No Advantage of Cell-Penetrating Peptides over Receptor-Specific Antibodies in Targeting Antigen to Human Dendritic Cells for Cross-Presentation¹

Dendritic cells (DCs)³ are the professional APCs of the immune system. They capture Ags, process them into peptides, and present those on MHC classes I and II to T lymphocytes (1). Their capacity to regulate T cell immunity allows for the use of DCs in vaccination strategies, and the immunogenicity of Ags delivered by DCs administered to cancer patients has now been demonstrated in clinical studies (2, 3). The technique most widely used to load DCs with tumor-associated Ag involves incubation of DCs with peptides that bind directly to MHC molecules on the cell surface. However, the half-life of these exogenously loaded MHC-peptide complexes is relatively short, resulting in suboptimal Ag presentation. Alternatively, DCs can be incubated with tumor-associated proteins or tumor lysates, resulting in Ag uptake and proteolytic processing into peptides, which are loaded onto MHC molecules endogenously (3). Unfortunately, MHC class I presentation of exogenous Ags taken up by DCs is relatively inefficient (4). Specific targeting of Ag to the DC, either in vitro or in vivo, increases the efficiency of Ag uptake and thereby cross-presentation (5). These targeting strategies involve the coupling of cell-penetrating peptides (CPPs) (6, 7, 8) or Abs directed against DC-specific receptors (9, 10, 11) to the Ag.

CPPs are positively charged peptides that deliver macromolecules including proteins, oligonucleotides, plasmid DNA, and even beads and liposomes into living cells (12). There is much controversy in the literature on how CPPs mediate intracellular delivery of their conjugated cargo. Initial studies suggested that CPPs directly translocate across cell membranes into the cytoplasm in a temperature- and receptor-independent manner, but these observations have been dismissed as an artifact resulting from redistribution of CPPs upon the fixation of cells (13). It is now generally accepted that CPPs enter cells via endocytosis, which probably involves the binding of CPPs to negatively charged heparan sulfate proteoglycans on the cell surface (12, 14). However, there is still much debate over whether CPPs deliver their conjugated cargo predominantly to the endosomal compartment, followed by degradation in the lysosomal compartment, or if they facilitate the endosomal escape of cargo into the cytoplasm and nucleus (12). Escape of CPP-Ag conjugates from endosomes following uptake by DCs would favor cross-presentation of the Ag (4). Indeed, CPP-Ag conjugates enhance cross-presentation of the Ag by DCs (6, 7), but it remains to be established whether CPPs merely increase Ag uptake or also facilitate endosomal escape.

In addition to CPPs, Abs directed against DC-specific cell surface receptors have been used to target Ag to DCs. The advantage of using such Abs over CPPs is their ability to specifically target DCs in vivo, whereas CPPs target virtually all cell types (15, 16). Several receptors of the C-type lectin family are promising candidates for Ab-mediated Ag delivery to DCs, as they play an important role in Ag capture (17). Many studies on Ag delivery to DCs exploit the C-type lectin receptor DEC-205. Ag conjugated to DEC-205 Abs induces Ag presentation via MHC classes I and II. Moreover, it induces CTL responses in mice, provided DC maturation stimuli are coadministered (9, 18, 19). However, it remains to be determined whether DEC-205 targeting will be efficient in human, as human DEC-205, unlike mouse DEC-205, is expressed by a wide variety of cells other than DCs (20). In contrast, expression of the human C-type lectin receptor DC-SIGN (DC-specific ICAM-grabbing nonintegrin) is mostly restricted to DCs and macrophages (21, 22). We have previously shown that Ag targeted to DC-SIGN on human DCs in vitro is rapidly endocytosed, transported to the lysosome, and presented to T cells. In addition, there were strong indications that Ag targeted to DC-SIGN was presented via both MHC class I and class II (23).

In this study, we directly compare the efficiency of CPPs and an Ab directed against DC-SIGN to induce cross-presentation of conjugated Ag. First, various targeting constructs were generated to determine whether Ab-mediated Ag targeting to DC-SIGN on human DCs results in cross-presentation. Subsequently, the intracellular fate of CPP-protein conjugates in human DCs was analyzed and compared with that of the DC-SIGN-targeting Ab. In addition, we analyzed the capacity of three CPPs, the HIV-derived Tat (24), the superoxide dismutase-derived DPV3 (25), and an artificial polyarginine peptide (polyR) (26, 27) to induce cross-presentation. Finally, protein complexes of similar size and composition harboring a class I-restricted tumor-associated Ag were targeted to DCs either through a CPP or the humanized DC-SIGN Ab hD1. The results demonstrate that CPPs and hD1 are equally efficient in mediating the cross-presentation of conjugated Ag when targeted to human DCs in vitro.

