Supramolecular Assembly of High-Density Lipoprotein Mimetic Nanoparticles Using Lipid-Conjugated Core Scaffolds

Abstract

Synthetic high-density lipoprotein (HDL) mimics have emerged as promising therapeutic agents. However, approaches to date have been unable to reproduce key features of spherical HDLs, which are the most abundant human HDL species. Here, we report the synthesis and characterization of spherical HDL mimics using lipid-conjugated organic core scaffolds. The core design motif constrains and orients phospholipid geometry to facilitate the assembly of soft-core nanoparticles that are approximately 10 nm in diameter and resemble human HDLs in their size, shape, surface chemistry, composition, and protein secondary structure. These particles execute salient HDL functions, including efflux of cholesterol from macrophages, cholesterol delivery to hepatocytes, support lecithin:cholesterol acyltransferase activity, and suppress inflammation. These results represent a significant step toward a genuine functional mimic of human HDLs.

High-density lipoproteins (HDL) are dynamic nanoparticles that circulate in the blood and transport cholesterol. Clinically, elevated HDL cholesterol levels are associated with reduced risk of atherosclerotic cardiovascular disease (ASCVD).^1,2 At the cellular level, HDLs target macrophages and hepatocytes to remove and deliver cholesterol respectively [a process known as reverse cholesterol transport (RCT)], which leads to reduced inflammation and atherosclerotic burden.^3,4 Due to the beneficial association between HDL and ASCVD, the cell-specific targeting properties of HDL, and the intrigue of using HDLs for targeted drug delivery, tremendous effort has focused on synthesizing HDLs that resemble their natural counterparts.⁵

There are multiple HDL species in the blood that serve as synthetic targets. Immature HDLs are cholesterol-poor and discoidal in shape, consisting mainly of the HDL-defining protein, apolipoprotein A-I (apoA-I), and phospholipids. Mature forms of HDL, which also harbor apoA-I and a surface layer of phospholipids, are rich in cholesterol and cholesteryl esters, and are roughly spherical (Figure 1). Spherical HDLs comprise the majority of HDLs in circulation.^6,7 Yet, synthetic forms of HDL used in the clinic have exclusively resembled immature, discoidal HDL, largely because these recombinant HDLs (rHDL) are relatively easy to fabricate by self-assembling phospholipids and apoA-I.^8–12 Clinical trials using rHDLs have shown some promise, but also reveal that there is significant room for improvement.^10–12 There remains ongoing debate regarding the relative antiatherosclerotic functions of discoidal and spherical HDLs, and hence a question about which subspecies should be mimicked by bioinspired therapeutics. The prevailing view has been that discoidal HDLs are responsible for the majority of RCT activity and ASCVD risk reduction.¹³ However, larger, spherical HDL species have specifically been shown to be associated with reduced ASCVD risk.^14,15 This correlative data, in combination with recent metabolic evidence,¹⁶ has led some investigators to propose an alternative model¹⁷ that suggests spherical HDLs play a prominent role in atheroprotection. To this end, we sought to develop a spherical HDL mimic to both address the dearth of synthetic, spherical HDLs in the literature, and provide a tool for interrogating the therapeutic functions of discoidal and spherical HDLs.

Figure 1.

Canonical reverse cholesterol transport pathway. Abbreviations: FC = free cholesterol, CE = cholesteryl esters, PL = phospholipids, and LCAT = lecithin:cholesterol acyltransferase.

Spherical HDLs are challenging to synthesize due to the enzymatic steps required to mature discoidal HDLs into cholesterol- and cholesteryl ester-containing spherical HDLs. Our group, and others, have attempted to circumvent these biological maturation steps by using inorganic nanoparticles as templates.^18–20 The templates can be functionalized with phospholipids and apoA-I, and successfully restrict the size of HDL mimics to < 15 nm in diameter.²¹ These HDL-like nanoparticles have demonstrated exquisite properties for modulating cellular cholesterol,^22,23 in vivo imaging,²⁴ and drug delivery.²⁵ However, these materials fail to recapitulate a critical feature of native HDLs, namely a robust, soft material core capable of dynamically loading and off-loading cholesterol and cholesteryl esters. We note that many other HDL-like platforms have been reported, including soft-core, discoidal particles using amphiphilic peptides,²⁶ rationally engineered HDL-like nanodiscs,^27,28 and drug-loaded rHDLs.^29–31

