Alternative Polyadenylation and Differential Regulation of Ucp1: Implications for Brown Adipose Tissue Thermogenesis Across Species

Brown adipose tissue (BAT) is a thermogenic organ owing to its unique expression of uncoupling protein 1 (UCP1), which is a proton channel in the inner mitochondrial membrane used to dissipate the proton gradient and uncouple the electron transport chain to generate heat instead of adenosine triphosphate. The discovery of metabolically active BAT in human adults, especially in lean people after cold exposure, has provoked the “thermogenic anti-obesity” idea to battle weight gain. Because BAT can expend energy through UCP1-mediated thermogenesis, the molecular mechanisms regulating UCP1 expression have been extensively investigated at both transcriptional and posttranscriptional levels. Of note, the 3′-untranslated region (3′-UTR) of Ucp1 mRNA is differentially processed between mice and humans that quantitatively affects UCP1 synthesis and thermogenesis. Here, we summarize the regulatory mechanisms underlying UCP1 expression, report the number of poly(A) signals identified or predicted in Ucp1 genes across species, and discuss the potential and caution in targeting UCP1 for enhancing thermogenesis and metabolic fitness.


INTRODUCTION
Brown and white adipocytes possess unique functions in thermogenesis and energy storage, respectively; however, to some extent, one could cover the function of the other. A high fat diet (HFD) can induce brown-to-white adipocyte conversion when brown adipocytes store excess lipids to appear unilocular. Some white adipocytes are known as beige adipocytes, which can be browned in response to adrenergic signaling to become thermogenic (1). Cold-induced metabolic activity of brown adipose tissue (BAT) is positively correlated with resting metabolic rate and negatively with body mass index and body fat percentage in humans (2,3). Therefore, a "thermogenic antiobesity" approach has been proposed to combat obesity if the metabolic furnace, BAT, can be pharmacologically activated to burn excess fat (4). Because BAT defends against fat accumulation via uncoupling protein 1 (UCP1)-mediated thermogenesis, extensive efforts have uncovered how UCP1 activity is controlled by allosteric conformation changes and gene expression to affect thermogenesis. Several polymorphisms in the promoter, non-coding, and coding regions of UCP1 gene were found associated with obesity and type 2 diabetes (5)(6)(7)(8). Of note, genomic variations in Ucp1 across species indicate that non-shivering thermogenic function and regulation are not necessarily conserved through evolution.
Discovery of Ucp1, Whose 3 ′ -UTR Is Processed Differently Among Species UCP1 was first isolated from BAT of rats and hamsters in 1980 and named for its heat-producing function by uncoupling the oxidative phosphorylation process in mitochondria ( Figure 1A) (9). At the basal status, the binding of purine nucleotide diand tri-phosphates on the outer facing cavity of UCP1 blocks its proton-translocating activity (10,11). In response to nutrients or cold environment, thermogenesis is activated via the release of norepinephrine from sympathetic circuits onto β3 adrenergic receptors (β3ARs) in brown adipocytes (4,12,13). β3AR signaling elevates cyclic adenosine monophosphate (cAMP) level to activate protein kinase A and lipolysis and acutely enhances UCP1 activity via free fatty acid-induced allosteric changes (14). Chronic activation of β3ARs promotes UCP1 synthesis by multiple transcriptional and posttranscriptional mechanisms to be discussed later.
Rat and mouse Ucp1 cDNAs were cloned in 1985 (15,16) and later sequenced to reveal their coding sequences and untranslated regions (UTRs) (17)(18)(19). Soon after, the rabbit Ucp1 cDNA sequence was reported (20). The 3 ′ -UTR of Ucp1 in all three species contains two poly(A) signals (i.e., AAUAAA). Northern blot analysis revealed two Ucp1 transcripts of ∼1.5 and 1.9 kb (hereafter called Ucp1S and Ucp1L) in mouse and rat BAT (15,16). The transcripts result from an alternative use of the poly(A) signal to yield 3 ′ -UTRs of 78 and 485 bp in mice, respectively (21) (Figure 1B). By contrast, only the distal poly(A) signal is used to generate Ucp1L in rabbit BAT (20). The human UCP1 coding sequence was identified by screening a genomic library with a rat Ucp1 cDNA probe, and a 1.9-kb UCP1 mRNA was detected in human perirenal BAT (22). More recently, the sequence of human UCP1 cDNA (ENST00000262999) revealed only one poly(A) site in the 3 ′ -UTR, and RT-PCR analysis confirmed that human UCP1 encodes only the long form carrying a 465-bp 3 ′ -UTR (21).

