Identification of the murine firefly luciferase-specific CD8 T-cell epitopes

Abstract

In vivo bioluminescence imaging of reporter enzymes has proven to be a uniquely powerful tool that allows the study of the biology of viral and nonviral gene transfer agents. Cost-effective, noninvasive, longitudinal gene transfer studies in individual animals yield important information, which can influence the design of subsequent preclinical studies. The broad and expanding use of luciferase transgenes, specifically firefly luciferase, has prompted the study of luciferase-specific T-cell activation following in vivo gene transfer. Herein, we report the mapping of the dominant T cell epitope in C57BL/6 mice (LMYRFEEEL) and the mapping of the dominant and minor T-cell epitopes in BALB/c mice (GFQSMYTFV and VPFHHGFGM, VALPHRTAC, respectively). These CD8 T-cell epitopes can be used to monitor cellular responses in vivo as well as be important tools in studies designed to suppress transgene-specific T cells.

Introduction

More than two decades have passed since the first gene transfer application of the luciferase transgene.¹ Since then the use of the firefly luciferase gene of Photinus pyralis as a reporter transgene (luc2) has had an enormous positive impact on the design of informative and cost-effective in vivo gene transfer studies.² Indeed, the combinatory use of the luc2 gene and bioimaging has proven to be a powerful technology, which allows for noninvasive longitudinal studies in individual vector-treated animals. More importantly, in vivo bioluminescence imaging can provide unique informative insights into vector biology ranging from onset of gene expression to biodistribution of transgene expression.

In viral-mediated gene transfer studies, the activation of destructive transgene or vector-specific T cells has been implicated in the loss of transgene expression in various tissue targets.³ It is now well established that the immunogenicity of a transgene product can be influenced by various parameters that include the genetic background of the experimental mice, vector dose and target tissue. The most studied reporter transgenes include β-galactosidase and enhanced green fluorescence protein (EGFP). For both of these reporter transgenes, the CD8 T-cell epitopes have been identified and mapped in C57BL/6^4,5 and BALB/c^6,7 mice, providing useful tools to study transgene-specific cellular responses in the context of gene transfer.

The escalating use of luc2 as a reporter transgene in gene transfer studies warrants the investigation of luc2-specific T-cell activation following in vivo gene transfer. Indeed, the identification of luc2-specific T-cell epitopes could provide a valuable tool to study in vivo luc2-specific cellular responses.

Results and discussion

The use of luc2 together with bioluminescence imaging is a powerful tool to noninvasively study long-term vector-mediated gene expression and vector biodistribution in vivo. It is possible, however, that immune recognition of luciferase-expressing cells may negatively impact the outcome of vector-mediated gene transfer conducted in the two most commonly studied experimental mouse strains, namely C57BL/6 and BALB/c mice, and compromise proper interpretation of negative results. We sought to investigate the immunogenicity of the luciferase transgene in vivo and to identify the dominant T-cell epitopes in C57BL/6 and BALB/c mice.

Identification of the H2-D^b-restricted T-cell epitope in the luc2 protein

An Ad.Hu5 vector encoding the luc2 protein product was injected intramuscular (i.m.) into C57BL/6 mice. Mice injected with Ad.Hu5.LacZ vector served as a negative control. Luc2 expression was confirmed by imaging at day 8 (Figure 1a), the peak of T-cell activation, and the spleen was harvested from the treated mice. Splenocytes were isolated and stimulated in vitro with an array of peptide pools (composed of 15-mers with 10 amino acid overlapping sequence) which comprised a peptide pool matrix (Figure 1b) that corresponded to the entire luc2 protein and analyzed using the ELISpot assay which determines T-cell activation by measurement of interferon-γ (IFN-γ) secretion. As predicted, there were no IFN-γ responses specific to the luc2 protein in the negative control (Ad.Hu5.LacZ vector treated) mice (data not shown). In C57BL/6 mice, luc2-specific T-cell responses were evidenced in pools 6, 7 and 18 (Figure 1c) and based on the matrix pool design neighboring peptides 66 and 67 (Figure 1b) were identified as containing the dominant luc2-specific T-cell epitope(s). Subsequent analysis of pools 6, 7, 18 and the peptides 66 (RVVLMYRFEEELFLR) and 67 (MYRFEEELFLRSLQD) by ELISpot (Figure 2a) confirmed that the T-cell epitope was restricted between these two peptides. To define the T-cell population which was expressing IFN-γ following stimulation, splenocytes harvested from the Ad.Hu5.luc2 i.m. treated C57BL/6 mice were stimulated in vitro with peptide pools 6, 7, 18 and peptides 66 and 67 and analyzed by intracellular IFN-γ staining (ICS) (Figure 2b). The T-cell response following stimulation was found to be CD8 mediated (Figure 2b). Staining with a CD4 T-cell antibody demonstrated the absence of CD4 T-cell population activated in response to luc2 (data not shown).

