Cross-presentation of exogenous Ags in MHC class I molecules by dendritic cells is the underlying basis for many developing immunotherapies and vaccines. In the phagosome-to-cytosol pathway, Ags in phagocytosed particles must become freely soluble before being exported to the cytosol, but the kinetics of this process has yet to be fully appreciated. We demonstrate with a yeast vaccine model that the rate of Ag release in the phagosome directly affects cross-presentation efficiency, with an apparent time limit of ∼25 min postphagocytosis for Ag release to be productive. Ag expressed on the yeast surface is cross-presented much more efficiently than Ag trapped in the yeast cytosol by the cell wall. The cross-presentation efficiency of yeast surface-displayed Ag can be increased by the insertion of linkers susceptible to cleavage in the early phagosome. Ags indirectly attached to yeast through Ab fragments are less efficiently cross-presented when the Ab dissociation rate is extremely slow.
Vaccines that stimulate Ab production have enjoyed success for the past century, but the development of vaccines that generate effective cellular immune responses, in particular, CD8+ cytotoxic lymphocytes, remains a challenge. Such vaccines provide hope for the prevention and treatment of cancer (1, 2, 3) as well as viral diseases like HIV (4), hepatitis C (5), and herpes simplex virus (6).
Part of this challenge arises from the fact that peptide-MHC class I complexes required to prime CD8+ T cells are generally produced from the endogenous proteins of APCs, principally dendritic cells (DCs)3 (3). Although alternative strategies such as adoptive transfer of lymphocytes (7), DNA vaccines (8), and vaccination with exact peptide epitopes exist (9), cross-presentation—the process by which peptides derived from exogenous Ags are displayed with MHC class I molecules—remains central to most immunotherapy strategies. Early reports of cross-presentation presented conflicting results as to whether this process required TAP and was sensitive to proteasome inhibitors (10), or was TAP-independent and sensitive to protease inhibitors (11, 12). At least two distinct mechanisms for cross-presentation emerged: the phagosome-to-cytosol pathway and the vacuolar pathway (13). In the vacuolar pathway, peptides are generated from internalized Ags by the action of endolysosomal proteases; the peptides may then meet up with recycling MHC class I molecules. In the phagosome-to-cytosol route, phagocytosed Ags escape to the cytosol to be processed by proteasomes and are loaded onto nascent MHC class I molecules with the aid of TAP.
Quantitative mechanistic details about these pathways for cross-presentation are still missing from the picture which, if discovered, could lead to the development of more effective vaccines. We hypothesize that with particulate Ags that are cross-presented via the phagosome-to-cytosol pathway, Ag release from the particles may be a rate-limiting step. To our knowledge, Ag release rates within the phagosome have not been comprehensively studied. In this study, we explore the effects of altering Ag release kinetics on cross-presentation efficiency using a yeast vaccine model.
Recombinant yeasts show promise as vaccine candidates in mouse models (14, 15) and in human blood cell assays (16, 17, 18). In particular, S. cerevisiae is attractive because it is nonpathogenic, well-characterized, and is a strong adjuvant—Zymosan, a cell wall preparation of this yeast, has been a valuable tool in immunology for >50 years (19). S. cerevisiae potently induced the maturation of murine bone marrow-derived DCs and secretion of IL-12 (14). In our hands (our unpublished data) and others (16), human monocyte-derived DCs were similarly activated. When the recombinant Ag is expressed intracellularly in yeast, the rate of Ag release is primarily dictated by the rate of yeast cell wall degradation. With yeast surface display technology (20), in contrast, Ag is expressed on the cell wall exterior, permitting the release kinetics to be manipulated.
Materials and Methods
Human HLA-A*0201 monocytes were obtained from two sources, purified either by counterflow centrifugal elutriation (Advanced Biotechnologies) or negative magnetic cell sorting (Biological Specialty Corporation). Similar results were obtained with both sources. The monocytes were aliquoted into vials and cryopreserved in 90% FBS, 10% DMSO. For each experiment, one or more vials were thawed and washed in C10 medium: RPMI 1640 with 10% FBS, 2 mM l-glutamine, 10 mM HEPES, 1 mM sodium pyruvate, 1× nonessential amino acids, 50 μM 2-ME, and Primocin (InvivoGen). Unless otherwise indicated, medium components were from HyClone; low endotoxin products were chosen where available. A total of 4–5 × 106 monocytes were cultured per well of a 6-well plate in 2.5 ml C10 medium supplemented with 1000 U/ml each of IL-4 and GM-CSF (C10GF; cytokines from R & D Systems). After 2 and 4 days of culture, each well was topped up with 0.5 ml C10GF; after 6 days of culture, floating and loosely adherent immature monocyte-derived DCs were harvested by gentle resuspension.
