Apolipoprotein AV (APOA5), a low-abundance plasma protein, has a huge impact on intravascular triglyceride metabolism (1–3). APOA5 deficiency causes hypertriglyceridemia in mice and humans (1, 4–8), whereas overexpression of human APOA5 in mice reduces plasma triglyceride levels (1). For more than two decades, the mechanism by which APOA5 influences plasma triglyceride metabolism remained a mystery, but recent studies have started to bring APOA5 function into focus.
Chen et al. (9) discovered that APOA5 binds to the ANGPTL3/ANGPTL8 complex (ANGPTL3/8) and suppresses its capacity to inhibit lipoprotein lipase (LPL) catalytic activity. Kumari et al.
(10) discovered that ANGPTL3/8 inhibits LPL activity by catalyzing the unfolding of LPL’s hydrolase domain. Yang et al. (11) showed that the unfolding of LPL by ANGPTL3/8 results in detachment of LPL from the surface of cells and that ANGPTL3/8-mediated LPL detachment is blocked by APOA5.
APOA5’s ability to block LPL detachment preserves amounts of LPL along the luminal surface of capillaries (11). In the setting of APOA5 deficiency, ANGPTL3/8 activity is unsuppressed, resulting in inactivation of LPL catalytic activity, reduced amounts of LPL at the luminal surface of capillaries, and hypertriglyceridemia (11).
Clinical observations provided a clue regarding functionally important sequences in APOA5. An APOA5 frameshift mutation that truncates APOA5 by 35 residues was identified in two patients with severe hypertriglyceridemia (chylomicronemia) (7, 8), implying that C-terminal APOA5 sequences were important for its function. That observation prompted Chen, Yang, and coworkers (12) to investigate the properties of a mutant mouse APOA5 lacking 40 C-terminal residues (APOA5Δ40).
We removed 40 residues (rather than 35) from mouse APOA5 because it contains an extra five residues at its C terminus. Surface plasmon resonance studies revealed that wild-type (WT) mouse APOA5 bound ANGPTL3/8 with high affinity, whereas there was no binding of APOA5Δ40 to ANGPTL3/8 (12). Also, WT-APOA5, but not APOA5Δ40, blocked ANGPTL3/8’s capacity to inhibit LPL catalytic activity and detach LPL from cells (12).
Human genetic observations have provided no further insights. Exome and genome sequencing efforts have uncovered many nonsynonymous polymorphisms in the last 40 residues of APOA5 (13), but in most cases the significance of these polymorphisms is unknown. AlphaFold3 modeling (14) suggested that C-terminal APOA5 sequences (residues S330–L352) form an amphipathic α-helix (2, 12), but the identity of functionally important residues within that segment has remained unclear.
Monoclonal antibodies (mAbs) have proven useful for defining functionally important sequences in key proteins in lipoprotein metabolism, including APOB, LDLR, and the ANGPTL proteins (15–19). We reasoned that a mAb against the C-terminal sequences in APOA5—particularly if the mAb blocked APOA5’s ability to suppress ANGPTL3/8 activity—could be helpful in delineating functionally important residues in APOA5. We further suspected that experiments with an inhibitory mAb could provide additional insights into how APOA5 regulates ANGPTL3/8 activity and plasma triglyceride metabolism.
In the current studies, we generated, by phage panning, a mAb (IBA707) that bound to evolutionarily conserved sequences within the last 40 residues of APOA5. We tested the capacity of IBA707 to block APOA5-mediated inhibition of ANGPTL3/8 activity in biochemical studies, cell culture experiments, and mouse models. IBA707 proved to be useful for defining functionally important residues in APOA5 and for further elucidating the mechanism by which APOA5 regulates ANGPTL3/8 activity and intravascular triglyceride processing.
Results
Since C-terminal APOA5 sequences are strongly conserved in mice and humans and because phage panning has proven useful for creating mAbs against evolutionarily conserved sequences (20, 21), we screened a human Fab bacteriophage library (22, 23) for phage that bound to WT mouse APOA5 but not APOA5∆40 (SI Appendix, Fig. S1). We also reasoned that an in vitro approach with Fab bacteriophage screening would allow us to identify binding reagents with the desired specificity.
After three rounds of selection, a high affinity binder was identified. Using phage DNA sequences of the human variable regions and expression vectors for the mouse lambda 1 constant region and mouse gamma 1 constant region (Fc region), we built a chimeric IgG1 mAb (IBA707) that bound APOA5 (SI Appendix, Fig. S1).
The affinity of IBA707 binding to mouse APOA5, as judged by surface plasmon resonance (SPR), was high (KD = 0.81 nM; SI Appendix, Table S1).
IBA707 Binds Mouse APOA5 and Blocks Its Activity.
The binding of IBA707 to WT mouse APOA5 and APOA5∆40 was assessed with an immunoassay. APOA5 proteins were captured on a 96-well plate coated with mAb IBA720 (developed against a peptide containing the first 23 residues of mouse APOA5). APOA5 binding was then assessed with HRP-IBA707 or HRP-3298 (a rabbit polyclonal antibody specific for mouse APOA5) (SI Appendix, Fig.
S2). IBA707 bound WT-APOA5 but not APOA5∆40; 3298 detected both proteins (Fig. 1A).
IBA707 blocked the ability of WT mouse APOA5 to suppress ANGPTL3/8-mediated inhibition of mouse LPL activity (Fig. 1B). IBA707 also blocked the ability of APOA5 to inhibit ANGPTL3/8-mediated LPL detachment from the surface of CHO-K1 cells.
In these studies, we first incubated the cells with recombinant LPL, allowing LPL to bind to cell-surface HSPGs. ANGPTL3/8 reduced cell-surface LPL by ~55% (n = 5 independent experiments). WT-APOA5 blocked ANGPTL3/8-mediated LPL detachment, and IBA707 (but not an irrelevant control mAb 11A9 against epiregulin) blocked APOA5’s ability to inhibit ANGPTL3/8-mediated LPL detachment (Fig.
