Targeting low-density lipoprotein receptors with protein-only nanoparticles

Low-density lipoprotein receptors (LDLR) are appealing cell surface targets in drug delivery, as they are expressed in the blood–brain barrier (BBB) endothelium and are able to mediate transcytosis of functionalized drugs for molecular therapies of the central nervous system (CNS). On the other hand, brain-targeted drug delivery is currently limited, among others, by the poor availability of biocompatible vehicles, as most of the nanoparticles under development as drug carriers pose severe toxicity issues. In this context, protein nanoparticles offer functional versatility, easy and cost-effective bioproduction, and full biocompatibility. In this study, we have designed and characterized several chimerical proteins containing different LDLR ligands, regarding their ability to bind and internalize target cells and to self-organize as viral mimetic nanoparticles of about 18 nm in diameter. While the self-assembling of LDLR-binding proteins as nanoparticles positively influences cell penetration in vitro, the nanoparticulate architecture might be not favoring BBB crossing in vivo. These findings are discussed in the context of the use of nanostructured materials as vehicles for the systemic treatment of CNS diseases.


Introduction
Cell-targeted drug delivery and personalized medicines strongly push towards the 55 development of biocompatible materials adapted to deliver cargo drugs to specific cell types. A critical point in such design process is the selection of intrinsically non-toxic materials, which while keeping high structural and functional tunability would not induce side effects upon administration. Because of their biodegradability, biocompatibility and functional and structural plasticity, proteins are highly convenient materials to construct 60 carriers for the delivery of both conventional and emerging drugs (Lohcharoenkal et al., 2014). On the other hand, drug vehicles, apart from exhibiting powerful targeting properties, should overcome the sequential biological barriers encountered previous to reaching the right cell compartment in the target organ. This is compulsory when targeting the central nervous system (CNS) that is protected by the blood-brain barrier 65 (BBB) and by the blood spinal cord barrier. Since in a therapeutic context, local administration into brain is not desirable because its invasiveness (Lockman et al., 2002), systemic administration is mandatory and empowering drugs to cross the BBB has become a major issue in current pharmacology and nanomedicine (Pardridge, 2010). BBB tightly controls the access of molecules and drugs to brain, either by 70 paracelullar or transcellular pathways, by using both functional and structural elements addressed to maintain brain homeostasis (Barbu et al., 2009). Hydrophilic and cationic small molecules show some spontaneous penetrability. However, usual chemical drugs and therapeutic proteins cannot cross the BBB or are targets for the efflux pumps acting in the BBB. A nanoparticulate organization of vehicles used for systemic drug 75 delivery increases drug stability and circulation time (Cespedes et al., 2014), what preventing renal filtration offers potential for sustained release of the cargo. Although these and other properties of nanostructured materials are highly desirable, paracelullar penetration of nanoparticles targeted to the central nervous system (CNS) is assumed to be especially problematical. Functionalization with ligands of hormone 80 receptors or transporters for transcytosis is then mandatory despite the unexpected BBB-crossing activities exhibited by a few polymers used for nanoparticle fabrication and coating (eg polysorbate 80 and poly-[ethylene glycol-co-hexadecyl]-cyanoacrylate (Kim et al., 2007;Kreuter et al., 2002a)).

85
A catalogue of potential BBB-crossing peptides and proteins for functionalization is available (Van et al., 2012) (http://brainpeps.ugent.be). Among them, ligands binding transferrin, insulin and low density lipoprotein receptors (LDLR) have been especially appealing because of their transcytotic properties. LDLR, in particular, are of additional 4 interest as they are overexpresed in several human conditions including lung, stomach 90 and cervical cancers. Several LDRL protein ligands, namely ApoB (Spencer and Verma, 2007b); ApoE (Re et al., 2011;Wagner et al., 2012) and Apo A-I (Fioravanti et al., 2012;Kratzer et al., 2007), have been already used to functionalize diverse types of drugs and nanoparticles to allow or enhance BBB crossing. Others, such as Kunitzderived peptides (Angiopeps), presented in plain protein-drug complexes, have entered 95 clinical trials addressed to brain tumors. (Kurzrock et al., 2012).
(http://clinicaltrials.gov/ct2/show/NCT01480583?term=ANG1005&rank=6). Although several of these LDLR ligands have proved to be promising, the ideal architecture for the drug-ligand complex to effectively cross the BBB and reach the brain remains to be elucidated. In particular, whether the ligand would be more effective when 100 functionalizing a nanostructured vehicle than when applied in plain ligand-drug complexes remains unsolved, being a critical issue that needs further investigation (Juillerat-Jeanneret, 2008).
In the present study we have selected several known peptidic LDLR ligands and 105 explored them as BBB crossers, in protein-only materials under several presentations.
Some of these constructs self-organize as nanoparticulate materials while others remain in monomeric, unassembled forms. The in vitro and in vivo analyses of cell penetrability, biodistribution and brain targeting provide new concepts about the BBB crossing properties of functional protein nanoparticles, and suggest divergent diffusion 110 properties when acting in cell culture and upon systemic administration.

