Generation of cytotoxic T-cell lines using overlapping pentadecapeptides derived from conserved regions of the adenovirus hexon protein

During the period of immune reconstitution following allogeneic haematopoietic progenitor-cell transplant (HPCT), patients are highly susceptible to infectious agents. Disseminated adenovirus (Ad) infection is among the most lethal post-HPCT complications, especially in paediatric patients (Flomenberg et al., 1994). The importance of T-cell immunity in preventing Ad infection is evidenced by higher disease incidence associated with targeted T cell-specific treatment for graft versus host disease (GVHD) (La Rosa et al., 2001) and in recipients of T cell-depleted grafts, where T-cell recovery may be delayed (Chakrabarti et al., 2002; Symeonidis et al., 2007). Antiviral agents, including cidofovir and ribavirin, may be effective in reducing viral loads in patients with at least partial T-cell immunity, but have been less effective in the absence of viral immunity, indicating the need for alternative therapies (Feuchtinger et al., 2007; Lenaerts et al., 2008; Neofytos et al., 2007). Adoptive immunotherapy has proven to be effective for treatment or prevention of cytomegalovirus (CMV) (Walter et al., 1995) and Epstein–Barr virus (EBV) (Rooney et al., 1995) infection, and more limited early studies have shown potential beneficial effects for Ad infection (Feuchtinger et al., 2006; Leen et al., 2006, 2009).

Most individuals with previous exposure to CMV or EBV develop potent T-cell immunity focused on limited immunodominant epitopes shared by clinically relevant strains of these viruses (McLaughlin-Taylor et al., 1994; Murray et al., 1992). However, Ad disease is associated with many of the six major Ad groups and 51 serotypes within these groups. Indeed, it is common for even healthy individuals to experience repeated infection by different Ad serotypes, even though serotype-specific immunity may be lifelong (Lenaerts et al., 2008). Shared epitopes that confer protective immunity across Ad serotypes are relatively uncharacterized. Cross-reactive T-cell responses among Ad serotypes can be generated ex vivo by priming with Ad-infected fibroblasts (Olive et al., 2002; Tang et al., 2006), Ad lysates (Feuchtinger et al., 2006) or a modified Ad vector (Leen et al., 2004a). In each case, the T-cell response was found to be predominantly against hexon conserved regions. In contrast to the CD8⁺ T-cell response elicited by most viruses, the methods used to date result in responses mediated predominantly or exclusively by CD4⁺ T cells.

Here we have used an approach to prime Ad-specific T cells that focuses the response to conserved regions of hexon so as to maximize cross-reactivity across Ad groups. For this purpose, we prepared a series of overlapping pentadecapeptides (15-mers) spanning four highly conserved regions within the Ad5 hexon. We hypothesized that priming with pentadecapeptides would increase the likelihood of expanding CD8⁺ memory T cells compared with currently used methods. The combination of CD8⁺ cytotoxic T lymphocytes (CTLs) and CD4⁺ T-helper cells may be more clinically effective in preventing or controlling Ad infection than CD4⁺ CTLs alone (Yewdell & Bennink, 1992).

CTL line generation
Cell lines from nine donors were generated. Six were prepared by using Ad peptide pool (PP)-pulsed dendritic cells (DCs) as the antigen-presenting cells (APCs). Lytic activity was monitored by using Ad PP-pulsed B lymphoblastoid cell lines (BLCLs) as targets. Three of six lines showed activity after three stimulations (week 3) (Fig. 1a). The three lines responding at week 3 showed an increase in activity after a fourth stimulation with DCs. However, even after five rounds of DC priming, the three lines that lacked activity at week 3 remained inactive.

(13K): Fig. 1. CTL activity of lines primed with DCs or DCs followed by BLCLs. (a) Results of six lines stimulated with Ad PP-pulsed autologous DCs only. (b) Results of nine lines stimulated with Ad PP-pulsed autologous DCs (week 3), followed by two stimulations using autologous Ad PP-pulsed BLCLs. (c) Results of lines shown in (b) tested against non-pulsed BLCL targets. All data are the percentage specific lysis at an effector : target cell (E : T) ratio of 25 : 1. Data in (a) and (b) are minus reactivity to non-pulsed BLCLs.

