Serpin-6 Expression Protects Embryonic Stem Cells from Lysis by Antigen-Specific CTL¹

Embryonic stem (ES)⁴ cells have recently attracted great attention with regard to cell replacement therapy as a treatment option for injured or dysfunctional organs. When mouse or human ES cells are cultured in suspension in the absence of factors that normally maintain their pluripotency (e.g., feeder layers derived from mouse embryonic fibroblasts and leukemia inhibiting factor (LIF)), they spontaneously develop into spherical structures called embryoid bodies (EB). These multicellular aggregates resemble early postimplantation embryos and contain differentiating cells of endodermal, ectodermal, and mesodermal origin. ES cells have been successfully used to generate neurons, endothelial cells, cardiomyocytes, hemopoietic precursors, keratinocytes, osteoblasts, and hepatocytes (1) and may become a valuable therapeutic modality for tissue regeneration in neurodegenerative disorders (2), spinal cord injury (3), heart failure (4), or diabetes mellitus (5).

In general, nonautologous grafts derived from ES cells will most probably be challenged by the very same mechanisms of allospecific immune responses that are responsible for rejection of non-MHC-matched classical organ transplants. Therefore, immunosuppressive therapy will be required to prevent rejection by alloresponsive host CTL. However, undifferentiated and differentiated ES cells seem to be of low immunogenicity and resistant to killing by activated NK cells (6). Mice reconstituted with human PBMC did not reject undifferentiated or differentiated human ES cells transplanted under the kidney capsule. Even immunization of the reconstituted mice with irradiated human ES cells did not result in subsequent rejection of the transplants (7). Moreover, intraportal injection of undifferentiated rat ES cell-like cells into fully MHC-mismatched rats induced immunologic tolerance, allowing the subsequent long-term acceptance of second-set transplanted cardiac allografts (8). However, the immunoprivileged phenotype of ES cells and their derivatives bears the risk of compromising the immunosurveillance of the transplant. Thus, it might favor expansion and spreading of malignantly transformed cells (e.g., of teratomas). Furthermore, immunologically noncontrolled ES cells may provide a niche for intracellular infectious agents to become a source of systemic seeding of pathogens, which may contribute to the vertical transmission of infectious diseases. How ES cells are dealt with by adaptive immunity has been a recurring issue, but information about cellular immune responses to ES cells is scarce.

To characterize adaptive cellular immune responses to undifferentiated murine ES cells and their differentiated derivatives in greater detail, we used the well-established system of acute murine infection with the lymphocytic choriomeningitis virus (LCMV); because this virus can productively infect almost any murine cell type, large quantities of Ag are produced in infected cells and large amounts of highly cytotoxic primary CD8⁺ CTL can be recovered from mice infected with LCMV (9, 10, 11). Besides, infections with the LCMV are of clinical relevance. Vertical transmission of the LCMV during the first trimester to human fetuses leads to abortions, while infection at later times during pregnancy is an increasingly recognized cause of early neonatal death, hydrocephalus, mental retardation, and chorioretinitis in infants (12, 13, 14). Furthermore, fatal transmission of LCMV by organ transplantation has been recently reported (15).

We show in this study that despite low-level expression of MHC class I molecules, undifferentiated ES cells and differentiated cells from EB are readily recognized by LCMV-specific CTL either after loading with exogenous synthetic viral epitopes or after productive infection with LCMV. However, neither ES nor EB cells are effectively lysed by otherwise highly cytotoxic virus-specific CD8⁺ T lymphocytes. The molecular mechanism of resistance to CTL-mediated cytotoxicity is shown to be conferred by high-level expression of serine protease inhibitor 6 (SPI-6) in ES and EB-derived cells. Knockdown of SPI-6 expression by RNA interference sensitized ES cells for CD8⁺ T cell-mediated cytotoxicity, suggesting that SPI-6 is crucial as an ES cell-autonomous protection mechanism against cellular immune responses.

