Recombinant Protein Micelles to Block Transduction by SARS-CoV-2 Pseudovirus.

The continuing emergence of variants of the SARS-CoV-2 virus requires the development of modular molecular therapies. Here, we engineered a recombinant amphiphilic protein, oleosin, to spontaneously self-assemble into multivalent micellar nanostructures which can block the Spike S1 protein of SARS-CoV-2 pseudoviruses (PVs). Short recombinant proteins like oleosin can be formulated more easily than antibodies and can be functionalized with precision through genetic engineering. We cloned S1-binding mini-protein genes called LCBx, previously designed by David Baker's laboratory (UW Seattle), to the N-terminus of oleosin, expressing Oleo-LCBx proteins in E. coli. These proteins largely formed 10-100 nm micelles as verified by dynamic light scattering. Two proteins, Oleo-LCB1 and Oleo-LCB3, were seen to completely and irreversibly block transduction by both wild-type and delta variant PVs into 293T-hsACE2 cells at 10 μM. Presented in multivalent micelles, these proteins reduced transduction by PVs down to a functional protein concentration of 5 nM. Additionally, Oleo-LCB1 micelles outperformed corresponding synthetic LCB1 mini-proteins in reducing transduction by PVs. Tunable aqueous solubility of recombinant oleosin allowed incorporation of peptides/mini-proteins at high concentrations within micelles, thus enhancing drug loading. To validate the potential multifunctionality of the micelles, we showed that certain combinations of Oleo-LCB1 and Oleo-LCB3 performed much better than the individual proteins at the same concentration. These micelles, which we showed to be non-toxic to human cells, are thus a promising step toward the design of modular, multifunctional therapeutics that could bind to and inactivate multiple receptors and proteins necessary for the infection of the SARS-CoV-2 virus.

The emergence of the highly infectious SARS-CoV-2 virus, which is responsible for the Covid-19 pandemic, has created an urgent need to develop therapeutics, treatments, and vaccines to treat the Covid-19 disease.[1] The pandemic has caused millions of infections, comorbidities, and deaths, with the virus continually mutating and producing new variants with differing infectivity and potency. While vaccines remain the best way to prevent the severity of disease, the development of general, modular strategies for blocking infection by new mutants of SARS-CoV-2 or by other respiratory viruses is eminently necessary.

The lipid bilayer membrane of the SARS-CoV-2 virus contains the spike protein S, which extends ∼20 nm beyond the viral surface.[2] The S1 subunit of the S protein is responsible for the first step in infection, containing the receptor binding domain (RBD) involved in binding to the type I transmembrane metallocarboxypeptidase angiotensin converting enzyme 2 angiotensin (ACE2) receptor on the surface of epithelial cells that line the pulmonary tract.[3] Cleavage of the S protein, followed by activation by human proteases like furin, TMPRSS2, and cathepsins, allow spike protein subunits to undergo conformational changes that promote viral entry through fusion with the host cell membrane.[3,4] Many of the new variants of the SARS-CoV-2 virion involve mutations in the S protein. An effective way to design therapeutics to block the entry of SARS-CoV-2 would involve inhibiting the spike protein or fusogenic viral or host proteases, or a combination of these factors. A multifunctional therapeutic equipped with molecules that block entry by hindering more than one of these processes might ultimately be the most potent.

A powerful strategy to block viral infection is to use inhibitors such as short-chain peptides and mini-proteins that bind to the S1 protein with high affinity. Such moieties have either been designed based on ACE2 receptor binding epitopes[5−9] or designed de novo to have very low dissociation constants (Kd) with the S1 RBD.[10,11] Peptides have several advantages over bulky antibodies in that they are much easier to design and manufacture.[12] However, making multifunctional peptides can be difficult, requiring long linkers and complex chemistry between two functionalities and a potential alteration of the conformation of the final peptidic structure.[13−15] Furthermore, the standard chemical process of manufacturing peptides involving synthetic polymerization is usually not amenable for large-scale manufacturing and can often lead to a considerable portion of the product having amino acid sequences that are slightly different than those intended. Production of such peptides also becomes considerably more difficult and complex with the length of sequence.[13,16,17] Finally, small peptide molecules are short-lived in vivo with a half-life of minutes, leaving a potentially insufficient time to fight the pathogen, necessitating an increase in administered dose, thus wasting material and possibly increasing the risk of toxicity.[18−20]

Colloids and nanoparticles which incorporate functionalities like peptides present an attractive platform to inactivate the SARS-CoV-2 virus.[21−23] The structure and functionality of soft nanoparticles like vesicles and micelles can be varied over a wide range by changing the identity and mixture of surfactant(s) and incorporating bioactive agents such as engineered transmembrane proteins.[24,25] Recent examples include liposomes that carry and release peptides designed to block respiratory syncytial viruses,[26] liposomes with subdomains designed to act as decoys to inactivate bacterial toxins in the lung,[27,28] and polymersomes and proteolipid vesicles that are functionalized with reconstituted leukocyte receptors to associate with targeted endothelium.[29,30] All of these strategies involve the incorporation of functional domains or molecules onto the surface of the nanostructure.