Thermogenic UCP1 Is Not Preserved in All Endothermic Placental Mammals
The functional Ucp1 gene consists of six exons and spans a region of ∼8-10 kb among species (an example of mouse Ucp1 is in Figure 1B). The phylogenetic relationship of Ucp1 across species has revealed that the non-thermogenic UCP1 in fish and frogs (23) was derived from a hypothetical proto-UCP hundreds of millions of years ago. Non-thermogenic UCP1 was first lost in Sauropsida (reptiles and birds) around 300 million years ago, then thermogenic UCP1 emerged in endothermic placental mammals about 100 million years ago (24,25). Although UCP1-mediated non-shivering thermogenesis is believed to provide a survival advantage in the cold and retain the body temperature of eutherian neonates after birth, eight mammalian clades, namely Xenarthra, Pholidota, Cetacea, Sirenia, Proboscidea, Hyracoidea, Equidae, and the most wellknown Suidae (pigs), showed inactivation of Ucp1 (26). The loss of exons 3 to 5 in pig Ucp1 occurred ∼20 million years ago, so pigs produce no functional UCP1 and depend on shivering for thermoregulation (27).
The previous studies aligned the genomic or amino acid sequences of Ucp1 across species to build up the phylogenetic tree and derive an evolutionary timeline for Ucp1 (24)(25)(26). We followed these evolutionary patterns by aligning the nucleotide sequences of the Ucp1 coding regions in selected mammals and analyzed whether alternative polyadenylation could follow any evolutionary path ( Figure 1C). The phylogenetic tree of Ucp1 grouped most species following the canonical taxonomy, classifying animals of the same order in the same branch. The three exceptions are Ucp1 in the small Madagascar hedgehog in Afrosoricida more related to rodents, in the gray mouse lemur in Primates more related to rodents and rabbits, and in the dog quite different from other animals in Carnivora (24).
The 3 ′ -UTR information for Ucp1 cDNAs in various species is often lacking or incomplete. We predicted "alternative polyadenylation" by searching the poly(A) signal (i.e., AATAAA) within 1 kb downstream of the stop codon from Ucp1 genomic sequences because the 3 ′ -UTR, including in mouse and human Ucp1 mRNAs, is limited in the last coding exon of most transcripts. Our analyses revealed many representative animals carrying two poly(A) sites, such as mice, rats, and squirrels in Rodentia; rabbits in Lagomorpha; and cats, dogs, cows, leopards, and bears in Laurasiatheria. However, microbats and greater horseshoe bats in Chiroptera are the exceptions, with only one poly(A) motif in the 3 ′ -flanking region. Notably, almost all primates, such as humans, chimpanzees, orangutans, and macaques, have only one poly(A) site in Ucp1, with the exception of mouse lemurs, which carry two poly(A) signals at the position similar to those of rodent Ucp1. The nucleotide sequence of the Ucp1 coding region in mouse lemurs was unexpectedly more similar to those in rodents than primates ( Figure 1C). Although the platypus, a unique placental mammal, has two adjacent poly(A) signals in Ucp1, alternative polyadenylation, even if it occurs, would generate two 3 ′ -UTRs of only 32 bp difference. The poly(A) site of Ucp1 in koala and gray short-tailed opossum in Marsupialia is located at 1,389 and 2,074 bp downstream of the stop codon, respectively, implying Ucp1 in both species may be subjected to more complex posttranscriptional regulation. Other than the 3 ′ -UTRs of rat Ucp1S, mouse Ucp1S and Ucp1L, and human UCP1 having been confirmed by cDNA sequencing (17,18,21), the actual usage and preference of the two putative polyadenylation sites in other species require further investigation.

