Immunobiology of naïve and genetically modified HLA-class-I-knockdown human embryonic stem cells Free

Human embryonic stem cells (hESCs) are pluripotent with the capacity to differentiate into any cell line. It is therefore conceivable that complex tissues can be rebuilt by hESC-based therapy after tissue injury. In the case of myocardial damage, therapeutic strategies currently pursued include the transplantation of hESC-derived cardiac progenitor cells to form contractile intracardiac grafts (Laflamme et al., 2007; Xue et al., 2005) and the transplantation of functional in-vitro-generated engineered heart tissue from hESCs (Domian et al., 2009). Both approaches have already shown substantial improvements in cardiac function. The reestablishment of contractile cells contributing to the cardiac mechanics is aspired, and great progress has recently been achieved in differentiating hESCs into cardiomyocytes (Burridge et al., 2007; Passier et al., 2005) and in sorting highly purified hESC-derived cardiomyocyte populations (Anderson et al., 2007; Hattori et al., 2010; Mummery, 2010). However, it has to be emphasized that improvements of cardiac function in experimental settings have so far only been achieved by transplanting hESC-derived cardiomyocytes into either immunodeficient (van Laake et al., 2007) or heavily immunosuppressed recipients (Caspi et al., 2007). Former beliefs of an immune privilege of hESCs or hESC-derived tissues have been thoroughly disproven (Drukker et al., 2006; Swijnenburg et al., 2008a) and human leukocyte antigen [HLA; the major histocompatibility complex (MHC) in humans] disparity between the hESC-derived donor cells and the recipient's immune cells during transplantation inevitably provokes immune rejection (Chidgey et al., 2008; Lui et al., 2009).

Although autologous induced pluripotent stem (iPS) cells can be generated from adult somatic cells as a potential pluripotent cell source for patients, thereby avoiding immune barriers, the generation of individualized cellular grafts for every patient might be impractical. This is because of the complexity of stem cell generation, requirements for quality assurance, timely and costly manufacturing process, the possibility of genetic abnormalities in donor cells and the uncertain availability at the time of need (Daley and Scadden, 2008). iPS-derived cells or tissues will more probably, at least in part, be transplanted in an allogeneic setting. Therefore, efforts are currently being undertaken to establish both hESC (Lin et al., 2009) and iPS cell banks (Tamaoki et al., 2010) in order to achieve at least partial HLA matching and reduce the risk for rejection. However, immunosuppressive protocols might still be needed at various levels.

Our group aims at generating hypoimmunogeneic hESC lines that would conserve immune non-responsiveness and might therefore serve as cell sources for generating universally compatible ‘off-the-shelf’ cell grafts or tissues. We herein present our studies evaluating the contributions of cellular and humoral immune components to the killing of hESCs and demonstrate that HLA class I (HLA I)-knockdown hESCs (hESC^KD) induce a substantially reduced immune activation and show extended survival.

hESCs showed modest HLA I and β₂-microglobulin surface expression, whereas these antigens were dramatically reduced in 7-day hESC^KD (Fig. 1). Both hESCs and hESC^KD were negative for HLA II and the co-stimulatory molecules CD80, CD86 and CD40. They were also negative for SSEA-1, a marker for maturity, but positive for SSEA-4 and the embryonic stem cell markers Oct-4, Sox-2, SSEA-3 and TRA-1-60. HLA phenotyping revealed the presence of A2, A3, B35, B44, Cw4, Cw7, DR15, DR16, DQ5 and DQ6.

The mean HLA I expression on hESC^KD over time was followed by fluorescence-activated cell sorting (FACS) (Fig. 2a). After around 7 days, HLA I expression reached its trough at 1.1±0.1 (P<0.001 compared with hESCs), which corresponds to ~1% of the surface expression of naïve hESCs. Between days 7 and 42, the lowered HLA I expression remained stable, at between 1.1±0.1 and 2.1±0.4.

To reveal whether the lowered HLA I surface expression was due to retainment of intracellular HLA I, the intracellular HLA I accumulation was quantified (supplementary material Fig. S1). Intracellular HLA I was hardly detectable in naïve hESCs and was not increased in hESCs after small interfering RNA (siRNA) application. In hESCs treated with intrabody (IB), however, accumulated intracellular HLA I was found. When IB was used as an adjunct to siRNA in hESC^KD, retainment of HLA I was not observed.

