An updated ANGPTL3-4-8 model as a mechanism of triglyceride partitioning between fat and oxidative tissues

Abstract

In mammals, triglyceride (TG), the main form of lipids for storing and providing energy, is stored in white adipose tissue (WAT) after food intake, while during fasting it is routed to oxidative tissues (heart and skeletal muscle) for energy production, a process referred to as TG partitioning. Lipoprotein lipase (LPL), a rate-limiting enzyme in this fundamental physiological process, hydrolyzes circulating TG to generate free fatty acids that are taken up by peripheral tissues. The postprandial activity of LPL declines in oxidative tissues but rises in WAT, directing TG to WAT; the reverse is true during fasting. However, the molecular mechanism in regulating tissue-specific LPL activity during the fed-fast cycle has not been completely understood. Research on angiopoietin-like (ANGPTL) proteins (A3, A4, and A8) has resulted in an ANGPTL3–4-8 model to explain the TG partitioning between WAT and oxidative tissues. Food intake induces A8 expression in the liver and WAT. Liver A8 activates A3 by forming the A3–8 complex, which is then secreted into the circulation. The A3–8 complex acts in an endocrine manner to inhibit LPL in oxidative tissues. WAT A8 forms the A4–8 complex, which acts locally to block A4’s LPL-inhibiting activity. Therefore, the postprandial activity of LPL is low in oxidative tissues but high in WAT, directing circulating TG to WAT. Conversely, during fasting, reduced A8 expression in the liver and WAT disables A3 from inhibiting oxidative-tissue LPL and restores WAT A4’s LPL-inhibiting activity, respectively. Thus, the fasting LPL activity is high in oxidative tissues but low in WAT, directing TG to the former. Future research on A3, A4, and A8 can hopefully provide more insights into human health, disease, and therapeutics.

1. Introduction

Evolution has resulted in a mechanism by which the human body stores energy during feasting while using the energy that has been stored during famine. Triglycerides (TG) are the main form of lipids to store and provide energy. As a fundamental physiological process, TG is stored in white adipose tissue (WAT) after food intake while during fasting it is routed to oxidative tissues (heart and skeletal muscle) for energy production, a process referred to as TG partitioning. Because TG is not water soluble, to enable it to circulate in the blood, TG is emulsified by proteins, forming TG-rich lipoproteins, which include chylomicrons and very-low-density lipoprotein (VLDL).

Chylomicrons are formed after a meal from dietary TG in mucosal cells within the villi of the duodenum, while VLDL is produced mainly during fasting in the liver by TG synthesis and is secreted directly into the bloodstream. These TG-rich lipoproteins transport and distribute TG to various tissues for either storage or oxidation. The partitioning of TG between WAT (TG storage) and oxidative tissues (TG oxidation) is largely dependent on lipoprotein lipase (LPL) [1–6].

Research in past decades has established that LPL, discovered in 1943 [6, 7], is a rate-limiting enzyme for hydrolyzing TG presenting in circulating lipoproteins, generating free fatty acids that are taken up by peripheral tissues [3, 8], including the heart [9–11], muscle [12–14] and fat [4]. Postprandial LPL activity rises in WAT but declines in oxidative tissues, directing TG to WAT for storage. During fasting, LPL activity declines in WAT but increases in oxidative tissues, directing TG to muscles for energy production. Therefore, LPL is considered a gatekeeper in tissue-specific substrate delivery and utilization, and LPL activity is carefully and accurately orchestrated in a tissue-specific manner through the fed-fast cycle [1, 2, 4, 15–17]. A major mechanism to coordinate the physiological variations of LPL activity is through its interactions with other proteins, including apolipoproteins and members of the angiopoietin-like protein family, such as ANTPTL3, 4, and 8 (A3, 4, and 8) [4, 18, 19].

2. ANGPTL3, ANGPTL4, and some unanswered questions

Two ANGPTL family members, A3 and A4, have been well established as being critical regulators of TG metabolism through the inhibition of LPL activity.

In 2002, Koishi et al. identified a null mutation in A3 in a mouse line (KK/San) that exhibits extremely low serum TG levels, suggesting that A3 deficiency results in the phenotype of low TG levels [20]. A3, a hepatokine, is specifically expressed in the liver, which secretes A3 into the circulation. In mice, A3 overexpression or its deletion leads to hypertriglyceridemia or hypotriglyceridemia, respectively [21, 22]. It has been well established that A3 regulates TG metabolism by inhibiting the activity of LPL [23]. A3 has two functional domains, an N-terminal coiled-coil domain and a C-terminal fibrinogen-like domain. A3 is proteolytically cleaved by proprotein convertases to yield the N-terminal domain, which is sufficient and necessary for LPL inhibition [24, 25]. In humans, loss-of-function mutations of A3 or its therapeutic antagonism cause reduced plasma levels of TG, LDL-C, and HDL-C [26–33].