1 C and D and SI Appendix, Figs. S3 and S4).
IBA707 Also Binds Human APOA5 and Blocks Its Activity.
To test the ability of IBA707 to bind human APOA5, we captured human APOA5 on wells of a 96-well plate that had been coated with the human APOA5–specific mAb 2G1H11. We then tested the ability of HRP-IBA707 and HRP-3298 to detect human APOA5. IBA707 bound human APOA5, whereas 3298 did not (SI Appendix, Fig.
S5A). The affinity of IBA707 binding to human APOA5 was high (KD = 0.12 nM, SI Appendix, Table S1). IBA707 did not bind to a truncated human APOA5 lacking 34 C-terminal amino acids (SI Appendix, Table S1).
IBA707 blocked the ability of human APOA5 to suppress ANGPTL3/8-mediated inhibition of human LPL activity (SI Appendix, Fig. S5B) and blocked the ability of human APOA5 to suppress ANGPTL3/8-mediated detachment of LPL from cultured cells (SI Appendix, Fig. S5 C and D).
IBA707 Triggers Hypertriglyceridemia in WT Mice.
After an 18-h fast, we gave Apoa5+/+ mice an intravenous (IV) injection of IBA707 (or the irrelevant control mAb 11A9) and then allowed the mice to consume a chow diet. We anticipated that refeeding would increase plasma triglyceride levels in both groups of mice, but refeeding had little effect on triglyceride levels in 11A9-treated mice (Fig. 2 A and E).
In contrast, IBA707 triggered hypertriglyceridemia (triglyceride levels of 370.6 ± 152.1, 506.9 ± 252.7, and 512.6 ± 307.0 mg/dL after 4 h, 8 h, and 24 h, respectively) (Fig. 2 A and E). At 48 h, the triglyceride levels were similar in IBA707- and 11A9-treated mice (Fig.
2 A and E). IBA707 did not alter plasma triglyceride levels in Apoa5–/– mice (Fig. 2 B and E).
IBA707 also increased plasma triglyceride levels in Apoa5–/– mice that express a human APOA5 transgene (Apoa5–/–;hAPOA5-Tg) (SI Appendix, Fig. S6A).
IBA707 Increases Plasma Levels of ANGPTL3/8.
We were confident that the plasma ANGPTL3/8 levels would increase in IBA707- and 11A9-treated mice after refeeding, but we suspected that the increase might be greater in IBA707-treated mice. Indeed, that was the case. In 11A9-treated Apoa5+/+ mice, the plasma ANGPTL3/8 levels increased ~twofold after feeding (Fig.
2 C and E). In contrast, the ANGPTL3/8 levels in IBA707-treated Apoa5+/+ mice increased more than fivefold (from 8.4 ± 3.5 ng/mL at baseline to 46.6 ± 20.6, 65.0 ± 28.6, and 51.9 ± 23.9 ng/mL after 4 h, 8 h, and 24 h, respectively) (Fig. 2 C and E).
At the 4-h time point, ANGPTL3/8 levels and triglyceride levels were positively correlated (R2 = 0.405; P = 0.008). After 48 h, ANGPTL3/8 levels were similar in IBA707- and 11A9-treated mice (Fig. 2 C and E).
ANGPTL3/8 levels in IBA707- and 11A9-treated Apoa5–/– mice were not statistically different (Fig. 2 D and E). In IBA707- and 11A9-treated Apoa5–/–;hAPOA5-Tg mice, baseline ANGPTL3/8 levels were similar but after 4 h the ANGPTL3/8 levels were ~twofold higher in the IBA707-treated mice (SI Appendix, Fig.
S6B).
IBA707 Increases Plasma Triglyceride and ANGPTL3/8 Levels in Fasted Mice.
Plasma ANGPTL3/8 levels are low during fasting, reflecting low levels of Angptl8 expression in the liver (24–26). To assess the impact of APOA5 expression on triglyceride and ANGPTL3/8 levels during fasting, Apoa5+/+ and Apoa5–/– mice were fasted for 18 h; then given an IV injection of 11A9 or IBA707; and then fasted for an additional 8 h. In 11A9-treated Apoa5+/+ mice, the continuation of fasting reduced plasma triglyceride and ANGPTL3/8 levels (SI Appendix, Fig.
S7 A and C). In contrast, triglyceride and ANGPTL3/8 levels increased significantly in the IBA707-treated Apoa5+/+ mice (SI Appendix, Fig. S7 A and C).
In fasted Apoa5–/– mice, the triglyceride and ANGPTL3/8 levels in IBA707- and 11A9-treated groups were not statistically different (SI Appendix, Fig. S7 B and D).
Elevated Plasma Levels of ANGPTL3/8 in Apoa5–/– Mice.
The higher plasma levels of triglycerides and ANGPTL3/8 in IBA707-treated Apoa5+/+ mice suggested that APOA5 lowers plasma triglyceride levels at least in part by reducing amounts of ANGPTL3/8 in the plasma. In support of this idea, plasma ANGPTL3/8 levels were higher in Apoa5–/– mice than in Apoa5+/+ mice (P = 0.003 in Fig. 2; P < 0.001 in SI Appendix, Fig.
S7). We also measured ANGPTL3/8 levels in the plasma from Apoa5–/– and Apoa5+/+ mice during ad libitum feeding, after a 15-h fast, and after 4 h of refeeding (11). The ANGPTL3/8 levels were ~two- to threefold higher in Apoa5–/– mice than in Apoa5+/+ mice during ad libitum feeding; ~55% higher in Apoa5–/– mice after a 15-h fast; and ~two- to threefold higher in Apoa5–/– mice after 4 h of refeeding (SI Appendix, Fig.
S8). In each setting, the plasma triglyceride levels were substantially higher in Apoa5–/– mice than in Apoa5+/+ mice (11). Also, in Apoa5–/– and Apoa5+/+ mice fed ad libitum, the Angptl3 and Angptl8 transcript levels in the liver did not differ (11).
Rapid Clearance of IBA707 in Apoa5+/+ Mice.