Based on our previous studies using a similar overlapping pentadecapeptide approach (Zhu et al., 2007), we retested the six donors along with three additional donors, but switched the lines to Ad PP-pulsed autologous BLCLs as APCs after the third DC stimulation. After a single use of Ad PP-pulsed BLCLs, eight of the nine lines had measurable CTL activity that increased further with additional rounds of stimulation (Fig. 1b). Only the RD0406 line failed to recognize Ad PP-pulsed targets even after five stimulations with DCs, or DCs followed by BLCLs. BLCLs as APCs resulted in modest lysis of non-pulsed BLCL targets for most donors (Fig. 1c). However, two donors showed strong EBV activity that coincided with a stronger Ad response for donor RD0407, but not for donor RD0406. The expansion rate of lines was similar using DCs or BLCLs as APCs. Typically, lines expanded between 2- and 5-fold after the first week of stimulation, then by approximately 1 log each subsequent week until week 4. Regardless of the APC used, expansion rates slowed after week 4 and most lines could not be maintained for longer than 6 weeks.

Line screening and peptide identification
Lines that lysed PP-pulsed BLCLs underwent small-pool screening generally after week 4 of culture. Candidate single peptides (SPs) were tested individually. One representative experiment for donor RD0504 is shown in Supplementary Fig. S1 (available in JGV Online). Sequences of SPs recognized by each of eight lines responding to Ad PP-pulsed targets are shown in Table 1.

Table 1. Summary of donor line screening and peptide characterization

HLA restriction and effector phenotype
Identified peptides were characterized for HLA restriction by using a panel of peptide-pulsed HLA-partially matched BLCLs. Data interpretation was complicated for lines from donors with tightly linked HLA haplotypes or rare alleles where appropriate panel BLCLs were not available and due to cross-reactive peptide presentation by certain HLA alleles, particularly class II alleles (Chicz et al., 1993). Nevertheless, we were able to identify restricting allele(s) clearly for most lines. HLA restriction was further confirmed by determining the subset of T cells responding, given that the CD8⁺ T-cell response is restricted by HLA class I and the CD4⁺ T-cell response by HLA class II. HLA-restriction elements and effector phenotype data are summarized in Table 1 and the full HLA phenotype of the donor lines is shown in Supplementary Table S1 (available in JGV Online).

Minimum peptide identification
RD0307.
The RD0307 line recognized eight peptides, including three pairs representing sequential overlapping peptides. Both SP31 and SP32 induced strong responses when presented by B*3501 that were weaker when presented by A*3301 (Fig. 2a). As expected, SP31 and SP32 stimulated CD8⁺ T cells with a similar response frequency to both peptides (Table 1). Based on the B*3501-binding motif, the overlap sequence MPNRPNYIAF probably contains the minimal peptide. This agrees with previous reports showing T-cell recognition of MPNRPNYIAF presented by B*3501 (Leen et al., 2008). Likewise, based on binding motifs, NRPNYIAFR is the most likely sequence in SP31 and SP32 to be presented by A*3301. In contrast, the next overlapping peptide, SP33, induced only CD4⁺ T cells to produce gamma interferon (IFN-γ) (Table 1). RD0307 was heterozygous for DQB*0501 and DQB*0503 and both alleles presented SP33 equally, and SP33 was only slightly less well presented by DQB*0502 (Fig. 2b). SP86 and SP87 induced similar reactivity in the screening assays, therefore only SP87 was tested for HLA restriction. The HLA-restriction pattern of SP87 was similar to that of SP33 (Fig. 2e), indicating that both peptides could be presented by all three DQB*05 alleles. SP46 and SP47 also stimulated only CD4⁺ T cells in line RD0307 and both were presented by DPB1*0501 (Fig. 2c). However, in all assays, SP47 stimulated a higher-frequency response, indicating that it contains the optimal sequence. Nearly 6 % of CD8⁺ T cells in the RD0307 line produced IFN-γ in response to SP60, which was A*1101-restricted, but with a lesser response when presented by the serologically cross-reactive A*0301 allele that was not present in RD0307 (Fig. 2d). Given the lack of response to SP59 or SP61, the peptide that is recognized must be contained in the sequence YLNHTFKKV. This is consistent with the known A*1101-binding motif.