In 2000, multiple groups independently identified A4 as a novel ANGPTL family member induced by fasting via the peroxisome proliferator-activated receptor (PPAR) in adipocytes [34–36]. A4 is a potent LPL inhibitor [22, 37] and plays an important role in regulating LPL activity under the conditions of fasting and exercise [38]. A4, which has a domain structure similar to that of A3, is cleaved in the conserved proprotein convertase recognition sequence to release the N-terminal coiled-coil domain, which potently inhibits LPL [39, 40]. In mice, loss- and gain-of-function mutations of A4 result in lower and higher TG levels, respectively [22, 37, 41–44]. In humans, the E40K substitution and loss-of-function mutations in A4 result in low plasma TG levels [45–47].

In summary, a wealth of evidence from in vitro and in vivo studies, animal models with gain- and loss-of-functions studies, as well as human genetic studies has demonstrated the critical roles of A3 and A4 in regulating TG metabolism through inhibiting LPL. However, the relationships among A3, A4, and LPL, as well as how they are coordinated during the fed-fast cycle was far from being clear, especially when A8 was not included in the analysis. Below are some specific questions.

1. Although it was generally accepted that A3 negatively regulates LPL, many studies showed that A3 only inhibits LPL at supraphysiological concentrations in vitro [23, 24, 48, 49], and its potency is less than that of A4 [48, 50–53]. If A3 is not a potent LPL inhibitor, why, then, does A3 deficiency in humans and animal models have strong LPL-related phenotypes?

2. A3 KO mice were found to have higher LPL activity, especially in the fed state [54]. However, A3 levels are relatively stable, regardless of nutritional states [49, 50, 54, 55]. Why, then, is A3’s LPL-inhibiting activity more pronounced in the fed state?

3. Adipose A4 expression is quickly induced by fasting, but in the fed state, although A4 mRNA is much suppressed, there is still a significant amount of remaining A4 protein (e.g., refer to Fig. 4A in the reference [56]). Indeed, compared to the rapid change of A4 mRNA, the turnover of A4 protein is slow [57]. To enable high postprandial LPL activity in WAT, what is the mechanism, then, to block the activity of A4 protein that is already in WAT?

4. A3 is a hepatokine, and A4 is an adipokine (and possible hepatokine). When circulating in the bloodstream, how do A3 and A4 establish their tissue specificity when inhibiting LPL in the capillary? Is it possible that A3 and A4 inhibit LPL in both WAT and oxidative tissues?

Overall, based on LPL, A3, and A4 only, without considering A8, the picture of how the tissue-specific activity of LPL is regulated during the fast-fed cycle was obscure.

3. ANGPTL8 and the ANGPTL3–4-8 model

In 2012 and 2013, we and others independently reported the functional roles of a previously uncharacterized gene, Gm6484, under names including lipasin [58], RIFL [59], ANGPTL8 [60], betatrophin, and C19ORF80 [61], which was later officially named ANGPTL8 [58–62]. Specifically, the proposed functions of A8 included regulation of TG metabolism [58], adipogenesis [59], A3-dependent TG metabolism [60], and autophagy [61]. Thus far, the most studied function has been the role of A8 in regulating TG metabolism, which is the focus of the current review.

A8 expression, highly enriched in the liver, WAT, and brown fat (BAT), is reduced by fasting and is highly induced by feeding in both the liver and WAT [58–60, 63, 64]. A8 plays a key role in regulating liver clock [65], and its expression is under the regulation of AMP-activated protein kinase [66] and hepatocyte nuclear factor 1 [67]. In BAT, A8 is up-regulated by cold exposure [68]. By using adenovirus-mediated overexpression, we overexpressed A8 in the mouse liver, and A8 overexpression dramatically increased serum TG levels [58]. Quagliarini et al. found that A8 interacts with A3, and that this interaction is needed for increased serum TG mediated by A8 overexpression. Consistently, A8-null mice have lower serum TG, especially in the fed state, likely due to enhanced plasma TG clearance resulting from increased post-heparin LPL activity [59, 69, 70]. Human circulating levels of A8 are increased in obesity, and are associate with metabolic parameters [50, 71–75]. These initial lines of evidence clearly showed a role of A8 in LPL-mediated TG metabolism [58, 76].

The discovery of A8 initially was puzzling. A8 was found to negatively regulate LPL activity, because A8 KO mice had higher post-heparin LPL activity [69]. Consistently, A8 KO mice had lower circulating TG [59, 69, 70], while A8 hepatic overexpression increased circulating TG [58]. These phenotypes are similar to those of mouse models with altered expressions of A3 and A4. Indeed, A3 KO mice had lower TG, while A3 overexpressing mice had higher TG [20–22]. Likewise, A4 KO mice had lower TG and A4 overexpressing mice had higher TG [22, 37, 41–44]. Thus, it was puzzling that the human genome encodes three LPL inhibitors, which seem somewhat redundant, because in terms of circulating TG, the phenotypes of the KO mice for all three are the same (lower TG levels), and those of overexpressing mice are the same (higher TG levels).

