TAP and ERAAP sculpt the class II-restricted self peptidome
Previous reports have documented an altered endogenous class I–associated self class II-restricted peptidome in the absence of the CAP components TAP or ERAAP [22-26]. Recently, increasing interdependence of the class I– and class II–restricted Ag-processing pathways and the identification of several class II–restricted peptides that require the activity of components of the CAP machinery have been reported [12-15, 27-31]. This led us to query whether the basal class II–associated self peptidome might also have a similar dependence on TAP and/or ERAAP. To this end, class II–associated peptides were eluted from affinity purified H2Ab molecules expressed by WT, B6.129-TAP−/−, B6.129-ERAAP−/−, and B6.129-H2Ab−/− splenocytes. Importantly, deficiency in either TAP or ERAAP did not alter the frequency of APCs within the spleen. Nor was the cell surface phenotype (e.g. class II and co-receptor CD80 and CD86 expression) different than WT (data not shown; [24, 25]). The recovered peptides were fractionated by reversed-phase chromatography (RPC) and their sequence deduced by LC-MS/MS tandem mass spectrometry (Fig. 1 and Supporting Information Fig. 1).
Figure 1. LC-MS/MS spectra of H2Ab-associated peptides commonly displayed by WT, TAP−/− and ERAAP−/− splenocytes. Peptides eluted from immunoaffinity purified H2Ab molecules expressed by splenocytes from 68 to 70 WT, TAP−/− and ERAAP−/− mice were separated by RPC and their amino acid sequence determined by LC-MS/MS. Representative mass spectra are presented. For each spectrum, the b- and y-ions are indicated along with the Sequest cross-correlation score (Cn) showing the degree of concordance between the observed and expected fragment ions. Within the spectrum, b1, b2, y1 and y2 refer to fragment ions that have mass/charge (m/z) +1 or +2. Below each spectrum are the +1 ion m/z values for each peptide (bold underlined, observed ion masses). Note: the +2 ion mass/charge values are provided in Supporting Information Fig. 1.
Download figure to PowerPoint
The mass/charge (m/z) pattern generated by MS/MS was compared against a dataset consisting of the m/z patterns of theoretical and known peptide sequences. The degree of concordance between these two patterns was assigned a cross correlation score Xcorr (Cn). Higher Cn values are assigned to those peptides whose m/z pattern showed greater concordance between the observed and expected m/z patterns . Only peptides with a Cn > 1.5 were considered to be possible peptide sequences. However, the larger the Cn value the more confidence is placed in the peptide sequence identification. In addition, greater differences in the Cn values between the top two most likely peptide sequence identifications (ΔCn) provides greater confidence in the identification. Therefore, peptides with a highly confident identification were considered to have a Cn score >3.0 and ΔCn >0.2. Overall, this dataset had an average Cn = 3.536 and ΔCn = 0.324. In addition, 44% of the peptides had only a single possible sequence identification for which no ΔCn can be calculated.
To ascertain the specificity of the bound peptides, materials eluted from control H2Ab-deficient cells were isolated and analysed by the same methods. We found that only ∼7% of the peptide sequences (Cn > 1.5) identified in WT, TAP−/− and ERAAP−/− samples were also present in the control H2Ab−/− eluates (data not shown). These were largely derived from three sources; (i) Ig – likely representing the Ab used for immunoaffinity purification or splenic Ig that bound to protein A Sepharose used to prepare the immunoaffinity column; (ii) complement – perhaps because they bind Ig; and (iii) fibronectin, fibrinogen and other secreted proteins – likely representing unspecific contaminants of the purification. Few peptides were derived from cytosolic/intracellular proteins. Hence, peptide sequences that matched those isolated from H2Ab−/− splenocytes were considered an artefact of the purification. Such peptide sequences with Cn > 1.5 when present in WT, TAP−/− and ERAAP−/− samples were removed from all downstream analyses.