Murine CGR8 ES cells, previously established from strain 129P2/Ola mouse embryos (16), were maintained on gelatin-coated tissue culture plates in Glasgow MEM supplemented with 10% FBS (v/v), 2 mM l-glutamine, 50 μM 2-ME (all from Invitrogen Life Technologies), and 100 U/ml LIF (ESGRO; Chemicon International). ES cell differentiation was initiated (day 0) by the hanging drop method with 500 ES cells/20-μl drop in IMEM plus GlutaMAX supplemented with 20% FBS, 100 μM 2-ME, and 1× nonessential amino acids. Murine transgenic ES cell line αPIG (17), derived from D3 ES cells established from blastocysts of a 129S2/SvPas mouse (18), was cultivated on mitomycin C-inactivated mouse embryonic fibroblast feeder cells in DMEM supplemented with 15% FBS (v/v), 2 mM l-glutamine, 50 μM 2-ME, and 1000 U/ml LIF. Their differentiation was performed by mass culture method starting on day 0 with 1 × 10⁶ ES cells in 14 ml/plate of the differentiation medium as detailed for CGR8 cells. The developmental stages of the cells were controlled by assessing expression of the pluripotency marker Oct4, the stem cell marker SSEA-1, the cardio-specific marker α-myosin H chain, and the early endodermal marker α-fetoprotein, as well as by demonstrating the appearance of beating cardiomyocytes in EB from day 8 of differentiation onward. Cells were defined to be undifferentiated when they were harvested directly from their respective maintenance culture conditions in medium containing LIF and found to express the pluripotency markers Oct4 and SSEA-1 and not differentiation-specific markers α-fetoprotein and α-myosin H chain. Differentiation stages of EB cells were defined by the culture time in differentiation medium and expression of the above-listed markers. To exclude artifacts that might result from specific properties of a given cell line, all experiments were performed with the independently generated CGR8 and αPIG ES cells. C57BL/6-SV (B6SV) or BALB/c-SV fibroblasts were used as syngeneic or allogeneic target cells, respectively.

LCMV, strain WE, was generated and titrated on L929 cells as PFU (19). LCMV was detected by immunofluorescence microscopy in cryosectioned and paraformaldehyde-fixed cells by using a LCMV-specific mAb (clone M-104; PROGEN Biotechnik).

C57BL/6 and BALB/c mice were kept under specific pathogen-free conditions. Animal experiments were approved by the ethics committee of the Government of Cologne and performed in accordance to German animal protection law.

If not otherwise indicated, all experiments were repeated at least three times and one representative experiment is shown. If applicable, group sizes are indicated in the figure legends.

Virus-specific CD8⁺ T lymphocytes were induced by i.v. inoculation of C57BL/6 or BALB/c mice with 10⁵ IU of the LCMV strain WE. On day 8 after infection, CD8⁺ T cells were immunomagnetically enriched from splenic single-cell suspensions. LCMV-specific cytotoxicity of primary CD8⁺ CTL was assessed in a 4-h standard chromium release assay. Target cells were infected 48 h before the assay with LCMV at a multiplicity of infection of 0.01 or loaded for 1 h with the synthetic peptide representing the immunodominant H-2D^b-restricted epitope (gp_33–41) of the LCMV glycoprotein 1 at the indicated concentrations.

OVA-specific hybridoma B3Z (20) carrying the lacZ reporter gene under control of the IL-2 promoter/enhancer were cocultured with peptide-loaded stimulator cells overnight. Activation of T cell hybridomas that were stimulated via TCR were visualized in situ by using a β-galactosidase (β-Gal) staining kit (Invitrogen Life Technologies) according to the manufacturer’s instructions. The B3Z hybridoma recognizes the H-2K^b-restricted epitope SIINFEKL. The peptide KAVYNFATC representing the H-2D^b-restricted epitope gp_33–41 of LCMV was used as a control for peptide specificity. Stimulator cells were loaded with peptides (Sigma-Aldrich) at a concentration of 10⁻⁶ M for 1 h and subsequently washed three times before addition of B3Z cells.

