Multivalent targeting of the asialoglycoprotein receptor by virus-like particles

Abstract

The asialoglycoprotein receptor (ASGPR) is expressed in high density on hepatocytes. Multivalent variants of galactosyl carbohydrates bind ASGPR with high affinity, enabling hepatic delivery of ligand-bound cargo. We previously demonstrated ASGPR-mediated small-molecule delivery using di- or tri-valent variants of a galactose-mimetic ligand. Virus-like particle (VLP) conjugates of this ligand were efficiently endocytosed by ASGPR-expressing cells in a manner strongly dependent on the nature and density of ligand display, with the best formulation using a nanomolar-, but not a picomolar-level, binder. Optimized particles were taken up by HepG2 cells with greater efficiency than competing small molecules or the natural multi-galactosylated ligand, asialoorosomucoid. Upon systemic injection in mice, these VLPs were rapidly cleared to the liver and were found in association with sinusoidal endothelial cells, Kupffer cells, hepatocytes, dendritic cells, and other immune cells. Both ASGPR-targeted and non-targeted particles distributed similarly to endothelial and Kupffer cells, but targeted particles distributed to a greater number and fraction of hepatocytes. Thus, selective cellular trafficking in the liver is difficult to achieve: even with the most potent ASGPR targeting available, barrier cells take up much of the injected particles and hepatocytes are accessed only approximately twice as efficiently in the best case.

Graphical Abstract

Delivery of drugs to particular cell types in the liver is a useful goal. For hepatocytes, this is typically done using molecules that bind to the asialoglycoprotein receptor. This paper describes the construction of protein nanoparticles with ultra-high affinity for cells bearing this receptor, but also that uptake by other cells in the liver is difficult to overcome if hepatocyte-specific delivery is desired.

INTRODUCTION

The asialoglycoprotein receptor (ASGPR) is a hepatocellular C-type lectin membrane protein that exhibits high ligand specificity, transport rate, and expression level – ideal properties for targeted drug delivery.^1,2 ASGPR recognizes and internalizes structures displaying galactose (Gal), N-acetylgalactosamine (GalNAc), and related glycomimetic ligands, and is especially sensitive to multivalent variants of such structures.^3-5 Small molecules,⁶ nucleic acids,^7,8 CRISPR-Cas9,^9-11 and nanoparticles of various classes (lipids,^12-14 inorganic,^15-18 polymer^19-21) have been successfully delivered using multivalent carbohydrate ligands targeting ASGPR. To our knowledge, only a few reports have appeared of the enhancement of protein nanoparticle uptake by surface functionalization targeting ASGPR. Earlier examples illustrated in vitro uptake by ASGPR-expressing cells of particles bearing high densities of lactose or lactobionic acid ligands.^{22, 23} A recent and more application-relevant example described enhanced hepatocyte transduction by chemically-derivatized adeno-associated virus.²⁴ In all cases, monovalent, weakly-binding ligands (lactose and GalNAc) were employed.

Qβ virus-like particles (VLPs) are composed of 180 copies of a self-assembling coat protein monomer; their tolerance for chemical modification and genetic point mutations makes them excellent scaffolds for chemical biology.^25-27 VLPs can be addressed with synthetic glycans to generate polyvalent ligand-displaying structures with high affinities for targeted cellular receptors.^26-28 Intravenously administered nanoparticles naturally traffic to the liver, but are mainly taken up by liver-resident macrophages.^29-33 We previously characterized a series of multivalent ASGPR ligands composed of a bicyclic bridged ketal (BC) variant of GalNAc and demonstrated efficient hepatocellular targeting of small-molecule cargo conjugated to derivatives of the ligands.^6,34 Here we explore the ability of the BC ligand and a related monocyclic trifluoracetamide motif (designated MC) reported earlier³⁴ to promote ASGPR-mediated internalization of macromolecular VLP cargo. An exploration of the effects of ligand valency showed the ability to significantly outcompete a natural multivalent protein ligand for receptor-mediated endosomal uptake. However, these particles were found to access different liver cell types in mouse after intravenous injection, captured by sinusoidal endothelial cells and distributed to Kupffer, hepatocyte, and other immune cell populations.

RESULTS

Preparation and characterization of ASGPR-targeted VLPs

Qβ VLPs were prepared as previously described, using the K46Q capsid protein mutant shown to exhibit significantly decreased nonspecific binding to cellular surfaces.³⁵ Ligands for the asialoglycoprotein receptor were attached to these particles by a two-stage process (Figure 1a) beginning with reaction with a small amount of an AlexaFluor NHS ester, introducing approximately 30 dye molecules per particle to facilitate VLP detection. An excess of azide-terminated NHS ester was then used to acylate most of the remaining amines and introduce a convenient handle for copper-mediated azide-alkyne cycloaddition (CuAAC). Following purification from excess reagent, alkyne-terminated mono-, di-, or trivalent ASGPR ligands (Figure 1b, BC1, BC2, BC3, MC2) were coupled to the VLPs, varying the stoichiometry of alkyne reagents to control the density of the displayed structures. Finally, CuAAC reaction with excess propargyl alcohol was performed to cap the remaining unreacted azides. The average number of displayed ligands per particle was determined by high resolution ESI-TOF mass spectrometry (Figure 1c). The conjugates were also characterized by analytical size-exclusion chromatography and dynamic light scattering (DLS). The former showed very similar elution profiles for particles before and after ASGPR ligand attachment (Figure 1d), but the latter occasionally gave evidence of a broadened size distribution (shifted to larger diameter) for ligand-addressed particles (Figure 1e, Table 1). DLS is very sensitive to aggregation and to changes in solvent interactions, both of which may play a role in the observed changes.