Figure 1.

Identification of peptide pools containing luc2-specific T-cell epitopes in C57BL/6 and BALB/c mice. (a) BALB/c and C57BL/6 mice were imaged by the Xenogen imaging system to confirm Luc2 expression at the time of necropsy (day 8), (b) The design of the peptide pool matrix. Splenocytes from (c) C57BL/6 and (d) BALB/c mice immunized with 5 × 10¹⁰ particles (p) of Ad.Hu5.luc2 were isolated from 3 mice at day 8 after injection. Splenocytes were stimulated in vitro with the peptide pools corresponding to the entire luc2 peptide library. Cells were seeded at 1 or 2 × 10⁵ cells per well. T-cell responses were evidenced in C57BL/6 mice as spots in peptide pools 6, 7, 18 and in BALB/c mice as spots in peptide pools 1, 4, 12, 16, 17 and 18. Controls: no peptide (negative), PMA/I (positive, not shown). Experiments were performed on three separate occasions and representative results are shown.

Figure 2.

Identification of the H2-D^b-restricted T-cell epitope in the luc2 peptide. Splenocytes (2 × 10⁵ cells/well) from C57BL/6mice immunized with 5 × 10¹⁰ particles (p) of Ad.Hu5.luc2 were isolated from 3 mice at day 8 after injection and (a) stimulated in vitro with the three positive peptide pools (6, 7 and 18), and the two neighboring 15-mer peptides (66 and 67) that contained the CD8 T-cell epitope LMYRFEEEL, (b) ICS was also performed on these splenocytes stimulated with peptide pools 6, 7, 18 and peptides 66 and 67. Staining was performed using an anti-mouse CD8⁺ antibody conjugated to fluorescein isothiocyanate (FITC) and an anti-mouse interferon-γ (IFN-γ) antibody conjugated to phycoerythrin (PE) to determine the frequencies of IFN-γ-producing CD8⁺ T cells. Negative control: no peptide. The intensity of CD8 T-cell activation by the mapped T-cell epitope (LMYRFEEEL) was also confirmed. Experiments were performed on three separate occasions and representative results are shown.

To identify the potential major histocompatibility complex (MHC) class I binders in RVVLMYRFEEELFLRSLQD, two different prediction programs were used. These were SYFPEITHI⁸ and HLA-BIND-Bio-Informatics and Molecular Analysis Section (BIMAS).⁹ The SYFPEITHI and BIMAS programs predicted CD8 T-cell epitopes restricted to H2-K^b and H2-D^b (Table 1). Some candidate epitopes were identified by both algorithms as shown in Table 1. The high scores of specific epitopes suggested that the CD8 T-cell response may be restricted by either or both of the class I molecules, H2-K^b and H2-D^b MHC. The highest-ranking candidate epitope sequence for the H2-K^b MHC class I molecule as predicted by BIMAS, albeit with a low score, was LMYRFEEEL, which was not predicted by the SYFPEITHI program. The highest-ranking candidate epitope sequence for the H2-D^b MHC class I molecule as predicted by both the SYFPEITHI and BIMAS programs was also LMYRFEEEL. The high scores by both prediction programs (Table 1) suggested a high binding affinity of LMYRFEEEL for H2-D^b.