120
Bacteria were harvested through centrifugation and resuspended in Tris buffer (20 mM Tris-HCl pH 8.0, 500 mM NaCl, 10 mM Imidazol) in the presence of EDTA-free protease inhibitor (Complete EDTA-Free; Roche).Then, cells were disrupted by a French press (Thermo FA-078A) at 1100 psi, and the soluble fraction separated from the mixture by centrifugation at 15,000 g for 30 min. The insoluble fraction from ApoB-

125
GFP-H6 was stored at -80°C for further use.

140
The pellet of ApoB-GFP-H6 IBs was washed with water twice, and resuspended with solubilizing buffer (40 mM Tris with 0.2 % N-lauroyl sacosine, pH 8.0) in a ratio 1:40 and incubated for 24 h at room temperature. After that, the sample was centrifuged at 15000 g for 30 min. Resuspended soluble protein from IBs was purified as described above with prior N-lauroyl sarcosine removal by using a Hitrap QFF ion exchange 145 column (GE healthcare).

165
deposited onto carbon-coated grids for 2 min, stained with uranyl acetate and observed in a Hitachi H-7000 transmission electron microscope.

Confocal microscopy
HeLa cells were seeded on Mat-Teck culture dishes (Mat Teck Corp., Ashland, MA, were collected at 0.5 m intervals.

Fluorescence determination and dynamic light scattering (DLS)
All proteins were diluted to 400 μg/ml; then GFP fluorescence was determined by Cary Eclipse Fluorescence Spectrophotometer (Variant) at detection wavelength of 510 nm, 7 by using an excitation wavelength of 450 nm. Volume size distribution of nanoparticles and monomeric GFP fusions were determined by dynamic light scattering at 633 nm (Zetasizer Nano ZS, Malvern Instruments Limited, Malvern, UK).

190
Permeability studies were performed at the USEF Drug Screening Platform (http://www.usc.es/en/investigacion/riaidt/usef). Briefly, CaCo2 cells were cultured in DMEM high in glucose supplemented with 10 % FBS, 1 % nonessential amino acids (100x), 1 % L-glutamine, 100 U/ml penicillin, 100 µg/ml streptomycin 95 % air and 5 % CO 2 and at 37 °C. The cells (CaCo2, passage 65) were seeded in the apical 195 compartment of a sterile 6-well transwell at a density of 250,000 cells / well in 1.5 ml of medium and 2.5 ml of fresh medium was then added to the basal compartment. Cells were maintained in this medium for 21 days until complete differentiation (renewing the medium every 2 days). After this time, the medium was changed to HBSS (0.9 mM CaCl 2 , 0.5 mM MgCl 2 , and 20 mM HEPES, pH 7.4).

200
Transepithelial resistance (TEER) measurement was conducted using a Millipore

In vivo model and biodistribution analyses
Five-week-old female Swiss nu/nu mice weighing between 18 and 20 g (Charles River, L-Abreslle, France) and maintained in SPF conditions, were used for in vivo studies. All 220 the in vivo procedures were approved by the Hospital de Sant Pau Animal Ethics Committee and performed according to EC directives. Proteins were injected 8 intravenously at a dose of 500 µg/mouse (n=3 mice), control mice was injected with NaHCO 3 buffer. At 5, 15, 30 min, 1 h and 2 h after injection, mice were anesthetized with isofluorane and whole body fluorescence was monitored using IVIS spectrum 225 equipment (Xenogen, France). After that, mice were sacrificed and brain, kidney, lung and liver collected and examined separately at 30 min and 2 h for GFP fluorescence in an IVIS Spectrum. The ex vivo fluorescent recording of the brain was performed sequentially, first measuring the emission from whole brain and then of sagittal sections to achieve a complete fluorescent signal characterization.