After injecting Apoa5+/+ mice with IBA707, the plasma triglyceride and ANGPTL3/8 levels were higher at 4, 8, and 24 h but by 48 h declined to levels comparable to those in the 11A9 control group (Fig. 2). This time course was mirrored by the plasma levels of IBA707, which were high at 4 and 8 h, lower at 24 h, and virtually undetectable at 48 h (Fig.
3). In contrast, the IBA707 levels in Apoa5–/– mice remained high at the 24- and 48-h time points (Fig. 3).
The delayed clearance of IBA707 in Apoa5–/– mice presumably reflected the absence of IBA707’s target, namely APOA5 (27).
IBA707 Reduces LPL Levels in Capillaries.
We suspected that the hypertriglyceridemia in IBA707-treated Apoa5+/+ mice reflected reduced amounts of LPL in capillaries. To test this possibility, Apoa5+/+ mice were fasted overnight (18 h); given an IV injection of either IBA707 or 11A9; and then fed a chow diet. After refeeding for 8 h, the mice were given an IV injection of three Alexa Fluor–labeled mAbs: 27A7 (against LPL), 11A12 (against GPIHBP1), and 2H8 (against CD31).
After 10 min, the mice were killed; the vasculature was perfused with PBS and 4% PFA; and tissue sections were prepared for microscopy. The amount of LPL in capillaries, relative to GPIHBP1 or CD31, was assessed by the intensities of Alexa Fluor signals on capillaries. The LPL/GPIHBP1 and LPL/CD31 fluorescence intensity ratios in heart capillary segments were 69.9 and 61.6% lower, respectively, in IBA707-treated mice than in 11A9-treated mice (Fig.
4A and SI Appendix, Fig. S9A). In brown adipose tissue (BAT), the LPL/GPIHBP1 and LPL/CD31 ratios were 68.8% and 53.9% lower, respectively, in IBA707-treated mice (Fig.
4B and SI Appendix, Fig. S9B). In quadriceps, the LPL/GPIHBP1 and LPL/CD31 ratios were 52.8% and 36.3% lower, respectively, in IBA707-treated mice (Fig.
4C and SI Appendix, Fig. S9C). We used the same methods to assess intracapillary LPL levels in IBA707- and 11A9-treated Apoa5–/–;hAPOA5-Tg mice.
The LPL/GPIHBP1 and LPL/CD31 fluorescence intensity ratios in heart capillaries were 24.1% and 27.5% lower, respectively, in IBA707-treated mice than in 11A9-treated mice (SI Appendix, Fig. S10 A and D). Similar differences in intracapillary LPL levels were observed in BAT (SI Appendix, Fig.
S10 B and E) and quadriceps (SI Appendix, Fig. S10 C and F).
Investigating the Amino Acid Sequences in APOA5 that Are Important for Interactions with IBA707 and ANGPTL3/8.
The last 40 residues of mouse APOA5 (residues D329–G368) are important for its function (12). AlphaFold3 modeling (14) predicts that 23 residues (S330–L352) within that segment form an α-helix containing multiple leucines, an isoleucine, and a tryptophan on one side of the helix (SI Appendix, Fig. S11).
Alphafold3 also predicts that the APOA5 helix binds to the ANGPTL3–ANGPTL3 groove within the N-terminal coiled-coil domain of the ANGPTL3/8’s complex (SI Appendix, Fig. S11). ANGPTL3/8’s coiled-coiled domain is recognized to be functionally important because mutations in that region have been identified in the setting of Familial Combined Hypolipidemia (28–30).
Given the importance of APOA5–ANGPTL3/8 interactions for regulating intravascular triglyceride metabolism, we suspected that APOA5’s C-terminal α-helix would be evolutionarily conserved. We therefore aligned the last 40 residues of mouse APOA5 with corresponding APOA5 sequences from 51 vertebrates (25 mammals, 19 birds, 7 reptiles) (Fig. 5).
In mammals, the highest level of conservation was within a 16-residue segment (L337 to L352 in mouse APOA5). L337, Q338, R340, L344, W345, E346, D347, and I348 were perfectly conserved in mammals (Fig. 5).
To assess the importance of APOA5’s conserved α-helix for IBA707 binding, we introduced nonconservative amino acid substitutions into mouse APOA5 residues S330–L352. WT and mutant APOA5 proteins (Table 1) were expressed in PgsA-745 CHO cells lacking hamster LPL expression, and the binding of IBA707 to the APOA5 in the medium was assessed with an ELISA. IBA707 binding to L337R and L341R mutants was markedly reduced; the binding to L344R, W345S, D347H, and L352R mutants was also reduced but to a lesser degree (Table 1).
IBA707 binding to a “quadruple mutant” with four substitutions (W345S, E346K, D347H, I348T) and a “double mutant” with W345S and D347H mutations was markedly reduced (Table 1). We also assessed IBA707 binding to WT and mutant APOA5 proteins in permeabilized cells by immunocytochemistry. IBA707 binding to L337R, L341R, L344R, W345S, and D347H mutants was markedly reduced (Fig.
6 and SI Appendix, Fig. S12), consistent with the ELISA findings.
We also assessed binding of WT and mutant APOA5 proteins to ANGPTL3/8 (Table 1). The binding of L337R, Q338D, R340D, L341R, L344R, W345S, and D347H mutants to ANGPTL3/8 was markedly reduced. That binding pattern closely resembled the IBA707 binding results, but there were several differences.
The Q338D, I348T, and A333S mutations had a greater impact on ANGPTL3/8 binding than on IBA707 binding (Table 1).
AlphaFold3 modeling predicts that APOA5 residues that interfered with ANGPTL3/8 binding (L337, Q338, R340, L344, W345, D347, I348, L352) interact with one or both ANGPTL3 protomers in ANGPTL3/8 (Table 1). The AlphaFold3 modeling also predicts that APOA5 binding to ANGPTL3/8 involves ANGPTL3 residues between L46 and K65 (Table 1). This same region of ANGPTL3 binds LPL (as judged by HDX-MS experiments) and triggers unfolding of LPL’s catalytic domain (10, 19).