Table 1.

Predicted MHC class I peptide binders

Strain	Peptide sequence	MHC class I	Score
C57BL/6	LMYRFEEEL	H2-Kb	0.66 (B); 0 (S)
		H2-Db	16.6 (B); 17 (S)
BALB/c	GFQSMYTFV	H2-Kd	345.6 (B); 22 (S)
		H2-Dd	0.9 (B); a (S)
	VALPHRTAC	H2-Kd	2.88 (B); 10 (S)
		H2-Dd	0.03 (B); a (S)
	VPFHHGFGM	H2-Kd	6.0 (B); 2 (S)
		H2-Dd	0.3 (B); a (S)

Abbreviations: B, BIMAS; S, SYFPEITHI; MHC, major histocompatibility complex.

The discrepancy of the scoring system as well as the nonidentification of an epitope by SYFPEITHI, which was scored by the BIMAS program can be attributed to the limitations of the epitope binding prediction programs in general.^10,11

^aNot identified.

A 9 amino-acid peptide spanning the LMYRFEEEL T-cell epitope was generated and tested following i.m. injection of Ad.Hu5.luc2 in C57BL/6 mice. The IFN-γ ELISpot and ICS assays confirmed that LMYRFEEEL was the luc2 CD8 T-cell epitope in C57BL/6 mice (Figures 2a and b) and resulted in higher T-cell frequencies than pools 6, 7, 18 and peptides 66 and 67, as expected. We then evaluated the relative affinity of LMYRFEEEL for the class I MHC molecule H2-K^b or H2-D^b by titrating the amount of peptide necessary for T-cell activation in vitro. Specifically, splenocytes isolated from C57BL/6 mice that were immunized with Ad.Hu5.luc2 8 days earlier were stimulated with LMYRFEEEL at concentrations ranging from 2 to 9.37 × 10⁻⁴ μg ml⁻¹ (Figure 4). T-cell activation was measured by counting the number of IFN-γ-secreting cells by ELISpot. The LMYRFEEEL epitope appeared to have moderate binding affinity that remained stable even when the peptide concentration was lowered to 0.250 μg ml⁻¹. However, the binding affinity began to steadily decrease from 0.125 μg ml⁻¹ and remained just above background when the peptide concentration was lowered to 9.37 × 10⁻⁴ μg ml⁻¹ (Figure 4).

Figure 4.

Dose titration of the H2-D^b and H2-K^d-restricted CD8 T-cell epitopes necessary for T-cell activation in vitro. Splenocytes isolated from C57BL/6 and BALB/c mice, immunized with Ad.Hu5.luc2 8 days previously, were stimulated with dose-decreasing peptide concentrations of the dominant 9-mer epitopes LMYRFEEEL and GFQSMYTFV, respectively. Peptide-specific IFN-γ-expressing splenocytes, assayed by ELISpot, were calculated by subtracting the number of spots in the negative control from the number of spots in peptide-containing wells. Triangles represent SFU/10⁶ cells following stimulation with LMYRFEEEL (average of duplicate samples), and squares represent SFU/10⁶ cells following stimulation with GFQSMYTFV (average of triplicate samples). Error bars are representative of the standard deviation. Experiments were repeated on two separate occasions.

Identification of the H2-K^d-restricted T-cell epitopes in the luc2 protein

BALB/c mice were injected i.m. with Ad.Hu5.luc2 vector. Luc2 expression was confirmed by imaging at day 8 (Figure 1a). The spleen was then harvested and splenocytes isolated and stimulated in vitro with the same array of peptide pools comprising the peptide pool matrix (Figure 1b) that corresponded to the entire luc2 protein and analyzed using the ELISpot assay.

