Exploiting the Yeast L-A Viral Capsid for the In Vivo Assembly of Chimeric VLPs as Platform in Vaccine Development and Foreign Protein Expression

A novel expression system based on engineered variants of the yeast (Saccharomyces cerevisiae) dsRNA virus L-A was developed allowing the in vivo assembly of chimeric virus-like particles (VLPs) as a unique platform for a wide range of applications. We show that polypeptides fused to the viral capsid protein Gag self-assemble into isometric VLP chimeras carrying their cargo inside the capsid, thereby not only effectively preventing proteolytic degradation in the host cell cytosol, but also allowing the expression of a per se cytotoxic protein. Carboxyterminal extension of Gag by T cell epitopes from human cytomegalovirus pp65 resulted in the formation of hybrid VLPs that strongly activated antigen-specific CD8+ memory T cells ex vivo. Besides being a carrier for polypeptides inducing antigen-specific immune responses in vivo, VLP chimeras were also shown to be effective in the expression and purification of (i) a heterologous model protein (GFP), (ii) a per se toxic protein (K28 α-subunit), and (iii) a particle-associated and fully recyclable biotechnologically relevant enzyme (esterase A). Thus, yeast viral Gag represents a unique platform for the in vivo assembly of chimeric VLPs, equally attractive and useful in vaccine development and recombinant protein production.


INTRODUCTION
Viral expression systems can be classified into three types based on the regulatory and/or structural viral component that drives protein expression: (i) plasmid-based vectors containing promoter elements from either pro-or eukaryotic viruses; (ii) infectious viral vectors in which the gene of interest is integrated into the viral genome and expressed from a viral promoter in an appropriate host; (iii) virus-like particles (VLPs), also called pseudovirions, representing subunit structures composed of multiple copies of a viral capsid and/or envelope protein capable to self-assemble into VLPs of defined spherical symmetry in vivo [1][2][3]. Currently, VLPs composed of a structural protein are often used as particulate antigen in the design of prototype vaccines as they possess several advantages over conventional monomeric protein immunogens [4]. Firstly, most VLPs can be produced in large quantity in a heterologous host. Secondly, due to their particle structure and high molecular weight, VLPs can be easily purified in a preparative scale. Thirdly, a number of particle forming proteins tolerate insertion of foreign amino acid sequences without affecting in vivo self-assembly competence. Such chimeric or hybrid VLPs, exploited as platform for the display of antigenic determinants in a polyvalent manner, have already been shown to be promising candidates in the development of various subunit vaccines [5].
Here, a novel expression system based on the non-infectious yeast (S. cerevisiae) dsRNA virus L-A was designed. This mycovirus represents an autonomously replicating, encapsidated dsRNA element that stably persists in the cytoplasm of an infected yeast cell without conferring a recognizable phenotype upon its host [6]. As member of the Totiviridae family, L-A contains a linear nonsegmented dsRNA genome (4.6 kb) comprising two overlapping ORFs, gag and pol. While gag encodes the major capsid protein Gag (76 kDa), pol specifies a multifunctional RDRP which is in vivo expressed as a 171 kDa Gag/Pol fusion protein by a [21] ribosomal frame-shift event [6,7]. As Gag has been shown to be sufficient to drive in vivo self-assembly into VLPs, Pol is dispensable for viral coat assembly [8]. However, N-acetylation of Gag (catalyzed by Mak3p of the host cell) is an essential prerequisite for VLP formation in vivo [9]. The 40 nm L-A capsid has a 120subunit structure composed of 118 Gag proteins and two copies of Gag/Pol configured as an icosahedron of triangulation class T = 1 [7,[10][11][12]. In the present study, we used Gag -and specifically designed variants thereof -for the in vivo assembly of VLP chimeras suitable for heterologous protein production and display of vaccine-relevant immunogens.