In this study, we designed purely proteinaceous micelles that are capable of binding to the S1 RBD of SARS-CoV-2 and hindering viral ability to infect the cell (Figure ). The micelles are made of variants of a protein called oleosin, which is a triblock copolymer-like molecule that stabilizes oil bodies in sunflower seeds.[31] In prior work, we successfully modified the wild-type oleosin gene to recombinantly formulate variants in a bacterial expression system.[31−33] Unlike most proteins, these variants are highly amphiphilic and spontaneously self-assemble into nanostructures like vesicles, fibers, sheets, and micelles based on the composition of their hydrophilic arms and hydrophobic core. We cloned the S1 binding peptide and mini-protein genes into the N-terminal end of a particular oleosin variant to produce a library of proteins. There are many advantages to this approach. First, because the proteins are made recombinantly, they are uniform. Second, the strategy is modular, in that any peptide of interest can be incorporated by genetic modification without any need for post-formulation conjugation chemistry. Third, because oleosin is a surfactant, it spontaneously self-assembles into micelles without the aid of additional lipids or block copolymers. Fourth, multiple functionalities can be incorporated in the same structure by blending different oleosins, leading to a multifunctional therapeutic.[34] Finally, functionalities which are otherwise sparingly water-soluble can be delivered in high concentrations via loading onto the oleosin backbone due to the tunable aqueous solubility of oleosin variants. We characterized the nanostructures assembled from various oleosins, tested their ability to block viral transduction in vitro, demonstrated the utility of multifunctionality in fighting off transduction, and explored the cell–micelle interactions in terms of toxicity and nanostructure cellular fate. The results from this study establish a promising modular platform for inactivation of viruses.

Figure 1

Schematic showing molecules (A) used to form micelles and mechanism (B) of micelle inactivation of SARS-CoV-2 pseudovirus.

Results and Discussion

Oleosin Gene Design and Expression

The naturally occurring wild-type oleosin has a triblock copolymer-like structure with 87 amino acid residues in the middle hydrophobic block.[31] Previously, variants of the oleosin, truncated at both the hydrophilic and hydrophobic portions, have demonstrated spontaneous self-assembly into nanostructures. Variants in which the hydrophobic part was reduced to 30 residues along with the addition of 5 glycine residues in the backbone to impart flexibility (30G) have been shown to assemble into micelles of diameters ranging from 10 to 100 nm.[32] 30G variants with only anionic, zwitterionic, and uncharged residues in the hydrophilic arms can be produced with both high purity and yield.[33,34] In the current study, we used an oleosin designated 25-30G-30(−) with 25 and 30 residues in the N- and C- terminal hydrophilic arms, respectively, with an isoelectric point of ∼4.2 at neutral pH as the base material. To label this oleosin protein with a fluorescent dye, a lysine was added through a point mutation near the N-terminus (Q4K); this protein is termed Control Oleosin.

Several short amino acid sequences (23–65 residues) that have been shown to bind to the SARS-CoV-2 spike protein receptor binding domain computationally and/or experimentally were cloned into the N-terminus of the 25-30G-30(−) (full sequences in Table S1). Mini-proteins known as LCB1, LCB3, and LCB7, originally designed de novo by Baker and co-workers,[10] were used to create oleosin sequences that we call Oleo-LCB1, Oleo-LCB3, and Oleo-LCB7 (or, Oleo-LCBx). ACE2 receptor derived small peptide sequences were used to produce Oleo-23mer and Oleo-31mer genes.[7,8,35] The 23-mer peptide was derived by selecting a sequence from the α-1 helix of the ACE2 peptidase domain; while the 31-mer peptide was designed by selecting two specific parts of the human ACE2 protein interface, lining them with glycine residues. After confirming the gene composition via T7 promoter sequencing, genes were expressed in the vector pBamUK with a polyhistidine (6H) tag at the C-terminus to enable purification by immobilized metal affinity chromatography (IMAC). The yields of each protein varied considerably, with Control Oleosin having the highest yield (50–70 mg per 1 L of starting bacteria stock) and Oleo-LCBx being produced at relatively lower amounts (10–15 mg per 1L). Protein expression was carried out in E.coli (BL21-DE3), and the purity of each protein was assessed via gel electrophoresis, as shown in Figure A (and Figure S1.A).

Figure 2

(A) SDS-PAGE images showing bands (red arrows) corresponding to Control Oleosin and each of the Oleo-LCBx proteins. Each protein was run on a gel using the same protein standard ladder on the left. (B) Distributions in the hydrodynamic diameter of 10 μM of protein samples obtained via dynamic light scattering. Each plot represents the average size distribution (n = 3) for micelles assembled from one particular type of protein, except the sample labeled “49-50-1_Control-LCB3-LCB1” which is a blend containing 1% Oleo-LCB1, 50% Oleo-LCB3, and 49% Control Oleosin.

Self-Assembly of Oleosin Variants

Similar to synthetic free-chain surfactants, oleosin can form micelles above its critical micelle concentration (cmc). Earlier reports of both the neutral and anionic 30G oleosin proteins and all of their variants have shown a cmc of 3–6 μM in 1× phosphate buffered saline (PBS).[32,34] To characterize the self-assembly of the oleosin variants used in this study, we conducted dynamic light scattering (DLS) with 10 μM of each protein. From Figure B, we see that Control Oleosin formed micelles with a mean size of ∼25 nm, while Oleo-LCBx formed larger micelles with mean hydrodynamic diameters from 30 to 40 nm. This could be due to the larger molecule size (∼10 kDa vs ∼14–18 kDa) or a difference in intermolecular forces stemming from altered charge in the hydrophilic arms. Oleo-LCBx constructs also displayed a small but distinct population of small-diameter micelles. Since it was our aim to show therapeutic dose response for the functional proteins, we carried out DLS of micelles in which Oleo-LCB1 and Oleo-LCB3 were blended with the Control Oleosin, as shown in Figure S2. Blends of Oleo-LCBx with Control Oleosin produced mostly bimodal curves, in which, as the mol % of Oleo-LCBx relative to that of Control Oleosin decreases, neither the right peaks (indicative more of Oleo-LCBx only) nor the left peaks (indicative more of Control Oleosin) increase or decrease proportionally to the change in concentration. For instance, the peak corresponding to larger micelles for 0.1% or 0.05% Oleo-LCBx is only about half the value of that for 100%. This suggests that a blend of proteins forms 3 populations of micelles: Control Oleosin only, Oleo-LCBx only, and micelles containing both (which is the desired population). Evidence of higher-diameter peaks in both Figure B, corresponding to Oleo-LCBx, and Figure S2, corresponding to all the blends, could also mean that some micelles are not spherical or that micelles stick together to form dimers or trimers.