Statistical analyses
Data were analyzed using one-way ANOVA and post hoc Tukey tests.

Results
Three chimerical genes were constructed to produce LDLR-binding recombinant 235 proteins (Table 1), based on the following modular organization; from N-to C-termini, ligand, linker, EGFP and H6 tail ( Figure 1A). Such organization had been previously proved useful in promoting the spontaneous formation of highly stable fluorescent protein nanoparticles, provided a sufficient positive electrostatic charge is present at the N terminus of the whole fusion (Cespedes et al., 2014;Unzueta et al., 2012).

240
Angiopep-2 and Seq-1 fusions were produced in E. coli as fully soluble versions while ApoB-GFP-H6 obtained from the soluble cell fraction was partially proteolized. In fact, protein sequencing by Edman degradation procedure of the soluble protein form revealed loss of the amino-terminal 34-mer peptide of ApoB (Table 1)

260
In a preliminary screening (Unzueta et al., 2012), Angiopep-2 and Seq-1 were observed as unable to promote the assembling of the fusion proteins in higher order nanoparticles, probably due to their low cationic amino acid content, although doubts remained about the potential influence of the composition of the different buffers used to store the proteins. All the proteins produced here were tested again for nanoparticle 265 formation under homogeneous buffer conditions as described above, in 166 mM NaHCO 3 , pH 7.4. The exclusive occurrence of unassembled forms of Seq-1-GFP-H6 and Angiopep-2-GFP-H6 was indeed confirmed (Figure 2A), with a particle size, determined by DLS, compatible with that of GFP monomers or dimers (as GFP naturally tends to dimerization). Contrarily, ApoB-GFP-H6 formed nanoparticles in both In this regard, we first wanted to explore cell penetrability of all constructs in cells 285 displaying and not displaying LDLRs. Uptake of protein constructs in LDLR -HUVEC was indeed negligible when comparing with that of closely related nanoparticles empowered by the unspecific but highly efficient cell penetrating peptide R9 (nine sequential arginines, (Vazquez et al., 2010)) ( Figure 3A). In contrast, penetrability was highly stimulated in LDLR + HeLa cells ( Figure 3B), especially in the case of ApoB-GFP-

290
H6IBs. In presence of chloroquine, internalization of ApoB-GFP-H6IBs protein in HeLa cell population dramatically increased (Figure 4), indicative of an endosomal route as expected for any receptor-mediated uptake (Vazquez et al., 2008). Interestingly, the penetrability of ApoB-GFP-H6s was always lower than that of ApoB-GFP-H6IBs. This fact suggests that an unstable nanoparticle might be less suitable for proper receptor 295 binding and cell internalization. Alternatively, the folding status of the protein (probably different as derived from the soluble cell fraction or from refolding) might influence ligand exposure and/or particle performance in a biologically significant way. The efficient cell penetration of ApoB-GFP-H6IBs was fully confirmed by confocal microscopy ( Figure 5). In general, the unassembled constructs were internalized by 300 cells in a less efficient way, and the uptake was not influenced by background protein precipitation in the extracellular medium that has been generally observed in GFPbased self-assembling proteins (Vazquez et al., 2010).
Considering these cell internalization results, the transepithelial crossing efficiency of 305 the LDLR-ligand functionalized modular proteins was determined in a fully stablished in vitro BBB model based on CaCo2 cells (Hellinger et al 2012) (Table 2). In the two protein concentrations tested, ApoB-GFP-H6IBs presented the highest penetrability in accordance with the internalization assays presented above (Figure 3). Angiopep-2-GFP-H6 and Seq-1-GFP-H6 also displayed minor but still important penetrability in this 310 BBB model at high protein concentration, thus suggesting a potential to effectively cross the BBB. However, when ApoB-GFP-H6s was challenged to the CaCo2 cell barrier, the apparent permeability was even lower than the negative control GFP, again indicating a failure of these protein nanoparticles to reach a fully functional status.
Indeed, the stability of the Caco2 cell monolayer is shown in the data related to Papp of 315 one the protein constructs (ApoB-GFP-H6s) at both protein concentrations maintained low through the experiment.  In a step further, and particularly encouraged by the good performance of ApoB-GFP-H6IBs nanoparticles we wanted to examine the biodistribution of the protein set and the potential influence of the supramolecular protein organization, upon systemic administration through the tail vein in healthy mice in which side events that affect brain 325 permeability such as enhanced permeability and retention (EPR) effect do not take place. We were specifically interested in this issue as at one side, LDLR are important targets in BBB-crossing for drug delivery into the CNS (Demeule et al., 2008;Kim et al., 2007;Spencer and Verma, 2007b), and also, cationic protein nanoparticles are biocompatible materials that fulfil most of the requests posed for vehicle-mediated drug 330 delivery into brain (Juillerat-Jeanneret, 2008). Therefore, we analysed ex vivo the signal from the whole brain to avoid the noise coming form the background of the whole body imaging followed by ex vivo recording of brain sagittal sections to complete evaluation of the extent of the emitted fluorescence. The analyses of these samples clearly indicated BBB-crossing properties of Angiopep-2-GFP-H6 and Seq-1-GFP-H6