N-succinimidyl-S-acetylthiopropionate, hydroxylamine-HCl, and sulfosuccinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate, immobilized iminobiotin, and streptavidin (SA) were purchased from Pierce. Goat anti-mouse IgG1 Alexa Fluor 488, SA Alexa Fluor 488, and Alexa Fluor 488 and 647 labeling kits were obtained from Invitrogen, Cy5-labeled SA was from Jackson ImmunoResearch Laboratories and Penta-His Alexa Fluor 488 came from Qiagen.

gp100(277–291)*SAMA is an N-terminal S-acetyl mercaptoacetic acid (SAMA)-derivatized peptide (THTYLEPGPVTAQVV) and was produced by Eurosequence. N-terminally biotinylated peptides gp100(272–288) (RALVVTHTYLEPGPVTA), MAGE (SKASSSLQLVFGIELMEVDPIGHLYYIFAT), Tat (GRKKRRQRRRPPQ), DPV3 (RKKRRRESRKKRRRES), and polyR (GGGRRRRRRRRRRR) were purchased from ServiceXS.

The composite IgG2/IgG4 humanized anti-human DC-SIGN Ab hD1 was generated by CDR grafting of mouse anti-human DC-SIGN Ab AZN-D1 hypervariable domains into human framework regions, as described previously (23). The Ab 5g1.1 (eculizumab; Alexion Pharmaceuticals), containing the same IgG2/IgG4 constant region (28), was used as a nontargeting negative control. To generate the anti-DC-SIGN single chain D1 Ab (scD1), the light and heavy chain variable regions (V_L and V_H) of hD1 were amplified individually by PCR. The Ab scD1 was assembled by an overlapping PCR to link V_L and V_H with a (GGGGS)₃ peptide linker and introduce six histidines 3′ of V_H. scD1 was produced by 293 cells and purified with Ni-NTA agarose columns.

scD1-gp100(280–288) and scD1-gp100(154–162) were generated by introducing nucleotide sequences encoding two copies of gp100(280–288) (YLEPGPVTA) for scD1-gp100(280–288) and two copies of gp100(154–162) (KTWGQYWQV) for scD1-gp100(154–162) between the C terminus of the single chain and the histidine tag.

scD1-gp100(277–291)*SAMA was generated by chemical cross-linking of scD1 to gp100(277–291)*SAMA. Therefore, scD1 was maleimide activated by 4-(N-maleimidomethyl)cyclohexane-1-carboxylate (sSMCC) according to the manufacturer's protocol (Pierce). Subsequently, activated scD1 was incubated with gp100(277–291)*SAMA in PBS (pH 7.4) with 0.67 M NH₂OH-HCl and 0.67 M NaOH for 2 h at room temperature. Unconjugated peptide was removed by size separation on a D-Salt column (Pierce). The coupling ratio of scD1 to gp100(277–291)SAMA was determined by SDS-PAGE analysis. Protein band intensities were determined using image analysis software, revealing that scD1-gp100(277–291)*SAMA is a protein mixture of 77% free scD1 and 23% conjugated scD1-gp100(277–291) molecules. The 23% conjugated molecules showed a SDS-PAGE mobility shift corresponding to a single peptide being introduced into each scD1 molecule (data not shown).

The targeting constructs hD1-SA and 5g1.1-SA were generated by conjugating SA to the hD1 and 5g1.1 Abs. Sulfhydryl groups were introduced into SA by N-succinimidyl-S-acetylthiopropionate (SATP), deprotected with hydroxylamine, and added to the maleimide-activated Abs according to the manufacturer's protocol. Unconjugated SA was removed by affinity chromatography on a HiTrap protein G column (1 ml of bed volume; Amersham Biosciences). Unconjugated Abs were removed by affinity purification of the mixture using immobilized iminobiotin according to the manufacturer's protocol. Where indicated, hD1-SA and 5g1.1-SA were incubated with a 2-fold molar excess of the biotinylated peptides Tat, DPV3, polyR, gp100(272–288), MAGE, or combinations thereof. Fig. 1 shows a diagram of the targeting constructs used in these studies.

Immature DCs (iDCs) were cultured from human PBMCs isolated from the blood of HLA-A2.1-positive individuals after confirmed consent, as reported elsewhere (21). The gp100-negative, HLA-A2.1-positive cell line BLM and BLM cells transfected with the melanocyte differentiation Ag gp100 were cultured as described previously (29). Jurkat T cells transduced with CD8 and the complete α- and β-chains of a TCR specific for the HLA-A2.1-restricted peptide gp100(280–288) have been described elsewhere (JE6.1 fl296) (30) and were a gift from Dr. R. Debets (Erasmus Medical Center-Daniel den Hoed, Rotterdam, The Netherlands).