An early decline in BLI signals over the first few days was observed in Balb/c and SCID-beige recipients of hESCs (Fig. 2b). This initial decrease in cell viability was attributed to the transplant procedure itself, whereas the subsequent course of the bioluminescence imaging (BLI) signal reflected the impact of immune rejection on the proliferating viable hESC fraction. All hESCs were rapidly rejected in Balb/c mice and BLI signals were lost beyond the fifth day. In immunodeficient SCID-beige mice, which lack both lymphocytic and natural killer (NK) cell killing, BLI signals steadily increased to regain their initial values by around day 21. On day 80, macroscopic tumor growth was observed at the site of injection (supplementary material Fig. S2a) with ~50 times increased BLI signals (supplementary material Fig. S2b). Upon histological evaluation, the local tumors that had formed were clearly identified as teratomas, containing tissues from all three germ layers (supplementary material Fig. S2c). The rejection of hESC^KD in Balb/c recipients was profoundly mitigated compared with that of naïve hESC and their survival was markedly prolonged (Fig. 2b). In six of the ten animals, the BLI signals slowly dropped into the background at around day 28, but in four animals the signals remained stable and lasted for more than 42 days. Therefore, long-term cell survival in a stringent xenogeneic setting was achieved with hESC^KD in 40% of the animals.

At 5 days after the transplantation, the BLI signals from hESCs in SCID-beige mice and hESC^KD in Balb/c mice were similar, whereas signals from hESCs in Balb/c mice were hardly detectable (Fig. 2c). Because SCID-beige mice lack lymphocytes and NK cells, Elispot assays cannot be performed. Balb/c Elispot assays performed on day 5 were strikingly different between the naïve hESC and hESC^KD groups (Fig. 2d). Interferon (IFN)-γ (P=0.002) and interleukin (IL)-4 spot frequencies (P<0.001) were significantly lower after hESC^KD transplantation. Within the regional lymph nodes, the regulatory T cell (Treg) fraction (CD25+ Foxp3+) of CD4+ cells significantly increased 5 days after transplantation of hESCs (P=0.002) or hESC^KD (P=0.010) compared with control mice. After 14 days, a substantial time after all hESC grafts had been rejected, the Treg population had decreased to basal values. However, all hESC^KD grafts still showed viable BLI signals and the Treg fraction remained elevated (P=0.033 among the 14-day groups). No differences were observed after 42 days (Fig. 2e).

Gastrocnemius muscles of Balb/c mice were harvested on day 5 after hESC or hESC^KD transplantation and the composition of the cellular infiltrates was analyzed by immunohistochemistry. hESC grafts had a mainly round cell morphology and were identified by their firefly luciferase (Fluc) positivity (Fig. 3a–c). The main cell populations were macrophages and CD3+ lymphocytes, and only scarce KLRA1+ NK cells were found (Fig. 3d). All inflammatory cell populations were significantly lower within hESC^KD grafts (Fig. 3e; P<0.001 each compared with hESC).

Untreated Balb/c mice (before transplantation) showed negligible amounts of naturally occurring anti-hESC xenoantibodies, but cell transplantation rapidly induced IgM donor-specific antibody (DSA) production (Fig. 4a). However, the percentage of IgM-antibody-loaded cells was found to be significantly higher after hESC transplantation compared with hESC^KD transplantation (P<0.001). Next, the specificity of these antibodies was determined in Luminex single-antigen assays for HLA I and II. Class I antibody production was only induced after hESC transplantation (P=0.021 compared with native animals), but not after transplantation of hypoantigeneic hESC^KD (Fig. 4b). There was no significant induction of class II antibodies in both transplant groups. hESC re-injection 1 week after the first hESC transplantation resulted in a massive boost of class I (P=0.007 compared with hESCs), but not class II, antibodies, verifying the specificity of the antibody response. In recipients of hESC, the mean Z-score of hESC-specific class I antibodies was significantly higher than that of non-hESC-specific antibodies (Fig. 4c; P=0.049) and all of the six hESC-specific HLA-A, -B and -C antibody Z-scores were above a threshold of three (Fig. 4d). In recipients of hESC^KD, no hESC-specific class I antibodies were significantly induced. All HLA-DR and -DQ antibody Z-scores remained far below the threshold after both hESC and hESC^KD transplantation.

