Fusion Proteins for Versatile Antigen Targeting to Cell Surface Receptors Reveal Differential Capacity to Prime Immune Responses Free

Vaccination is a highly efficient strategy for protection against the human and social cost associated with infectious pathogens. Although current vaccines confer efficient protection against many pathogens, there is a need for novel vaccines protecting against pathogens causing chronic infection, such as HIV or hepatitis C virus (HCV), or stimulating vigorous responses against tumor Ags. In these cases, rational design of vaccines capable of inducing effective CD8⁺ and CD4⁺ T cell in addition to humoral responses is an objective of vaccine design (1).

Targeting of protein vaccines to defined APCs has recently emerged as a promising vaccination strategy (2). Targeting should reduce greatly the amount of vaccine required and allow for more precise manipulation of the ensuing immune response but can also be used for targeted delivery of toxins (3). Both the type of presenting cell and its activation/maturation status affect the outcome of Ag presentation with respect to stimulation of CD8⁺ and/or CD4⁺ T and B cells and to acquisition of proinflammatory or regulatory effector functions. In the mouse, CD8⁺ dendritic cells (DCs) are particularly efficient in “cross-priming” of cytotoxic T cells, whereas CD8⁻ DCs excel in Ag presentation to Th cells (4). Moreover, although fully activated conventional DCs prime and expand protective immune responses required for pathogen control, immature conventional DCs, but also subsets, such as Langerhans cells, plasmacytoid DCs, and intestinal CD103⁺ DCs, have been proposed to induce regulatory T cell responses protecting from autoimmunity (5).

The nature of the receptor targeted may also affect the ensuing response. For example, routing into different intracellular compartments of OVA internalized via different receptors results in preferential presentation by MHC class I and class II molecules (6). Given that different compartments recruit different signaling molecules, it is conceivable that the nature of the receptor engaged by an internalized Ag may determine the type of effector T cells produced (7). Thus, by targeting Ags to a variety of receptors, it may be possible to finely tune not only the efficacy but also the quality of the ensuing immune response.

Several approaches rendering Ags suitable for targeting have been described [reviewed by Tacken et al. (8)]. Targeted Ags stimulate T and B cell responses with higher efficacy than nontargeted Ags and possess protective potential against infectious agents and tumors. The potential of targeted Ags for vaccination has been documented impressively for hybrid Ab–Ag proteins targeting the macrophage/DC receptor DEC205 (2). Other targeted receptors include the a_Mβ₂ integrin as well the CD40, CD207, CD209, Fcγ, and mannose receptor (MR) (8) and the DC NK lectin receptor-1 (9). However, the relative suitability of individual receptors for inducing desired B and T cell responses remains unknown, because current strategies do not allow for a direct comparison of targeting to a variety of receptors. Moreover, the laborious production of Ab–Ag hybrid proteins is an obstacle to rapid screening of the panoply of APC receptors remaining to be explored for targeting.

In this study, we report production and characterization of fusion proteins suitable for screening and quantitative comparison of Ag targeting to an unlimited number of cell surface receptors. The design of these proteins allows for expression of hydrophobic viral proteins as soluble proteins. Using an OVA model fusion protein, we rank 10 different receptors with respect to efficacy of CD8⁺ T cell priming, demonstrate substantial differences between the receptors with respect to T cell subset and B cell stimulation, and show that the proteins can induce effector CD8⁺ T cells with extremely high efficacy.

cDNA constructs encoding the various fusion proteins were assembled stepwise from PCR products, generated using high-fidelity enzymes (Advantage HF, Clontech, Mountain View, CA). First, a cassette encoding enhanced GFP (eGFP) linked to human ubiquitin (Ub) or Ub alone was amplified from vector pGene eGFP/Ub, kindly provided by F. Lévy (Ludwig Insitute for Cancer Research, Épalinges, Switzerland) (10). Both products were cloned in pCR-Blunt (Invitrogen, Cergy Pontoise, France). In the second step, cDNAs from several sources encoding the chosen Ags were amplified and also cloned into pCR-Blunt. The influenza proteins matrix 1, neuraminidase (extracellular), nucleoprotein, and hemagglutinin were ordered as insect cell-optimized DNA sequences from Geneart (Regensburg, Germany). The nonoptimized template was a viral RNA (H1N1, A/PR/8/34 mouse-adapted, a gift from N. Escriou, Unité de Génétique Moléculaire des Virus Respiratoires, Institut Pasteur, Paris, France). A sequence encoding HCV core (strain 1b), truncated by 15 hydrophobic residues at its C terminus, was amplified from plasmid pBac 1-E2Con (a gift from R. Bartenschlaeger, Department of Molecular Virology, University of Heidelberg, Mainz, Germany). A HIV-1 negative factor (nef) cDNA (strain Bru) was amplified from plasmid pBSnef-Bru (a gift from G. Niedermann, Radiation Therapy Clinic, University of Freiburg, Freiburg, Germany) and OVA from a pcDNA3 plasmid (a gift from C. Théry, Inserm U932, Institut Curie, Paris, France). The type 1 diabetes autoantigens mouse 65-kDa glutamic acid decarboxylase (GAD) and human insulinoma-associated protein 2 (IA2) were cloned from vectors previously constructed in the laboratory. Mouse proinsulin (PI) was ordered as an insect cell-optimized DNA sequence (Geneart). In the third cloning step, a cDNA encoding the first Ig-binding domain of protein G was constructed by inserting annealed complementary oligonucleotides into the vector pCR-Blunt. In the fourth step, a sequence encoding the gp64 insect protein signal peptide was excised from the published vector pAcUW51-DRB1*0404 (11) and ligated into the baculovirus transfer vector pVL1392 resulting in pVL1392-gp64. All of the plasmids were sequenced to ensure the absence of errors introduced in the PCR. Fusion proteins were then assembled sequentially. In the final step, the entire cassette including two or three protein G domains was transferred into the pVL1392-gp64 plasmid. To produce fusion proteins with mutations at amino acid positions 48 and 76, we used the QuikChange Mutagenesis Kit (Stratagene, Amsterdam, The Netherlands), followed by sequencing.