In BALB/c mice, luc2-specific T-cell responses were evidenced in pools 1, 4, 12, 16, 17 and 18 (Figure 1d) and based on the design of the peptide pool matrix, peptides 40, 52 and neighboring peptides 60 and 61 (Figure 1b) were identified as containing the dominant and minor luc2-specific T-cell epitope(s). Subsequent analysis of pools 1, 4, 12, 16, 17 and 18 and the peptides 40 (DYQGFQSMYTFVTSH), 52 (PKGVALPHRTACVRF), and the neighboring peptides 60 (ILSVVPFHHGFGMFT) and 61 (VPFHHGFGMFTTLGY) by ELISpot (Figures 3a and c) confirmed that the dominant T-cell epitope was restricted in peptide 40 and that peptides 52, 60 and 61 contained minor epitopes. To define the T-cell population which was expressing IFN-γ following stimulation, splenocytes harvested from the Ad.Hu5.luc2 i.m. treated BALB/c mice were stimulated in vitro with peptide pools 1, 4, 12, 16, 17, 18 and peptides 40, 52, 60 and 61 and analyzed by ICS. The T-cell response following stimulation was found to be CD8 mediated (Figures 3b and d). Staining with a CD4 T-cell antibody demonstrated the absence of CD4 T-cell population activated in response to luc2 (data not shown).

Figure 3.

Identification of the H2-K^d-restricted T-cell epitope in the luc2 protein. Splenocytes (2 × 10⁵ cells/well) from BALB/c mice immunized with 5 × 10¹⁰ (p) of Ad.Hu5.luc2 were isolated from 3 mice at day 8 after injection and (a and c) stimulated in vitro with the six positive peptide pools (1, 4, 12, 16, 17 and 18), and peptides 40, 52, 60, 61 that contained the dominant and minor CD8 T-cell epitopes. (b and d) ICS was also performed on these splenocytes stimulated with peptide pools 1, 4, 12, 16, 17 and 18 and peptides 40, 52, 60 and 61. Staining was performed using an anti-mouse CD8⁺ antibody conjugated to FITC and an anti-mouse IFN-γ antibody conjugated to PE to determine the frequencies of IFN-γ-producing CD8⁺ T cells. Negative control: no peptide. The intensity of CD8 T-cell activation by the mapped T-cell epitopes (GFQSMYTFV, VPFHHGFGM and VALPHRTAC) was also confirmed. Experiments were performed on three separate occasions and representative results are shown.

SYFPEITHI and BIMAS were used to identify the potential MHC class I binders in peptide sequences DYQGFQSMYTFVTSH (peptide 40), PKGVALPHRTACVRF (peptide 52) and ILSVVPFHHGFGMFTTLGY (peptides 60 and 61). Some candidate epitope sequences were identified by both programs as shown in Table 1. The highest-ranking candidate epitope sequence for the H2-K^d MHC class I molecule as predicted by BIMAS, with a very high score, was GFQSMYTFV which was also predicted by the SYFPEITHI program. The next highest-ranked epitope by BIMAS was VPFHHGFGM. Although this epitope was also predicted by SYFPEITHI, it scored lower than the third predicted epitope VALPHRTAC. On the basis of the BIMAS and SYFPEITHI programs the dominant epitope in BALB/c mice appears to be H2-K^d restricted.

All three 9-mer CD8 T-cell epitopes GFQSMYTFV, VALPHRTAC and VPFHHGFGM were generated and tested in BALB/c mice. The IFN-γ ELISpot and ICS assays confirmed that GFQSMYTFV is the dominant luc2-specific CD8 T-cell epitope in BALB/c mice (Figures 3c and d), with frequencies higher than those of pools 4, 16 and peptide 40. VPFHHGFGM and VALPHRTAC were found to be the minor CD8 T-cell epitopes (Figures 3c and d). The relative affinity of the dominant GFQSMYTFV epitope for the class I MHC molecule H2-K^d or H2-D^d was determined by titrating the amount of peptide necessary for T-cell activation in vitro. Specifically, splenocytes harvested from BALB/c mice that were immunized with Ad.Hu5.luc2, 8 days earlier, were stimulated with GFQSMYTFV at concentrations ranging from 2 to 9.37 × 10⁻⁴ μg ml⁻¹ (Figure 4). T-cell activation was measured by counting the number of IFN-γ-secreting cells by ELISpot. The GFQSMYTFV epitope had a very strong binding affinity evidenced by high T-cell responses even when stimulated with 3.75 × 10⁻³ μg ml⁻¹ (Figure 4).