Chimeric Gag assembles into yeast VLPs
Since in the natural L-A virus, Pol (as C-terminal part of Gag/Pol) extends into the interior of the capsid to ensure replication and transcription of the viral genome [11], we replaced Pol by a truncated version of the immunodominant phosphoprotein pp65 from human cytomegalovirus (HCMV) to modify the inner surface of the capsid. The truncated protein (Dpp65) comprised the Cterminal amino acids 358-561 of pp65 flanked by the CD8 + T-cell epitopes AE44 and AE45 [13] at its N-and C-terminus, respectively. The resulting Gag/Dpp65 protein fusion (101 kDa) as well as non-modified (naked) Dpp65 (24.9 kDa) were separately expressed in yeast and analyzed for expression level and protein stability. In the Gag/Dpp65 protein fusion, Dpp65 is fused in the [0]-frame to the 39-end of gag resulting in a protein fusion that is ought to self-assemble (via its Gag domain) into VLPs encapsulating Dpp65 as C-terminal cargo ( Figure 1A). Western analysis of cell extracts from yeast expressing either naked Dpp65 or Gag/ Dpp65 revealed only a weak signal for non-fused Dpp65 in contrast to an intense signal seen in cells expressing Gag/Dpp65 ( Figure 1B). The observed instability of the naturally short-lived Dpp65 protein in the multiple protease-deficient mutant strain S86c could not even be prevented in mutant hosts defective in components of the ubiquitin-proteasome-system (UPS) nor in a yeast Dpep4 mutant devoid of vacuolar proteases (data not shown). Interestingly however, Dpp65 was significantly stabilized and effectively protected from proteolytic degradation when expressed in a particulate manner as C-terminal protein fusion to Gag ( Figure 1B). The competence of Gag/Dpp65 for in vivo selfassembly into hybrid VLPs was demonstrated by analyzing its sedimentation profile during sucrose gradient centrifugation and by electron microscopy of gradient-purified VLPs: Gag/Dpp65 formed isometric particles which showed a similar sedimentation behaviour as natural L-A virions (Figure 2A and 2B).

Gag/Dpp65 activates HCMV-specific CD8 + memory T-cells ex vivo
Antigenicity of recombinant Gag/Dpp65 particles was determined in an ex vivo stimulation assay allowing quantification of HCMVspecific memory T cell responses in human whole blood by FACS analysis [14,15]. For such an assay, both VLPs (Gag and Gag/ Dpp65) were expressed and assembled in yeast and partially purified by centrifugation through a sucrose cushion ( Figure 3A). T cell stimulation assays performed on whole blood of a HCMV seropositive donor indicated that pp65-specific CD4 + and CD8 + T-lymphocytes were strongly activated to maximal frequencies of 2.64% and 0.22% by HCMV positive control antigens, while no immune response was seen in the negative control. Interestingly, cushion-purified Gag/Dpp65 as well as non-modified Gag only poorly activated CD4 + cells (,0.1%, Figure 3B), while chimeric Gag/Dpp65 particles induced a pronounced CD8 + T-cell response in a dose-dependent manner that was even higher than in the positive control (0.35% versus 0.22%; Figure 3C). In contrast to Gag/Dpp65, unmodified Gag did not significantly activate HCMVspecific CD8 + cells, not even at the highest concentration tested ( Figure 3C). To demonstrate that the observed CD8 + T cell response was caused by the Dpp65 moiety of the chimeric particles, recombinant VLPs were isolated from yeast, purified by sucrose gradient centrifugation and subsequently analyzed by SDS-PAGE and Coomassie-Blue staining. As shown in Figure 4A, gradientpurified VLPs only contained two protein species representing Gag and Gag/Dpp65, thereby demonstrating that both preparations were of high purity (.95%). For T cell stimulation, whole blood of HCMV-seropositive donors was supplemented with 5 mg gradientpurified VLPs; a lysate from HCMV-infected fibroblasts served as positive control, a HCMV-seronegative blood sample as negative control to demonstrate antigen specificity of the immune response. As expected, no T cell response was detectable in the negative control, while a significant CD4 + T lymphocyte response was seen in seropositive samples against the positive control antigen ( Figure 4B and 4C). In contrast to unmodified Gag, chimeric Gag/Dpp65-VLPs caused a significant activation of CD4 + T cells only in donor 2 ( Figure 4B). As shown before for cushion-purified particles, gradient-purified Gag/Dpp65 showed significantly elevated frequencies of activated CD8 + T cells that were up to 25-fold increased over unmodified Gag ( Figure 4C). Qualitatively the same result was obtained by analyzing chimeric VLPs in which the antigenic Dpp65 moiety was expressed on the outer VLP surface by in-frame insertion into surface-exposed loops of Gag immediately upstream of amino acid position S182 and flanked by flexible spacers (Powilleit and Schmitt, unpublished). These data demonstrate that Gag/Dpp65 expressed and assembled into yeast VLPs exposing Dpp65 either inside the particle or at the outer VLP surface, both possess antigenic properties (in particular to activate CD8 + memory T cells) that are due to their HCMV-specific Dpp65 moiety. To further investigate the potential of Gag/Dpp65 particles as unique yeast vaccine, we are currently analyzing native Gag/ Dpp65 particles in a murine HCMV model of HLA transgenic mice for the induction of a protecting in vivo immune response. That antigens exposed inside chimeric yeast VLPs are indeed efficiently processed by immune cells in vivo resulting in a humoral immune response was demonstrated by using native Gag/K28a VLPs (assembled in and isolated from a GTXa expressing yeast strain) which induced K28a-specific antibodies in rabbit ( Figure 5A and  ). In addition, since the in vivo expression of the a-subunit of the viral a/b toxin K28 is known to be toxic (in particular when expressed in the ER lumen [16]), successful expression of Gag/ K28a particles demonstrates that chimeric Gag-VLPs are also suitable for the expression of a per se lethal protein.