Recombinant baculoviruses were produced and cloned using standard methods. For production of recombinant proteins, baculovirus-infected Sf9 (Spodoptera frugiperda) or Hi-5 (Trichoplusia ni) insect cells were incubated in serum-free Ex-Cell 420 (Sigma-Aldrich, Saint-Quentin Fallavier, France; for Sf9) or Ex-Cell 405 (Sigma-Aldrich; for Hi5) medium for 1–3 d. In a typical large-scale purification of a fusion protein, 100–200 ml Sf9 cell supernatant was passed over a 3 ml rabbit Ig/Sepharose 4B column. Bound protein was eluted using a buffer containing 20 mM N-cyclohexyl-3-aminopropanesulfonic acid (pH 11.5) and 1% 3-[(3-cholamidopropyl) dimethylammonio]-1-propane sulfonate. Eluted fractions containing the fusion proteins were neutralized, pooled, dialyzed against PBS with 10% glycerol, and stored in aliquots at −80°C.

Standard denaturing SDS-PAGE analysis was performed using minigel equipment. For immunoblots analysis, proteins were transferred onto nitrocellulose in a buffer containing 10 mM N-cyclohexyl-3-aminopropanesulfonic acid and 10% methanol. Membranes were blocked overnight with 2% nonfat dry milk in TBS (20 mM Tris, 150 nM NaCl [pH 7.5]). Abs were diluted in TBS with 0.05% Tween 20, and Ab binding was visualized by enhanced chemiluminescence (ECL Plus, GE Healthcare, Munich, Germany). The following primary Abs were used: supernatant (1:50) of hybridoma M2-1C6-4R3 (American Type Culture Collection, Manassas, VA) specific for influenza matrix protein 1; polyclonal rabbit anti-chicken OVA (Immunology Consultants Laboratories, Newberg, OR). Peroxidase-coupled goat anti-mouse or anti-rabbit Abs (Jackson ImmunoResearch Laboratories, Cambridgeshire, U.K.), respectively, were used as secondary reagents.

Human DCs were prepared according to Sallusto et al. (12). Monocytes were purified from fresh human blood by positive (CD14) selection, using paramagnetic microbeads (Miltenyi Biotec, Paris, France), cultured in RPMI supplemented with 10% male human AB serum, 250 ng/ml GM-CSF (Leucomax, Schering-Plough, Levallois-Perret, France), and 10 ng/ml IL-4 (R&D Systems, Abingdon, U.K.), and used starting on day 6. To examine internalization of Ab/fusion protein complexes, 1 × 10⁵ DCs were preincubated for 60 min at 4°C with the eGFP-containing fusion protein P2EUM (5 μg/ml), or with complexes of P2EUM with anti–DC-specific ICAM3-binding nonintegrin (DC-SIGN) mAbs (5 μg/ml fusion protein and 10 μg/ml mAb, in PBS/glycerol, preincubated 1 h at 4°C) before shifting to 37°C for various periods. At the end of 37°C incubations, cells were fixed in 1% paraformaldehyde (PFA; Sigma-Aldrich), stained with 5 μg/ml PE-labeled goat Abs to mouse Ig (Beckman Coulter, Roissy, France), and analyzed on a FACSCalibur machine (BD Biosciences, Le Pont de Claix, France).

DCs (1 × 10⁵ per staining) were allowed to adhere during 45 min to coverslips pretreated with 0.033% poly-l-lysine. Then fusion protein alone or preformed complexes of fusion protein (5 μg/ml) and mAb RK113 (10 μg/ml) were added to cells for 1 h at 4°C. DCs were incubated for 1 h at 37°C to allow internalization, washed, fixed in 4% PFA in PBS, and neutralized with 100 mM glycine in PBS. Then cells were permeabilized, using PBS with 0.2% BSA and 0.05% saponin for 30 min at room temperature, stained for 45 min at room temperature with a mAb against DC lysosomal membrane-associated protein (DC-LAMP; clone 104.G.4; Beckman Coulter), previously labeled with Alexa Fluor 594 using a Zenon kit (Invitrogen). Images were acquired on a DMI 6000B fluorescence microscope (Leica, Rueil Malmaison, France) equipped with a piezoelectric translator and a Coolsnap HQ charge-coupled device camera (Roper Scientific, Duluth, GA). Images were deconvolved using Metamorph 6.3.7 software (Universal Imaging Corporation, Downington, PA).