Alexa Fluor 488-labeled hD1-SA or 5g1.1-SA were preincubated with a 4-fold molar excess of biotinylated CPPs. iDCs were incubated for 1 h with SA constructs (3 μg/ml) or the hD1 or 5g1.1-SA constructs (10 μg/ml) in RPMI 1640 medium supplemented with 10% FCS, at either 4°C or 37°C. Subsequently, iDCs were washed, resuspended in PBS and analyzed on a FACSCalibur flow cytometer using CellQuest software (BD Biosciences). In other experiments, various concentrations of peripheral blood leukocytes (PBLks) were incubated with 0.1 μg/ml Alexa Fluor 647-labeled hD1 or 0.5 μg/ml Cy5-labeled SA-polyR (subsaturating conditions; data not shown) in RPMI 1640 for 20 h at 37°C. Subsequently, the medium was transferred to wells containing 5 × 10⁴ iDCs, followed by 1 h of incubation at 37°C.

Competitive binding and uptake studies with ¹¹¹In-labeled hD1 and SA-polyR were performed essentially as described previously (31). In brief, the chelator diethylenetriamine pentaacetic acid (DTPA) was conjugated to hD1 and SA. Subsequently, proteins were labeled with ¹¹¹InCl₃ (Tyco Mallinckrodt). Radiochemical purity was >95%. DCs (10⁶) were incubated for 5 h at 4°C or 37°C in 2 ml of RPMI 1640 medium supplemented with 10% FCS containing 0.5 ng of radioligand in either the presence or absence of various concentrations of unlabeled ligand. Next, DCs were washed in PBS and harvested and cell-associated radioactivity was counted in a 3″ well-type gamma counter (PerkinElmer). The concentration of unlabeled drug blocking half the specific radioligand cell association was calculated using GraphPad Prism version 4.0 software (GraphPad Software). Maximum binding and association levels were calculated according to the manufacturer's instructions.

Binding and uptake of SA-polyR relative to hD1 was determined by flow cytometry. Hereto, Alexa Fluor 488-labeled SA was loaded with a 2-fold molar excess of biotinylated polyR. hD1 and isotype control Ab were labeled using an Alexa Fluor 488 labeling kit (Invitrogen). Alexa Fluor dye labeling efficiency was determined according to the manufacturer's protocol (Invitrogen). iDCs were incubated with 10 μg/ml Alexa Fluor 488-labeled hD1 or 3 μg/ml Alexa Fluor 488-labeled SA or SA-polyR for 5 h at 4°C or 37°C in RPMI 1640 medium supplemented with 10% FCS. Subsequently, iDCs were washed, resuspended in PBS, and analyzed by flow cytometry. Data were corrected for variations in Alexa Fluor dye labeling efficiency by dividing mean cell fluorescence values by the amount of dye incorporated per mole of SA and hD1. Specific binding and uptake were determined by subtracting SA and isotype control values from SA-polyR and hD1 values, respectively.

Alexa Fluor 488-labeled SA or Cy5-labeled SA was loaded with CPPs by incubation with a 4-fold molar excess of biotinylated DPV3, polyR, or Tat peptide. iDCs were incubated with 5g1.1-SA or 5g1.1-SA preincubated with a 4-fold molar excess of biotinylated CPP, combinations of various SA-CPPs, or combinations of SA-CPPs and Alexa Fluor 647-labeled hD1 at 10 μg/ml for 45 min at 37°C. Internalization of the targeting constructs was confirmed by confocal laser scanning microscopy. Cells were adhered to poly-l-lysine-coated glass slides and fixated with 1% paraformaldehyde in PBS. 5g1.1-SA and 5g1.1-SA-CPP were detected by methanol permeabilization followed by the addition of goat anti-human IgG secondary Abs. Cells were imaged with a Bio-Rad MRC 1024 confocal system operating on a Nikon Optiphot microscope and a Nikon ×60 Plan-Apochromat 1.4 oil immersion lens (Bio-Rad). Pictures were analyzed with Bio-Rad Lasersharp 2000 and Adobe Photoshop 7.0 (Adobe Systems) software.

iDCs were labeled with LysoTracker Red (Invitrogen) in RPMI 1640 without phenol red (Invitrogen) for 2 min. Subsequently, iDCs were washed and incubated with Alexa Fluor 488-labeled SA-polyR (3 μg/ml) and Alexa Fluor 647-labeled hD1 (10 μg/ml) at 4°C for 1 h, washed, and adhered to fibronectin-coated glass slides in RPMI 1640 without phenol red. Cells were analyzed at 37°C for 75 min with a Zeiss LSM 510 microscope equipped with a type S heated stage CO₂ controller and a Plan Apochromat ×63 1.4 oil immersion differential interference contrast lens (Carl Zeiss). Cells were imaged using Zeiss LSM Image Browser version 3.2 (Carl Zeiss) and processed with NIH ImageJ version 1.32j software (rsb.info.nih.gov/ij). Pearson correlation coefficients between the intensities of fluorescent labels were determined by the ImageJ Red Green Correlator plugin.

iDCs were incubated with the various targeting constructs at the indicated concentrations in RPMI 1640 medium supplemented with 10% FCS and the TLR3 ligand polyinosinic-polycytidylic acid (poly(I:C)) (Sigma-Aldrich). After 3 h, the TLR8 ligand R848 (PharmaTech) was added at 4 μg/ml to further induce DC maturation and enhance cross-presentation over a period of 20 h (32, 33). Subsequently, DCs were washed and cross-presentation of the Ag was assessed by the Ag presentation assay.