Chimeric VLPs as platform for protein expression and purification
In contrast to monomeric protein fusions, hybrid VLPs are of high molecular weight and can be easily prepared from crude cell extracts by ultracentrifugation [4]. In this context, we exploited Gag as particle-forming carrier to express and purify the green fluorescent protein GFP as model polypeptide. For this purpose, a Gag variant was constructed encoding a 105 kDa protein fusion containing Gag at its N-terminus (to ensure in vivo VLP assembly and GFP encapsulation), followed by an 11 amino acid T7 epitope tag (for immunological detection) and a factor X a cleavage site to release mature GFP from Gag ( Figure 6A). Yeast transformants expressing such a construct (GTXG) were used for VLP preparation, and western analysis of sucrose gradient fractions revealed that GTXG (105 kDa) assembled into VLPs that showed   a sedimentation profile portraying that of natural yeast VLPs ( Figure 6B). To check whether the protein fusion is accessible to factor X a cleavage and subsequent release of its GFP moiety, GTXG-VLPs (288 mg protein/ml) were treated with Triton X-100, thereafter incubated with factor X a and subsequently subjected to SDS-PAGE and western analysis. As shown in Figure 6C, detergent-treated GTXG-VLPs were efficiently processed by factor X a , liberating two protein moieties from the GTXG precursor (both absent in negative controls) whose calculated molecular weights are consistent with the presence of monomeric (26.8 kDa) and dimeric (53.6 kDa) GFP. In direct support, dimeric GFP was only seen in non-reducing SDS-PAGE and completely disappeared under reducing conditions in the presence of b-mercaptoethanol (data not shown). To investigate the efficiency of preparative GFP purification, detergent-treated GTXG particles (274 mg) were incubated in the presence of factor X a , VLP debris was removed by high-spin centrifugation (100.0006g) and the resulting supernatant was treated with Xarrest agarose to eliminate residual endoproteinase. Soluble GFP released from the VLPs was precipitated by the addition of ammonium sulfate, and the resulting pellet and supernatant fraction was subsequently analyzed by SDS-PAGE probed with anti-GFP and anti-T7. By this procedure, monomeric GFP (26.8 kDa) and the larger GTX cleavage fragment (77.8 kDa) were successfully released from the 105 kDa GTXG precursor ( Figure 6D). Furthermore, GFP could be purified in a single step by hydrophobic interaction chromatography (HIC) as judged by SDS-PAGE and western analysis ( Figure 6D). GFP containing fractions from the HIC column were 100-fold concentrated by Amicon ultrafiltration, and the pooled HIC fractions as well as filtrate and retentate after ultrafiltration were analyzed by SDS-PAGE and Coomassie-Blue staining. As expected, GFP concentration in the retentate was higher than in the pooled HIC fractions or in the filtrate ( Figure 6E). Moreover, the preparation was highly pure, only showing two GFP-specific signals on nonreducing SDS gels, a major 26.8 kDa protein representing monomeric GFP and a minor 53.6 kDa species representing dimeric GFP ( Figure 6E). Based on the signal intensity after Coomassie-Blue staining, the overall yield of GTXG-derived GFP after sucrose gradient centrifugation and HIC purification was in the range of 0.2 mg purified protein from 1 liter yeast culture (and a density of 5610 8 cells/ml).

