The role of ADP-ribose metabolism in metabolic regulation, adipose tissue differentiation, and metabolism

In this review, Szanto et al. summarize the metabolic regulatory roles of PARP enzymes and their associated pathologies.


Brief introduction to ADP-ribose metabolism
The field of poly(ADP-ribose) polymerases (PARPs or ARTDs) has come a long way since the discovery of a nuclear poly(ADP-ribosyl)ating (PARylating) enzyme in 1963 (Chambon et al. 1963). PARPs now constitute a superfamily of at least 17 members in human that share a conserved catalytic domain (Amé et al. 2004;Hottiger et al. 2010). ADP-ribosylation is a posttranslational modification, during which the ADP-ribosylation enzymes cleave NAD + and attach the resulting ADP-ribose (ADPR) units to acceptor proteins. ADP-ribosylation is referred to as mono(ADP-ribosyl)ation (MARylation), oligo (ADP-ribosyl)ation, or poly(ADP-ribosyl)ation (PARylation), based on the number of the ADPR units added to the acceptor protein (Amé et al. 2004;Hottiger et al. 2010). Although all PARPs inherited the family name of the founding member, PARP-1, the PARP "polyenzymes" include only PARP-1, PARP-2, and the tankyrases (PARP-5a and PARP-5b) (Gibson and Kraus 2012). Other members perform only MARylation or oligo(ADP-ribosyl) ation, while PARP13 possesses no enzymatic activity ). To our current understanding, the majority of PARP activity is attributed to PARP1 (80%-85%), while the rest is largely attributed to PARP2 (Amé et al. 1999;Schreiber et al. 2002;Szanto et al. 2011). In most cases, the major acceptor of PAR is PARP1 itself (termed auto-PARylation); nevertheless, with the use of state-of-the-art proteomics a large set of PARylated or ADP-ribosylated proteins were identified and this process is termed trans-PARylation (Chapman et al. 2013;Gibson et al. 2016;Abplanalp et al. 2018;Leslie Pedrioli et al. 2018;Palazzo et al. 2018) (for a comprehensive database of ADP-ribosylated proteins see Vivelo et al. (2017).
PARP enzymes have widespread biological functions ranging from DNA repair and chromatin structure (Javle and Curtin 2011;De Vos et al. 2012;Dantzer and Santoro 2013), RNA transcription, protein translation, and degradation (Kraus and Hottiger 2013;Bai 2015), cell division, tumor biology (Curtin and Szabo 2013), immune processes (Fehr et al. 2020) metabolism, and mitochondrial biology (Bai and Cantó 2012;, oxidative stress biology, and cell death and differentiation, and aging (Mangerich et al. 2010;Burkle and Virag 2013;Fatokun et al. 2014). In this review, we focus on the metabolic properties of PARP enzymes.

PARP enzymes in metabolism
PARP enzymes impact metabolism at multiple points, exerting regulatory functions on higher order organismal and basic cellular processes. From another perspective, PARPs impact both central and peripheral metabolic regulation. Frequently, PARP activation represent pathological disruptive metabolic signals. Here, we briefly review PARP-mediated pathways in metabolic regulation. Metabolic pathologies associated with PARP activation are listed in Table 1.
PARPs in regulating central and peripheral organismal metabolic homeostasis PARP enzymes are widely expressed in almost all tissues and cells of the human organism, including metabolic tissues and organs, such as the liver, skeletal muscle, hormone glands, adipose tissue (white, brown, and beige), and the nerve system (Bai 2015). Central metabolic regulation encompasses the coordinated regulatory activity of the central nervous system and the hormonal system, which allows the organism to adjust to environmental and internal metabolic challenges. Such signals are integrated into the nuclei of the ventromedial hypothalamus, which serve as a central orchestrator and zeitgeber for other organs through hypothalamic neurohormonal changes (Cedernaes et al. 2019). Whole body genetic deletion of PARP1 alters feeding entrainment in mice and changes spontaneous locomotor activity (Bai et al. 2011b), suggesting a role for PARP1 in the circadian phase of entrainment. PARP1 expression and PARP1 activity show circadian changes in murine models and humans that contribute to circadian entrainment of transcriptional programs in skeletal muscle, the liver, and in the cells of the immune system (Mocchegiani et al. 2003(Mocchegiani et al. , 2004Asher et al. 2010;Zhao et al. 2015). PARP1 can achieve circadian regulation of gene transcription through the following actions: (1) interacting with 11-zinc-finger protein or CCCTC-binding factor (CTCF) and converting parts of the chromatin to heterochromatin in a time-dependent fashion (Zhao et al. 2015) and (2) interacting with and ADP-ribosylating Clock protein (Asher et al. 2010). Yetuncovered pathways may also be active. PARP1 activation seems to be vital for sensing or mediating NAD + /NADH levels to be integrated into cellular energy sensing and signaling. Although, the aforementioned pathways were described in nonneuronal models, PARPs are abundantly expressed and active in the nervous system (Komjati et al. 2004;Fatokun et al. 2014) and feeding and locomotion behavior changes in the PARP1 knockout mice (Bai et al. 2011b), making it likely that these processes are active in neurons and other cellular elements of the nervous system. It is important to note that disrupting circadian entrainment increases the risk for obesity and the consequences of obesity (Kettner et al. 2015); however, this has not been studied in the context of PARP activation.