Presentation of the gp100(280–288) peptide by BLM-gp100 cells and targeted DCs was assessed by the TCR-transduced Jurkat JE6.1 fl296 cells in combination with an NFAT luciferase assay that was performed essentially as described previously (30). In brief, JE6.1 fl296 cells were transfected with an NFAT-luciferase reporter construct (Stratagene) using an Amaxa Cell Line Nucleofector kit V in combination with program G10 on an Amaxa Nucleofector. Transfected JE6.1 fl296 cells were cultured for 18 h, washed, and resuspended in RPMI 1640 medium supplemented with 1% FCS. Subsequently, DCs or BLM cells were added to the transfected JE6.1 fl296 cells at a ratio of 1:10 (DC) or 1:1 (BLM). After 6 h, Jurkat cells were harvested and lysed with cell culture lysis reagent (Promega). Luciferase activity in cell lysates was measured using a luciferase assay system from Promega. Samples were analyzed in a Gen-Probe Leader 50 luminometer and luciferase activity was expressed as light units relative to nontargeted DC controls.

The CPPs polyR, Tat, and DPV3 have been described as targeting proteins to cells (25, 26). To compare the efficiencies with which these three CPPs deliver protein to DCs, biotinylated polyR, Tat, and DPV3 were produced and conjugated to Alexa Fluor 488-labeled SA. DCs incubated with Alexa Fluor 488-labeled SA for 1 h at 4°C did not display cell-associated fluorescence, showing that SA itself does not bind to DCs (Fig. 2,A). However, performing the experiment at 37°C did reveal cell-associated fluorescence, suggesting that DCs take up SA nonspecifically (Fig. 2,A). Preloading of the Alexa Fluor 488-labeled SA with Tat, DPV3, or polyR enhanced cell-associated fluorescence, indicating that the CPPs delivered SA to the DCs (Fig. 2,A). In addition, the humanized anti-DC-SIGN Ab hD1 targeted DCs. No binding of the isotype control Ab was detected at 4°C while samples incubated at 37°C only revealed a minor peak shift, suggesting that aspecific uptake of the isotype control Ab was very low (Fig. 2 B).

Upon administration in vivo, CPPs and Abs will encounter a variety of cell types before encountering a DC. To mimic this in vitro and to study how it affects targeting, directly labeled SA-polyR or Ab hD1 were added to the culture medium of an increasing number of PBLks, followed by incubation at 37°C. Subsequently, the culture medium was transferred to wells containing a fixed amount of DCs. As expected, prior exposure to increasing amounts of PBLks did not affect association of the hD1 Ab with DCs, whereas it decreased the association of SA-polyR with DCs (Fig. 2 C). Flow cytometric analysis revealed that SA-polyR was associated with the PBLks, whereas hD1 was not (data not shown). Thus, the effectiveness of CPPs to deliver Ag to DCs in vivo will be hampered the fact that they also target other cells.

To directly compare CPP-mediated to Ab-mediated cross-presentation of Ags, targeting constructs of similar size and composition were required that could either be targeted via CPPs or to DC-SIGN. Therefore, SA was chemically cross-linked to the DC-SIGN targeting Ab hD1 (hD1-SA) and its isotype control 5g1.1 (5g1.1-SA). We have previously demonstrated that hD1 targets conjugated proteins to DCs (23). The hD1-SA protein conjugate could now be loaded with combinations of biotinylated peptides and targeted to DC-SIGN while 5g1.1-SA could be loaded with combinations of biotinylated peptides, including CPPs.

In addition to targeting SA to DCs, it was determined whether CPPs target the relatively large 5g1.1-SA protein complex, with a molecular mass of >200 kDa, to DCs. iDCs incubated with Alexa Fluor 647-labeled 5g1.1-SA for 1 h at 4°C displayed no cell-associated fluorescence whereas cells incubated with the construct at 37°C did, indicating that the DCs take up low levels of 5g1.1-SA nonspecifically (Fig. 2,D). Preloading of the Alexa Fluor 647-labeled 5g1.1-SA with the CPPs Tat, DPV3, or polyR enhanced cell-associated fluorescence (Fig. 2,D). From the results it was concluded that polyR was most potent in targeting both SA and 5g1.1-SA to DCs, followed by DPV3 and Tat (Fig. 2, A and D).

Because flow cytometric analysis does not discriminate between proteins bound to the cell surface and internalized proteins, cellular uptake of the various constructs was analyzed by confocal microscopy.

Internalization or binding of 5g1.1-SA without CPP attached could not be confirmed by confocal microscopy despite the fact that flow cytometric analysis revealed low levels of cell association (Fig. 2,D). Probably, the relatively high sensitivity of the FACS allowed for the detection of low levels of 5g1.1-SA nonspecifically taken up by DCs. Internalization of 5g1.1-SA coupled to CPP could be confirmed by confocal microscopy (Fig. 3 A). The strongest intracellular staining was detected when polyR was attached to 5g1.1-SA, whereas less intracellular signal was detected when DPV3 or Tat was used (data not shown). The internalized 5g1.1-SA-CPP proteins localized within the endosomal compartment, and no evidence for cytoplasmic localization was found.

