1. ¹Institute for Virology, University of Mainz, Mainz, Germany
2. ²Department of Medical Virology, University of Tübingen, Tübingen, Germany

Correspondence: B Plachter, Institut für Virologie, Johannes Gutenberg-Universität Mainz, Obere Zahlbacher Str. 67, 55101 Mainz, Germany.

Received 17 April 2002; Accepted 1 August 2002.

Abstract

Direct protein delivery is an emerging technology in vaccine development and gene therapy. We could previously show that subviral dense bodies (DB) of human cytomegalovirus (HCMV), a beta-herpesvirus, transport viral proteins into target cells by membrane fusion. Thus these non-infectious particles provide a candidate delivery system for the prophylactic and therapeutic application of proteins. Here we provide proof of principle that DB can be modified genetically. A 55 kDa fusion protein consisting of the green fluorescent protein and the neomycin phosphotransferase could be packed in and delivered into cells by recombinant DB in a functional fashion. Furthermore, transfer of protein into fibroblasts and dendritic cells by DB was efficient, leading to exogenous loading of the MHC-class I antigen presentation pathway. Thus, DB may be a promising basis for the development of novel vaccine strategies and therapeutics based on recombinant polypeptides.

Direct transfer of biologically active proteins into cells is an emerging technology both for basic research as well as for the development of therapeutics and vaccines. For therapeutic applications that do not require sustained and regulated transgene expression, potential side effects of gene therapy might be circumvented by direct delivery of the gene products. In addition, as cross-priming has been identified as a major mechanism for the induction of protective cellular immune responses against viral infections,^1,2 direct targeting of professional antigen presenting cells (pAPC) by particulate protein vaccines has become an attractive alternative to other immunization strategies.

Here we describe the delivery of heterologous antigens by subviral particles derived from the human cytomegalovirus (HCMV), termed dense bodies (DB). These large, non-infectious structures are one of three forms of enveloped particles released by HCMV-infected culture fibroblasts.³ The DB, which are surrounded by a viral envelope, can easily be purified from culture supernatant using gradient centrifugation.³ Their internal structure mainly consists of the tegument protein pp65 (pUL83), while lacking viral capsids or DNA. DB enter human and non-human cells by membrane fusion, leading to delivery of their proteinaceous content into the cytoplasm.^4,5,6 We have previously shown that DB elicit strong humoral and cellular immune responses in mice.⁵ Thus, by virtue of their biophysical properties, DB can be considered a potentially rewarding system for the development of both vaccines and therapeutics. However, for both applications, it is mandatory to provide methodology for the modification of the protein content of these particles.

The goal of the work described here was (i) to provide evidence that DB can be altered in their protein content to harbor additional proteins and (ii) to demonstrate that DB can deliver their proteinaceous content in a physiologically active form into eukaryotic cells.

The first set of experiments was designed to prove that additional proteins can be packaged into DB. For this, a DNA construct encoding the green fluorescent protein from Aequoria victoria (GFP), fused in-frame to the gene of the bacterial neomycin-phosphotransferase II (Neo) gene was used (Figure 1a). This construct was inserted into the carboxiterminus of the pp65 (pUL83) to result in a pp65–GFP–Neo fusion protein, driven by the autologous pp65 promoter as shown by plasmid transfection and subsequent superinfection with a pp65-deletion mutant of HCMV (data not shown).

Figure 1.

Generation of viral recombinant RVArDB expressing a pp65–GFP–Neo fusion protein. (a) Schematic representation of the strategy used for the generation of RVArDB. A SpeI fragment from cosmid AD6 (a kind gift of C. Sinzger, Tübingen, Germany) was inserted into the SpeI site of vector pbluescript, resulting in pAD6-SpeI. This plasmid contained 12 kB of genomic sequences of HCMV strain Ad169 from nucleotides 115926 to 127969.²⁹ The GFP–Neo expression cassette was amplified by PCR using the vector pQBI-PGK (Quantum Biotechnologies Inc., Quebec, Canada) as template. As no information about the requirements and mechanisms of DB-formation were available, the GFP-Neo expression cassette was inserted in-frame in the 3'-portion of the pp65-gene, using a SrfI site of plasmid pAD6-SpeI. This cloning provided plasmid pAD6-GFPNeoI. Sequence analyses confirmed that all elements had been cloned in-frame. pAD6-GFPNeoI was subsequently used to generate RVArDB by homologous recombination with strain Ad169 in HFF under G418 selection. For this, the plasmid was transfected into human foreskin fibroblasts (HFF) using the Fugene-TM6 transfection reagent (Roche, Mannheim, Germany). The cells were subsequently superinfected with HCMV strain Ad169. Cultures were supplemented with 200 g/ml G418 to enrich for viral recombinants. Infectious supernatant was passaged in G418 containing medium as described before.³⁰ (b) Direct fluorescence analysis of RVArDB-infected fibroblasts at different time points after infection. The pp65–GFP–Neo fusion protein was located in the cytoplasm throughout the replicative cycle. (c) Immunoblot analysis of infected cell lysates and extracellular particles of RVArDB and Ad169. Extracellular particles were purified by glycerol-tartrate gradient ultracentrifugation from the infectious supernatant of infected HFF. The preparations used for the analyses are indicated on top of the panels. The location of the pp65–GFP–Neo fusion protein is indicated by an arrow to the left. The antibodies used for detection are indicated at the bottom. Molecular mass standards are given in the central part of the figure.