Thus, APOA5 binding to ANGPTL3/8 inhibits ANGPTL3/8–LPL interactions and protects LPL from unfolding and inactivation.
We tested the ability of recombinant WT mouse APOA5 and two mutant APOA5 proteins (L337R and a L344R-W345S-D347K “triple mutant”) to suppress ANGPTL3/8 activity. WT-APOA5, but not the APOA5 mutants, blocked ANGPTL3/8-mediated suppression of LPL catalytic activity (Fig. 7A).
Also, WT-APOA5 sharply reduced plasma triglyceride levels in Apoa5–/– mice, whereas three different APOA5 mutants (APOA5Δ40, L337R, L344R-W345S-D347K) did not lower triglyceride levels (Fig. 7B).
IBA707 Binds to the APOA5 in Lower Vertebrates.
Several mouse APOA5 residues that were important for IBA707 binding (L337, L341, L344, W345, D347, and L352) are conserved in the APOA5 of chicken and Chinese softshell turtle (Fig. 5). We found, by ELISA and immunocytochemistry, that IBA707 binds chicken and turtle APOA5.
This binding was specific because replacing a conserved leucine (corresponding to L341 in mouse APOA5) with an arginine abolished IBA707 binding (SI Appendix, Fig. S13 A and B).
Discussion
With the goal of delineating physiologically important APOA5 sequences and improving our understanding of how APOA5 regulates triglyceride metabolism, we used phage panning to create a mAb against C-terminal sequences in APOA5. The antibody that we generated, IBA707, bound to a stretch of 12 conserved residues (L337 to I348 in mouse APOA5; L340 to I351 in human APOA5). The very same stretch of amino acids was crucial for APOA5 binding to ANGPTL3/8.
In biochemical studies, IBA707 blocked APOA5’s ability to suppress ANGPTL3/8-mediated inhibition of LPL catalytic activity. In cell culture studies, IBA707 blocked APOA5’s ability to suppress ANGPTL3/8-mediated detachment of LPL from the surface of cells. In Apoa5+/+ mice, an IV injection of IBA707 reduced amounts of intracapillary LPL and triggered hypertriglyceridemia, mimicking the impact of a genetic deficiency in APOA5.
Our studies with IBA707 yielded important insights into APOA5 function. The first was that APOA5 regulates plasma levels of ANGPTL3/8. When APOA5 function in Apoa5+/+ mice was blocked with IBA707, plasma ANGPTL3/8 levels increased sharply (mirroring the sharp increase in plasma triglyceride levels).
Consistent with that observation, plasma ANGPTL3/8 were higher in Apoa5–/– mice than in Apoa5+/+ mice. The elevated ANGPTL3/8 levels, both in the setting of Apoa5 deficiency and IBA707 treatment, imply that APOA5 regulates plasma triglyceride metabolism by regulating plasma levels of ANGPTL3/8 (and of course also by binding to ANGPTL3/8 and suppressing ANGPTL3/8-mediated unfolding of LPL). The second insight is that APOA5-mediated regulation of plasma triglyceride metabolism is not confined to feeding (when plasma levels of ANGPTL3/8 levels are high).
APOA5 also regulates plasma triglyceride metabolism during fasting. IBA707-mediated blockade of APOA5 function increases ANGPTL3/8 levels in fasted mice and triggers higher plasma triglyceride levels. Furthermore, the plasma ANGPTL3/8 levels are higher in fasted Apoa5–/– mice than in fasted Apoa5+/+ mice.
Thus, ANGPTL3/8 is active during fasting and is subject to inhibition by APOA5. These findings explain why plasma triglyceride levels remain elevated in Apoa5–/– mice after a prolonged fast and never approach the low levels in fasted Apoa5+/+ mice (11) and explain why a neutralizing ANGPTL3/8-specific mAb reduces fasting plasma triglyceride levels in human subjects with mixed hyperlipidemia (31).
After giving Apoa5+/+ mice an IV injection of IBA707, the triglyceride and ANGPTL3/8 levels in the plasma were elevated at 4 h but fell to baseline levels by 48 h. This time course was mirrored by the plasma levels of IBA707 (high at 4 h, extremely low at 48 h). Of note, IBA707 pharmacokinetics differed in Apoa5+/+ and Apoa5–/– mice.
The plasma IBA707 levels in Apoa5+/+ mice after 48 h were low (and often undetectable), but they remained high in Apoa5–/– mice. The virtually complete disappearance of IBA707 from the plasma of Apoa5+/+ mice within 48 h is noteworthy because IgG1 antibodies typically have a half-life of 6 to 8 d in mice (32). The rapid disappearance of IBA707 in Apoa5+/+ mice can be explained by target-mediated drug disposition (TMDD) (27), a pharmacokinetic phenomenon in which high-affinity binding of a drug (i.e., IBA707) to its target (i.e., APOA5) influences drug clearance.
In the realm of mAb therapeutics for plasma proteins, TMDD is often mediated by the physiologic clearance mechanisms for the target protein (27). The tissue responsible for APOA5 clearance and the molecular mechanism for APOA5 clearance have not been defined, but we speculate that the same mechanism is responsible for the rapid disappearance of IBA707 in Apoa5+/+ mice. We further speculate that the clearance mechanism for APOA5 is involved in removing ANGPTL3/8 from the plasma—explaining why plasma ANGPTL3/8 levels are elevated when APOA5 is absent or when APOA5 function is blocked by IBA707.
An Alphafold3 model (2, 14) predicts that amino acids 330–354 in mouse APOA5 form an α-helix with multiple hydrophobic residues on one side of the helix and predicts that this α-helix interacts with the hydrophobic ANGPTL3–ANGPTL3 groove in the coiled-coil domain of the ANGPTL3/8 complex. We suspected that the conserved sequences in APOA5’s C-terminal α-helix were important for APOA5 binding to ANGPTL3/8. Also, because IBA707 blocks APOA5 function, we imagined that the same sequences would be important for APOA5–IBA707 interactions.