Vials of a human CD8+ T cell line specifically recognizing the peptide NLVPMVATV in the context of HLA-A*0201 were purchased from ProImmune. Each vial was thawed and cultured overnight in RPMI 1640 with 10% FBS and 5 ng/ml IL-2 and used the next day.
Yeast surface display
Plasmids for yeast surface display were based on pCT-CON (21) and were transformed into EBY100 (20), a strain that expresses Aga1p under galactose induction, using the Frozen EZ Yeast Transformation II kit (Zymo Research). Yeast colonies were cultured to mid-log phase at 30°C in selective SD-CAA medium (2% dextrose, 0.67% yeast nitrogen base, 0.5% casamino acids, 0.1 M sodium phosphate, pH 6.0) and then induced in SG-CAA (SD-CAA with galactose replacing dextrose) for 48 h at 20°C. Single copies of some expression cassettes were integrated into the EBY100 yeast chromosome using the integrating shuttle vector pRS304 (22). The resulting yeast strains were grown up in rich YPD medium (1% yeast extract, 2% peptone, and 2% dextrose) and induced in YPG (1% yeast extract, 2% peptone, and 2% galactose) for 36 h at 20°C. Yeast medium nitrogen sources were obtained from BD Biosciences. Surface display levels were measured by flow cytometry with chicken α-c-myc (Invitrogen) or 9e10 mAb (Covance). The number of copies per yeast cell was estimated by comparison with Quantum Simply Cellular beads (Bangs Laboratories).
After 6 days of differentiation, immature monocyte-derived DCs were seeded in 96-well round-bottom plates at 1 or 2 × 105 cells in 200 μl C10GF per well. Appropriate numbers of yeast cells (measured by OD at 600 nm with 1 OD ≈ 107/ml) were rendered nonviable by UV-irradiation (2 × 1000 J/m2 in a Stratalinker; Stratagene), pelleted by centrifugation, and added to the DCs. For inhibition experiments, DCs were preincubated with Z-FL-COCHO (10 μM; Calbiochem), lactacystin (5 μM; Calbiochem), or chloroquine (25 μM) for 1 h before yeast samples were introduced. Twenty-four hours later, half the medium was replaced with a T cell suspension, with 0.7–1 × 105 T cells per well. Following 4 h of coculture, the contents of each well were transferred to tubes for labeling with Miltenyi’s IFN-γ secretion assay kit according to the recommended protocol. Briefly, cells were labeled with a bispecific Ab that captures secreted IFN-γ on the cell surface during a 45 min incubation period in medium at 37°C, and then labeled on ice for 30 min with α-CD8-FITC (BD Biosciences) and α-IFN-γ-PE (Miltenyi Biotech). In experiments involving FITC-conjugated yeast, α-CD8-Alexa Fluor 647 (BD Biosciences) was substituted. The percentage of CD8+ cells that were IFN-γ+ was determined by flow cytometry (Coulter Epics XL or BD FACSCalibur). The cut-off PE fluorescence was set for each experiment such that ∼0.5% of T cells were IFN-γ+ in a negative control sample (no yeast or peptide). The positive control with 1 μM of the extended peptide ARNLVPMVATVQGQN (synthesized by GenScript) resulted in 45–70% IFN-γ+ T cells.
Yeast intracellular expression
Intracellular expression of the same fusion protein as is expressed by surface display was achieved by deleting the signal peptide of Aga2p, followed by transformation into BJ5464α (Yeast Genetic Stock Center). BJ5464α is isogenic to the parent strain of EBY100 and lacks the galactose-inducible Aga1p gene. The resulting colonies were grown up in SD-CAA and induced in SG-CAA for 12 h at 30°C.