The sensing of hormones is also regulated by PARPs. Nuclear hormone receptors use PARPs as cofactors (Table  2). Therefore, nuclear hormone receptor activation is PARP-dependent. Insulin-like growth factor (IGF)-1 signaling is potentiated by PARP inhibition (Amin et al. 2015). Furthermore, PARP1 interferes with GLP-1 signaling that may interfere with insulin secretion from β cells (Liu et al. 2011). PARP1 and PARP2 activation were Continued shown to be a key step in the development of insulin resistance (for review, see Bai and Cantó 2012). Hormones, such as insulin (Horvath et al. 2008), estrogens (Mabley et al. 2005;Jog and Caricchio 2013;Joshi et al. 2014), androgens (Shimizu et al. 2013), progesterone (Ghabreau et al. 2004), artificial steroids, and vitamin D (Marton et al. 2018b) can modulate the expression and activity of PARP1 and PARP2. Endocrine disruptors were also shown to modulate PARP activity (Chen et al. 2013;Guerriero et al. 2018). These observations suggest feedback loops where PARPs interfere with hormonal signaling and hormones regulate PARP availability and activity.
PARPs interplay with energy sensor systems in cells (for review, see . These systems assess the energy charge of cells (NAD + /NADH or ATP/(ADP + AMP) ratio) and the availability of nutrients (amino acids, oxygen, etc.) and shape cellular metabolism to meet these challenges.

PARPs in carbohydrate metabolism
PARPs regulate points in glycolysis (Hopp et al. 2019), the core pathway of glucose catabolism. PARP1 activation hampers glycolytic flux, inducing metabolic dysfunction (Ying et al. 2002(Ying et al. , 2003Devalaraja-Narashimha and Pada-nilam 2009;Módis et al. 2012;Robaszkiewicz et al. 2014). Tankyrase 1 and Tankyrase 2 (TNK1, TNK2) regulate glucose transporter 4 (Glut4) translocation to the cytoplasmic surface in an ADP-ribosylation-dependent manner and, thus play a vital role in regulating glucose (and glutamine) availability and glycolytic flux (Yeh et al. 2007). The next step in glucose catabolism is the phosphorylation of glucose by hexokinase to form glucose-6-phosphate, which represents a commitment to glycolysis. Hexokinase is localized to the mitochondrial surface to help synchronize glycolytic flux and mitochondrial oxidation (Andrabi et al. 2014). PARP1 activation disrupts this synchronized function, reducing glycolytic influx (Andrabi et al. 2014;Fouquerel et al. 2014). This observation is further underlined by the observation that the supplementation of pyruvate, the end product of glycolysis, can alleviate cellular dysfunction and cell death upon PARP1 activation (Ying et al. 2002(Ying et al. , 2003Suh et al. 2005;Zeng et al. 2007). In agreement with these observations, the down-regulation of PARP1 supports glycolysis (Regdon et al. 2019). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is an NAD + -dependent enzyme in glycolysis. PARP1 can PARylate and hence inhibit GAPDH (Du et al. 2003). Furthermore, since GAPDH is NAD + -dependent, NAD + breakdown by cytoplasmic PARPs can limit  GAPDH activity and, consequently, glycolytic flux (Hopp et al. 2019). These results were confirmed by the observation that PARP1 knockout mice have higher respiratory quotient, suggesting a shift toward glucose oxidation (Bai et al. 2011b). Although pyruvate dehydrogenase complex is not considered as a member of the glycolytic enzymatic machinery, it is important to note that three subunits of the complex (PDPR, PDHA1, and PDHX) are subject to poly-ADP-ribosylation, which may regulate the fate of pyruvate, whether it can enter the TCA cycle, convert to lactate, or undergo gluconeogenesis (Hopp et al. 2019). PARP10 and PARP14 are two poorly characterized members of the PARP family. Nevertheless, they seem to be connected to carbohydrate metabolism. Silencing of PARP10 induces glycolysis and mitochondrial oxidation, rendering cells hypermetabolic (Márton et al. 2018a). PARP14 can support glycolysis in lymphoma cells, although the molecular mechanism has not been elucidated (Cho et al. 2011). Another interesting feature of PARP14 is its physical interaction with phosphoglucose isomerase, an enzyme that enables the entry of fructose into glycolysis (Yanagawa et al. 2007). The actual consequence of this interaction is unknown.