Single-cell suspensions of ES and EB cells were prepared by trypsin/EDTA digestion. Expression of MHC class I surface molecules was determined by flow cytometry using H-2K^b-specific (clone AF6-88.5; BD Pharmingen) or isotype control Ab, each conjugated to PE. Degranulation of CTL was assessed by exposure of CD107a/Lamp1 (clone 1D4B; BD Pharmingen) on the plasma membrane (21). Data were collected with a FACScan and analyzed by CellQuest Pro software. IFN-γ secreted by LCMV-specific CD8⁺ CTL was quantified by a specific ELISA according to instructions of the manufacturer (R&D Systems).

ES cells (10⁶) were left untreated, or irradiated with UV (20 mJ/cm², 253–255 nm), or incubated with 0.5 μM staurosporine (Sigma-Aldrich). Caspase-3 activation was detected by immunofluorescence using rabbit mAb against activated caspase-3, clone 5A1 (Cell Signaling Technology). Enzymatic caspase-3 activity was quantified in the cytosolic extracts by fluorometry using DEVD-AFC (22).

Whole-cell extracts were prepared for immunoblotting by sonication of cells in ice-cold homogenization buffer (20 mM Tris-HCl (pH 7.5), 1% Triton X-100, 0.5 mM EDTA, 1 mM DTT, 1 mM PMSF, 10 μg/ml aprotinin, 5 μg/ml pepstatin A, and 25 μg/ml leupeptin) followed by incubation for 30 min on ice. Homogenates were spun at 15,000 × g for 10 min at 4°C, and the resulting supernatants were used for further analysis. Protein concentration was determined with Roti-Quant reagent (Carl Roth). Ten micrograms of each sample was separated on SDS-polyacrylamide gel and immunoblotting was performed as described previously (23) using anti-H-2K^b H chain serum (anti-p8, provided by J. Nefjees, Netherlands Cancer Institute, Amsterdam, The Netherlands) or commercial Abs for Oct4, β₂-microglobulin (Santa Cruz Biotechnology), cathepsin B (R&D Systems), β-actin (Sigma-Aldrich), PI-9/SPI-6 (MBL Woburn, JM-3544-100; this mAb is specific for human PI-9 and cross-reacts with murine SPI-6 (24)), and GAPDH (ab9485; Abcam). Blots were developed with ECL using the AP substrate CDP-Star (Applied Biosystems).

Expression of SPI-6 in ES and EB cells was determined by using TRIzol reagent for the extraction of RNA from 1 × 10⁶ cells, the Reverse-IT RTase Blend Kit (ABgene) for synthesis of cDNA, and a real-time PCR master mix (Applied Biosystems) for real-time PCR according to the manufacturer’s specifications. The values obtained were normalized to that of the gene encoding 18S rRNA before the fold change was calculated using crossing point values. Primers were as follows: mouse SPI-6: 5′-ctctgcatcatgaatact-3′, and reverse, 5′-ccttaaaggtttggagga-3′. The 18S rRNA was quantified with a 18S rRNA control kit (Applied Biosystems).

For gene expression analyses in ES cells and EB at different stages of differentiation, total RNA was isolated by a RNAqueous kit according to the manufacturer’s instructions (Ambion) and 1 μg was reverse-transcribed using Superscript II RTase (Invitrogen Life Technologies). Semiquantitative PCR was performed using a RedTaq DNA Polymerase kit (Sigma-Aldrich). PCR were stopped at the exponential phase of amplification, which was determined experimentally for each primer pair, and products were analyzed by agarose gel electrophoresis.

