Short-Term In-Vitro Expansion Improves Monitoring and Allows Affordable Generation of Virus-Specific T-Cells against Several Viruses for a Broad Clinical Application

Adenoviral infections are a major cause of morbidity and mortality after allogeneic hematopoietic stem cell transplantation (HSCT) in pediatric patients. Adoptive transfer of donor-derived human adenovirus (HAdV)-specific T-cells represents a promising treatment option. However, the difficulty in identifying and selecting rare HAdV-specific T-cells, and the short time span between patients at high risk for invasive infection and viremia are major limitations. We therefore developed an IL-15-driven 6 to 12 day short-term protocol for in vitro detection of HAdV-specific T cells, as revealed by known MHC class I multimers and a newly identified adenoviral CD8 T-cell epitope derived from the E1A protein for the frequent HLA-type A*02∶01 and IFN-γ. Using this novel and improved diagnostic approach we observed a correlation between adenoviral load and reconstitution of CD8+ and CD4+ HAdV-specific T-cells including central memory cells in HSCT-patients. Adaption of the 12-day protocol to good manufacturing practice conditions resulted in a 2.6-log (mean) expansion of HAdV-specific T-cells displaying high cytolytic activity (4-fold) compared to controls and low or absent alloreactivity. Similar protocols successfully identified and rapidly expanded CMV-, EBV-, and BKV-specific T-cells. Our approach provides a powerful clinical-grade convertible tool for rapid and cost-effective detection and enrichment of multiple virus-specific T-cells that may facilitate broad clinical application.


Introduction
Adenovirus (HAdV), cytomegalovirus (CMV), Epstein-Barr-Virus (EBV), and polyoma-Virus (BKV) are responsible for serious morbidity and mortality in patients after hematopoietic stem cell transplantation (HSCT) [1,2,3,4]. HAdV represents one of the most frequent and dangerous infections post transplant [5,6], especially after haploidentical HSCT [1,5,7] and, therefore, is a front-ranking target for early preemptive antiviral therapy [8]. Unfortunately, prophylactic treatment with anti-viral drugs is of limited effectiveness, expensive and associated with substantial toxicity, and may result in overtreatment of patients [1,6,9,10]. Recently, it has been shown that reconstitution of HAdV-specific T-cell response correlates with clearance of ADV infection [11,12,13,14]. In patients who showed no virus-specific immune reconstitution after HSCT, donor-derived virus-specific T-cells against different viruses including HAdV were administered with impressive clinical results [15,16,17,18,19,20,21,22]. As a prereq-uisite for the monitoring of virus-specific T-cells in donors and patients, immunodominant viral epitopes have to be identified. Altough we focused only on the monitoring of HAdV-specific T cells, new epitopes could also be used for adoptive therapy, i.e. for the magnetic isolation of HAdV-mulitmer+ T cells [22]. Certain sequences of the major capsid protein hexon are highly conserved among human HAdV which currently comprise more than 55 sybtypes divided into 7 different species (A-G) [23]. This provides the basis for ''cross-reactivity'' of HAdV-specific T-cells facilitating broad recognition and protection against several species [24]. It is known that most CD4 + and CD8 + ADV-specific T-cells recognize predominantly hexon protein structures or overlapping 15-mer peptide pools. The IFN-c secretion induced by appropriate stimulation enables their detection by the IFN-c -cytokine secretion assay (CSA) [25,26]. Alternatively, virus-specific T-cells can be identified and isolated using different types of MHC class I multimers including tetramers, pentamers or streptamers [24]. To date, only few HAdV-specific immunodominant CD8 + T-cell epitopes have been identified that are presented in the context of the common HLA-types A*01, A*24, B*07 and B*35 [14,27] thus greatly limiting the number of available HAdV-multimers. Using these four multimers, the probability to detect ADV-specific Tcells within the Caucasian population is about 73%. According to an algorithm presented by Schipper et al [28], this percentage could be increased to 95%, if a functional A*02-based multimer were available. Our primary aim was therefore to identify new promising ADV-specific epitopes for the HLA-types A*01 and A*24, and particularly for the frequent HLA-type A*02, by analyzing the main structural proteins of the virus, including hexon and protein II, as well as the E1A protein expressed very early after infection.
The utility of HAdV-specific multimers for diagnostic applications is further supported by the recent observation that, in patients who cleared HAdV-infection after HSCT, apart from CD4 + -, also CD8 + T-cells were present. [14]. However, in most healthy donors and HSCT-patients, HAdV-specific T-cells were reliably detectable only after in vitro culture with HAdV-antigen [14,29]. Due to the low frequency of circulating the HAdVspecific T-cells, their exact phenotype remains to be elucidated.
Current clinical immunotherapy protocols are based on either long-term in vitro expansion, excluding [15] or including transfected antigen-presenting cells (APCs) [16,17]. Alternatively, direct magnetic selection of virus-specific T-cells using the IFN-c -CSA [18,20], tetramers [19] or pentamers [22] is employed. More recently good manufacturing practice (GMP)-compliant removable streptamers became available that represent the only therapy presently not considered as an ''Advanced Therapy Medicinal Product (ATMP)'' [21]. Although all studies referenced above reported prevention of overt viral disease and only mild or no graft versus host disease (GvHD), they have a number of important limitations: some are very time-consuming (10-14 weeks), technically demanding and cost intensive, others involve gene therapy, require large volumes of blood, or are limited to those patients that express HLA alleles for which multimers are available. These constraints represent a major impediment to broad clinical application of these adoptive immunotherapy approaches [30]. The first attempt to use only synthetic peptide mixes and cytokines to rapidly generate virus-specific T-cells within 9-16 days was recently published [31]. A major focus of this study was to evaluate optimal conditions for T-cell expansion by testing different viral peptide concentrations and cytokines (preferential IL-4 and IL-7). However, relevant cytolytic activity (,10%) of e.g. expanded HAdV-specific T-cells was only shown after 16 days of expansion. In our study, fresh/frozen PBMCs were stimulated twice within only 12 days by using GMPcompliant adaptable peptide mixes and a consciously delayed supplementation of IL-15, which resulted in high numbers of functional and cytolytically active virus-specific T-cells against HAdV, CMV, EBV and BKV. In addition, for the first time, no or only low alloreactivity was evaluated very detailed by several different assays to further proof the safety of these cells.