Although peptide 52 contained a CD8 T-cell epitope (Figures 3c and d), when the predicted VALPHRTAC epitope (Table 1) was generated and used to stimulate splenocytes harvested from mice immunized with the Ad.Hu5.luc2 vector there was no evidence of a T-cell response either by ICS or by the more sensitive IFN-γ ELISpot assay (Figures 3d and c, respectively). This example serves to highlight the limitation of the prediction programs (BIMAS and SYFPEITHI) used to predict the CD8 T-cell epitope sequence. Although it is more than likely that an epitope(s) other than VALPHRTAC in PKGVALPHRTACVRF (peptide 52) was the minor epitope(s), we opted against further investigation since the immunodominant H2-K^d epitope was identified (GFQSMYTFV).

It has been reported that transgene products are more immunogenic in BALB/c mice than C57BL/6 mice. Indeed this has been reported for two of the most widely used transgenes LacZ⁷ and EGFP.⁶ In our work, based on the IFN-γ ELISpot and ICS analysis, we also found that luc2 is more immunogenic in BALB/c mice than C57BL/6 mice (P<0.05).

The technology of imaging luciferase transgene expression in vivo is rapidly evolving and with it comes the ability to perform longitudinal gene transfer studies designed to address important issues that include persistence of gene expression, vector biology and host responses to the gene transfer agent and transgene product. We demonstrated that luciferase is immunogenic in two of the most commonly used experimental mouse strains, C57BL/6 and BALB/c mice. Although immunogenicity of a transgene is dependent upon many factors, including vector dose and site of injection, it is warranted that when interpreting differences in the profile of vector-mediated gene expression with time by bioimaging the immunogenicity of the transgene is also considered. In fact, the advantage of bioimaging together with the immunogenicity of luc2 may allow for the development of strategies designed to suppress cellular responses or even tolerize the host to ‘non-self’ transgene expression.

Materials and methods

Firefly luciferase peptide library

A peptide library of the firefly luciferase (luc2) protein sequence (NCBI protein reference AY738222) was synthesized as 15-mers with a 10 amino acid overlap with the preceding peptide (Mimotopes, Clayton, Victoria, Australia) and dissolved in dimethylsulphoxide (DMSO) at approximately 100 mg ml⁻¹. A peptide matrix was created for the luc2 peptide library. Specifically 24 peptide pools were created with peptide pools 1–3 and 13–23 containing 12 peptides each, peptide pools 4–12 contained 11 peptides each and peptide pool 24 contained 3 peptides (Figure 1a). The peptide pools and subsequently the individual peptides were used at a concentration of 2 μg ml⁻¹ to avoid the toxicity associated with higher peptide/DMSO concentrations. In all experiments DMSO concentrations were kept below 0.1% (v/v) in all final assay mixtures.

Mice

For experiments C57BL/6 and BALB/c mice were purchased from Charles River Laboratories (Wilmington, MA, USA) and used at 6–8 weeks of age. Mice were housed under specific pathogen-free conditions at the University of Pennsylvania’s Translational Research Laboratories. All animal procedure protocols were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania.

Adenovirus vectors

E1/E3-deleted replication deficient, Ad.Hu5 vectors expressing firefly luciferase (luc2) or beta-galactosidase (LacZ) (5 × 10¹² particles (p) ml⁻¹) were created as previously described.^12,13

Virus instillations

Groups of three to five mice were anesthetized with ketamine/xylazine and injected with 5 × 10¹⁰ p per mouse of Ad.Hu5 vector diluted in 50 μl phosphate-buffered saline (PBS) given i.m. in the right hindlimb as two 25 μl doses. Mice were killed after 8 days and spleens harvested. All experiments were performed in duplicate (or triplicate) on three separate occasions.