After incubation with CD3+ CD56− lymphocytes, hESCs (P=0.001 and P=0.002), but not hESC^KD (P=1.0 and P=0.428), significantly increased the spot frequencies for IFN-γ (Fig. 5a) and IL-4 (Fig. 5b), respectively, compared with resting responder lymphocytes. hESC^KD were therefore incapable of inducing relevant in vitro lymphocyte activation. Elispots were then performed with CD3− CD56+ NK cells. NK-susceptible K562 cells (P<0.001 compared with responder NK cells) induced substantial NK cell activation that was similar to the effect of incubation with PMA plus ionomycin (P<0.001 compared with responder NK). Neither hESCs nor hESC^KD induced relevant IFN-γ release (Fig. 5c). Similarly, CD107a surface expression on NK cells was only significantly increased upon either incubation with PMA plus ionomycin (P<0.001) or K562 challenge (P=0.001). hESC and hESC^KD incubation (P=1.0 compared with responder NK cells) did not alter CD107a expression (Fig. 5d).

Differences in expression patterns of stimulatory and inhibitory NK ligands, more than differences in HLA I expression, might account for the differences in NK cytotoxicity between K562 cells and hESC^KD (Table 1). The stimulatory danger signals MHC class I polypeptide-related sequence A (MICA) (P=0.012), MICB (P<0.001) and, most obviously, Hsp70 (P<0.001) were expressed at significantly higher levels on K562 cells. The stimulatory NK ligands ULBP1, 2 and 3 and the NK inhibitory ligands HLA-E and HLA-G were similarly low on hESC, hESC^KD and K562 cells and might therefore not account for the different NK susceptibilities of these cells.

The present study aimed at revealing the mechanisms of hESC rejection and describing the survival benefits of HLA I knockdown (hESC^KD). The immunological nature of cell fate after xenogeneic transplantation could be demonstrated by the rapid hESC death in immunocompetent Balb/c mice and the survival in severely immunodeficient SCID-beige recipients, exhibiting defective T cell, B cell and NK cell responses. We could clearly demonstrate strong Th1 (IFNγ release) and Th2 (IL-4 release) responses in Balb/c recipients of hESC transplants. In addition, we showed that those T cell activations were mainly triggered by hESC HLA I expression because hESC^KD largely failed to induce T cell responses. Differences in the completeness and durability of HLA I knockdown might therefore determine which graft undergoes slowed rejection or proceeds to long-term survival. We could further show that stem cell transplantation increased the fraction of CD4+ CD25+ Foxp3+ Tregs in regional lymph nodes and that only recipients of hESC^KD showed a prolonged polarization of T cells toward a regulatory phenotype. The induction of transplantation tolerance for ESCs by host conditioning with antibodies specific for the CD4 and CD8 receptors (Robertson et al., 2007) or by costimulatory blockade (Grinnemo et al., 2008; Pearl et al., 2011) has been successful in prolonging cell survival. Tregs have been identified as central regulators for this state of acquired immune privilege. Interestingly, hESC^KD seem to induce similar tolerogenic responses in non-conditioned immunocompetent hosts. The regulatory properties of Tregs might be most crucial in the induction phase of tolerance early after cell transplantation (Muller et al., 2010).

Strategies involving the gene transfer of an anti-HLA I single chain IB (Beyer et al., 2004) or plasmids containing siRNA (Gonzalez et al., 2005) were recently developed to downregulate surface HLA I. Using the IB technology as a second line of capture mechanism for the very few HLA I molecules that still were synthesized despite the use of siRNA caused no retainment of intracellular HLA I molecules. Such retainment has been reported with IB alone (Busch et al., 2004) and might paradoxically increase the antigenicity of cell transplants.

Although T cells are crucial for hESC recognition (the afferent limb of immune activation), other effector cells might participate in the killing. In a study with allogeneic leukemia cells, the blocking of both perforin- and FasL-dependent killing, two major cytotoxic effector mechanisms of T cells, did not prevent cell rejection. Macrophages had been identified as the effector cell population (Nomi et al., 2007). It has been further demonstrated that T cells can mediate rejection of allogeneic marrow progenitor cells through alternative effector pathways independent of perforin-, FasL-, TNF-like weak inducer of apoptosis (TWEAK)-, and TNF-related apoptosis-inducing ligand (TRAIL)-dependent cytotoxicity (Komatsu et al., 2003; Zimmerman et al., 2005). Other cytolytic mechanisms involving antibodies (Drukker and Benvenisty, 2004) and complement (Koch et al., 2006) have also been suggested to contribute to the killing of hESCs in vivo.