Magnetic selection of ADV-specific T-cells from patients
Five to 8610 6 PBMCs were stained with HAdV-specific streptamers according to manufactures instructions (IBA Technologies, Göttingen, Germany), incubated with anti-phycoerythrin(PE) MicroBeads (Miltenyi), magnetically selected by MScolumns (Miltenyi) according to manufacturers instructions (Miltenyi) and analyzed by flow cytometry.

HLA typing of blood donors and patients
Low and high resolution HLA typing of healthy blood donors and patients was performed at the Institute of Transfusion Medicine (Tübingen, Germany) or the Department of Blood Group Serology (Vienna, Austria) with the donors and patients written consent.

Virus-specific and control antigens
Peptides used for multimer analyses are shown in the Table S2 in File S1. Peptide pools for EBV (EBNA-3A) and BKV (LT-Ag) were purchased from JPT (JPT Peptide Technologies, Berlin, Germany) and used at a final concentration of 10 mg/ml for stimulation or pulsing of cells. The final concentration of HAdV (subgroup C-derived Hexon AdV5), EBV (EBNA-1, BZLF-1 and LMP-2) and CMV (pp65) PepTivator (Miltenyi) in the cell suspension was 0.6 nmol for each pepide per ml.

Quantitative real-time-PCR analysis of viruses from patients
Routine HAdV virus screening of patientss stool and blood samples was performed by real-time quantitative (RQ) PCR. Viral DNA isolation followed by RQ-PCR were done as described [1].
For the generation of IL-15-based short-term expanded virusspecific T-cells (seVirus-T-cells), fresh or frozen PBMCs were cultured in AIM2V+ and stimulated with the appropriate peptide pools from HAdV (AdV5-PepTivator), EBV (BZLF-1-and EBNA-1-PepTivator), CMV (pp65-Peptivator), or BKV (LT-Ag-pepmix) antigens for 6 days. On day 6, cultured cells were added to 5610 6 post-thaw and adherent monocytes and re-stimulated with a peptide pool. In addition, IL-15 (R&D Systems) was added at 5 ng/ml on days 3 and 9. On day 12, seVirus-T-cells were harvested and used for several analyses. To determine crossreactivity with other strains of adenovirus, PBMCs were stimulated and expanded as described for the seVirus-T-cells, with the exception that different subgroup-specific peptides were used prior to streptamer analysis. For the monitoring of virus-specific T-cells in donors and patients, PBMCs were mostly stimulated for 6 days, without a second re-stimulation step and IL-15 treatment.