(10K): Fig. 2. Determination of HLA restriction of the RD0307 line. The line was tested by ELISPOT assay using the indicated SPs presented on BLCLs selected to share limited HLA alleles with the responders. Alleles likely to be cross-reactive with donor alleles are shown in parentheses. Results of overlapping SPs to which the line responded are shown on the same graph. (a) SP31 (filled bars) and SP32 (open bars); (b) SP33; (c) SP46 (filled bars) and SP47 (open bars); (d) SP60; (e) SP87. Data are no. spots per 105 cells plated.

RD0309.
The RD0309 cell line recognized 12 peptides, including five pairs where the minimal peptide is probably in the overlapping sequence (Table 1). The frequency of CD4⁺ T-cell response to SP2 and SP3 was similar. SP3 was shown to be DRB1*0101-restricted. The SP2/SP3 overlapping sequence, WSYMHISGQDA, contains motif peptides at each of four DRB1*0101 anchor positions (underlined). SP38, but not the flanking peptides, induced a CD4⁺ T-cell response, indicating that the sequence unique to SP38, MGVLAGQASQLNA, may be the minimal peptide. The class II allele(s) presenting SP38 could not be distinguished. The RD0309 line also recognized three sequential peptides, SP45, SP46 and SP47. The response to SP45 and SP46 in screening assays was similar (not shown), indicating that the shared sequence was recognized. Further characterization of SP46 showed a response only by CD4⁺ T cells, indicating HLA class II restriction, but the restricting antigen could not be distinguished. SP47 was clearly DRB1*0101-restricted and elicited the strongest response of any SP in the study, with nearly 30 % of CD4⁺ cells responding. Candidate sequence TRYFSMWNQAV fits the DRB1*0101-binding motif best and is unique to SP47. SP73 and SP74 each stimulated a similar CD4⁺ T-cell response, and further characterization of SP73 showed that either DQB*0501 or DQB*03 could serve as the restricting antigen. The SP81/SP82 pair induced a CD8⁺ T-cell response restricted by Cw*0401, together with a CD4⁺ T-cell response that was strongest for SP81 and was DPB1*0401-restricted. Considering the Cw*0401-binding motif, FPYPLIGTA is the most likely minimal overlapping sequence. Due to logistical and technical reasons, the T-cell subset responding to SP45, SP86 and SP87 was not tested and, although SP87 was assessed for HLA restriction, no clear pattern emerged that distinguished between possible class II alleles in the panel. SP87 stimulated approximately twice the response in enzyme-linked immunospot (ELISPOT) screening assays as SP86 (304 per 10⁵ cells versus 155 per 10⁵ cells, respectively), indicating that the optimal binding sequence was contained in SP87.

RD0407.
Two separate lines were prepared from cells obtained from donor RD0407. The first line was screened by using CTL assays in part, because a strong response to EBV after switching to BLCLs as APCs made ELISPOT screening difficult to interpret. Three peptides were identified in this line and two of these (SP28 and SP29) shared the overlapping sequence KPYSGTAYNAL, previously identified as being HLA-B*0702-restricted (Leen et al., 2004b). Using SP29, we confirmed that RD0407 also recognized KPYSGTAYNAL presented by B*0702. One additional peptide, SP51, was identified as DPB1*0201-restricted. Although individual SPs were not tested, both CD4⁺ and CD8⁺ T cells responded to PP well in excess of the response to non-pulsed BLCLs (Table 1). The second line generated from donor RD0407 used DCs alone and recognized two additional peptides using ELISPOT, although fewer cells responded than the first line. SP60 elicited a CD4⁺ T-cell response that, as for SP51, was DPB1*0201-restricted. SP98 contained two minimal peptide sequences that were not shared with adjacent peptides, one of which was DQB*0402-restricted and the second Cw*0702-restricted. Review of the Cw*0702-binding motif indicates that EPTLLYVLF is the most likely sequence recognized.

RD0501.
The RD0501 line was screened by using ⁵¹Cr-release assays and recognized SP28 and SP29 (Fig. 3a) presented by HLA-B*0702 (not shown). However, after an additional week of priming, HLA-restriction studies using Ad PP revealed that both B*0702- and DRB1*0701-matched targets were recognized (Fig. 3b). Intracellular cytokine (IC) assays confirmed that both CD4⁺ and CD8⁺ T cells produced IFN-γ in response to Ad PP, whereas only CD8⁺ T cells responded to SP28 and SP29 (Fig. 3c). RD0501 typed for the uncommon DP alleles, DPB1*17 and DPB1*03, neither of which had been included in the HLA screening panel; therefore, it is unclear whether other DP-restricted peptides might also be present. Unfortunately, no additional cells were available from this donor for further characterization.