Bone marrow cells from C57/BL6 (H-2^b), 129Sv (H-2^b), or BALB/c (H-2^d) mice were cultured in IMDM for 6–13 d with 10% FCS, supplemented with 20% supernatant from J558 transfectants secreting GM-CSF. OVA-specific T cells recognizing a peptide in the context of H-2K^b (OT-I) (13) or of I-A^b (OT-II) (14), respectively, were prepared from lymph nodes of transgenic mice.

To study Ag presentation, DCs suspended in AIM V medium (Invitrogen) with 10% FCS were seeded at 50 μl and 100,000 cells per well in 96-well plates and incubated overnight with Ag. mAb/fusion protein complexes were formed by preincubating the two components, at a 1:1 molar ratio, for 1 h at 4°C. The following Abs were used: rat anti-mouse DEC-205 (clone NLDC-145), hamster anti-mouse CD11c (clone N418), and control mAb mouse anti–HLA-A2 (clone BB7.2) were purchased from American Type Culture Collection and purified from hybridoma supernatant; rat anti-mouse TLR2 and rat anti-mouse CD206 (MR1) were from AbD Serotec (Cergy Saint-Christophe, France); control mAb mouse anti-human TAP1 (clone 148.3) was purified from ascites. After overnight pulsing with complexes or with OVA (Worthington, Lakewood, NJ), DCs were washed twice in medium before addition of 100,000 naive or effector T cells in 100 μl AIM V to each well. Cytokine secretion by T cells was analyzed using commercial sandwich ELISA (BD Biosciences) as follows: OT-I effectors, IFN-γ after 24 h; naive OT-I, IL-2 after 24 h or IFN-γ after 72 h; naive OT-II, IL-2 after 48 h.

A human CD4⁺ T cell line recognizing the type 1-diabetes autoantigen IA2 was produced by stimulating PBLs from a HLA-DR4/DR15 healthy donor repeatedly with 5 μg/ml of the recombinant intracellular domain of human IA2. Cells were cloned by limiting dilution and shown to recognize peptide IA2 694–704 presented by HLA-DRB5*0101 (P.v.E., unpublished observations). Cells were stimulated biweekly with autologous PBL pulsed with 1 μg/ml cognate epitope. To study the effect of Ag targeting, 1 × 10⁵ cloned T cells were added 17 d after restimulation to 2 × 10⁵ autologous DCs in RPMI with 7.5% human serum and incubated for 16 h in the presence of titrated amounts of recombinant IA2 or of complexes between Ab RK113 and fusion protein P3UhIA2. T cell activation was examined by staining with an APC-labeled CD4 Ab together with a PE-labeled CD69 Ab and analysis on a FACSCanto II machine. In parallel, 5 × 10⁴ T cells labeled with 10 μM CFSE were incubated with 1 × 10⁵ DCs and titrated Ag amounts. CD4⁺ T cell proliferation was read out as a decrease in the CFSE staining intensity after 3 d.

Naive OT-I or OT-II T cells were labeled at 2.5 × 10⁸ cells per milliliter with 10 μM CFSE for 12 min at 37°C, washed, and adjusted to a concentration of 1–2 × 10⁶ cells per milliliter and injected i.v. (100 μl per mouse). Twenty-four hours later, mice were injected under anesthesia in the four footpads with Ag targeted with various Abs. In addition to Abs studied in in vitro experiments (see above), Abs with the following specificities were used: CD40 (clone FGK45; American Type Culture Collection), CD207 (Langerin; clone eBioL31; eBioscience, San Diego, CA), CD209 (CIRE; clone 5H10; eBioscience), CD169 (clone 3D6.112; AbD Serotec), Siglec H (clone 440c; HyCult, Uden, The Netherlands), polyclonal hamster Ig (AbD Serotec), rat IgG2a (Invitrogen), and polyclonal mouse Ig (Invitrogen). Three days later, lymph node cells were stained with anti–CD8-allophycocyanin plus anti–Vβ5-PE for OT-I or with anti–CD4-allophycocyanin plus anti–Vβ5-PE for OT-II (Abs from BD Biosciences). Proliferation of CFSE-labeled cells was quantified as the number of cell divisions relative to the precursor transgenic cell number according to Angulo et al. (15), using a FACSCalibur flow cytometer and FlowJo 6.4.7 (Tree Star, San Carlos, CA) software. Briefly, the division index = (100 − Y)Y where Y(%) = x₀ + x₁/2 + x₂/3 + x₄/8 + x₄/16 + x₅/32 + x₆/64 + x₇/128, x₀ represents the percentage of OT-I cells that have not divided, and x_1–7 represents those within progressive CFSE division gates (15). Animal use and experimentation was approved by the Comité régional d’éthique pour l’expérimentation animale Ile de France—René Descartes. The results of the experiments in Fig. 4 were analyzed by the Kruskal-Walls test, using Prism software (GraphPad, San Diego, CA).