Transcriptional Regulation of Ucp1 Gene
Cold exposure markedly increases both Ucp1L and Ucp1S mRNA levels in mouse and rat BAT (15,16), so elaborative efforts have identified the molecular mechanisms controlling Ucp1 transcription (28)(29)(30). DNA footprinting and promoter assays were performed to identify a 212-bp distal enhancer (−2,494 to −2,283 bp from the transcription start site) and a 555-bp proximal promoter (−611 to −57 bp) in the 5 ′ non-coding region of rat Ucp1 gene (28,31). Many studies used in vitro assays to identify the trans-acting factors for Ucp1 transcription, including protein kinase A-activated cAMP-response element binding protein (CREB) to compete with the negative regulator, Jun, for binding to the −139-to −122-bp region, and CCAAT/enhancerbinding protein α and β (C/EBPα and β) to activate through regions −457 to −440 and −335 to −318 bp [for details, see the review (32)]. Zfp516 binds to the promoter region (−70 to −45) with PR domain-containing 16 (PRDM16), which is essential for BAT differentiation and promotes browning of beige adipocytes to increase body temperature and energy expenditure (33) (Figure 2A). Notably, the absence of the CCAAT sequence in the human UCP1 promoter suggests evolutionary variations in regulating Ucp1 transcription (22).
In addition to positive regulators, receptor-interacting protein 140 (RIP140) mediates methylation at CpG sites in both the promoter and enhancer to silence Ucp1 transcription (39) (Figure 2A). In pregnant mice with streptozotocin-induced diabetes, intrauterine exposure to hyperglycemia impaired fetal BAT development by altering methylation in several metabolic genes including Ucp1. Such epigenetic changes can last and affect BAT function and glucose homeostasis later in life (40). Besides CpG methylation, epigenetic modification of histone 3 (H3) affects chromatin structure to regulate transcription. Histone deacetylase 3 (HDAC3) controls the acetylation of H3 on lysine 27 (H3K27ac) in the enhancers of Ucp1 and Pparα (Figure 2B), so its deficiency facilitates Ucp1 and Pparα transcription and browning in beige adipocytes (41). Of note, pregnant mice fed omega-3 polyunsaturated fatty acids resulted in epigenetic changes, including decreased tri-methylation of H3 on lysine 27 (H3K27me3), increased H3K27ac and dimethylation of H3 on lysine 9 (H3K9me2) to affect BAT development, and increased Ucp1 and Ppargc1a transcription in mouse offspring even at 8 weeks after post-weaning (42). H3K9me2 is usually defined as a repressed histone mark (43), so increased H3K9me2 may not directly associate with the Ucp1 promoter. Nevertheless, this study reported that a maternal dietary effect could have long-lasting benefits on the development and function of the offspring's BAT. Moreover, the silencing mark, H3K9me2, was identified at the Ucp1 enhancer in white adipose tissue (WAT), whereas the active mark, H3K4me3, at the Ucp1 promoter was found in BAT in response to cold stimulation (43). The ubiquitously transcribed tetratricopeptide repeat on chromosome X (UTX) is a histone demethylase recruited to Ucp1 upon the activation of β3ARs to reduce H3K27me3 level (Figure 2B), which consequently leads to increased H3K27ac level via unidentified histone acetyltransferases (44). Therefore, epigenetic changes on Ucp1 are dynamically regulated for temporal and spatial control of UCP1 synthesis, and intrauterine nutrient exposure has a long-term effect on metabolic gene expression via epigenetic regulation.