Indeed, nonconservative amino acid substitutions in highly conserved APOA5 residues (L337R, L341R, L344R, W345S, D347H) blocked APOA5 interactions with both ANGPTL3/8 and IBA707. The substitutions that interfered with IBA707 binding closely resembled those that interfered with ANGPTL3/8 binding, although there were a few discrepancies (a greater impact of Q338D, I348T, and A333S substitutions on ANGPTL3/8 binding). The existence of those discrepancies was not surprising because there was no reason to suspect that the IBA707 and ANGPTL3/8 interactions with APOA5’s C-terminal α-helix would be identical at the molecular level.
Of note, the mutations that blocked IBA707 binding to APOA5 also blocked APOA5’s ability to inhibit the physiologic activity of ANGPTL3/8. WT-APOA5, but not APOA5 mutants (L337R, L344R-W345S-D347K), suppressed ANGPTL3/8-mediated inhibition of LPL catalytic activity. Also, WT-APOA5, but not APOA5 mutants (APOA5∆40, L337R, L344R-W345S-D347K) reduced plasma triglyceride levels in Apoa5–/– mice.
We emphasize that our insights into APOA5 interactions with ANGPTL3/8 and IBA707 were based on site-directed mutagenesis findings and therefore are subject to caveats. Our choice of nonconservative amino acid substitutions in APOA5 probably influenced IBA707 and ANGPTL3/8 binding results. Also, it is not clear whether our APOA5 amino acid substitutions altered IBA707 and ANGPTL3/8 binding directly or indirectly by altering the stability of APOA5’s C-terminal α-helix.
A more refined understanding of APOA5–IBA707 and APOA5–ANGPTL3/8 interactions will require hydrogen–deuterium exchange/mass spectrometry and structural biology studies.
Because IBA707 reduces intracapillary LPL levels and triggers hypertriglyceridemia, we do not foresee therapeutic possibilities for this antibody. On the other hand, our mapping of the IBA707 epitope and the ANGPTL3/8 binding site could inspire the development of therapeutic APOA5 mimetic peptides (i.e., peptides that inhibit ANGPTL3/8 activity and thereby lower plasma triglyceride levels). Of note, peptides spanning APOC2’s C-terminal α-helix have been shown to activate LPL activity and reverse the hyperlipidemia in Apoc2-deficient mice (33).
In our studies, screening a nonimmune human Fab bacteriophage library for APOA5 binders made it possible to create a mAb (IBA707) against evolutionarily conserved sequences in mouse and human APOA5. The APOA5 residues that were crucial for IBA707 binding to mouse APOA5 (L337, L341, L344, W345, D347) are conserved in mammals, birds (e.g., chicken, golden eagle, turkey), and reptiles (e.g., Chinese softshell turtle, common wall lizard, three-toed box turtle). We found that IBA707 binds chicken and turtle APOA5 and that the binding was eliminated by a leucine-to-arginine substitution corresponding to the L341R mutation in mouse APOA5.
An earlier study documented APOA5 in chicken lipoproteins and found, based on ligand blotting experiments, that chicken APOA5 interacts with an LDL receptor–related protein (LDL receptor relative 8, LR8) (34). That observation prompted speculation that chicken APOA5 influences lipid metabolism by mediating the uptake of lipoproteins by LDL receptor–related proteins. That scenario remains a possibility.
However, given that APOA5 residues governing APOA5–ANGPTL3/8 interactions in mammals are conserved in lower vertebrates, we suspect that the APOA5 in lower vertebrates might function to modulate the activity of ANGPTL proteins. ANGPTL8 and the ANGPTL3/8 complex are unique to mammals, but ANGPTL3 is expressed in lower vertebrates. It is tempting to speculate that APOA5 in lower vertebrates regulates ANGPTL3 (and thereby regulates intravascular triglyceride metabolism).
Many biochemical and in vivo experiments will be needed to address that possibility. For those studies, we suspect that IBA707 will be a valuable experimental tool.
Materials and Methods
Purified Proteins, Antibodies, and Plasmids.
A list of recombinant proteins, antibodies, and mutagenesis primers is provided in the SI Appendix, Tables S2–S4.
Phage Display Panning.
To create a mAb against the last 40 amino acids of APOA5, we screened a human Fab bacteriophage library (HuCAL PLATINUMÒ, Bio-Rad) (22, 23) for binders of full-length WT mouse APOA5 (WT-APOA5) but not APOA5∆40. The bacteriophage library was incubated with a molar excess of APOA5∆40 (12) and then incubated with biotinylated WT-APOA5 immobilized on a streptavidin plate. Phages capable of binding WT-APOA5 were eluted and amplified.
This selection process was repeated three times, allowing us to identify a high-affinity binder to WT-APOA5 that did not bind APOA5∆40. That clone was sequenced, and the sequences from the variable heavy chain and light chain regions were synthesized and cloned into murine antibody constant region expression vectors, followed by transfection into Chinese hamster ovarian cells to generate a recombinant chimeric IgG1 (IBA707).
Sandwich ELISAs to Assess IBA707 Binding to Mouse and Human APOA5.
We assessed the binding of IBA707 to WT mouse APOA5 and APOA5∆40 with a sandwich ELISA. Wells of 96-well EIA plates (Costar) were coated with a mAb against amino-terminal sequences in mouse APOA5 (IBA720, 0.5 µg/well) at 4 °C. On the next day, the wells were washed with PBS/Ca/Mg containing 0.05% Tween 20 and 0.1% bovine serum albumin (BSA) and then blocked in StartingBlock (ThermoFisher Scientific) at room temperature (RT).
4 h later, serial twofold dilutions of mouse WT-APOA5 and APOA5∆40 (0 to 32.5 ng in 100 µL StartingBlock buffer/well) were added to the wells and incubated overnight at 4 °C. After washing, HRP-labeled IBA707 or an HRP-labeled rabbit polyclonal antibody against mouse WT-APOA5 (3298) was added to wells (100 ng/well) and incubated at RT for 1.5 h. The wells were then washed, and HRP activity was assessed by adding 50 µL of 1-step Ultra TMB ELISA Substrate Solution (ThermoFisher).