Figure 1.

Bioconjugation of ASGPR ligands to Qβ nanoparticles. (a) Synthesis of ASGPR-targeted Qβ conjugates. “CuAAC” = CuSO₄, THPTA, sodium ascorbate, aminoguanidine, 0.1 M potassium phosphate buffer, pH 7.4. Particles were purified by desalting and centrifugal ultrafiltration after the second and fourth steps. (b) Structures of the ASGPR-targeting bicyclic ligand-alkynes. Because the MC1 structure was prepared as an azide derivative, it was installed on VLP-alkynes in analogous fashion to panel (a); details in Supporting Information. (c) Representative high-resolution mass spectra of denatured final products, showing the relative amounts of functionalized subunits, thereby allowing for a quantiattive estimate of the average ligand loading per particle. For simplicity, the masses of Qβ-BC2-low or Qβ-BC2-high are shown prior to treatment with excess propargyl alcohol (see SI for MC of final conjugates including PraOH groups). (d) Representative size-exclusion FPLC and (e) dynamic light scattering data.

Table 1.

ASGPR-targeted Qβ conjugates, all particles bearing an average of 30 AlexaFluor647 molecules.

VLP	Ligand	#Lig a	#GalNAc b	A/Z c	r (nm) d
K46Q	--	--	--	--	16.1
Qβ-BC1-high	BC1	400	400	7.5	17.7
Qβ-BC2-low	BC2	20	40	0.8-1.25	17.9
Qβ-BC2-high	BC2	200	400	6.0-7.5	18.7
Qβ-BC3-high	BC3	130	400	4.0	19.7
Qβ-MC1-med	MC1	80	80	2.0	17.6
Qβ-MC1-high	MC1	400	400	6.0	>300 e
Qβ-MC2-low	MC2	30	60	1.25	19.0
Qβ-MC2-med	MC2	80	160	4.5-5	25.3
Qβ-MC2-high	MC2	>120	>240	7.8	n.d. e
Qβ-capped	Pra-OH	500	0	2.5	18.2

^(a)Average number of triazole-linked attachments of the indicated ligand per particle.

^(b)Average number of GalNAc-mimetic ligands attached per particle.

^(c)Alkyne:azide molar ratio used in CuAAC reactions for each conjugate.

^(d)Observed hydrodynamic radius by dynamic light scattering; MC1-high and MC2-high VLPs were prone to immediate aggregation.

^(e)Irreversible particle aggregation rendered these samples unusable.

Uptake of ASGPR-targeted VLPs by HepG2 cells

Particle internalization via ligand mediated uptake was measured in hepatocyte cell lines expressing either low (SK-Hep1) or high (HepG2) levels of ASGPR. HepG2 is estimated to have approximately 76,000 asialoglycoprotein receptors per cell,³⁶ fewer than primary hepatocytes but significantly more than Sk-Hep1 cells.⁶ Particles displaying the BC ligand interacted with these cells to a greater degree than control particles, as determined by flow cytometry (Figure 2a). Fluorescence microscopy revealed two general patterns of particle association with cells, illustrated in Figure 2b: (i) substantial and dominant punctate staining, characteristic of robust endosomal uptake (exhibited by the natural asialoorosomucoid (ASOR) ligand as well as Qβ-BC1-high, Qβ-BC2-high, and Qβ-BC3-high, the VLPs densely decorated with the BC ligand); and (ii) both diffuse and punctate staining, indicating both binding to the cell membrane and endosomal localization, exhibited by particles with less dense display of ASGPR ligands (Qβ-BC2-low). The control particle, subjected to the same bioconjugation protocol but displaying a high density of simple alcohol groups (Qβ-capped), bound to a significantly lesser degree to HepG2 (Figure 2b) or SK-Hep cells (Figure S8), and was not appreciably internalized by either cell-type.

Figure 2.

