fat mass, TNP-470 treated mice exhibited glucose intolerance [149]. Subsequent studies focusing on VEGF revealed a relationship between vasculature and insulin sensitivity. Activation of angiogenesis via VEGF in adipose tissue is involved in cold-induced beige adipogenesis [150, 151]. In adipose tissue, overexpression of VEGF protects diet-induced obesity and insulin resistance [131, 152, 153]. VEGF overexpression in adipose tissue improves hypoxia and prevents metabolic disorders, whereas VEGF deletion leads to hypoxia, apoptosis, inflammation, and metabolic dysfunction in response to HFD [152]. Intriguingly, VEGF repression protects against diet-induced obesity as seen in VEGF overexpression models [154]. One explanation is that proangiogenic activity during at the early stages of the HFD challenge causes WAT browning, whereas antiangiogenic action in pre-existing adipose tissue clears the dysfunctional proinflammatory adipocytes [131]. However, the paradoxical effects of proangiogenic and antiangiogenic chemicals on obesity need more efforts to understand the underlying mechanism and explore effective vasculature-based therapy for obesity treatment.

Vitamin A biology and metabolism

Dietary vitamin A is available in the form of pro-vitamin A carotenoids from plant sources, or preformed vitamin A retinyl esters from animal tissues [9]. Both carotenoids and retinyl esters absorbed by the intestine will be converted into ROL and esterified with long-chain saturated fatty acids [9]. The liver is the major site of retinoid storage and metabolism. Hepatocytes are involved in the uptake and processing of ROL and the hepatic stellate cells (HSCs) are responsible for hepatic retinoid storage [9]. The chylomicron retinyl ester is hydrolyzed to ROL within the hepatocytes and transferred to the HSCs where it is re-esterified and stored [155]. The lipid droplets in the cytoplasm of HSCs are specialized organelles for retinoid storage [156]. The retinyl ester stored in the HSC lipid droplet is hydrolyzed to ROL and transported to peripheral tissues when the dietary retinoid is insufficient.

ROL binding protein (RBP) is responsible for the delivery of ROL from the liver to peripheral tissues, cells take up the ROL-RBP4 complex through STRA6, a multi-transmembrane domain protein that is specific for RBP [157]. As demonstrated in RBP-deficient mice, dietary ROL can be stored in the liver but can’t be mobilized to the peripheral tissues without RBP [158]. The majority (66–75%) of dietary retinoids is taken up by the liver, the remainder is directly taken up by peripheral tissues [159]. In rats, about 15% of tissue retinoids are stored in WAT [160]. Humans and mice that lack RBP are physiologically normal because the postprandial retinoid delivery pathway can provide sufficient retinoids when dietary retinoids are provided regularly [155]. ROL is transported in the form of lipoproteins, by RBP in plasma, and by cellular RBPs (CRBPs) [9]. In WAT, RBP is predominantly expressed in adipocytes while CRBP1 is predominantly expressed in stromal vascular cells (SVCs) [160].

Inside the cells, ROL converts into RAL by microsomal ROL dehydrogenase (RDH) or cytosolic alcohol dehydrogenase (ADH), which is a reversible process, and then retinal irreversibly converts into RA by RAL dehydrogenase (RALDH) or aldehyde dehydrogenase ALDH) [161]. These enzymes have several subtypes, RDH10 is identified to be essential for the synthesis of embryonic RA and is required for organ development [162]. ADH1 and ADH7 are responsible for removing excess ROL rather than RA synthesis [163]. The conversion of RAL to RA is catalyzed by three aldehyde dehydrogenases, ALDH1A1, ALDH1A2, and ALDH1A3. ALDH1A2 is involved in embryogenesis [164]. In mice and humans, Aldh1a1 is expressed predominately in WAT [165].

Transcriptional regulation by retinoic acid

RA is the most active derivative of vitamin A. It serves as a ligand for RA receptors which has 3 subtypes including RAR-α, RAR-β, and RAR-γ. RAR form heterodimers with retinoid X receptors including RXR-α, RXR-β, and RXR-γ [166], and the complex binds to RA response elements (RAREs) on target genes [167]. Depending on the availability of RA, RAR/RXR heterodimers interact with either nuclear co-repressor proteins including silencing mediators of RA and thyroid hormone receptor (SMRT) and nuclear receptor corepressor (NCoR), or with coactivators such as p160 steroid receptor co-activator (SRC/p160) family and p300/CREB binding protein (CBP) [168]. RA also binds to other nuclear receptors like PPARβ/ δ to regulate the transcription of related genes [169].