IFN-γ ELISpot assay

Spleens were harvested from the treated mice and transferred into 1 × Liebowitz’s-15 (L-15; Cellgro; Mediatech Inc., Herndon, VA, USA) at room temperature. The tissues were homogenized and filtered through a 70-μm cell strainer. Cells were centrifuged for 5 min at 1600 r.p.m. at 24 °C and the cell pellet resuspended in fresh L-15 media and centrifuged for 5 min at 1600 r.p.m. at 24 °C. The cell pellet was resuspended in complete media (DMEM, 10% heat-inactivated fetal bovine serum, 1% Pen-Strep, 10 mm 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 0.1 mm nonessential amino acids and 10⁻⁶ m 2-mercaptoethanol) and the cells were overlaid onto a Ficoll-Paque gradient layer and centrifuged for 20 min at 2000 r.p.m. at 24 °C. The resulting highly pure population of spleen-derived lymphocytes was harvested from the interface, washed in PBS and centrifuged for 5 min at 1600 r.p.m. at 24 °C. The splenocytes were resuspended in complete media. The assay was performed using the ELISpot Mouse Set (BD Pharmingen, San Diego, CA, USA) according to the manufacturer’s instructions. Briefly, a 96-well ELISpot plate was precoated with 5.0 μg ml⁻¹ of anti-mouse IFN-γ capture antibody overnight (o/n) at 4 °C. Wells were then blocked with complete culture medium for a minimum of 2 h at room temperature (RT). Splenocytes were added to wells at a density of 1 or 2 × 10⁵ cells per well along with the luc2-specific peptide pools. Cells were incubated at 37 °C, 5% CO₂ for 20 h. As positive control phorbol myristate acetate (PMA; 0.05 μg ml⁻¹); ionomycin (I; 1 μg ml⁻¹) (PMA/I)¹⁴ was used. For PMA/I cells were seeded at 2 × 10⁴ per well providing a reproducible cell count of ~100 SFU/2 × 10⁴ cells, thus controlling variations of cell density in each well. Negative control cells were incubated in the absence of peptide. Following o/n incubation, wells were vigorously washed with water, followed by PBS/0.05% Tween-20 and subsequently incubated with 2.0 μg ml⁻¹ of biotinylated anti-mouse IFN-γ detection antibody for 2 h at RT. Following three PBS/0.05% Tween-20 washes, the wells were incubated with 5 μg ml⁻¹ of streptavidin-horseradish peroxidase antibody for 1 h at RT. Wells were washed with PBS/0.05% Tween-20, followed by PBS and developed using the AEC substrate set (BD Pharmingen). Color development was monitored and stopped 5 min later by washing well with distilled water. After drying o/n at RT, spots were counted using an ELISpot reader.

Intracellular IFN-γ staining

Splenocytes from vector-treated mice were stimulated with the luc2 peptide pool(s) for 5 h at 37 °C, 10% CO₂ in the presence of 1 μg ml⁻¹ Brefeldin A (GolgiPlug; BD Pharmingen) and 1 μl ml⁻¹ mouse IL-2 (BD Pharmingen). Control cells were incubated without peptide. After washing, cells were stained with a FITC-labeled anti-mouse CD8 antibody (BD Pharmingen). Cells were then washed and permeabilized in Cytofix/Cytoperm (BD Pharmingen) for 20 min at 4 °C. Cells were washed well, and stained with a PE-labeled anti-mouse IFN-γ antibody (BD Pharmingen) in the presence of Cytoperm wash buffer (BD Pharmingen). After extensive washing, cells were examined by two-color flow cytometry and data were analyzed by WinMDi software.

Imaging

Mice were injected intraperitoneally with 200 μl of 15 mg ml⁻¹ d-luciferin (Caliper, USA) (that is 10 μl luciferin per 1 g body weight). After 5 min mice were anesthetized with ketamine/xylazine and imaged within 10 min of anesthesia using the IVIS Xenogen imaging system. Quantitation of signal was calculated using the Living Image 3.0 Software.

Statistical analysis

Statistical analysis of the data presented was performed using the SigmaStat 3.1 program (SPSS, Chicago, IL, USA). Statistical significance was set at P = 0.05 and statistical power at 0.80. Results are presented as a mean±s.d. Student’s t-test was used for two-group comparisons.