Efficiency of Oleosin Variant Micelles in Blocking Transduction by SARS-CoV-2

To assess the ability of functional oleosin-based micelles to block viral entry, we used SARS-CoV-2 pseudovirus particles (PVs) as a model to test the inactivation effectiveness of different micellar compositions. PVs are essentially liposomes with the SARS-CoV-2 Spike S1 protein at the membrane, making them antigenically alike to the wild-type virus.[36] PVs have a second-generation lentiviral core containing a green fluorescence protein (GFP) gene, which is expressed upon transduction.[37] PVs have been proven to serve as a viable, non-replicating model for testing transductions by dengue, zika, and SARS-CoV-2.[38−40] 293T-hsACE2 cells express ACE2 receptors and can be transduced by PVs via ACE2-mediated endocytosis and/or via virus-cell membrane fusion; these cells were thus used as the in vitro cell model. PVs are able to infect 293T-hsACE2 and induce a fluorescence signal at 488 nm excitation, as shown in Figure . Approximately 104 transduction units (TUs) of PVs were used per well containing the 293T cells. When 10 μM of micelles of different anti-S1 oleosin variants are co-incubated with PVs, the fluorescent signal is reduced to varying degrees. The quantification of the signal represents a combination of both the number of cells that were infected as well as the intensity of fluorescence in different host cells (Figure ). While presence of Control Oleosin did not produce any significant change in fluorescent signal, Oleo-LCB1 and Oleo-LCB3 were seen to block transduction by PVs completely; this was determined by way of a lack of a finite quantifiable signal from images for these samples. Oleo-LCB7 reduced transduction by approximately 30-fold when compared to positive control. Oleo-31mer and Oleo-23mer by themselves led to relatively limited signal change (Figure S1.B). The K of free LCB1 and LCB7 binding the RBD were theoretically predicted to be <1 nM, below what could be experimentally determined with certainty. The Kd of LCB7 was experimentally determined to be ∼20 nM while that of the 31mer and 23mer were around 40 nM and greater.[8,35] However, the difference in signal between LCB7 and LCB1 is disproportionately larger than the difference in Kd, suggesting that Kd is not the sole determinant in blocking transduction by PVs. For remaining experiments, Oleo-LCBx proteins were used. To confirm results from microscopy, flow cytometry and Immunospot[41] imaging were utilized for a 10 μM sample of Oleo-LCB1 micelles and controls, which both showed a complete absence of transduction signal with Oleo-LCB1 relative to controls (Figures S3 and S4).

Figure 3

Representative images and fluorescence quantification (bottom right) of 293T-hsACE2 cells co-incubated with only PVs (positive control) and with PVs and 10 μM micelles of Control Oleosin or Oleo-LCBx. Quantification was done with approximately n ≈ 10 images for each sample from at least 2 different trials. Error bars represent standard deviation. Scale bar: 50 μm.

Dose Response of Oleosin Micelles

To assess the therapeutic dose required to elicit significant inactivation of the PVs, structures that combine nonfunctional Control Oleosin and functional Oleo-LCBx were prepared, with the total protein concentration fixed at 10 μM to form micelles.[32,34] The fraction of Oleo-LCBx in the blend was changed from 0.05% (5 nM) to 100% (10 μM). Images (Figure S5) and corresponding signal quantifications (Figure ) showed a marked distinction between samples with Control Oleosin and those with Oleo-LCBx as low as 0.05% (5 nM). In blends, Oleo-LCB1 proved to be the most effective at the same concentration.

Figure 4

Therapeutic dose response of Oleo-LCBx against transduction by PVs. As Oleo-LCB1 (black), Oleo-LCB3 (orange), Oleo-LCB7 (purple) fractions decrease relative to Control Oleosin (red) in micelles that come from a total protein concentration of 10 μM, the fluorescent signal increases monotonically. Each data point represents n ≈ 10 images from multiple samples. #: This sample with 10% Oleo-LCB1 had finite signal for only n = 2 images, resulting in the level of fluorescence reported here. All other images for this sample had zero fluorescence. All other samples had either 100% images with a finite signal or 100% images with zero signal. Error bars represent standard deviation.

Efficiency of Co-incubation versus Pre-incubation of Micelles in Blocking Transduction by SARS-CoV-2

In the trials described thus far, PVs were co-incubated with micelles. However, to gain a better insight into PV-micelle attachment and PV inhibition, micelles containing a high (10 μM, 100%) and low (10 nM, 0.1%) amount of Oleo-LCB1 were preincubated with PVs separately before dilution and addition to the cells. Figure A (and Figure S6) shows that at high concentration of Oleo-LCB1 (10 μM), preincubation is at least as efficient as co-incubation in blocking transduction. Moreover, with 10 nM (0.1%) Oleo-LCB1 in the blended micelle, the signal corresponding to transduction completely disappeared as well, thus outperforming co-incubation with the 0.1% sample (which showed a low, but finite signal).

