Multipotent adult germ-line stem cells, like other pluripotent stem cells, can be killed by cytotoxic T lymphocytes despite low expression of major histocompatibility complex class I molecules

Background Multipotent adult germ-line stem cells (maGSCs) represent a new pluripotent cell type that can be derived without genetic manipulation from spermatogonial stem cells (SSCs) present in adult testis. Similarly to induced pluripotent stem cells (iPSCs), they could provide a source of cellular grafts for new transplantation therapies of a broad variety of diseases. To test whether these stem cells can be rejected by the recipients, we have analyzed whether maGSCs and iPSCs can become targets for cytotoxic T lymphocytes (CTL) or whether they are protected, as previously proposed for embryonic stem cells (ESCs). Results We have observed that maGSCs can be maintained in prolonged culture with or without leukemia inhibitory factor and/or feeder cells and still retain the capacity to form teratomas in immunodeficient recipients. They were, however, rejected in immunocompetent allogeneic recipients, and the immune response controlled teratoma growth. We analyzed the susceptibility of three maGSC lines to CTL in comparison to ESCs, iPSCs, and F9 teratocarcinoma cells. Major histocompatibility complex (MHC) class I molecules were not detectable by flow cytometry on these stem cell lines, apart from low levels on one maGSC line (maGSC Stra8 SSC5). However, using a quantitative real time PCR analysis H2K and B2m transcripts were detected in all pluripotent stem cell lines. All pluripotent stem cell lines were killed in a peptide-dependent manner by activated CTLs derived from T cell receptor transgenic OT-I mice after pulsing of the targets with the SIINFEKL peptide. Conclusion Pluripotent stem cells, including maGSCs, ESCs, and iPSCs can become targets for CTLs, even if the expression level of MHC class I molecules is below the detection limit of flow cytometry. Thus they are not protected against CTL-mediated cytotoxicity. Therefore, pluripotent cells might be rejected after transplantation by this mechanism if specific antigens are presented and if specific activated CTLs are present. Our results show that the adaptive immune system has in principle the capacity to kill pluripotent and teratoma forming stem cells. This finding might help to develop new strategies to increase the safety of future transplantations of in vitro differentiated cells by exploiting a selective immune response against contaminating undifferentiated cells. Reviewers This article was reviewed by Bhagirath Singh, Etienne Joly and Lutz Walter.

and if specific activated CTLs are present. Our results show that the adaptive immune system has in principle the capacity to kill pluripotent and teratoma forming stem cells. This finding might help to develop new strategies to increase the safety of future transplantations of in vitro differentiated cells by exploiting a selective immune response against contaminating undifferentiated cells.
Reviewers: This article was reviewed by Bhagirath Singh, Etienne Joly and Lutz Walter.

Background
Pluripotent stem cells could provide the basis for restoring tissue functions by transplantation of cellular grafts in a broad variety of diseases. Embryonic stem cells (ESCs) are the best characterized pluripotent stem cells. They clearly have the capacity to self-renew and to give rise to any cell type of the body [1]. However, the use of human ESCs is restricted due to severe ethical concerns. Therefore, new pluripotent cell types that are derived from adult organisms hold great promises for regenerative medicine. Recently, it has been shown that pluripotent cells can be obtained from spermatogonial stem cells (SSCs) of neonatal [2] and even adult mice [3][4][5]. Although SSCs are known to be unipotent [6] they can undergo reprogramming during in vitro cultivation and give rise to multipotent adult germ-line stem cells (maGSCs) that can differentiate into various cell types in vitro, and form teratomas in vivo [3,7]. The maGSCs have significant similarities to ESCs [8,9] and their differentiation capacity into functional cardiomyocytes [10] and neuronal cells [11,12] has been shown. Therefore, maGSCs are expected to have a high potential for regenerative medicine [13]. The generation of pluripotent cells from adult human testis has been recently reported [14][15][16]. Since maGSCs can be obtained without the need of genetic manipulation, they might have advantages even compared to induced pluripotent stem cells (iPSCs) that were obtained by successful reprogramming of mouse and human somatic cells after expression of a limited number of defined transcription factors [17][18][19].
The possibility to generate autologous pluripotent stem cells from adults could also reduce immunological problems that are expected to occur when allogeneic ESCs are used for regenerative therapies [20,21]. However, in the case of maGSC-based therapies only male patients could theoretically benefit from fully autologous stem cells. The complexity of maGSC generation and requirements for quality assurance for individual stem cell lines might be a further hurdle for the therapeutic use of individualized maGSCs, as has been discussed in the context of iPSCs [22,23]. Thus, although maGSCs and iPSC could theoretically allow the generation of autologous pluripotent stem cells, they might also be of use for therapeutic purposes in partially allogeneic settings. However, even when autologous stem cells are utilized for transplantation therapies, cytotoxic T lymphocytes (CTLs) might recognize "oncofe-tal" antigens [24] expressed by pluripotent stem cells or differentiation antigens [25] as has been documented and exploited in the context of tumor immunology. A further concern is that syngeneic or autologous transplantations might be associated with a higher risk of teratoma growth than allogeneic transplantations if grafts are derived from pluripotent cells and are contaminated with undifferentiated cells [26]. Therefore, it seems important to determine whether new pluripotent stem cell types such as maGSCs and iPSCs are protected against CTLs, similarly to what has been reported for ESCs [27].
In this study, we have analyzed the teratoma growth of maGSCs depending on the conditions used for their culture and in vitro differentiation. Furthermore, the susceptibility of three maGSC lines (maGSC Stra8 SSC5, maGSC 129/Sv, maGSC C57BL) to CTLs has been determined in comparison to corresponding ESC lines (ESC Stra8, ESC 129/Sv R1, ESC C57BL), iPSCs, and F9 teratocarcinoma cells.