Six-week-old C57BL/6 mice (Janvier, Le Genest-Saint-Isle, France) were injected s.c. in the footpads with 3 μg Ag or an equivalent amount of TLR2-targeted fusion protein P3UOrv or P3UmGADrv. Twenty hours later, draining lymph nodes were removed and incubated for 15 min in PBS containing 1.6 mg/ml collagenase D (Roche, Meylan, France) at room temperature. CD11c⁺ DCs were purified using paramagnetic beads (Miltenyi Biotec) and incubated at 10⁵ cells per well overnight with 10⁵ naive OT-I cells at 37°C. The IL-2 (and IFN-γ, data not shown) secretion was analyzed by commercial sandwich ELISA (BD Biosciences) on day 1 (or day 4 for IFN-γ).

Anesthetized C57BL/6 mice were injected in hind footpads with preformed Ab/fusion protein complexes. One to 72 h later, mice were sacrificed, and popliteal and inguinal lymph nodes removed and frozen at −80°C in optimum cutting temperature compound (Sakura, Zoeterwoude, The Netherlands). Five- to six-μm sections prepared on a cryostat were placed on glass slides, dried, and fixed in methanol for 5 min or in 2% PFA in PBS for 15 min at room temperature. The slides were washed in distilled water, dried, and blocked with 5% BSA, 10% FCS, and 0.1% saponin in PBS for 30 min. For CD19 staining, the sections were blocked with the Biotin Blocking System (Dako, Trappes, France). Sections were incubated for 1 h with primary Abs and for 30 min with secondary Abs (both diluted in blocking buffer). Next to Abs described above, primary Abs with the following specificities were used: CD3 (clone KT3), CD19 (clone 6D5; EuroBioSciences, Friesoythe, Germany), CD35 (clone 8C12; BD Pharmingen, San Diego, CA).

Two million naive OT-I cells (CD45.2⁺) were injected i.v. in CD45.1⁺ C57BL/6 mice, followed 24 h later by s.c. immunization with CD11c-targeted P3UOrv, PBS, or OVA with or without adjuvants LPS (1 μg), polyinosinic:polycytidylic acid [poly(I:C), 25 μg], and anti-CD40 (50 μg). The mice were boosted 2 wk later s.c. with the same Ag without adjuvant, and draining lymph nodes were removed 3 d later. Single-cell suspensions were analyzed by flow cytometry for CD8 and CD45.2 staining. For cytokine secretion assays, C57BL/6 mice were immunized s.c. with CD11c-targeted P3UOrv with or without adjuvant poly(I:C) (25 μg) and anti-CD40 (25 μg s.c. or 50 μg i.p.). Seven days later, the draining lymph nodes and the spleen were removed, and the obtained single-cell suspensions were restimulated for 6 h with cognate or control peptide (10⁻⁶ M) in the presence of 10 μg/ml brefeldin A. Cytokine secretion was detected by staining of permeabilized cells with anti–IFN-γ (clone XMG1.2; BD Biosciences), anti–IL-2 (clone JES6-5H4; BD Biosciences), and anti–TNF-α (MP6-XT22; BioLegend Europe, Uithoorn, The Netherlands) and analyzed using a FACSCanto II cytometer (BD Biosciences) and FlowJo 6.4.7 software.

One million naive OT-I cells were injected i.v. in 12-wk-old C57BL/6 mice, followed 24 h later by s.c. immunization with targeted P3UOrv with or without adjuvants. Twelve days later, 3 × 10⁶ syngeneic splenocytes pulsed with 10⁻⁶ M cognate peptide SIINFEKL (S8L) or control peptide TSYKFESV (T8V) and loaded with 0.5 μM (T8V) or 5 μM (S8L) CFSE were injected. Sixteen hours later, the draining lymph nodes and the spleen were removed, and surviving CFSE-loaded cells were quantified by flow cytometry.

Six- to eight-wk-old C57BL/6 mice were injected once s.c. with P3UOrv alone or complexed with targeting Abs or with OVA, together with adjuvants: 25 μg poly(I:C) and 25 μg anti-CD40 in 120 μl PBS. Mice were boosted with the same Ags and amounts without adjuvant 5 wk later, and serum Ab titers were tested 1 wk later by ELISA, using plates coated overnight with 100 μl OVA (10 μg/ml) at 4°C. Bound Ab was detected using peroxidase-coupled goat Abs against different mouse Ig classes and isotypes (1:400; Southern Biotechnology Associates, Birmingham, AL) and BD OptEIA tetramethylbenzidine substrate (BD Biosciences).

Our aim was to produce a wide range of Ags, including hydrophobic vaccine Ags, in a form allowing for versatile targeting to surface receptors on professional APCs and inducing efficient priming of T and B cell responses. The components assembled to achieve this are displayed in Fig. 1A. The signal peptide chosen mediates efficient secretion of recombinant proteins by insect cells (11). Tandem Ig binding domains of protein G (P2 or P3), which has broad Ig binding capacity across various species and isotypes (16), were inserted to allow for binding of the fusion proteins to Abs. Ub (U) was inserted upstream of the Ags because it can enhance proteasomal degradation of proteins linked to its C terminus (17). Some fusion proteins were also produced with an eGFP (E) to facilitate monitoring of fusion protein binding to and internalization by cells using flow cytometry and microscopy.