(38K): Fig. 3. RD0501 peptide identification, HLA restriction and IC production. (a) SP screening at week 4 by 51Cr-release assay. (b) HLA restriction at week 5 using Ad PP-pulsed targets. (c) IFN-γ production after presentation on autologous BLCLs pulsed with Ad PP, SP28 or SP29. Lytic data are expressed as lytic units (LU) per 106 cells after subtraction of non-pulsed targets.

RD0504.
The RD0504 line recognized nine peptides, including three sequential pairs (Supplementary Fig. S1; Table 1). Pooled sequential pairs (i.e. SP9 pooled with SP10) showed a single HLA-restriction pattern (Table 1), therefore only one of each pair was screened for IC production (SP9, SP38 and SP98). All except one SP tested stimulated a CD4⁺ T-cell response and were class II-restricted (Table 1). The DPB1*0401-restricted sequence found in the SP98/SP99 overlap, PTLLYVLFEVF, contains the minimal epitope, TLLYVLFEV, identified as optimal for presentation by DPB1*0401 (Tang et al., 2004). SP31 stimulated CD8⁺ T cells and was A*0301-restricted. As RD0504 did not respond to SP32, the minimal peptide cannot be contained completely within the SP31–SP32 overlap. The peptide preceding SP31 in the hexon sequence was not part of our PP, therefore specificity cannot be narrowed further considering the overlap. The preferred binding motif for HLA-A*03-restricted peptides includes tyrosine (Y) at the last anchor position, making QSMPNRPNY the most likely minimal sequence in SP31.

RD0505M.
The RD0505M line recognized four peptides, consisting of two pairs with overlapping sequences (Table 1). Stimulation with PP-pulsed autologous BLCLs showed a CD4⁺ T-cell response. SP32 and SP33 were both DRB1*0401-restricted. The overlap PNYIAFRDNFI contained four of five amino acids identified at anchor or auxiliary anchor positions of the DRB1*0401-binding motif (underlined). SP6 and SP7 stimulated a high frequency of IFN-γ-producing cells (not shown). However, none of the panel cells used for HLA restriction stimulated a significant response. The panel lacked BLCLs sharing HLA-DRB1*0810 with RD0505M, so it is likely that this allele is the restricting element.

RD0607.
The RD0607 line recognized five SPs in the small-pool screening, including two pairs of overlapping peptides, both of which were DRB1*0701-restricted (Fig. 4a, b), along with SP60, found to be DPB1*0201-restricted (Fig. 4c). SP33 induced 6.5 % of CD4⁺ T cells to produce IFN-γ, compared with 0.9 % for SP32, indicating that the optimal peptide sequence is within SP33 (Fig. 4e). Indeed, the sequence PNYIAFRDNFIG contains three of four motif anchor or auxiliary anchor positions (underlined) versus only two shared positions for the overlapping sequence, indicating that either a second DRB1*0701 sequence is contained within SP32, or the shared sequence binds only partially to DRB1*0701 (Fig. 4f). The overlap between SP56 and SP57, YYTYSGSIPYL, matches the DRB1*0701-binding motif at all anchor positions and, as both peptides induced a similar response CD4⁺ T-cell response, this is probably the minimal sequence (Fig. 4e). After the initial small-pool screening, the RD0607 line was stimulated twice more and it became apparent from IC and HLA-restriction data from PP-pulsed autologous BLCLs that minimally two additional peptides were recognized (Fig. 4d), at least one of which was recognized by CD8⁺ T cells (Fig. 4e). The HLA-restriction panel indicated that one or more A*0101-restricted and DRB1*1101-restricted peptides were recognized. Review of the initial small-pool screening showed weak stimulation by SP91 and SP92, and both contain the previously identified A*0101-restricted epitope, TDLGQNLLY (Leen et al., 2004b). Insufficient cells remained for further characterization.

