Transport of Molecular Cargo by Interaction with Virus-like Particle RNA

Abstract

Virus-like particles (VLPs) derived from Leviviridae virions contain substantial amounts of cellular and plasmid-derived RNA. This encapsidated polynucleotide serves as a reservoir for the efficient binding of the intercalating dye thiazole orange (TO). Polyethylene glycol (PEG) molecules and oligopeptides of varying length, end-functionalized with TO, were loaded into VLPs up to approximately 50% of the mass of the capsid protein (hundreds to thousands of cargo molecules per particle, depending on size). The kinetics of TO-PEG binding included a significant entropic cost for the reptation of long chains through the capsid pores. Cargo molecules were released over periods of 20–120 hours following simple reversible first-order kinetics in most cases. These observations define a simple general method for the noncovalent packaging, and subsequent release, of functional molecules inside nucleoprotein nanocages in a manner independent of modifications to the capsid protein.

Entry for the Table of Contents

Pick ‘em up and drop ‘em off. The binding of intercalating dyes such as thiazole orange (TO) to polynucleotides allows for a simple method to load RNA-rich virus-like particles with molecular cargo, and then to release that cargo by diffusion. TO-labeled molecules come out much slower than they go in because of the high RNA concentration inside the shell; the surprise is how much cargo these easily-produced nanocontainers can carry.

Introduction

Nanoparticles have long been regarded as having great potential for drug delivery,^[1] energized by a vision of vehicles that can be directed to different locations in the body for the transport and release of different functional cargo.^[2–7] In addition to liposomes, synthetic polymers and dendrimers, and inorganic nanoparticles, viruses and other natural nanoparticles represent highly promising materials for this purpose.^[8] We focus here on virus-like particles (VLPs), self-assembled structures from recombinantly-produced viral capsid proteins,^[9–11] and particularly those derived from well-known E. coli bacteriophage (Leviviridae) such as Qβ, MS2, and PP7.^{[12, 13]} VLPs in general, and these examples in particular, possess several advantageous features such as high physical and chemical stability, easy and scalable production, and well-defined structures known to atomic resolution. As a result, VLPs have been used as carriers for nucleic acids,^[14] polymers,^{[15, 16]} and therapeutic proteins and enzymes.^[17–19]

Since natural viruses efficiently package their genomic cargo, virus-like particles derived from capsid proteins capture whatever forms of polynucleotide are available under the conditions of their expression and self-assembly. VLP-packaged nucleic acids have in turn been used to noncovalently capture small-molecule imaging agents and drugs,^[20–25] but the effects of VLP porosity and cargo size on the properties of entrainment and release have remained largely unexplored. Here we show the use of a well-known intercalating dye as the key component of a modular encapsulation strategy in VLPs derived from the Leviviridae Qβ. This system has allowed us to probe the influence of VLP porosity and cargo size on loading and release.

The 29 nm diameter Qβ capsid forms an icosahedral (T = 3 symmetry) structure from 180 copies of a 14.3 kDa coat protein (CP).^[26] The apparent pore diameter is 1.3–1.4 nm at the threefold icosahedral symmetry axis and 0.7 nm at the fivefold axis.^[26] As these estimates are derived from the electron density visible in x-ray crystallography, they do not take into account the dynamic mobility of the protein around the pores, which is likely to be considerable. While these holes allow the free diffusion of small molecules (MW <1000 Da) in and out of the capsid,^[17] larger molecules such as enzymes are excluded, thereby protecting packaged protein from proteolytic degradation under a variety of conditions.^[27] When expressed in E. coli, the Qβ VLP randomly packages cellular and plasmid-derived RNA, but the entrained polynucleotide can be biased toward sequences that include a high-affinity RNA hairpin sequence that the natural virus uses to selectively encapsulate its own genome.^[17]