The reaction was stopped with 50 µL 2 M sulfuric acid. The plate was read with a SpectraMax iD3 plate reader (Molecular Devices) at 450 nm.
A similar assay was used to assess the ability of IBA707 to bind to human WT-APOA5. Wells of 96-well EIA plates were coated with a human APOA5 mAb 2G1H11 (0.5 µg/well; Abcam ab115772). After washing, the wells were incubated with 100 µL of twofold dilutions of human WT-APOA5 (0 to 32.5 ng).
The wells were washed again. We then assessed the binding of HRP-IBA707 and HRP-3298 to human APOA5.
Binding Kinetics between IBA707 and Human and Mouse APOA5.
We determined the real-time binding kinetics between His-HSA–hAPOA5 or His-MSA–mAPOA5 in solution and surface-captured human monoclonal IBA707 with a Biacore T200TM system (Cytiva) as described (35, 36). First, we immobilized a polyclonal rabbit antibody against the human IgG Fc-gamma (Invitrogen) as the primary capture layer (30 µg/mL in 10 mM sodium acetate, pH 5.0) on a CM5 sensor chip with N-hydroxysuccinimide and N-ethyl-N-(3-(diethylamino)propyl)-carbodiimide (Cytiva). This yielded a surface density of 1,100 resonance units (RU) (36).
An injection of 1 M ethanolamine inactivated excess NHS-esters. To prepare the capture layer, we injected 50 nM of IBA707 for 350 s at 10 µl/min in the active flow cell only. Kinetic rate constants for the IBA707–APOA5 interactions were determined with single-cycle protocols in which five serial twofold dilutions of the APOA5 (2.5 to 40 nM) were injected for 220 s at a flow rate of 45 µL/min without intervening regeneration and followed by a dissociation phase of 4,500 s.
These interactions were measured in 10 mM HEPES, 150 mM NaCl, 3 mM EDTA, and 0.05% (v/v) surfactant P-20 at pH 7.4 at 20 °C. Injection of with 10 mM glycine/HCl, pH 1.7, for 2 × 20 s at the end of each single cycle regenerated the chip. Fitting of the double blank referenced data by nonlinear regression to a bimolecular interaction model, assuming pseudo-first-order reaction conditions, yielded the association (kon) and dissociation (koff) rate constants and the KD (koff/kon) of the interactions (36).
We used the evaluation software supplied with the instrument for global fitting (BiacoreT200 Evaluation 3.2).
Assessing the Ability of IBA707 to Block APOA5 Activity.
We assessed the impact of IBA707 on mouse APOA5’s capacity to suppress mouse ANGPTL3/8–mediated inhibition of mouse LPL activity as previously described (12). In brief, mouse LPL–expressing HEK293 cells were maintained in DMEM/F12 (3:1) (Invitrogen), 10% FBS (Hyclone), and 5 µg/mL blasticidin (Invitrogen) and seeded in 96-well plates (50,000 cells/well). On the next day, the medium was replaced with 80 µL of medium containing increasing concentrations of mouse ANGPTL3/8 alone or ANGPTL3/8 containing 20 nM mouse APOA5, 20 nM mouse APOA5 + 0.5 µM IBA707, or 20 nM mouse APOA5 + 0.5 µM IBA720 (preincubated for 1 h at 37 °C).
After 1 h at 37 °C, 20 µL of freshly prepared 5× substrate solution [EnzChek lipase substrate BODIPY-dabcyl-labeled triglyceride analog from Invitrogen and 0.05% Zwittergent detergent 3-(N,N-dimethyl-octadecylammonio)-propanesulfonate from Sigma] was added to each well to achieve a final substrate concentration of 1 µM. Fluorescent signal intensities were read at 1 and 30 min with a Synergy Neo2 plate reader with excitation and emission wavelengths of 485 nm and 516 nm, respectively. The reading at 30 min was corrected for the background reading at 1 min.
We modified this method to assess the effect of IBA707 on human APOA5’s ability to suppress human ANGPTL3/8–mediated inhibition of human LPL activity. In brief, human LPL–expressing HEK293 cells were treated with medium containing increasing concentrations of human ANGPTL3/8 that had been preincubated for 1 h at 37 °C with medium alone, medium containing 20 nM human APOA5, or medium containing 20 nM human APOA5 + 0.5 µM IBA707. After 1 h at 37 °C, freshly prepared substrate was added to the wells to measure human LPL catalytic activity.
We also assessed the impact of IBA707 on the ability of mouse APOA5 to suppress ANGPTL3/8-mediated detachment of LPL from the cell surface (11, 12). CHO-K1 cells (ATCC CCL-61) were maintained in F12 medium (ThermoFisher) containing 10% FBS (Gemini) and 1% L-glutamine (Gibco, ThermoFisher) and seeded at 150,000 cells/well in 24-well plates (Costar). On the next day, the cells were washed with PBS/Ca/Mg and incubated with purified human LPL (50 nM) at RT for 10 min.
After washing, cells were treated for 15 min at RT with medium alone, 100 nM mouse ANGPTL3/8, or 100 nM mouse ANGPTL3/8 containing 1 µM mouse WT-APOA5 that had been preincubated with 4 µM IBA707 or the irrelevant control mAb 11A9. After washing, the cells were cooled on ice for 15 min and blocked in 10% donkey serum in PBS/Ca/Mg at 4 °C for 1 h. Amounts of human LPL remaining on the cell surface were assessed by incubating cells with Alexa Fluor 555–labeled mAb 5D2 (5 µg/mL) in 3% donkey serum/PBS/Ca/Mg at 4 °C for 1 h.
After washing the cells three times with 3% donkey serum/PBS/Ca/Mg and twice with PBS/Ca/Mg, the cells were fixed, stained with DAPI, rinsed, and mounted (12). Images were recorded with a LSM980 microscope (Zeiss) with a 20× objective. 5D2 fluorescence intensity was assessed in 80 to 149 cells/group.
