Anti-tumor MHC class Ia-unrestricted CD8 T cell cytotoxicity elicited by the heat shock protein gp96



In Xenopus as in mammals, gp96 stimulates MHC-restricted cellular immunity against chaperoned minor histocompatibility (H) antigens (Ag). In adult Xenopus, gp96 also elicits peptide-specific effectors against MHC class Ia-negative 15/0 tumors. To determine whether gp96 can generate functionally heterogeneous CD8+ effectors (CTL that kill MHC class Ia+ minor H-Ag-disparate lymphoblasts and MHC class Ia tumor targets), LG-6 isogenetic frogs were immunized with gp96 purified either from MHC-identical but minor H-Ag-disparate LG-15 normal tissues or from the MHC class Ia-negative 15/0 tumor line (derived from LG-15 frogs). LG-15 normal liver-derived gp96 did not induce detectable CD8+in vitro killing against 15/0 tumor cells. However, 15/0-derived gp96 did induce killing against both MHC class Ia+ LG-15 lymphoblasts and the MHC class Ia 15/0 tumor, but not against another MHC class Ia tumor (B3B7) or against LG-6 lymphoblasts. Tumor killing was better when 15/0 rather than normal LG-15 irradiated stimulators were used, but in vitro stimulation without prior in vivo immunization was ineffective. These data suggest that (1) 15/0-derived gp96 chaperones minor H-Ag shared with normal LG-15 lymphocytes and elicits MHC-restricted CTL, and (2) 15/0-derived gp96, but not normal liver-derived gp96, generates CD8+ effectors that kill 15/0 tumor cells in the absence of MHC class Ia expression.


Histocompatibility antigen


Heat shock protein

1 Introduction

gp96 was the first heat shock protein (hsp) shown to evoke potent protective tumor immunity in mammals 1. At least two different immunological properties of gp96 bear on this observation. The first is the peptide-dependent capacity of gp96 to chaperone and elicit adaptive CTL responses against antigenic peptides through receptor-mediated endocytosis by antigen-presenting cells (APC) and representation by APC MHC class Ia 2, 3. The second property is a peptide-independent capacity of gp96 to stimulate APC (i.e. macrophages and dendritic cells) to produce pro-inflammatory cytokines and up-regulate costimulatory molecules 46. These two different immunological properties of gp96 appear to result from its interaction with different surface receptors. CD91, an endocytic receptor for α2-macroglobulin, has been shown to mediate the uptake of gp96 (as well as hsp90, hsp70, and calreticulin) by APC and the subsequent representation of chaperoned peptides 7, 8. Interaction of gp96 with Toll-like receptors (TLR)2/4 and subsequent signaling through the NF-kB pathway has also been reported 5.

The complexity of the gp96-mediated anti-tumor response is also apparent at the level of the cellular effectors involved. Although studies in mammals involving CD8 depletion 9, 10 and research using tumor-secreted or tumor membrane-associated gp96 11, 12 implicate CTL, increasing evidence suggests that natural killer (NK) cells 10, 13 and other less well-characterized cell types 14, 15 may also play a critical role.

Given the high degree of phylogenetic conservation of the gp96 structure, its involvement in both innate and adaptive immune responses is also likely to be evolutionarily conserved. In this regard, we have developed a model to investigate the immunological properties of gp96 in the frog Xenopus, a genus whose common ancestors with mammals diverged some 350 million years ago 16. The immune systems of Xenopus and mammals are fundamentally similar (e.g. rearranging TCR and Ig genes, as well as MHC class I- and class II-restricted T cell recognition 17). We have shown that, like mammalian hsp, the Xenopus gp96 protein homologue stably and noncovalently complexes peptides in vitro and that such chaperoned peptides are efficiently represented by MHC class I of mouse macrophages to specifically stimulate a murine CTL line 18. Additionally, both Xenopus gp96 and hsp70 generate MHC-restricted thymus-dependent immunity in vivo and CTL against chaperoned anti-minor H-Ag in vitro 19. To study the ability of gp96 to generate anti-tumor immunity, we took advantage of a Xenopus transplantable tumor, namely the 15/0 lymphoid tumor line which is derived from a thymic tumor that spontaneously arose in a LG-15 cloned frog 20, 21. 15/0 cells do not express MHC class I mRNA or protein 20, but they do express non-classical MHC class Ib (XNC) mRNA 2022 as well as β2-microglobulin (unpublished observations). Immunization of Xenopus with gp96 purified from this highly tumorigenic 15/0 tumor generates potent anti-tumor immunity in the absence of MHC class Ia restriction, whereas immunization with gp96 purified from normal tissue only has a minimal anti-tumor effect 18. Interestingly, a weak but significant gp96 anti-tumor response has also been obtained in naturally MHC class Ia-deficient but immunocompetent Xenopus larvae. In this instance, however, the lack of any difference between tumor- and normal tissue-derived gp96 suggests that in larvae, gp96 induces an innate type of anti-tumor response independent of chaperoned peptides 19. This peptide-independent effect is likely due to gp96 itself rather than to an LPS contaminant, since relative to mammals, Xenopus larvae and adults are minimally responsive to purified LPS 2325.