Plasmids encoding the fusion proteins listed in Table I were assembled from PCR-amplified building blocks and transfected into Sf9 insect cells to produce recombinant baculoviruses. Secreted fusion proteins were detectable by immunoblot in serum-free culture supernatants starting at 20 h after baculovirus infection of Sf9 insect cells (Fig. 1B). To obtain purified proteins, supernatants, collected at time points when optimal amounts of nondegraded protein were detected (e.g., 36 h for P2UM; Fig. 1B), were passed over a rabbit Ig immunoaffinity column. More than 90% pure full-length fusion proteins could be recovered from serum-free culture supernatants by this single-step procedure (Fig. 1C). Yields were between 0.25 and 3.5 μg/ml supernatant for all of the proteins. Remarkably, the system allowed for expression of hydrophobic proteins, such as influenza matrix 1 and nucleoprotein, HIV nef, and HCV core, all of interest as vaccines. When expressed in unmodified form in insect cells, these proteins require 8 M urea or strong detergents for solubilization (R.K. and P.v.E., unpublished observations).

To find out whether the fusion proteins were suitable for Ab-mediated targeting to APCs, we first studied binding to Ig from different species and of different isotypes. Fusion protein P2EUM bound quantitatively or near quantitatively to bead-immobilized polyclonal rabbit Ig and an IgG1 mouse mAb (Supplemental Fig. 1). Equivalent results were obtained with other murine mAbs of the IgG1, IgG2a, and IgG2b isotypes and with human, hamster, and rat polyclonal Ig (data not shown). Next, we investigated internalization of the fusion protein by purified DCs. For these experiments, we used mAbs produced in our laboratory (R. Kratzer and P. van Endert, unpublished observations) (Supplemental Fig. 2) recognizing DC-SIGN. Incubation at 37°C of human DCs precoated with complexes formed between P2EUM and anti–DC-SIGN mAb RK113 (which can be internalized) resulted in rapid internalization of the fusion protein, whereas complexes between P2EUM and a second anti–DC-SIGN mAb (RK526, which cannot be internalized) remained on the cell surface (Fig. 2A). Internalization of RK113/P2EUM complexes was confirmed by fluorescent microscopy. Although the fusion protein alone remained surface-associated after incubation for 1 h at 37°C (Fig. 2B, left panel), RK113/P2EUM complexes were found in endolysosomal compartments (Fig. 2B, right panel) identified by simultaneous staining with an Ab to DC-LAMP. Thus, efficient uptake of fusion protein requires targeting by an Ab-recognizing cell surface receptor. Incubation of human or murine DCs with fusion protein and/or targeting Ab did not result in DC activation, as assessed by staining for CD40, CD80, CD83, and CD86 (Supplemental Fig. 3).

To study delivery of fusion proteins to Ag processing compartments and resulting epitope presentation, we used OVA fusion proteins, taking advantage of OT-I and OT-II transgenic T cells that recognize OVA peptides presented by murine MHC class I and class II molecules, respectively, of the H-2^b haplotype. Although OVA concentrations of 200–500 μg/ml were required for optimal stimulation of OT-I and OT-II cells by bone marrow DCs in vitro, substantially lower concentrations of the P3UO fusion protein were sufficient for the same effect (Fig. 2C, 2D). Complexes between Abs to CD11c or TLR2 (the latter for OT-I) and P3UO were most efficient (~100-fold better than OVA); however, several other Abs, including control Abs, and P3UO alone when used in medium containing Ig were also far superior to OVA in inducing MHC class I and class II-mediated Ag presentation. The enhancing effect of complex formation with control Abs or serum Ig is consistent with the well-known capacity of FcRs to mediate efficient Ag uptake and presentation. Moreover, targeting to DC-SIGN increased presentation by human DCs of the type 1 diabetes autoantigen IA2 to a human CD4⁺ T cell clone by a factor of 10–50, as assessed by monitoring of the expression of the early activation marker CD69 and of proliferation reflected in CFSE dilution (Fig. 2E, 2F).

Next, we sought to determine whether in vitro findings reflected the targeting efficacy of fusion proteins in vivo. In preliminary experiments, we tested the stability of mAb/fusion protein complexes in the presence of an excess of polyclonal serum Ig. Incubation with a 10-fold excess of serum Ig for 30 min at 37°C reduced the amount of complexes by only 20% (Supplemental Fig. 4), suggesting that the vast majority of mAb/fusion protein complexes will remain stable during transport during drainage in lymphatic vessels. We analyzed proliferation of adoptively transferred, CFSE-labeled naive OT-I and OT-II T cells, reflected in dye dilution 3 d after Ag administration (Fig. 3A). These experiments confirmed the vastly superior performance of the fusion proteins and established a ranking among targeting Abs with respect to T cell priming. Fig. 3B shows the amount of different Ag forms required to induce two divisions of adoptively transferred OT-I T cells. Priming with 130 ng CD11c-targeted Ag or 225 ng TLR2-targeted Ag induced the same proliferative response as priming with 60 μg OVA, with targeting to the MR, DEC-205, or via a control mAb, all requiring ~1 μg (Fig. 3B). Thus, CD11c-targeted OVA was 460-fold more efficient than nontargeted OVA. CD11c^hi conventional DCs purified from mice injected with TLR2-targeted P3UOrv primed OT-I T cells in vitro, demonstrating that the fusion protein was indeed delivered to lymph node DCs (Fig. 3C).