The importance of antibody-mediated rejection after cell transplantation is gaining increasing awareness and has already been reported for hematopoietic (Spellman et al., 2010) and mesenchymal stem cell (MSC) transplantation (Beggs et al., 2006; Cho et al., 2008) and for islet transplantation (Mohanakumar et al., 2006). FACS-based DSA quantification using living cells identifies all bound antibodies both against HLA and non-HLA epitopes. In our murine model, we did not find naturally occurring anti-hESC antibodies, but cell transplantation induced a rapid surge of DSAs. Antibody production after hESC transplantation was restricted to the HLA class I and was mostly specific for hESC HLA I. HLA II was absent on hESCs and could not be induced by stimulation. Consequently, there was no significant antibody production against HLA II. hESC re-injection induced a boost in HLA I IgG antibodies, as has been reported after MSC transplantation in swine (Cho et al., 2008). Even non-cytotoxic alloantibodies detected by flow cross-match, which are negative in conventional complement-dependent lymphocytotoxic cross-match, have been reported to be sufficient to cause immunologic damage of transplanted islets (Mohanakumar et al., 2006). The antibody surge in our study occurred rapidly (within 5 days) and accompanied the strong cellular immune activation. Reports that short-term T-cell-specific immunosuppression prevented the antibody response against transplanted MSCs support the hypothesis of a T-cell-driven humoral response against cellular grafts (Poncelet et al., 2008). Ongoing humoral rejection of islet grafts was successfully treated with an anti-CD20 monoclonal antibody (Kessler et al., 2009). Epitope spreading might have caused some concomitant increase of non-hESC-specific antibodies in our study and has also been recognized after clinical hematopoietic stem cell transplantation (Leffell et al., 2009; Papassavas et al., 2002). The antigenicity of hESC^KD was low enough to prevent the induction of hESC-specific antibody production.

