CP of Ags on MHC I molecules is a well-established phenomenon that has been demonstrated both in vitro
and in vivo
). However, to what extent CP contributes to overall TCD8+
responses is hotly debated (9
), and this is partly due to the fact that CP could not be demonstrated for some antigens that, nonetheless, induce strong TCD8+
). Work by Norbury et al. has shown that the cross-presentation of a model Ag was lost if rapidly degraded within the donor cell (17
). In addition, Wolkers et al. showed that two different determinants were cross-presented when they localized to the mature domain of recombinant protein but not to the signal sequence, which is normally rapidly degraded. On the other hand, DP was not affected by the localization of either determinants to one or the other region of the protein (34
). These works provided the first evidence that changes in the primary structure of an Ag affecting speed of degradation can have strong consequences for efficient CP. However, that the primary structure of an Ag can affect CP and not DP while not involving rapid degradation has not been previously described.
Previous work in other laboratories has shown that altering the flanking sequences of some Ags may affect DP. For example, Shastri et al. (35
) showed that specific residues immediately following the SIINFEKL can affect DP. In the same vein, Eisenlohr and co-workers demonstrated specific negative effects on DP of point mutations immediately preceding or following an influenza nucleoprotein and that deletion of an AA immediately following this determinant was deleterious in the DP of short but not long constructs (36
). In addition, Mo et al. (39
) showed that the specific sequences of the 8 and 5 AA preceding and following SIINFEKL, respectively, can affect DP. All these results imply that altering the immediate flanking sequences of a determinant can affect their processing by the proteasome and other peptidases. Here, our focus has been different. While some of other constructs included changes in the specific AA sequence (most of which did not affect DP) we have mostly focused on changes in the length of the antigen protein and the differential effects on DP vs. CP. For this purpose, we have systematically altered the length of the N and C terminal extensions flanking the dominant Kb
-restricted determinant of OVA to demonstrate that extensions of minimal length at both sides are essential for efficient CP but not for DP of SIINFEKL. Our work shows that CP of SIINFEKL does not occur when the determinant localizes exactly at the N or C terminus of the Ag and that extensions at both termini are essential for efficient CP. In the case of the N-terminal extension we did not determine the exact length requirement but clearly showed maximal CP for a construct with a 46 AA extension and poor CP when the extension was 18 AA long. In the case of the C-terminal extension we demonstrated the exact requirement for 22 AA because removal but not replacement of the C-terminal AA in 46-SIINFEKL-22 was sufficient to strongly decrease CP in a construct that, otherwise, is cross-presented with high efficiency. Remarkably, removal of this AA did not result in a protein with decreased stability, which is thought to play an important role in determining whether a protein is cross-presented or not.
While the length of the extensions were important for CP, the OVA specific sequences were not absolutely essential because even we do not know whether the cut-off length is exactly the same (and most likely is not), replacement of the natural OVA sequences at either termini for sequences of the same length but derived from GFP resulted in constructs with the same CP patterns as those with the original OVA sequence. However, substitutions of the natural 46 AA N-terminal extension in 46-SIINFEKL-121 with sequences from influenza virus NP [46(C-NP)-SIINFEKL-121] resulted in the loss of CP indicating that some particular sequences may be deleterious. The reason for this remains unknown.
With rapid degradation ruled out, the reason why SIINFEKL must have N and C terminal extensions of minimal length but not particular sequence is not clear. We have made numerous attempts to unveil the mechanism by appending commonly used tags (HA, poly-His) to constructs that are not cross-presented as a way to track them intracellulary by confocal microscopy or biochemically in co-immunoprecipitation experiments. Unfortunately, we did not reach any conclusive results because Ag tagging affected the CP of the constructs most likely because the tags alter the length and sequence of the constructs). Still, we speculate that size may affect the specific enzymes that participate in protein degradation, modification, or association with scaffolding proteins such as chaperones. Also, the extensions may themselves protect the determinant from proteolytic destruction. For example, an N-terminal extension could protect the amino terminus of the determinant from being rapidly chewed-off by amino peptidases. Nonetheless, our finding of differences in CP efficiency that cannot be explained by differential half-life of the Ag add an additional factor that may affect CP (i.e., length of the flanking AA chains) but do not contradict the work of Norbury et al. (17
) showing that short antigenic half-life can decrease CP because we did not identify constructs with short-half lives that are nonetheless well cross-presented.
While most of our work was performed in tissue culture using bone marrow derived Mϕ as pAPC, we have used some of the constructs in vivo
to show that the basic pattern of CP is maintained in an in vivo
setting and without a predetermined pAPC. Clearly, 46-SIINFEKL-21 reproducibly induced less OT1 proliferation in vivo
than 46-SIINFEKL-22. For example, the total number of OT1 cells (events, in ) in mice immunized with infected donor cells expressing 46-SIINFEKL-21 was decreased ~20% in the LN on day 4 and ~55% in the spleen on day 6 after immunization when compared with those immunized with infected donor cells expressing 46-SIINFEKL-22. The differences between these two constructs in vivo
, however, were not as pronounced as in vitro
(~80% reduction) but this may most likely be explained by the high sensitivity of the OT1 system and that measuring OT1 proliferation in vivo
requires several days. Nevertheless, similar to the in vitro
assay (not shown) CP of 46-SIINFEKL-16 was completely abolished in vivo
. It is also important to note that in our in vitro
and in vivo
experiments the synthesis of the Ag is driven but not encoded by the infecting VACV and can only occur in the donor cells because they express the T7 polymerase (16
). Thus, even though VACV can replicate in A9 cells and spread to other cells, our system allowed us to measure CP without having to inactivate the virus. Hence, in the in vitro
experiments feeding Mϕ with infected cells resulted in their activation (not shown) while the in vivo
experiments reflect CP as it occurs during the inflammatory conditions that are normally induced by live VACV infection and have been shown to utilize TLR2 (40
In this report we exclusively use variants of the model antigen OVA for which there are very well defined reagents that allow this type of study. Specific details of our findings should not be considered as rules since it is very likely that the exact overall length of the expressed polypeptide and that of the extensions are intrinsic to the OVA model and will not be identical for other antigens. Indeed, for some determinants, flanking sequences may not be required as recently shown for a Db
restricted peptide from influenza virus (18
). However, our studies show by clear example that the primary structure of the antigen may strongly affect the efficiency of CP and this possibility should be considered for vaccines that use CP as the main mechanism for TCD8+
stimulation such as DNA vaccines.
To date little is known about the relative contributions of CP and DP to anti-viral TCD8+ responses and whether this may vary for each type of virus. Our work demonstrates that CP can be much more limited by the primary structure of the Ag than DP and may explain why some Ags that are presented directly are not cross-presented. Our studies identified several OVA constructs with minimal differences in their sequence that are excellent substrates for DP but that vary widely as substrates for CP. Side-by-side comparison of these constructs expressed in recombinant viruses for their ability to initiate TCD8+ responses in vivo may provide direct evidence about the relative importance of CP and DP for the priming of anti-viral TCD8+.