IFN-c ELIspot assays
Five 610 5 IL-2 (20 IU/ml)-expanded T-cells/well (see above and Fig. 1A) were seeded and stimulated for 24 h with pools of ADV-derived peptides (1 mg/ml each) or HIV peptide (1 mg/ml) as negative control. As positive control, 10 mg/ml PHA-L (Roche Applied Science, Indianapolis, IN) was used. IFN-c was detected using the Human IFN-c ELIspot kit (MabTech, Nacka Strand, Sweden) according to the manufacturer's instructions. Pools eliciting positive responses were split into reactions containing individual peptides and tested again against the respective donor. The cut-off value for a positive response was more than 5 spots per 10 5 cells exceeding background levels.

General gating strategy and cut-off values
First, beads (if used) were defined. Viable cells were addressed by their appropriate position in the SSC versus FSC plot. Notably, for multimer analysis, CD19 + B cells were excluded to avoid false positive results. The analyses were performed either on a FACS LSRII or a LSRFortessa, and the FACSDiva (all BD, Biosciences, CA) was used for data evaluation. The limit of detection for the multimers and the IFN-c-CSA was defined as .10 positive events and a 5-fold increase compared to the individual negative control.

CFSE labeling
PBMCs or PHA blasts (10 7 /ml) were resuspended in PBS (0.1% BSA,, Sigma-Aldrich) and labeled with 3 mM CFSE (Sigma-Aldrich) for 10 min at 37uC. The reaction was quenched with 1 ml of Octaplas for 5 min at room temperature (RT). Cells were washed twice with PBS and adjusted to a density of 10 6 cells/ml in AIM-V+. After incubation o/n at 37uC/5% CO 2 they were used for proliferation or cytotoxicity assays.

Proliferation assay
CFSE-labeled PBMCs (2.5610 6 ) were expanded for 12 days according to the IL-15-based in vitro expansion protocol described above (Fig. 1D), with the exception of using Cell Proliferation Dye (CPD) eFluor 670 (Ebioscience, San Diego, CA)-labeled monocytes instead of unlabeled monocytes, which enabled exclusion by gating. On day 12, cells were scraped, washed and analyzed by flow cytometry using Trucount TM tubes to determine the percentage and absolute cell number of viable proliferating cells.
positive, gray bars those negative for the respective allele. Significant differences are indicated (*, p.0.05, Fishers exact test). D) Schematic drawing of the IL-15-based in vitro expansion protocol for FACS analysis. E) Percentages of A*02 ADV-streptamer + T-cells including representative dot plots are given. F) Specific lysis is shown of autologous and allogeneic A*02 mismatched PHA-blasts (target cells), unloaded-or HAdV-A*02 (LLD) peptideloaded, induced by A*02-multimer-sorted HAdV-T-cells. The total number of ''dying'' target cells/well and percentage values were evaluated from each sample. Two representative dot plots are shown. G) A*01, A*24, B*07, and B*35 HLA-dependent ADV-specific multimer + (streptamer or pentamer) T-cells among CD8 + T-cells, before (day0) and after 6 and 12 days of expansion are shown. Notably, for some donors, multimer analyses were only performed at day 0, 6 and/or 12. H) Subgroup C-derived streptamer staining of PBMCs stimulated with specific peptides for 12 days, representing different adenoviral subgroups, as indicated. The percentage of streptamer + T cells among CD8 + is shown. Asterisks represent subgroup recognition. doi:10.1371/journal.pone.0059592.g001 MLR 10 5 CFSE-labeled PBMCs or seHAdV-T-cells were incubated with 10 5 autologous or allogeneic 30Gy-irradiated PBMCs in 100 ml AIM2V+ in a 96 well (round bottom) microtiter plate. On day 7, the cell suspension was transferred into Trucount TM tubes and residual alloreactivity, represented by the total number of viable proliferating (CFSE low) cells, was analyzed by flow cytometry.

Cytotoxicity assay
The cytolytic activity of seVirus-T-cells was assessed by flow cytometry. Notably, only viable cells (.70%), sorted on a FACSAria, were used. CFSE-labeled PHA targets were pulsed with the appropriate viral peptides or peptide pools for 2 hours or o/n, respectively. Unpulsed and control peptide-pulsed targets were used as negative controls. Autologous and allogeneic targets (1.25610 4 ) were mixed with seVirus-T-cells at a ratio 1:20. Of note, due to low numbers of A*02-sorted HAdV-specific T-cells, only 1.9610 3 target cells were used. 4 h after incubation at 37uC, the cell suspension was transferred to Trucount TM tubes and stained with DAPI (0.03 mg/ml). The absolute number of late apoptotic/necrotic targets (CFSE+/DAPI+) was analyzed. Thawed seHAdV-T-cells were cultured for 3 days before cytolytic acitivity was tested.