(36K): Fig. 4. RD0607 HLA-restriction analysis, IC production and identification of probable minimal peptide based on binding motif. (a–d) Results of HLA restriction using Ad PP individual SPs. Data are no. spots per 105 cells plated. (e) Results of intracellular IFN-γ production. (f) Overlapping shared sequences between SP32 and SP33 (SP32/33), the full sequence of SP33 and the overlapping shared sequences between SP56 and SP57 (SP56/57) aligned with the core binding motif for DRB1*0701. Anchor sequences in the Ad peptides are underlined.

RD0801M.
The RD0801M cell line recognized three pairs of overlapping peptides (Table 1). SP15/SP16 and SP91/SP92 showed similar reactivity during screening, so only one of each pair was evaluated further (SP16 and SP92, respectively). A high-frequency response (>300 spots per 10⁵ cells plated) was seen when SP16, SP19 or SP20 was presented by the single panel BLCL that shared both DPB1*0301 and DQB*0402 with RD0801M. This BLCL was also positive for DRB1*0811 and was the only line available that expressed an allele similar to the DRB1*0801 of responding cells. Therefore, which of these three alleles was the restricting element could not be distinguished. SP91, in contrast, stimulated CD8⁺ T cells when presented by A*0101. The minimal sequence TDLGQNLLY shared by SP91/SP92 has previously been demonstrated to be A*0101-restricted (Leen et al., 2004b).

Ability of peptide-primed CTLs to kill Ad-infected cells
Experiments were performed to demonstrate that peptide-primed CTLs could recognize targets infected with Ad. For this purpose, we utilized a chimeric vector with an Ad5 backbone and fiber protein partially from Ad35, thus making it infectious to haematopoietic cells (Yotnda et al., 2001). Autologous BLCLs were tested in ⁵¹Cr-release and ELISPOT assays after infection with the Ad5f35 vector. Data from three such experiments demonstrated that lines responded best to Ad PP-pulsed BLCLs, but did recognize Ad5f35-infected cells (Fig. 5). The RD0407 and RD0501 lines exhibited some response to EBV on non-pulsed targets, which was detected by ELISPOT assay, whereas the RD0504 line had very little EBV response at the time of assay.

(19K): Fig. 5. Recognition of Ad5f35 vector-infected targets or APCs by Ad PP-pulsed donor lines. Lines from the indicated donors were tested in (a) 51Cr-release assays and (b) IFN-γ ELISPOT assays for reactivity against Ad5f35 vector-infected targets or APCs, respectively. Data are shown as the percentage specific lysis at the indicated E : T ratio, or as no. spots per 105 cells plated.

Location of peptides identified in this and other published studies
We compared peptides identified in this and other published reports with their location in the consensus hexon sequence (based on Ad5) to determine which peptides were most likely to be cross-reactive with other Ad serotypes. This was complicated by the fact that the minimal sequence recognized was not defined definitively in all studies. We used the likely minimal sequence of peptides characterized in this study, as shown in Table 1 in bold, underlined and italic, with peptides to the level identified by other groups (Hanley et al., 2009; Heemskerk et al., 2003, 2006; Leen et al., 2004b, 2008; Onion et al., 2007; Tang et al., 2006). Because we used 15-mers, compared with 20- or 30-mers used by others for peptide identification, the sequences containing the minimal peptides could be narrowed in several cases. Overlay of the diversity plot of all 51 Ad serotype hexon sequences (Ebner et al., 2005) on the identified sequences shows that most fall within regions of limited diversity. Note that isolated high-frequency differences in sequence where peptides have been identified may or may not prevent cross-reactivity, dependent upon whether the difference affects a critical anchor residue or whether those differences are within the minimal peptide. However, peptides located in areas of high diversity, such as the peptide identified in this study at position H813–823 or the peptide identified by Leen et al. (2008) at position H201–221, are likely to react only to group C Ad. This map reveals several regions containing multiple peptides, some of which have been demonstrated to represent peptides with overlapping sequences, such as the sequences at H320–338 (five peptides), H376–420 (five peptides), H691–730 (eight peptides), H865–884 (five peptides) and the well-characterized area where the first hexon peptide recognized by T cells was found (Olive et al., 2002), H906–928 (nine peptides). HLA class I-restricted peptides (shown in bold) not infrequently overlap sequences also containing identified class II-restricted peptides (see Supplementary Fig. S2, available in JGV Online). Similar to CMV (Trivedi et al., 2005) and Aspergillus (Ramadan et al., 2004, 2005a, b; Zhu et al., 2007), the use of protein-spanning overlapping pools of pentadecapeptides proved highly effective at priming T-helper 1 cellular immune responses to targeted conserved regions of Ad hexon. Consistent with our experience with Aspergillus (Zhu et al., 2007), optimal results were obtained by switching to BLCLs as APCs after initial use of DCs. BLCLs had the further advantage of reducing requirements for large numbers of DCs that can be limiting as the line expands in numbers. For most donors, BLCL APCs increased the frequency of Ad-reactive T cells while inducing a minimal EBV response, despite typically high numbers of EBV-specific T-cell precursors in peripheral blood (Koehne et al., 2002). Eight of nine lines contained Ad-specific responses using DCs followed by BLCLs as APCs, compared with only three of six lines using DCs alone. This finding could be in part due to the use of DCs at no more than a 1 : 10 DC : T-cell ratio, when more DCs may have been more effective. Responding lines recognized at least 34 of the 105 pentadecapeptides that were used. Minimally, three additional peptides were recognized by two lines (RD0501 and RD0607) that were not identified, but had clear HLA-restriction patterns and cytokine-production profiles in response to PP that were not explained by the SPs that were identified. Thirteen SPs were recognized by more than one donor, including ten by two donors, two by three donors and one by four donors, although the sequence within the pentadecapeptide that constituted the minimal sequence was probably not the same in eight cases (Table 1).