Cyanine dyes are well known to intercalate into RNA and DNA in single- or double-stranded forms. Thiazole orange (TO) is a prototypical example, the fluorescence intensity of which increases dramatically when bound to nucleic acid, making it an often-used probe for the detection of DNA and RNA.^{[28, 29]} Compared to most other cyanine dyes, TO exhibits higher nucleic acid affinity (K_d = 10⁻⁵-10⁻⁶ M)^[30] and less toxicity.^[31] TO variants are readily synthesized, and functional groups can be attached without significantly perturbing the intercalating and fluorescence properties of the dye.^[31–33]

We used the ability of TO to diffuse through the Qβ capsid and bind to entrained RNA as a probe of capsid permeability and a method to package and release attached molecules (Scheme 1). The intercalation interaction allows for rather efficient packaging, in numbers inversely proportional to cargo size. We demonstrated the generality of this strategy by successful application to small molecule and oligopeptide cargo.

Scheme 1.

Schematic representation of the strategy used for packaging cargo within VLPs and the general structure of TO-cargo molecules used in this work.

Results and Discussion

The thiazole azide derivative 1^[34] was employed as a common building block for the preparation of PEG, biotin, and peptide conjugates by copper-catalyzed azide-alkyne cycloaddition (CuAAC) as described in Supporting Information. These “TO_Q” structures, as noted by Kelley and coworkers, exhibit more intense fluorescence than corresponding molecules functionalized by attachment to the benzthiazole fragment.^[35] The PEG products were designated as TO-PEG_n, where n denotes the average number of glycol repeat units in the chain (8, 45, 113, or 227); peptide adducts were designated as TO-Pep_n where n designates the number of amino acids (18, 19, or 29) in the peptide.

Loading of TO into Qβ VLP

Bacteriophage Qβ is composed of a 180-subunit icosahedral capsid packaging a 4.2 kilobase single-stranded RNA assisted by a high-affinity interaction of the interior surface of the coat protein shell with a hairpin RNA sequence. Expression and purification of the Qβ VLP was performed as previously described,^[17] providing typical yields of 120–150 mg of pure VLPs per liter of E. coli culture. All batches packaged approximately the same amount of RNA as indicated by UV-vis spectroscopy (A₂₆₀/A₂₈₀ ratio = 1.9 ± 0.1). Binding of the parent TO molecule by VLP-packaged RNA, and the stability of the resulting Qβ@TO complex, were examined as a benchmark. Qβ particles were incubated in the presence of increasing concentrations of TO (up to 8000 times molar excess with respect to particle) at 4 °C. An increase in the fluorescence intensity of these solutions provided evidence of dye intercalation, reaching maximum values within seconds, and remaining unchanged for at least 12 hours, suggesting the rapid achievement of a binding equilibrium. Each mixture was then subjected to sucrose gradient (10–40%) ultracentrifugation to remove excess dye. VLPs appeared in a bright greenish-orange band at the standard density in these gradients, consistent with the expected retention of TO by the particles. Particle recoveries, determined by Bradford assay, were 70–85%, typical for these types of manipulations.

Size-exclusion chromatography showed a standard particle size (Figure 1a vs. 1c) and co-elution of the dye (absorbance at 510 nm) with nucleoprotein (260 and 280 nm). Removal of >95% of the packaged RNA by Pb-mediated hydrolysis^[16] before exposure to TO eliminated dye binding to the particle (Figure 1b vs. 1c). When approximately 80% of the RNA was removed by shorter duration hydrolytic treatment, the amount of VLP-associated TO was reduced by the same percentage (Fig. S8), both observations supporting the assumption that TO is trapped by RNA intercalation rather than by association with capsid protein. The observation of a consistent A₂₆₀:A₂₈₀ absorbance ratio for all RNA-containing samples, TEM analysis (Fig. 1f), and dynamic light scattering (Fig. S5) all illustrated the intact nature of the particles.

Figure 1.

Characterization of Qβ VLPs: (a) SEC and (e) representative negative-stained TEM image of Qβ particles before treatment with TO. (b) SEC of Qβ-RNA@TO⁴ showing almost no packaging of TO. (c) SEC and (f) TEM of Qβ@TO⁴, showing the intact nature of the particles and co-packaging of the intercalating dye. (d) UV-vis absorption spectra of Qβ@TO samples prepared by incubation with different concentrations of TO, normalized with respect to absorbance at 260 nm.