Slot blot comparison of Ag levels
A total of 6 OD.ml of each yeast culture was washed with PBS, resuspended in 300 μl 25 mM Tris(2-carboxyethyl)phosphine hydrochloride (TCEP; Soltec Ventures) in PBS, and incubated for on ice for 30 min. The proteins released into solution by the reducing agent were pooled with those from a second 30-min extraction with 25 mM TCEP. The yeast pellets were then washed with spheroplast buffer (50 mM Tris-HCl, pH 7.5, 1.4 M sorbitol, and 40 mM 2-ME), incubated with 2.4 U Zymolyase (Zymo Research) in 120 μl spheroplast buffer containing a protease inhibitor mixture (Roche) for 15 min at 37°C, and boiled in 2% SDS for 5 min. The protein extracts were blotted onto nitrocellulose membrane with a slot-blotting apparatus (Bio-Rad). The membrane was blocked with 5% milk powder, incubated with 9e10 ascites fluid (Covance) followed by goat α-mouse-HRP (Pierce), developed with SuperSignal West Dura substrate (Pierce), and imaged on a FluorS Imager (Bio-Rad).
Surface display Ag dose normalization
In experiments where different linkers were used to surface-display Ag, several cultures of each yeast sample were induced, and cultures with mean Ag levels within 10% of each other were selected to minimize the effect of variable Ag dose on cross-presentation. However, the variability in expression level across the panel of initial constructs (deleted linker, unchanged, and C1–5) was too high for this approach to be satisfactory. Therefore, each yeast sample was mixed with the appropriate amount of EBY100 yeast to normalize the Ag dose while maintaining the 20:1 ratio of yeast to DCs.
Measuring linker susceptibility to Cathepsin S
A total of 0.2 OD.ml of each yeast sample was washed and incubated with the indicated amounts of recombinant human Cathepsin S (CatS; Calbiochem) in 100 μl PBS at 37°C. The yeast samples were washed and labeled with 12CA5 mAb (α-HA; Roche) and chicken α-c-myc, followed by goat α-mouse-PE (Sigma-Aldrich) and goat α-chicken-Alexa Fluor 488 (Invitrogen). The mean c-myc fluorescence of the HA+ population was compared against that of yeast samples that had not been treated with CatS.
DCs (2 × 105/well) were seeded in 96-well round-bottom plates, with separate plates for each time point. After adding the yeast samples (5 × 105/well), the plates were immediately centrifuged briefly (200 × g, 1 min) to settle the yeast and were returned to the incubator. At each time point, a plate was placed on ice and 90% of the medium in each well was replaced with cold radioimmunoprecipitation assay (RIPA) buffer (Sigma-Aldrich). The well contents were moved to tubes, vortexed to promote cell lysis, and centrifuged to pellet the released yeast. The yeast was washed with RIPA buffer and PBS with 0.1% BSA before being labeled for HA and c-myc epitopes as described above.
Fluorescein-binding single-chain variable fragment (scFvs)
The fluorescein-binding scFvs used here were products of directed evolution for decreased dissociation rate using yeast surface display (23). These scFvs were subcloned into pRS316-based plasmids with an improved α mating factor pre-pro sequence (J. A. Rakestraw, A. Piatesi, and K. D. Wittrup, unpublished observations). Codons encoding the extended peptide ARNLVPMVATVQGQN were inserted between the C terminus and the c-myc epitope. The resulting constructs were transformed into the protein disulfide isomerase-overexpressing yeast strain YVH10 (24) together with a dummy plasmid bearing the trp nutritional marker. Transformants were grown up in SD-CAA and induced in YPG containing 0.1 M sodium phosphate, pH 6.0 for 3 days at 20°C. The culture supernatants containing ∼10 mg/L of scFv-Ag were adjusted to pH 7.4 and dialyzed against PBS.