DC-SIGN targeting and CPP-mediated targeting were compared by incubating DCs for 45 min with combinations of directly labeled hD1 and SA-CPP conjugates. Subsequently, DCs were subjected to fixation and analyzed by confocal microscopy. SA-DPV3 and SA-polyR showed a high degree of colocalization when targeted to DC. In addition, SA-polyR colocalized with SA-Tat. In contrast, the degree of colocalization between the anti-DC-SIGN Ab hD1 and SA-polyR was very low (Fig. 3, B and C). Because fixation of cells might introduce artifacts and alter the intracellular localization of CPPs (13), we confirmed these data by live imaging of cells upon the addition of directly labeled SA-polyR and hD1 (Fig. 3,D). In addition, the lysosomal compartment of the cells was visualized with LysoTracker. Ten minutes after SA-polyR and hD1 were added to the DCs, endosomal vesicles harboring either SA-polyR or hD1 could be seen traveling from the cell surface toward the lysosomal compartment. Although hD1 and SA-polyR were transported within separate endosomal vesicles, they started colocalizing upon reaching the lysosomal compartment ∼75 min after the targeting constructs were added to the culture medium (Fig. 3, D and E).

To quantify the number of molecules entering the DC via Ab- and CPP-mediated targeting, hD1 and SA-polyR were labeled with ¹¹¹In. The advantage of this technique over other labeling techniques is that the ¹¹¹In label, once internalized, is very poorly excreted by the cells, allowing accurate measurements of label accumulation (34). Competitive binding experiments on ice showed that ∼100,000 hD1 molecules are bound per DC. Performing competitive uptake experiments for 5 h at 37°C increased the number of cell-associated hD1 molecules by ∼38% (Table I). Unfortunately, polyR binding experiments failed to produce a standard competitive binding curve (data not shown). Such binding behavior is best explained by a mechanism in which the positively charged polyR interacted with multiple, distinct, negatively charged binding sites on the cell, each with its own binding constant. Therefore, the maximum binding capacity for polyR could not be determined. Flow cytometry experiments were designed to directly compare the binding and uptake of SA-polyR relative to hD1. Under saturating conditions, maximal SA-polyR binding comprised 70% of maximal hD1 binding (Table I). However, DCs incubated at 37°C for 5 h accumulated more SA-polyR than hD1 (Table I). The accumulation of hD1 measured in DCs in the flow cytometry experiments was comparable to that in the radioligand competition studies (154 and 138% of maximal binding, respectively). This indicates there was no major loss of ligand-associated fluorescence due to degradation in the flow cytometry experiments, as this would have resulted in decreased accumulation levels compared with those obtained by the relatively degradation-insensitive radioligand studies.

Our previous studies show that Ag targeted to DCs via the humanized anti-DC-SIGN Ab hD1 is presented to T cells (23). The targeting Ab hD1 was based on the mouse anti-human DC-SIGN Ab AZN-D1 and contains a composite human IgG2/IgG4 constant domain, which prevents binding to Fc receptors (35). The data showed that targeting resulted in presentation via MHC class II and possibly class I. To prove that Ab-mediated targeting of Ags to DC-SIGN results in cross-presentation, we now generated a series of targeting constructs consisting of the Ab hD1, either as a single chain or as a whole Ab, conjugated to the class I-restricted, melanoma-associated peptide gp100(280–288) (Fig. 1). Similar to what was shown previously for the hD1 Ab (23), the single chain hD1 Ab, scD1, was internalized by iDCs as efficiently as the “parent” AZN-D1 Ab (data not shown). To determine whether Ab-mediated targeting of Ag to DC-SIGN results in cross-presentation, DCs were incubated with various targeting constructs harboring the tumor-associated, HLA-A2.1-restricted peptide gp100(280–288), in combination with the TLR ligands poly(I:C) and R848 to induce DC maturation. Jurkat T cells transduced with a HLA-A2.1/gp100(280–288)-specific TCR were used as a readout for peptide presentation by targeted DCs. These TCR-transduced Jurkat T cells were transfected with an NFAT/luciferase reporter construct, resulting in luciferase expression upon engagement of the TCR (30, 36). The specificity of the TCR-transduced Jurkat T cells for gp100(280–288) was confirmed by incubating the cells with the HLA-A2.1-positive, gp100-negative, melanoma cell line BLM, either transfected (BLM-gp100) or untransfected (BLM) with a gp100 expression construct. The luciferase activity in Jurkat T cells incubated with BLM-gp100 cells was a factor 8.5 higher as compared with Jurkat T cells incubated with BLM cells (Fig. 4,A). This level of Ag presentation should be sufficient for T cells to recognize their target cell, because BLM-gp100 cells are efficiently lysed by gp100-specific CTLs (29, 36). As a second control, Jurkat T cells were incubated with HLA-A2.1-positive mature DCs (mDCs) loaded with gp100(280–288) or a control gp100(154–162) peptide. The Jurkat T cell response to mDCs externally loaded with gp100(280–288) peptide was 67-fold higher than the response to gp100(154–162) peptide-loaded mDCs (Fig. 4 A).