Generation of CTLs and 51 Chromium release assays
To obtain peptide-specific CTLs, spleen cells from naive OT-I mice [33] were stimulated in vitro in the presence of 1 nM SIINFEKL peptide (Ovalbumin 257-264; Bachem Biochemica, Heidelberg, Germany) as previously described [34]. Target cells were labeled by incubating 1 × 10 6 cells in 200 μl N-2-hydroxyethylpiperazine-N'-2ethanesulfonic acid (HEPES)-buffered DMEM containing 100 μl FCS and 50 μCi Na 2 51 CrO 4 (Hartmann Analytic, Braunschweig, Germany) for 1 h at 37°C and washed three times with HEPES-buffered DMEM. Effector cells were added to 5 × 10 3 51 Cr-labeled target cells in triplicate at various ratios in 200 μl HEPES-buffered DMEM/10% FCS per well of round-bottomed microtiter plates. Spontaneous release was determined by incubation of target cells in the absence of CTL. To allow for a peptide-specific killing the respective wells were supplemented with 0.5 μg/ml SIINFEKL. To determine calcium-dependency of killing 2 mM ethyleneglycol-bis(b-aminoethyl ester)-N,N,N',N'-tetraacetic acid (EGTA) and 4 mM MgCl 2 were added. The microtiter plates were centrifuged for 5 min at 40 × g, incubated at 37°C for 4 h, and then centrifuged again. Supernatant and sediment were separately taken to determine radioactivity in each well using a Wallac MicroBeta Trilux counter (PerkinElmer Life Sciences, Köln, Germany). Percentage of specific lysis was calculated by subtracting percent spontaneous 51 Cr release [35].

Flow cytometry
Flow cytometry was performed on a FACScan™ flow cytometer (BD Biosciences, Heidelberg, Germany) using CellQuest™ data acquisition and analysis software. The cell surface expression of major histocompatibility complex (MHC) class I molecules was analyzed using locus and haplotype-specific antibodies (Ab) anti-H2K b (clone CTKb, mouse IgG 2a , phycoerythrin (PE)-conjugated, Caltag Laboratories, Hamburg, Germany), and anti-H2D b (clone CTDb, mouse IgG 2a , PE-conjugated, Caltag). The isotype control (mouse IgG 2a , PE-conjugated) was purchased from Caltag. To test whether the MHC class I expression can be stabilized at a lower temperature, the stem cells were incubated at 28°C for 24 h before analysis. In order to test for the stabilization of H2K b cell surface expression by external peptides, the H2K b -restricted peptide SIINFEKL or a non-binding control peptide (NGLT-LKNDFSRLEG) were added for 24 h to cell cultures. The expression analysis of pluripotency markers was done as described previously [3] using mouse anti-SSEA-1 (clone MC480, Developmental Studies Hybridoma Bank, IA, USA) and mouse anti-Oct4 (clone 9E3.2, Chemicon) mAbs.

Gene expression analysis
Total RNA was extracted from cell lines using the TRIZOL ® reagent (Invitrogen, Carlsbad, USA) according to the manufacturer's instructions. The RNA was then treated with RNase free DNase (RQ1, Promega, Madison, USA) for 20 min at 37°C and purified by phenol-chloroformisoamyl alcohol (25:24:1) extraction and precipitated with 1/10 volume 300 mM sodium acetate (pH 4.8) and 1 volume 2-propanol before washing with 70% ethanol and solving in RNase free water. The quantity of the extracted RNA was determined with a ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, USA) and the quality was analyzed using a Bioanalyzer 2100 (Agilent Techonolgies, Santa Clara, CA, USA). For synthesis of cDNA random oligo primers (Promega, Madison, USA) were used. The reverse transcription of RNA was performed for 60 min at 37°C with M-MLV RT polymerase (Promega, Madison, USA) in a total volume of 25 μl. Gene expression levels were analyzed by quanti-tative real time polymerase chain reaction (qRT-PCR) assays using the following forward and reverse primers generated according to the indicated reference sequences: . The PCR reaction plates were preheated for 2 min at 50°C and for 10 min at 95°C followed by 40 cycles of denaturation (15 s at 95°C) and amplification (1 min at 60°C). All reactions were performed in technical triplicates using an ABI 7500 Real Time PCR System. For the data analysis, the ABI 7500 SDS software (Applied Biosystems) was used. The variations in cDNA concentration in different samples were normalized to the housekeeping gene Hprt1. The cycle threshold (ct) values obtained for the genes of interest were corrected by the ct value obtained for Hprt1 in the same sample. The relative level of transcripts was then expressed as Δct value (ct for Hprt1 minus ct for the gene of interest).