UV-irradiated BJ5654α yeast cells were washed three times in 0.4 M sodium carbonate, pH 8.4, and resuspended in 10 μl/OD.ml of a freshly prepared 0.15 mg/μl solution of fluorescein-PEG-NHS (MW 5000; Nektar) in sodium carbonate buffer. The reaction was allowed to proceed for 30 min at room temperature, after which the yeast was washed six times with PBS containing 0.1% BSA. Fluorescein-conjugated yeast was loaded with Ag by incubation with scFv-Ag culture supernatants (1 ml/107 yeast) for 1 h on ice. Flow cytometry analysis (c-myc labeling) of the loaded yeast showed that the Ag levels mediated by 4M2.3, 4M3.12, and 4M4.5 were within ∼5% of each other, but the level of 4M5.3-Ag was ∼15% higher. Labeling fluorescein-conjugated yeast with 4M5.3 fusion protein for 30 min followed by 30 min, 37°C incubation in pH 5.4 PBS containing 0.1% BSA and 1 μM fluorescein-biotin resulted in a final Ag level comparable to that mediated by the other scFvs. This method of Ag level normalization was performed for the cross-presentation assay. In addition, to reduce Ag loss before phagocytosis, the plate was centrifuged (at 200 × g for 1 min) immediately after addition of the yeast to the DCs.
The mathematical model consisted of the following equations describing the amounts of yeast-bound Ag (Ab), free Ag within the phagosome (Af), and cytoplasmic Ag (Ac) relative to the initial amount:
For 0 < t ≤ tpre, where tpre is the time before phagocytosis For tpre < t ≤ tpre + tmat, where tmat is the time taken for a phagolysosome formation For tpre + tmat < t ≤ 24 h These equations were solved in Matlab with the initial conditions Ab = 1, Af = Ac = 0. For a given set of parameter values, the final value of Ac (at 24 h) was determined for a wide range of koff values. Ten logarithmically spaced values spanning three orders of magnitude were tested for each of the following parameters: c1 (0.1 – 10), c2 (10 – 1000), c3 (10 – 1000), kesc (0.01 – 1 min−1), and kdeg (0.01 – 1 min−1). An optimal t1/2 = ln(2)/koff was found to exist between 10 and 105 min for all 105 possible combinations of parameter values, which should span all reasonable biological values. The time parameters, tpre and tmat, were fixed at 30 min and 20 min, respectively. The former value is an estimate but is not a critical value because it is always multiplied by c1, which was varied widely. The latter value was deemed reasonable based on the postphagocytosis time course analysis we performed.
Yeast surface-displayed Ag is cross-presented to CD8+ T cells
We selected the well-characterized HLA-A*0201-restricted peptide NLVPMVATV (N9V), derived from CMV phosphoprotein pp65 as our model Ag, for which cognate CD8+ T cells are available commercially. To ensure proper Ag processing, we included its native flanking sequences in the yeast surface display construct, in the form of the 15-mer ARNLVPMVATVQGQN that was consistently immunogenic in HLA-A*0201, CMV-positive individuals (25). The yeast surface display construct consisted of a fusion of this extended peptide to the yeast mating adhesion receptor subunit Aga2p via a (G4S)3 linker, with HA and c-myc epitope tags for detection purposes (Fig. 1⇓A). We created the yeast strain EBYN9V with coinducible chromosomal copies of this construct and Aga1p, with expression resulting in ∼120,000 copies/cell of the Aga2p-N9V fusion anchored to the yeast cell wall by disulfide bonds.
To test for cross-presentation, EBYN9V yeast were added to HLA-A*0201 monocyte-derived DCs at various ratios. The DCs avidly phagocytosed the yeast with an average maximum “capacity” of ∼20 yeast per DC (numbers of unphagocytosed yeast rose sharply at higher ratios). Twenty-four hours later, the DCs were cocultured for four hours with a CD8+ T cell line specifically recognizing the N9V/HLA-A*0201 complex. An IFN-γ secretion cell capture FACS assay was performed on the T cells to quantify the percentage of cells that had been activated as a result of cross-presentation by the DCs. As shown in Fig. 1⇑B, EBYN9V yeast resulted in dose-dependent cross-presentation at levels much higher than the background caused by EBY100 yeast lacking the N9V surface display construct. We decided to use the 20:1 yeast:DC ratio for future cross-presentation experiments to minimize the signal-to-noise ratio; note that the concentration of peptide equivalents at this dose is only 4 nM.