For targeting DC-SIGN, three single chain Ab constructs were generated. scD1-gp100(277–291)*SAMA was generated by chemically cross-linking a 15-mer peptide harboring the HLA-A2.1-restricted gp100(280–288) epitope to scD1. scD1-gp100(280–288) was produced by genetically inserting two gp100(280–288) peptides into scD1. As a negative control, scD1-gp100(154–162) was produced by genetically inserting two gp100(154–162) peptides into scD1. iDCs were targeted for 20 h with targeting constructs in the presence of the TLR ligands R848 and poly(I:C) to induce DC maturation. DCs targeted with the scD1-gp100(277–291)*SAMA construct induced the most potent response by the TCR-transduced Jurkat T cells (Fig. 4,B). At a concentration of 4.3 μM, which corresponds to 1 μM peptide, scD1-gp100(277–291)*SAMA induced a 5-fold increase in luciferase activity. In contrast, the gp100(277–291)*SAMA peptide without the scD1 attached did not result in a response. Similar to scD1-gp100(277–291)*SAMA, scD1-gp100(280–288) induced cross-presentation of the gp100(280–288) peptide, although the T cell response was less pronounced while the concentration of the targeted peptide was higher (4 μM, corresponding to 2 μM targeting construct; Fig. 4,B). As expected, DCs targeted with the scD1 construct harboring the control gp100(154–162) peptide did not elicit a response by the Jurkat T cells. When targeting was conducted in the presence of the DC-SIGN-blocking Ab AZN-D1, a significant reduction in cross-presentation of the scD1-gp100(277–291)*SAMA construct was observed (Fig. 4 C). These data show that specific uptake of the construct via DC-SIGN was required to induce cross-presentation.

Subsequently, the hD1-SA and 5g1.1-SA constructs were evaluated for their ability to induce cross-presentation of a biotinylated peptide linked to the SA moiety. Because SA harbors four biotin binding sites, these targeting constructs can be loaded with various combinations of biotinylated cell-penetrating, tumor-associated, and control peptides. hD1-SA and its isotype control 5g1.1-SA were loaded with a 2-fold molar excess of biotinylated gp100(272–288) peptide in combination with a 2-fold molar excess of a control peptide (hD1-SA-gp100-MAGE and 5g1.1-SA-gp100-MAGE). This control peptide was added so that it could be exchanged with CPP in subsequent experiments to study the effects of CPP-mediated targeting without disturbing the amount of gp100 peptide loaded on the targeting construct. Although DCs incubated with 0.05 μM 5g1.1-SA-gp100-MAGE presented the gp100(280–288) peptide to Jurkat T cells, DCs targeted with hD1-SA-gp100-MAGE presented the gp100(280–288) peptide more efficiently (Fig. 4,D). These findings demonstrate that DCs nonspecifically take up the Ab-SA constructs, probably by macropinocytosis, resulting in low levels of cross-presentation that can be enhanced by DC-SIGN targeting. The nonconjugated gp100(272–288) was cross-presented as efficiently as the 5g1.1-SA-gp100-MAGE. In the presence of the DC-SIGN-blocking Ab AZN-D1, cross-presentation of the hD1-SA-gp100 construct was significantly reduced and was comparable to that of the nontargeted constructs 5g1.1-SA-gp100 and gp100(272–288) (Fig. 4 D).

To evaluate the potential of Tat, DPV3, and polyR in mediating cross-presentation of Ag, SA was loaded with a 2-fold molar excess of a biotinylated CPP together with a 2-fold molar excess of a biotinylated gp100(272–288) or control (MAGE) peptide. As a control for cross-presentation of nonspecifically endocytosed, nontargeted Ag, SA was loaded with a 2-fold molar excess of biotinylated gp100(272–288) and a 2-fold molar excess of control MAGE peptide (SA-gp100-MAGE). iDCs were incubated for 20 h with the targeting constructs at 0.05 μM (corresponding to 0.1 μM gp100 and/or MAGE peptide) in the presence of the TLR ligands R848 and poly(I:C). SA-gp100-MAGE was cross-presented to TCR-transduced Jurkat cells, and cross-presentation efficiency was comparable to that of the free gp100(272–288) peptide (Figs. 5 and 4,D, respectively). As expected, DCs targeted with a construct harboring the CPP polyR instead of gp100 (SA-MAGE-polyR) failed to induce a T cell response. Surprisingly, replacement of the control MAGE peptide with Tat (SA-gp100-Tat) failed to enhance cross-presentation of the gp100 peptide, while the slight increase in cross-presentation seen with DPV3 (SA-gp100-DPV3) was not significant. In contrast, replacement of the control MAGE peptide with polyR (SA-gp100-polyR) increased the efficiency of cross-presentation of the gp100 peptide (Fig. 5). This corresponded to the finding that polyR was the most efficient CPP for targeted delivery of protein to DCs (Fig. 2, A and D).