Animal experiments
All animals were bred in the central animal facility of the Medical Faculty of the University of Göttingen. Severe combined immunodeficient SCID/beige mice (C.B-17/ IcrHsd-scid-bg) were kept under pathogen-free conditions. All animal experiments had been approved by the local government. For the analysis of subcutaneous tumor growth undifferentiated maGSC and in vitro differentiated cells, respectively, were suspended in 100 μl PBS (2 × 10 6 cells/100 μl) and injected into the flank of the animals. Tumor growth was monitored every second day by palpation and size was recorded using linear calipers. The tumor volume was calculated by the formula V = πabc/2, where a, b, c are the orthogonal diameters. Animals were sacrificed at day 100, or before when a tumor volume of 1 cm 3 was reached, when a weight loss of more than 10% occurred, or when any behavioral signs of pain or suffering became evident. Autopsies were performed for all animals. Tumor tissue was immediately frozen in liquid nitrogen or placed in phosphate-buffered 4% formalin for 16 h and then embedded in paraffin. Tissue sections (2 μm) were stained with hematoxylin and eosin (HE) for histological examinations.

Statistics
The data were analyzed using the non-parametric Kruskal Wallis test and the U test for subgroup analysis with the WinSTAT software. A significance level of α = 0.05 was used.

Maintenance of maGSC Stra8 SSC5 cells under different culture conditions
MaGSC lines can be grown in vitro under different cell culture conditions [3,7], of which we used four in our study. Condition I consists of basic medium, condition II is the basic medium plus LIF, condition III is the basic medium plus a feeder layer of MEFs, and condition IV is the basic medium plus LIF plus MEFs. The condition IV represents the gold standard for ESC culture. The maGSC line SSC5 expressed the pluripotency markers stage-specific embryonic antigen 1 (SSEA-1) and Oct3/4 under all culture conditions similarly to the ESC line MPI-II ( Figure 1A). This maGSC line was generated from the Stra8-EGFP/Rosa26 transgenic mouse, which expresses the enhanced green fluorescent protein (EGFP) under control of the Stra8 promoter [3]. In male mice the Stra8 promoter is selectively active in spermatogonia and their immediate descendants (preleptotene spermatocytes) [36][37][38][39] before it becomes silenced upon further differentiation. Therefore, the undifferentiated maGSC Stra8 SSC5 cells are characterized by EGFP expression [3]. We determined the proportion of EGFP expressing cells under the four culture conditions and after in vitro differentiation for 5 plus 10 days by flow cytometry as exemplified in Figure 1B. Under all four culture conditions more than 75% of the maGSC Stra8 SSC5 cells expressed EGFP ( Figure 1C) and the conditions did not differ significantly (p = 0.1753, Kruskal-Wallis test). Thus, maGSCs retain expression of pluripotency markers even under culture conditions without either LIF or feeder cells, at least for more than 16 passages. After spontaneous in vitro differentiation using the hanging drop method, on average less than 20% of the cells remained positive for EGFP at day 10 after plating EBs at day 5 ( Figure 1C).

Teratoma formation of maGSC Stra8 SSC5 cells cultured under different culture conditions
We have previously shown that maGSC Stra8 SSC5 cells cultured in basic medium on a feeder layer (condition III) have the potential to form teratomas in immunodeficient SCID/beige mice [3]. To determine whether the various cell culture conditions have an effect on the pluripotency of maGSCs, we compared the tumor growth of maGSC Stra8 SSC5 cells cultured under the conditions I to IV after subcutaneous injection into SCID/beige mice. The passage numbers of maGSC Stra8 SCC5 cells used for these experiments was between 12 and 22. Tumors were found in high frequency within 100 days after injection of maGSC Stra8 SSC5 cells cultured under all conditions (Table 1). In most cases, progressively growing tumors occurred within the first 50 days ( Figure 2). However, some tumors remained small and stable for the whole observation time ( Figure 2); others even regressed before the end of the experiment ( Figure 2, Table 1). Usually, the tumors were found only subcutaneously at the site of injection. In a few cases, autopsy revealed that tumors had grown invasively and given rise to tumors in the peritoneal cavity ( Table 1). The histopathological evaluation of tumors indicated that teratomas had formed from the maGSC Stra8 SSC5 cells cultured under four different conditions. This is exemplified in Figures 3A-C representing a teratoma derived from maGSC Stra8 SSC5 cells cultured under condition I. The teratomas contained derivatives of all three embryonic germ layers, including epithelium with intestinal differentiation (endoderm, Figure 3A), striated muscle ( Figure 3B), smooth muscle, fat, bone and cartilage (mesoderm), and neural tissue (ectoderm, Figure  3C). Thus, the cells cultured under all four conditions were undifferentiated, as indicated by EGFP expression and expression of pluripotency markers SSEA-1 and Oct3/ 4 ( Figure 1), and also pluripotent as shown by the capability to form teratomas.