Herein, we describe the synthesis of spherical HDL-like nanoparticles using lipid-conjugated core scaffolds. Three organic scaffolds varying in size and composition were investigated. The first scaffold is a highly hydrophobic small molecule-phospholipid conjugate (PL₄) that was synthesized using copper-free click chemistry. Specifically, a headgroup-modified phospholipid harboring a ring-strained alkyne, 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-dibenzocy-clooctyl (DBCO PL), was click-coupled to tetrakis(4-azidophenyl)methane, a small molecule with four terminal azides (SM-Az₄) (Figure 2a; SI, Figure S6) to yield PL₄. The second (9-DNA-PL₄) and third (18-DNA-PL₄) core scaffolds are amphiphilic DNA-linked PL₄ cores (Figure 2b) with different DNA linker length (9- vs 18-mer dsDNA). While DNA linkers enhance the solubility of the cores in aqueous solution, externally oriented phospholipid tails enable hydrophobic interaction with apoA-I and phospholipids. The arm lengths of these core scaffolds can be easily tuned by adjusting the length of DNA linkers. Both DNA-PL₄ cores were synthesized in a two-step sequence (SI, Section S2a). First, we synthesized DNA-phospholipid conjugates (ssDNA-PL), and tetrahedral small molecule-DNA hybrids (SMDH₄). Second, ssDNA-PL and SMDH₄ with complementary base-pairing sequences were hybridized to yield the final DNA-PL₄ cores (Figure 2b; SI, Figures S1–4, Table S1).

Figure 2.

(a–c) Synthesis schemes for PL₄ and DNA-PL₄ core scaffolds, and LC HDL NP assembly. (d) Schematic of native and synthetic HDLs. (e) Composition of native and synthetic HDLs by mass %. *Au HDL NP composition is reported from a previously published data set.²² (f) Circular dichroism spectra for LC HDL NPs and human HDLs. (g) TEM imaging of DNA-PL₄ HDL NPs. Scale bar = 50 nm.

Using the resulting core scaffolds, we carried out the assembly of lipid conjugate HDL-like nanoparticles (LC HDL NPs). We first prepared liposomes from 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) (SI, Section 2c). PL₄ scaffolds were then prepared as thin films, while DNA-PL₄ scaffolds were prepared in aqueous buffer. ApoA-I protein and DPPC liposomes were added to the core scaffolds and the mixture was subjected to three rounds of sonication, allowed to relax on ice, then filtered and concentrated using a 50 kDa MWCO spin column (Figure 2c). Schematic models of LC HDL NPs, gold nanoparticle-templated HDL NPs (Au HDL NPs), and human HDLs are shown in Figure 2d.

Transmission electron microscopy (TEM) imaging revealed a relatively monodisperse population of nanoparticles for the assemblies using PL₄ and 9-DNA-PL₄ scaffolds (TEM diameter = 10 ± 2 nm and 9 ± 2 nm, respectively) (Figure 2g; SI, Figure S8). Notably, the morphology of these particles was not characteristic of immature, discoidal HDLs, or rHDLs, both of which exhibit a hallmark rouleaux formation of stacked phospholipid discs.^21,26,32 Instead, the particles appeared roughly spherical, with a peripheral hypodense region. Dynamic light scattering experiments revealed hydrodynamic diameters resembling those of human HDLs (D_H = 13.8 ± 3.9 and 13.3 ± 4.6 nm for PL₄ HDL NPs and 9-DNA-PL₄ HDL NPs, respectively) (SI, Table S2). Size-exclusion chromatography (SEC) demonstrated peak retention volumes of 8.22 and 8.58–8.88 mL for 9-DNA-PL₄ HDL NPs and PL₄ HDL NPs, respectively (SI, Figure S14; see also Section S11f2 for additional discussions). SEC traces of LC HDL NPs also revealed right-sided tails consistent with a small amount of free apoA-I (10.13 mL retention volume), which was also observed in human HDL controls (see SI, Section 11d for discussion).