VLP chimeras as recyclable biocatalyst
To demonstrate the flexibility of the viral carrier for the expression of a biotechnologically relevant enzyme, the GFP moiety in GTXG was replaced by the carboxylesterase EstA from Burkholderia gladioli [17], the resulting Gag/EstA protein fusion was expressed in yeast and electron microscopy of sucrose gradient-purified Gag/EstA particles confirmed in vivo assembly into recombinant VLPs ( Figure 7A). To demonstrate esterase activity in the VLP chimeras, gradient-purified Gag/EstA particles were analyzed in an enzyme activity assay using 4-nitrophenylacetate as substrate. Since EstA is located inside the particle, the substrate must pass the capsid pores to be converted into acetate and 4-nitrophenol. Under the assay conditions used, the release of 4-nitrophenol was monitored through its absorption at 405 nm that was shown to be linearly correlated to a concentration of up to 1 mM (regression coefficient = 0.9974; Figure 7B). Based on these parameters, gradient-purified Gag/EstA (70 ng and 280 ng) and non-modified Gag (560 ng) were subsequently analyzed for esterase activity; a VLP-free sample served as negative control to detect autohydrolysis and unspecific breakdown of the ester substrate. In contrast to unmodified VLPs, Gag/EstA chimeras catalyzed the release of 4-nitrophenol and under steady state conditions (reaching reaction equilibrium within 48 min), 67.9% of the initial substrate were enzymatically converted into 4nitrophenol ( Figure 7C).

Gag/EstA particles allow multiple substrate conversions
To investigate whether particle-associated esterase can be recycled and reused in multiple rounds of substrate conversion, gradientpurified Gag/EstA particles (22.5 mg) were used in an enzyme reaction (5 ml), isolated by ultracentrifugation, subsequently subjected to SDS-PAGE and western analysis, and compared to the same VLP charge prior to substrate conversion. Based on the esterase signal intensity obtained after SDS-PAGE and Coomassie-Blue staining, approximately one-third of the original particle preparation had been recovered in the pellet fraction after a single ultracentrifugation step ( Figure 7D). Since Gag/EstA particles were significantly diluted prior to ultracentrifugation, esterase protein remaining in the final supernatant was not detectable in immunoblots. Most interestingly however, catalytic activity of Gag/EstA-VLPs after recovery was not negatively affected and rather resembled EstA activity in the original non-recycled VLPs. Using an equal volume of Gag/EstA particles before and after recycling, absolute esterase activity was three-fold lower in recycled VLPs; however, given that one-third of the initial VLP amount (750 ng) had been recycled, specific esterase activity in the Gag/EstA chimeras before and after recycling was almost identical (Figure 7E), demonstrating that VLP-associated EstA can be recycled and repeatedly used in multiple rounds of enzyme-catalyzed substrate conversion. Most interestingly however, the remarkable efficacy of the VLP-associated enzyme becomes evident when specific esterase activity of Gag/EstA particles (20.8 U mg 21 protein) is compared to that after EstA cell surface expression in either yeast (S. cerevisiae) or bacteria (E. coli): in both cases, esterase activity was significantly lower and ranged from 1.3 to 2.7 U mg 21 protein in yeast [18] and 0.001 to 0.023 U mg 21 protein in bacteria [19].

DISCUSSION
Viral expression systems are not only useful in gene transfer experiments, but also in heterologous protein production. In most cases, structural or regulatory elements of animal and human viruses represent the key elements in these systems, restricting their   application to higher eukaryotic cells as host. In the present study we engineered the yeast totivirus L-A and demonstrated its feasibility for being used as unique expression system in a lower eukaryotic host. The potential of its capsid as platform for the presentation of immunogens was demonstrated by using the HCMV tegument protein pp65 as model antigen. This structural protein represents the major target of cellular immune response during HCMV infection [20,21], and also in vitro HCMV-infected cells are recognized by 70-90% of cytotoxic T lymphocytes (CTLs) [22]. Besides inducing a strong CD8 + T cell response, pp65 can also activate CD4 + T cells [23] making it an ideal candidate in developing an HCMV vaccine ensuring both, humoral and cellular immunity.