In 2015, we found that mice with A8 antibody (Ab) injections had higher cardiac LPL activity in the fed state [70]. This result immediately suggested an ANGPTL3–4-8 model that explains the partitioning of the circulating TG between WAT and oxidative tissues [70]. The model states that following food intake, liver A8 activates A3 by forming the A3–8 complex, which, after being secreted into the circulation, inhibits LPL in oxidative tissues, while WAT LPL is activated due to diminished A4 expression. Therefore, postprandial LPL activity is low in oxidative tissues, but is high in WAT, promoting circulating TG to WAT for storage. During fasting, reduced A8 expression inactivates A3, enabling elevated LPL activity in oxidative tissues, while WAT LPL activity is inhibited due to induced A4 expression. Therefore, fasting LPL activity is high in oxidative tissues but is low in WAT, promoting TG to oxidative tissues for oxidation [18, 70]. Thus, rather than being redundant, all the three proteins are essential to TG metabolism.

The model, to some extent, was an attempt to address the aforementioned questions regarding A3, A4, and LPL. Specifically, below are answers to the questions.

1. The reason why in vitro A3 is not as potent as A4 in inhibiting LPL is that A3 needs to be activated by A8.

2. The reason why A3 KO mice have more pronounced TG phenotypes in the fed state, despite A3 levels being stable regardless of nutritional states, is that A3 is activated by A8, which is induced by food intake.

3. The answer to question 3 is discussed in another section below. Briefly, A8 disables WAT A4 from LPL inhibition.

4. The model stated that A3-A8 complexes inhibit LPL in oxidative tissues, while A4 mainly acts locally to regulate TG uptake in adipose tissues [18, 70].

4. The missing evidence in the original ANGPTL3–4-8 model

The model was supported by some evidence that was available in 2015. These lines of evidence included that A8 is a feeding-induced hepatokine, that A8 negative regulates LPL activity, and that an A8 Ab increases LPL activity in oxidative tissues. In 2015, however, the model was, in fact, quite hypothetical due to limited evidence. Below is some critical evidence that was missing at the time.

1. The mechanism of action of A8 was unclear. It was shown that A3 and A8 interact [60], but the functional implication of this interaction was unclear. That is, it was not known whether this interaction is needed to regulate LPL activity. An alternative hypothesis at the time was that A8 facilitates A3 cleavage, releasing its active N-terminal domain for LPL inhibition [60].

2. The origin of circulating A8 was not clear. A8 is highly enriched in the liver and fat, but the origin of circulating A8 was not known. Is it secreted by the liver alone or from both the liver and fat? In other words, does A8 function in an endocrine, autocrine, or paracrine manner?

3. The functional roles of A8 in WAT were indeed puzzling. Food intake dramatically induces A8 expression in WAT, in which postprandial LPL activity is increased. Yet, if A8 negatively regulates LPL activity, as in the case of A3–8 interaction, then why is WAT A8 so robustly induced by food intake?

4. A challenging question was–and still is–about the tissue specificity of the A3–8 complex. The complex circulates in the blood, where it meets LPL in both WAT and oxidative tissues. Why, then, does it only inhibit LPL in oxidative tissues but not in WAT?

Since 2015, when the model was originally proposed [70], significant progress has been made in revealing the functional roles and the mechanism of action of A8 in particular and the relationship among A3, A4, A8, and LPL in general. The above issues, although apparently still being actively studied, have been answered to some extent. Below, we discuss the questions based on the emerging evidence.

5. Progress in elucidating ANGPTL8 biology

5.1. Functional implications of ANGPTL3–8 interaction.

A8 is an atypical member of the ANGPTL family, because it does not have the C-terminal domain that is found in all other ANGPTL family members [60, 68]. Specifically, A3 is cleaved, releasing the N-terminal domain, which can bind and inhibit LPL. A8 shares homology with A3’s N-terminal domain that possesses an LPL binding region. Hobbs’s group found that A3 and A8 interact and form a complex, suggesting that they function in the same pathway. A3 and A8 can be pulled down together from the plasma, suggesting their in vivo interactions [60]. Of note, A8 was found to enhance A3 cleavage [60]. This result suggests that one possibility is that A8 activates A3 by enhancing its cleavage, releasing its N-terminal domain, which in turn inhibits LPL [60]. Unexpectedly, however, in A8 KO mice, A3 cleavage was not decreased, likely due to a more complex in vivo regulation [69]. These results make the hypothesis that A8 functions by enhancing A3 cleavage unlikely.