A B
In addition to the undifferentiated cells, in vitro pre-differentiated cells (5 + 10 days of differentiation culture) were injected into 12 SCID/beige mice. In 6 cases an early progressive tumor growth was observed, 2 tumors remained small and stable, and 3 regressed before day 100 ( Figure 2, Table 1). Thus, the in vitro pre-differentiation protocol used here is not sufficient to deplete tumorigenic cells from the cultures. This finding is in agreement with the observation that a significant proportion of the in vitro differentiated cells still expressed EGFP, indicating the presence of undifferentiated cells ( Figure 1B, C). The histopathological analysis of these tumors showed that the in vitro differentiated cells, similar to undifferentiated maGSCs, formed teratomas ( Figure 3D-F). The teratomas contained derivatives of three embryonic germ layers, including epithelium with intestinal differentiation (endoderm, Figure 3D), striated muscle (mesoderm; Figure 3E), smooth muscle, fat, bone and cartilage (mesoderm), and neural tissue (ectoderm, Figure 3F).
The maGSC Stra8 SSC5 cells cultured under condition III (on feeder cells) were also injected subcutaneously into immunocompetent C57BL/6 mice. None of 9 animals receiving 2 × 10 6 cells developed a teratoma within 100 days. At autopsy no residues of the injected cells were found (data not shown). Thus, maGSCs were readily rejected in immunocompetent allogeneic hosts.

Expression of MHC class I molecules on maGSC Stra8 SSC5 cells cultured under various conditions
Mechanisms of the immune system play an important role for teratoma growth in immunocompetent animals.
To analyze the immunogenicity of maGSC Stra8 SSC5 cells, we determined the expression of MHC class I molecules and their susceptibility to CTL-mediated killing. A proportion of the maGSC Stra8 SSC5 cells expressed MHC class I molecules ( Figure 4A, B). Between 35 and 60% of the cells cultured under the four different conditions expressed H2K b molecules and less than 20% of the cells expressed H2D b molecules as detected by flow cytometry. The proportion of H2K b expressing cells differed between the cells cultured under the four conditions (p = 0.0127, Kruskal-Wallis test) and was higher in cultures without LIF (conditions I and III) compared to the respective condition with LIF suggesting an effect of the culture conditions on the immunologic phenotype of the maGSC Stra8 SSC5 cells. The results were similar (p = 0.0126, Kruskal-Wallis test) when the analysis was restricted to the undifferentiated Stra8-EGFP expressing cells ( Figure 4C). The expression levels as determined by the mean fluorescence intensity (MFI) on maGSC Stra8 SSC5 cells was rather low (on average <50) and much lower than on RMA cells (on average >350) that were used as positive control for the staining ( Figure 4A). However, the MFI for H2K b also differed between the conditions (p = 0.0151, Kruskal-Wallis test) and was higher in cultures without LIF. The in vitro differentiation of SSC5 cells reduced the proportion of H2K b expressing cells ( Figure 4A).

MaGSC Stra8 SSC5 cells cultured under different conditions are killed by CTLs whereas in vitro differentiated cells acquire resistance
Next we tested whether the maGSC Stra8 SSC5 cells can be killed by CTLs. We used CTLs from TCR-transgenic OT-I mice as effector cells, which recognize the ovalbuminderived peptide SIINFEKL in an H2K b -restricted manner [33]. The target cells, which do not express ovalbumin, were pulsed with the peptide to allow for binding to H2K b molecules and peptide-dependent killing. RMA cells served as positive control. The results of a representative individual experiment are shown in Figure 4D. RMA cells as well as maGSC Stra8 SSC5 cells were not killed when MaGSC Stra8 SSC5 cells cultured under various conditions (I to IV) and cells differentiated in vitro for 5 plus 10 days (diff.) were injected subcutaneously into the flank of SCID/beige mice (2 × 10 6 cells in PBS/animal). The percentage and number of animals in which tumors were found during autopsy or in which tumors were palpable (at least during 3 consecutive observations) at the side of injection before day 100 after injection is indicated. In addition, the percentages and numbers of animals are given in which tumors were present but regressed before day 100 and in which an invasive growth was observed. The differentiated cells (diff.) were obtained as described in the Materials and Methods section from maGSC Stra8 SSC5 cells that had been cultured under condition III.
they were not pulsed with the peptide. They were readily killed in an effector dose-dependent manner when the SIINFEKL peptide was present. In addition, the killing could be blocked by EGTA indicating that the lysis of the targets was mediated by the calcium-dependent granule exocytosis pathway. A summary of the results of 3 experiments with maGSC Stra8 SSC5 target cells cultured under the four conditions and differentiated cells, respectively, is shown in Figure 4E. The lysis of the maGSC Stra8 SSC5 targets did not directly correlate with the H2K b expression levels since the maGSC Stra8 SSC5 cells cultured under condition III showed the highest expression levels of H2K b but the lowest killing in comparison to the other three culture conditions. Interestingly, the in vitro differ-entiated cells (5 + 10 days of differentiation culture) were hardly killed by CTLs.