Incubation of particles with increasing concentrations of TO give rise to both greater uv-visible absorbance and a shift to shorter wavelength, as shown in Figure 1d. These effects are consistent with an increase in the average number of dye molecules entrained in each particle, the blue shift being similar to previous reports of dye aggregation in aqueous solutions at low micromolar concentrations (Fig. S3).^{[36, 37]} Because the molar absorptivity of the dye is not accurately represented by calibration under non-aggregating conditions, absorbance measurements of the dye-loaded particles significantly underestimated TO concentration. Similarly, the fluorescence emission from Qβ@TO samples showed sharp quenching at TO incubation concentrations above 40 µM (Fig. S4). Accordingly, we determined the amount of TO encapsulated in Qβ particles by first disrupting the capsid to release packaged TO into dilute solution, followed by absorbance measurement at λ_max = 510 nm.

Paired with measurement of capsid protein concentration by Bradford assay, we found that the particle could retain dye up to 52% of the protein mass when incubated with 4 mM TO (8000 times molar excess relative to VLP, Table 1, Table S1). Thus, for every mg of capsid protein (comprising 0.4 micromoles of VLP), 0.52 mg of thiazole orange could be retained, comprising about 1 mmol of dye, or approximately 2800 molecules per particle. (This is approximately two-thirds the number of RNA bases previously estimated by analytical ultracentrifugation to be entrained in these recombinant VLPs.^[17]) This level of loading was achieved within five minutes and was unchanged after 24 h incubation with TO (Fig. S6). The resulting VLPs appeared unchanged in size and shape (Fig. S7). Particles resulting from this treatment are designated Qβ@TO^m, where m = the concentration of TO in mM used during the incubation. Upon long term storage of Qβ@TO^m particles in buffer at room temperature or 4 °C, TO was slowly released over time. However, lyophilized samples of Qβ@TO^m particles were found to have lost none of the entrained dye and to have retained full structural integrity after storage for several weeks and rehydration in buffer. The kinetic and thermodynamic aspects of cargo release are discussed in detail below.

Table 1.

Incubation concentrations and average numbers of packaged TO and TO-PEG_n molecules with increasing size of PEG chain. Values are average number of molecules bound per particle; incubation time before sucrose gradient purification = 12 h in all cases; [Qβ] for all experiments = 1.25 mg/mL. Error is standard deviation for at least three independent replicates.

cargo conc. (mM)	TO	TO-PEG8	TO-PEG45	TO-PEG113	TO-PEG227
0.005	10 ± 2	8 ± 1.3	2 ± 0.6	2 ± 0.5	1 ± 0.2
0.02	40 ± 6	25 ± 5	5 ± 1.6	3 ± 1	2 ± 0.4
0.2	301 ± 20	99 ± 13	20 ± 2	14 ± 2	8 ± 2
0.4	630 ± 32	298 ± 17	38 ± 6	26 ± 4	15 ± 5
1.0	1487 ± 161	732 ± 56	69 ± 11	47 ± 6	35 ± 8
2.0	2577 ± 276	984 ± 81	101 ± 12	57 ± 8	50 ± 7
4.0	2769 ± 237	1056 ± 162	195 ± 18	120 ± 6	114 ± 12
10.0	–	1505 ± 127	390 ± 29	186 ± 15	191 ± 14

Entrainment of TO-PEG_n conjugates as function of cargo size

To test the use of TO as an anchor for the trapping of neutral cargo molecules in Qβ particles, we prepared TO-PEG_n conjugates of different chain lengths (Scheme S2) and subjected them to the same loading and release operations as the parent dye. Overnight incubation of Qβ with TO-PEG_n at concentrations up to 20000 times molar excess with respect to VLP was followed by sucrose gradient purification. Visual examination of the gradients confirmed the expected inverse relationship of chain length to packaging efficiency: the maximum number of dyes that could be sequestered inside the VLP (determined as described above) declined from approximately 2800 per particle for unfunctionalized thiazole orange to approximately 190 for TO-PEG₁₁₃ and TO-PEG₂₂₇ when incubated with the VLP at 10 mM concentration (Table 1).