The other hypothesis is that A8 forms a complex with A3 to enhance its LPL-inhibiting activity. Haller et al. found that co-expression of A3 and A8 leads to a far more efficacious increase in serum TG levels in mice than A3 alone. An Ab specific to the C terminus of A8 reversed LPL inhibition by A8 in the presence of A3. Collectively, these data show that A8 has a functional LPL inhibitory motif but only inhibits LPL and increases plasma TG levels in the presence of A3 [77].

Davies’s laboratory examined the functional implication of this interaction in terms of LPL activity [49]. They found that A3–8 complexes have a dramatically increased ability to inhibit LPL compared to either protein alone. At physiological levels, A3 or A8 alone had minimal LPL-inhibiting activity, especially when LPL was bound to its endothelial cell receptor/transporter GPIHBP1. A3–8 complexes bind to LPL much better than either protein alone. Hepatic A3 overexpression in mice increased serum TG only in the presence of A8 [49]. Therefore, A8 and A3 are interdependent when regulating LPL activity.

Konrad’s laboratory comprehensively examined the A3–8 interactions, their LPL inhibiting activities, and their concentrations in human plasma [50]. LPL inhibitory activity of A3–8 is more than 100-fold higher than A3 alone. And they further demonstrated that in the human blood, there are circulating A3–8 complexes, and that food intake significantly increased the amount of circulating A3–8 complexes [50].

The evidence from the above studies has demonstrated that A3 and A8 form a complex, which is a potent LPL inhibitor (Fig. 1).

Fig. 1. A8 activates A3 but inactivates A4 to regulate LPL activity.

A) A3 by itself only has low LPL-inhibiting activity. A8 activates A3 by forming A3–8 complexes, which have much increased LPL-inhibiting activity (more than 100-fold more active, compared to A3 alone). A4 by itself strongly inhibits LPL. A8 inactivates A4 by forming A4–8 complexes, which have much decreased LPL-inhibiting activity (more than 100-fold less active, compared to A4 alone). B) Food intake strongly induces A8, which, in turn, forms A3–8 and A4–8 complexes to balance LPL activity in different nutritional states.

5.2. Origin of circulating ANGPTL8.

A8 is highly enriched in the liver and fat, and A8 is present in the circulation [58–60]. It was clear that A8 can be secreted from the liver in the circulation [60], but is A8 also secreted from the fat? In other words, A8 is a hepatokine, but is it also an adipokine? An effective approach to address this question is to study phenotypes of mice with tissue-specific deletion of A8. Hobbs’s group generated mice with liver-specific (LKO) and adipose-specific (AKO) deletion of A8 [78]. Mice lacking hepatic A8 have phenotypes similar to those of the whole-body KO mice, that is, low plasma TG levels and high intravascular LPL activity, strongly suggesting that liver A8 functions in an endocrine manner. Importantly, mice lacking A8 specifically in adipose tissues have no detectable circulating A8. These results strongly suggest that liver A8 functions in an endocrine manner, while adipose A8 functions in an autocrine/paracrine manner.

5.3. Relationship between ANGPTL8 and ANGPTL4.

A8 is highly induced by food intake in WAT [58, 59]. The function of A8 in WAT was puzzling, because food intake up-regulates LPL activity in WAT. Why, then, is WAT A8 so robustly induced?

Soon after A8 was identified, we noticed that A8 and A4 seem to be reciprocally regulated and inherently related [68, 79]. For example, in the liver and WAT, food intake strongly induces A8 but suppresses A4; conversely, fasting reduces A8 but strongly induces A4 [58–60]. In the liver, insulin resistance induces A8 but suppresses A4 (refer to Fig. 2 in the reference [79]). In BAT, cold exposure induces A8 but reduces A4 [68]. The reciprocal regulation of A8 and A4 seems to suggest that the two molecules work together and are involved in the same cellular processes [79].

Fig. 2. An ANGPTL3–4-8 model for triglyceride partitioning between adipose and oxidative tissues.

A) Food intake induces A8 expression in the liver and white adipose tissue (WAT). Liver A8 activates A3, forming the A3–8 complex (protein ratio, 3:1), which is then secreted into the circulation. As a potent LPL inhibitor, the A3–8 complex in turn inhibits LPL in oxidative tissues. WAT A8 mainly acts locally to inactivate A4 by forming the A4–8 complex (protein ratio, 1:1), which disables A4 from LPL inhibition. That is, the A3–8 complex, secreted by the liver, acts in an endocrine manner to inhibit oxidative-tissue LPL, while the A4–8 complex acts in an autocrine/paracrine manner to activate WAT LPL by disabling A4’s activity. Meanwhile, the A4–8 complex blocks the LPL-inhibiting activity of the circulating A3–8 complex, thus enabling it to specifically target LPL in oxidative tissues. Consequently, LPL activity is low in oxidative tissues but is high in WAT, directing TG to WAT for storage. B) During fasting, A4 is induced in WAT, and A8 expression is diminished, enabling A4 to inhibit WAT LPL. Reduced circulating A3–8 complex levels, due to reduced liver A8 expression, disable A3 from LPL inhibition, resulting in high LPL activity in oxidative tissues. Therefore, fasting LPL activity is low in WAT but high in oxidative tissues, enabling TG uptake into the latter for oxidation.