When considering carbohydrate metabolism, the regulatory mechanisms should also be mentioned. PARPs inter-act with HIFs, GSK3b, and AMPK, sensors that regulate glycolytic flux and the coupling of glycolysis to mitochondrial oxidation. These pathways are reviewed in . A high-glucose or high-fructose diet can induce the expression of PARP1 (Choi et al. 2017;Huang et al. 2019). The interplay between carbohydrate metabolism and PARPs was extensively reviewed in Hopp et al. (2019).

PARPs in lipid metabolism
There is an ever-growing body of evidence for the involvement of PARPs in lipid metabolism. As a prime example, PARP2 was found to be connected to cholesterol and triglyceride metabolism in a genome-wide association study (Manunza et al. 2014).
Cellular and organismal fatty acid homeostasis are regulated by PARPs. Erener et al. (2012b) reported hypercholesterolaemia in PARP1 knockout mice. The pattern of polyunsaturated fatty acid metabolites is dysregulated in PARP1 knockout mice (Kiss et al. 2015) and there seems to be a correlation between PARP1 activity and erythrocyte membrane composition (Bianchi et al. 2016). Furthermore, the composition of membrane-constituent lipids was altered upon the deletion of PARP2 (Marton et al. 2018b). Fatty acid absorption and fatty acid biosynthesis had not been studied in the context of PARP enzymes and poly-ADP-ribosylation; however, the involvement of PARPs is likely, as suggested by scattered data in the literature. For example, the deletion of PARP2 reduces the expression of fatty acid synthase in the white adipose tissue (Bai et al. 2007). The expression of the fatty acid transporters, FABP7, FABP3, CD36, and aP2 (FABP4), are regulated by PARP1, PARP2, and tankyrases (Bai et al. 2007;Yeh et al. 2009;Erener et al. 2012a;Kiss et al. 2015). The deletion of PARP1, PARP2, or PARP10 induces mitochondrial fatty acid oxidation (Bai et al. 2011a,b;Márton et al. 2018a). Upon the genetic deletion of PARP2, the respiratory quotient decreases, suggesting a preference for fatty acid oxidation both in the active and in the sleeping period of the daily cycle (Bai et al. 2011a). Acylation of histone proteins by fatty acids may serve as epigenetic marks, a recent study suggested the PARP-sirtuin interplay may be a key factor in regulating acyl epigenetic marks (Faraone-Mennella et al. 2019).
Certain fatty acid-type lipid species can regulate the expression of PARPs. Serum deprivation of a plethora of lipid species (Sun et al. 2019) can inhibit PARP2 expression, similar to lipoic acid . Caloric restriction reduces, while a high-fat diet induces the expression of PARP1 (Bai et al. 2011b;Salomone et al. 2017;Huang et al. 2019). In a similar fashion, fatty acid synthase activation or overexpression can also induce PARP1 expression (Wu et al. 2016).
Another arch of lipid metabolism is cholesterol homeostasis and the metabolism of cholesterol derivatives. The central organ for cholesterol biosynthesis is the liver, although other organs, such as skeletal muscle, also possess functional enzymatic machinery for cholesterol biosynthesis. Dietary cholesterol is taken up from the intestines and is then transported to the liver by chylomicrons. Excess cholesterol is excreted in the bile that is subsequently emptied into the intestines. Collectively, this is called the enterohepatic circulation of cholesterol. The liver can excrete cholesterol into low-density lipoprotein (LDL) that are then sent to the periphery to supply cholesterol to cells. Peripheral cholesterol is returned to the liver by high-density lipoproteins (HDL). This is the peripheral circulation of cholesterol in humans. Mice have little HDL, therefore, LDL performs the functions of HDL in mice. Cholesterol is a starting compound for the synthesis of steroid hormones, vitamin D, and bile acids.
PARP2 negatively regulates de novo cholesterol biosynthesis through suppression of sterol-regulatory element-binding protein expression. The deletion of PARP2 induces increased cholesterol biosynthesis in the liver and skeletal muscle (Szántó et al. 2014;Marton et al. 2018b). A fraction of excess cholesterol seems to be incorporated into biomembranes (Marton et al. 2018b). The deletion of PARP2 does not affect the enterohepatic circulation of cholesterol. However, PARP2 deletion reduces the expression of hepatic ATP-binding cassette subfamily A member 1 (ABCA1), a major transporter of cholesterol to lipoproteins (Szántó et al. 2014). In line with this, serum HDL levels are lower in PARP2 knockout mice (Szántó et al. 2014). However, it is not easy to translate this finding into the human situation.