The apparent size of the Qβ@TO-PEG_n particles was found to increase with increasing PEG chain length, as indicated by decreasing retention volume on size-exclusion chromatography (12.3 mL for Qβ and Qβ@TO-PEG₈¹⁰ particles vs. 12.1 mL for Qβ@TO-PEG₄₅¹⁰ and Qβ@TO-PEG₁₁₃¹⁰ vs. 11.5 mL for Qβ@TO-PEG₂₂₇¹⁰) and increasing hydrodynamic radius (R_h) by dynamic light scattering (Figure 2, Fig. S10-13). In the latter measurements, R_h was also found to increase with increasing TO-PEG_n concentration during the loading step, going from approximately 15 to 16 nm for Qβ@TO-PEG₄₅ and Qβ@TO-PEG₁₁₃ incubated at 0.4 mM vs. 10 mM, respectively, and from 14.9 to 18.0 nm for Qβ@TO-PEG₂₂₇ over the same concentration range. In contrast, particles containing different amounts of different TO-PEG_n cargo molecules appeared the same in negative-stain electron microscopy (mean diameter = 26 ± 1 nm for all cases: Fig. 2, Fig. S10-13).

Figure 2.

Characterization data for Qβ@TO-PEG_n particles. (a) plot of encapsulation efficiency (EE) and (b) loading efficiency (LE) of TO and TO-PEG_n at different incubation concentrations of these molecules. (c) Size exclusion FPLC and (d) negative-stain (uranyl acetate) transmission electron microscopy of the Qβ@TO-PEG₁₁₃ particle as a representative case. (e, f) Absorption spectroscopy, normalized with respect to absorbance at 260 nm of the indicated particles in the presence of increasing concentrations of the TO derivative. Complete data are provided in Figures S9-S13; a brief additional discussion regarding panel (f) appears in Supporting Information.

These data suggest that longer PEG chains protrude through holes in the Qβ capsid shell: by interacting with solvent and changing particle diffusion properties, dangling PEG chains can influence hydrodynamic radius but are not stained by uranyl acetate and so are invisible by electron microscopy. A pulldown experiment with streptavidin (SA) coated beads (Figure 3) showed a small amount of nonspecific adsorption to the beads by untreated Qβ VLPs and particles loaded with the short bifunctional cargo molecule TO-PEG₈-biotin, contrasting with much greater capture of VLPs loaded with a much longer biotinylated thiazole-PEG molecule (TO-PEG₂₂₇-biotin). In other words, while none of the approximately 735 TO-PEG₈-biotin chains contained in each VLP exposed enough biotin on the exterior surface of the particle to bind to SA-coated beads, at least some of the approximately 41 TO-PEG₂₂₇-biotin molecules associated with Qβ presented surface-accessible biotinylated chain ends, presumably by virtue of the longer PEG chains threading through capsid pores while thiazole orange anchors in the packaged RNA. Fluorescence microscopy and size-exclusion chromatography measurements described in Figures S15-S16 were consistent with this conclusion.

Figure 3.

Pulldown of Qβ@TO-PEG_n-biotin¹ particles (n = 8 and 227) with streptavidin (SA) coated beads (blue, not to relative scale) with microfluidic electrophoretic analysis. Lanes: a = Qβ, ; b = Qβ + SA beads, c = Qβ@TO-PEG₈-biotin¹, d = Qβ@TO-PEG₂₂₇-biotin¹. “QβCP“ = VLP coat protein, “monomer” = monomeric CP, “dimer” = noncovalent dimer of CP.

Under standard conditions, equilibrium entrainment of all TO conjugates, from smallest to largest, was observed by eye within five minutes (Fig. S6). More precise measurements of the fluorogenic signals produced upon TO-RNA binding were obtained by stopped-flow kinetics, allowing for millisecond-resolution quantitation beginning 2 ms after mixing of VLP with cargo (Figure 4, Fig. S17). Saturation equilibrium was achieved in the presence of a large excess of TO-PEG_n cargo at rates inversely proportional to the square of the cargo molecular weight (Fig. 4f), consistent with the dependence expected for the reptation of linear polymers through holes.^[38]

Figure 4.