Figure 5

(A) Quantification of transduction by PVs into 293 T cells after PVs were either co-incubated (black) or premixed (for 1 h, RT) with 10 μM oleosin micelles containing fractions of 100% or 0.1% Oleo-LCB1 in Control Oleosin. There was finite signal only from co-incubation of PV with 0.1% Oleo-LCB1. (B) Quantification of transduction by PVs into 293T cells when PVs were co-incubated with Oleo-LCB1 micelles (black) or Free LCB1 (green). Error bars represent standard deviation for n ≈ 10 images.

Efficiency of Oleosin Micelles versus Free Mini-Proteins for Blocking SARS-CoV-2 Transduction

We assessed the efficiency of free LCBx mini-proteins in blocking trasduction by PVs and compared the results to that for Oleo-LCBx micelles. Because of the insolubility of higher LCB1 concentrations, a final concentration of 100 nM and 10 nM mini-protein samples (corresponding to 1% and 0.1% Oleo-LCB1) were tested against transduction by PVs. We found that at these concentrations, the performance of LCB1 mini-protein was inferior to that of Oleo-LCB1-containing micelles, especially for the 10 nM concentration sample, as seen in Figure B (and Figure S7). This showed that for the current system, oleosin micelles can be better inhibitors than free mini-proteins. Additionally, by virtue of the high aqueous dispersibility of oleosin, these micelles also enable introducing peptides/mini-proteins at high concentrations at which those free peptides/mini-proteins would fall out of aqueous solution.

Efficiency of Composite Micelles in Blocking Transduction by SARS-CoV-2

One of our goals was to demonstrate the advantage of combining multiple binding motifs into nanostructures to combat the virus. To that end, we formulated multifunctional structures by blending Control Oleosins and select concentrations of Oleo-LCB1 and Oleo-LCB3, with all three proteins present in the same formulation. The hypothesis was that multiple functionalities on a micelle can perform better than a single functionality to bind to a greater number of spike protein RBDs. Therefore, concentrations of functional oleosins that showed high levels of PV inactivation were chosen to work in tandem. For instance, in a total concentration of 10 μM protein, 0.1% Oleo-LCB1, 1% Oleo-LCB1, 10% Oleo-LCB3, and 50% Oleo-LCB3 (with the remainder being Control Oleosin) were all effective in near-complete blocking of transduction by PVs. When Oleo-LCB1 and Oleo-LCB3 were blended, the combination resulted in superior blocking, as illustrated in Figures and S8. Combining the individual molecules at their own highest functional levels—1% Oleo-LCB1 and 50% Oleo-LCB3—led to a complete elimination of any signal, akin to the results from 100% Oleo-LCB1 or Oleo-LCB3. This combination likely formed a mixture of micelles of different sizes, as evidenced by the fairly broad hydrodynamic diameter distribution illustrated in Figure B.

Figure 6

Representative images and quantification (bottom right) of transduction by PVs into 293T cells when co-incubated with Control Oleosin micelles containing Oleo-LCB1 or Oleo-LCB3 or both, at a final total concentration of 10 μM. Error bars represent standard deviation for n ≈ 10 images. Scale bar: 50 μm.

Cell–Micelle Interactions and Effect on Resistance to Viral Transduction

The results in this study indicate that micelles assembled from oleosin linked to peptides designed to bind to the spike protein can interact with PVs and interfere with their transduction into cells. On co-incubation, we expect a certain population of functional micelles to bind to PVs, preventing their entry into the cytoplasm. After 24 h of co-incubation, any unbound micelles or PVs in the ambient media were removed. We questioned whether any micelles were taken up by the host cells over a period of 24 h. To test the cellular fate of micelles, Control Oleosin and Oleo-LCB1 were used as model proteins. These two variants were functionalized with an NHS-Cy5 dye via the lysine molecules on the N-terminus. The NHS ester and the primary amine in the lysine formed an amide bond rendering the micelles fluorescent. After incubation for 24 h with and without PVs, washing, and further incubation for 2 days, the cells were stained with the cell membrane-permeable Hoechst 33342 dye to identify the cell nuclei. When imaged, it was seen that there was considerable fluorescence signal from oleosin in the vicinity of the cell nuclei. (Figure A). This was the case regardless of the type of oleosin used or the presence or absence of PVs (Figure S9). For instance, while the green fluorescence from PVs was not seen in the Oleo-LCB1 incubated sample, the far-red fluorescence from the Cy5 attached to the Oleo-LCB1 was significant. It follows that a portion of the micelles therefore are taken up by cells based on the properties of the base oleosin itself. The LCB1 portion or PVs have no discernible differential effect on the uptake.

Figure 7

(A) Representative fluorescent images of micelle (red) retention by cells (nuclei stained blue) in the presence of PVs (green) and 10 μM Oleo-LCB1 or Control Oleosin. Images were taken ∼72 h after protein + PV addition. (B) Quantification of PV signal when cells were first incubated for 30 min, 4 h, or 24 h with either 0.1% (dark blue) or 100% (gray) Oleo-LCB1 micelles before PV addition. The left of the green dividing line represents co-incubation of micelles and PVs for 24 h, as previously reported in Figure . Samples in which 100% Oleo-LCB1 micelles were pre-incubated had approximately 12% (30 min), 2% (4 h), and 35% (24 h) of available images show zero quantifiable signal, which were not included in the displayed mean fluorescence value (indicated by the symbol #). Error bars represent standard deviation for n ≈ 10 images. Scale bar: 50 μm.