To obtain a more comprehensive view of the efficacy of targeting to diverse receptors, we studied CD8⁺ T cell priming upon OVA targeting to additional receptors expressed strongly by Langerhans cells (CD207), activated APCs (CD40), subcapsular sinus macrophages (CD169), plasmacytoid DCs (Siglec H), macrophages/DCs (CD209), and FcR-expressing cells (rat or mouse Ig complexes) (Fig. 3E). The receptors studied can be divided into three groups displaying decreasing efficacy of CD8⁺ T cell priming: CD11c; TLR2 and CD40; and all other receptors, with the latter group still being ~50-fold more efficient than OVA alone. Interestingly, coadministration of anti-CD40 adjuvant had no effect on the priming efficacy of CD11c-targeted P3UO but increased the efficacy for all of the other tested receptors (DEC205, MR, and TLR2) to the level of CD11c-targeted Ag (Fig. 3E).

To rule out that the increase in T cell stimulation mediated by targeting Abs was due to systemic Ab effects, such as generalized activation due to LPS contamination of Abs, we examined the effect of fusion protein injection in two footpads together with targeting Ab injection in the contralateral footpads (Supplemental Fig. 5). Contralateral injection of anti-CD11c did not enhance OT-I priming by nontargeted P3UO. Moreover, passing the anti-CD11c mAb over a column for LPS removal did not change its targeting potential, providing additional evidence against a role of LPS contamination in the effect of this mAb.

To compare the effect of fusion protein targeting on MHC class I-restricted presentation with that on MHC class II-restricted presentation, we injected mice with the amounts inducing two divisions of OT-I cells ranging from 130 ng to 60 μg (see Fig. 2B) and examined proliferation of adoptively transferred OT-II cells (Fig. 3D). MHC class II-restricted OVA presentation to OT-II cells was also enhanced >100-fold by Ag targeting to DCs. However, the relative efficacy of CD8⁺ and CD4⁺ T cell priming differed between the receptors studied. Thus, whereas the P3UO fusion protein targeted to the MR primed OT-I and OT-II T cells with equivalent efficacy, OVA alone (60 μg) and CD11c-targeted P3UO (130 ng) primed OT-II cells more efficiently than OT-I cells, whereas the opposite was observed for TLR2-targeted P3UO (Fig. 3D).

The efficacy of CD8⁺ T cell priming by the OVA fusion protein was enhanced by modulation of two parameters: the presence of three rather than two protein G domains, which presumably enhances stable Ig binding (Fig. 4A), and substitution of Gly⁷⁶ at the C terminus of Ub with valine, which prevents cleavage by deubiquitinating enzymes and promotes Ag degradation in the “Ub fusion pathway” by the proteasome (18) (Fig. 4B). Ub substitution had no effect on OT-II priming (data not shown). Processing of targeted fusion proteins not only involved the proteasome but also TAP, because transferred OT-I cells did not proliferate in fusion protein-injected TAP-deficient mice (Supplemental Fig. 6).

To identify the cell populations accounting for highly efficient T cell priming in vivo, we monitored localization of P3UO fusion protein from between 1 and 72 h after s.c. injection by fluorescent microscopy. Both CD11c-targeted and nontargeted P3UO were detected in the subcapsular zone and in large clusters in the cortical zone of draining lymph nodes, although with different staining intensities (Fig. 5, top panels). From 1 to ~8 h after injection, P3UO was found in the marginal zone in cells staining for CD169, a marker for subcapsular sinus macrophages. At later time points (8–24 h), P3UO was detected in the center of B cell follicles, within follicular dendritic cells staining for CD35, whereas Ag could not be detected in paracortical T cell zones. Although Ag was rarely seen in areas close to APCs staining for the macrophage/DC marker DEC205 (Fig. 5, bottom right panel), colabeling of DEC205⁺ APCs for OVA was not detectable. This was not due to failure of targeting of injected anti-CD11c Ab to DCs, because lymph nodes could not be stained for CD11c between 4 and 24 h after injection, presumably because of saturation and/or downregulation of the receptor by injected Ab (Supplemental Fig. 7). Thus, CD11c-targeting Abs effectively reached conventional DC populations in lymph nodes.