(a-e) Stopped-flow kinetics of the appearance of TO fluorescence emission (540 nm) upon mixing of the indicated TO molecule (20 µM) with Qβ particle (0.2 µM). Pseudo-first order kinetic fitting (black lines) produced the indicated association rates and half-times of formation of the maximal signal (t_1/2). Fluorescence intensity (y-axis) was plotted by normalization to the maximum signal. (f) Plot of association rate constants, derived from panels a-e, vs. the inverse square of the molecular weight of the TO-PEG_n molecule.

Release of thiazole orange-functionalized cargo from VLP-encapsidated RNA

Since thiazole orange binding to RNA is noncovalent in nature, the entrained dye can be released by diffusion when excess dye is removed. Thus, incubation of various Qβ@TO-cargo particles with excess salmon sperm DNA (average length ∼2000 base pairs)^[39] for 12 h extracted much, but not all, of the packaged dye. In this experiment, virus-like particles and DNA were cleanly separated after incubation by sucrose gradient ultracentrifugation, and the amount of TO or TO-PEG_n conjugate was determined by uv-visible absorption (after denaturation of the biomolecule and extraction, as noted above). The results, summarized in Figure 5a, showed that the longer PEG conjugates were more resistant to release, but also that packaged RNA was competitive for all TO-containing molecules against a large excess of duplex DNA, presumably due to the much greater local concentration inside the VLP. Separate gel-shift electrophoresis analysis established that the PEG chain of the largest adduct (TO-PEG₂₂₇) does not abrogate intercalation to DNA (Fig. S18).

Figure 5.

(a) Percent of TO or TO-PEG_n released from Qβ particles after incubation in the presence of salmon sperm DNA (average length ∼2000 base pairs) for 12 h. The concentration of DNA was constant (1 mg/mL) and the concentration of Qβ@cargo was varied to provide a ratio of base pairs to TO of approximately 57 in all cases. Error bars are standard deviation from three independent replicates. (c-k) Representative data for release of cargo from freshly-loaded Qβ particles to 100 mM PBS at the indicated temperature. For panels (d) and (e), the dialysis buffer contained salmon sperm DNA or bovine serum albumin, respectively. Dots represent observed data as the mean of triplicate measurements; curves are lines fitted to these data using the formula shown next to panel (b), giving approximate apparent values of first-order rates of release (k₁) and re-binding (k_-1) in a reversible process. The term k₀* captures a slow simultaneous zero-order release process in some cases; when it does not appear on the plot, k₀* = 0. The last data point in each plot is labelled with a percent value noting how much of the entrained compound was released from the VLP during the experiment. [X]₀ = starting concentration of packaged cargo, calculated as though the cargo were in free solution. Data from all such experiments are summarized in Table 2.

The practical rate of molecular release from Qβ@TO-cargo (where TO-cargo is the unfunctionalized dye or TO-PEG_n conjugates) was measured by dialyzing the freshly-loaded particle (cellulose dialysis membrane, 50 kDa cutoff, 0.25 mL) into 100 mL PBS buffer. Release of the thiazole orange-containing molecule was quantified by absorption spectroscopy of the bulk buffer solution. The dialysis membrane provided no hindrance to the diffusion of TO itself; TO-PEG₈ was slowed somewhat but passed this barrier much faster than the observed rate of release from VLPs. In contrast, TO-PEG₁₁₃ diffused through the dialysis membrane only very slowly and TO-PEG₂₂₇ to an insignificant extent, eliminating them from this assay (Fig. S23).

Representative results are shown in Figure 5 (raw data and the full set of plots may be found in Supporting Information, Fig. S19A,B-S21A,B). Fitting the data to a simple reversible first-order equation shown in Figure 5b produced apparent rates of release from, and reassociation to, the particle (k_out and k_in*, respectively); these values are collected in Table 2. Key observations are as follows.

Table 2.