Lipids can be stored physiologically or pathophysiologically in multiple organs, where excess lipids cause damage to the tissue. Lipid-mediated activation of PARP1 may have a crucial role in organ or cellular damage (Diestel et al. 2003;Kiss et al. 2006;Hans et al. 2008;Bai and Csóka 2015;Chen et al. 2015). Ectopic lipid deposition to the walls of arteries happens in atherosclerosis. PARP inhibition or genetic deletion of PARP1 alleviates the symptoms of atherosclerosis by reducing plaque area, lipid deposition, inflammation, and the HDL/LDL ratio (Martinet et al. 2002;Kiss et al. 2006;Ambrose et al. 2009;Liu et al. 2011;Sunderland et al. 2011;Shen et al. 2012;Wei et al. 2013;Xu et al. 2014).
The liver, although it has limited lipid storage, is also a site for abnormal lipid deposition in alcoholic and nonalcoholic fatty liver disease (AFLD and NAFLD, respectively). Alcohol consumption induces PARylation (Nomura et al. 2001). Logically, pharmacological PARP inhibition confers protection against steatosis, inflammation, and liver tissue injury in AFLD (Mukhopadhyay et al. 2017). While the genetic deletion of PARP2 is protective against nonalcoholic hepatic lipid accumulation (Bai et al. 2011a), there is apparent ambiguity in the literature on the role of PARP1 concerning whether the genetic ablation of PARP1 exacerbates NAFLD (Erener et al. 2012b) or pharmacological PARP inhibition protects against steatosis, inflammation, and liver tissue injury in NAFLD (Bai et al. 2011b;Gariani et al. 2017;Mukhopadhyay et al. 2017;Huang et al. 2018). The differences have not been elucidated yet.
General outline of adipogenesis "Professional" lipid storage cells in mammals are adipocytes classified as white, brown, and beige adipocytes.
Brown or multilocular (referring to the numerous intracellular lipid droplets) adipocytes are localized to specific regions, including the interscapular and perirenal regions and lining the large arteries (Cannon and Nedergaard 2004). Brown adipocytes are characterized by high mitochondrial content and high uncoupling protein-1 (UCP1) expression (Kajimura 2015). This tissue is vital in human newborns and in rodents for maintaining core body temperature through uncoupled respiration and through that, in maintaining organismal energy balance, regulating fatty acid and glucose oxidation, and preventing or alleviating obesity and its consequences (Cannon and Nedergaard 2004).
Beige adipocytes are localized within white adipose tissue depots mixed with white adipocytes . Beige cells share the morphological characteristics of white adipocytes; nevertheless, beige cells respond to adrenergic stimuli by mitochondrial biogenesis, induction of UCP1 expression, fatty acid breakdown, and heat generation. Beige adipocytes are characterized by a futile creatine cycle Bertholet et al. 2017;Kazak et al. 2017) that is not present in brown cells and is vital for heat generation. Importantly, a mutation in the fto gene was associated with impaired beige adipogenesis and, consequently, impaired mitochondrial biogenesis and organismal energy balance (Claussnitzer et al. 2015).
White adipocytes are cells specialized for fat storage. Morphologically, these cells are unilocular and when stimulated respond with triglyceride breakdown through hormone-sensitive lipase (HSL). There are multiple adipose tissue depots in the body and their metabolic behavior is quite different in terms of lipid mobilizing capacity or heat generation (Garaulet et al. 2006;Roca-Rivada et al. 2011;Sacks et al. 2013;Luche et al. 2015). The switching on of beige adipocytes in white adipose depots or the transdifferentiation of white adipocytes to brown or beige cells is termed "browning" (Kajimura 2015).
According to the classical scheme of adipocyte differentiation, Pax7 + Myf5 + brown cell precursors segregate from the dermatomyotome, while Pax7 − Myf5 − stem cells differentiate to white and beige adipocytes (Rosen and Spiegelman 2014). This picture is, in fact, more complex ( Fig. 1). Lineage tracing studies revealed that there are multiple lineages giving rise to white adipocytes. The majority of these are of mesenchymal origin; nevertheless, depots in the head region stem from the neural crest (Sox10 + , Wnt1 + precursors) (Billon et al. 2007; Sanchez-Gurmaches and Guertin 2014a). Mesenchymal precursors can be Myf5 + or Myf5 − . The proportion of white adipocytes derived from Myf5 + or Myf5 − precursors vary between the adipose tissue depots (Sanchez-Gurmaches and Guertin 2014a). Beige adipocytes can differentiate from the same precursors as the white adipocytes, except for neural crest-derived precursors (Sanchez-Gurmaches and Guertin 2014a). Finally, brown adipocytes differentiate from Pax7 + Myf5 + dermatomyotomal precursors (Sanchez-Gurmaches and Guertin 2014a).