(a) Analysis of binding of HepG2 cells by the indicated particles (20 nM VLP, 1 h, 37 °C), reported as mean fluorescence intensity determined for 5000 cells per condition by flow cytometry, normalized to the value of the best performing particle (Qβ-BC2-high, set to 100%). Since all particles bore the same number of dye molecules, this represents an assessment of the overall amount of binding and internalization (taken together) of each particle. Error bars represent the standard deviation for two independent experiments); one-way ANOVA with Dunnett’s multiple comparisons test was used for data analysis (p ≤ 0.01 = **; p ≤ 0.05 = *) . (b) Representative fluorescence microscopy images 1 hour after incubation of HepG2 cells with fluorescent ASGPR-targeted or non-targeted VLPs (50 nM VLPs or 100 nM ASOR). Scale bars = 11 μm. (c) Cell viability assessed by MTS assay after incubation with the indicated particles for 3 h.

We also prepared and tested mono- and divalent versions of a GalNAc trifluoroacetamide derivative (denoted here as the monocyclic, or MC, structure) closely related to compounds previously found to be relatively potent (low-micromolar) binders of ASGPR (Scheme S1-S2, Figure S9).³⁴ Due to its more hydrophobic nature, particles bearing dense arrays of the MC ligand (Qβ-MC1-high, with 400 MC1 structures per particle, or Qβ-MC2-high, with more than approximately 100 MC2 structures per particle) suffered irreversible aggregation and were therefore unusable. Less dense loadings were tolerated, and the resulting particles engaged HepG2 cells in a manner showing similar, but less potent, ASGPR-mediated function compared to the BC ligand. Thus, (Qβ-MC1-med, “med” = medium loading, 80 ligands per particle) was not internalized in HepG2 cells (Figure S10), while Qβ-MC2-med was taken up significantly better than Qβ-MC2-low (Figure 2b, Figure S10). However, both of these particles showed cell-surface adhesion as well as punctate endosomal uptake, similar to Qβ-BC2-low particles. Thus, the BC ligand conjugates outperform the MC conjugates for ASGPR targeting, in spite of the fact that a close analogue of MC1 binds more tightly to the carbohydrate-binding domain of ASGPR than BC (1/K_ads values of approximately 1 μM²⁸ vs. 6 μM,⁶ respectively). We suggest that this functional difference derives from a greater degree of nonspecific membrane association by particles bearing the more hydrophobic MC ligand.

Treatment of HepG2 with either the ASGPR-targeting VLPs or ASOR was found to have no discernable impact on cell viability (Figure 2c). A shorter 15-minute incubation of HepG2 with BC-labeled particles showed significantly less particle internalization, and a greater relative contribution from membrane-bound particles across all treatment groups (Figure S11). Total VLP uptake by ASGPR-deficient SKHep1 cells was much lower than HepG2, as expected (Figure S9). Pretreatment of HepG2 cells with inhibitors of calcium-dependent ASGRP-ligand binding (EDTA) or clathrin-mediated endocytosis (chlorpromazine) each abrogated particle uptake and only allowed for extracellular binding to hepatocytes (Figure S12). It should be noted that both chlorpromazine and EDTA treatment resulted in changes in cell morphology, as have been previously described for HepaRG and other cells.^{37, 38} Particle internalization in HepG2 was further confirmed by spinning disk confocal microscopy (Supplementary Information), although ASGPR-targeted particles were not found to predominantly colocalize with either EEA1 (a marker of early endosomes) or LAMP1 (a lysosomal marker) (Figure S13). Taken together, these data are consistent with the expected ASGPR-driven endosomal mechanism of particle uptake.

Competitive binding

To further assess the role of polyvalency in uptake efficiency, cultured HepG2 cells were simultaneously incubated with different concentrations of the free dimeric BC2 ligand and the polyvalent ASGPR conjugates. Under these conditions, uptake of ASOR, which bears multiple galactosamines and is a highly efficient substrate for ASGPR endocytosis (K_i = 1.7 nM; 20 Gal/protein),³⁹ was inhibited by approximately 50% in the presence of 100 equivalents of BC2 and completely inhibited in the presence of 400 equivalents (Figure 3f). ASOR was more resistant to competition against MC2 ligand, requiring approximately 1000 equivalents of this molecule to inhibit uptake by approximately 75% (Figure S15). Only a modest excess of the free ligand, significantly less than for ASOR, was required to completely abrogate the binding of the relatively low-valent Qβ-BC2-low (Figure 3d). While both Qβ-BC2-low and ASOR bear approximately 20 Gal-mimetic units each, Qβ VLPs are significantly larger, and so the effective density of displayed ligand on the particles is less. Virus-like particles with a high density of the monovalent ligand were modestly more resistant to small molecule competition than ASOR, requiring 1000 molar equivalents of the BC2 ligand to significantly inhibit uptake (to approximately 25%, Figure 3c).

Figure 3.

Representative fluorescence microscopy images illustrating competitive inhibition of the HepG2 uptake of dye-labeled (a) Qβ-BC3-high, (b) Qβ-BC2-high, (c) Qβ-BC1-high, (d) Qβ-BC2-low, (e) Qβ-capped or (f) ASOR using increasing concentrations of BC2 ligand. Values are the molar ratio of the small molecule ligand vs. protein (particles or ASOR). HepG2 cells were simultaneously incubated with the BC2 ligand and with dye-labeled VLP conjugate (a = Qβ-BC3-high, b = Qβ-BC2-high, c = Qβ-BC1-high; d = Qβ-BC2-low, e = unmodified Qβ) or f = ASOR for 1 hour at 37 °C in a humidified incubator. Images in each horizontal panel are contrast-matched to each other, as are rows a, b, and c. The contrast in other rows was adjusted independently. Scale bars = 20 μm.