Retinoic acid regulates adipogenesis

Much of our knowledge about transcriptional activities during adipogenesis is obtained through in vitro studies using cell lines, especially 3T3L1. RA inhibits lipid accumulation in 3T3L1, C3HT1/2, and stromal vascular fraction (SVF) cells isolated from murine WAT [13]. RA blocks terminal white adipogenesis by upregulating the expression of preadipocyte genes Pref-1, Sox9, and Klf2 [12], as well as inhibiting C/EBPβ mediated transcription [170]. RA also inhibits terminal white adipogenesis by interrupting the DNA methylation process and inhibiting the transcription of Zfp423 [13], a key gene for white adipocyte differentiation [99, 171] and identity [122]. During white adipogenesis, ING1 directs GADD45A to the locus enriched with H3K4me3 on the promoter region of Zfp423 [13]. GADD45A promotes Zfp423 DNA demethylation and terminal white adipogenesis [13]. By targeting RAR/RXR binding to RARE on the promoter of Zfp423, RA switches the partnership of ING1 from GADD45A to RAR, reducing GADD45A available for catalyzing Zfp423 DNA demethylation [13]. Thus, RA blocks terminal white adipogenesis. A recent study identified CD142⁺ adipose stem and precursor cells (ASPCs) as an anti-adipogenic population in SVF cells, and its anti-adipogenic capacity is mediated by RA signaling [172]. Another in vitro study found that RA suppresses adipogenesis of 3T3L1 cells when cultured in a normal glucose medium but enhances lipid accumulation of these cells when cultured in a high glucose medium [173]. However, this extracellular glucose concentration-dependent modulation of adipogenesis by RA has not been confirmed in vivo.

In contrast to its inhibition of white adipogenesis, RA promotes brown/beige adipogenesis. ZFP423 is a switcher of white and brown/beige adipogenesis. Upregulation of ZFP423 induces white adipogenesis while losing Zfp423 promotes WAT browning [122]. Thus, RA is also an important modulator of brown/beige adipocytes via suppression of Zfp423. Besides Zfp423, RA also targets Prdm16, a key regulator of both beige [78] and brown [79] adipogenesis. RA binds to RAR on two RAREs of the Prdm16 promoter and upregulates the Prdm16 expression [15]. RA also activates p38MAPK which further phosphorylates RARα and enhances the binding of RARα to the Prdm16 promoter, which further enhances the effect of RA on Prdm16 transcription [15]. Thus, RA promotes the differentiation of PDGFRa⁺ preadipocytes into brown/beige adipocytes by regulating Prdm16.

Retinoic acid regulates angiogenesis

Studies using preadipocyte cell lines or primary cells cultured on a flat surface only reflect the effects of RA on terminal adipogenic differentiation or the process of lipid accumulation. The in vitro induction of embryonic stem cells differentiation into adipocytes involves the formation of embryoid bodies in hanging media drops. RA treatment on these embryoid bodies promotes adipogenic lineage commitment before terminal adipogenesis initiated by PPARG functions [174]. Under a similar culturing procedure, SVF cells extracted from WAT form cell spheroids floating in hanging media drops or on Matrigel and develop capillary sprouts [15, 175–179] (Figure 3). RA promotes the formation of cell spheroids and capillary sprouting, a procedure that is important for later adipogenesis and lipid accumulation [15, 175]. As mentioned previously, adipogenesis is spatially and temporally associated with vascular development [123]. The 3D culture model where cells grow in hanging drops or on Matrigel undergoes angiogenesis and adipogenesis subsequently [15, 175, 179]. In vitro culture of vascularized adipose tissue mimics the physiological conditions in vivo and is a better model for adipose studies.

RA has been long known to inhibit carcinoma angiogenesis [180, 181], however, RA showed the opposite effects on normal angiogenesis. In vitro, RA stimulates angiogenesis in human umbilical vein endothelial cells (HUVECs) [182] and human retinal pigment epithelial cells (APRE-19) [183]. In vivo, RA regulates lung vascular and alveolar development by induction of VEGFA and FGF18 [184]. RA indirectly stimulates VEGF secretion by the production of reactive oxygen species (ROS) [185]. Moreover, RA increases nitric oxide (NO) production of vascular endothelial cells (ECs), which may contribute to vascular endothelial function [186]. In adipose tissue, RA activates RAR that binds with the Vegfa promoter to upregulate Vegfa transcription [15]. As a major activator of angiogenesis, VEGFA promotes adipose tissue vascularization and increases the population of PDGFRa⁺ adipose progenitors [15]. During RA-induced WAT browning, VEGFA is required for PDGFRa⁺ adipose progenitor replenishment as well as the differentiation of PDGFRa⁺ cells into beige adipocytes [15]. Not only in adult animals but also in fetuses, maternal vitamin A or RA treatment increases adipose capillary density and the population of PDGFRa⁺ adipose progenitors [15]. PDGFRα⁺ adipose progenitors can differentiate into both beige and white adipocytes [139]. Thus, at a particular stage of adipose development, RA can increase the commitment of both white and brown/white adipocytes. However, the underlying mechanism of RA-inducing adipose commitment needs further investigation.