While assemblies using PL₄ and 9-DNA-PL₄ scaffolds successfully generated sub-20 nm particles, the assemblies using 18-DNA-PL₄ scaffolds persisted as large vesicular structures after the addition of DPPC and apoA-I (SI, Figure S9). We hypothesize that this is a result of the hybrid’s size being prohibitively large for apoA-I assembly. Interestingly, 9-DNA-PL₄ HDL NPs were not significantly larger than PL₄ HDL NPs. This result, in conjunction with the lack of assembly using 18-DNA-PL₄, suggests that LC HDL NP size may be determined primarily by apoA-I protein folding, irrespective of core size below a certain threshold. For convenience henceforth, we refer to the particles formed using 9-DNA-PL₄ scaffolds simply as DNA-PL₄ HDL NPs.

To characterize particle structure in terms of apoA-I protein conformation, circular dichroism measurements were carried out on LC HDL NPs and controls, and analyzed using three distinct reference protein databases. PL₄ HDL NPs and DNA-PL₄ HDL NPs were found to closely resemble human HDLs in secondary structure, while Au HDL NPs exhibited excess α helicity and less β sheet and turn content (Table 1 and Figure 2f). To characterize LC HDL NP structure further, we investigated the apoA-I oligomerization state by crosslinking LC HDL NPs with bis(sulfosuccinimidyl)suberate followed by immunoblot.²² ApoA-I interaction with LC cores was found to induce stable oligomer formation (SI, Figure S13). Notably, the surface zeta potentials of LC HDL NPs were more negative (ζ = −21 ± 2 mV for PL₄ and −19 ± 1 mV for DNA-PL₄) than those for apoA-I alone (ζ = −10 ± 3 mV) or an [apoA-I + DPPC] mixture (ζ = −12 ± 2 mV), but quite similar to human HDL₂ (ζ = −19 ± 2 mV) and HDL₃ (ζ = −21 ± 5 mV). Finally, we found that the composition of LC HDL NPs closely mirrors that of HDL₂ and HDL₃, with the exception that LC HDL NPs do not contain cholesterol or cholesteryl esters (Figure 2e; see SI, Section S11e,f for molar composition data (Table S4) and additional discussions).

Table 1.

Structural Analysis of Native and Synthetic HDLs via Circular Dichroism

	α-helix (%)	β-sheet (%)	Turn (%)	Unordered (%)
apoA-I	57 ± 3	4 ± 2	14 ± 3	25 ± 5
Au HDL NP	73 ± 1	0 ± 0	6 ± 1	21 ± 2
HDL2	50 ± 2	7 ± 1	18 ± 4	24 ± 2
HDL3	42 ± 1	15 ± 1	15 ± 1	28 ± 3
PL4 HDL NP	50 ± 1	7 ± 1	16 ± 5	27 ± 1
DNA-PL4 HDL NP	40 ± 1	20 ± 2	16 ± 2	25 ± 7

Complementary biological assays were carried out to investigate whether LC HDL NPs could recapitulate salient HDL functions. A critical property for HDL mimicry is the ability to efflux cholesterol from lipid-laden macrophages, which we evaluated using an in vitro radiolabeled cholesterol efflux assay (SI, Section S7a).³³ Briefly, J774 macrophages were loaded with tritium-labeled cholesterol ([³H]-chol), cultured with cAMP to upregulate cholesterol efflux receptors, and treated with nanoparticles or controls for 4 h. The media supernatant was subjected to liquid scintillation counting to quantify percent effluxed [³H]-chol, revealing that both PL₄ HDL NPs and DNA-PL₄ HDL NPs facilitated robust cholesterol efflux (PL₄: 6.2 ± 0.7%, DNA-PL₄: 6.5 ± 0.2%; 100 nM protein) in a dose-dependent manner (Figure 3a).