Surface-displayed Ag is cross-presented much more efficiently than intracellular Ag
We next compared cross-presentation of yeast surface-displayed Ag to Ag expressed inside the cytosol of yeast. By deleting the signal peptide, the same Aga2p-N9V fusion protein was expressed intracellularly in yeast. At the same 20:1 yeast:DC ratio, cross-presentation resulting from intracellular Ag was only half that from surface-displayed Ag (Fig. 1⇑C). This result was obtained even though the expression level of intracellular Ag was 20–30 times the surface display level. In the slot blot in Fig. 1⇑D, the amount of Ag in an intracellular extract of 1 × 106 cytosolically expressing yeast is equivalent to the amount of Ag reduced off the cell walls of 2–4 × 107 surface-displaying yeast. The blot also demonstrates that with both yeast cultures, Ag expression was restricted to the intended location.
To try to understand the marked difference in cross-presentation efficiency between surface-displayed and intracellular Ags, we studied the effects of inhibitors of either the phagosome-to-cytosol route or the vacuolar route. Cross-presentation of surface-displayed Ag was strongly inhibited by lactacystin, a proteasome inhibitor, whereas chloroquine, which raises the endolysosomal pH, had no inhibitory effect and was actually slightly beneficial (Fig. 1⇑C). We deduced that the phagosome-to-cytosol pathway is the major mechanism of cross-presentation with yeast surface-displayed Ag. Chloroquine has been observed to increase the cross-presentation efficiency of soluble Ags, possibly because it increases membrane permeability and, hence, Ag escape into the cytosol (26), and may be having a similar subtle effect here. Cross-presentation of intracellular Ag was inhibited by both lactacystin and chloroquine (Fig. 1⇑C). It is unclear whether cross-presentation of intracellular Ag proceeds by a combination of the phagosome-to-cytosol and vacuolar routes, or whether only the phagosome-to-cytosol route is involved, with chloroquine reducing the rate at which the yeast cell wall was breached, thus slowing Ag export into the DC cytosol. In any case, it is clear to us that having Ag exposed on the yeast surface provides a significant advantage for cross-presentation due to greater accessibility to the DC cytosol compared with having Ag trapped by the thick yeast cell wall.
Manipulating the kinetics of Ag release with different linkers
We hypothesized that the rate at which Ag is released from a phagocytosed particle influences the efficiency of cross-presentation, because Ag release is a necessary step before export into the cytosol can occur. The yeast surface display model provided an excellent means to test this hypothesis. We conjectured that with EBYN9V yeast, the N9V antigenic peptide could be released from the yeast cell wall by proteolysis in the phagosome or by reduction of the disulfide bonds tethering Aga2p to Aga1p. The rate of the former mechanism could potentially be manipulated by including protease recognition sites N-terminal to the antigenic peptide. We targeted Cathepsin S (CatS) because unlike most other cathepsins that are active only in acidic conditions found later in phagosomal maturation, its operating range extends from pH 5.0 to 7.5 (27). Furthermore, phagosomes in macrophages and DCs fuse preferentially with endocytic compartments enriched in CatS, with CatS activity detected in ten-minute-old phagosomes (28).
Five potential CatS recognition sites culled from the literature are listed in Table I⇓. In some cases, four amino acid residues on either side of a known CatS cleavage point were used. These sequences, termed C1 to C5, were each inserted individually between the (G4S)3 linker and the extended antigenic peptide. An additional construct was created where the (G4S)3 linker, a suspected CatS cleavage site, was deleted. To test whether these sequences were recognized in their new context, yeast expressing the modified plasmid constructs were incubated with recombinant CatS and analyzed for loss of the c-myc epitope. CatS had negligible effect on HA epitope levels, indicating that the polypeptide chains linking together HA, Aga2p, Aga1p and the cell wall remained intact. Although the addition of C1, C2, and C5 increased CatS cleavage, C3 and C4 had the opposite effect and were apparently not recognized and/or disrupted a preexisting recognition site (Fig. 2⇓A). Deleting the (G4S)3 linker altogether conferred the greatest resistance to CatS cleavage. When yeast with these different linker sequences were phagocytosed by DCs, the resulting pattern of cross-presentation was strikingly similar to the pattern of CatS cleavage (Fig. 2⇓B). Performing Spearman’s rank correlation on the rankings listed in Table I⇓, CatS susceptibility and cross-presentation efficiency were found to be positively correlated at the significance level of p < 0.05, supporting our hypothesis that faster Ag release within the phagosome results in more efficient cross-presentation. Note that one would not expect the rank order correspondence to be perfect because it is likely that other cathepsins, which may share some degree of substrate specificity with CatS, also play a role in Ag release toward cross-presentation.