Defects in MHC class Ia expression are frequently found in malignancies 26. This is usually considered to be a poor prognosis because of the critical role of MHC class Ia molecules in immune recognition. In this regard, the potential ability of gp96 to induce responses against tumors that do not express MHC class Ia molecules has yet to be considered. Our Xenopus model with its MHC class Ia-negative 15/0 tumor, different minor H-Ag-disparate but MHC identical clones of animals, and the poor response of this species to LPS provides the opportunity to explore this issue. We report here that in addition to the ability of gp96 to effect a sustained NK-like anti-tumor activity that is independent of the tissue source from which it was purified, tumor-derived gp96 has the unique ability to specifically stimulate classical MHC class Ia-unrestricted cytotoxic CD8+ T lymphocyte anti-tumor effectors.

2 Results

2.1 Anti-tumor cytotoxic activity generated by gp96 in syngeneic LG-15 cloned Xenopus

As mentioned, the 15/0 tumor line of LG-15 genotype does not express any MHC class I protein or mRNA and is highly tumorigenic (poorly immunogenic 20). We have shown that gp96 immunization generates a potent immune response in vivo against the 15/0 tumor 18. To further characterize effectors involved in this MHC class I-unrestricted gp96-mediated response, we determined their cytotoxic activity in vitro. Total splenocytes from LG-15 frogs immunized with 15/0 tumor-derived gp96 and restimulated in vitro with irradiated 15/0 cells consistently and significantly killed 15/0 tumor cells (Fig. 1B). No significant killing was obtained in the absence of restimulation (Fig. 1B) or by stimulation with splenocytes from naïve LG-15 animals (data not shown). Interestingly, splenocytes from LG-15 frogs bearing a 15/0 tumor also showed killing activity against 15/0 tumor targets.

Figure 1.

Cytotoxicity generated by 15/0 tumor-derived gp96 immunization in LG-15 syngeneic frogs. (A) Splenocytes from outbreds (naïve or immunized with several MHC-disparate LG-15 skin grafts) were restimulated in vitro with irradiated LG-15 splenocytes and assayed for killing of LG-15 lymphoblast targets. (B) Splenocytes from 15/0-derived gp96-immunized LG-15 or (C) 15/0 tumor-bearing frogs were restimulated for 6 days with or without (unstim.) irradiated 15/0 tumor cells and assayed for killing of 15/0 targets. Effectors were pre-incubated with isotype control, anti-MHC class Ia (TB17), anti-IgM (10A9), anti-CD8 (AM22), or anti-NK (1F8) mAb (5–10 µg mAb/well). PHA-lymphoblasts (25,000) or 15/0 tumor targets (10,000) labeled with [3H]thymidine were incubated with effectors (E/T ratio 10:1) in triplicate for 4 h at 26°C in 96-well plates. Thymidine uptake was determined by scintillation spectrometry. The percentage of killing activity was calculated by determining the loss of apoptotic DNA relative to target alone control ([cpm control – experimental]/[cpm control] × 100). Variation within triplicate groups was less than 10% of group mean cpm. The cpm value for the 10,000 tumor targets ranged between 5,000 and 6,000 cpm, and for the 25,000 PHA-lymphoblasts between 1,000 and 1,200 cpm.