335
( Figure 6A). Angiopep-2-GFP-H6, in particular, was observed as accummulating in the brain parenchyma 30 min after administration, a fact that was fully assesed by quantitative analysis of fluorescence under conditions that not allowed GFP-H6 background signal ( Figure 6B,C). Surprisingly, ApoB-GFP-H6s but also ApoB-GFP-H6IBs failed to accummulate into brain (Figure 6), indicating that the ApoB ligand was 340 unable to drive the crossing of BBB under the presentation offered by the resulting nanoparticles.
To understand better the stability in circulation and the potential renal clearance of both BBB-crossing and failing constructs, GFP fluorescence was also determined in kidney.

345
All the constructs that did not form nanoparticles (namely Angiopep-2-GFP-H6 and Seq-1-GFP-H6, and the parental GFP-H6) and also the unstable ApoB-GFP-H6s nanoparticles accummulated in kidney ( Figure 7A, B), indicative of renal clearance and consequently, of a material size under 8 nm (Cespedes et al., 2014). This is in agreement with the unability of Angiopep-2-GFP-H6 and Seq-1-GFP-H6 to self-350 assemble, and it also suggests that the ApoB-GFP-H6s nanoparticles, observed in vivo as unstable, probably dissasamble once in the bloodstream (maybe due to the high salt content of the biological fluid). No fluorescence was recorded in lung and liver, in any case (not shown).

355
These data indicates that a nanoparticulated architecture of ligand-containing proteins, promoting efficient cell penetrability and transcytosis, is neither sufficient nor necessary to reach the brain under systemic administration, and that unassembled soluble proteins, even when undergoing an effective renal clearance, are able to cross the BBB in a significant fraction.

Discussion
Proteins are excellent functional carriers for therapeutic nucleic acids and conventional drugs (Aris and Villaverde, 2004;Nehate et al., 2014). When fused to the amino terminus of a His tagged GFP, the cationic peptide ApoB promotes the formation of nanoparticles that are only composed by the modular protein acting as self-interacting 365 building block. This is based on a recently proposed protein engineering principle that allows designing protein nanoparticles by the fusion of cationic peptides to polyhistidine tagged polypeptides, and that act irrespective of the nature and sequence of the core protein (Cespedes et al., 2014;Unzueta et al., 2012). Nanoparticle formation is promoted by the hydrostatic contacts between the resulting dipolar monomers, but the 370 whole supramolecular structure is largely stabilized by additional forces such as Van der Waals, hydrogen bond interactions (Cespedes et al., 2014;Unzueta et al., 2014), and protein-DNA interactions if used as non-viral gene therapy vehicle . Interestingly, the amino terminal cationic peptide (ApoB in case of the current study) acts as an architectonic tag but also as a LDLR ligand with known BBB-crossing 375 properties (see Table 1). Under the same conditions, the less cationic Seq-1 and Angiopep-2 peptides, also LDLR ligands, fail in promoting nanoparticle formation ApoB-GFP-H6s, and also the differential cell penetrability of these constructs ( Figure   3B, 4, 5), can be only attributed to different conformations of the protein as resulting 14 from either the soluble cell fraction or from refolding from protein aggregates. For instance, the ApoB tail in ApoB-GFP-H6s might be more involved in crossmolecular contacts between building blocks and less available for cellular interactions. Of course, 400 the heterogeneity in protein bands detected in the Western blot analysis of the soluble E. coli cell fraction, probably resulting from selective proteolysis ( Figure 1B), could also contribute to this fact. Therefore, the conformational and structural status of protein building blocks of de novo designed nanoparticles, and the influence of the cell factory in the quality and properties of the final supramolecular assemblies is a currently 405 neglected field that deserves deeper exploration . This is especially relevant in the context of emerging biomaterials resulting from in vivo fabrication , the rising number of conventional and nonconventional cell factories for protein and polymer production  and the new bio-engineering strategies to 410 improve microbial biosynthesis regarding industrial and biopharma applications (Chen, 2012;Lee et al., 2012;Rodriguez-Carmona and Villaverde, 2010).
On the other hand, ApoB-GFP-H6IBs nanoparticles internalized cultured cells more efficiently than ApoB-GFP-H6s nanoparticle versions and than Seq-1-GFP-H6 and