T cell proliferation upon priming by targeted protein Ags is commonly followed by activation-induced cell death unless strong adjuvants are used (19, 20). To find out whether this also applied to CD11c-targeted OVA, we examined survival of adoptively transferred OT-I cells immunized on day 0 and boosted on day 14 with Ag in the presence or absence of adjuvant (Fig. 6A). Cells from mice immunized with 50 μg OVA or 130 ng CD11c-targeted P3UO alone, or in conjunction with LPS and poly(I:C), had disappeared by day 17. In contrast, when injected together with an Ab to CD40, an extremely low amount (15 ng) of targeted P3UO was sufficient for survival of a substantial number of OT-I cells (Fig. 6A), and dose-dependent generation of effector T cells was observed. CD11c-targeted Ag in conjunction with anti-CD40 adjuvant also generated effector T cells derived from the endogenous repertoire with high efficiency. For example, after s.c. injection of 50 ng targeted P3UO, 23% of OVA-specific CD8⁺ T cells in draining lymph nodes were triple producers secreting IL-2, TNF-α, and IFN-γ (Fig. 6B). Effector T cells were more abundant in the spleen than in draining lymph nodes. However, when adjuvant was administered s.c., the percentage of double or triple effectors was much higher in lymph nodes. The essential role for anti-CD40 adjuvant and the extreme efficacy of targeting to CD11c were further demonstrated in in vivo kill assays. OT-I cells immunized with 130 ng CD11c-targeted P3UO plus adjuvant completely eliminated target cells injected 13 d later, whereas no killing was observed when adjuvant was omitted (Fig. 6C).

Substantial differences between the different targeting Abs were also observed when the amount of Ig with different isotypes was determined in sera from mice primed with targeted P3UO plus anti-CD40 adjuvant and boosted once (Fig. 6D). Targeting to CD11c stimulated production of the whole array of Ig isotypes and nontargeted or MR-targeted P3UO production of IgG1 and IgG2. Thus, targeting of vaccine Ags to these receptors in the presence of adjuvants stimulates production of multiple Ig isotypes associated with Th1 and Th2 responses and may therefore be of interest for protective vaccination. In contrast, targeting to TLR2 resulted in production of low Ig amounts predominantly of the IgG1 isotype, consistent with the poor stimulation of CD4⁺ Th cells by Ag targeted to this receptor. CD11c-targeted and nontargeted P3UO also induced IgG1 in the absence of adjuvant and without boosting, whereas OVA alone did not induce Ab formation under these conditions (Supplemental Fig. 8).

The fusion proteins described in this study possess several properties that are of interest for studies evaluating the relative potentials of different receptors on selected cells for vaccine Ag delivery and that may also be useful for recombinant protein vaccines. Specifically, the stability in aqueous solution of fusion proteins containing strongly hydrophobic Ags, the ease of purification, the ability of binding to a large variety of targeting Abs, the capacity of being internalized, and efficient induction of MHC class I and class II-restricted T cell as well as B cell responses are relevant in this context. Although we observed induction of a full range of T and B cell responses by low fusion protein amounts in the OVA system, the efficacy of the system for induction of protective immunity as well as for tolerization strategies remains to be tested.

Different from published approaches, we used noncovalent Ag coupling to targeting Abs, taking advantage of the high affinity of streptococcal protein G (e.g., 5 × 10⁻¹⁰ M for human Ig) and of its binding to Ig of a comprehensive range of species and isotypes (16). To attain high binding affinity for rodent Ig, we engineered proteins containing three tandem Ig-binding domains, similar to native bacterial protein G. Proteins with two domains were less efficient in vivo, and proteins with a single domain bound rabbit but not murine Ig with good affinity (data not shown).

Noncovalent coupling of fusion proteins and targeting Abs had three potential drawbacks: 1) complexes might dissociate in vivo in the presence of circulating Ig; 2) Ab/fusion protein complexes might preferentially be targeted to FcγRIII (CD16), the low-affinity receptor for immune complexes, limiting effectively the versatility of the approach; and 3) repeated immunization with fusion proteins may induce production of Abs against the protein G moieties and targeting Abs, limiting responses against the targeted Ag. However, the results obtained after in vivo priming, as well as direct examination of complex stability in the presence of excess serum Ig, suggest that neither of the two former problems is encountered to a significant extent. Specifically, we find that mAb/fusion protein complexes are almost entirely stable in the presence of excess serum Ig; thus, specific targeting to the selected receptors is unlikely to be compromised by formation of complexes with nonspecific, FcR-binding Ig upon injection in vivo. Moreover, the superior priming of CD8⁺ T cells upon Ag targeting to CD11c and TLR2 (also observed after i.v. priming; data not shown), relative to targeting to FcRs using polyclonal mouse Ig, is not consistent with complex dissociation and/or targeting to CD16 in vivo. Along the same line, the distinct patterns of relative CD8⁺ and CD4⁺ T and B cell priming by individual Abs suggest that these problems are of limited relevance. Concerning the production of “neutralizing” Abs against protein G and targeting Abs, we have found that a single preimmunization with CD11c Abs (a hamster mAb) complexed with an unrelated fusion protein (P3UmGAD) does not reduce OVA-specific subsequent T cell and Ab responses upon injection of anti-CD11c/P3UO complexes. However, two preimmunizations with anti-CD11c/P3UmGAD complexes reduce subsequent OVA-specific responses upon anti-CD11c/P3UO immunization to the level observed for targeting of P3UO to FcRs in naive mice (data not shown). Note that this level is still far superior to that induced by nontargeted Ag.

Next to conferring high affinity for Ig, fusion of Ags to protein G had the important advantage of allowing for expression of several hydrophobic proteins as secreted soluble proteins. Both protein G and Ub are exceptionally stable proteins that can stabilize covalently linked proteins (21). Although most of the fusion proteins listed in Table I remain to be tested for functionality and T cell stimulation, the fact that 10 out of 12 tested proteins, including at least four strongly hydrophobic Ags, could be expressed as soluble secreted proteins suggests that the system will be permissive for many other Ags of interest as vaccine components.