The cellular mediation of this cytotoxicity (e.g. NK and/or CTL) was further explored using an antibody inhibition protocol. In a typical CTL response (Fig. 1A) generated by immunization of an outbred frog with an MHC-disparate LG-15 skin graft and in vitro restimulation of the outbred splenocytes with irradiated LG-15 splenocytes, the strong killing against LG-15 lymphoblast targets is inhibited by pre-incubation of the effectors with anti-CD8 mAb; pre-incubation with anti-NK (1F8) mAb has little inhibitory effect. In contrast, when LG-15 frogs are immunized with 15/0-derived gp96 and restimulated in vitro with irradiated 15/0 tumor cells, killing activity is decreased by either anti-CD8 or anti-NK mAb (Fig. 1B). These results suggest that the gp96-stimulated anti-tumor response involves both CD8+ T cells and NK splenocytes. Interestingly, the killing activity of splenocytes from tumor-bearing LG-15 animals is also sensitive to anti-CD8 and anti-NK mAb inhibition (Fig. 1C). Since we did not immunize these tumor-bearing frogs with gp96, they must have been exposed to the relevant immunogens (15/0 tumor Ag) in vivo.

2.2 Cytotoxic activity of CD8+ and CD8 effectors generated by gp96 in syngeneic LG-15 Xenopus

To better distinguish between the relative contribution of NK and CD8+ T cell effectors in gp96-mediated 15/0 killing, we separated CD8+ and CD8 splenocytes with mAb-coated micromagnetic beads, before assaying their killing activity in vitro. CD8+ T cells were 95% pure and CD5 positive (pan T cell marker in Xenopus27; Fig. 2). No consistent expression of the NK marker 1F8 28 was observed on purified CD8+ T cells (Fig. 2A), whereas a minor NK cell population was detected in the CD8-negative fraction after 6 days of culture (Fig. 2B). As shown in Fig. 3, immunization of LG-15 frogs with 15/0 tumor-derived gp96 did not result in consistent anti-15/0 CD8 cytotoxicity. Rather, the potent killing activity resulting from gp96 immunization was confined to the CD8-depleted effector population. These effectors readily killed MHC class Ia 15/0 tumor targets but did not kill MHC class Ia+ lymphoblast targets from either outbred or LG-15 frogs. Similar CD8 NK-like killing activity was also obtained when gp96 purified from normal LG-15 tissue (i.e. no tumor Ag) was used as the immunogen (Fig. 3). In vitro restimulation of naïve frogs (Fig. 4), or LG-15 frogs immunized with total protein lysate 19, was insufficient to generate such potent NK-like killing. Although 1F8+ NK splenocytes were detected by flow cytometry (Fig. 2B), their relatively small numbers spoke against their purification and subsequent use in killing assays.

Figure 2.

Surface phenotype of cell effectors after in vitro restimulation. In two different experiments (A/B and C), splenocytes from LG-15 cloned frogs that had been immunized with gp96 purified from 15/0 tumors were harvested and restimulated in vitro for 6 days with irradiated 15/0 stimulators. (A, C) CD8+ cells purified by MACS (see Sect. 4) or (B) CD8-depleted cells were stained and analyzed with a FACSCalibur. Events (10,000) were collected, and dead cells stained with propidium iodide were gated out. The following mAb were used: biotinylated anti-CD8 (AM22) or anti-MHC class II (AM20) mAb, PE-conjugated anti-CD5 (2B1) and FITC-conjugated anti-NK (1F8) mAb. FITC-conjugated streptavidin was used for biotinylated mAb. The thick line in AM22 histograms represents the isotype-negative control.

Figure 3.

gp96 priming generates potent NK cytotoxicity. Splenocytes from LG-15 cloned frogs immunized with gp96 purified from normal LG-15 liver (A, B) or 15/0 tumor (C, D) were harvested and restimulated in vitro for 6 days with irradiated 15/0 stimulators. Various numbers of CD8+-enriched and CD8-depleted effectors (selected by MACS} were then incubated in triplicate (4 h at 26°C in 96-well plates) with 10,000 MHC class I 15/0 or 25,000 MHC class I+ LG-15 or outbred (OB) PHA-lymphoblast targets. The percentage of killing activity was calculated as in Fig. 1.

Figure 4.

Irradiated 15/0 stimulators are not sufficient to generate potent LG-15 NK cytotoxicity. Splenocytes from naïve LG-15 cloned frogs or from LG-15 frogs immunized with gp96 purified from normal LG-15 liver were harvested and restimulated with irradiated 15/0 stimulators. CD8-depleted effectors (selected by MACS} were assayed for killing of MHC class I 15/0 or MHC class I+ LG-15 or LG-6 PHA-lymphoblast targets as in Fig. 3.