To analyze recombinant yeast VLPs for their potential as nonreplicating particle vaccine, we fused an N-terminally truncated pp65 variant (Dpp65) of HCMV containing immunodominant T cell epitopes to the C-terminus of Gag and showed that it selfassembled into VLP chimeras when expressed in the yeast cell cytosol. Electron microscopy revealed a spheric symmetry of the recombinant particles (Gag/Dpp65) in which the Dpp65 moiety was buried inside the capsid, analogous to Pol in the natural L-A virus [11]. The ''in viro'' localization of Dpp65 either inside or outside the particle was judged in two ways: (i) by immunogold labelling and electron microscopy, and (ii) by analyzing cosedimentation profiles of Gag/Dpp65 and monoclonal anti-pp65 in a sucrose density gradient (Powilleit and Schmitt, unpublished). In contrast to non-modified (naked) Dpp65 which was only weakly expressed and subject to proteolytic degradation in vivo, particleassociated Dpp65 was highly stable and effectively protected against the action of host cell proteases. In an ex vivo stimulation assay in which memory T cell stimulation can be quantified and characterized in human whole blood [14], purified Gag/Dpp65 particles -in contrast to non-modified Gag -resulted in a significant activation of CD8 + T lymphocytes, while frequencies of activated CD4 + helper T lymphocytes (HTLs) were generally low. The same holds true for chimeric VLPs in which the Dpp65 moiety was exposed at the outer VLP surface by insertion into surface-exposed loops of Gag immediately N-terminal to position S182 and flanked by flexible spacer elements (Powilleit and Schmitt, unpublished). In all cases the observed bias in activation of CD8 + T lymphocytes by exogenous antigen points to an alternative presentation pathway favouring association of pp65 epitopes with MHC I. In the classical pathway of antigen presentation, peptides derived from exogenous proteins or particles are exposed in complex with MHC II molecules on the cell surface where interaction with complementary T cell receptors leads to an activation of CD4 + HTLs. This mechanism is apparently true for the positive control antigen used in this study, a lysate of HCMV-infected fibroblasts containing both soluble as well as virion-associated pp65. In one out of four blood samples, Gag/Dpp65 also induced a CD4 T cell response, indicating that Dpp65 peptides can also be presented in complex with MHC II. Due to the N-terminal truncation in Dpp65, HTL epitopes in pp65 such as peptides 11, 71 and/or 72 are lacking [23], probably attenuating its ability to induce a more frequent CD4 + T cell response.
In contrast to MHC II, MHC I-associated presentation of peptides is considered to be restricted to endogenously synthesized proteins, initiating with proteasomal processing in the cytoplasm. Upon targeting on the cell surface, the MHC I/peptide complex can activate CD8 + T cells through interaction with the corresponding T cell receptor [24]. More recent studies have indicated that proteins taken up by phagocytosis can also be presented by MHC I molecules, thereby promoting CD8 + CTL proliferation [25]. Such alternative antigen presentation (also known as cross-presentation) has been observed in phagocytic cells upon engulfment of bacterial cells or viral particles [26][27][28][29][30]. Since recombinant Gag/Dpp65-VLPs share the particulate nature with these antigens, a cross-presentation pathway might also exists for them as well as for pp65 associated to intact HCMV virions (present in the positive control).
Current strategies in HCMV vaccine development imply the application of live, attenuated virus strains, DNA vectors coding for immunodominant HCMV proteins and/or genetically modified carrier viruses [31]. Although these approaches might be well-tolerated and immunogenic, they bear the risk of reconverting to original virulence, inducing anti-DNA antibodies or recombining with the host cell genome [32,33]. In addition, the production of most of these vaccines in human cell lines is costly, timeconsuming and hardly suitable to an industrial scale-up [34]. In contrast, yeast is regarded as safe due to its GRAS status and widely accepted as a profitable host to produce biotechnologically and pharmaceutically relevant proteins [35]. As proof of principle for a yeast vaccine based on recombinant VLP chimeras, we intend to use Gag/Dpp65 particles and HLA transgenic mice in a murine HCMV model to analyze in vivo immune responses and to evaluate vaccine potential of chimeric yeast VLPs.