The same method was used to assess the impact of IBA707 on human APOA5’s ability to suppress ANGPTL3/8-mediated detachment of LPL. Human LPL–loaded CHO-K1 cells were incubated at RT for 15 min with medium, 100 nM mouse ANGPTL3/8, or 100 nM mouse ANGPTL3/8 containing 1 µM human WT-APOA5 that had been preincubated with either 4 µM IBA707 or mAb 11A9. Amounts of cell-surface LPL (90 to 139 cells/group) were assessed by confocal microscopy after incubating the cells with Alexa Fluor 555–labeled 5D2.
Sandwich ELISAs to Assess APOA5 Binding to IBA707 and ANGPTL3/8.
CHO PgsA-745 cells lacking hamster LPL expression (37) were maintained in F12 medium (ThermoFisher) containing 10% FBS (Gemini) and 1% L-glutamine (Gibco, ThermoFisher) and transfected with vectors for HIS-tagged WT-APOA5 (NC_000075.7) or mutant APOA5 proteins that had been modified by site-directed mutagenesis. After transfection, the cells were seeded into 6-well plates in F12 medium containing 2 U/mL heparin, 1% L-glutamine, and 0.05% protease inhibitors. After 2 d, medium was collected and stored at –80 °C.
The binding of IBA707 to WT and mutant APOA5 proteins in the cell culture was assessed with a sandwich ELISAs. One ELISA measured total amounts of APOA5 in the sample, and the other measured IBA707-bound APOA5. Wells of 96-well plates were coated with a HIS-tag mAb (AD1.1.10, 0.5 µg/well) overnight at 4 °C; washed with 5 U/mL heparin, 0.1% BSA, 1× PBS/Ca/Mg; and blocked for 4 h at RT with StartingBlock buffer (ThermoFisher).
Next, serial twofold dilutions of medium (1:2 to 1:64) were added to duplicate sets of AD1.1.10-coated wells and incubated overnight at 4 °C. After washing, amounts of APOA5 in the medium were assessed in two ways. One set of wells was incubated with 100 ng/well of HRP-labeled mAb 9H10 (which binds to APOA5 residues 75–95) to assess amounts of APOA5 in the medium; the other set was incubated with 100 ng/well of HRP-labeled IBA707.
After 1.5 h, the wells were washed; HRP activity in the wells was detected with 1-step Ultra TMB substrate (50 µL/well, ThermoFisher); and the reaction was stopped with 2 M sulfuric acid (50 µL/well). Optical density (OD) was recorded at 450 nm with a SpectraMax iD3 plate reader (Molecular Devices). IBA707 binding to mutant APOA5 proteins was calculated from the ratio of the HRP-IBA707 OD reading to the HRP-9H10 OD reading in the same sample at the same dilution.
In each experiment, we used HRP-IBA707/HRP-9H10 ratios from samples containing comparable amounts of APOA5 (as judged by the HRP-9H10 OD readings).
The binding of APOA5 (from the medium of transfected cells) to ANGPTL3/8 was assessed with two sandwich ELISAs, one to assess amounts of APOA5 in the medium and the other to assess amounts of APOA5 binding to ANGPTL3/8. 96-well plates (Costar) were coated with a HIS-tag mAb (AD1.1.10, 0.5 µg/well) or with a FLAG-tag mAb (Millipore Sigma, clone M2, 0.5 µg/well) overnight at 4 °C; washed with 1× PBS/Ca/Mg containing 5 U/mL heparin and 0.1% BSA; and then blocked for 4 h at RT with StartingBlock buffer (ThermoFisher). In the first ELISA, serial twofold dilutions of medium from transfected cells were added to wells coated with a HIS-tag mAb overnight at 4 °C.
After washing, HRP-labeled mAb 9H10 (100 ng/well) was added to the wells to assess amounts of HIS-tagged APOA5 in the medium. The OD reading from mock-transfected cells was used as the blank. In the second ELISA, the twofold dilutions of medium were preincubated with FLAG-tagged mouse ANGPTL3/8 (250 ng) at 4 °C for 20 min.
The mixture was then added to wells coated with a FLAG tag mAb (Millipore Sigma, clone M2) and incubated overnight at 4 °C. HRP-labeled mAb 1F3 (which binds to APOA5 residues 1–23) was added to the wells to assess amounts of ANGPTL3/8-bound APOA5. The binding of ANGPTL3/8 to WT-APOA5 and the mutant APOA5 proteins was calculated from the HRP-1F3/HRP-9H10 OD ratio in wells containing comparable amounts of APOA5 (as judged by the HRP-9H10 OD readings).
In this ELISA, the OD reading obtained with the medium from cells expressing APOA5Δ40, which cannot bind ANGPTL3/8 and does not lower plasma triglyceride levels in Apoa5–/– mice (12), was used for background correction.
An Immunocytochemistry Assay to Assess Binding of IBA707 to APOA5.
Transfected CHO PgsA-745 cells lacking hamster LPL expression were seeded at 200,000 cells/well in wells of 24-well plates in F12 medium (ThermoFisher) containing 10% FBS (Gemini) and 1% L-glutamine (Gibco, ThermoFisher). On the next day, the cells were washed with PBS three times; fixed with 4% PFA; and blocked with 5% donkey serum at RT. After 1 h, the cells were incubated with 1 µg/mL Alexa Fluor 488–IBA707 and 1 µg/mL Alexa Fluor 647–labeled AD1.1.10 at RT for 1 h in 5% donkey serum with 0.2% Triton X-100.
After washing, the cells were stained with DAPI and mounted. Images were recorded with an LSM980 microscope (Zeiss) with a 20× objective. The fluorescence intensities of AF488-IBA707 and Alexa Fluor-labeled AD1.1.10 were quantified in 55 to 272 cells/group.
Testing the Capacity of APOA5 to Inhibit ANGPTL3/8 Activity.