To silence SPI-6 expression, double-stranded small hairpin RNA (shRNA) for SPI-6 and control shRNA (scrambled) were obtained from Operon Technologies (ID1002175914 and 1002175915). Corresponding DNA oligonucleotides were cloned into pENTR (Invitrogen Life Technologies) containing the polymerase III-H1-RNA promoter as described in the pSUPER RNAi system manual (OligoEngine). CGR8 cells were transfected with 20 μg of pENTR/shSPI-6 and 2 μg of pPGK-puro (Invitrogen Life Technologies) in serum-free RPMI 1640 using the Gene Pulser Xcell nucleofection device (Bio-Rad) applying the following parameters: 950 μF, 200 Ω, 250 V, and 4-mm electroporation cuvettes. After transfection, CGR8 cells were plated in 10 ml of culture medium and were allowed to recover for 24 h. Subsequently, the cells were selected for 72 h with 3 μg/ml puromycin before being used as target cells in cytotoxicity assays.

Murine ES cells and cells of EB at early differentiation stages (days 5 and 8 postinduction of differentiation) are readily infected by the LCMV and produce high virus titers as illustrated by immunofluorescence analysis using a LCMV-specific mAb and by plaque assays (Fig. 1), respectively.

However, as shown in Fig. 2, neither two types of LCMV-infected murine ES cells (αPIG, derived from D3 cells, and CGR8 cells), nor LCMV-infected EB-derived cells at various stages of differentiation were lysed by virus-specific CD8⁺ CTL despite productive infection with LCMV. In contrast, syngeneic control fibroblasts were efficiently lysed by the same effector CTL population.

Analysis of MHC expression revealed that undifferentiated murine ES cells do express mRNA coding for MHC class I H chain molecules and β₂-microglobulin, which was also observed with EB-derived cells at early and late stages of differentiation (Fig. 3,A). In Western blot analysis, H-2K^b molecules were not detectable in undifferentiated ES cells (Fig. 3,B). However, EB cells from day 2 of differentiation onward were positive for H2-K^b molecules, and expression was strongly enhanced by pretreatment with IFN-γ (Fig. 3,B). ES cell differentiation was proven by the disappearance of the stem cell marker Oct4 and expression of the tissue-specific markers α-fetoprotein (liver, endodermal origin) and α-myosin H chain (heart, mesodermal origin) (Fig. 3, A and B; the strong signal for Oct4 in untreated day 16 EB (Fig. 3,B) is probably due to a contamination, because it was not observed in several other experiments). Differentiation was also confirmed by assessing the levels of SSEA-1 by flow cytometry, which progressively decreased with time, reaching background levels between days 5 and 7 postinduction of differentiation (data not shown). To confirm MHC class I expression by an independent method, flow cytometry analyses were performed. Undifferentiated mouse ES cells express only low levels of MHC class I molecules on their cell surface (Fig. 3,C). MHC class I molecule expression remained low in EB-derived cells up to day 20 postinitiation of differentiation (Fig. 3,C). IFN-γ did not enhance the surface expression of MHC class I molecules by undifferentiated ES cells, while it strongly induced MHC class I molecules in EB cells from day 4 of differentiation onward (Fig. 3,C). Infection with the LCMV did neither in control fibroblasts nor in ES or EB cells change expression levels of H-2K^b (Fig. 3 D).

Productive infection with the LCMV and cell surface MHC class I expression do not ensure that viral epitopes are properly processed and presented by ES and EB cells from endogenously synthesized viral Ags. To bypass possible defects in Ag processing and presentation in ES and EB cells, MHC class I molecules were externally loaded with the gp_33–41 epitope. Preincubation of syngeneic C57BL/6-SV fibroblasts with the synthetic peptide resulted in efficient killing by LCMV-specific CD8⁺ CTL, while undifferentiated ES cells loaded with gp_33–41 were lysed just slightly above background levels, and differentiated cells of EB loaded with gp_33–41 were not significantly lysed by the same effector cells (Fig. 4 A).

To test whether lack of killing of ES and EB-derived cells by CTL was secondary to low-level MHC I expression, ES and EB-derived cells were pretreated with IFN-γ for 48 h and subsequently loaded with the LCMV gp_33–41 peptide. However, IFN-γ-induced up-regulation of MHC I resulted only in day 8 EB-derived cells in clearly increased lysis (Fig. 4 B), which yet remained more than about 4-fold reduced compared with target control fibroblasts as revealed by horizontal comparison of E:T ratios required to achieve equivalent lysis.