Besides being attractive in vaccine development, the yeast viral expression system described here is also interesting in foreign protein production. This was demonstrated for a gene fusion in which the 39-end of gag was sequentially extended by a T7 epitope, a factor X a cleavage site, and the coding sequence of GFP as model protein. After in vivo expression, the protein fusion selfassembled into hybrid yeast VLPs from which GFP could be entirely released from its Gag carrier by factor X a cleavage. Singlestep purification via hydrophobic-interaction chromatography and subsequent ultrafiltration resulted in a highly pure GFP preparation with an overall yield of 0.2 mg purified and biologically active GFP from 1 liter yeast culture (and a density of 5610 8 cells/ml). The overall yield in heterologous GFP production via chimeric yeast VLP expression falls within the broad-range levels of GFP fusion protein production which has been shown to range over 3 orders of magnitude, from 4 mg/liter to 4 mg/liter yeast cell culture [36,37]. Furthermore and in contrast to expression systems based on yeast Ty retrotransposons exposing foreign proteins at the outer VLP surface [38,39], recombinant Gag particles described here contain their cargo within the inner capsid, thereby effectively preventing proteolytic degradation. Especially for the production of short-lived and unstable proteins (such as pp65 from HCMV), the L-A-derived expression system might be superior as it efficiently protects its cargo from proteolytic degradation in the host cell cytosol. Furthermore, by using the a-subunit of K28 toxin -which is cytotoxic when expressed in yeast [16] -we could demonstrate that recombinant Gag-VLPs are also suitable for the in vivo expression of a protein which is per se toxic. In addition, Gag was also shown to be effective in the expression of a particle-associated and recyclable biotechnical enzyme, carboxylesterase A from B. gladioli [18]. Gag/EstA protein fusions expressed in yeast assembled into VLP chimeras that were catalytically active and effectively converted 4nitrophenylacetate into 4-nitrophenol and acetate. A hallmark of this VLP-based ''bioreactor'' is its reusability in multiple substrate conversions without loss in enzyme activity and its overall yield in particle-associated specific esterase activity, being significantly higher than esterase activity after cell surface display in E. coli or S. cerevisiae [18,19]. In sum, these data demonstrate the efficiency of the yeast L-A viral expression system in the production and purification of recombinant proteins/enzymes in a particleassociated manner, providing substantial yields of a functional protein in sufficient quality without the need of time-consuming purification procedures. In addition, the ease of fermentation in low-cost media makes S. cerevisiae and its chimeric Gag-VLPs attractive for foreign protein production.

MATERIALS AND METHODS
Strains, oligonucleotides and plasmids E. coli strain TOP10 [F 2 mcrA D(mrr-hsdRMS-mcrBC) W80lacZDM15 DlacX74 recA1 deoR araD139 D(ara-leu)7697 galU galK rpsL (Str R ) endA1 nupG] (Invitrogen) used for plasmid propagation was grown at 37uC in Luria Broth supplemented with 100 mg/ml ampicillin. All plasmids and oligonucleotide primers used in this study are listed in supplementary Table S1 and Table S2, respectively. Target gene amplification was performed using High-Fidelity Taq polymerase (Roche) according to the manufacturer's instructions. PCR products were subcloned into pCRHII-TOPO (Invitrogen) and checked by DNA sequencing using primers M13for and/or M13rev (59-labeled with infra-red dye 800; MWG). For Dpp65 amplification, template JW4303 and primers 59pp65epi+39CMVepi were used. To obtain plasmid pPGK-Dpp65, the Dpp65 fragment was inserted into pPGK via EcoRI/BamHI. The gag-ORF, amplified using pTIL05 [40] as template and primers 59L-A ORF1+39L-A ORF1, was cloned as HindIII/BamHI fragment into vector pPGK to give pG. To obtain expression plasmids pGAG/Dpp65 and pGTXG, PCR reactions were carried out using the template/primer combinations JW4303/ 59CMVepi+39CMVepi and pUG36/59T7Xa-GFP+39GFP, respectively. Upon subcloning, both fragments were inserted as SacI/ BamHI fragment into pG. Plasmid yGTXG was constructed by introducing the HindIII/BamHI GTXG fragment from pGTXG into YEp352. The Xa gene (encoding the a-subunit of killer toxin K28) was amplified using vector pM28-SL [16] as template and primers 59SpeXal/39altaaBgl. Subsequently, the 59-terminal Spe I/ Bam HI fragment was integrated into yGTXG [Spe I/Bam HI] to give yGTXaD. GTXaD [Hin dIII/Bam HI] was then inserted into pPGK, and the GTXa fusion was completed by inserting the 39terminal Bgl II/Bam HI fragment of Xa into the Bam HI digested vector pGTXaD. For amplification of the Ce fragment, template JW4303 and primers 59pp65epi+39CMVepi were used. To obtain plasmid pCe, the Ce fragment was inserted into pPGK via Eco RI/ Bam HI. The multiple protease-deficient S. cerevisiae strain S86c [MATa ura3-2 his3 pra1 prb2 prc1 cps1 L-0 M-0] represents a heatcured, virus-free derivative of strain S86 [41] that was employed for the in vivo assembly of hybrid VLPs as well as for the expression of soluble (naked) Dpp65 in the yeast cell cytosol. If not otherwise stated, cells were cultivated in YPD at 30uC. Yeast cells were transformed by the lithium acetate method [42] and transformants were selected on synthetic complete medium lacking uracil (Ura-d/ o). Since in a yeast super-killer ski3 mutant (defective in exosome complex components) translation efficacy of the poly(A) 2 transcript of L-A is more effective and dsRNA copy number is significantly increased [43,44], a Dski3 variant of strain BY4741 [MATa his3D1 leu2D0 met15D0 ura3D0 Dski3] (Euroscarf) was used to prepare natural L-A virions.