Since micelles were observed to be retained by cells, it is imperative to ensure that they do not cause toxicity. Cell viability in the presence and absence of micelles was measured via an XTT assay where degradation of the XTT tetrazolium salt by enzymes from metabolically active cells produces a highly colorimetric, water-soluble formazan salt.[42] When micelles of Control oleosin and of different variants were incubated with 293T-hsACE2 T cells at the highest protein concentration used (10 μM), the viability of treated cells with respect to untreated cells was not affected to any significant degree (Figure S10), indicating the feasibility of oleosin as a safe inactivation agent for SARS-CoV-2.

Finally, the effect of retention of micelles by cells on transduction by PVs was tested. Cells were first incubated with a high and a low concentration of Oleo-LCB1 (10 μm, 100% and 10 nM, 0.1% in micelles) for 30 min, 4 h, or 24 h, with subsequent washing to remove all micelles. Next, PVs were added for 24 h, and subsequent washing and cell imaging were carried out as before. The data for these trials are shown in Figure B (and Figure S11). We found that cells gain resistance to transduction by PVs when pre-incubated with micelles: the longer the incubation of Oleo-LCB1 micelles, the lower the PV signal. For the low concentration sample (0.1%), the decrease in PV tranduction signal with incubation times from 30 min to 4 to 24 h is consistent, indicating that the pre-incubation time of cells with micelles dominates the kinetics of PV–micelle binding. However, with 10 μM Oleo-LCB1, much of the signal reduction is within the first 30 min, indicating that kinetics is dictated more by concentration than preincubation time. While pre-incubation with 10 μM micelles did not completely eliminate PV signal as it did with the co-incubation protocol, the reduction in signal was significant—the pre-incubation time of cells with micelles dominates the kinetics of PV transduction. However, at 10 μM Oleo-LCB1, much of the reduction in signal occurs within the first 30 min, indicating that kinetics is dictated more by concentration than pre-incubation time. While pre-incubation with 10 μM micelles did not completely eliminate PV signal as it did with the co-incubation protocol, the reduction in signal was significant.

Micelle Action against Delta Variant PVs

Finally, to demonstrate the versatility of the platform, Oleo-LCB1 and Oleo-LCB3 micelles were tested against PVs expressing the spike protein from the Delta variant of SARS-CoV-2 (B.1.617.2). Dose response curves in Figure (and Figure S12) show that both oleosin variants are highly effective in blocking transduction by Delta PVs, at efficacies comparable to the Wuhan variant PVs. Previously, free LCB1 and LCB3 mini-proteins have been seen to be considerably less effective against the Alpha and Beta variants of SARS-CoV-2 in vivo.[11] The delta variant has been theorized to have multiple means to evade immune response and exacerbate infection in in vivo cells. For instance, the P681R mutation on the Delta spike protein has been shown to enhance furin-mediated cleavage of the spike protein helping fusion,[43] a phenomenon which may occur when the in vitro cell line expresses fusion-specific receptor proteases absent in the current system. Our results establish that when it comes to the first line of defense, i.e., blocking virus S1 RBD-ACE2 receptor interaction, these micelles are well-suited to perform as blocking agents.

Figure 8

Therapeutic dose response of Oleo-LCB1 against transduction by Delta variant PVs. As Oleo-LCB1 (black) fraction decreases relative to Control Oleosin fraction in a total of 10 μM micelle sample, the fluorescent signal increases monotonously. Each data point represents n ≈ 10 images from multiple samples. #: This sample with 0.1% Oleo-LCB1 had 20% images with zero signal, which were not included in the displayed mean fluorescence value. Error bars represent standard deviation. Scale bar: 50 μm

Discussion

As an alternative to large-molecule treatments like monoclonal antibodies, peptides, smaller molecules like small proteins, sybodies, and nanobodies have been designed and tested both computationally and experimentally to block the SARS-CoV-2 virus from attaching to human cells because of their relative ease of production and low costs.[6,7,10,35,44,45] One approach to design short-chain peptides involves molecular dynamics and Rosetta modeling to use part or whole of the human ACE2 receptor helix to which the spike protein S1 binds.[7,35,46] As models, we have utilized oleosins functionalized with a 23-mer and a 31-mer with Kd values in the tens of nM range. However, while these sequences performed well in binding to the S1 RBD computationally, they were unable to block transduction by PVs to any appreciable degree. A second approach involves designing sequences ab initio, which allows for a potential wider range of functionalities from which to choose. Baker and co-workers generated mini-binders corresponding to different areas of the S1 RBD surrounding the region that binds to the ACE2 receptors by utilizing rotamer interaction field (RIF) docking with large in silico mini-protein libraries.[10] They found two mini-proteins (54 and 65 residues), LCB1 and LCB3, to be the most potent with Kd < 1 nM and IC50 values of 20–50 pM for blocking 102 PFUs (plaque forming units) of the virus from infecting Vero E6 cells. While our system is different, we observed that our Oleo-LCB1 and Oleo-LCB3 variants proved to be highly effective in completely inactivating ∼104 transduction units (TUs) of the PVs when co-incubated at 5–10 μM micelle concentration, in keeping with the trends seen for the free mini-proteins. Calculated IC50 of Oleo-LCBx micelles under these specific conditions range from ∼3–9 nM with the lowest dose tested being 5 nM functional Oleo-LCBx. This indicates that the conformation of the mini-proteins was well preserved on an oleosin backbone and when asssembled into a micelle. We also observed that pre-incubation of Oleo-LCB1 micelles with PVs and subsequent dilution of the PV–micelles rendered PV inhibition even more effective than co-incubation. This shows that along with micelle concentration, incubation time with PVs is also an important factor determining binding efficiency. This would also suggest that micelle–PV binding is irreversible, that the micelles do not detach after dilution, and that blocking transduction by PVs involves irreversible conformational changes to the spike protein after micelle binding. We also saw that Oleo-LCB7, while not as effective as Oleo-LCB1 and Oleo-LCB3, still performed significantly better than the ACE2-derived 23-mer and 31-mer peptides. LCB7 has a Kd of 10–20 nM, which is not much higher than the other two peptides; hence, the difference in results point out that other factors like final peptide conformation and areas of RBD targeted likely play an important role along with Kd to determine peptide-S1 binding.