In contrast, the highly loaded multivalent ligand-decorated virus-like particles [Qβ-BC2-high, Qβ-BC3-high] were remarkably resistant to small-molecule competition, showing substantial punctate endosomal uptake in the presence of 1000 molar equivalents of the BC2 ligand relative to the VLP capsid, and requiring 10,000 equivalents of BC2 to completely (or nearly completely) block particle uptake (Figure 3a,b). ASOR itself proved to be a poor inhibitor of both cell-surface binding and endosomal internalization by lightly- or heavily-loaded Qβ conjugates: up to a 100-fold molar excess of ASOR relative to VLP had little effect (Figure 4c,d). The fluorescence intensity of HepG2 cells treated with particles and competing ligand in each experiment was manually quantified and plotted against the concentration of competing ligand to estimate IC₅₀ values for the inhibition of particle uptake by BC2 ligand (Figure 4a,b). The results (Qβ-BC2-high > Qβ-BC3-high ~ Qβ-BC1-high > Qβ-BC2-low) mirror the relative particle uptake as measured by flow cytometry under very similar conditions, but in the absence of competing ligand (Figure 2a).

Figure 4.

(a) Relative protein binding and uptake from Figure 3 (measured as the normalized fluorescence intensity per cell) plotted as a function of the molar excess in competing ligand. Error bars represent the standard error of the mean for each sample, for two to four independent measurements. (b) IC₅₀ values based on the curves in panel (a), fit in GraphPad Prism, using a one-phase decay model. (c) Representative fluorescence microscopy images of HepG2 cells simultaneously incubated with ASOR and dye-labeled (top) Qβ-BC2-high or (bottom) Qβ-BC2-low. Scale bars = 20 μm. (d) Relative protein binding and uptake from panels (c) plotted as a function of the molar ratio of excess ASOR to VLP, from two independent experiments

While direct comparisons are difficult to make since different competitive ligands and nanoparticle concentrations have been used by others (Table S1), VLP display of these ligands appears to be much more resistant to competitive binding than other multivalent platforms displaying Gal or GalNAc moieties.

Liver biodistribution of ASGPR-targeted VLPs

To determine the relative liver targeting and clearance of these particles, we intravenously administered dye-labeled VLPs to outbred CD1 mice. Collection of plasma at several time points established that two particles bearing the BC2 ligand at different densities were cleared from circulation faster than capped particles, although clearance was rapid in all cases (within approximately 30 minutes, Figure 5a). All three particles were very similar in the degree of chemical functionalization and surface charge, indicating that the differences are due to the loading of ASGPR-binding ligands on the particle surface and not simply amine acylation or the presentation of triazole groups. Liver clearance of protein has previously been reported to be enhanced by ASGPR targeting.⁴⁰

Figure 5.

In vivo plasma clearance, biodistribution, and liver subcellular targeting of ASGPR-targeted VLPs. (a) Plasma clearance of 30 μg AF647-Tag-labeled VLPs from CD1 mice. Data points from individual mice are plotted (n=2). Liver-associated fluorescence from (b) fluorescent intensity from IVIS imaging of the liver (n=4) or (c) FNIR fluorescence intensity from extracted liver at the indicated time points. Half-lives calculated from the curves in panel (b) in GraphPad Prism, using a one-phase decay model (error bars = standard deviation of the mean, n = 4). (d-f) Flow cytometry analysis of liver cells harvested after 1 h (75 μg AF647-labeled VLP, C57BL/6J mice, n = 5 mice per group). (d) Normalized mean fluorescence intensities of dye-positive cells. (e) Number of cells showing significant uptake of dye-labeled VLPs. (f) Percent of dye-VLP-positive cells of each indicated type. (g) Ratio from data in panel f. (h) Targeting efficiency of each particle, calculated from data in panel e as the ratio of hepatocytes to all cells positive for particle binding. Panels (d-g): Statistics by two-way ANOVA with Šidák’s multiple comparisons test. **** = p ≤ 0.0001; *** = p ≤ 0.001; ** = p ≤ 0.01; * = p ≤ 0.05; ns. = not significant. Panel (h): statistics by un-paired t-test; * = p ≤ 0.05.

Whole-animal fluorescence imaging showed that ligand-targeted particles accumulated in the liver to a greater extent than control particles (Figure 5b) and were cleared from the liver more rapidly (liver residence half-life of ~13.4 or 14.0 hours for Qβ-BC2 particles vs. ~30 hours for Qβ-capped); all particles were mostly eliminated from the liver within 96 hours (Figure 5c).