Full figure and legend (208K)

The fusion cassette was inserted into the viral genome by homologous recombination, thereby replacing the wild-type pp65 gene. Recombinant viruses, termed RVArDB, were repeatedly plaque purified. However, as determined by PCR analysis, residual wild-type virus was still present. Yet, as wild-type pp65 was not considered to interfere with the outcome of further experiments, this stock of RVArDB was used.

The intracellular localization of the pp65–GFP–Neo fusion protein was visualized at different time points after infection using direct fluorescence analysis (Figure 1b) as well as indirect immunofluorescence staining (data not shown). In infected cells, pp65–GFP–Neo was mainly localized in the cytoplasm throughout the entire replicative cycle, consistent with the destruction of the major nuclear localization signal of pp65 by the insertion of the GFP-Neo coding cassette.⁴ Expression was first seen 3 days p.i. and increased until 6 days p.i.

Since important steps of the assembly process of herpesviruses occur in the nucleus or at the nuclear membrane, we next addressed the question whether the cytoplasmic retention of the fusion protein allowed its packaging into recombinant viral particles. Virions and DB were purified from culture supernatant infected with RVArDB or the parental strain Ad169 using glycerol-tartrate gradient centrifugation and were subsequently analyzed by immunoblotting.³ Using monoclonal antibody 65-33 directed against pp65 (a kind gift of W. Britt, UAB, Birmingham, AL, USA), a band of the expected molecular mass of 120 kDa could be detected in preparations of recombinant virions and DB, but not in wild-type particles (Figure 1c). Using an antibody directed against Neo (5'-3' Prime, Boulder, CO, USA), the fusion protein could be detected in recombinant particles but not in wild-type particles (Figure 1c). The 120 kDa band was also detectable in HFF infected with recombinant virus RVArDB but not in Ad169-infected HFF (Figure 1c). These experiments demonstrated that the morphogenesis of viral particles was unimpaired by the fusion of the additional 55 kDa polypeptide to pp65, rendering the protein almost twice its natural size.

In a second set of experiments, we wished to address the question of efficiency and functionality of protein transfer by recombinant DB. Initially, primary human fibroblasts were incubated with gradient-purified recDB. Using direct fluorescence analysis, no signal was obtained in cells exposed with recDB carrying the pp65–GFP–Neo protein (data not shown). Since the immunoblot analyses had shown that the fusion protein was present in recDB (Figure 1c), the recombinant protein was either not transferred in significant amounts into cells or the autofluorescent properties of GFP were not retained during packaging. The latter explanation was considered possible, as light emission of GFP depends on the accurate folding of the protein, which might have been compromised by fusing both C- and N-terminus to a heterologous partner, ie pp65 and Neo, and subsequent packaging into particles. To test this, fibroblasts were exposed to recDB and wtDB and stained with an antibody specific for pp65, using indirect immunofluorescence analysis. A representative section of each slide in two different magnifications is shown in Figure 2a. In both cases, almost all cells showed specific staining. Using wtDB, pp65 can be seen exclusively in the nucleus as a consequence of its rapid nuclear import mediated by two nuclear import signals.⁴ In contrast, the pp65–GFP–Neo protein is distributed between the cytoplasm and the nucleus, following the destruction of one of the major NLS of pp65 in the very C-terminus of the protein during cloning of the expression cassette (Figure 1a). These experiments showed that transfer of the fusion protein by recDB is efficient in terms of the number of cells targeted.

Figure 2.