In an attempt to further increase Ag release rates by CatS, we created constructs with tandem repeats of C1 and C2 sequences. Tandem repeats of C2 did not further enhance CatS susceptibility (data not shown), but the rate of CatS cleavage increased with the number of tandem copies of C1 (Fig. 2⇑C). Consistent with our hypothesis, there was a corresponding increase in cross-presentation efficiency (Fig. 2⇑D).
Ag released by CatS is processed by proteasomes
Yeast strains were created with chromosomally integrated expression cassettes for the constructs with the deleted (G4S)3 linker, with a single C1 insertion, and with four tandem C1 repeats, selected for being representative of the entire range of CatS susceptibilities. These yeast strains displayed the expected rank order of cross-presentation efficiency across a range of yeast to DC ratios: (C1)4 > C1 > deleted (Fig. 3⇓A). The differences in cross-presentation efficiency were largely diminished when the DCs were pretreated with a specific CatS inhibitor (10 μM Z-FL-COCHO) (Fig. 3⇓A), showing that CatS cleavage was indeed primarily responsible for these differences. Because the disparities in cross-presentation efficiency were not completely eliminated, it is possible that other proteases may have played minor roles in Ag release; alternatively, CatS was not completely inhibited. The gains in cross-presentation efficiency with increased CatS susceptibility were not due to the vacuolar route becoming dominant; instead, cross-presentation of all three strains remained inhibited by lactacystin and unaffected or slightly improved by chloroquine (Fig. 3⇓B), suggesting that the Ag released by CatS moved from the phagosome to the cytosol.
Evidence of a time window for productive Ag release
With these integrated expression yeast strains, at least 98% of the yeast cells expressed the surface-displayed Ag (compared with ∼75% for transformed yeast subject to plasmid loss), so Ag loss that occurred after phagocytosis could be clearly distinguished. We developed an assay for monitoring in vivo Ag processing involving lysing the DCs at various time points after phagocytosis was initiated, followed by labeling the released yeast with α-HA and α-c-myc Abs. Ag release by proteolytic cleavage of the linker C-terminal to the HA epitope (or less likely, cleavage of the antigenic peptide or the c-myc epitope) results in HA+, c-myc − yeast. We observed that between 5 and 25 min postphagocytosis, this population was largest with the (C1)4 linker and smallest with the deleted linker (Fig. 3⇑C). This is consistent with CatS attacking the linkers at different rates during this early stage of phagosomal maturation, and supports the notion that early proteolytic release was responsible for the variation in cross-presentation efficiency. During the first 20 min or so, very little Ag was released in a way that would cause the loss of both epitopes (Fig. 3⇑D), such as disulfide bond reduction or enzymatic attack of the yeast cell wall, Aga1p, or Aga2p. Between 25 and 30 min postphagocytosis, the double-negative population started rising rapidly, suggesting that the phagosomes had fused with late endosomes or lysosomes that provided a more acidic environment and a larger complement of active proteases. The differences in Ag loss levels between the three strains diminished at these later time points, and presumably, all the yeast cells would eventually lose their attached Ag. This suggests that Ag release rates early in phagosomal maturation are key to cross-presentation efficiency; Ag released after the 25 min time point may be mostly degraded by lysosomal proteases rather than cross-presented.
Manipulating the kinetics of Ag release using fluorescein-binding scFvs
One of the earliest applications of yeast surface display was to perform directed evolution of a fluorescein-binding scFv of an Ab to select for mutants with increased affinity (23). The existence of a pool of mutants spanning over four orders of magnitude in dissociation rate provided the opportunity to manipulate Ag release kinetics in a manner distinct from proteolytic release. We first loaded yeast expressing these scFv mutants with fluorescein-tagged extended peptides, but the surface display levels of these scFvs were low and variable. We then inverted the topology and loaded fluorescein-conjugated yeast with scFv-ARNLVPMVATVQGQN-c-myc fusion proteins, with the attendant advantage that the delivered Ag dose would be independent of the protein expression level. Four scFvs, with attributes listed in Table II⇓, were selected from the mutant pool to be produced as scFv-Ag fusions.