Ethics Statement
Cells from donors and patients were obtained upon approval from the local Ethics Committees of Tübingen and Vienna (EK Nr.514/2011) and (EK Nr.024/2011) and informed consent.

Statistical analysis
Fisher's exact test, p = 0.05 was used for IFN-c -ELIspot data and students t test analysis was employed to determine the statistical significance (P) of all other findings. Data are shown as mean value with standard deviation and/or range.

Prediction, identification and characterization of novel and known human ADV-epitopes
In order to identify new adenoviral epitopes for CD8 + T-cell responses, we first predicted epitope candidates for the frequent HLA-types A*02, A*01, and A*24 using the SYFPEITHI software (www.syfpeithi.de). We chose two widespread adenoviral strains, Ad2 and Ad5, and focused on three proteins: protein II (hexon, major capsid protein), protein VIII, (minor capsid protein) [35], and E1A (an early antigen). We selected the top-scoring 2% of sequences for synthesis, and included published epitopes, resulting in 29 peptides for the HLA-A*02, 21 for HLA-A*01 and 21 for HLA-A*24 (Table S1 in File S1). All peptides were analyzed for their capacity to stimulate CD8 + T-cells as defined by IFN-c-ELIspot detection. To avoid overlooking HAdV-specific T-cells with very low frequency, PBMCs from at least 16 appropriate donors sharing the HLA allele presenting the peptide were stimulated with the peptide together with IL-2, and expanded for 13 days (Fig. 1A). Peptides were defined as immunodominant if the ELIspot response was specifically positive in more than 50% of appropriate, and negative in most inappropriate donors. These criteria were fulfilled for the novel E1A-based peptide LLD (A*02) as well as for the known hexon-peptides LTD (A*01), and TYF (A*24) which showed an IFN-c response in 49/74, 68/73, and 44/ 58 cases, respectively (Fig. 1B, and Table S1 in File S1). All other known hexon-based peptides for YVL and TFY (both A*02) mediated either low responses or nonspecific recognition (Fig. 1C).

Applicability of multimers to detect very rare HAdVspecific T-cells in healthy donors
Using the LLD-based HLA-A*02 streptamer we assessed the functionality of this novel and of four other known epitopes complexed in multimers of HLA-types A*01, A*24, B*07 and B*35. Due to the very low frequency of HAdV-specific T-cells in freshly drawn blood, we cultured PBMCs for 6 to 12 days using HAdV-specific 9-mer peptides or 15-mer pepmixes and IL-15 as stimulants (Fig. 1D). After 12 days of expansion, A*02 streptamerpositive T-cells were reliably detectable in 5/5 HLA-A*02 positive donors, with frequencies ranging from 0.6 to 8.6% of CD8 + T cells, representing an increase of 0.8 to 2 logs as compared to day 0 (Fig. 1E). The specificity of the new A*02-streptamer was confirmed by its failure to bind to CD4 + T-cells in HLA-A*02 matched, and to CD8 + T-cells in HLA-A*02 mismatched donors (data not shown). In addition, specific killing between 38% and 74% of A*02-multimer purified-T-cells could be achieved, if LLD peptide-loaded target cells were used with different peptide concentrations. No specific killing was observed when A*02mismatched target cells were used ( Figure 1F). Regarding the known epitopes, only 5/25 donors were determined positive with multimers on day 0 (range 0.06-0.2% of CD8 + T-cells). On days 6 and 12, 12/13 and 25/25 of the matched donors were clearly positive, with a range from 0.09-2.1% to 0.64-28% of CD8 + T cells (Fig. 1G), representing a 1-3 log increase in the frequency of multimer-positive T-cells.
To address cross reactivity, we tested whether common subgroup-C-derived HAdV-streptamers detect HAdV-specific Tcells specific for subgroups A-F. Therefore, PBMCs were expanded for 12 days with peptides derived from hexon proteins of different HAdV subgroups (A-F), followed by staining with subgroup-C-derived HAdV-streptamers. The subgroup-C-derived streptamers restricted to A*01 and B*07 did detect HAdV-specific T-cells from all six different subgroups. For the A*24-and B*35subgroup-C-derived streptamers, HAdV-specific T-cells were neither detectable from subgroups A and B nor from A and F (Fig. 1H). Notably, the new E1A-derived A*02 peptide is not conserved in other subgroups, as determined by the basic local alignment search (blast), and can therefore not be employed to  detect HAdV-specific T-cells from other subgroups (data not shown). Beside HAdV, similar results with streptamers were seen for the detection of EBV and BKV-specific T-cells, before and after expansion ( Figure A and B in Figure S1 in File S1). Whereas all 6 EBV-streptamers showed reliable staining results ( Figure. A in Figure S1 in File S1) for BKV, only the B07-restricted streptamer ( Figure B in Figure S1 in File S1) was functional. These data indicate that novel and known multimers, in combination with a short in vitro expansion period, is a reliable tool to monitor virusspecific T-cells.
In analogy to the multimer results, a short expansion period was also necessary to detect HAdV-, EBV-and BKV-specific T-cells when using the IFN-c-CSA (Figure C -E in Figure S1 in File S1).