In 2019, Kovrov et al. reported that A8 can not only form complexes with A3, but can also form complexes with A4. In contrast to the A3–8 complex, the LPL inhibiting activity of which is higher than A3 alone, A4–8 LPL inhibiting activity is lower than that of A4 alone. In other words, A8 activates A3 but inhibits A4, in terms of LPL-inhibiting activity [68].

Konrad’s laboratory comprehensively examined the relations among A3, A4, and A8, and generated quantitative and informative results [50]. LPL-inhibitory activity of A3–8 complexes is more than 100-fold more potent than that of A3 alone, and LPL-inhibitory activity of A4–8 complexes is more than 100-fold less potent than that of A4 alone. Furthermore, the A4–8 complex also dramatically inhibits the LPL-inhibiting activity of A4. Using mass spectrometry, Konrad’s group further determined that the protein ratios in the endogenous A3–8 and A4–8 complexes were 3:1 and 1:1, respectively [50].

The result that A8 inactivates A4 has important physiological implications. After food intake, even though A4 mRNA is diminished, a significant amount of protein still remains. For example, refer to Fig. 4A in the reference [56]. Indeed, it was shown that in rat WAT, A4 mRNA levels responded quickly to fasting and refeeding, but the turnover of A4 protein was slow. Specifically, six hours after the injection of the transcription blocker actinomycin D in fasted rats, although A4 mRNA levels decreased by 93%, its protein levels remained unchanged [57]. Therefore, there should be a mechanism for blocking the residual A4 protein from inhibiting LPL during the transition from fasting to refeeding. A8, induced by refeeding, forms the A4–8 complex to greatly inhibit the activity of residual A4 protein. An alternative mechanism is that A8 promotes A4 protein degradation [78]. Consistent with A4’s functional roles in WAT, Davies’s laboratory showed that A4 KO mice have increased uptake of radiolabeled TG in WAT, suggesting that A4 primarily functions locally in adipose tissue to inhibit LPL and regulate TG uptake [80, 81].

5.4. The mechanism that determines tissue specificity of the ANGPTL3–8 complex.

The A3–8 complex circulates in the bloodstream, and presumably, it encounters LPL in both oxidative tissues and WAT. Why, then, does the A3–8 complex only inhibit LPL in oxidative tissues but not in WAT? In other words, what determines the tissue specificity A3–8 activity on LPL inhibition? Obviously, this was–and still is–a challenging question. But Konrad et al. suggested a potential mechanism [50].

Konrad’s group comprehensively examined the relationship among A8, A4, and A3, as well as their complexes, in terms of LPL-inhibiting activities. Interestingly, the A4–8 complex dramatically inhibits the LPL-inhibiting activity of the A3–8 complex, and this result has profound implications. Locally expressed A4 forms a complex with A8, which binds the adipose-associated LPL, and therefore the A4–8 complex can block the ability of circulating A3–8 to inhibit LPL in WAT. That is, the WAT A4–8 complex protects LPL from being inhibited by circulating A3–8 complexes. However, A4–8 complexes are not present in oxidative tissues, where A8 is not expressed, and thus the A3–8 complex can inhibit LPL in oxidative tissues. This elegant hypothesis suggested by Konrad’s group provides a potential mechanism that explains the tissue specificity of the A3–8 complex.

Food intake highly induces WAT A8 expression, forming the A4–8 complex. Therefore, it is likely that the A4–8 complex protects WAT LPL from being inhibited by A4 (to enable high WAT LPL activity) and at the same time also protects WAT LPL from being inhibited by the circulating A3–8 complex, to ensure its inhibition of oxidative-tissue LPL. Consistently, previous studies showed that both intracellular A4 [82, 83] and extracellular A4 [84, 85] can function to regulate LPL activity.

Collectively, these results strongly suggest that A8 activates A3 but inactivates A4 by forming complexes with the two proteins. Therefore, A8 can be regarded as a molecular switch balancing the flow of circulating TG according to the needs of the body during the fed-fast cycle [50, 86] (Fig. 1).

6. An updated ANGPTL3–4-8 model

Significant progress has been made in elucidating the function and mechanism of action of A8, providing new evidence and information for an updated ANGPTL3–4-8 model.