We expressed recombinant mouse APOA5 proteins as described by Chen, Yang, and coworkers (12). We tested the ability of WT-APOA5 and two mutant APOA5 proteins (L337R, L344R-W345S-D347K) to suppress the capacity of ANGPTL3/8 to inhibit the triglyceride hydrolase activity of mouse LPL (12). We also tested the ability of WT-APOA5, APOA5Δ40, L337R-APOA5, L344R-W345S-D347K-APOA5 to reduce plasma triglyceride levels in Apoa5–/– mice (n = 3 to 4 mice/group) (12).
In these studies, the mice were given an IV injection of APOA5 (300 μg per mouse) via the tail vein. Plasma triglyceride levels were measured at baseline and 5, 10, and 24 h after the APOA5 infusion.
Modeling APOA5–ANGPTL3/8 Interactions.
The stoichiometry of human ANGPTL3/8 (i.e., two ANGPTL3 molecules and 1 ANGPTL8 molecule) was documented experimentally by Kumari et al. (10) using mass photometry, and the 2:1 stoichiometry was subsequently modeled by AlphaFold3. Here, we used AlphaFold3 to replicate the ANGPTL3/8 modeling with mouse ANGPTL3 (Uniprot: Q9R182) and mouse ANGPTL8 (Uniprot: Q8R1L8).
The interaction between mouse ANGPTL3/8 complex and mouse APOA5 (Uniprot: Q8C7G5) was initially modeled by AlphaFold3 (2, 14) and then refined with ChimeraX software (38), which made it possible to remove clashes and predict contacts between the ANGPTL proteins and APOA5. Contacts were defined by interactions that did not exceed 4 Å.
Mouse Experiments.
WT and Apoa5–/– mice (FVB/NJ) have been described previously (11, 12). Mice expressing a human APOA5 transgene (hAPOA5-Tg) were purchased from MMRRC (009981-UCD). We then bred Apoa5–/–;hAPOA5-Tg mice.
The mice were maintained in a barrier facility with a 12-h light–dark cycle. Animal studies were approved by UCLA’s Animal Research Committee. Plasma triglyceride levels were quantified with a Serum Triglyceride Determination Kit (Sigma TR0100).
ANGPTL3/8 Levels in Mouse Plasma.
We measured plasma ANGPTL3/8 levels with an immunoassay (12, 24). Streptavidin plates were washed with TBST (10 mM Tris pH 7.4, 150 mM NaCl, and 0.1% Tween 20) and blocked with TBS containing 1% BSA for 1 h. After washing, the wells were incubated with a biotinylated ANGPTL3 polyclonal antibody (R&D AF136) for 1 h and then washed again.
Plasma samples (50 µL/well) were added to the wells at a 1:10 dilution in assay buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 1% Triton X-100, 5 mM EDTA, 5 mM EGTA). A standard curve was generated with recombinant mouse ANGPTL3/8. After a 2-h incubation, the wells were washed, incubated with 50 μL of a ruthenium-labeled ANGPTL3/8-specific mAb (IBA492) for 1 h, washed again, and incubated with 150 μL of an MSD read buffer.
Ruthenium electrochemiluminescence was detected with an MSD QuickPlex SQ120 reader. This assay is minimally affected by the presence of APOA5. Even when 5 µg/mL of recombinant mouse APOA5 is added to plasma samples (an amount >10 times greater than physiological levels of APOA5), the immunoassay remained effective in quantifying ANGPTL3/8 levels.
Levels of IBA707 in Mouse Plasma.
We developed an immunoassay to measure plasma levels of IBA707. An MSD plate was washed with TBST (10 mM Tris pH 7.4, 150 mM NaCl, and 0.1% Tween 20) and coated with mouse WT-APOA5 (5 µg/mL in PBS) at RT with gentle shaking. After a 1-h incubation, the plate was blocked in TBS containing 1% BSA for 1 h.
Next, 50 μL plasma samples (1:100,000 diluted in assay buffer) were added to wells. A standard curve was generated with serial dilutions of IBA707. After 2 h, the wells were washed, incubated in ruthenium-labeled goat mouse lambda antibody (50 μL) for 1 h.
The plate was washed again and incubated with 150 μL of MSD read buffer. Ruthenium electrochemiluminescence was detected with an MSD QuickPlex SQ120 reader.
Assessing Amounts of LPL Along the Luminal Surface of Capillaries.
We assessed amounts of LPL, GPIHBP1, and CD31 along the luminal surface of capillaries with a previously described method (11, 12). After an 18-h fast, Apoa5+/+, Apoa5–/–, and Apoa5–/–;hAPOA5-Tg mice were given an IV injection of IBA707, IBA720, or 11A9 and then fed a chow diet. 6 or 8 h later, the mice were given an IV injection of Alexa Fluor–labeled mAbs 27A7 (against LPL), 11A12 (against GPIHBP1), and 2H8 (against CD31).
After 10 min, the vasculature was perfused with PBS and 4% PFA, and sections of the heart, BAT, and quadriceps were examined by confocal microscopy with a LSM980 microscope (Zeiss) with a 20× objective. The fluorescent signal intensities in capillary segments (reflecting amounts of LPL, GPIHBP1, and CD31 on the luminal surface of endothelial cells) were quantified by confocal microscopy.
Statistical Analyses.
Data show mean ± SD or mean ± SEM. A four-parameter logistic nonlinear regression model was used to fit LPL activity curves (12). For ANGPTL3/8 immunoassays, MSD software was used for curve fitting (5-parameter fit with 1/y2 weighting).
A one-way ANOVA with a post hoc test or a two-way ANOVA with a post hoc test was used for multiple group comparisons. A Student’s t test was used to compare data from two groups. Statistical analyses were conducted with JMP 13 (SAS Cary, NC), Prism 9.0 (GraphPad), and Stata SE 18.1 (StataCorp, College Station, TX) software.
For all studies, a two-sided P value < 0.05 was considered significant. All data in this paper can be found in SI Appendix.
Supplementary Material