The naturally low HLA I expression of hESCs could make them susceptible to NK killing. However, the histological sections of immunocompetent hESC-rejecting Balb/c mice revealed macrophages as the dominant cell population, followed by CD3+ lymphocytes and revealed only scattered NK cells. The early surge of hESC-specific antibodies and the direct correlation between the level of HLA I knockdown and the strength of the T cell response, as well as the pace of cell rejection, are further proof of an adaptive T-cell-directed immune response. There are reports of substantial (Dressel et al., 2010; Dressel et al., 2008) and negligible (Bonde and Zavazava, 2006; Koch et al., 2008; Mammolenti et al., 2004) in vitro NK responses against murine ESCs, presumably depending on the experimental conditions. Although even syngeneic NK cells have been reported to kill murine ESCs after adequate stimulation in vitro (Dressel et al., 2010), murine ESCs have uniformly been reported to form teratomas in syngeneic (Dressel et al., 2008; Koch et al., 2006; Kolossov et al., 2006) but not allogeneic (Dressel et al., 2008; Koch et al., 2006) or xenogeneic recipients (Dressel et al., 2008). Because the cytotoxic activity of mouse NK cells in vitro strongly depended on sufficient activation, SCID mice, deficient in T and B cells but with functional NK cells, were treated with poly(I:C) and were used as recipients for murine ESCs. The in vivo NK activity was found to be increased and the frequency of teratoma formation significantly reduced (Dressel et al., 2010). In vivo rejection of murine ESC derivatives by NK cells has also been reported in Rag2^−/− mice (Tabayoyong et al., 2009). Both mouse models lack functional lymphocytes and contain greatly increased numbers of NK cells (Tabayoyong et al., 2009). The impact of NK cells in immunocompetent hosts might, however, be less important. hESCs have been shown not to be effectively recognized by human NK cells in vitro (Drukker et al., 2002) but only after additional in vitro NK stimulation (Tseng et al., 2010). Besides the lack of MHC I on target cells, the activation of additional stimulatory NK receptors is necessary to trigger NK killing (Bryceson et al., 2006), and the expression of stimulating NK ligands can lead to lysis even in the presence of MHC I (Cerwenka et al., 2001). Murine ESCs have been shown to express high levels of the NKG2D ligands REA-1, H60, MULT-1 (Bonde and Zavazava, 2006; Dressel et al., 2008; Frenzel et al., 2009) and ICAM-1 (Frenzel et al., 2009), and blocking of these receptors prevents cell lysis (Frenzel et al., 2009; Preynat-Seauve et al., 2009). Murine ESC-derived cardiomyocytes, however, which are negative for MHC I but also negative for NKG2D ligands, were spared by NK cells (Frenzel et al., 2009). Human equivalents for NKG2D ligands are MICA and MICB and ULBP1, 2 and 3. Here, we have shown that hESCs and hESC^KD express NKG2D ligands at negligible levels and might therefore be less susceptible to human and murine NK killing (Drukker et al., 2006) than their murine counterparts. Furthermore, hESC and hESC^KD express Hsp70 at much lower levels than the HLA-I-negative and NK-susceptible erythroleukemia cell line K562. Hsp70-blocking experiments with K562 cells strongly reduced NK cytotoxicity (Stangl et al., 2008), demonstrating the important activating role of Hsp70. Because of the 94.5% homology between human and mouse Hsp70, murine NK cells have been previously shown to kill Hsp70-expressing MICA-positive human tumors (Elsner et al., 2007). NK depletion in nude mice has been demonstrated to facilitate tumor growth of K562 xenografts and the development of metastases (Yoshimura et al., 1996). hESCs have been shown previously to only very weakly express ligands for the natural cytotoxicity receptor (NCR) NKp44, and ligands for NKp30, NKp46 and CD16 were not found (Drukker et al., 2002). Blocking experiments with NCR antibodies did not inhibit in vitro NK killing of hESC derivatives and made NCR appear less important for NK activation of hESCs (Preynat-Seauve et al., 2009). The low levels of HLA-E and HLA-G on hESCs and hESC^KD do not support an NK-protective role in these cell populations. However, HLA-G is upregulated in hESC-derived mesenchymal progenitor cells and has been shown to contribute to the immunosuppressive properties of the mesenchymal stem cell phenotype (Yen et al., 2009).

We demonstrate a central role for HLA I in the recognition of hESCs by xenogeneic or allogeneic T cells. The HLA-I-knockdown hESC^KD experience strongly diminished cellular and antibody responses. Despite HLA-I-knockdown, hESC^KD do not trigger substantial NK activity, which might be related to their negligible expression of stimulating NK ligands. We believe the generation of hypoantigeneic cell sources is a requisite for future ex vivo generation of hypoantigeneic tissues for cell replacement therapies.

Male Balb/c (n=72) and SCID-beige (CB17.Cg-Prkdc^scidLyst^bg/Crl; n=8) mice were purchased from Charles River Laboratories (Sulzfeld, Germany) and received humane care in compliance with the Guide for the Principles of Laboratory Animals.

H9 hESCs (Wicell, Madison, WI) were maintained on inactivated murine embryonic fibroblast (MEF) feeder layers (Millipore, Billerica, MA) in hESC medium. hESC colonies were separated from MEF feeders by incubation with dispase (Invitrogen) and subcultured on feeder-free Matrigel-coated plates (Matrigel from BD Biosciences) before use. The hESC and iPS cell characterization kit (Applied StemCell, Sunnyvale, USA) was used to demonstrate the expression of Oct-4, Sox-2, SSEA-3 and TRA-1-60. HLA immunotyping was performed by using LiPA HLA-DRB1, -DQB1, -A, -B and -C kits according to the manufacturer's protocol (Innogenetics, Alpharetta, GA).

Gene silencing both at the post-transcriptional level using small interfering RNA (siRNA) (Caplen et al., 2001) and at the post-translational level with intrabody (IB) technology (Beyer et al., 2004) were used to generate HLA-I-knockdown hESCs (hESC^KD).