Following food intake, A8 expression is induced in both the liver and WAT. In the liver, A8 activates A3 by forming the A3–8 complex (3:1 ratio), which is then secreted into the circulation. The A3–8 complex, a potent LPL inhibitor, inhibits LPL in oxidative tissues, making circulating TG available for uptake by WAT. In WAT, A4 mRNA is diminished, despite a significant amount of A4 protein remaining. A8, highly induced in WAT, forms the A4–8 complex (1:1 ratio), which blocks A4’s LPL-inhibiting activity. At the same time, the A4–8 complex blocks the LPL-inhibiting activity of the circulating A3–8 complex, thus enabling it to specifically target LPL in oxidative tissues. Therefore, the postprandial LPL activity is low in oxidative tissues but high in WAT, enabling the uptake of circulating TG into WAT for storage (Fig. 2).

Conversely, during fasting, A8 expression in WAT is diminished, and thus there is less A8 and consequently fewer A4–8 complexes. A4’s LPL-inhibiting activity is then restored, and therefore WAT LPL activity is low. At the same time, liver A8 expression is reduced, resulting in a lower level of circulating A3–8 complexes and hence less inhibition of LPL in oxidative tissues. Therefore, the fasting LPL activity is low in WAT but high in oxidative tissues, enabling the uptake of circulating TG into the latter for oxidation (Fig. 2).

7. Explanation of the phenotypes of ANGPTL8 tissue-specific KO mice

We used the ANGPTL3–4-8 model in explaining the phenotypes of A8-overexpressing and whole-body KO mice [18, 70]. Two dramatic phenotypes are noteworthy. 1) In WT mice the serum TG level is reduced by fasting; in A8-overexpressing mice, however, it is dramatically increased by fasting. We reasoned that this is because in the A8-overexpressing mice during fasting, exogenous A8 still activates A3 to inhibit LPL in oxidative tissues, and thus LPL activity is low in both oxidative tissues and WAT, leading to TG accumulation [18, 70]. 2) In WT mice the serum TG level is generally increased by feeding; in A8-KO mice, however, it is reduced by feeding. We reasoned that in the KO mice after food intake, LPL activity is high in both WAT and oxidative tissues (due to the absence of the A3–8 complex), hence the reduced TG levels [70].

Hobbs’s group recently generated A8 liver-specific KO mice (LKO) and adipose-specific KO mice (AKO) [78]. These mouse lines provide critical information about tissue-specific functions of A8. According to the ANGPTL3–4-8 model, we would expect that in LKO mice, since A8 is deleted in the liver, no A8 is secreted from it. The absence of A8 in the circulation inactivates A3, hence the higher LPL activity in oxidative tissues. That is, LPL activity is high in both WAT and oxidative tissues in the fed state (Fig. 3A). Post-heparin LPL activity reflects an overall intravascular LPL activity in both WAT and oxidative tissues, and thus we expect that post-heparin LPL activity is high and serum TG levels are low in LKO mice. In other words, the LKO mice have a phenotype similar to that of whole-body KO mice in terms of serum TG levels (Fig. 3A). This is consistent with the phenotypes of the LKO mice, which have higher post-heparin LPL activity and lower TG levels in the fed state [78].

Fig. 3. Phenotypes of tissue-specific A8 KO mice explained by the ANGPTL3–4-8 model.

A) Liver-specific A8 KO (LKO) mice show hypotriglyceridemia in the fed state. In LKO mice, A8 is specifically deleted in the liver, and therefore the liver fails to secrete A8 in the fed state, resulting in the absence of circulating A3–8 complexes and consequently the inability to inhibit LPL in oxidative tissues. Because adipose A8 is still induced, forming A4–8 complexes that block A4 from LPL inhibition. Therefore, LPL activity is high in both oxidative and adipose tissues, resulting in high post-heparin LPL activity and hypotriglyceridemia. B) Adipose-specific A8 KO (AKO) mice show hypertriglyceridemia in the fed state. Following food intake, liver A8 forms A3–8 complexes, which, after being secreted into circulation, inhibit LPL in oxidative tissues. Because A8 is specifically deleted in adipose tissues, A8 absence activates A4, hence reducing adipose LPL activity. Therefore, LPL activities are low in both oxidative and adipose tissues, resulting in low post-heparin LPL activity and TG accumulation (hypertriglyceridemia).

The AKO mice, in contrast, have increased plasma TG levels, a phenotype that was described as “the most perplexing phenotype”, leading to a conclusion that “the increase in plasma TGs in As-A8^−/− [AKO] mice cannot be explained by previous models from our laboratory or others” [78].

According to the ANGPTL3–4-8 model, in AKO mice, because A8 is still expressed in the liver, it continues to form the A3–8 complex, which, after being secreted into the circulation, inhibits LPL in oxidative tissues. However, because A8 is deleted in adipose tissues, A4 cannot be inactivated in the fed state, and thus A4 activity remains high to inhibit LPL. Therefore, adipose LPL activity becomes lower, compared to that of WT mice. In other words, in the fed state, LPL activity is low in both WAT and oxidative tissues in AKO mice, resulting in circulating TG to accumulate (hypertriglyceridemia) (Fig. 3B). Therefore, we would expect a lower post-heparin LPL activity (due to low LPL activity in both WAT and oxidative tissues) and higher serum TG levels in AKO mice (Fig. 3B). And this is exactly the phenotype of AKO mice (lower post-heparin LPL activity and hypertriglyceridemia) [78]. Therefore, these phenotypes, although described as “perplexing” [78], do fit the ANGPTL3–4-8 model.