The in vitro models of (human) adipose tissue-derived stem cells (hADMSCs), (embryonic) fibroblasts, or immortalized cell lines (e.g., 3T3-L1, 3T3-F442A, etc.) (Ruiz-Ojeda et al. 2016) are useful tools in understanding transcriptional control over adipogenesis. The differentiation protocol usually involves a complete stop of proliferation by growing cells at confluency, followed by the induction of differentiation by a cocktail of hormones including insulin, a synthetic glucocorticoid, dexamethasone, and 3-isobutyl-1-methylxanthine (IBMX), a phosphodiesterase inhibitor. After the induction of differentiation, cells undergo commitment and committed cells undergo a few rounds of cellular division, called mitotic clonal expansion (Fig. 2). It is not known whether clonal expansion also characterizes the in vivo differentiation of adipocytes. After clonal expansion, cells begin accumulating lipids in lipid droplets (in vitro differentiated adipocytes are multilocular), termed terminal differentiation (  Table 3. Classically, clonal expansion of white adipocytes was shown to be mediated by the self-amplifying activation of C/EBPδ and C/EBPβ that subsequently induces the expression of C/EBPα and, finally, the expression of peroxisome proliferator activated receptor (PPAR) γ1 and PPARγ2 expression (Fajas et al. 1998).
The induction of the expression of PPARγ isoforms is a common denominator of beige and brown adipogenesis, similar to white adipogenesis. Mitochondrial biogenesis is a key factor for the differentiation of beige and brown adipocytes. The concerted action of the energy stress sensor web is vital for the induction of mitochondrial biogenesis, including the activation of AMPK or SIRT1 (Qiang et al. 2012;Shan et al. 2013;Wang et al. 2015;Abdul-Rahman et al. 2016;Nagy et al. 2019).   Nozaki et al. 2003;Idogawa et al. 2005;Mariotti et al. 2017 Most of these interactions were shown in model systems other than adipogenesis.

The role of PARP enzymes in adipogenesis
The first observation that PARPs modulate adipogenesis came in 1995 by Smulson et al. (1995) using the 3T3-L1 model system and 3AB, a rather unspecific PARP inhibitor. This study showed that pharmacological PARP inhibition hampers 3T3-L1 differentiation (Smulson et al. 1995). Indeed, PARPs play a role in the regulation of adipogenesis and adipose tissue function. Since this first observation, much data has emerged along with numerous controversial issues.
Early commitment and clonal expansion PARP1, PARP2, and PARP7 have pivotal roles in decision making between retaining stem cell properties and differentiation in nonadipogenic models (Yélamos et al. 2006;Farrés et al. 2013Farrés et al. , 2015Nozaki et al. 2013;Roper et al. 2014;Vida et al. 2016). Therefore, PARPs may be crucial in the early commitment of cells toward preadipocytes and adipose lineages (Fig. 1). To date, no studies have been published concerning the role of PARPs in commitment to adipocyte lineages in an in vivo setting (e.g., as in Sanchez-Gurmaches and Guertin 2014b). However, PARP1 has a crucial role in preadipocyte commitment to white adipocyte differentiation in in vitro systems (Luo et al. 2017;Ryu et al. 2018).
In the in vitro differentiation of 3T3-L1 preadipocytes, a characteristic PARylation pattern was detected (Luo et al. 2017). In confluency (growth arrest), PARP1 auto-PARylation dominates cells, after which the PARylation signal is low in the clonal expansion phase and boosts again in terminal differentiation (Luo et al. 2017). In terminal differentiation, PARP1 auto-PARylation returns, nevertheless, lower molecular weight PARylation signals are also detected (Luo et al. 2017).
As noted in the previous chapter, the clonal expansion phase is dominated by the self-intensifying loop between C/EBPβ and C/EBPδ. This loop is vital for the subsequent transcription of C/EBPα and PPARγ transcription factors that then transcribe the "executors" of lipogenesis. PARP1 can PARylate C/EBPβ on K133, E135, and E139 residues, resulting in decreased binding of C/EBPβ to the promoters of C/EBPα or PPARγ2. Hence, genetic or pharmacological inactivation of PARP1 supports adipocyte differentiation (Luo et al. 2017). The deletion of these PARylation sites enhance C/EBPβ binding to target promoters and renders C/EBPβ resistant to PARP inhibitors. These findings provide a physiological explanation for reduced PARylation during the clonal expansion phase.
Another mechanism for the regulation of PARP1 activity and clonal expansion is the compartment-specific NAD + biosynthesis through NMNAT enzymes. Ryu et al. (2018) showed that blocking nuclear NMNAT-1 induces adipocyte differentiation through limiting nuclear NAD + for PARP1. In other words, PARP1 activation and fueling PARP1 activation by NMNAT-1 can keep preadipocytes undifferentiated. The cytosolic NMNAT-2 is induced early in adipocyte differentiation (4 h after induction) and shifts nuclear NAD + biosynthesis to the cytosol to support glycolysis (Ryu et al. 2018). As a "side effect," nuclear PARylation is reduced, supporting white adipocyte differentiation (Ryu et al. 2018).