Distribution in the liver was assessed by intravenous administration of 75 μg of Qβ-BC2-high VLPs or capped VLPs to C57BL/6J mice, extraction of liver tissue at certain time points up to 6 h, and flow cytometry analysis using the following markers: endothelial cells (CD45− CD31+), Kupffer cells (CD45+ CD68+), dendritic cells (CD45+ CD31− CD11c+), other immune cells (CD45+ CD31− CD68−), and hepatocytes (CD45− CD31− ASGPR+). The results are summarized in Figure 5d-f, and a representative gating strategy is included in Figure S19.

Similar numbers of endothelial cells and Kupffer cells appeared to bind or take up either the targeted (Qβ-BC2-high) or non-targeted (capped) VLPs (Figure 5e), and nearly all endothelial cells were associated with enough particle to be detected (Figure 5f). Endothelial cells were associated with the greatest amounts of particle per cell (Figure 5d), but functionalized particles distributed to approximately 1.9 times as many hepatocytes as endothelial cells; and functionalized particles additionally distributed to 2.3 times as many hepatocytes as capped particles (Figure 5e). Capped VLPs exhibited equal or greater degrees of association with endothelial cells compared to ASGPR-targeted particles, whereas inclusion of the ASGPR-targeting ligand significantly enhanced (by a factor of up to 2.5) the fraction of dendritic cells, other immune cells, and hepatocytes reached by the particles at an early time point (1 h, Figure 5f,g.). Ligand-targeted particles were taken up by hepatocytes approximately 1.6 times more efficiently than were capped particles (Figure 5h). At 6 h (Figure S21), the extent of endothelial cell uptake relative to other cells in the liver was decreased, although particles were still predominantly associated with the liver endothelium. In addition, the preference of BC2-bearing VLPs for dendritic cells, other immune cells, and hepatocytes largely disappeared at 6 h.

No histological evidence of active inflammation or any significant histopathologic abnormalities were observed in association with VLP administration. Mouse weights remained stable for at least three days following intravenous administration of VLPs (Figure S19) and no cellular abnormalities were observed by histology (Figure S18), indicating that particles did not cause acute toxicity. The dominance of liver sinusoidal endothelial cell (LSEC) uptake was confirmed by histology (30 μg VLPs, BALB/c mice, cryosectioning of liver taken at 2 and 4 h time points, Figure S16-S17). Particle accumulation in LSECs rather than hepatocytes was evidenced by the distribution at the periphery of the vessels and pattern of discrete fluorescence at the periphery of the vascular channels.

DISCUSSION

Hepatocytes perform many important tasks in the body and are therefore key targets for the treatment of disease. An exciting area in this regard is gene therapy, often using viral vectors such as adeno-associated virus.⁴¹ Capsid engineering,⁴² in vivo selection,^{43, 44} and chemical modification strategies²⁴ can be used to actively target hepatocytes to achieve specificity.⁴⁵ Gene editing through viral vectors can also be made specific through the use of promoters that limit gene expression to hepatic cells.^{41, 46, 47}

The asialoglycoprotein receptor provides an obvious high-capacity point of entry to hepatocytes if endosomal delivery (or delivery and subsequent escape) is desired. As models of modified viral capsids such as AAV, we designed a series of virus-like particle conjugates bearing GalNAc variants of the highest affinity for ASGPR and assessed the ability of these structures to be taken up by cultured HepG2 cells, which share with hepatocytes a high density of ASGPRs. Uptake and accumulation of these particles was observed in an ASGPR-dependent manner, with interesting trends regarding ligand number and density.

The ASGPR has long been known to be responsive to multivalent ligands, and nanoparticles targeting ASGPR take advantage of this property.^{6, 48, 49} Excessive ligand density on nanoparticles, however, needs to be carefully balanced to avoid off-targeting binding.⁵⁰ A conceptually simple way to do this is to use a smaller number of higher affinity ligands⁵¹ or a greater number of lower affinity ligands to balance overall platform avidity and optimize binding to cells expressing a targeted amount of surface receptors.⁵² Ligand spacing will also affect apparent affinity (avidity), and so these factors are often difficult to disentangle; one recent report highlights that, in a protein nanoparticle system with the same number of targeting ligands, a clustered rather than even presentation of ligands is more effective at inducing scaffold uptake in vitro or in vivo.⁵³

For small molecule conjugates and glycopolymers, the optimal ligand spacing has been well-studied: estimates of 2-3 nm spacing between ligands allow multivalent molecules to effectively bind multiple receptors for internalization.^{2, 54} These same parameters are often ignored in nanoparticle systems, with only a limited number of reports attempting to understand how ligand valency and density impacts nanoparticle uptake by ASGPR (Table S2). Overall, higher-density ligand display is generally found to promote endosomal entrapment of viruses, but this effect tends to reach a limit past which no further improvement in uptake is observed. In addition, biological performance varies with the nature of the polyvalent display platform. Here we add observations of structure-activity relationships concerning cell-surface binding vs. uptake in vitro and liver trafficking in vivo.