2.3 Anti-tumor cytotoxic activity of CD8+ T cell effectors generated by gp96 in a minor-Ag-disparate LG-6 Xenopus clone

In vivo anti-CD8 mAb treatment markedly impaired the LG-15 anti-tumor effector response 21, 29. Although in the present experiments we did not detect consistent in vitro CD8+ killing of 15/0 tumor cells by LG-15 splenocytes, it is possible that the anti-tumor CD8+ T cell precursor frequency is too low to be revealed by our killing assay. To increase the specific response against 15/0 cells, we took advantage of our system of alloimmunity directed against minor H-Ag (Fig. 5, top panel). The 15/0 tumor is as perfectly transplantable into allogeneic LG-6 frogs as it is into LG-15, the clone from which the 15/0 line originated. These two Xenopus clones display the identical heterozygous (a/c) MHC haplotype, but differ from each other by multiple minor-H loci.

Figure 5.

15/0 tumor-derived gp96 immunization of minor H-Ag-disparate LG-6 frogs generates CD8+ MHC class Ia-unrestricted cytotoxicity. Upper panel: Schematic of the experiments; lower panel: results of the experiments. Splenocytes from minor H-Ag-disparate LG-6 cloned frogs that were either naïve (A) or had been previously immunized with 15/0 tumor-derived gp96 (B) were restimulated in vitro for 6 days with either normal LG-15 splenocytes or irradiated 15/0 tumor-stimulators (as indicated in parentheses in the figure). CD8+ T cells were assayed for killing of MHC class Ia 15/0 tumor cells, MHC class Ia+ minor H-Ag-disparate LG-15 PHA-lymphoblasts, or cognate LG-6 PHA-lymphoblasts, as in Fig. 3.

Results from several experiments (Fig. 5 depicts a representative experiment) indicate that CD8+ splenocytes from LG-6 animals immunized with 15/0 tumor-derived gp96 consistently kill both MHC class Ia 15/0 tumor targets and MHC class I+ minor H-Ag-disparate LG-15 lymphoblasts, but not cognate MHC class Ia+ LG-6 blasts. Killing of 15/0 targets is more pronounced when irradiated 15/0 tumor rather than irradiated normal LG-15 splenocyte stimulators are used, but in vitro stimulation with irradiated 15/0 tumor cells in the absence of prior immunization is not sufficient to generate such CD8+ MHC-unrestricted cytotoxicity (Fig. 5A). Interestingly, immunization of LG-6 frogs with gp96 derived from normal minor H-Ag-disparate LG-15 tissues does not result in any anti-tumor killing activity by CD8+ T cell effectors (data not shown). This is in sharp contrast to potent NK-like killing activity of the CD8-depleted fractions shown in Fig. 3.

To further explore the specificity of gp96-mediated anti-15/0 CD8+ T cell effectors, we used as target cells the NK-sensitive Xenopus B3B7 tumor line 30 derived from the Xenopus inbred F strain. Like the fully allogeneic 15/0 line, the B3B7 line is thymus derived and MHC class Ia negative. Unlike 15/0 targets, however, B3B7 tumor cell targets were not killed by CD8+ T cell effectors generated by immunization with gp96 purified from the 15/0 tumor or from normal outbred or LG-15 tissue (Fig. 6). However, they were killed to the same extent as 15/0 tumor targets by CD8-depleted NK-like effectors (data not shown). Further evidence that anti-tumor CD8+ T cells generated by gp96 immunization are not NK-like cells comes from the absence of a response obtained by immunization and restimulation with either heat-denatured gp96 (30 min at 95°C) or LPS (Fig. 6C, D).

Figure 6.

Specificity of the CD8+ MHC-unrestricted anti-tumor responses stimulated by gp96. (A–E) Splenocytes from LG-6 cloned frogs that were either naïve (C) or previously immunized with either 15/0 tumor-derived gp96 (Tugp96) or LG-6 liver-derived gp96 (Ngp96) were restimulated with irradiated 15/0 tumor cells. Cytotoxicity of CD8+ T cells was assayed against MHC class Ia tumor targets (15/0 or B3B7 tumor cells) or MHC class Ia+ minor H-Ag-disparate PHA-lymphoblast targets (minor H-Ag disparate LG-15 or cognate LG-6), as in Fig. 3. (C, F) Splenocytes from LG-6 cloned frogs were immunized with LPS or heat-denatured 15/0 tumor-derived gp96 and restimulated for 6 days with the same immunogen (50 µg/ml). CD8+ T cells were assayed against 15/0 (C) or B3B7 (F) targets.