We developed a mathematical model based on the schematic in Fig. 4⇓A to predict the cross-presentation outcome. In this model, the scFv-Ag fusion dissociates from yeast cells in three stages, the first encompassing the handling steps and time lag before phagocytosis by DCs, the second being the estimated 20-min window in phagosome maturation before the transition into a phagosolysosome, the third stage. We made the simplifying approximation that for all scFvs, the dissociation rates during these three stages could each be expressed as a proportionality constant multiplied by the measured dissociation rate at neutral pH and 25°C (koff). ScFv-Ag released before phagocytosis is assumed to be lost, whereas scFv-Ag released in the phagosome escapes to the cytosol at a rate kesc. When the phagosome matures into a phagolysosome, proteases degrade both yeast-bound and free Ag with the rate constant kdeg. Protease activity that could cause Ag release rather than destruction of the epitope was neglected. We assumed that the final level of cross-presentation is proportional to the amount of Ag that escapes to the DC cytosol. The solution to the ordinary differential equations comprising this model describes a bell-shaped curve for cytosolic Ag vs koff (Fig. 4⇓B). The existence of a koff value optimal for cross-presentation was a property of the model that was robust to simultaneous parameter variations spanning three orders of magnitude. At very high koff values, most of the Ag is lost before phagocytosis, whereas at very low koff values, little Ag is freed in the phagosome and the majority is degraded in the phagolysosome.
If Ag release rates in the phagosome did not affect cross-presentation, we would expect to see cross-presentation efficiencies rising monotonically with decreasing koff due to the increased dose taken up by the DCs. Instead, the results of three independent cross-presentation experiments confirmed the existence of the model-predicted optimum, with the femtomolar fluorescein binder 4M5.3 resulting in less cross-presentation than the lower affinity 4M4.5 (Fig. 4⇑C). When the yeast was extracted from lysed DCs 15 min postphagocytosis, Ag loss was shown to decrease with increasing affinity (Fig. 4⇑D). Although the dissociation half-time of 4M5.3 is almost two hours in vitro even at pH 5.4, 37°C, proteolysis by CatS or other early phagosomal proteases contributed to a baseline level of Ag release not accounted for in the model; hence, the 4M5.3-Ag fusion gave rise to higher than expected levels of cross-presentation.
We have shown by three distinct types of experiments that Ag release kinetics from phagoctosed yeast cells influences the efficiency of cross-presentation occurring via the phagosome-to-cytosol route. First, protease-accessible Ag exposed on the yeast external surface is cross-presented much more efficiently than Ag trapped inside the tough cell wall. Second, increasing the susceptibility to CatS cleavage of the linker between the Ag and its cell wall anchor resulted in increased cross-presentation efficiency. Third, there exists an optimal affinity for Ab fragments used to attach Ag to the yeast surface, with extremely low dissociation rates being detrimental for cross-presentation efficiency.
Although this evidence was obtained with yeast, we believe that the hypothesis applies to other particulate vaccines, because Ag release in the phagosome logically precedes phagosome-to-cytosol translocation. Indeed, in one of the earliest demonstrations of cross-presentation, it was observed that OVA passively adsorbed to latex beads was more efficiently cross-presented than OVA conjugated to the same beads, although no explanation was put forth at that time (29). Particulate vaccines have advantages over soluble vaccines in that they are not diluted by diffusion, and are targeted to phagocytic professional APCs. It behooves particulate vaccine developers to take Ag release kinetics into consideration and to perform research into methods of optimizing these kinetics for their vaccine systems. An alternative strategy is to bypass the requirement for freeing Ag in the phagosome, as illustrated by microparticles that released Ag directly in the cytosol (30).