To directly compare CPP-mediated targeting to DC-SIGN targeting, we loaded the anti-DC-SIGN Ab conjugate hD1-SA and the control Ab conjugate 5g1.1-SA with a 2-fold molar excess of gp100(272–288) in combination with a 2-fold molar excess of either the MAGE control peptide or polyR. Subsequently, iDCs were incubated for 20 h in the presence of TLR ligands R848 and poly(I:C) with various concentrations of the targeting constructs. DCs cross-presented the nontargeted control 5g1.1-SA-gp100-MAGE to Jurkat T cells with a lower efficiency than the targeted hD1-SA-gp100-MAGE and 5g1.1-SA-gp100-polyR (Fig. 6). No differences in Ag presentation were detected between DCs targeted with the DC-SIGN Ab (hD1-SA-gp100-MAGE) and DCs targeted with the CPP polyR (5g1.1-SA-gp100-polyR). Combining CPP-mediated targeting and DC-SIGN targeting in one construct (hD1-SA-gp100-polyR) did not further enhance cross-presentation of the gp100 peptide. These data imply that CPPs merely target conjugated Ags to the cell surface and do not enhance class I presentation by facilitating delivery of Ags to the cytosol.

Our data show that the most potent targeting CPP used in our studies, polyR, is as efficient as a targeting Ab against DC-SIGN in inducing cross-presentation of a conjugated Ag by DCs. In addition, adding polyR to the DC-SIGN targeting construct does not further enhance cross-presentation of the conjugated Ag. Thus, there seems to be no advantage in using CPPs for cytosolic delivery of Ag over targeting the Ag to a DC surface receptor such as DC-SIGN, and this suggests that CPPs do not significantly affect endosomal escape of Ags.

Several studies report successful transduction of protein Ag into DCs using CPPs (6, 7, 8). DCs incubated ex vivo with Tat-containing OVA protein induce OVA-specific CTL responses in mice, in contrast to DCs incubated with OVA alone (6, 7). In addition, a study by Batchu et al. demonstrates that human DCs incubated with the tumor-associated Ag NY-ESO-1 are less potent in inducing Ag specific CTL in vitro than DCs incubated with a Tat-NY-ESO-1 fusion protein (8). It was suggested that Tat enhanced cross-presentation in these studies by facilitating endosomal escape of conjugated Ags, which seems consistent with numerous studies reporting efficient CPP-mediated delivery of functional proteins into nucleus and cytoplasm (for a review, see Ref. 12 by Melikov and Chernomordik). In contrast, our findings together with several other studies (37, 38, 39, 40, 41) show that fluorescently labeled CPP-protein conjugates localize in endosomes and not in the cytoplasm and nucleus of cells. These apparent inconsistencies most likely reflect differences in the sensitivity of the detection assays used to detect CPP-protein within cellular compartments. A relatively large number of fluorescently labeled protein molecules must be present in the cytoplasm to be able to successfully detect them. In contrast, many studies reporting CPP-mediated delivery of active protein to cell cytoplasm and nucleus are based upon powerful amplification cascades that will efficiently detect endosomal escape of only a few active molecules (12).

In our studies, Tat targets SA to DCs but does not enhance cross-presentation of a conjugated gp100 peptide, whereas polyR significantly increases cross-presentation of this peptide. This probably reflects more efficient internalization of polyR constructs as indicated by the flow cytometry and confocal microscopy data. Apparently, the additional effect of targeting via Tat does not increase the basal level of Ag uptake at these peptide concentrations to such an extent that it affects Ag presentation.

DC-SIGN has been shown previously to promote class I-restricted presentation of exogenous HIV-1 Ag (42). In this study, we convincingly show that Ab-mediated targeting of Ag to DC-SIGN results in cross-presentation. In our in vitro experiments, DCs exogenously loaded with peptide induce stronger T cell responses than DCs targeted with targeting constructs, reflecting the inefficiency by which endocytosed Ags are shuttled into the MHC class I presentation pathway. However, Ag targeted to the DC receptor DEC-205 has been shown to induce stronger CD8⁺ T cell responses in vivo than DCs exogenously loaded with class-I restricted peptide (9). These differences between in vitro and in vivo experiments might reflect the relatively short time period during which exogenously loaded MHC-peptide complexes remain on the cell surface, whereas Ags targeted to DCs are gradually degraded and their derived peptides are loaded onto MHC molecules over a prolonged period of time (3).