Adipocyte terminal differentiation
Adipocyte terminal differentiation in in vitro models is characterized by increasing C/EBPα and PPARγ protein expression and lipid accumulation. This phase of terminal differentiation is associated with the accumulation of PARP1 and PAR formation (Erener et al. 2012a;Luo et al. 2017). In the studies of Erener et al. (2012a,b), pharmacological and genetic PARP inhibition blocked the differentiation of 3T3-L1 cells. When PARP1 was blocked in the course of 3T3-L1 differentiation, a major reduction in the expression of C/EBPα and PPARγ2 and a set of PPARγdependent transcripts was observed, in stark contrast to the previously discussed studies (Luo et al. 2017;Ryu et al. 2018).
Lower adipocyte differentiation was linked to a slower resolution of transcription-coupled topoisomerase II-inflicted double strand breaks and the consequent slower initiation of RNA polymerase II-mediated transcription in the absence of PARP activity (Pavri et al. 2005;Erener et al. 2012a;Lehmann et al. 2015). Furthermore, PARP inhibition supported the binding of NCoR-1 (an inhibitory cofactor of PPARγ), while decreasing the binding of p300 (an activating cofactor of PPARγ) (Lehmann et al. 2015). In a cardiomyocyte model, pharmacological, and genetic PARP1 inhibition led to increased PPARγ activity (Huang et al. 2009), in contrast to the observations detailed above.
There is apparent contradiction between the results showing that PARP1 and NAD + biosynthesis during the commitment phase blocks (Luo et al. 2017;Ryu et al. 2018), while during terminal differentiation PARP1 supports adipocyte differentiation (Erener et al. 2012a,b;Lehmann et al. 2015). To date, no explanation is given to the discrepancies that is backed by experimental proof. Nevertheless, the visibly contradictory results may be both true. The contradictory reports do observe PARP auto-PARylation in confluent and in terminally differentiated cells (Erener et al. 2012a;Luo et al. 2017) suggesting that similar processes may take place in all cases; however, the dependence of the cells on early commitment may be different. In our hands different clones of the 3T3-L1 cells have different behavior in differentiation and response to PARP inhibitors (unpublished data).
The genetic silencing of PARP2 led to lipodystophy in chow diet-fed mice, which was mirrored when primary fibroblasts were differentiated to mature adipocytes (Bai et al. 2007). Decreased adipocytic differentiation was a result of blunted PPARγ activation. PARP2 binds to PPARγ-mediated promoters (e.g., aP2) and supports mRNA transcription. Reduced expression of the PPARγdependent genes in the PARP2 knockout mice points toward hampered PPARγ activation in the absence of PARP2 (Bai et al. 2007).
In the above-mentioned studies (Bai et al. 2007;Huang et al. 2009;Erener et al. 2012a,eb;Lehmann et al. 2015;Luo et al. 2017;Ryu et al. 2018), PARP inhibition or the genetic deletion of PARP1 or PARP2 modulated genes involved in fatty acid uptake (lipoprotein lipase [LPL], fatty acid binding protein 4 [FABP4,aP2], and CD36), lipid storage (perilipin), fatty acid biosynthesis (fatty acid synthase [FAS]), and adipokines (leptin, adiponectin, and resistin) in white adipocyte differentiation models. The deletion of tankyrase-1 induced leptin and adiponectin expression and secretion from white adipose tissue (Yeh et al. 2009). These genes are PPARγ-dependent and encompass all processes needed for triglyceride uptake and storage. To date, no studies have reported fatty acid release disorders in relation to the modulation of PARP1 or PARP2 activity (Bai et al. 2007;Erener et al. 2012b).

Switch between white, brown, or beige adipogenesis
PARPs may have a role in selecting between the differentiation to white, brown, and beige adipocytes. PARP1 and PARP2 were shown to modulate skeletal muscle myoblast differentiation and health (Butler and Ordahl 1999;Vyas et al. 2001;Hu et al. 2013;Chacon-Cabrera et al. 2015). Therefore, it is also likely that PARPs can influence white/brown/beige diversion. This hypothesis is further supported by the widespread interactions between energy stress sensors, mitochondrial biogenesis regulators, and PARPs .
The deletion of PARP1 or PARP2, as well as the pharmacological inhibition of PARP, supports mitochondrial biogenesis (Virag et al. 1998a;Bai et al. 2011aBai et al. ,b,2015Szanto et al. 2011;Mohamed et al. 2014) via the preservation of cellular NAD + pools and the subsequent activation of the SIRT1-PGC1α axis (Cantó et al. 2013;. In agreement with this, Nagy et al. (2019) found that in vitro treatment of hADMSC cells, differentiated to white adipocytes, with olaparib induced browning of the cells, marked by mitochondrial biogenesis and UCP1 induction. In the olaparib-treated cells, beige cell markers were not induced, suggesting browning induced transdifferentiation to brown adipocytes. In good agreement with that observation, in PARP1 knockout mice, we detected more active brown adipose tissue (lower lipid deposition, induction of UCPs, increased fatty acid oxidation, and higher mitochondrial content), increased energy expenditure, and improved capacity to withstand cold exposure (Bai et al. 2011b). We detected increased cellular NAD + content and SIRT1 activity in both models (Bai et al. 2011b;Nagy et al. 2019). Interestingly, the brown adipose tissue of the PARP2 knockout mice was not more active (Bai et al. 2011a). To date, no thorough studies were performed to assess the contribution of PARPs to beige and brown adipocyte differentiation. These findings are in agreement with the observations that better NAD + availability (Yamaguchi et al. 2019) or SIRT1 activation supports brown and beige differentiation (Qiang et al. 2012;Khanh et al. 2018).