Both Oleo-LCB1 and Oleo-LCB3 were almost equally effective in negating transduction by the Delta variant as they were against the original Wuhan strain. This is an exciting result since in previous studies, even the alpha variant (B.1.1.7) with the fewest significant mutations decreased the efficacy of free LCB1 50-fold.[11] If we were to equate PVs with live virus and PFUs with TUs, this would make the micelles similarly effective against delta variant as the mini-proteins were against the alpha variant but less effective against the Wuhan strain than free LCB1. This suggests that only a small number of micelles need to be engaged; this could be due to only a few micelles attaching to each PV because of the large size of micelles, while free LCB1 molecules are sufficiently small to completely cover the virus surface. The comparable efficacy of the micelles against both variants could also stem from steric effects wherein multiple micelles binding to the PV surface physically block PV-ACE2 interaction. However, this may also not be a perfect comparison since the experimental designs of the previous and current studies are different. This can be seen from the results in this study where free LCB1 mini-proteins were less efficient at PV-blocking than Oleo-LCB1 micelles at the same functional concentration. This could originate from steric effects due to micelle size or due to the multivalency of micelles.[47,48] Both in vitro and in vivo studies are therefore necessary to compare micelles and free mini-binders for different variants under the same experimental setup.

Another crucial observation from this study involves the continued blocking of PVs from transduction into cells pre-exposed to micelles. Cells likely retain oleosin molecules in their membranes because of the amphiphilic nature of oleosin causing self-assembly mediated membrane incorporation. Subsequently, the cell-incorporated oleosins remain functional and work to prevent viral entry into the cell long after micelles have been removed in the surrounding media. Since oleosin is not toxic to cells, this feature can have favorable applications in prolonged resistance to viral infection and in prophylactic treatment of patients before viral exposure.

The immediate goal of this study was to develop structures that can bind and clear SARS-CoV-2. However, our strategy of putting functional groups on oleosin is modular, in that domains of proteins embedded in the nanostructures can be swapped with other bioactive domains of different specificity as new infectious agents, such as mutant/variant strains, emerge. As a proof-of-concept, several oleosin variants were blended, leading to structures that are more effective than micelles with only one type of protein. With variants like Omicron having some 30 mutations, few treatments are highly effective; even many monoclonal antibodies are not as potent, and there is a huge shortage due to high demand.[49] Compared to expensive and cumbersome production of antibodies, these recombinant proteins can be manufactured directly in E. coli and scaled-up with relative ease.[10] It will be imperative to attack the virus from multiple fronts, including targeting different areas of the spike protein, incorporating moieties that block other pathways of viral entry, and packing small-molecule drugs in the core.[4,50] For instance, fusion inhibitory peptides which have been discovered in recent years to block SARS-CoV-2 infections can be used to genetically functionalize oleosin chains.[4,51−54] This approach can be extended to other nanostructures like lipid vesicles and droplets where there will be much greater control in incorporating specific oleosins distributed uniformly within the vesicle population.[26,55,56]

Conclusion

In the current study, we made functional surfactant proteins through recombinant modification of a naturally occurring amphiphilic protein, oleosin, and assembled functionalized oleosins into different structures to block the entry of SARS-CoV-2. Different functional peptides and mini-proteins (LCBx) were genetically incorporated into the N-terminus of oleosin; after expression in E. coli and purification, these functional proteins were seen to spontaneously self-assemble into micellar nanostructures as indicated by DLS. In particular, oleosin variants Oleo-LCB1 and Oleo-LCB3 with Kd < 1 nM were highly effective in irreversibly blocking the entry of SARS-CoV-2 pseudovirus particles down to Oleo-LCBx concentrations of 5 nM when incorporated in micelles where the total oleosin concentration was 10 μM. Oleo-LCB1 micelles proved to be more inhibitory than free LCB1 mini-proteins, with the former also enhancing the aqueous solubility of the latter. Also, certain blends of Oleo-LCB1 and Oleo-LCB3 in micelles were more effective at inactivation together than individually, providing a proof-of-concept feasibility of making multifunctional micelles and nanoparticles that could counter multiple mechanisms of viral entry into 293T-hsACE2 cells. None of the micelles were found to be toxic to cells in vitro at the highest doses used. A significant population of micelles was seen to be taken up by the cells regardless of protein composition or the presence of PV, which caused cells to retain their ability to resist infection from the virus 24 h after pre-incubation. Finally, Oleo-LCB1 and Oleo-LCB3 micelles were seen to be equally effective against the Delta variant of the virus with an excellent therapeutic dose response.