HLA class I siRNA (Ambion, Darmstadt, Germany), a target-specific 20–25 nucleotide siRNA, was used at a concentration of 20 μM. A total of 10 μl of 20 μM siRNA solution was incubated with 200 μl OptiMEM (Invitrogen) containing 0.3 μl DharmaFECT 1 (Thermo Scientific). Then, 1.2 ml of prepared siRNA in OptiMEM was added into each hESC well and incubated for 24 hours. To assess the efficiency of siRNA delivery into cells, the localization of the siGLO green transfection indicator (Thermo Scientific), which localizes to the nucleus, was checked. hESC were transfected with siGLO green (20 nM), complexed with DharmaFECT 1 transfection reagent (0.1 μl/well), fixed with 3.7% formaldehyde solution and stained with phalloidin–Alexa-Fluor-546.

E1-deleted human Ad5 recombinant adenovirus containing the anti-human-HLA-I single-chain intrabody fragment (AdScFv), driven by the cytomegalovirus (CMV) promoter, was constructed, purified and assayed for plaque forming units (pfu) per ml, as described previously (Kolls et al., 1994). The intrabodies recognize a non-polymorphic HLA I epitope. Transduction of cells was performed with a multiplicity of infection (MOI) of 150 pfu. To assess transduction efficiency, E1-deleted, human Ad5 recombinant adenovirus encoding for the enhanced green fluorescent protein (AdvGFP), as a reporter construct, was generated and served as the control construct. The AdvGFP transduction dilutions were prepared at a MOI of 150 pfu.

hESCs were dissociated by trypsinization, and 5 × 10⁵ cells were stained for 45 minutes at 4°C in 100 μl of 0.1% BSA in PBS containing an appropriate dilution of a desired phycoerythrin (PE)-conjugated antibody. Primary antibodies used were against: human-SSEA-1 (clone MC-480), SSEA-4 (clone MC813-70), IgG3 isotype control from R&D Systems (Minneapolis, MN), HLA DR/DP/DQ (clone WR18), IgG2a isotype control from Abcam (Cambridge, MA), β2-microglobulin (clone TÜ99, HLA -A, -B and -C (clone DX17), CD80 (clone L307.4), CD40 (clone 5C3), CD86 [clone 2331(FUN-1)] and IgG1 (clone MOPC-21) from BD Biosciences. Sample measurement was performed on a FACSCalibur system (Becton Dickinson, Heidelberg, Germany) and analysis was performed using FlowJo (Tree Star, Ashland, OR). In the histograms, the isotype controls are represented by solid gray areas and the antigens of interest by black lines. To quantify surface molecule expression, mean fluorescence intensities (MFI) of the above molecules were compared to those of the associated negative isotype controls. Normalized expression was calculated as (MFI of molecule) divided by (MFI of isotype control). Therefore, a normalized expression of 1 corresponds to no molecule expression.

Before transplantation, hESCs were trypsinated and resuspended in sterile PBS at 1 × 10⁶ cells per 50 μl. Previous studies have shown that injections of 1 × 10⁶ hESCs provide reliable signals for in vivo imaging and are sufficient to form stable grafts in immunocompromised recipients (Lee et al., 2009). hESCs viability was ~95% as determined by Trypan Blue staining. hESC transplantation was performed by direct injection into the right gastrocnemius muscle of recipient mice using a 27-gauge syringe. For histology, gastrocnemius muscle or teratomas were formalin-fixed and paraffin-embedded. Hematoxylin and eosin (H&E) or periodic acid Schiff (PAS) stained slides were interpreted by an expert pathologist (J.V.) blinded to the study. The density of MAC2+ macrophages, CD3+ lymphocytes and KLRA1+ NK cells was quantified in eight animals receiving hESC or hESC^KD, respectively using Image J (NIH, Bethesda, MD).

hESCs were transduced with pLenti CMV/TO V5-LUC Puro reporter gene at a multiplicity of infection (MOI) of 10 to stably express firefly luciferase (Fluc). Viral aliquots were kindly provided by Eric Campeau (Campeau et al., 2009). The Fluc-tagged hESC line is resistant to puromycin and cells were purified using a concentration of 0.2 μg/ml puromycin for 5 days followed by plating on feeder layer cells for culturing. BLI was performed using the Xenogen in vivo imaging system (Caliper Life Sciences, Hopkinton, MA) as reported earlier (Cao et al., 2006). This imaging modality allows for quantitative, longitudinal and non-invasive studies of cell viability and location within the same animals (Sun et al., 2009; Swijnenburg et al., 2008b). Mice were anesthetized with 2% isoflurane and D-luciferin (Caliper Life Sciences) was administered intraperitoneally at a dose of 375 mg per kg of body weight. At the time of imaging, animals were placed in a light-tight chamber, and photons emitted from Fluc-expressing hESCs were counted. The recipient mice were scanned repetitively on days 0, 1, 3, 5, 7, 14, 21, 28, 35 and 42, or until the signal dropped to background levels. BLI signals were quantified in units of photons per second (total flux) and are presented on a logarithmic scale.