We found that surface binding and endocytosis of the Qβ VLP was each achieved by different formulations. Endocytosis required a high density of ASGPR ligands, presented as monomers, dimers, or trimers. Noting that such dimers and trimers engage the ASGPR with nanomolar or better association (compared to micromolar binding of monovalent ligands),⁶ it was interesting that scattered presentation of such high-affinity ligands (as in Qβ-BC2-low) promoted cell-surface binding but not uptake (Figure 2), suggesting that properties of both binding and spacing may be important. These apparently have functional consequences as well.

For example, high-density arrays of BC1 and BC3 ligands on VLPs provided somewhat inferior particle uptake compared to the divalent (BC2) analogue, all three particles carrying approximately the same number of BC units (Figure 2a, Table 1). This may reflect an interesting balance between ASGPR engagement and release, enabled by differences in the binding power of individual ligands (micromolar, nanomolar, and picomolar, respectively, for BC1, BC2, and BC3).⁶ Thus, the Qβ-BC3-high particle should be slow to detach from the endosomal membrane or may inhibit the recycling of the receptor back to the cell surface, and the Qβ-BC2-high particle would then represent a “sweet spot” for particle transport. We previously observed an analogous preference for BC2-mediated delivery of a small-molecule cargo.⁶

Macrophages and endothelial cells in the liver are regarded as the greatest barriers to nanoparticle targeting, although their relative contributions to nanoparticle uptake are complex and contested.^{31, 55-58} While the asialoglycoprotein and related galatose-binding C-type lectin receptors are found on non-hepatic cell types,⁵⁹ Kupffer cells are usually identified as the most avid competitors to hepatocytes for ASGPR-targeted proteins and other cargo.^60-63 The picture that emerges from our studies, however, is one of competition between galactose-independent mechanisms of nanoparticle uptake, which direct viruses and virus-derived nanoparticles to endothelial cells,^{29, 64-68} and C-type lectin-mediated binding/uptake which includes hepatocytes, Kupffer cells, dendritic cells, and other immune cells.^{69, 70} A relatively small number of endothelial or Kupffer cells accounted for a significant degree of uptake of either particle, but glycosylated particles were able to evade barrier cell uptake to a much greater extent than non-functionalized particles, and distribute to greater numbers of hepatocytes. The magnitude of these effects, however, are on the order of 2-3-fold at most. We also note that engagement of a greater number of hepatocytes (Fig. 5e,f) is not accompanied by uptake of a greater number particles by those cells (Fig. 5d).

These findings resemble a previous report in which galactosylated BSA was observed to distribute more to hepatocytes in normal mouse but equally to hepatocytes and endothelial cells in ASGPR-deficient animals.⁴⁰ We also add an awareness of the importance of kinetics in receptor-based particle distribution. While attention is usually focused on clearance from blood circulation, we found that high-affinity ASGPR targeting gave rise to significantly faster clearance (of the covalently-attached dye) from the liver itself, and a change in distribution relative to untargeted particles at 1 h vs. 6 h. The disappearance of this apparent preference at 6 hours suggests that these cells (Kupffer macrophages, dendritic cells, other immune cells) degrade particles faster than LSECs, although endothelial cells also have this function.⁶⁷

Thus, particles of this size (approximately 30 nm diameter) encounter and are trapped by LSECs efficiently by ASGPR-independent processes (such as by scavenger receptors^71-73), regardless of their hydrophilic surface functionality (here the negative-control alcohol vs. GalNAc mimetic). However, the ASGPR ligand was able to direct the VLPs to partition better to GalNAc-binding cells.

CONCLUSIONS

Nanoparticles targeting with galactose-mimetic ligands are often assumed to distribute exclusively to hepatocytes. Indeed, studies often show liver-specific delivery, but delivery to non-parenchymal cells in vivo is not always carefully excluded.^{13, 19, 24, 74} Our results suggest that the extent of non-hepatic delivery in the liver should be carefully examined for nanoparticles and therapeutics using ASGPR-directed targeting.

Furthermore, the results described here, along with others in the literature, show that multivalent targeting of the asialoglycoprotein receptor is not achieved solely by enhancing the avidity of receptor binding. Indeed, high avidity may be counterproductive. For example, ASGPR targeting has found an interesting niche in the field of targeted protein degradation, where the signature properties of the receptor – its high capacity and turnover; expression on liver-associated cells (assisting in filtering); and the lysosomal fate of trafficked cargo – are effectively harnessed.^75-77 The parameters of intrinsic ligand affinity and the density and spacing of polyvalent display must also be carefully considered along with other constraints such as the stability of cargo and vehicle to modification and the potential for adverse effects of chemical modification on other desired function.²⁴ Overall, it is important to consider ASGPR residence time as a functional parameter, and to study nanoparticle trafficking within the liver, as well as clearance to the liver, in time-resolved fashion.

MATERIALS AND METHODS

Expression and purification of VLPs.