Of responding lines, 50 % recognized SP60 (H710–724), making this the most immunogenic SP in our study. Two donors (RD0407 and RD0607) recognized a DPB1*0201-restricted peptide contained within SP60, while RD0504, who was negative for DPB1*0201, recognized SP60 presented by DRB1*1501. SP60 also contained an A*1101-restricted epitope recognized by donor RD0307, who was negative for both DPB1*0201 and DRB1*1501. Leen et al. (2008) identified a class II-restricted response narrowed to a 20-mer peptide, H711–730, that may represent the SP60 sequence. This group also identified an A*02-restricted peptide at position H711–721 that completely overlaps the A*1101-restricted sequence H713–720, further demonstrating the immunodominance of this region.

The region H906–928 has also been shown to contain sequences that elicit responses from both CD4⁺ and CD8⁺ T cells. The well-characterized DPB1*0401-restricted epitope TLLYVLFEV (H913–921) is contained within this region, along with minimally four additional peptides recognized by Ad-specific T cells (Supplementary Fig. S2) (Tang et al., 2006). We confirmed that SP98/SP99, each containing the TLLYVLFEV sequence, were recognized in the context of both DPB1*0401 and DPB1*0402 and further identified epitopes restricted by DQB*0402 and Cw*0702 within H906–928. Indeed, multiple investigators have reported hexon-specific responses within this region after natural infection (Tang et al., 2004, 2006), after priming with intact Ad (Olive et al., 2002) or a modified Ad vector (Leen et al., 2008), and using 30-mer hexon peptides (Veltrop-Duits et al., 2006).

Predominantly CD4⁺ T cells respond to Ad in healthy adults and paediatric patients post-HPCT, which contrasts with the largely CD8⁺ T-cell response to most viral pathogens (Appay et al., 2008; Yewdell & Bennink, 1992). This probably reflects effects of the Ad early region 3 (E3) glycoprotein, E3-19K, which prevents class I antigen transport out of the endoplasmic reticulum to the cell surface, thus inhibiting presentation of viral antigens to CD8⁺ T cells during natural infection (Burgert & Kvist, 1985). CD4⁺ T cells recognize exogenous viral proteins degraded into peptides in endosomal vesicles, where they bind to class II antigens; therefore, presentation is not affected by E3-19K (Germain & Hendrix, 1991). Ex vivo generation of Ad-specific CD8⁺ T-cell responses by intact virus requires the use of E3 deletion mutants that allow for presentation of Ad peptides by class I molecules (Tang et al., 2006). In these studies, only targets infected with mutant virus or pretreated with IFN-γ to upregulate class I expression were recognized. Smith et al. (1996) were first to demonstrate that virus-infected DCs could also prime CTL responses to Ad, and both IFN-γ-treated wild-type and E3 deletion mutant-infected DCs were effective. These studies indicated that low numbers of CD8⁺ CTL precursors are present after natural infection and that viral capsid proteins, rather than early gene products, are predominantly recognized.