Further pluripotent cell lines are killed by CTLs despite being negative for MHC class I expression in flow cytometry
We then analyzed the killing of two further maGSC lines of the haplotype H2 b (maGSC 129/Sv and maGSC C57BL) by CTL and compared them to ESC targets (ESC 129/Sv R1, and ESC C57BL). These cells were cultured in the presence of LIF on feeder cells (condition IV) and transferred two days before the experiment to gelatinecoated dishes (condition II). A summary of the results of 3 experiments is shown in Figure 5A. In parallel to the killing assays, the MHC class I expression on the target cells was determined by flow cytometry. Interestingly, the maGSC lines maGSC 129/Sv and maGSC C57BL, in contrast to the maGSC Stra8 SSC5 cell line, did not express MHC class I molecules at a level detectable by flow cytometry (Figure 5C, D). They were very similar to the MHC class I negative ESC lines 129/Sv R1 and C57BL ( Figure 5C, D). In addition, also the iPSCs and F9 teratocarcinoma cells were negative for MHC class I molecules in flow cytometry ( Figure 5C, D). The external peptide pulsing of targets was apparently so efficient that even maGSCs, ESCs, iPSC and F9 cells, on which no H2K b molecules could be detected by flow cytometry, were readily killed in a peptide-specific manner.

The expression of MHC class I peptides on maGSCs and ESCs cannot be rescued by decreased temperature or external peptides
The pluripotent stem cells could either completely fail to express MHC class I molecules or have very few MHC molecules at the cell surface due to a deficiency in peptide loading, as e.g. neuronal cells that express low levels of TAP molecules [40] or RMA-S cells that have a mutation

test). (D)
A representative experiment is shown in which the susceptibility of maGSC Stra8 SSC5 cells cultured under conditions II and III to CTLs was determined. RMA cells served as highly CTL susceptible controls. The target cells were pulsed with the SIINFEKL peptide P (0.5 μg/ml) and exposed to CTLs derived from TCR-transgenic OT-I mice. The mean of specific lysis plus SD at different effector:target ratios (10:1 to 0.16:1) in the presence or absence of the SIINFEKL peptide measured in a 51 Cr release assay is shown. To confirm granule exocytosis dependency of killing EGTA was added to the test. (E) The means of relative lysis ± SD of the various target cell lines by OT-I CTLs are shown as determined in three independent experiments. The mean percentage of specific lysis of RMA cells (serving as positive control) at the highest effector to target ratio (10:1) was adjusted to 100% in each test and the relative lysis of the various target cells at different effector to target ratios was calculated.  in the Tap2 gene [41,42]. We analyzed the expression of H2K, H2D, Tap1, Tap2, and β2-microglobulin (B2m) at the mRNA level by qRT-PCR ( Figure 6). The pluripotent stem cell lines expressed H2K and B2m mRNA although at lower levels than RMA and RMA-S cells. The H2D mRNA was less expressed in the stem cell lines. The expression of Tap1 and Tap2 was indeed low in most stem cell lines.
Only maGSC Stra8 SSC cells that expressed H2K b molecules at the cell surface ( Figure 4A, B) had similar Tap1 mRNA levels as RMA cells. F9 teratocarcinoma cells were the only pluripotent cell type that expressed all these mRNAs at rather high levels. Interestingly, iPSCs expressed much less of these mRNAs than MEFs that were used for their generation.
The mRNA expression data are in accordance with the hypothesis that the peptide loading might be impaired in the pluripotent stem cell lines. Thus, we tested the expression of MHC class I molecules at decreased temperature and in the presence of external peptides. Incubation at 28°C for 24 h stabilized unloaded MHC class I molecules on RMA-S cells ( Figures 7A, B), whereas a subsequent culture at normal temperature (2 h 37°C) decreased the cell surface expression again. A culture for 24 h at 37°C in the presence of the SIINFEKL peptide (0.5 μg/ml) did not increase the H2K b expression as compared to untreated cells or cells that were incubated with a peptide that cannot bind to H2K b (Figures 7A, B). The combination of culture at 28°C and presence of the SIINFEKL peptide further augmented the H2K b but not H2D b expression and the peptide-loaded H2K b molecules were found to be more stable when the cells were transferred afterwards for 2 h to 37°C ( Figures 7A, B). However, on the stem cell lines maGSC 129Sv and ESC 129Sv/MPI-II that were tested in parallel, MHC class I molecules were not similarly stabilized by these treatments. Thus, a low expression of TAP molecules is unlikely to be the main reason for the low MHC class I cell surface expression on these cell lines.