Future work will involve visualizing detailed micellar structures with single and multiple protein blends using cryo-electron microscopy. We will also incorporate anti-fusogenic peptide-modified oleosins into micelles or nanovesicle shells in addition to ACE2-dependent anti-Spike protein peptides studied here and test them for their ability to hinder PV entry into cells expressing both ACE2 receptors and relevant proteases. Additionally, micelles and vesicles active against PVs will be tested against the live SARS-CoV-2 virus to validate their therapeutic efficiency.

Experimental Section

Materials

Phosphate-buffered saline (PBS), lysozyme, Luria–Bertani (LB) media, and HisTrap Ni-NTA columns were purchased from Sigma-Aldrich (St. Louis, MO). In-fusion cloning kit with Stellar cells was obtained from Takara Bio (San Jose, CA). E. coli (BL21-DE3 strain) and cloning buffers were obtained from New England Biolabs Inc. (Ipswich, MA). Tris-buffered saline (TBS) was purchased from Bio-Rad Laboratories (Hercules, CA). Protease Inhibitor Tablets were obtained from Roche (Indianapolis, IN), and NHS-Cy5 was purchased from Lumiprobe Corp (Cockeysville, MD). Kanamycin Sulfate, B-PER Bacterial Protein Extraction Reagent, Penicillin–Streptomycin, Trypsin-EDTA, Dulbecco’s Modified Eagle Medium (DMEM, with high glucose), Puromycin Dihydrochloride, 4% paraformaldehyde in PBS, and Hoechst 33342 dye were purchased from Thermo Fisher Scientific (Waltham, MA). Fetal bovine serum was purchased from Hyclone (Logan, UT). Pseudovirus particles (PV) of SARS-CoV-2 (both Wuhan-Hu-1 and Delta variant B.1.617.2 strains) and 293T-hsACE2 cells were purchased from Integral Molecular (Philadelphia, PA). Free LCB1 mini-protein was custom-made and sold by Genscript Biotech Corp (Piscataway, NJ).

Gene Design and Cloning

The base oleosin gene has been used in our lab as mentioned previously.[33] Genes are contained in the expression vector, pBamUK, a pET derivative. To design mini-protein oleosin conjugates, Snapgene software was used, and the gene sequences of the mini-proteins were cloned with the base oleosin sequence at the N-terminus. Cloning was carried out using an In-Fusion protocol according to the manufacturer’s specifications. For cloning of shorter peptides, the short peptide DNA sequence is added to the base oleosin sequence, and the entire sequence is cloned into an empty pBamUK vector using the same protocol. The final product was transformed into Stellar cells, and the colonies were sequenced with Genewiz (Azenta Life Sciences) to verify gene sequences.

Protein Expression and Purification

All oleosin variants were expressed in an E. coli strain BL21 DE3 (New England Biolabs) controlled by the lac promoter. The base oleosin protein has been previously designed to have a 6-Histidine tag for purification using immobilized metal affinity chromatography (IMAC). Bacterial cultures were grown in Luria–Bertani (LB) media (for Control Oleosin and LCB1-Oleosin) and in Terrific Broth (TB) media for all other variants (TB is used to increase final yield for the other variants because of relatively lower levels of expression) along with kanamycin (50 μg/mL) at 37 °C. Protein production was induced using a final concentration of 1 mM isopropyl-β-D-1-thiogalactoside (IPTG) when culture optical density reached around 0.8. After overnight growth at 18 °C, cell pellets were separated via centrifugation at 4500g for 10 min and stored at −20 °C until further use.

Cell pellets obtained from each 1 L of culture were lysed using B-PER, lysozyme, protease inhibitor, and DNase for 30 min, as detailed previously.[31] The supernatant from cell lysis was obtained by centrifuging at 15000g for 15 min. The supernatant was filtered through a 0.22 um syringe filter and incubated with 3–4 mL of HisPur Ni-NTA beads for 1 h in a 50 mL falcon tube with gentle rotation. The beads were then washed 8–10 times with wash buffer (25 mM Imidazole in 1X TBS with 500 mM NaCl). Next, 500–700 mM Imidazole in the same TBS was used as elution buffer to elute out fractions of the desired protein from the beads in a column. The eluted protein was then dialyzed overnight against 1X PBS; the purified proteins were stored at −80 °C for characterization for further use.

Characterization of Micelles

All experiments were carried out with a total 10 μM protein concentration. Proteins were first filtered through a 0.22 μm nylon syringe filter to remove any aggregates. Micelles were expected to form spontaneously. To break up any aggregates, protein samples were bath-sonicated for a minute before use. For blends of different proteins together to form composite micelles, a 10 s vortexing was carried to facilitate mixing. Micelle size distribution was measured using dynamic light scattering (DLS) using a Malvern Zetasizer Nano ZS. Three different readings were taken with 1 mL of any given sample in a semi-micro cuvette at 25 °C. The resulting intensity versus size plots were then averaged to produce the final hydrodynamic diameter size distribution.