Reconstitution of HAdV-specific T-cells in the context of HAdV infection following HSCT
By applying the expansion protocol for diagnosis, we assessed whether the presence of CD8 + and/or CD4 + HAdV-specific Tcells in patients correlated with clearance of adenoviral load in stool and blood. All blood samples were from 10 patients positive for HAdV in stool (range 6610 2 -2610 10 copies/gram), 3 of them were also viremic (range 7610 2 -6610 7 copies/ml). T-cell analyses performed in 2 patients with multimers and in 1 patient with the IFN-9-CSA during HAdV-infection showed no detectable HAdV-specific T-cells ( Fig. 2A and B). After clearance, however, samples from 9/9 patients stained with multimers (after expansion or magnetic selection) showed positive results. In 4/4 cases, CD8 + and CD4 + HAdV-specific T-cells were also detectable by the IFNc-CSA ( Fig. 2A and B and Table 1). Of note, HAdV-specific T cells had been detectable in all respective donors prior to stem cell donation (data not shown). More detailed analysis of patient 3 demonstrated that, shortly after viral clearance, multimer + HAdVspecific CD8 + T-cells were clearly detectable at day 23 post HSCT and further increased during the following days (Fig. 2C). After 18 months, however, a clear population of HAdV-specific T-cells was only seen after short term cell expansion (Fig. 2C).
To determine the phenotype of CD8 + HAdV-specific T-cells in patients who had cleared adenoviral infection several months before, multimer + HAdV-specific T-cells were magnetically enriched prior to analysis. HAdV-multimer + T-cells revealed clear populations of central memory T-cells (TCM) (median: 30%) and effector memory T-cells (TEMs) (median: 65%) emphasizing a prominent role of TCMs in maintaining long-term immunity in patients (Fig. 2D). Taken together, these results show that CD4 + and CD8 + HAdV-specific T-cells include high proportions of TCMs, and that their presence correlates with clearance of HAdV load in patients.

Generation and phenotypic characterization of seHAdV-T-cells for potential clinical use
Based on the protocol used for diagnosis (Fig. 1D), we generated seHAdV-T-cells (short-term expanded human adenovirus-specific T cells) with GMP-compliant adaptable peptide mixes and analyzed the absolute cell number and function in more detail. Whereas the starting cell number of 5610 6 PBMCs was slightly reduced to a median of 3.7610 6 cells after 12 days of expansion (Fig. 3A), the total number of HAdV-streptamer + T-cells increased 435 fold from 270 to 117513 (Fig. 3B). For clinical use, the number of starting fresh/frozen PBMCs can be easily scaled up to 25610 6 PBMCs (instead of 5610 6 ),which should result in sufficiently high cell numbers for both treatment and quality control analyses.
To assess the percentage values and total cell numbers of viable proliferating seHAdV-T-cells after in vitro expansion in 24 well plates, PBMCs were labeled with CFSE prior to culture. Compared to unstimulated controls (total cell number (61000): 21613.8 or percentage value: 16.8), HAdV-peptide pool and IL-15-driven stimulation resulted in high numbers (61000) (479,56194) and percentage values (85.5%) of proliferating viable T-cells (Fig. 3C). But also the number of HAdV-streptamer + T-  cells/ml within the culture was high (mean: 33.2/ml) in HAdVpeptide pool-and IL-15 stimulated cells, whereas HAdV-specific cells were not detectable in unstimulated or control peptide(-MAGE-A1)-stimulated PBMCs (data not shown).
Phenotypic analysis revealed that, after expansion, 71%63.8 were CD3 + T-cells including 24%62 CD8 + T-cells and 42.4%65.3 CD4 + T-cells (Fig. 4A). Whereas the percentage of TCMs (21.6%63.5) and TEMs (56.8%63.9) was slightly increased in CD8 + T-cells, no significant change was seen within CD4 + T-cells ( Fig. 4B and C). The percentage of TCMs within the streptamer + population was also not significantly altered (Fig. 4D), indicating that the expansion procedure did not lead to terminal cell differentiation. Of note, cultures supplemented with IL-15 showed highest numbers of HAdV-multimer + T-cells/ml (3.5-fold increase) compared to cultures stimulated in the absence of cytokines, and a 2-fold increase compared to IL-2 or IL-7 stimulation, but no detectable influence on their phenotype (data not shown).