Analysis of the peptides identified with high confidence (Cn > 3.0 and ΔCn > 0.2) that were eluted from WT, TAP−/− and ERAAP−/− splenocytes surprisingly revealed little overlap between the peptides displayed by WT cells and either TAP−/− or ERAAP−/− cells (Fig. 2 and Supporting Information Table 1). Only 22.5% of the H2Ab-restricted self peptide sequences displayed by WT cells were also presented by TAP−/− or ERAAP−/− cells (Fig. 2A). In a different project, replicate MS samples that consisted of peptides with similar confidence levels eluted from MHC molecules, demonstrated a 63% concordance (SBC, CTS, AJL and SJ, unpublished data). Since class II–associated peptides expressed by WT- and CAP-deficient cells have only 22.5% overlap, the differences in the WT and CAP peptidomes are likely real and not caused by irreproducibility in the experiment. Conversely, 18.4% of self peptide sequences displayed by TAP−/− cells were presented by WT cells, while 33% of self peptide sequences displayed by ERAAP−/− cells were presented by WT cells. This lack of identity was not due to bias in selecting peptides with Cn > 3.0 as datasets which included peptides identified with either moderate (Cn > 2.5 and ΔCn > 0.2; Fig. 2B) or low (Cn > 1.5 and ΔCn > 0.2; Fig. 2C) confidence also demonstrated little overlap in peptide sequence. However, to maintain focus on relevant naturally processed self peptides using this unbiased approach, all downstream analyses were performed on peptides with Cn > 3.0 and ΔCn > 0.2. Importantly, this peptide set was found to have a false discovery rate (FDR; described in the Materials and Methods) of 0, i.e. no peptides were identified by random similarity.
Figure 2. TAP and ERAAP deficiency alters the basal H2Ab-restricted self peptidome. The prevalence of H2Ab-restricted self peptide sequences was compared between WT, TAP−/− and ERAAP−/− strains. Venn diagrams indicate the number of unique and common peptide sequences identified amongst the peptidomes displayed by the indicated strains. Cn > 3.0 (A), Cn > 2.5 (B) or Cn > 1.5 (C) indicates decreasing spectral confidence (see Materials and Methods). ΔCn ≥ 0.2 distinguishes between the top two peptide sequences predicted from the spectrum; this criterion allows identification of the best peptide sequence that matches the observed spectrum. (D) Using the LOCATE database, the number of peptides derived from cytosolic and secreted proteins was compared amongst the peptidomes consisting of peptides with Cn > 3.0.
Download figure to PowerPoint
Notably, the average length of H2Ab-associated peptides increased from 14–16 amino acid residues in WT cells to 18–20 amino acids in TAP−/− and ERAAP−/− cells (Supporting Information Table 1 and Fig. 2). This was consistent with peptide length changes previously observed for class I–associated peptides displayed by ERAAP−/− cells . In addition, we observed numerous groups of nested peptides arising from the same protein (Supporting Information Table 2) as would be expected from class II–associated peptides expressed by WT cells [37, 38]. These nested peptides contained both N- and C-terminal extensions, consistent with previous reports on class II–associated peptides expressed by WT cells [37, 38]. Moreover, only two peptides identified in this study have been previously reported (Supporting Information Table 1) [37, 38]. The lack of overlap in peptides identified in previous studies and this one may have resulted from the analysis of different cell populations. We used unmanipulated APCs isolated directly ex vivo in this study compared with B-cell lymphomas, LPS-induced B-cell blasts, IFN-γ–induced BMC2.3 cell line and Flt3-induced cells used in the earlier reports [37, 38]. In addition, although we found thousands of peptides by LC-MS/MS, we have focused solely on those with the highest Cn values. It is conceivable that the few hundred peptides previously reported were excluded based on the criteria used for sequence determination and validation and may be present in the larger dataset. Hence the differences observed in the different reports do not detract from the novel peptides reported herein as similar results were observed with the larger datasets as well (Fig. 2B and C).