MaGSCs as well as in vitro differentiated cells do not express the serine protease inhibitor 6 that can protect target cells against CTLs
It has been reported that mouse ESCs express SPI-6 that confers protection against cellular cytotoxicity mediated by the granule exocytosis pathway [27,43]. We analyzed the expression of SPI-6 in maGSC Stra8 SSC5 cells cultured under the four conditions (I to IV) described above.
The protein was expressed in activated CTLs from OT-I mice in which it contributes to the self protection against granzymes [44]. However, we did not detect the SPI-6 protein in the maGSCs ( Figure 8A). In contrast to undifferentiated maGSCs, cells differentiated in vitro from maGSCs were rather resistant to lysis by CTLs (see Figure 4E). Therefore, we analyzed the expression of SPI-6 in cells differentiated from maGSC Stra8 SSC5 cells and included two ESC lines for comparison. The protein was neither detected in the in vitro differentiated cells nor in the ESC lines ( Figure 8B). Serpinb9 that encodes SPI-6 was also hardly expressed in these cell lines at the mRNA level (Figure 8C). Similarly, we did not detect SPI-6 expression in any of the other pluripotent cell lines analyzed in this study (data not shown). Thus, expression of SPI-6 is apparently not responsible for the relative resistance of the in vitro differentiated maGSCs cells to CTLs that we observed.

Discussion
One advantage of maGSCs over ESCs is that they could theoretically be a source of autologous pluripotent stem cells for individualized medicine. One criterion for assessing the pluripotency of stem cells is the capacity to give rise to teratomas in immunodeficient mice [45]. Therefore, the potential of maGSCs and their differentiation products to form teratomas has to be analyzed very carefully before any therapeutic application can be envisaged.
In syngeneic or immunocompromised recipients, the risk of teratoma growth is inherent to transplantation therapies that are based on pluripotent cells. In immunocompetent allogeneic mice, transplanted ESCs do not usually form teratomas [26], unless very high numbers of pluripotent cells are injected [46]. We have compared the capacity of the maGSC line Stra8 SSC5 to give rise to teratomas depending on the culture conditions used to propagate them, and after differentiation. Cells cultured under four different conditions that varied in the presence of LIF and MEFs as feeder cells did not differ with respect to teratoma growth. This result was consistent with the presence of a similar proportion of undifferentiated cells under the four culture conditions as demonstrated by expression of pluripotency markers SSEA-1 and Oct-3/4, as well as EGFP expression driven by the Stra8 promoter in undifferentiated cells. Thus, the pluripotency of maGSCs was preserved under a variety of culture conditions for over 16 passages including those conditions that do not require feeder cells and might, therefore, have advantages in view of future therapeutic applications. Furthermore, cells differentiated in vitro (5 + 10 days of differentiation culture) without lineage selection still contained about 20% undifferentiated cells and gave rise to teratomas at high frequency in immunodeficient recipients, suggesting that lineage selection is necessary before the differentiated cells can be used for therapeutic application [47,48].
The risk of teratoma growth after transplantation of pluripotent cells does not only depend on the pluripotency of the injected cells but also on their immunogenicity and their susceptibility to cytotoxic effector mechanisms of the immune system because maGSC Stra8 SSC5 cells did not form teratomas in allogeneic immunocompetent mice. Therefore, we have started to character-