The Ab-mediated targeting experiments demonstrate that the design of the targeting construct markedly affects the level of cross-presentation induced. In contrast to the gp100(272–288) peptide, cross-presentation of the gp100(277–291)*SAMA peptide could not be detected, suggesting the gp100(277–291) peptide was less efficiently processed. Furthermore, the minimal epitope gp100(280–288) incorporated into scD1 hardly induces a T cell response, whereas scD1-gp100(277–291)*SAMA and especially hD1-SA-gp100-MAGE induce stronger responses even at lower concentrations. This likely reflects differences in processing of the various constructs by the cell, as the processing efficiency of the presented gp100(280–288) peptide is not only determined by the sequence of the peptide itself, but also by its flanking amino acid residues, which differ between the various constructs (43). Furthermore, the processing efficiency could have been affected by the various methods of peptide-Ab conjugation, either by peptide bond, biotin-SA binding, or a chemical cross-linker. It is of crucial importance for vaccination strategies to optimize the methods by which Ags and targeting moieties are combined. CPPs have the advantage in that they are produced at relatively low cost and are easily introduced into multiepitope or protein Ags. Although it is technically feasible to produce Abs harboring multiple peptides or complete proteins (10), large-scale production often proves to be difficult. This might be overcome by conjugating the Ab to Ag delivery systems such as liposomes or polymer particles. In this way the targeting moiety, multiple Ags, and DC maturation stimuli could be combined into a single particle (44).

Remarkably, over a period of 5 h DCs accumulate merely 1.4–1.5 times the amount of DC-SIGN specific Ab that they can maximally bind. Thus, ligand processing by DC-SIGN seems rather slow, especially because the related C-type lectin receptor DEC-205 and the mannose receptor have been described to resurface within 1 h following the binding of ligand (45). However, Guo and coworkers have shown that rat fibroblasts transfected with human DC-SIGN process 10 times more glycosylated protein within 2 h than the number of receptors expressed on the cells (46). Possibly, the ligand processing speed of DC-SIGN is ligand specific or dependent on variations in intracellular DC-SIGN routing between different cell types. Our experiments show that SA-polyR accumulates in DCs more rapidly than hD1, although this does not result in increased Ag presentation. Likely, this is a consequence of the distinct mechanisms responsible for internalizing CPPs and the targeting Ab. Upon uptake by DCs, internalized SA-Tat, SA-DPV3, and SA-polyR constructs showed a high degree of colocalization. Many types of endocytosis have been implicated in the uptake of CPPs by cells, including clathrin-dependent endocytosis (47), caveolin-dependent endocytosis (38, 48), and raft-dependent macropinocytosis (49). These contrasting findings might be explained by variations in experimental setup, such as the type of cargo attached to the CPP and the cell type used. Our data suggest that Tat, DPV3, and polyR use identical or at least largely overlapping endosomal routes to enter DCs. Internalized hD1 Ab was localized in a different set of endosomes than SA-polyR, demonstrating that proteins endocytosed via DC-SIGN follow a different route. These findings are supported by two recent publications showing that DC-SIGN ligands enter cells via a clathrin-dependent (50) while the CPP polyR enters cells via a clathrin-independent, flotillin- and dynamin-dependent internalization process (51). In our study we showed that both endosomes carrying SA-polyR and hD1 eventually delivered their antigenic cargo to the lysosomal compartment.

The aim of our study was to compare the efficiency of DC cross-presentation by using different targeting strategies. The results do not predict whether this level of cross-presentation is sufficient for cross-priming of CTL responses in vivo. To initiate a proliferative response, naive T cells require a higher amount of MHC-peptide complexes on the APC surface than primed T cells (52) and, possibly, our Jurkat T cells. However, various preclinical studies on Ab-mediated targeting of C-type lectins show the priming of CTL responses in vivo (44). Furthermore, DC-SIGN targeting induces primary T cell responses in mice reconstituted with human immune cells (53). Future studies will establish whether DC-SIGN targeting induces the priming of CTL responses in vivo.

Clinical trials involving ex vivo loading of DCs with tumor-associated Ags have shown that this technique induces CTL responses against the Ags used (2). Our results show that ex vivo loading of DCs with class I-restricted Ags is improved by conjugating them to polyR or humanized Abs directed against DC-SIGN. As ex vivo loading of DCs with Ags is a very laborious and costly procedure, future DC vaccination therapies will benefit from strategies involving the targeting of Ags to DCs in vivo (44). The high specificity of Abs for their target molecules allows the targeting of conjugated Ags to DCs specifically, which cannot be accomplished by using CPPs. Thus, Abs directed against receptors preferentially expressed by DCs hold great promise for future vaccination strategies.

We thank Dr. R. Debets at the Erasmus Medical Center-Daniel den Hoed, Rotterdam, The Netherlands for providing the JE6.1 fl296 cells and the Microscopic Imaging Center of the Nijmegen Centre for Molecular Life Sciences for use of their facilities.

Dayang Wu and Anita Reddy are employed by Alexion Pharmaceuticals and Anke Kretz-Rommel is employed by Alexion Antibody Technologies, companies whose potential products were studied in the present work.