Viral Transduction into Cells

293T-hsACE2 cells were cultured according to the manufacturer’s protocol. Briefly, cells were grown in DMEM media with 10% FBS and 1% P/S in a T75 flask. One day after seeding, media was replaced and a final concentration of 5 μg of puromycin was added to induce ACE2 receptor expression on the cell membrane. 2–3 days later, cells were detached with 0.05% trypsin-EDTA, washed, and plated at ∼1.5 × 104 per well (at a final volume of 100 μL) in a 96-well plate and allowed to adhere and grow for 24 h. Following this, 25 μL of PVs were added with or without 10 μL of oleosin proteins at a final concentration of 10 μM. Within this 10 μM, different functionalities of oleosin variants were incorporated as required. After 24 h of incubation, the cells were washed to remove residual PVs and micelles in the media. Media was replaced after 24 h again. Finally, after 24 h more (a total of 72 h after adding PVs), cells were imaged in the 96-well plate under bright field and at 488 nm excitation using an inverted Nikon Eclipse TE300 phase contrast microscope. All areas in the wells having a near-100% coverage by cells were imaged. For pre-incubation studies, 10 μL of micelles containing 100% or 0.1% Oleo-LCB1 (with the rest being Control Oleosin) were incubated with 25 μL of PV for 1 h at RT. Following this, the mixture was added to the cells, and subsequently, the samples were washed and imaged in time as mentioned before. For resistance to transduction by PVs studies, 10 μL of Oleo-LCB1 micelles was incubated with cells at a final concentration of 10 μM for 30 min, 4 h, and 24 h. Following incubation, cells were washed 5× with PBS, and 25 μL of PVs was added.

Free LCB1 Mini-Protein Assay

The commercially purchased, synthetically made LCB1 mini-protein was mildly water-soluble. A solution of 3% acetic acid in PBS was used to dissolve 100 μM LCB1. This was subsequently diluted in PBS to concentrations of 1 μM and 100 nM LCB1, and 10 μL of each was added to each well with PVs and 100 μL cell media to attain final LCB1 concentrations of approximately 100 nM and 10 nM, respectively (with final acetic acid concentrations being 0.003% and 0.0003%). Subsequent washing and imaging was carried out as mentioned in Viral Transduction into Cells.

Image Analysis

Multiple images taken from across multiple wells over at least 2 separate trials were considered to quantify the fluorescent signal. Only images with a near-100% cell coverage (as confirmed by the corresponding bright-field image) were used to ensure uniformity of cell density for comparison (Figure S13). ImageJ (NIH) was used to quantify both the frequency (number of fluorescent spots) and the intensity (how bright the spots were) of the images. In doing so, 10 random regions of interest from the background were selected for each image, and an average mean gray value from those regions was subtracted from mean gray values for each fluorescent ROI so as to remove the effect of background noise in the image. The sum of the product of background-subtracted mean gray values of each fluorescent ROI and the corresponding ROI area was calculated for each image; an average of the sum values from approximately 10 images was then used as the final signal from each image. A value of zero was allotted for samples with no images with any quantifiable fluorescence. IC50 values were calculated using a standard 4-parameter model online at AAT Bioquest.

Cell Viability Assay

Cells were allowed to grow for 24 h in a 96-well plate. Next, the media was replaced with fresh media, and in line with the experiments conducted, 10 μL of different Oleosin variants (to a final concentration of 10 μM in the well) were incubated with the cells for another 24 h. Following this, 40 μL of a freshly prepared XTT reagent was added to each well, and the samples were incubated at 37 °C for 4 h. A Tecan Infinite M200 microplate reader was used to measure absorbance at 475 nm with a reference reading at 660 nm. The absorbance values from multiple wells for each sample were then averaged and expressed as a percentage of the positive control sample (no proteins), where it was assumed that the positive control had a 100% viability.

Micelle Uptake by Cells

Control Oleosin and Oleo-LCB1 proteins were labeled with NHS-Cy5. Briefly, 100 μL of NHS-Cy5 was mixed with 1 mL of proteins in 1X PBS at an 8:1 dye-to-protein molar ratio, and the samples were gently agitated for 4 h at room temperature. The mixture was the dialyzed against 1X PBS with 3 changes of PBS to remove any unbound dye. Labeled proteins were flash frozen using liquid nitrogen and stored at −80 °C until further use.

For cell–micelle interaction studies, 10 μL of labeled micelles was incubated with cells in 100 μL of media (to a final protein concentration of 10 μM) for 24 h, with and without 25 μL of PVs, same as the procedure mentioned before. The wells were then washed with PBS to remove proteins and PVs remaining in the media. After a further 48 h (with 1 change of media in between), the cell nuclei stained with Hoechst 33342 (final concentration of 8 μM in 100 μL media) for 30 min at 37 °C. Imaging was done under both 488 nm and 647 nm excitation to visualize the signal from PVs and micelles, respectively.

Flow Cytometry

After treatment with PVs with and without 10 μM Oleo-LCB1 or Control Oleosin as mentioned in Viral Transduction into Cells, each well was trypsinised and resuspended in complete growth media in 1.5 mL microcentrifuge tubes. Cells were then centrifuge-washed in PBS and resuspended in 0.5 mL of PBS with 5% FBS for each well. Cells were then injected into a BD Accuri C6Plus flow cytometer and tested against a 488 nm excitation channel.

Immunospot Imaging

After treatment with PVs with and without 10 μM Oleo-LCB1 or Control Oleosin as mentioned in Viral Transduction into Cells, each well was washed with PBS and 40 μL of 4% paraformaldehyde was added to each well and left at RT for 15 min. Following fixing of cells, each well was washed 3 times with PBS. The plate was then imaged with CTL ImmunoSpot S6 Core Analyzer, and the number of fluorescent clusters in each image was counted using the corresponding Immunospot software.