Release of molecular cargo from Qβ particles assayed by dialysis against buffer. “to DNA” or “to BSA” refers to experiments in which DNA or BSA were present in the dialysis reservoir, but not in the VLP solution.

Qβ@ a	T (°C)	[X]0 (µM) b	max. release c	kout (h−1)	kin* (h−1)	k0* (µM•h−1)
TO1	25	1.83	30%	0.016	0.04	0
TO2	25	3.17	27%	0.025	0.07	0
TO1	37	1.83	76%	0.04	0.0015	0
TO2	37	3.17	80%	0.04	0.0013	0
TO1 to DNA	25	1.89	51%	0.03	0.09	0.0041
TO1 to DNA	37	1.89	88%	0.045	0.025	0.0038
TO1 to BSA	37	1.56	100%	0.09	0	0
TO-PEG81	25	0.90	79%	0.052	0.042	0.0030
TO-PEG82	25	1.21	77%	0.050	0.055	0.0052
TO-PEG810	25	1.85	79%	0.060	0.048	0.0060
TO-PEG81	37	0.90	90%	0.15	0.026	0
TO-PEG82	37	1.21	88%	0.125	0.02	0
TO-PEG810	37	1.85	92%	0.11	0.014	0
TO-PEG451	37	0.052	31%	0.1	0.22	0
TO-PEG4510	37	0.341	25%	0.055	0.18	0
TO-PEP180.5	37	0.207	92%	0.165	0.13	0.0025
TO-PEP190.5	37	0.259	90%	0.13	0.18	0.0034
TO-PEP290.5	37	0.234	87%	0.06	0.009	0

^[a]Cargo molecule packaged inside VLP.

^[b]Starting concentration of cargo, calculated as though the packaged molecules were free in solution.

^[c]Percentage of starting cargo released at the end of the experiment (after 100–150 h).

(i) No difference was found in the release rates of TO from Qβ particles incubated with 1 mM TO for various times (5 min – 24 h) and then separated from excess TO (Fig. S22), showing that no functional “annealing” of entrapped TO occurs inside the particle: the equilibrium established initially is the final equilibrium.

(ii) With the possible exception of a small effect on TO itself, the presence of the Qβ capsid (with most of its packaged RNA removed) had no influence on the dialysis method used to assess cargo release (Fig. S23).

(iii) The values of k_out estimated in this manner are lower (slower) by factors of approximately 10⁶ than the binding rates (k_b) established by stopped-flow kinetics above (Figure 4), consistent with the overall high binding affinity of packaged RNA for TO. Thus, release of these cargo molecules required several days to approach or reach equilibrium.

(iv) The values of k_in* do not represent real rate constants, but rather the relative extent of re-binding to the VLP in the context of the dialysis experiment that starts with all of the cargo entrained in the particle. For example, only approximately 30% of packaged thiazole orange was released at equilibrium for two different dye loadings (Fig. 6a-b), so k_in* was found to be greater than k_out in the simple first-order model. In reality, the situation is more complex.

TO-cargo escapes the particle into a zone of greater dilution outside the particle (250 µL volume), and then to much greater dilution (100 mL) once through the dialysis membrane (the passage of which is inhibited for larger molecules, as noted above). Incubation at 37°C liberated significantly more thiazole orange (up to approximately 80% of the packaged dye) compared to 25°C (approximately 30%), consistent with the greater contribution of entropy at higher temperatures. The apparent k_in* value was therefore much smaller at the higher temperature, while k_out increased only modestly (Fig. S19B panels a,b vs. panels c,d).

Similarly, the presence of a high concentration of BSA (40 mg/mL, approximating that present in serum) as a binding sink for TO induced complete release at 37°C, and therefore a decrease in the apparent value of k_in* to near zero (Fig. 5e). DNA was not quite as powerful a trap, giving rise to a greater degree of TO release at each temperature, but in a different manner (example in Fig. 5d). The data was best fit by the addition of a slow zero-order release term (k₀*), as suggested by the linear nature of the observed concentration-vs-time plot at long incubation times.