VLP preparation
Transformants of the indicated yeast strain were incubated in 400 ml Ura-d/o at 220 rpm (30uC, in a 1 l-Erlenmeyer flask) to a density of 5610 7 -5610 8 cells/ml, harvested by 10 min centrifugation at 5,0006g (4uC), washed in prechilled H 2 O, thereafter in 1 M sorbitol, and finally resuspended in 50 ml cold PBSES (150 mM NaCl, 10 mM Na 2 HPO 4 pH 7.4, 10 mM EDTA, 1 M sorbitol). Subsequently, 2-mercaptoethanol (1:2,000) and 2.5 mg zymolyase 20T (Seikagaku, Japan) were added. Upon 1.5 h incubation at 120 rpm (30uC), spheroplasts were collected by 15 min centrifugation at 5006g (4uC) and washed in cold PBSES. Thereafter, cells were resuspended in 10 ml PBSE (150 mM NaCl, 10 mM Na 2 HPO 4 pH 7.4, 10 mM EDTA) and disrupted by vortexing seven times for 1 min (with 1 min breaks in between to cool samples on ice) in the presence of 12 g glass beads (0.45-0.55 mm in diameter). The resulting raw extracts were supplemented with 10 ml PBSE and centrifuged at 10,0006g for 1 h (4uC) to sediment glass beads and cell debris. The supernatant was adjusted with PBSE to 23 ml and then layered onto a cushion of 15 ml 45% sucrose. During ultracentrifugation at 69,2606g overnight (4uC; Beckman SW28 rotor) only structures of high molecular weight pass the cushion and form a pellet. Subsequently, the cushion pellet was resuspended in 1 ml PBSE and layered onto a linear density gradient (38 ml) of 20-70% sucrose. Upon further ultracentrifugation at 76,7406g overnight (4uC) the gradient was fractionated into 18-20 fractions (each 2 ml) while the gradient pellet was resuspended in 2 ml PBSE. Aliquots of each fraction were subjected to SDS-PAGE followed by western analysis or Coomassie blue staining. For reisolation of recombinant VLPs, a maximum of 12 fusion protein containing gradient fractions was pooled, supplemented with PBSE to 38 ml, and again ultracentrifuged at 76,7406g overnight (4uC). Finally, the VLP pellet was resuspended in 100-500 ml PBSE. The procedure described above was also used to prepare natural L-A particles from yeast strain BY4741 starting from a 200 ml YPD culture grown to a density of 5610 8 cells/ml.

Rapid extraction (S80 method) and detection of intracellular proteins
Yeast cells from a 1 ml overnight culture were collected by 3 min centrifugation at 8,5006g (20uC) and washed in H 2 O. Upon removal of the supernatant, the pellet was resuspended in the residual liquid and incubated at 280uC for 15 min. Cells were thawed on ice, mixed with 100 ml 36 TT sample buffer (0.15 M Tris/HCl pH 6.8, 12% SDS, 30% glycerol, 0.03% Coomassie Blue R250, 0.6% 2-mercaptoethanol) and heated at 100uC for 5-10 min with occasional vortexing. Thereafter, samples were centrifuged at 17,0006g for 3 min (20uC) to pellet cell debris. Subsequently, the supernatant was transferred into a fresh reaction tube, and 5-20 ml aliquots were separated by SDS-PAGE. Aliquots of the gradient fractions and the cushion and VLP pellets, as well as HIC fractions (20 ml each) were mixed with 10 ml 36TT sample buffer, boiled for 3 min and subjected to SDS-PAGE. Protein samples were applied onto 7.5% SDS-polyacrylamide gels and run in Tris/Tricine buffer [45]. Upon separation, proteins were either stained with Coomassie Brilliant Blue R250 (Roth) or blotted onto polyvinyl difluoride membranes [46]. Blots were probed with monoclonal anti-pp65 (Novocastra), anti-T7 (Novagen), anti-GFP (Roche) or polyclonal antibodies raised in rabbit against native Gag-VLPs (anti-Gag) or chimeric Gag/K28a particles (anti-Gag/K28a) followed by treatment with an alkaline phosphatase-coupled secondary anti-mouse immunoglobulin (Sigma). For colorimetric signal detection blots were covered with NBT/BCIP solution (Roche) according to the instructions of the manufacturer. Protein concentration was determined by using a bicinchoninic acid assay kit (Sigma). Alternatively, defined amounts of bovine serum albumin (BSA; Sigma) served as standard for semi-quantitative determination of protein concentration after SDS-PAGE and Coomassie Blue staining.