K46Q VLPs were produced and characterized as previously described.³⁵ Briefly, BL21 (DE3) chemically competent E. coli cells (Lucigen) were transformed with the pET28-CP(K46Q) plasmid according to the manufacturer’s protocol. Cells were plated onto selective super optimal broth (SOB-streptinomycin) agar and grown overnight. A single colony was inoculated into selective SOB media for overnight growth at 37 °C. After 12h, cells were diluted into fresh media (500 mL, SOB-strep) and incubated at 37 °C to mid-log phase growth (OD600 ~ 0.9). Protein expression was induced by addition of IPTG (1 mM final concentration). Cells were maintained at 30 °C for 16 h, then harvested by centrifugation (6k rpm, 10 min). Cell pellets were resuspended in 1x PBS, lysed via probe sonication (10 min lysis, 75 W, 5 s intervals) in an ice bath, and centrifuged to remove cellular debris (14k rpm, 10 min). VLPs were precipitated from the clarified cell lysate by the addition of 30% ammonium sulfate and incubation (1 h, 4 °C) followed by centrifugation (14k rpm, 10 min). The resulting protein pellet was resuspended in 1x PBS and treated with 1:1 n-BuOH:CHCl₃ to extract water-soluble protein from lipids and aggregates. Following centrifugation (14k rpm, 10 min), the aqueous protein-containing layer was collected and loaded onto 10-40% sucrose density gradients (28k rpm, 4 h), for further purification. VLPs bands were isolated via syringe and pelleted by ultracentrifugation (68k rpm, 2 h). Pellets were dissolved in 0.1 M KPO₄, sterilized via 0.2 μm syringe filters, and subsequently characterized.

Characterization of Qβ VLPs.

Protein concentration was determined by Bradford assay (Pierce, Coomassie Plus) against BSA standards. Particles were characterized by FPLC (Superose 6 size exclusion) to determine particle purity and aggregation, and by dynamic light scattering (Wyatt Dynapro plate reader) to determine hydrodynamic diameter.

VLP bioconjugation.

Amine-reactive succinimidyl ester chemistry was used to install both dye and azide (the latter by means of the NHS-PEG₁₂-azide linker shown in Figure 1a), followed by ASGPR-ligand-alkyne ligation via copper-catalyzed azide-alkyne cycloaddition, as follows. AlexaFluor647 NHS ester (0.1 μmol) was added to a solution of Qβ VLPs (2 mg, 0.14 μmol in capsid protein subunit) and, after 2 h of gentle mixing at 4°C, NHS-PEG₁₂-azide (21 μmol) was added. The reaction was allowed to proceed overnight at 4°C. The reaction mixture was purified using a PD-10 desalting column, followed by centrifugal filtration using an Amicon Ultra 100k MW cut-off device. The extent of particle modification was determined via ESI-TOF HRMS. The resulting azide-modified particles were treated with alkyne-terminated ASGPR ligand in the presence of copper-sulfate, tris((1-benzyl-4-triazolyl)methyl)amine (THPTA), aminoguanidine, and sodium ascorbate as previously described.⁷⁸ Ligand loading was controlled by addition of different concentrations of the alkynes (Supporting Information). Propargyl alcohol was used as a negative control. Upon completion, reactions were purified by PD-10 desalting columns, followed by centrifugal filtration using an Amicon Ultra 100k MW cut-off device. Protein recovery was determined via a Bradford assay and particles were characterized as described above to determine particle stability, purity, and the extent of modification.

Cell culture.

All culture media were purchased from Life Technologies (Carlsbad, CA), unless noted otherwise. Human hepatocytes expressing high (HepG2; human hepatocarcinoma cells, ATCC HB-8065) or low (SK-Hep; human liver adenocarcinoma cells; ATCC HTB-52) levels of ASGPR were cultured in supplemented RPMI medium (10% FBS, 1X sodium pyruvate, 1X GlutaMax, 1X Pen-Strep) and maintained in a humidified incubator (37 °C, 5% CO₂). Cells were routinely cultured in phenol red containing RPMI media but were switched to phenol red-free media for all imaging and flow cytometry experiments.

Live-cell fluorescence microscopy.

HepG2 cells or SK-Hep cells were seeded at density 5 x 10⁴ cells/well on glass-bottom 8-well μ-Slide (Ibidi; Munich, Germany) for 24 hours and then dosed with AlexaFluor647 labeled virus-like particles diluted in culture media. Cells were incubated with the particles for 1 hour at 37 °C in all procedures unless mentioned otherwise. Incubation was followed by washing with PBS and nuclear staining with Hoechst 33342 (Invitrogen). Cells were imaged immediately using either a Nikon Ti-U inverted microscope or a PerkinElmer UltraView spinning disk confocal microscope.

Procedure for competitive inhibition assay.