Subsequently, several groups have further characterized Ad-specific human CTLs and each has confirmed that responses are restricted primarily to conserved regions of hexon (Leen et al., 2004a, b, 2008; Olive et al., 2001, 2002; Tang et al., 2004, 2006; Veltrop-Duits et al., 2006). Recent data have also revealed that two early-region proteins, DNA polymerase and DNA-binding protein, may also be recognized by Ad-specific CTLs (Joshi et al., 2009).

The most extensive characterization of hexon response used lines expanded with the Ad5f35 vector (Hanley et al., 2009; Leen et al., 2004b, 2008). CD4⁺ T-cell responses were partially characterized by using 20-mer peptides overlapping the entire coding region of Ad5 hexon to narrow the sequence recognized. Custom-made peptides were used to identify minimal epitopes recognized by less frequent CD8⁺ T-cell responders (Leen et al., 2004b, 2008). Altogether, nine class I-restricted peptides were identified, only one of which was known previously. Three of these nine peptides with identical HLA restriction were confirmed in our study (H114–124, H320–329 and H866–894), showing that whole virus, as well as 15-mer peptides, can prime identical responses. We also identified five novel class I-restricted peptides. Two of these peptides were presented by HLA-C, which has not previously been shown to restrict hexon-specific CTL responses. Considering the two unique peptides identified by Tang et al. (2006), Ad hexon is now known to contain 15 unique class I-restricted peptides; of these, eight are presented by HLA-A, five by HLA-B and two by HLA-C. It remains to be seen whether peptide presentation by HLA-B results in CTLs with a greater functional avidity than those restricted by HLA-A or HLA-C, as has been shown for human immunodeficiency virus and EBV (Bihl et al., 2006).

Thus far, only two approaches to prepare Ad-specific T-cell lines for immunotherapy have proceeded to clinical trials. These include lines primed by using Ad5f35 vectors with or without CMV pp65 and expanded by using BLCLs as a source of EBV (Leen et al., 2006, 2009), and Ad-specific T cells isolated from peripheral blood by using an IFN-γ-capture system after overnight activation with crude Ad extracts (Chatziandreou et al., 2007). Both approaches demonstrated the capacity of transferring immunity to viruses to which lines were primed, although Ad-specific immunity may not be apparent until Ad reactivation occurs (Leen et al., 2006). Some, but not all, Ad5F35-primed lines include Ad-specific CD8⁺ T cells and, while fresh IFN-γ-captured cells contained approximately 25 % CD8⁺ T cells, expansion in medium containing interleukin (IL)-2 expands nearly exclusively CD4⁺ T cells (Feuchtinger et al., 2004, 2006). Furthermore, of six patients who could be evaluated for response to IFN-γ-captured cells, one showed worsening of skin GVHD associated with infusion, suggesting that T cells with GVHD potential may still be present. In this regard, using products that are antigen-expanded ex vivo has advantages, as such lines contain few or no alloreactive T cells (reviewed by Leen & Heslop, 2008). More recent preclinical studies used purified Ad5 hexon protein as the stimulus for IFN-γ capture, but this resulted in low cell recovery and purity and even fewer Ad-specific CD8⁺ cells (Feuchtinger et al., 2008).

An approach similar to ours has been explored, but used 30-mer overlapping peptides spanning the entire hexon to prime T-cell responses. Initial experiments used ten donors to select peptides to which the majority of donors responded. Five 30-mers were chosen for additional studies (Veltrop-Duits et al., 2006). Lines generated with the five peptides had low alloreactive potential but, consistent with the longer length of the peptides, the response was nearly exclusively by CD4⁺ T cells (Comoli et al., 2008). These lines were not characterized for HLA restriction or for specificity more narrowly than the 30-mers used for priming.

It is unclear whether class I-restricted responses to hexon offer greater protection from infection than class II-restricted responses. However, given that CD8⁺ T cells mediate the primary cellular response to most other viruses (Appay et al., 2008) and clearly play a role in the immune response to Ad vectors (Sumida et al., 2004), a method that consistently primes both CD4⁺ and CD8⁺ T cells seems desirable. Use of smaller 15-mer peptides instead of 20- or 30-mers increases the likelihood that both CD8⁺ and CD4⁺ T cells will expand in other viral systems (Maecker et al., 2001). This finding also appears to hold true for Ad, as evidenced by our identification of nearly as many class I-restricted responses in our eight donor lines as were identified in 25 lines primed with Ad5f35 (Leen et al., 2008). Most lines generated with pentadecapeptides not only include a CD8⁺ T-cell response, but also contain CD4⁺ T cells that respond to multiple hexon peptides, increasing the likelihood of broad protection across Ad serotypes.