Posttranscriptional Regulation of Ucp1 mRNA
A pre-mRNA undergoes several processing steps, including splicing, 3 ′ -end cleavage, and polyadenylation, to become a mature mRNA. Alternative polyadenylation generates Ucp1 mRNA carrying long (Ucp1L, ∼10%) or short (Ucp1S, ∼90%) 3 ′ -UTR in mice and rats, but rabbits and humans express only the long form. The regulatory sequences in the 3 ′ -UTR dictate posttranscriptional efficiency for protein production (45), but the effect of different Ucp1 3 ′ -UTRs on UCP1 synthesis and thermogenesis has not been addressed until recently. The study of Ucp1 L mice, with 10% Ucp1L converted to Ucp1S, reported that the protein but not the mRNA level of UCP1 was reduced ∼50-60% in Ucp1 L BAT, so Ucp1L is translated ∼15 times more efficiently than Ucp1S in BAT (21). Moreover, the same study also identified that cytoplasmic polyadenylation element binding protein 2 (CPEB2) signaling through β3ARs binds to and activates polyadenylation-induced translation of only Ucp1L but not Ucp1S (Figure 2C), so both CPEB2-knockout (KO) and Ucp1 L mice show reduced UCP1 level and impaired BAT thermogenesis (21).
The other posttranscriptional mechanisms identified to date inhibit UCP1 synthesis. Reporter assay revealed that insulin-like growth factor 2 mRNA-binding protein 2 (IGF2BP2/IMP2) is a type 2 diabetes-and obesity-associated gene that suppresses Ucp1 mRNA translation likely via both 5 ′ -and 3 ′ -UTRs. Thus, IMP2null mice show increased energy expenditure, insulin sensitivity, and glucose tolerance and defend against cold temperature better than wild-type mice (46). HFD reduces the amount of UCP1 in obese adipose tissues by diminishing the stability of mouse Ucp1 mRNA via the CCR4-NOT deadenylase complex. Moreover, HFD induces the expression of several subunits of the CCR4-NOT complex, including Cnot1-3, Cnot6 and Cnot7, and Cnot7interacting Tob in inguinal WAT (iWAT) of obese mice. The binding of butyrate response factor 1 (BRF1) to the 3 ′ -UTR of Ucp1L mRNA recruits Tob and Cnot7 to cause deadenylationinduced Ucp1 decay ( Figure 2C). Thus, Cnot7-or Tob-KO mice are resistant to HFD-induced obesity with increased Ucp1 mRNA level in iWAT and BAT (47). Although IMP2 and BRF1 suppressed the translation and stability, respectively, of a reporter appended to the Ucp1L 3 ′ -UTR in HEK293 cells, both studies did not examine Ucp1L and Ucp1S separately when comparing polysomal distribution (46) or mRNA stability of Ucp1 (47) between wild-type and KO adipose tissues. MicroRNAs (miRNAs) are a conserved class of small RNAs of about 20-22 nt that regulate the expression of their target mRNAs via the miRNA-induced silencing complex (48,49). Despite the relatively short 3 ′ -UTR (485 bp) of mouse Ucp1L mRNA, a previous study used miRanda to identify miR-9 and miR-338-3p as putative miRNAs targeting the 3 ′ -UTR of Ucp1L mRNA (50). Activation of β3AR signaling decreased both miRNA levels and increased Ucp1 mRNA level in epididymal WAT (eWAT) but not iWAT and BAT. Despite a negative correlation between miR-9 and miR-338-3p and Ucp1 expression in eWAT, no further experiments were performed to determine whether both miRNAs affect Ucp1 mRNA stability by binding to the predicted target sites in the 3 ′ -UTR (50). Although cold exposure did not change the proportion of Ucp1L and Ucp1S in BAT (21), no studies have measured the proportion of Ucp1L and Ucp1S in WAT and whether it can be changed under HFD or cold temperature. The alternative use of proximal and distal poly(A) signals in mouse Ucp1 mRNA may vary in different adipose tissues.
A recent paper demonstrated that methyltransferase like 3 (METTL3)-catalyzed m 6 A modifications in Ucp1 mRNA increased its mRNA and protein levels during postnatal BAT development ( Figure 2C). In mice with BAT-specific ablation of Mettl3, the loss of m 6 A-enhanced UCP1 expression reduced energy expenditure and cold tolerance (51). Moreover, these mice showed accelerated HFD-induced obesity, impaired glucose intolerance, and insulin resistance (51). Depending on the reader, m 6 A modification on mRNA can affect translation efficiency or mRNA stability (52), but how this epitranscriptomic modification mechanistically affects UCP1 expression has not yet been answered.