Strongly reduced proliferative capacity of seHAdV-T-cells upon alloantigen stimulation
Next we tested the alloreactive potential of seHAdV-T-cells. In 11 out of 14 allogeneic-pairs, the mean proliferative response of allogeneic PBMCs was about 1.2 log higher compared to that of seHAdV-T-cells ( Fig. 5A and B). Only 3 combinations showed comparable residual alloreactivity to allogeneic-PBMCs (combination: L, M, N) (Fig. 5A). Of note, a closer look at HAdVmultimer + T-cells within the culture revealed that they divided only once representing low if any alloreactivity (data not shown). In addition, we could show that the median alloreactive potential of control short-term expanded T-cells (stimulated with MAGE-1A and IL-15) was similar to that of PBMCs, and 0.9 log higher than that of seHAdV-T-cells (Fig. 5C). These results support our observation that the alloreactive potential of seHAdV-T-cells is strongly reduced after in vitro expansion.

seHAdV-T-cells are highly functional and specific, and fail to kill unpulsed-allogeneic target cells
The capacity of seHAdV-T-cells to express activation and cytotoxic markers, such as IFN-c, CD137 and CD107a, was highly increased after restimulation with HAdV-pepmix-pulsed monocytes (Fig. 6A-D). Furthermore, seHAdV-T-cells produced both IFN-c and TNF-a ( Fig. 6E and F). The lysis of peptide or of peptide-pool-pulsed autologous target cells by seHAdV-T-cells was about 4-fold higher as compared to unpulsed targets (Fig. 6G). Similar lysis was seen when HAdV-pulsed-allogeneic target cells, matched in only one MHC class I or II antigen, were used (Fig. 6G). Unpulsed or control (CMV)-pulsed autologous or allogeneic targets were not recognized, which strongly supports their specificity and the loss of alloreactivity of seHAdV-T-cells ( Fig. 6G and data not shown). This cytotoxic activity was also maintained in post-thaw seHAdV-T-cells (Fig. 6H).

Generation of short-term-expanded virus-specific T-cells against several viruses for potential clinical use
Our protocol to generate seHAdV-T-cells was adapted to expand also EBV-, CMV-and BKV-specific T-cells. Whereas the median absolute number of cells was not significantly altered after expansion, the total cell number of streptamer + T-cells was 1-2 log increased, irrespective of the type of virus-specific T-cells ( Figure  A, D, and G in Figure S2 in File S1). As seen for HAdV, the proportions of TCM and TEM of CD4 + and CD8 + virus-specific T-cell lines hardly changed during the expansion period ( Figure B, E, and H in Figure S2 in File S1). Specific lysis by all cell lines was only observed for peptide-loaded autologous and allogeneic, but not for unloaded allogeneic target cells mismatched or matched in only 3 alleles. (Figure C, F, and J in Figure S2 in File S1).