Analysis of the transfer efficiency of the pp65–GFP–Neo fusion protein into fibroblasts. (a) Efficiency of protein delivery into fibroblasts by wtDB and recDB using indirect immunofluorescence analysis. HFF were incubated for 16 h with DB from parental strain Ad169 or from recombinant virus RVArDB. Cells were fixed with methanol for 5 min and incubated with pp65-specific monoclonal antibody 65-33 (kind gift of W. Britt, UAB, Birmingham, AL, USA), followed by incubation with a secondary FITC-coupled anti-mouse IgG conjugate. Micrographs were taken at a magnification of 40 fold. The upper panels provide an overview of the number of positive cells, the lower panels show the differences in intracellular location of wild-type pp65 in comparison to the pp65–GFP–Neo fusion protein in a 100-fold magnification. (b) Verification of Neo activity provided by the expression of the pp65–GFP–Neo fusion protein in fibroblasts infected with RVArDB using in vitro phosphorylation of kanamycin and subsequent thin layer chromatography. Cells were infected with RVArDB or, for control, with the parental strain Ad169 (negative control) or the pp65 deletion mutant RVAd65 (positive control). Cells were lysed 4 days after infection and assayed for Neo activity in the presence of [-³²P]ATP using kanamycin as a substrate. Phosphorylated kanamycin was identified by thin layer chromatography according to published procedures.⁷ Except for the 3rd lane, all assays were performed in the presence of kanamycin. (c) Analysis of the Neo activity in the lysates of cells incubated with purified extracellular particles. HFF were incubated for 16 h with the indicated particle types and were subsequently lysed. In vitro phosphorylation of kanamycin and thin layer chromatography were performed as described in section (b). The position of phosphorylated kanamycin is indicated by an arrow.

Full figure and legend (383K)

To analyze whether these particles transported the heterologous protein in biologically active form, the functionality of Neo after transfer into cells was analyzed. To first test whether the kinase was functional in the context of the pp65–GFP–Neo fusion protein, fibroblasts were infected with RVArDB. Cell lysates obtained 4 days after infection were assayed for Neo activity in the presence of [-³²P]ATP using kanamycin as a substrate. Phosphorylated kanamycin was identified by thin layer chromatography according to published procedures.⁷ Cells infected with RVArDB expressed Neo activity consistent with and exceeding the activity of the positive control RVAd65 ⁴ (Figure 2b). Subsequently recombinant viral particles from infected cell culture supernatant were purified and tested for their ability to transfer functional Neo into fibroblasts. Cells were incubated with DB or virions from RVArDB or Ad169 overnight in the presence of cycloheximide to prevent de novo protein synthesis. Inhibition of viral protein synthesis was verified by the lack of detectability of the viral immediate–early protein expression using indirect immunofluorescence analysis (data not shown). After 16 h, cells were collected and lysed. The cleared extracts of cells that had been incubated with recombinant DB or recombinant virions showed distinct Neo activity (Figure 2c). In contrast, lysates of cells incubated with wild-type particles did not contain Neo activity. These results proved that recDB transferred a heterologous protein in functionally active form into target fibroblast cells and that this transfer was efficient.

In a next step we wished to address the question, whether DB could transfer protein into antigen presenting cells and whether cells could be sensitized to CTL-mediated immune responses by exogenous loading of the major histocompatibility complex class I (MHC-class I) presentation pathway. Primary DC cultures were established from the PBMC of blood donors in the presence of granulocyte macrophage colony stimulating factor (GM-CSF) and Interleukin-4 (IL-4) and were exposed overnight with DB. Uptake of the viral tegument protein pp65 was visualized by staining with a specific antibody using indirect peroxidase staining (Figure 3). The pp65 could be detected in the cytoplasm of DC after exposure with increasing amounts of DB. In contrast, infection with the endothelial cell-propagated strain TB40/E of HCMV, which is able to replicate in DC,⁸ showed primary nuclear localization of pp65, indicating that nuclear import was impaired after exogenous uptake of the protein by DB. The reason for this lack of transport of pp65 in DC is unknown, but was also seen in other pAPC exposed to DB.⁹ These experiments showed that DC, being an important pAPC, were loaded exogenously with pp65 by DB. To prove that this loading was immunologically relevant, ie that cells were sensitized to become CTL targets after loading, ELISPOT- and chromium release analyses were performed. Since for both the neomycin transferase and the green fluorescent protein used as markers in our study, no epitopes for the cytolytic T-lymphocyte response were known, we used a well-defined CTL-target epitope from the pp65 protein of HCMV for testing. Stably growing CTL clones were established in CD8/HLA-A2 double transgenic mice, following established procedures,^10,11 against the peptide epitope pp65_495-503, known to be presented by HLA-A2.^12,13 These clones were subsequently used to analyze the presentation of a defined A2-presented peptide of pp65 after transfer of this tegument protein by DB into both fibroblasts and DC.