Retinoids regulate the energy expenditure of adipose tissues

As discussed above, RA promotes the commitment of stem cells to adipose progenitors. By regulating Zfp423 and Prdm16 transcription, RA switches lipid-accumulating white adipocytes to thermogenic brown/beige adipocytes. Besides that, RA also increases energy expenditure by regulating genes involved in lipid metabolism and mitochondria uncoupling. RA activates both RA receptors (RARs) and peroxisome proliferator activator receptor delta (PPARδ) in mature adipocytes to enhance lipid oxidation and energy dissipation [187] which reduces lipid accumulation and leads to weight loss [12]. RA upregulates Ucp1 expression through RARs in murine white and brown adipocytes [188]. RA showed no effect on UCP1 expression of the human adipocyte cell line or primary white adipocyte [188] but upregulated the expression of UCP1 during adipogenesis of human adipose-derived stem cells [189].

Overall, RA inhibits lipid accumulation in white adipocytes and promotes thermogenesis by increasing and activating brown/beige adipogenesis and thermogenesis, which prevents adipocyte hypertrophy and adipose metabolic dysfunction. Figure 4 summarizes the mechanism by which vitamin A/RA prevents metabolic dysfunction.

Retinoids-related proteins and adipose tissue metabolism

The roles of retinoid transporters and metabolic enzymes on obesity have been extensively studied. ROL is transported in the form of lipoproteins, by ROL-binding proteins (RBPs) in plasma [9]. Serum RBP4 level is correlated with insulin resistance in obesity and type 2 diabetes [190–192]. RBP4 overexpression leads to insulin resistance whereas its deletion increases insulin sensitivity [190]. However, several subsequent studies proved that RBP4 is not associated with diet-induced changes in insulin sensitivity [14, 193, 194]. A possible reason for these discrepancies is that ROL-bound RBP4 (holo-RBP4) and free RBP4 (apo-RBP4) were not separately analyzed. Apo-RBP4 (ROL-free RBP4) triggers inflammation and insulin resistance [195, 196]. Obesity is associated with higher apo-RBP, but normal holo-RBP [197], and RBP4/ROL ratios are higher in the serum of insulin-resistant lean and obese humans [196, 198]. Considering the low level of retinoids in obese or diabetic subjects [199, 200], the increase of RBP4 in obese subjects may be a feedback up-regulation in response to the demand for more retinoids for peripheral tissues.

Intracellular ROL is transported by cellular RBPs (CRBPs) [9]. Deletion of CRBP1 leads to increased adiposity [201]. The accessibility of RA to different receptors is mediated by lipid-binding proteins, cellular RA-binding proteins (CRABPs), and fatty acid-binding proteins (FABPs) [202]. CRABP2 delivers RA to RARs, while FABP5 transports it to PPARβ/δ [203, 204]. Recently, cellular RA binding protein 1 (CRABP1) was found to protect mice from HFD-induced obesity by decreasing adipocyte hypertrophy [205].

Up to now, the roles of retinoid metabolic enzymes on adiposity remain controversial. ROL dehydrogenase 1 (rdh1) is the enzyme catalyzing the conversion of ROL to RAL. Ablation of rdh1 decreases UCP1, impairs BAT function [206], and causes diabesity [207]. However, the deletion of the enzyme that catalyzes the conversion of RAL to RA, ALDH1, represses adipogenesis and diet-induced obesity due to RAL overproduction [208]. A later study showed that ALDH1A1 deficiency raised RAL to induce WAT browning [165]. Interestingly, RAL induces Ucp1 promoter activity by activating RAR and recruiting the coactivator PGC1α [165], similar to RA. Considering that RA is the most active retinoid, determining whether RAL or RA plays a better role in obesity prevention is of interest. Of note, the WAT browning phenotype of ALDH1A1 deficiency mice is only seen in females but not in males, in visceral fat but not subcutaneous fat [209], suggesting a sex and depot-specific difference in retinoids and adipose metabolism due to ALDH1A1 deficiency, which requires further studies. In addition, these studies used whole-body permanent ALDH1A1 knock-out mice, considering the important role of RA in adipose tissue development, ALDH1A1 mutation might have inhibited adipocyte formation at the developing stage which contributed to adipose metabolic dysfunction observed in these mice. For purpose of treating obesity, specific inhibitors of ALDH1A1 were developed, among which WIN18446 suppressed weight gain in mice fed a HFD but caused increased hepatic lipidosis and reversible male infertility [210]. Another ALDH1A1-specific inhibitor, N42, increases serum TG levels [210]. To date, the notion that preventing obesity by inhibiting RA synthesis has not been proven to be effective. Considering that ALDH1A1 not only catalyzes the dehydrogenation of retinal but also other aldehydes in the body, the results of previous studies using Aldh1a1 knock-out mice may have been interfered by other neglected factors.

In summary, although the experiments using ALDH deficiency mice raised some controversial issues, most research during the past decades reported the inhibitory effects of retinoids on lipid accumulation. Besides, retinoids contribute to adipose tissue homeostasis by increasing energy expenditure promoting healthy adipose tissue expansion. Thus, vitamin A supplementation is a promising nutritional strategy in obesity treatment.