Discussion
PCR screening for HAdV following allogeneic HSCT allows early detection of impending invasive HAdV-infections, and timely preemptive antiviral treatment [8,36]. Recently, it has been shown that the reconstitution of HAdV-specific T-cells plays a pivotal role in the clearance of HAdV infection [11,12,13,14]. Few groups suggested that combined monitoring of viral load and virus-specific immunity by ELIspot, tetramer staining or the IFNc-CSAs has a clinical impact on the therapeutic intervention for pediatric allogeneic HSCT patients [36,37,38,39]. One of the most sensitive and fastest tools to monitor virus-specific T cells are multimers. Therefore, efforts have been made over the past years to identify HAdV-derived MHC class I-restriced as well as MHC class II-restricted epitopes [14,27,40,41]. Nevertheless, the number of published class I epitopes remained rather low, and the focus has been on hexon, the main capsid protein of the virion. Our analyses of different proteins identified an immunodominant A*02-restricted HAdV subgroup C-specific epitope derived from the E1A protein. Based on this epitope, a functional multimer was produced that showed reliable staining results. To our knowledge, this is the first functional A*02-restricted multimer specific for HAdV. Other promising previously published A*02-restricted candidates failed to be useful for the production of functional multimers. Although in a pilot study the treatment of a single patient with pentamer+ CD8+ T-cells specific for HAdV was not successful [22], more studies will be necessary to determine whether polyclonal CD4 + and CD8 + HAdV-specific T-cells are necessary for successful immunotherapy.
However the detection of HAdV-specific T-cells is hampered by the low frequency of HAdV-specific T-cells in peripheral blood [14,17,18]. Even with the more sensitive ELIspot technique (procedure needing 3 days), false negative results cannot be excluded if uncultured PBMCs are used [29]. The prevalence of HAdV in the Caucasian population is supposed to be above 80% [32]. We demonstrated that in most cases a 6 day expansion period was sufficient to obtain clearly positive multimer-based results in 90% of donors. Similar results were seen when the IFNc-CSA instead of streptamers was used after expansion. Moreover, in addition to the generally known cross-reactivity of reactive Tcells between HAdV-subgroups [24], we could show that A*24 matched in only a single MHC I, or in only a single MHC II, unloaded (white bar), HAdV-peptide-(grey bar) or HAdV-peptide pool (black bar)-loaded, induced by seHAdV-T-cells. The total number of ''dying'' target cells/well was evaluated from each sample. Based on these, a summarizing graph shows the percentage of dying target cells related to unloaded autologous (auto) target cells which is set to 100%. Representative dot plots, indicating ''dying'' cells/well, are shown. H) Specific lysis of allogeneic, unloaded (white bar) or HAdV-peptide pool (black bar)-loaded target cells induced by post-thaw seHAdV-T-cells. Based on total cell number of dying cells, a summarizing graph shows the percentage of dying target cells related to unloaded target cells, which is set to 100%. Significance vs unloaded autologous targets: ns = not significant, *, p, = 0.01.***, p, = 0.001. doi:10.1371/journal.pone.0059592.g006 and B*35 streptamers are not able to recognize subgroup A-or subgroup A and F-derived HAdV-specific T-cells, respectively. This is a very important piece of information, if HAdV-multimers are supposed to be used for diagnosis or future multimer-based therapies.
So far, only one study including seven pediatric and six adult patients showed that, besides CD4 + [7,12,13], also CD8 + virusspecific T-cells [14] are detectable after HAdV clearance in patients after HSCT. We confirmed these results with additional 10 pediatric patients, using the IFN-c-CSA or multimer analysis of expanded or magnetically selected cells. Out of 7 patients who were HAdV positive in stool and HAdV-negative in plasma, two had transient enteritic symptoms, which were attributed to Clostridium difficile infection and gut GvHD respectively, and five remained asymptomatic. HAdV-specific T-cells were detectable in all patients after clearance of infection. Of note, 3/10 patients who cleared HAdV infection, had received T-cell depleted grafts from haploidentical donors (,10 5 T-cells/kg BW). This may indicate that, even in the context of profound lymphopenia, viral antigen can trigger virus-specific immune responses. None of these patients reactivated HAdV infection within several months after HSCT, which could be due to the presence of residual HAdV-specific Tcells with characteristics of TCM and TEMs, as seen in all patients analyzed. It further strengthens the suggested important role of TCMs in mediating long-term protection [16,42]. Although the optimal time point for a T-cell immunotherapy in clinical practice is unknown, recent data suggest that the generation or isolation of virus-specific T-cells should occur within the first weeks after detection of a high viral load (.10 6 copies) in stool in order to allow prompt treatment in case of invasive HAdV infection [8,18,43,44].