Besides these posttranscriptional regulations, interleukin 33 (IL-33) and ST2 (also known as IL-1 receptor 4, a receptor of IL-33) are required for accurate splicing of Ucp1 mRNA between exons 4 and 5 by using the splice site at nucleotide 864 on exon 5 ( Figure 2C). Without IL-33 or ST2, the splicing machinery prefers the cryptic nucleotide 972 site on exon 5 to produce a non-functional transcript (53). Depletion of either protein results in the loss of UCP1 in brown and beige adipocytes, so IL-33/ST2 signaling is important to register brown and beige adipocytes for uncoupled respiration and thermogenesis during the perinatal period (53).

DISCUSSION
Treating lean healthy men with a specific β3AR agonist, CL316243, increased insulin sensitivity and fat oxidation without affecting body weight (54), but 4-week administration of another β3AR agonist, L796568, in obese men did not confer evident metabolic benefits (55). The lack of chronic effect on shifting energy balance for weight loss may be due to few β3ARresponsive adipose tissues in humans. Indeed, two recent studies found a very low level of β3AR mRNA in human BAT (56,57). Using cultured human brown adipocytes, one study reported that β2AR activation increased UCP1 level and lipolysis (56), but the other study identified β1AR signaling as responsible for upregulating thermogenic gene expression (57). Because the activation of β1AR and β2AR has a profound impact on cardiopulmonary function, the positive regulatory mechanisms downstream of β-adrenergic signaling must be targeted to promote UCP1 synthesis, energy expenditure, and metabolic fitness in humans. Notably, the mouse model has been used to reveal mechanisms controlling UCP1 synthesis, thermogenesis, and possible implications for human obesity. However, the loss of the CCAAT sequence and Sp1-binding motif in the proximal promoter (22) and alternative polyadenylation in human UCP1 (21) supports that UCP1 expression is regulated by speciesdependent variations. For example, despite comparable UCP1 activity in mouse and human BAT (58), we reported that the expression of human Ucp1 mRNA in five different BAT samples is only 1/30-1/300 the level in mice (21). Our result agrees with a Northern blot study, which detected 1.9-kb UCP1 mRNA by using 5 µg poly(A) RNA from human perirenal BAT. Thus, absolute RT-qPCR was used to compare the amount of Ucp1 mRNA between human and mouse BAT (21). Because Ucp1L is translated 15 times more efficiently than Ucp1S in mouse BAT, we predicted that CPEB2-mediated translational activation, if fully functioning in human BAT, can make up the protein level to ∼1/2-1/20 of that in mouse BAT (21). Similarly, if the mechanism of BRF1-mediated deadenylation-induced Ucp1L decay (47) is conserved in humans, it is expected that HFDinduced UCP1 mRNA degradation in human beige adipocytes should become more prominent. Thus, humans, and perhaps also other primates, may rely more on posttranscriptional regulation to control UCP1 synthesis because they express a low level of only long 3 ′ -UTR UCP1 mRNA. Moreover, although ectopic expression of mouse Ucp1 in white adipocytes of pigs also enhances thermogenic capacity and lipolysis to decrease fat deposition (59), why UCP1-mediated metabolic plasticity is lost in some mammals through evolution remains unclear. Therefore, the results from studying gene-modified mice in BAT thermogenesis need to be translated with caution into humans and other species.

AUTHOR CONTRIBUTIONS
W-HL wrote the manuscript and illustrated the figures. Y-MC analyzed the poly(A) sites in Ucp1 across species. Y-SH co-wrote the manuscript and is responsible for its content. All authors contributed to the article and approved the submitted version.

ACKNOWLEDGMENTS
The authors thank Hao-Wen Chen for the illustration in Figure 1A and Hsi-Nien Chen from the Taipower Research Institute, Taiwan Power Company, for helping with the construction of the phylogenetic tree in Figure 1C.