H2Ab-associated peptides were derived from both secreted/extracellular and cytosolic/intracellular proteins as defined in the LOCATE database . However, the majority (∼70%) were processed from cytosolic/intracellular proteins (Fig. 2D), including proteins associated with endosomes. Comparing individual genotypes, the presentation of cytoplasmic/intracellular protein-derived peptides was increased in TAP−/− and ERAAP−/− splenocytes. Consistent with previous reports , ∼63% of the H2Ab-associated self peptidome presented by WT cells were generated from cytosolic/intracellular proteins. In contrast, 87.5% and 80.2% of the H2Ab-associated peptides displayed by TAP−/− and ERAAP−/− splenocytes, respectively, were derived from cytosolic/intracellular proteins (Fig. 2D). These data demonstrate that numerous cytoplasmic/intracellular proteins, including endosomal proteins, are processed and presented by H2Ab in TAP−/− and ERAAP−/− mice. From these analyses, we conclude that CAP components can impact the H2Ab-associated self peptidome.
TAP and ERAAP deficiency alter the CD4+ TCR repertoire
As the self peptidome instructs the developing TCR repertoire, we compared TCR Vβ usage by CD4+ CD62LhiCD44lo naive T (Tn) cells between WT mice and for TAP−/− or ERAAP−/− animals using a panel of Vβ-specific antibodies. As previously reported , the frequencies of TCR Vβ usage between WT-, TAP−/−- or ERAAP−/−-derived CD4+ Tn cells were quite similar, although not identical (Fig. 3A). Likewise, TCR Vβ usage within Lm-reactive CD4+ CD62LloCD44hi effector T (Teff) cells of WT, TAP−/− or ERAAP−/− mice were similar as well (Fig. 3B).
Figure 3. Differential self peptidome display has little impact on the TCR Vβ usage. WT, TAP−/− and ERAAP−/− mice were inoculated with Lm or not and the TCR Vβ usage of the indicated CD4+ T-cell population was determined by flow cytometry after staining with a panel of Vβ-specific antibodies. The cumulative bar graphs indicate the proportion of each Vβ segment present within the (A) CD4+ Tn (CD44loCD62Lhi) or (B) Lm-immune Teff (CD44hiCD62Llo) population in replicate experiments.
Download figure to PowerPoint
Since Ag recognition is mediated by the highly variable CDR3, we specifically examined this region of the TCR β-chains. CDR3β sequence diversity can be estimated by analysing the number of amino acids spanning the V-D-J recombination site by spectratyping the nucleotides that encode them [41, 42]. Although different sequences may have equivalent lengths, thereby underestimating the true diversity, differences in the number of amino acids, nonetheless, provide a high throughput estimate of Ag receptor diversity. The diversity of the TCR of flow sorted CD4+ Tn cells were analysed by spectratyping 52 Vβ-Jβ pairings. This analysis revealed extensive alterations in some but not all CDR3β length profiles in the naive TCR β-chain repertoire expressed by WT, TAP−/− or ERAAP−/− mice (Fig. 4 and Supporting Information Fig. 3A). Similar analysis of flow sorted Lm-responsive CD4+ Teff cells revealed extensive differences in the CDR3β length profiles between WT and TAP- or ERAAP-deficient CD4+ Teff cells (Fig. 5 and Supporting Information Fig. 3B). These data suggest that, despite similarities in Vβ usage, which was serologically determined, CD4+ T cells utilize different CDR3β sequences in the absence of the CAP machinery. Since the CDR3β region of the TCR is predominantly involved in Ag recognition, sequence differences in this region could potentially lead to alterations in the CD4+ T-cell responses to microbial challenge.
Figure 4. The TCR repertoire of naïve CD4+ T cells in both TAP−/− and ERAAP−/− mice is substantially different from that of WT mice. Total RNA was isolated from purified CD4+ Tn cells from naive, uninfected mice and amplified by RT-PCR using primers specific for the indicated VβJβ rearrangements. CDR3β length diversity was detected by capillary-gel electrophoresis and quantified by calculating the area under each peak. Representative spectrograms of four TCR VβJβ CDR3 length distributions are shown from WT, TAP−/− and ERAAP−/− derived CD4+ Tn cells. Bar graphs depict the fraction of specific CDR3β lengths present in the total population. Data presented are from one experiment, which is a representative of two performed (n = 3–5 mice per strain per experiment). Replicates sometimes displayed minor alterations in the absolute frequencies of CDR3 lengths but no alteration in their presence or relative frequencies were observed within a sample.