Figure 3.

Visualization of DB uptake by dendritic cells. DC were generated according to published procedures.³¹ 110⁶ cells were incubated with the indicated concentration of DB for 2 h at 37°C. Subsequently, culture medium was added and incubation was continued for another 24 h. After that, cells were washed with PBS/2 mM EDTA and were centrifuged onto glass coverslips using a cytocentrifuge. After air-drying, cells were fixed in 1% paraformaldehyde for 10 min at room temperature and permeablized in PBS/0.5 % Igepal/10% Sucrose/1% FCS for 5 min at room temperature. Staining of pp65 transferred by DB was performed by indirect immunoperoxidase technique using antibody 28-77 (a kind gift of W. Britt, UAB, Birmingham, AL, USA). Non-specific binding was blocked with rabbit serum for 20 min at room temperature followed by incubating the cells with cell culture supernatant of the antibody for 30 min at 37°C. Binding was visualized using a peroxidase-conjugated secondary antibody (goat anti-mouse IgG, DeBeer Medicals, Diessen, Netherlands) and diaminobenzidine as chromogen. The slides were washed in PBS for 5 min after each incubation step. Nuclei were counterstained with hematoxylin. For negative and positive control, cells were either mock-infected or infected with TB40E for 24 h prior to staining. A positive result consisted of brown nuclear or cytoplasmic staining.

Full figure and legend (197K)

Transfer of pp65 into fibroblasts, which are the primary in vitro target cells of HCMV infection, proved to be surprisingly efficient in terms of exogenous loading of the MHC-class I pathway. Using ELISPOT analysis, we could show that DB-loaded target cells were presenting the pp65 test peptide as efficiently as infected or peptide-loaded cells to pp65-specific CTLL (Figure 4a). This presentation was shown to be functionally relevant, as loaded cells were killed very efficiently when tested in chromium release assays (Figure 4c). Presentation of pp65_495-503 was also seen when DC as pAPC were used as target cells. In this case, about 30% of the DB-treated DC could present pp65_495-503 in a way sufficient to induce IFN- secretion by specific CTLL (Figure 4b). Although the number was lower than in fibroblasts, presentation after DB loading of DC was still as effective as the internal control, namely the infection with TB40/E strain of HCMV. These experiments showed that DB can transfer antigens into both pAPC and non-professional APC efficiently, verifying that these particles are a promising basis for vaccine development.

Figure 4.

Recognition of DB-treated target cells by specific CTL. (a) ELISPOT analysis of HFF, incubated with DB. HLA-A0201 positive MRC-5 fibroblasts (110⁶) (ATCC CCL-171) were either mock treated, incubated for 16 h with 20 g DB, infected with HCMV strain Ad169 over night at a moi of 10, or labeled with synthetic peptide pp65_495-503 at 10^-8 M, respectively, and used as target cells in Interferon- (IFN-) specific ELISPOT analyses. 110⁵ Target cells were seeded per well of a microtiter plate and incubated with 100 effector cells for 20 h. As effectors, a murine polyclonal CTL line (CTLL) specific for the pp65-derived peptide epitope pp65_495-503 presented by human HLA-A0201 (pp65-CTLL) was used. CTLL were generated by immunizing HLA-A2/CD8 double-transgenic mice (a kind gift of L. Sherman, The Scripps Institute, LaJolla, USA) with DB resuspended in incomplete Freund's adjuvant, followed by spleenectonomy 10 days later, similar to procedures described previously.^5,10,11 Splenic cells were restimulated weekly with irradiated, pp65_495-503-loaded feeder cells and TCGF according to established protocols.^10,11 IFN--based ELISPOT analyses were carried out as detailed by others.³² Values given represent the mean number of spots of triplicate wells and the standard deviation. (b) ELISPOT analysis of dendritic cells (DC), incubated with DB. HLA-A0201-positive DC were generated from PBMC of HCMV seronegative donors by adherence to plastic and stimulation with GM-CSF and IL-4 as described previously.^33,34 DC (110⁶) were either mock treated, incubated for 16 h with 20 g DB, infected with HCMV strain TB40/E overnight at a moi of 10, or labeled with synthetic peptide pp65_495-503 at 10^-8 M, respectively, and used as target cells in IFN- specific ELISPOT analyses. pp65-CTLL were used as effector cells. (c) CTL-mediated cytolysis of DB-treated target cells. HLA-A0201-positive MRC-5 cells were either mock treated (MRC5), incubated for 16 h with 20 g DB, infected with HCMV strain Ad169 overnight at a moi of 10, or labeled with synthetic peptide pp65_495-503 at 10^-8 M, respectively, and used as targets in standard 5 h chromium-release assays at the indicated effector/target ratios. A constant number of 10^{3 51}chromium-labeled target cells and graded numbers of effector cells were used in the assay. Throughout, reported cytolytic activities represent the mean percent specific lysis from three replicate microcultures.