415
Angiopep-2-GFP-H6 proteins ( Figure 3B, 5). The penetration of ApoB-GFP-H6IBs took place, as expected, in LDLR + cells but not in LDLRcells ( Figure 3A). The control R9-GFP-H6 nanoparticles, which are empowered by a potent Tat-inspired unspecific cell penetrating peptide (R9), do not shown any LDLR-linked preference in internalization ( Figure 3). LDLR-dependent internalization is dramatically enhanced by chloroquine 420 (Figure 4), indicative of an endosomal pathway. Under these conditions, ApoB-GFP-H6IBs but no other constructs was essentially found in all cells among the population exposed to the nanoparticle, even when applied at moderate doses (1 µM).
Although based on the good performance in in vitro experiments, ApoB-GFP-H6IBs 425 particles were highly promising regarding BBB-crossing, none ApoB-derived protein version was found in the brain parenchyma up to two hours after iv administration ( Figure 6). Surprisingly, Seq-1-GFP-H6 and Angiopep-2-GFP-H6 proteins were detected in brain in ex vivo imaging, with an occurrence that peaked at around 30 min.
BBB-crossing of these two proteins occurred even with important renal filtration ( Figure   430 7), while skipping renal clearance did not promoted, by itself, brain localization of ApoB-derivatives. Being ApoB a well-known BBB-crossing peptide for soluble drugs (Kreuter et al., 2002b) and also when linked to nanoparticles (Kim et al., 2007), failure in a proper activity when empowering protein nanoparticles might be due to 15 inappropriate presentation of the ligand in these kind of constructs. In fact, due to its 435 cationic nature, ApoB acts as both architectonic and targeting agent with limited solvent exposure when compared to ligands in monomeric proteins. Although such a dual activity is not by itself an obstacle for proper biodistribution of protein nanoparticles (as exemplified by the peptide T22 in similar GFP-based constructs) (Cespedes et al., 2014;Unzueta et al., 2012) and also for ligand-mediated cell penetrability (Figure 3 and 440 4), the most complex biological barriers imposed by brain vessels might represent a tighter bottleneck to proper biodistribution.

Conclusions
The results presented here upon exploration of three recombinant protein-only LDLR 445 ligands, presented in a total of four versions, reveal that high cellular penetrability in cultured cells does not guarantee efficient BBB-crossing and brain targeting mediated by transcytosis-associated receptors. Interestingly, protein versions in form of nanoparticles do penetrate cultured cells more efficiently than unassembled constructs, while the contrary is true regarding in vivo BBB-crossing. Such a divergent 450 performance prompts to evaluate the use of nanoparticulate materials for BBB-crossing therapies, which even being highly efficient in cell culture might find in vivo bottlenecks essentially distinguishable from those encountered when aiming to targets other than brain.
16 455 LEGENDS Figure 1. Structure of the fusion proteins. A) EGFP was used as the core of the fusions (green), flanked by a cell ligand at the N-terminus (blue) and a hexahistidine at the C-460 terminus (brown). A linker segment (orange) was placed between the ligand and GFP.
Residues in green indicate the end terminal amino acids of GFP in the joining regions.
The sequences of the fused N-terminal ligands are depicted in Table 1