Addressing these criteria, we used 5610 5 fresh or frozen PBMCs (including 2.5 to 3.5610 6 CD3+T cells) as starting material and managed to increase the number of HAdVstreptamer + T-cells 435-fold within 12 days. The slightly reduced cell number from 5 to 3.7610 6 could be explained by the delayed stimulation with IL-15 and loss of unspecific T cells during the culture period. Of note, prophylactic cryo-preservation of PBMCs from appropriate donors prior to HSCT proved helpful to safe time. Culture of both fresh and frozen PBMCs resulted in high values of total proliferating cells (up to 85.5%) and at least 68.868%, if only highly proliferating cells were gated (data not shown). By using a similar but more strict gating strategy, an approach by Gerdemann et al., who also used peptide mixes but preferentially combined with IL-4 and IL-7, resulted in 52.4% of proliferating cells. In contrast to this work, natural killer (NK) and cytokine-induced killer (CIK) cells, which are assumed to be beneficial in the prevention and treatment of relapses [45,46], are still present after 12 days of expansion. Although our work shares certain features with the protocol by Gerdemann et al., it differs considerably with regard to the stimulation procedure itself, and consequently also with regard to the nature of expanded T-cells and focuses mainly on the safety of our seHAdV-T-cells, as mentioned above.
In our experiments, all seHAdV-T-cells were highly functional, and not only able to lyse antigen-pulsed autologous but also antigen-pulsed allogeneic target cells, if at least one MHC class I or II was matched, which, to our knowledge, had not been shown in that detail before and could be of high relevance for haploidentical or third party donors. For IL-4-and IL-7-driven expansion of HAdV-specific T-cells, at least a 16 day expansion period was necessary to obtain sufficient cytolytic activity [31]. In contrast to previous studies, the killing of MHC class I-matched HAdV-pulsed targets was much higher compared to MHC II-matched targets, indicating that mostly HAdV-specific CD8 + T-cells are involved. This discrepancy might be explained by the use of 15mer instead of 30mer peptides, since the 30mer peptides were described to contain predominantly CD4 + epitopes [29,47]. In addition, even post-thaw seHAdV-T-cells were able to kill partially matched allogeneic target cells albeit to a lower extend compared to fresh seHAdV-T-cells. This finding would enable the prophylactic generation and infusion on demand.
To address the potential risk of seHAdV-T-cells to induce GvHD, the percentage of residual unspecific T-cells -represented by non-proliferating T-cells during the culture period -was analyzed. Although about 15% of the seHAdV-T-cell population did not proliferate as indicated by their unchanged high CFSE load, MLRs showed that, in at least 11/14 donor/recipient pairs, the alloreactive potential of seHAdV-T-cells was reduced by 1.2 log compared to unmanipulated PBMCs. Only 1/14 cases (combination L) showed alloreactivity signals similar to those of the control PBMCs, although in the cytotoxic assay no significant alloreactivity was seen (data not shown). For the other two, the MLR might have failed since the alloreactivity of PBMCs was near the background level. However, residual alloreacivity in some combinations could also be explained by the fact that, in contrast to other studies [16,29] only one MHC allele was matched. Reduced or even absent alloreactivity of seHAdV-T-cells was further confirmed by comparison with control T-cells expanded with MAGE-A1-peptide pools, or by the failure to recognize and lyse allogeneic target cells. The fact that residual alloreactivity, despite expansion, can never be completely excluded, was also shown by other groups [16,18,29]. In addition, Chen et al. showed that the capacity of in vitro expanded alloreactive T-cells to survive and expand in vivo is limited [48]. This was also supported by Melenhorst et al. who showed no correlation between in vitro results and in vivo data concerning alloreactivity [49]. Clinical evidence supports that even a small number of virus-specific Tcells (like 10 3 to 10 4 /kg body weight), which is easily achievable with our protocol, is sufficient to attain therapeutic efficacy, and infusion of such low lymphocyte numbers would further minimize the risk for GvHD [18]. By adapting our short-term expansion protocol, we were also able to generate high numbers of functional virus-specific T-cells directed against CMV, EBV, and BKV. Also these cells were able to lyse autologous and allogeneic peptideloaded target cells, although, in some cases, significant lysis was hindered by the high intra-individual variability.
In conclusion, we optimized tools for diagnosis of HAdVspecific T-cells and underline the importance of CD8 + HAdVspecific T-cells in the clearance of HAdV-load in patients. In addition, we were able to generate efficient virus-specific T-cells mainly against HAdV but also CMV, EBV, and BKV within 12 days. The usage of fresh or frozen PBMCs further enables an immunotherapy protocol within a short time span, with low cost and effort, and with the potential for broad clinical application.