Transmission electron microscopy
An aliquot of gradient-purified VLPs was layered onto a cupper grid (mesh 300-400; coated with poly-L-lysine) and allowed to bind for 5 min at room temperature. Upon washing three times with 30 ml TBS (150 mM NaCl, 100 mM Tris/HCl pH 7.5) the grid was incubated in uranyl acetate/methyl cellulose (1.8%/ 0.2%) for 5 min at room temperature (negative staining) before it was slowly dried. For analysis and documentation of VLP samples, a transmission electron microscope type TECNAI G 2 (FEI) equipped with a MegaView III camera (Olympus) was used.
Protein processing using factor X a and hydrophobic interaction chromatography (HIC) To release the GFP moiety from chimeric particles, GTXG-VLPs were prepared from gradient fractions and resuspended in 0.9 ml H 2 O. Upon addition of Triton X-100 (1% final concentration) the sample was rotated overnight at 20uC. Digestion was carried out using 55 mU factor X a /mg GTXG protein in a total volume of 4.5 ml at 20uC under rotation for 4 d. To remove residual intact capsid and factor X a , the sample was ultracentrifuged at 102,0006g (4uC) for 1 h, and the supernatant was treated with Xarrest agarose (Novagen) according to the manufacturer's instructions. The resulting sample (10 ml) was carefully supplemented with 850 mM ammonium sulfate and subsequently centrifuged at 12,4006g (4uC) for 1 h. The pellet was resuspended in 10 ml H 2 O while 2.5 ml aliquots of the supernatant were applied onto a column of Phenyl Superose HR5/5 (2 ml; Amersham Pharmacia) equilibrated in HIC buffer (pH 7.4) containing 850 mM ammonium sulfate and 100 mM KH 2 PO 4 . The column was washed with the same buffer and bound proteins were eluted in a linear gradient (15 ml) from 850 to 51 mM ammonium sulfate/100 mM KH 2 PO 4 (pH 7.4). The column was run at 0.5 ml/min, and 1 ml fractions were collected and analyzed by SDS-PAGE.

Microtiter activity assay
Catalytic activity in esterase-coupled VLPs was determined in microtiter plates (96 flat-bottom wells; Nunc) using an MF reader V2.9-0 (EMS). A stock solution of 80 mM 4-nitrophenylacetate (dissolved in dimethyl sulfoxide) was diluted to the indicated concentrations to serve as substrate solution. To start the reaction, recombinant protein (5 ml) was added to 90 ml PBS 50 buffer (150 mM NaCl, 50 mM Na 2 HPO 4 , pH 7.0), finally completed with 5 ml substrate solution to give a total volume of 100 ml. The release of 4-nitrophenol was photometrically measured at 405 nm (25uC) in 30 s intervals. Immediately prior to each measurement, the samples were automatically shaken at 600 rpm. The absolute activity [U] of the enzyme was calculated by the ratio of released 4-nitrophenol [mmol] within 1 min, while the specific activity is given by the quotient of absolute activity to the protein amount [U/mg].

Whole blood assay and flow cytometry
The procedures were performed as breviously described [15]. Briefly, T cell stimulations were carried out using heparinized human blood (450 ml) of four HCMV-seropositive and one HCMV-seronegative European individuals and different amounts of antigen in a standard volume of 45 ml (adjusted with PBSE buffer). Lysates of HCMV-infected and non-infected fibroblasts (CMV-Ag and control-Ag, BioWhittaker) served as controls. To determine the frequency of antigen-specific T cells by flow cytometry, at least 25,000 cells -each of CD4 + and CD8 + lymphocytes -were analyzed on a FACScan (Becton Dickinson) using the Cellquest software. CD4 + and CD8 + T cells were identified by gating on the lymphocyte population via cell size and granularity and their high CD4 or CD8 expression level, respectively. Specifically activated T lymphocytes were identified and quantified as CD69 and IFNc double-positive cells. Due to the observation that negative controls can lead to a minimal activation of T cells [15], the threshold of a significant response was defined to be 0.05% of counted T lymphocytes.