Lipid accumulation, obesity, insulin sensitivity
A role for PARP enzymes in obesity has been reported. In a study of monozygotic twins, higher PARP activity was found in the subcutaneous white adipose tissue of the heavier cotwin . Furthermore, in weight loss adipocytic PARP activity is reduced, while SIRT1 activity is up-regulated ). In murine studies, PARP1, PARP2, and tankyrase-1 were shown to be involved in modulating energy balance and obesity. Similar to the ambiguity in the role of PARP1 in adipocyte differentiation, the studies on the organismal role of PARP1 in obesity and its consequences are also contradictory. In our studies, PARP1 knockout mice were leaner when kept on chow diet that was accentuated on high-fat feeding (Bai et al. 2011b). This study was backed by a study from another laboratory. PARP1 knockout mice had lower body weight and white adipose tissue mass when on a highfat diet (Erener et al. 2012b). Furthermore, treatment of mice with an orally administered PARP inhibitor, MRLB-45696, (PARP1 is responsible for 80%-85% of total cellular PARP activity) (Schreiber et al. 2002;Szanto et al. 2011) prevented weight gain on a high-fat diet (Lehmann et al. 2015). In contrast to these studies, a report by Devalaraja-Narashimha and Padanilam (2010) reported a complete opposite phenotype; the PARP1 knockout mice became seriously obese as compared with their wild-type counter partners upon high-fat feeding. In all studies, a hypercaloric high-fat diet was used.
Obesity is a complex pathology and cannot be solely attributed to the dysfunction of white adipocytes; a complex deregulation of organismal energy homeostasis is involved (Rosen and Spiegelman 2014). In the above-mentioned studies that reported a lean phenotype, an energy expenditure phenotype was described due to mitochondrial biogenesis in the brown adipose tissue and the skeletal muscle, attributed mainly to the activation of the NAD + -SIRT1 axis (Bai et al. 2011b;Pirinen et al. 2014;Lehmann et al. 2015). The improved metabolic fitness yielded improved glucose tolerance and insulin sensitivity, with skeletal muscle being responsible for glucose clearance both in chow-fed and high-fat-fed mice (Bai et al. 2011b). In the monozygotic twin study, the activation of the NAD + -SIRT1 axis and the consequently lower PARP activity was associated with a leaner, metabolically healthier phenotype . The contradictory study (Devalaraja-Narashimha and Padanilam 2010) reported an opposing rearrangement of energy homeostasis characterized by lower oxygen consumption, energy deliberation, worsened glucose clearance, and insulin resistance.
These are again opposing results without good experimental explanation. A root cause for the disagreement between the studies could be that these studies were conducted on two different knockout PARP1 mouse strains. One of the strains was generated by Wang et al. (1995) and deposited at Jackson Laboratories; the other strain was generated in the laboratory of de Murcia et al. (1997). The mice generated by Wang et al. (1995) were on an SV129 background, while the mice generated by de Murcia et al. (1997) were on a C57/Bl6J background. The metabolic behavior of the two different backgrounds is profoundly different (Andrikopoulos et al. 2005;Berglund et al. 2008) and might be the explanation for the differing results. A solution for these issues could be the use of a transgenic PARP1loxP mouse strain that will bypass developmental issues and enable the study of interorgan interactions (JAX 2019).
Induction of mitochondrial biogenesis by enhancing the NAD + -SIRT1 axis in the skeletal muscle after the genetic deletion of PARP2 brought about a lean phenotype (Bai et al. 2011a;Mohamed et al. 2014). Interestingly, the brown adipose tissue of the PARP2 knockout mice was not involved in the energy expenditure phenotype, in contrast to the PARP1 knockout mice (Bai et al. 2011a,b). In chow-fed mice, the deletion of PARP2 improved insulin sensitivity and glucose clearance. While on a high-fat diet, the ablation of PARP2 improved insulin sensitivity, but insulin secretion and glucose clearance were blunted due to inhibition of compensatory β-cell proliferation (Bai et al. 2011a).