8. Conclusion and perspective

Although a great amount of information has been obtained in the past 9 years regarding A8’s function and mechanism of action as well as the relationship among A8, A3, A4, and LPL, more questions still need to be answered. Many of these questions have been excellently summarized in a recent review [19]. These questions include mechanism of action, such as the structural basis for interactions among A8, A3, A4 and LPL, and the mechanism by which A8 activates A3 but inactivates A4. These questions also include A8’s role in physiological and pathological states, such as cold-induced thermogenesis [68, 79, 87, 88], exercise, starvation, inflammation, and aging [19]. Here, we would like to emphasize three areas.

8.1. The mechanism by which the ANGPTL3–4-8 system regulates LPL activity.

One question is the mechanism of action for LPL inhibition. In past decades, much knowledge has been gained regarding the mechanisms by which A3 and A4 inhibit LPL. Regarding A3, one mechanism is that A3 promotes dissociation of LPL dimers into inactive monomers [52]. LPL can be cleaved to become inactive by furin-like proprotein convertases [89, 90], and thus another mechanism is that A3 increases LPL cleavage [91]. A3 can also inhibit LPL by unfolding or sterically changing the active domain of LPL [92]. Regarding A4, it was shown that A4 catalytically inactivates LPL [93–96] and can convert LPL into inactive monomers [93]. A4 can inhibit LPL by unfolding or sterically changing the active domain of LPL [92].

Although common mechanistic schemes have been proposed for A3 and A4, such as converting LPL dimer into monomers, it has also been shown that A3 and A4 use distinct mechanisms to inhibit LPL [51]. LPL monomers have been found to be catalytically active [95–98]. These results make the mechanism of converting dimers into monomers open to debate. Therefore, the mechanisms by which A3 and A4 inhibit LPL are far from being completely understood. It is now clear that A8 interacts with A3 and A4 to activate and inactivate the former and the latter, respectively. A recent study showed that A3 and A8 interact with LPL to induce its cleavage [99]. Therefore, it would be more physiologically relevant to re-investigate the LPL inhibition mechanisms by including A8.

It should be noted that TG hydrolysis by LPL happens on the platform of GPIHBP1, an endothelial cell receptor for LPL [100–102]. Indeed, GPIHBP1 protects LPL from being inhibited by A3 and A4. [48, 94, 95, 98]. Therefore, it appears that, physiologically, A3, A4, A8, LPL, and GPIHBP1 work together, and future models, therefore, may include all these proteins.

8.2. The mechanism by which the ANGPTL3–4-8 system regulates endothelial lipase activity.

Both A3 and A8 have been implicated in regulating HDL-C metabolism, likely through the regulation of the activity of endothelial lipase (EL). A3 regulates plasma HDL-C levels by inhibiting EL, which hydrolyzes phospholipids of HDL [103–105]. In humans, ANGPTL3 deficiency or its therapeutic antagonism reduces plasma HDL-C levels, due to reduced inhibition of EL [28]. Likewise, elevated ANGPTL3 levels increase plasma HDL-C levels due to enhanced inhibition of EL [103].

Human genome-wide association studies (GWAS) have robustly linked A8 gene variants to HDL-C plasma levels. The SNP rs2278426 leads to an amino acid (AA) change at residue 59 from arginine to tryptophan (R59W), and the 59W variant has been linked to lower HDL-C levels in multiple ethnic groups [60, 106]. The SNPs rs145464906 and rs760351239 are characterized by premature stop codons leading to A8 truncations, 120 AA and 130 AA, respectively, while the full-length protein is 198 AA. Both SNPs are linked to increased HDL-C levels in multiple ethnic groups [107, 108]. Two recent studies have investigated how A8 regulates A3 in terms of EL inhibition [109, 110], providing a mechanism by which A3 and A8 coordinately regulate HDL-C metabolism.

A4 gene variants have also been robustly linked to human plasma HDL-C levels by GWAS. For instance, one A4 variant (E40K) is associated with higher HDL-C levels [45, 111]. Interestingly, Konrad’s group found that A4 is an even more potent EL inhibitor than A3, and that A8 reduces A4’s while increasing A3’s EL-inhibiting activity [109], providing a potential mechanism by which A4 influences HDL-C levels. These lines of emerging evidence suggest that the ANGPTL3–4-8 system also coordinately regulates EL’s activity [109], in addition to regulating LPL’s. Therefore, an important future direction is to study the mechanisms by which the ANGPTL3–4-8 system regulates EL activity under different physiological and pathological conditions, and these studies may shed light on the metabolism of HDL-C.