The spleen of recipient animals was harvested 5 days after hESC transplantation, to isolate recipient splenocytes. Elispot assays using 1 × 10⁵ mitomycin-treated hESC as stimulator cells and 1 × 10⁶ recipient splenocytes as responder cells were performed according to the manufacturer's protocol (BD Biosciences) using IFN-γ and IL-4-coated plates. Spots were automatically enumerated using an Elispot plate reader (CTL, Ohio, USA) for scanning and analyzing.

Inguinal lymph nodes were harvested from untreated Balb/c control mice and from recipients of hESC or hESC^KD transplants after 5, 14 and 42 days. The percentage of CD25+ Foxp3+ cells among the CD4+ population was assessed by FACS analysis using the FoxP3 Staining Kit (BD Biosciences).

hESC-specific mouse xenoantibodies were detected by FACS analysis. The serum of Balb/c mice before and 5 days after transplantation of hESCs or hESC^KD was incubated with graft cells and the binding of graft-specific antibodies was quantified. Only IgM antibodies were analysed because of their known rapid surge within 5 days upon allogeneic stimulation. Sera from 6 recipient mice per group were decomplemented and equal amounts of sera and hESC suspensions (1 × 10⁶ cells per ml) were incubated. IgM antibodies were stained by incubation of the cells with a PE-conjugated goat antibody specific for the Fc portion of mouse IgM (BD Biosciences). Cells were washed and then analyzed on a FACSCalibur system (BD Biosciences). Fluorescence data were expressed as MFI.

Sera from untreated native Balb/c mice or Balb/c recipients of hESC or hESC^KD 5 days after cell transplantation were centrifuged (8,000–10,000g) for 10 min. Then, 20 μl of undiluted serum supernatant per test well was incubated with 5 μl of LABScreen single antigen beads (HLA Class I, LS1A04 Lot 006 and HLA Class II, LS2A01 Lot 008, One Lambda, Canoga Park, CA) for 30 minutes in the dark at 20–25°C. The specimens were then washed before being incubated with PE-conjugated anti-mouse IgG for a further 30 minutes at room temperature. Beads were washed again and re-suspended by adding 80 μl of PBS to each well. Fluorescence was measured with the Luminex100 flow analyzer (Luminex, Austin, TX) in six animals per group and the data were expressed as MFI. A MFI Z-score, indicating how many standard deviations a specific antibody was above or below the mean of native mice was introduced to describe the humoral response.

For allogeneic in vitro Elispot assays, human peripheral blood mononuclear cells (PBMCs) were separated from 50 ml peripheral blood drawn from healthy blood donors and CD3+ CD56− lymphocytes and CD3− CD56+ NK cells were purified by MACS sorting (Miltenyi, Auburn, CA). Elispot assays were performed with 5 × 10⁵ hESCs and 5 × 10⁵ human CD3+ CD56− lymphocytes or CD3− CD56+ NK cells.

The cell-mediated cytotoxicity of sorted CD3− CD56+ human NK cells was evaluated in a CD107a degranulation assay. This is a sensitive assay for the surface expression of CD107a as a result of activation-induced degranulation and secretion of the lytic granule contents. NK cells were incubated for 6 hours with FITC-labeled anti-CD107a monoclonal antibody (BD Biosciences). K562, hESCs or hESC^KD at ratios of 1:10, or PMA+ionomycin was added and the surface expression of CD107a was determined by FACS.

Data are presented as means ± s.d. unless stated otherwise. Comparisons between groups were performed by independent sample Student's t-tests or ANOVA with Bonferroni post-hoc tests, where appropriate. Differences were considered significant for P<0.05. Statistical analysis was performed using SPSS statistical software for Windows (SPSS, Chicago, IL, USA).