Download figure to PowerPoint
Figure 5. The Listeria-immune TCR repertoire is altered in TAP−/− and ERAAP−/− CD4+ Teff cells when compared with WT Teff cells. WT, TAP−/− and ERAAP−/− mice were inoculated with Lm. After 7 days, CD4+ Teff cells were flow sorted to ≥95% purity and total RNA was isolated, processed and analysed as described in Figure 4. Representative spectrograms of four TCR VβJβ CDR3 length distributions are shown for WT, ERAAP−/− and TAP−/− CD4+ Teff cells. Bar graphs depict the fraction of specific CDR3β lengths present in the total population. Data presented are from one experiment, which is a representative of two performed (n = 3–5 mice per strain per experiment). Replicates sometimes displayed minor alterations in the absolute frequencies of CDR3 lengths but no alteration in their presence or relative frequencies were observed within a sample.
Download figure to PowerPoint
TAP-deficiency alters class II–restricted microbial Ag recognition
Previously, we reported that the magnitude of the CD4+ T cell response to minor histocompatibility Ag HY and Lm-derived LLO and p60 peptides were increased in animals deficient in TAP or ERAAP . Here, we have shown that TAP and ERAAP impact the quality of the H2Ab-restricted self peptidome and alter the TCR repertoire. Therefore, we queried whether the CAP machinery could destroy and/or create class II–restricted microbial peptides recognized by CD4+ T cells. To this end, WT, H2Ab−/− and TAP−/− mice were inoculated with VACV and, 7 days later, the Th response tested against a panel of 448 15-mer peptides. This panel consisted of putative H2Ab-restricted peptides from VACV ORFs . An initial screen of these peptides revealed few shared specificities and significant alterations in the magnitude of CD4+ T-cell responses to these shared peptides in TAP−/− mice when compared to WT animals (data not shown). In addition, the loss of response to some peptides and novel responses to others was suggested (data not shown). To confirm these results, WT, TAP−/− and H2Ab−/− mice were inoculated with VACV. After 7 days, splenocytes were restimulated in vitro with increasing amounts of select peptides identified from the initial screen. This interrogation confirmed our previous observation  that TAP−/− Th cells responded to certain peptides with increased magnitude (Fig. 6A). In addition, the reactivity against other peptides was lost when compared to the response elicited in WT mice, suggesting they are dependent on the activity of TAP (Fig. 6B). Still other peptides were uniquely recognized only by TAP−/− Th cells and not WT Th cells (Fig. 6C) suggesting that in WT animals those epitopes are destroyed by the action of TAP. Importantly, VACV-immune spleen cells from H2Ab−/− mice recognized none of the peptides tested (Fig. 6) indicating H2Ab-restricted recognition of these epitopes by Th cells and not CD8+ T cells. Hence, these data demonstrate that the CAP machinery profoundly affected the Th-cell response. The altered T-cell response is a reflection of both altered Ag processing and presentation as well as an altered CD4+ T-cell repertoire.
Figure 6. WT and TAP−/− CD4+ T cells recognize a different subset of vaccinia viral epitopes. WT, TAP−/− or H2Ab-/− mice were inoculated with 5 × 105 pfu VACV. After 7 days, 106 splenocytes were restimulated in vitro with the indicated class II–restricted VACV peptides and the number of IFN-γ–producing cells determined by ELISPOT. Peptide recognition was either (A) enhanced, (B) lost or (C) uniquely generated in TAP−/− animals compared with WT responses. Data are shown as mean ± SD of n = 3–5 mice and are from one experiment representative of two performed. Peptides were derived from (A) I1L, (B) K ORF B and (C) D13L and E ORF B, respectively.
Download figure to PowerPoint