(v) The release of TO-PEG₈ (cation MW ∼ 753) at room temperature was also characterized by this same two-component process (reversible first-order plus zero-order release), as shown in Figure 5f,g (Figure S20B). At 37°C, the extent of release at equilibrium went from 80% to 90%, and the zero-order contribution to the release profile disappeared (Figure S20B). The apparent release and reassociation constants were very similar at each temperature for different TO-PEG₈ loadings in the Qβ VLP, and the overall release of this PEG conjugate was modestly faster than for TO alone.

(vi) TO-PEG₄₅ (cation MW ∼ 2380) showed no evidence of a zero-order release mechanism, but reached equilibrium after loss of only approximately 30% of the packaged cargo at 37°C (Figure S20B). The diffusion of free TO-PEG₄₅ through the dialysis membrane occurred at a rate approximately equal to the measured value of VLP release by this method, so the reported k_out and k_in* values should be regarded as a lower limit.

Loading and Release of Peptides

While PEG is a simple structure that we used to explore the physical properties of intercalation-based packaging, oligopeptides are a much more interesting class of cargo. We chose as representative structures peptides of 18, 19, and 29 amino acids in length, each including a PraGS tripeptide at the N-terminus (Pra = L-propargylglycine). The first two were taken from the S1 subunit of the SARS-CoV-2 spike protein and the last was excerpted from ovalbumin, chosen for differences in charge and solubility (see Table 3 for sequences and calculated properties). As with the PEG units, the alkyne group on each peptide was used to connect them to the common azidopropyl TO derivative 1 (Scheme S2).

Table 3.

Packaging of TO-peptide molecules. N = average number of molecules bound per particle; EE = entrapment efficiency; LE = loading efficiency; incubation time before sucrose gradient purification = 12 h in all cases; [Qβ] for all experiments = 1.25 mg/mL. Error is standard deviation for at least three independent replicates. The values for pI and charge are calculated predictions, not experimentally determined.

cargo	MW	water soluble?	conc (mM) a	N	EE	LE
TO-Pep18	2462	Yes	0.15	176 ± 29	58%	17%
TO-Pep18			0.50	508 ± 47	50%	42%
TO-Pep19	2538	sparingly	0.15	122 ± 32	40%	12%
TO-Pep19			0.50	432 ± 65	43%	36%
TO-Pep29	3275	Yes	0.15	132 ± 23	40%	15%
TO-Pep29			0.50	346 ± 51	34%	45%

Pep₁₈ = Pra-GSVGGNYNYLYRLFRKS; pI = 10.3, +3.0 at neutral pH

Pep₁₉ = Pra-GSATRFASVYAWNRKRIS; pI = 12.2, +4.0 at neutral pH

Pep₂₉ = Pra-GSGSDGSPLEFISQAVHAAHAEINEAGR; pI = 12.2, −2.8 at neutral pH

Incubation of Qβ particles with TO-Pep_n molecules at concentrations up to 1000 times molar excess with respect to VLP for 12 h followed by sucrose gradient purification provided particles loaded with high efficiency regardless of the properties of the cargo (Table 3; EE = 34–58%; LE = approx. 40% using 0.5 mM peptide, the highest concentration tested in the loading step). SEC and TEM analyses showed particle sizes and shapes to be similar to native Qβ particles (Fig. S14). The similar entrainment of both positively- and negatively-charged peptides supports the assumption that the TO-RNA interaction is the main driver of packaging, in contrast to most other reported cases of VLP loading which take advantage of charge complementarity between packaged nucleic acid and polycationic cargo.^[20–25] Pep19 was efficiently packaged in spite of its hydrophobic (and therefore poorly water-soluble) nature.

The in vitro release of these peptides from VLPs was studied by dialysis of freshly-loaded particles as described above, showing approximately 90% release over 100–150 hours. However, the positively-charged Pep₁₈ and Pep₁₉ conjugates exhibited the type of dual (first-order and zero-order) release profiles described above for TO and TO-PEG₈ under certain conditions, whereas the larger, negatively-charged, TO-Pep₂₉ molecule was released in a purely first-order reversible process (Figure 5i-k). In all three cases, the reversible first-order treatment identified an apparent re-binding constant k_in*, but this was much larger for the positively-charged Pep₁₈ and Pep₁₉ than for the negatively-charged Pep₂₉. In the presence of RNA-depleted capsids, no zero-order behavior or re-binding term was observed (Fig. S23), and the k_out values for peptide translocation of the dialysis membrane were modestly greater than in the VLP-packaged experiment.