Full figure and legend (127K)

The primary goal of prophylactic vaccination against infectious agents is the prevention of infection or disease. In contrast, therapeutic vaccine approaches focus on the induction of immune effector mechanisms targeted against antigens, eg against tumor antigens, expressed in the course of disease. Common to both strategies is the lack of requirement for sustained antigen expression. In addition, in some therapeutic settings, permanent expression of a transgene is not necessary or even undesired. Thus direct protein translocation into cells has been favored for some clinical applications. Several systems for direct protein transduction across the plasma membrane have been described relying on carrier peptides like the antennapedia peptide,^14,15 fusions with VP22 of herpes simplex virus,¹⁶ the HIV tat protein^17,18,19,20 or bacterial toxins.²¹ Major drawbacks of these methods are the requirement for protein purification prior to transduction, possible interference of peptides carriers, like, eg an HIV-TAT-derived peptide, with cellular gene expression or inadequate subcellular targeting of the transduced protein (reviewed in ref. 22).

DB of HCMV may provide an alternative to these techniques. The most important properties of DB are their particulate nature and their ability to deliver functional protein across the membrane of eukaryotic cells, including pAPC. This renders them attractive especially for the design of recombinant prophylactic or therapeutic vaccines. In a mouse model, DB have been shown to be highly immunogenic.⁵ In contrast to virus-like particles from small animal viruses, used for vaccine purposes, DB are relatively large structures. Consequently, they have a high packaging capacity of protein molecules per particle. As shown here, pAPC like DC can be loaded with these particles and the protein delivered can be detected with methods like immunocytochemistry, which have a limited inherent assay sensitivity. Thus, considerable amounts of proteins must be delivered by DB to pAPC. These pAPC presented a test peptide to specific CTL, verifying that protein can be translocated into cells in a way to be presented on MHC restriction elements. Moreover, we could show that the protein was delivered into target cells in a physiologically active form. Most of the fibroblasts tested were hit by recombinant particles and the enzymatic activity of Neo was retained after transfer. In addition, exogenous loading of the MHC-class I presentation pathway proved to be highly efficient in these cells.

We have not determined the upper size limit of a heterologous polypeptide that still can be packaged in recDB. The fusion protein carried in the construct tested here comprised 55 kDa, which is less than the maximal size of proteins reported to be transduced by fusogenic peptides.²³ Therefore, with respect to the size of the proteins that are transferred into cells, DB do not appear to be superior to transducing peptides. However, more extensive studies comparing the two methods using identical protein constructs and similar amounts of test materials are required to approach this issue appropriately.

DB can be purified from supernatants of cell cultures, infected with wild-type HCMV or with recombinant HCMV using standard gradient purification. Residual infectivity can easily be removed, eg by UV irradiation. Extensive protein purification is not required. The particles are stable for several years when stored at –70°C (unpublished observation). For pp65 itself, no toxic regulatory functions have been described, rendering this protein a safe carrier molecule. Furthermore, DB, contained in vaccine preparations, proved to be safe in humans.^24,25,26s

These properties of DB render them a rewarding candidate for the development of novel recombinant vaccine strategies. Such recDB may, however, also be suitable for application of pharmacologically active polypeptide molecules. Inclusion of internal protease recognition sequences such as ubiquitin^27,28 may help to release the protein of interest from pp65.

Taken together, these recombinant particles are the basis for the design of delivery systems for an array of different purposes of in vitro and in vivo targeting of functional polypeptides including intracellular delivery of chemotherapeutic agents, ex vivo expansions of stem cells through introduction of stimulatory molecules or the efficient antigen loading of APC, eg for vaccination. Additional studies are under way to refine strategies for the construction of recDB and to compare the efficiency of the DB system with other application methods.