Addition of a Ub moiety to the N terminus of proteins can enhance their degradation in the proteasome-dependent MHC class I Ag processing pathway under several circumstances. Nonremovable (by substitution of Gly⁷⁶) Ub induces the so-called Ub fusion pathway (18). To act as a degradation signal, Ub must form poly-Ub chains; although these are generally coupled to Lys⁴⁸, coupling to Lys²⁹ can be equally or even more important (17, 22). Ub coupling, resulting in enhanced degradation (17, 22), has been successfully used to enhance CTL responses to Ags encoded by DNA vaccines (23) but, to our knowledge, not so far in recombinant protein vaccines. Our data show that introduction of a noncleavable junction to Ub enhances Ag processing and priming of CD8⁺ T cells. Coincidentally, these data as well the complete lack of CD8⁺ T cell priming upon fusion protein injection in TAP deficient mice demonstrate that processing of targeted fusion proteins is strictly dependent on the major, proteasome- and TAP-dependent cross-presentation pathway (24).

Similar to published studies [reviewed by Steinman (2)], targeted fusion proteins primed CD8⁺ and CD4⁺ T cells with much higher efficacy than Ag alone. Optimal stimulation of both cell types was achieved after injection of 0.2–0.5 μg of CD11c-targeted Ag; a similar range (0.05–0.25μg) has previously been reported for OVA coupled to a monovalent mAb against DEC205 (20), suggesting that fusion protein binding to mAb is compatible with efficient DC attachment, internalization, and intracellular routing of mAb and coupled Ag. Another feature shared between targeted fusion proteins and targeting through recombinant Abs is the role of DC licensing by anti-CD40 Abs (20), which in both cases is required for acquisition of CD8⁺ T cell effector functions, inhibition of activation-induced cell death and Ab production, but also enhance T cell priming by Ag targeted to all of the receptors but CD11c.

The initial studies reported in this paper also revealed some striking differences between individual receptors. A ranking of efficiency in CD8⁺ T cell priming could be established. Moreover, targeting to TLR2, although very efficient for CD8⁺ T cell priming, performs relatively poorly with respect to CD4⁺ T cell priming and Ab production. However, the most striking observation clearly is the extremely high efficiency of targeting to CD11c with respect to all of the parameters tested. The excellent suitability of CD11c for T cell priming has recently been noted in another study (25). It was speculated that the exceptional efficiency of CD11c targeting may be related to the uptake of the Ab by CD11c⁺ splenic marginal zone cells. Our observation of early massive colocalization of injected fusion proteins with CD169 in the marginal lymph node zone may help to explain why CD11c outperforms other targeting receptors. CD169 is expressed by subcapsular sinus macrophages, which have recently been shown to capture particulate Ags, including virions, and deliver them to follicular B cells (26), consistent with our observations for targeted OVA protein. These cells express CD11c, are strategically located in lymph nodes and spleen for Ag capture, and have been shown to capture fusion proteins comprising the cysteine-rich domain of the MR as well as immune complexes and transfer them to follicles, driving B cell affinity maturation (27–29). Another recent study concluded that subcapsular sinus macrophages primed virus-specific CD8⁺ T cells (30). Together with our observations, this study as well as an older one (31) raise the possibility that marginal zone and subcapsular sinus macrophages may contribute to cross-priming CD8⁺ T cells against Ags internalized through CD11c and possibly other receptors. It is also conceivable that subcapsular sinus macrophages and metallophilic macrophages regulate Ag uptake by CD8⁺ DCs, as recently suggested (32).

In contrast to CD169⁺ subcapsular metallophilic macrophages, conventional DCs in T cell zones showed little staining for fusion proteins, which might indicate that the fusion proteins are more rapidly degraded in CD11c^hi DCs than in CD169⁺ subcapsular macrophages. Rapid degradation of fusion proteins may be related to the presence of Ub domains. Whatever the explanation for the difficulty of fusion protein detection in T cell zone DCs, in vitro priming of OT-I cells by CD11c^hi DCs purified from mice injected with anti-TLR2/P3UO complexes demonstrates that the fusion represents an efficient strategy for Ag targeting to conventional cross-presenting DCs.

We are grateful to F. Lévy (Épalinges, Switzerland), R. Bartenschlaeger (Mainz, Germany), G. Niedermann (Freiburg, Germany), and C. Théry (Paris, France) for providing plasmids and/or Abs, F. Geissman (Inserm U838, Paris, France) for cDNA from human DCs, and P. Guermonprez (Inserm U932, Institut Curie, Paris, France) for providing OT-II transgenic mice. We thank F. Cordelières (Institut Curie, Orsay, France) and the Microscopy Facility of the Curie Institute at Orsay, France, F. Castelli (Institut National de la Santé et de la Recherché Médicale, Unité 1013) for help with fluorescent microscopy, and L. Saveanu (Institut National de la Santé et de la Recherché Médicale, Unité 1013) for cloning a human DC-SIGN cDNA.

Disclosures The authors have no financial conflicts of interest.