Figure 5 Pluripotent cells can be killed by CTLs despite very low expression of MHC class I molecules. (A)
The means of relative lysis ± SD of maGSC and ESC lines by CTLs are shown as determined in three independent experiments. The target cells were pulsed with the SIINFEKL peptide (0.5 μg/ml) and exposed to CTLs derived from TCR-transgenic OT-I mice. For normalization between experiments, the mean percentage of specific lysis of RMA cells (serving as positive control) at the highest effector to target ratio (10:1) was adjusted to 100% in each test and the relative lysis of the various target cells at different effector to target ratios was calculated. (B) The means of relative lysis ± SD of iPSCs and F9 teratocarcinoma cells by CTLs are shown as determined in three independent experiments. The target cells were pulsed with the SIINFEKL peptide (0.5 μg/ml) and exposed to CTLs derived from TCR-transgenic OT-I mice. The relative lysis was calculated as described above.   ize the immunological features of maGSCs and analyzed their susceptibility to killing by CTLs. CTLs recognize peptides presented by MHC class I molecules, which are highly polymorphic in the population. Alloreactive CTLs recognize MHC class I molecules either by direct or indirect allorecognition [20,21]. The most important immunological barrier for transplantation of pluripotent stem cell-derived allografts is presumably represented by MHC class I molecules [20,21], but minor histocompatibility antigens have also been shown to contribute to the rejection of ESC-derived grafts by CTL [49]. Interestingly, MHC class I molecules are not detectable by flow cytometry on mouse ESCs [26,27,[50][51][52][53][54]. This is in contrast to human ESCs [55][56][57]. However, the expression of MHC class I molecules has been reported to increase during in vitro [55] or in vivo differentiation of mouse ESCs [54,58]. MHC class I molecules are inducible by interferon-γ in mouse ESCs [51,54] or in differentiated cells [27]. Thus, ESCs and their differentiation products might become targets for CTLs after transplantation [49,53,54,59].
Whilst the mouse maGSC line Stra8 SSC5 was found to express comparatively low levels of MHC class I molecules by flow cytometry, the other maGSC lines (maGSC 129/ Sv and maGSC C57BL), ESCs, and iPSCs that were used for comparison were negative for these molecules, resembling the phenotype known for mouse ESCs and also F9 teratocarcinoma cells [60]. At the mRNA level H2K and B2m transcripts were detected in all pluripotent stem cell lines, although less abundantly than in RMA cells or MEFs. A low expression of Tap1 and Tap2 mRNA in most pluripotent stem cell lines suggested that a failure of peptide loading could explain the lack of MHC class I molecules at the plasma membrane. However, a decrease of temperature to 28°C or incubation with an H2K brestricted peptide did not stabilize the MHC class I molecules on maGSC 129Sv or ECS 129Sv MPI-II cells. A previous report showed a marked increase of expression of H2D d molecules on mouse embryonic cell lines (including one ESC line) that were transfected with an H2D dexpression construct and incubated with H2D d -binding peptides [61]. Thus, the endogenous H2K and B2m transcription in the stem cell lines analyzed here might still be insufficient to produce enough MHC class I molecules that can be detected by flow cytometry after stabilization by external peptides and decreased temperature.
However, all pluripotent cell lines analyzed here were killed by CTLs derived from TCR-transgenic OT-I mice that recognize the peptide SIINFEKL in an H2K b -restricted manner [33]. This finding is in accordance with studies showing that extremely low numbers of ligands on target cells are sufficient for CTL recognition and killing [62][63][64].
In addition, it has been reported previously that the MHC class I expression on mouse ESCs is sufficient for recogni-tion by the SIINFEKL-specific T cell hybridoma B3Z [27,65]. We now show that the MHC class I expression level on ESCs, maGSCs, and iPSCs is functionally relevant since it is also sufficient for peptide-specific CTL killing, although the presence of the peptide did not markedly increase the H2K b molecules at the cell surface of maGSCs and ESCs. The iPSCs appeared to be less susceptible to CTLs than the ESC and maGSC lines. Whether this is characteristic for iPSCs in general or only for the iPSC line analyzed here has to be determined. Interestingly, in vitro differentiation of maGSC Stra8 SSC5 cells appeared to reduce the sensitivity to CTLs because the in vitro differentiated target cells became much less sensitive to killing compared to their undifferentiated counterparts. The experiments using CTLs from TCR-transgenic mice demonstrated clearly that pluripotent stem cell lines can be killed by this mechanism. However, it has to be further clarified whether the antigen processing machinery of pluripotent stem cells is capable of generating peptide loaded MHC class I molecules for recognition by activated CTLs. Furthermore, it remains to be determined whether alloreactive CTLs that recognize their targets directly can kill the pluripotent stem cells with similar efficacy.
The observation that killing of the maGSCs, ESCs, iPSCs, and F9 cells could be inhibited by EGTA strongly suggests that the granule exocytosis pathway which involves perforin and granzymes [66,67] was used for killing.
Recently, it has been reported that the mouse ESC line CGR8 is resistant to antigen-specific CTLs due to the expression of SPI-6 [27], which is an endogenous inhibitor of granzyme B [43]. We therefore looked for SPI-6 by western blot in the maGSC and ESC lines but could not detect expression of this protein in any of these cells, even after in vitro differentiation of maGSC Stra8 SSC5 cells.
Their relative resistance to CTL cytotoxicity must therefore be caused by another mechanism. Our results are in accordance with recent reports by us [26] and others [65] showing that several ESC lines can be readily killed by NK cells. Although the mechanisms of target cell recognitions are different, NK cells and CTLs share the granule exocytosis pathway of target cell killing [66,67]. Thus, individual ESC lines might differ with respect to SPI-6 expression and resistance to CTL and possibly NK cells.
Several studies have suggested that mouse or human ESCs are immune-privileged. They may suppress immune responses by contact dependent [51,57] and contact independent mechanisms [46] or induce apoptosis in T cells [51]. It has also been reported that ESCs are resistant to CTL and NK cell-mediated cytotoxicity [27,46,51]. Here we have demonstrated that maGSC, ESC, and iPSC lines can be killed by activated peptide-specific CTLs despite levels of MHC class I expression that were below the detection level of flow cytometry. Therefore, it has to be assumed that these pluripotent cells can become targets for CTLs after transplantation.

Conclusion
MaGSCs represent a new pluripotent cell type derived from adult mouse testis. They retain pluripotency under several cell culture conditions and form teratomas in immunodeficient mice. No expression of MHC class I molecules could be detected by flow cytometry in most maGSC lines and the iPSCs analyzed here, a phenotype shared with ESC lines and F9 teratocarcinoma cells. However, in immunocompetent recipients the immune response could suppress growth of the maGSCs lines as teratomas, and all analyzed pluripotent cells including maGSCs, ESCs, iPSCs, and F9 teratocarcinoma cells could be killed efficiently by activated CTLs. Thus, maGSCs, ESCs, and iPSCs can be recognized as targets for CTLs and they might be rejected by this mechanism after transplantation in immunocompetent recipients. However, the capacity of the immune system to kill pluripotent and teratoma forming stem cells could also be used to develop a new strategy to increase the safety of transplantation of in vitro differentiated cells. Indeed, it might be possible to selectively induce an adaptive immune response in recipients against antigens expressed selectively in pluripotent stem cells in order to kill any teratoma forming cells that could not be depleted before grafting. Further studies are required to validate or disprove this hypothesis.