We postulate that the zero-order release term (observed for TO release to DNA, TO-PEG₈ at room temperature, and TO-peptides at 37°C) derives from nonspecific adhesion or aggregation of cargo in or on the particle that provides a reservoir of material that bleeds into solution at a constant rate. For the small molecules (TO and TO-PEG₈), these are low-avidity interactions, disrupted by an increase in temperature to 37°C. However, for peptides, this component seems to be correlated to the molecular charge. Positively-charged TO-functionalized peptides thus appear to interact with VLP-packaged RNA in two ways, by thiazole intercalation and by charge complementarity. These interactions differ in their molecular nature (single-point intercalation vs. coulombic attraction or ion pairing over a long oligomer-polymer surface) and therefore may be expected to exhibit different kinetics of release. Indeed, the positively-charged peptides show a rebinding term to the particle that far exceeds that of the negatively-charged example.

Conclusion

Intercalating dyes like thiazole orange are commonly used to stain nucleic acids or to bring functional components to nucleic acids for the purposes of labeling or cleavage.^[40–44] Here we combine this property with two natural features of virus-like particles derived from the Leviviridae class of RNA bacteriophages: (1) the substantial amounts of RNA entrained by these particles during normal recombinant expression and self-assembly and (2) the pores that exist at the six-fold symmetry axes of the icosahedral capsid shell. By attaching thiazole orange to molecules that can diffuse or thread through the capsid, we created structures that noncovalently anchor in the packaged RNA, making it easy to both load the nanoparticles with molecular cargo and to monitor their encapsidation and release. About the same TO as RNA?

Both the kinetics and thermodynamics of binding to particle-entrained RNA were found to depend on the length of the linear PEG conjugates used. Such behavior is consistent with linear diffusion (reptation) of long chains through the pores, a process that has an entropic cost well understood in polymer science and applicable here. Long PEG chains were found to remain dangling through the pores, exposing the other ends to the bulk solution. Importantly, oligopeptides, more interesting linear molecules than PEG, were also efficiently encapsidated by the same TO-anchoring strategy. For peptides of 18–29 amino acids, the maximum loading was approximately 500 peptides per particle, comprising up to 45% of the capsid mass. To our knowledge, the operationally simple method of loading peptides into intact virus-like particles demonstrated here is unique. While there must be limitations to the size of polypeptide that can diffuse through virus capsids (for example, VLP-encapsidated molecules are protected from the action of external enzymes), we have not yet explored enough examples to know what those limits are for TO-driven packaging.

While the loading of thiazole-functionalized PEG and peptide molecules reached equilibrium within seconds, the diffusional release of these cargos was much slower, presumably because of the very high local concentration of RNA inside the particle. Because entrainment of all the cargo molecules studied here relies on the same molecular interaction, the kinetics and thermodynamics of release were fairly similar: cargo ranging from a few hundred to thousands of daltons in molecular mass were all released within a time frame of approximately 20–120 hours, a difference that can be functionally important but does not span a large range of rate constants. The thermodynamics of the release process were predictably inversely dependent on molecular size, highly sensitive to temperature (consistent with a large entropic difference in the packaged vs. released states), and responsive to the presence of a competitive binding agent outside the capsid.

Overall, intercalative binding to naturally-packaged RNA was shown here to be a convenient way to encapsulate a remarkably large amount of material inside an easy-to-produce protein nanoparticle. While thiazole orange derivatives have shown moderate toxicity toward a cultured macrophage cell line,^[45] we and others^{[46, 47]} use it as a safer alternative to ethidium to visualize nucleic acids and nucleic acid containing material. The toxicity of TO conjugates, as well as structure-activity relationships in the use of intercalation for packaging and release, remain to be further explored.