It should be noted that other proapoptotic stimuli like UV irradiation or staurosporine readily induced caspase-3 activation and apoptotic cell death (Fig. 4 C), indicating that ES cells are not generally defective in cell death signaling. Thus, ES cells and their early differentiated derivatives appeared to be selectively resistant to CTL-mediated cytolytic effector functions.

Despite low-level expression of MHC class I molecules, ES and EB-derived cells can be recognized by CTL in an Ag-specific and MHC class I-restricted manner. This is demonstrated with a T cell hybridoma, B3Z, that specifically recognizes the OVA-derived epitope SIINFEKL (25), presented by MHC class I molecules H-2K^b (Fig. 5). Undifferentiated ES cells, as well as day 8 or day 14 EB cells loaded with SIINFEKL, were readily recognized by B3Z hybridoma as indicated by activation of the β-Gal reporter system. In contrast, neither allogeneic BALB/c-derived fibroblasts expressing H-2^d molecules incubated with SIINFEKL nor ES or EB-derived cells loaded with a control peptide KAVYNFATC representing the H-2D^b-restricted immunodominant epitope gp_33–41 of the LCMV (26) were recognized by the SIINFEKL-specific B3Z hybridoma, indicating MHC-restricted recognition of ES and EB-derived cells by SIINFEKL-specific CTL.

The markedly impaired lysis of peptide-loaded ES and EB cells, even after IFN-γ-induced high-level expression of MHC class I molecules, raised the question about the efficiency of E:T cell interaction. CTL exert cytotoxic effector function by means of the pore-forming perforin and of apoptosis-inducing granzymes (grz) that are secreted at the immunologic synapse formed between the CTL and the target cell. As shown in Fig. 6 A, microscopic analysis revealed that LCMV-immune CTL physically interact with virus-infected ES cells. Furthermore, staining with a grzA-specific antiserum revealed proper direction/polarization of cytolytic granules to the immunologic synapse.

To test for secretion of the granules containing grz and perforin, a Lamp1 (CD107a) degranulation assay was performed (21). CD107a/Lamp1 is an integral protein of the inner leaflet of cytotoxic granules, the exposure of which at the plasma membrane of CTL is an indicator for degranulation and correlates with cytotoxic activity. The emerging cell surface exposure of Lamp1 on CD8⁺ T cells coincubated with Ag-presenting ES or EB-derived cells (Fig. 6 B) indicated an intact fusion of cytolytic granules with the plasma membrane, a prerequisite for the release of cytotoxic effector molecules.

As shown in Fig. 6,C, ES cells as well as cells of EB infected with the LCMV or externally loaded with gp_33–41 induced syngeneic, virus-specific CD8⁺ CTL cells to secrete large amounts of IFN-γ, indicating that ES cells and cells of EB effectively present viral epitopes. Ag specificity of the target cell-induced secretion of IFN-γ is indicated by the low background levels of IFN-γ detected in cocultures of CD8⁺ CTL with noninfected target cells. MHC class I restriction is revealed by the nonresponsiveness of LCMV-specific allogeneic, H-2^d-restricted, CD8⁺ CTL derived from BALB/c mice. In these assays, peptide not bound by target cells was thoroughly washed away before coincubation with effector cells. In comparison to ES cells or EB-derived cells, CD8⁺ target cells required an ∼100-fold load of gp_33–41 to induce secretion of comparable amounts of IFN-γ (Fig. 6 D). Thus, the remote possibility that CD8⁺ effector CTL pick up peptide from the target cells and subsequently function as APCs themselves could be largely ruled out.