Competing interests
The authors declare that they have no competing interests.     1) Based upon the data presented the authors conclude that these cells can become targets for CTL even if the expression level of MHC class I molecules is below the detection limit by using flow cytometry. This is not surprising because little as 100 molecules of MHC-peptide complex are sufficient for CTL killing of target cells. Therefore, pluripotent cells might be rejected after transplantation by this mechanism and provides the hope that transplantations of in vitro differentiated cells may be possible and the undifferentiated cells may be killed by their immunogenicity. However, immunologically this approach does not provide any particular benefit as both differentiated cells and undifferentiated cells would be killed in allogeneic conditions. In syngeneic situation it is unclear if there is an advantage to the approach. Authors have to demonstrate this in vivo.  6) should be confirmed by RT-PCR.

Authors
Authors' response: As suggested, we have analyzed the MHC class I (see new Figure 6) and SPI-6 expression (see Figure 8) by quantitative real time PCR.
3) Finally, it is very important that authors provide a wellbalanced analysis to the differences where their results differ from previously published information about the immune-privileged nature of mouse or human ESC as quote in the last paragraph of their Discussion section.
Authors' response: We have altered the discussion in order to clarify these issues.

4)
Overall the manuscript is well written and the question posed, methods used are well defined although not new. Additional data is needed to validate the study. The discussion and conclusions need to be refined as outlined above. The title and abstract are well constructed but conclusions need to be modified to identify the shortcomings of the approach and the usefulness of the findings.
Authors' response: We tried also to further clarify the advantages but also the limitations of our present approach in the discussion.

Comment to the revised manuscript
The authors have considerably revised and updated the manuscript based on my comments. Additional experiments have been done and added as new figures. Authors have discussed these results in the text and in my view, the revised manuscript provides a more evidence for the observations and conclusions.
Authors' response: We would like to thank Dr. Singh for his thoughtful review of our manuscript and the helpful suggestions for improvement.

Reviewers' report 2
Dr. Etienne Joly, Equipe de Neuro-Immuno-Génétique Moléculaire, IPBS, UMR CNRS 5089, Toulouse, France 1) Have you considered the possibility that all these pluripotent cells express very low levels of TAP, and have thus very few properly folded MHC molecules on their surface, but probably quite a few empty, peptide-receptive ones. On this subject, I refer you to Bikoff et al. [61] and to my own paper on neuronal cells [40].
Authors' response: We have analyzed now the expression of Tap1 and Tap2 at the mRNA level (see the new Figure 6) and found rather low levels in most of the pluripotent stem cell lines. H2K and B2m transcripts were more abundant suggesting that indeed a failure of peptide loading could be responsible for the low MHC class I cell surface expression levels. However, culture at 28°C or in the presence of an H2K b -binding peptide did not markedly increase the H2K b cell surface expression (see the new Figure 7). Thus, it is unlikely that a low TAP expression alone is the reason for the low MHC class I expression on maGSCs and ESCs.
2) Considering this previous point, and the fact that the effectors in graft rejection would be alloreactive CTL rather than peptide specific ones, I think that you may have found quite a different picture if you had used anti-H2 b alloreactive CTLs obtained either after an in vitro mixed lymphocyte reaction, or after priming non-H2 b mice with H2 b splenocytes. Indeed, whereas OT-I cells have very high affinity TCRs that can kill upon recognition of just a handful of peptide-MHC complexes, the TCRs of alloreactive CTLs would be much more likely to be of low affinity, and would thus not have been triggered by the low levels found at the surface of the various stem cells used in your study.
Authors' response: We agree that using alloreactive CTL might give different results. Therefore, it will be of interest to determine whether the pluripotent cells used in this study can induce an alloreactive immune response and whether they are good targets for alloreactive CTL. However, we think that this question is beyond the scope of the present manuscript. In the present study we wanted to clarify whether the pluripotent cells can in principle be killed by CTL since it has been proposed previously that ESCs are resistant to cytotoxicity mediated by CTL [27]. Therefore, we used a well-controlled and very potent system of peptide-specific CTL derived from T cell receptor-transgenic mice instead of less well-defined alloreactive CTL. Using this system we could clearly demonstrate that the targets were killed and that the killing was peptide-dependent. We clarified the advantages but also the possible limitations of our present experimental approach in the discussion.
3) The last issue is that of the presentation of the FACS data. I find that histograms showing percentages of positive cells or even means of fluorescence are really much less informative than the actual plots. If they all look similar, then you should at least provide one example for each case. And if they are not so similar (for example different shapes), then you should definitely show the plots, and not the percentage of positive cells.