Cells were cultured at a density of 5 x 10⁴ cells/well on glass-bottom 8-well μ-Slide Ibidi dishes and dosed with increasing concentrations of unlabeled BC2 ligand for 15 min. Without removing the ligand from the well, cells were then treated with AlexaFluor647 labeled Qβ-BC2-high, Qβ-BC2-low, Qβ-capped or ASOR. Cells that received no BC2 ligand were used as controls. Images were captured using an Eclipse Ti-U fluorescence microscope (Nikon) equipped with a 60x oil immersion objective and a CoolSNAP CCD camera system (Photometrics).

Image-based analysis of nanoparticle binding and uptake.

Images from competitive binding experiments were analyzed using Fiji.⁷⁹ Cell boundaries were determined from the brightfield images, and the total fluorescent signal corresponding to internalized and surface-bound fluorescent protein was measured for each competitive binding condition. The fluorescent intensity per cell in each image was background corrected, and the average signal derived from each particle or ASOR was normalized against signal from a no-competition control (lacking the BC2 ligand). These data were plotted against increasing concentrations of excess non-labeled competitive binding ligand in GraphPad Prism, and fit using a one-phase decay model to determine the apparent IC₅₀.

Flow cytometry for assessing nanoparticle uptake.

Cells were cultured at a density of 2.5 x 10⁵ cells/well in 12-well plates and grown for 24 hours. Cells were washed and dosed with 20 nM of of AlexaFluor647 labeled Qβ-BC2-high, Qβ-BC2-low, Qβ-capped or 50 nM of ASOR. Cells were then washed and detached using accutase, followed by resuspension and washes in cation-free PBS. Cells were labelled with LIVE/DEAD^™ Fixable Aqua Dead Cell Stain Kit, for 405 nm excitation (Invitrogen) for 30 mins. Cells were then fixed in the dark for 15 min (PBS + 2% paraformaldehyde), before being washed and resuspended in PBS + 0.5% FBS. Cells were analyzed using an Aurora flow cytometer (Cytek) and the data were analyzed using FlowJo version 10.0.8 (Tree Star).

MTS cell viability assay.

Cells were seeded at a density of 5 x 10⁴ cells/well in 96 well plates and grown for 24 hrs. Cells were dosed with 50 nM of Qβ-BC2-high, Qβ-BC2-low, Qβ-capped, Qβ, or ASOR. Cells were incubated for 3 hours. MTS reagent (Promega) was added to each well at 1:10 dilution and incubated for 2 hours. Absorbance was measured at 490 nm.

In vivo biodistribution, cell isolation.

For the in vivo biodistribution study shown in Figure 5, mice were sacrificed 1 hour after administration of the VLPs and immediately perfused with 20 mL of 1× PBS through the right atrium. The liver was isolated immediately following perfusion, minced with scissors, and then placed in a digestive enzyme solution with collagenase type I (Sigma Aldrich),v collagenase XI (Sigma Aldrich) and hyaluronidase (Sigma Aldrich) at 37 °C and 750 r.p.m. for 45 min. Digested tissues were passed through a 70 μm filter and red blood cells were lysed.

In vivo biodistribution, cell staining.

Cells were stained to identify specific cell populations and analyzed using a BD FacsFusion analyzer. The antibody clones used for staining were anti-TER119 (Biolegend cat. no. 116220) anti-CD31 (Biolegend cat. no. 102407), anti-CD45.2 (Invitrogen cat. no. 78-0454-82), anti-CD68 (Invitrogen cat. no. 46-0681-82), anti-CD11c (Biolegend cat. no. 117349), anti-ASGPR (Clone 8D7, Santa Cruz Biotechnology, cat. no. sc-52623 AF488), and live/dead cell stain kit (Thermo Fisher, cat. no. L23105). Representative gating strategies for liver cell populations are included in Figure S19.

In vivo biodistribution, IVIS imaging and plasma sampling.

CD-1 IGS mice (6 weeks) were purchased from Charles River Laboratories and housed in the animal facility at the Georgia Institute of Technology on an alfalfa-free diet. All animal care and experimental procedures were approved by the Institutional Animal Care and Use Committee of the Georgia Institute of Technology. Mice were injected intravenously (tail vein) with either 30 μg of FNIR labeled VLPs or PBS, and whole animals were imaged using an IVIS SpectrumCT imaging cabinet (PerkinElmer) equipped with 745Ex/800Em filters under anesthesia. Reference images from mice immunized with PBS were acquired for all experiments to establish detection thresholds and to adjust for background fluorescence. For plasma clearance, mice were administered 30 μg of AF647-labeled VLPs by tail vein injection, and plasma collected at the submandibular vein at 5 mins, 30 mins, and 60 mins post-injection. Plasma AF647 signal was measured using a BioTek Synergy H4 Hybrid plate reader (excitation: 645/9.0, emission: 675/20.0).

Statistical analysis.

Image analysis was performed using Fiji, and statistics were performed using GraphPad Prism. Data processing steps are described in the appropriate methods subsections above, and statistical tests used are described in each figure caption.

Data availability statement

Data are available in the main text and in the supporting information.