The amount of IFN-γ secreted by CTL in response to ES and EB cells loaded with gp_33–41 or infected with the LCMV ranged between 65 and 90% of that induced by LCMV-infected or gp_33–41-loaded control fibroblasts (Fig. 6 C), indicating proper recognition by, and activation of, CTL. Intact formation of the immunologic synapse between Ag-loaded ES cells and virus-specific CD8⁺ T cells, proper polarization and degranulation of cytolytic granules, and proper secretion of IFN-γ by CD8⁺ T cells all together indicated that undifferentiated ES and differentiated ES-derived target cells are readily recognized by and trigger effector mechanisms in virus-specific CTL, leaving the possibility that ES and EB-derived target cells are resistant to cytotoxic effector molecules.

To unravel the molecular mechanism protecting ES and EB-derived cells from CTL-mediated lysis, we analyzed these cells for expression of cathepsin B and serpins, which have been shown to protect several differentiated cell types (e.g., CTL, dendritic cells, and various tumor cells) against the activity of perforin and grz. Like activated CTL, ES cells and days 5 and 8 EB-derived cells were found to express cathepsin B at high levels (Fig. 7,A). The specific protease inhibitor l-trans-epoxysuccinyl-lle-pro-OH propylamide (CA074) used as a pharmacologic inhibitor of cathepsin B (27) enhanced suicidal/fratricidal death of CD8⁺ CTL significantly (Fig. 7,B). However, both ES cells (Fig. 7,C) or EB-derived cells (Fig. 7 D) incubated with CA074 were as resistant to CTL-induced lysis as untreated controls.

In addition to cathepsin B, ES and EB-derived cells abundantly express SPI-6 comparable to that observed in activated CTL, dendritic cells, and placenta as revealed by RT-PCR analysis and immunoblot (Fig. 8,A). To test whether the robust resistance to CTL-mediated killing is caused by SPI-6 expression, SPI-6 was effectively down-regulated in ES cells by transfection with SPI-6 small interfering RNA (siRNA) as shown by PCR and immunoblot (Fig. 8,B). As shown in Fig. 8 C, SPI-6 siRNA markedly sensitized ES cells for CTL-mediated killing, while RNA interference with control (scrambled) siRNA had no effect. These results show that ES cells protect themselves by expression of SPI-6 against lysis by CD8⁺ CTL.

In previous studies about the immunogenicity of ES cells, murine ES cells were found to be devoid of MHC class I expression at any passage number (28, 29). Furthermore, human ES cells lack cell surface expression of MHC class II and costimulatory molecules (6, 7, 30) and, correspondingly, neither undifferentiated nor differentiated human ES cells elicited a proliferative or alloreactive human T cell response in vitro (31) or in vivo (7). In the present study, we report expression of mRNA coding for MHC class I molecules in undifferentiated murine ES cells and in EB-derived cells. Although expression levels of MHC class I molecules on the plasma membrane of ES cells were below the level of detection by flow cytometry, both LCMV-infected or epitope-loaded ES cells were readily recognized by virus-specific CD8⁺ CTL as determined by secretion of IFN-γ by CTL coincubated with the ES cells. Moreover, the B3Z hybridoma specifically recognized ES cells and EB-derived cells presenting the OVA-derived epitope SIINFEKL. ES cell recognition by CD8⁺ CTL was Ag specific as well as MHC class I restricted. Low-level MHC class I expression does not contradict Ag-specific and MHC class I-restricted recognition by CTL, because it is well documented that T cells can be triggered by <100 MHC-peptide complexes on a target cell (32, 33, 34).

Despite being recognized by Ag-specific CTL, both ES cells and EB-derived cells were not significantly lysed by highly effective CTL. One could envision that the contact between CTL and ES cells might be too weak to trigger in CTL the secretion of cytotoxic granules at the immunologic synapse. However, the following lines of evidence suggest that Ag-presenting ES and EB-derived cells provide sufficient signals to stimulate secretion of cytotoxic granules by Ag-specific CTL. We observed that virus-specific CTL contact Ag-presenting ES cells and polarize their cytotoxic granules previously scattered in the cytoplasm within 10 min toward the contact zone (i.e., toward the immunologic synapse). Moreover, virus-specific CD8⁺ T cells coincubated with either Ag-expressing ES cells or control fibroblasts expose CD107a/Lamp1 with comparable kinetics and densities on their plasma membrane. In light of proper recognition, the failure of CTL to lyse ES and EB-derived cells was concluded not to be secondary to low antigenicity of ES cells, but rather to be brought about by resistance against CTL-mediated killing.