Our analysis of Ag loss occurring postphagocytosis suggests that there exists a limited time window for productive Ag release. In the case of yeast surface display, it appears that Ag freed after the 25-min time point did not contribute significantly to cross-presentation. The 25-min time point coincided with a sudden rise in Ag loss by means other than cleavage of the linker, possibly indicating phagosome fusion with late endosomes or lysosomes. With macrophages that had phagocytosed yeast, the phagosomal pH took 20–25 min to decrease to a minimum of ∼5.0 (31). A time course analysis of protease activity in murine DC phagosomes showed that Cathepsins B, L, and Z jumped in activity between 20 and 30 min after phagocytosis (28). In contrast, Savina et al. (32) recently showed that phagosomes in murine DCs do not acidify significantly for 3 h. The connection between phagosome acidification and acquisition of proteolytic activity remains unclear, as well as how these may be affected by DC status and the nature of the particle.
The three constructs with different linkers that we compared displayed the greatest variation in Ag loss during the 10–20 min window, so it is likely that the major source of N9V peptide ultimately cross-presented was Ag freed during this time frame. With endocytosed Ag, it has been suggested that Ag destined for cross-presentation exited early from the endosomal pathway; the bulk of the Ag was colocalized with late endosomal/lysosomal markers after 25 min but did not contribute to cross-presentation (33). In the same study by Palliser et al. (33), aggregated protein Ag was observed to be cross-presented almost ten times less efficiently than the monomeric form. A possible explanation for this result is that aggregates that were not dispersed soon after uptake could not egress to the cytosol and were not cross-presented.
The narrow time window available for Ag release suggests that CatS, unusual among cathepsins for being active at up to neutral pHs, may play a special role in phagosome-to-cytosol cross-presentation. Roles for CatS in the vacuolar route of cross-presentation (30) and class II presentation (34) have previously been identified.
In our mathematical model, after 20 min of phagosome maturation, proteolytic degradation of Ag competed with Ag export into the cytosol. Thus, only a small fraction of Ag released after 25 min or so contributed to cross-presentation. However, we further speculate that phagolysosome formation may in some way close off the means for Ag egress to the cytosol, thus imposing another limit on the time window for productive Ag release. Teleologically, it would make sense for the class II epitopes generated in phagolysosomes to be retained rather than exported to the cytosol. The nature of this phagosomal “pore” remains a mystery, although it appears to have a size limit, with 40 K but not 500 K dextran being translocated (35). Several years ago, at least three studies provided evidence that membranes of the endoplasmic reticulum (ER) contribute to the nascent phagosome (36, 37, 38), leading to speculation that ER-resident proteins like the Sec61 translocon or Der1p (39) may be responsible. ER-phagosome fusion has since been disputed (40), but Cresswell et al. (41) recently demonstrated the involvement of ER retrotranslocation mechanisms in cross-presentation. Solving this mystery would represent a significant advance in the state-of-the-art and may permit the development of techniques to either increase the rate of Ag export to the cytosol or extend the time window for which this mechanism is active. Increasing the transport of Ag to the cytosol in this manner, combined with more complete Ag release during the critical time window, could lead to the development of more effective vaccines designed to raise cellular immunity against virus-infected cells and cancer cells.
We thank J. A. Rakestraw, A. S. Gai, D. J. Irvine, and Jianzhu Chen for their technical suggestions, and H. N. Eisen for critical reading of the manuscript.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 This work was funded in part by National Institutes of Health Grant AI065824, the Ludwig Institute for Cancer Research, and the Massachusetts Institute of Technology-Harvard Centers of Cancer Nanotechnology Excellence Consortium. S.W.H. is a recipient of a scholarship from Singapore’s Agency for Science, Technology and Research, and of the Margaret A. Cunningham Immune Mechanisms in Cancer Research Fellowship.
↵2 Address correspondence and reprint requests to Dr. K. D. Wittrup, Departments of Biological Engineering and Chemical Engineering, Massachusetts Institute of Technology, Building E19-551, 400 Main Street, Cambridge, MA 02139. E-mail address:
↵3 Abbreviations used in this paper: DC, dendritic cell; CatS, Cathepsin S; scFv, single-chain variable fragment; ER, endoplasmic reticulum; RIPA, radioimmunoprecipitation assay.
- Received March 28, 2007.
- Accepted November 19, 2007.
- Copyright © 2008 by The American Association of Immunologists