8.3. Therapeutic potential of targeting ANGPTL8

From the perspective of evolution, insufficient caloric intake was a major survival threat during human evolution [112]. The thrifty gene hypothesis suggests that carriers of the thrifty genes survived because they deposited more fat between famines, and therefore evolutionary selection favored genes and genetic variants that led to accumulation of adipose stores [113]. A major function of A8 is to promote fat storage after food intake, and thus A8 can be considered a thrifty gene. As correctly pointed out by Konrad’s group, in a world of caloric abundance, the same A8 protein that likely protected human ancestors from starvation now predisposes humans to metabolic syndromes, because constant feeding chronically increases circulating A8 levels, leading to increased adipose storage (obesity) and elevated circulating TG (hypertriglyceridemia) [50].

The idea that A8 is a thrifty gene suggests that A8 antagonism has the potential to reverse the thrifty phenotypes, i.e., to reduce obesity, circulating TG levels, and metabolic syndromes. Indeed, A8 deficiency in mice leads to decreased serum TG levels [59, 69, 70] and decreased adiposity [69]. Consistently, injection of an A8 Ab in mice reduced circulating TG levels [70, 114] and adiposity [114].

Of note, TG and HDL-C metabolism are interconnected [115], and the ANGPTL3–4-8 system regulates both, as mentioned above. We recently proposed that A8 antagonism has the potential to simultaneously reduce TG and increase HDL-C plasma levels [116]. The rationale is that the circulating A3 can be classified as A8-associated A3 and free A3, where the former inhibits LPL and the latter inhibits EL. An A8 Ab or inhibitor that disrupts or precludes A3–8 complex formation results in fewer A3–8 complexes but more free A3, where the former leads to lower plasma TG (due to reduced LPL inhibition), and the latter leads to higher plasma HDL-C levels (due to enhanced EL inhibition) [116].

Consistently, injection of an A8 monoclonal Ab in spontaneous dyslipidemic cynomolgus monkeys reduced circulating TG levels by 65% and increased HDL-C levels by 30% [114]. Naturally occurring loss-of-function human genetic variants provide an in vivo model of human gene inactivation and are a valuable source to guide the selection of drug targets [117]. Carriers of the SNP rs145464906 (truncated A8; 120 AA) had lower TG (−15%) and 10 mg/dL higher HDL-C levels than did noncarriers in a population composed of European ancestry [107]. In an independent dataset (UK Biobank), carriers of this SNP showed 18.9 mg/dL lower TG and 6.1 mg/dL higher HDL-C plasma levels [108]. Carriers of the SNP rs760351239 (truncated A8; 130 AA) had 24.0 mg/dL lower TG and 9.1 mg/dL higher HDL-C levels in a cohort sampled from the Finnish population [108]. These lines of evidence strongly suggest that A8 antagonism results in lower TG and higher HDL-C plasma levels.

Because both elevated TG and reduced HDL-C plasma levels are risk factors for atherosclerosis and cardiovascular disease [118], a drug that simultaneously achieves both goals—reducing TG and increasing HDL-C plasma levels—has the potential to prevent and treat these diseases. Consistently, in the FinnGen study, carriers of the A8 truncating variant (rs760351239) had a reduced possibility of coronary artery disease (OR = 0.53) and were less likely to use statin therapy (OR = 0.52) than noncarriers [108]. These results support the hypothesis that A8 antagonism reduces cardiovascular disease risks. Future studies to identify human SNPs that result in a complete A8 deficiency—rather than truncating variants, which are likely hypomorphic—and to conduct clinical trials with A8 Abs or inhibitors will confirm the hypothesis of A8 antagonism-based therapeutics [116].

In conclusion, we here update the ANGPTL3–4-8 model by providing additional supporting evidence that spatially and temporally regulated A3, A4, and A8 act coordinately to balance TG partitioning between WAT and oxidative tissues during the fed-fast cycle. Future research on A3, A4, and A8 can hopefully provide more insights into human health, disease, and therapeutics.

Abbreviations:

A3

ANGPTL3

A4

ANGPTL4

A8

ANGPTL8

AA

amino acid

AKO

adipose-specific KO mice

ANGPTL

angiopoietin-like protein

BAT

brown adipose tissue

GPIHBP1

glycosylphosphatidylinositol anchored high density lipoprotein binding protein 1

GWAS

genome-wide association studies

HDL

high-density lipoprotein

HDL-C

high-density lipoprotein cholesterol

KO

knockout

LDL-C

low-density lipoprotein cholesterol

LKO

liver-specific KO

LPL

lipoprotein lipase

OR

odds ratio

PPAR

peroxisome proliferator-activated receptor

SNP

single nucleotide polymorphism

TG

triglyceride

VLDL

very-low-density lipoprotein

WAT

white adipose tissue