Figure 3.

(a) LC HDL NPs efflux [³H]-chol from J774 macrophages. (b) % Influx of [³H]-chol to HepG2 hepatocytes in a tandem efflux-influx assay. (c) Quantification of [³H]-chol in all fractions from tandem assay. Two-tailed Student’s t test: **p < 0.01. *** p < 0.001.

Encouraged by the aforementioned results, we then designed an experiment to simulate the entire RCT process in a single assay. We accomplished this by carrying out the standard radiolabel efflux assay followed by an influx step where conditioned efflux media was introduced to cultured hepatocytes (HepG2) and influx was allowed to proceed for an additional 4 h. Liquid scintillation counting analyses of all three fractions from macrophages, hepatocytes, and media supernatant revealed that LC HDL NPs exhibited superior cholesterol transport capacity compared to human HDLs, apoA-I, and Au HDL NP (Figure 3b,c). This assay demonstrates not only that LC HDL NPs are capable of robust cholesterol efflux and delivery, but also that a single-particle cohort can execute these functions sequentially by on- and off-loading cholesterol in a dynamic fashion.

Next, we investigated the efficacy of LC HDL NPs as cholesterol delivery agents independently of efflux, a function carried out in the native setting by mature HDLs. HepG2 cells were cotreated with fluorescent cholesterol (NBD-chol) and LC HDL NPs or controls, and subsequently processed for flow cytometry or confocal microscopy. Strikingly, LC HDL NPs facilitated efficient delivery of NBD-chol in only 30 min (Figure 4), with 23% of PL₄ HDL NP-treated cells being NBD-positive compared to 8.5%, 6.8%, and 5.7% for HDL₂, apoA-I, and Au HDL NP, respectively.

Figure 4.

LC HDL NPs facilitate delivery of NBD-chol to hepatocytes in 30 min via (a,b) flow cytometry and (c) confocal microscopy. Two-tailed Student’s t test: *p < 0.05. ***p < 0.001.

Native HDLs also support the enzymatic esterification of free cholesterol, a reaction carried out in the blood by the enzyme lecithin:cholesterol acyltransferase (LCAT). We found that both PL₄ and DNA-PL₄ HDL NPs supported LCAT-mediated esterification (Figure 5a). However, PL₄ HDL NPs supported substantially more esterification than DNA-PL₄ HDL NPs and Au HDL NPs, converting 87% of bound cholesterol to cholesteryl esters. This is likely due to the increased hydrophobicity of the PL₄ core, which enables efficient sequestration of cholesteryl esters.

Figure 5.

(a) LC HDL NPs support LCAT-mediated cholesterol esterification. (b) LC HDL NPs suppress pro-inflammatory NF-κB signaling. Two-tailed Student’s t test: *p < 0.05. ** p < 0.01.

Finally, inflammation promoted by macrophage NF-κB activity is a hallmark of ASCVD that drives disease progression and increases morbidity and mortality. To this end, we treated human monocyte NF-κB reporter cells (THP1-Dual) with lipopolysaccharide to stimulate NF-κB activity prior to treatment with LC HDL NPs or controls. While apoA-I alone exhibited no capacity to reduce NF-κB activity under these low-dosing conditions, both PL₄ HDL NPs and DNA-PL₄ HDL NPs favorably reduced NF-κB activity in a dose dependent manner, by 31% and 16% respectively at concentrations of 150 nM (Figure 5b).

In summary, we report the synthesis of soft-core, HDL-like nanoparticles with the size, shape, surface chemistry, composition, protein structure, and cholesterol transport properties that are consistent with mature human HDLs. LC HDL NPs target relevant cell types to remove and deliver cellular cholesterol and reduce inflammation, closely mimicking the salient functions of human HDLs. Because synthesis of LC HDL NPs is straightforward and does not require time-intensive and costly enzymatic maturation steps, we anticipate these nanoparticles will be strong candidates for the next generation of HDL-based therapeutics.