Indeed, ES cells and EB-derived cells express high levels of cathepsin B and SPI-6, two potent inhibitors of CTL-derived cytotoxic effector molecules. Cathepsin B cleaves perforin, thereby impairing its pore-forming activity required for entry of grzB into the target cell. Although use of the specific inhibitor of cathepsin B, CA074, increased suicide/fratricide among CD8⁺ CTL, it did not enhance the susceptibility of ES cells or EB-derived cells to the cytotoxic mechanisms of CTL. This might be due to inherent problems of pharmacologic inhibitors, e.g., incomplete inhibition of cathepsin B by CA074. Alternatively, the protease cathepsin B might not be involved in protection of ES cells against cytolysis at all, but rather take part in tissue degradation during invasion of the uterine stroma by the embryonic trophoblast that abundantly expresses cathepsin B (35, 36). More work is needed to elucidate the biologic relevance of cathepsin B in embryonic development.

One of the key findings of our study was that SPI-6 is expressed in ES cells and EB-derived cells at levels almost as high as in activated CD8⁺ CTL. The human homolog of this specific antagonist of grzB, serpin PI-9, was previously shown to be expressed by dendritic cells, where it prevents premature destruction of professional APCs, and at immune-privileged sites, where degranulation of CTL is potentially deleterious (37, 38). In addition, murine SPI-6 was reported to play a crucial role in the protection of CTL against the suicidal effects of their own grzB (39). Aberrant expression of the human serpin PI-9 and its murine homolog SPI-6 has been detected in various types of melanoma, carcinomas, and lymphomas (24, 38, 40) and was realized to be linked with the escape of tumor cells from immunosurveillance. Strikingly, knockdown of mRNA coding for SPI-6 by RNA interference abolished protection of ES cells against CTL-mediated lysis, providing compelling evidence that it is SPI-6 that mediates the CTL-resistant phenotype of ES cells.

The detection of the cytoprotective molecules cathepsin B and SPI-6 in murine ES and EB cells and the demonstration that SPI-6 mediates efficacious protection against CTL-mediated lysis raises several questions. For example: 1) what is the impact of SPI-6 expression on the course and outcome of vertically transmittable viral diseases? 2) do SPI-6 and/or other cytoprotective molecules prevent rejection of the blastocyst by maternal lymphocytes during attachment of the trophoblast to the endometrium (for review, see Ref. 41)? and 3) does SPI-6-mediated CTL resistance pose previously unrecognized risks for stem cell-based therapy? Undifferentiated ES cells or EB cells of early differentiation stages will most probably not be used for therapeutic approaches. Notwithstanding, the physiologic expression or the pathophysiologic reactivation of cytoprotective molecules like SPI-6 in tissue-specific derivatives of ES cells might establish immunoprivileged properties, which provides previously unrecognized obstacles in regenerative medicine, such as persistent microbial infection of an ES cell-derived transplant or the escape of malignantly transformed cells from immunosurveillance.

We thank Melanie Trapp, Lena Wilkens, Madlen Strauß, Devi Mariappan, Naidu Kamisetti, Martina Bessler, and Lukas P. Frenzel for their technical assistance and Eva-Maria Menke for expert work at the animal facility. H-2K^b H chain polyclonal Ab anti-p8 was a gift from Jacques Neefjes (Netherlands Cancer Institute). B3Z hybridoma T cells were provided by Nilabh Shastri (University of California, Berkeley, CA) and Marcus Gröttrup (University of Konstanz, Germany).

The authors have no financial conflict of interest.