Tankyrase expression is among the highest in the white adipose tissue and the brain (Yeh et al. 2009). White adipose tissue and energy homeostasis changes were observed in tankyrase knockout mice (Yeh et al. 2007(Yeh et al. , 2009. Interestingly, tankyrase expression may also affect brown adipose tissue (Yeh et al. 2009). However, the involvement of tankyrase in brown adipose tissue function was not investigated yet. Tankyrase knockdown was shown to impair Glut4 translocation and hence insulinstimulated glucose uptake, resulting in down-regulation of glucose metabolism in differentiated 3T3-L1 adipocytes (Yeh et al. 2007). These effects were dependent on tankyrase activity (Yeh et al. 2007). In tankyrase knockout mice, the relative mass of the epididymal white adipose tissue decreased in parallel to enhanced energy expenditure marked by increased oxygen consumption (Yeh et al. 2009).

Future directions
PARP enzymes and PARP inhibition interfere with adipose tissue biology at multiple points. There are obesityassociated processes (e.g., inflammation) that are also PARP regulated, but their interplay had not been assessed. We give a brief overview of these processes below.
Inflammation plays diverse roles in obesity and adipose tissue homeostasis. Obesity is associated with inflammation and fibrosis of the adipose tissue (Reilly and Saltiel 2017). Preventing adipose tissue inflammation is a key step toward the "metabolically healthy" obese phenotype (Vishvanath and Gupta 2019). Furthermore, inflammatory signaling seems to be a player in diverting toward the beige lineage (Sun et al. 2018). PARP enzymes are involved in the regulation of inflammation; usually, the absence of PARP1 or PARP2 or pharmacological PARP inhibition is anti-inflammatory (Fehr et al. 2020), except for Th17-mediated processes (Kiss et al. 2019). Furthermore, increases in SIRT1 activity, which can be elicited by PARP inhibition, can suppress adipose tissue inflammation, and hence support its function (Gillum et al. 2011;Chalkiadaki and Guarente 2012). Importantly, there is evidence that the results of murine PARP inhibitor studies are likely translatable to humans (Morrow et al. 2009). PARP1 and PARP inhibition regulate IL6 (Lehmann et al. 2015), IL12m, IL13ra, SAA3, pu1, and MPEG1 (Erener et al. 2012b) expression. In the adipose tissue of PARP2 knockout mice, signs of inflammation were detected, including F4/80 positive cells and dilated capillaries, that were absent in their wild-type counter partners (Bai et al. 2007). Whether inflammatory processes are the cause or consequence of the distortion of adipose tissue function is unknown.
Recent studies showed that the loss of microbiome diversity hampers adipose tissue browning (Suárez-Zamorano et al. 2015;Li et al. 2019). Intriguingly, the genetic deletion of PARP1 enhances the diversity of the gut microbiome (Larmonier et al. 2016;Vida et al. 2018), suggesting a possible link between PARP1 and adipose tissue browning. Disruption of circadian entrainment of feeding can also contribute to obesity (Hatori et al. 2012;Zarrinpar et al. 2016;Chaix et al. 2019) and, as noted earlier, the disruption of PARP1 leads to changes in the diurnal cycle of feeding and metabolism (Asher et al. 2010;Bai et al. 2011b). PARP activation can be a go/no-go signal in cell death (Virág et al. 1998b;Fatokun et al. 2014;Dawson and Dawson 2017) and PARPs regulate cellular proliferation (Bai 2015), two vital steps to adipocyte differentiation and selection between beige, brown, or white lineages. Similarly, PARP1 and PARP10 were implicated in the regulation of autophagy and mitophagy (Muñoz-Gámez et al. 2009;Kleine et al. 2012), processes that shape adipocyte differentiation (Kim and Lee 2014). PARPs affect nuclear structure and the epigenetic code (Wacker et al. 2007;Krishnakumar et al. 2008;Hottiger 2015;Zhao et al. 2015). PARP1 deficiency was shown to modulate H3K9me3 and H3K4me3 methylation during adipogenic differentiation (Erener et al. 2012a). Nevertheless, large scale studies are missing. There are genes reported to be PARP-mediated (e.g., MDH1) (Hopp et al. 2019) that regulate adipocyte differentiation. Again, the involvement of these genes in adipogenesis in the context of PARylation had not been assessed.
All adipose tissue depots are characterized by secretion of bioactive compounds such as peptide hormones (adipokines), bioactive lipids (lipokines), and RNA molecules with local (paracrine) and systemic (endocrine) effects on multiple metabolic tissues and the cardiovascular system. These bioactive compounds are synthesized and secreted as a function of the energy status of adipose tissues, which in turn regulates appetite, thermogenesis, glucose, and lipid metabolism (Scheja and Heeren 2019). The role of PARPs had not been studied in this direction. Along the same lines, large-scale endocrine studies are also missing.
The role of PARPs in adipogenesis and metabolism will clearly have practical applications not only in the strict sense of metabolism and metabolic diseases, but also from the perspective of cancer and cancer cachexia (Chacon-Cabrera et al. 2015Barreiro and Gea 2018;Doles et al. 2018). These outstanding issues warrant further studies in the future.