Studies to characterize Ad hexon T-cell responses extensively are needed for rational design of culture conditions that result in consistent generation of lines containing a mixture of CD4⁺ and CD8⁺ T cells that react across Ad serotypes. Availability of such lines will improve feasibility, efficacy and safety of Ad immunotherapy.

Peptide selection and preparation.
We compared hexon sequence maps from 51 Ad serotypes to identify regions with limited diversity. Four highly conserved regions were selected and, using Ad5 as the prototype, 15-mer peptides, each overlapping the preceding peptide by 11 aa, were constructed to span the selected sequences (Table 2).

Table 2. Adenovirus hexon sequences from which peptides were made

In total, 105 pentadecapeptides were synthesized (NMI Technologietransfer GmbH, Reutlinger, Germany) and were reconstituted in DMSO. A complete PP was prepared to contain 1 ng ml^–1 of each peptide. Twenty-one smaller pools containing 10–11 SPs each at 1 ng ml^–1 were also prepared and arranged in a matrix such that each peptide was contained in two pools. Single-use aliquots were stored frozen until use.

Cell-line generation.
Starting blood was obtained under institutional review board-approved protocols. The research complied with all relevant federal guidelines and institutional policies. Cells were collected from a healthy research donor (RD) in seven cases or from a granulocyte colony-stimulating factor-mobilized healthy HPCT donor as de-identified CD34⁺ cell-reduced cells in two cases (RD0505M and RD0801M).

Unless otherwise indicated, monocyte-depleted peripheral blood lymphocytes were stimulated three times at 7–10 day intervals using immature DCs, as described previously (Ramadan et al., 2004), then subsequently stimulated with Ad PP-pulsed autologous BLCLs as described previously (Zhu et al., 2007). Cultures were initiated in X-VIVO 15/5 % AB serum (BioWhittaker) and 500 pg IL-12 ml^–1, at a ratio of 1 : 10 DCs : T cells. IL-2 was added on day 3 and was maintained through subsequent feedings. No additional IL-12 was added.

Activity to PP was tested starting 1 week after the third stimulation, using ⁵¹Cr-release assays. Subsequent testing used ⁵¹Cr-release assays or ELISPOT assays for IFN-γ production, as described previously (Zhu et al., 2007).

Peptide identification.
SPs recognized by hexon PP-stimulated T cells were identified by first screening with a panel of 21 small pools. Peptides shared by small pools were screened individually to determine which pentadecapeptides were recognized. Identification of the minimal sequence was based on reactivity to flanking overlapping peptides, HLA restriction and consideration of the most likely binding motif using the MHC ligand- and peptide motif-prediction program SYFPEITHI (Rammensee et al., 1999).

HLA restriction.
Recognized SPs were pulsed onto panel BLCLs selected to express one or more HLA alleles shared with the donor. BLCLs in these experiments included well-characterized homozygous BLCLs from the 10th International Histocompatibility Workshop, along with BLCLs prepared under approved protocols from patients or donors undergoing HPCT at our centre.

IC production.
Cell lines were washed minimally 1 week after the last stimulation, then stimulated overnight with peptide-pulsed autologous BLCLs before detection of IFN-γ using Fastimmune IC Detection kits (BD Biosciences). Data are reported as the percentage of CD4⁺ and CD8⁺ T cells expressing IFN-γ minus the unstimulated control.

Hexon sequence.
The hexon sequences of all 51 Ad serotypes were obtained from GenBank and are those sequenced or referenced by Ebner et al. (2005). Sequences were compared with the group C Ad5 hexon from which the pentadecapeptides were derived and the number of differences at each position was calculated.

This work was supported in part by a Medical College of Wisconsin Cancer Center New Interdisciplinary Grant to C. A. K.-T. and F. Z., a North-west Mutual Liberty Foundation award to C. A. K.-T., and the Greater Milwaukee Foundation, John E. Julien Fund to C. A. K.-T. Peptides were purchased by using money from the Sam Beres Memorial Fund.