As just mentioned, our results indicate that in vitro restimulation is essential for generating anti-tumor cytotoxic CD8 T cell effectors, presumably because it effects expansion of primed precursors as is the case for bona fide CTL in mammals 31 and in Xenopus19, 32. Since in mammals, antigenic peptides chaperoned by hsp can be taken up and cross-presented by APC in vitro, we asked whether in Xenopus, in vitro stimulation with 15/0 tumor-derived gp96 is sufficient to generate these anti-tumor MHC-unrestricted CD8+ effectors. CD8+ splenic T cells harvested from LG-6 adults previously immunized with 15/0 tumor-derived gp96 were restimulated in vitro with this same tumor-derived gp96 rather than with irradiated intact tumor cells. These CD8+ cells killed 15/0 tumor targets but did not kill MHC class I+ lymphoblast targets from either MHC-disparate outbreds, minor H-Ag-disparate LG-15, or cognate LG-6 frogs (Fig. 7). Similar results were obtained in two different experiments. It should be noted that the overall cell recovery after in vitro restimulation, which is usually low with irradiated cell stimulators (no recombinant IL-2 is available in Xenopus to facilitate cell growth), was even lower when only 15/0 tumor-derived gp96 was used. Nevertheless, these results clearly support the proposition that 15/0 tumor-derived gp96 chaperones a pool of Ag that provides the proper in vitro stimulation for these MHC-unrestricted populations of CD8+ anti-tumor effectors. The cytotoxic characteristics of these classical MHC class Ia-unrestricted cytotoxic T cells and NK cells are summarized in Table 1.

Figure 7.

15/0 tumor-derived gp96 is sufficient to restimulate CD8+ MHC-unrestricted effectors in vitro. Splenocytes from LG-6 cloned frogs immunized with 15/0-derived gp96 were restimulated in vitro for 5 days with 150 µg/ml of 15/0 tumor-derived gp96 protein and assayed for killing activity of 15/0 tumor, LG-15, LG-6, or outbred lymphoblast targets.

Table 1. Characteristics of the three types of effectors generated by 15/0 tumor-derived gp96a)
LG-6 immunized with:In vitro re-stimulationCell effectorsTargets (killed)
  1. a) Immunization with normal LG-15 tissue-derived gp96 generates only effector cell types 1 and 3; immunization with LG-6 tissue-derived gp96 generates only effector cell type 3.

15/0 tumor-derived gp96 Irrad. 15/0 cells or 15/0-derived gp96(1) Anti-minor H-Ag CTLLG-15 lymphoblasts (yes), LG-6 lymphoblasts (no), 15/0 (no), B3B7 (no)
(2) Anti-tumor class Ia-unrestricted CTLLG-15 lymphoblasts (no), LG-6 lymphoblasts (no), 15/0 (yes), B3B7 (no)
(3) CD8-depleted (NK-like) cellsLG-15 lymphoblasts (no), LG-6 lymphoblasts (no), 15/0 (yes), B3B7 (yes)

3 Discussion

3.1 gp96 generates both adaptive and innate anti-tumor immune responses

MHC class I 15/0 tumor cells are highly tumorigenic when transplanted into MHC-identical but minor H-Ag-incompatible cloned LG-6 Xenopus. Solid 15/0 tumors appear earlier and grow more rapidly when hosts are treated with anti-CD8 or anti-NK antibody just before transplantation 21, 29. Gp96 purified from the 15/0 tumor elicits in adult Xenopus a potent anti-tumor response that can be quantified by a significant delay in the appearance and diminished size of tumors after challenge 18. The weaker, but nonetheless reproducible, reduction in the size of tumors resulting from immunization with gp96 from normal tissues suggests that in Xenopus as in mammals, gp96 generates anti-tumor responses that are both peptide specific (adaptive) and non-peptide specific (innate).

Little attention has been paid to the ability of gp96 or other hsp to elicit responses against tumors that have down-regulated MHC class Ia molecules. Such down-regulation is a common feature of many tumors and may provide a mechanism by which such tumors escape immune surveillance 33. In the present study, we have investigated, in Xenopus, the effectors involved in the gp96 response against the MHC class Ia-negative 15/0 tumor. Our results provide strong evidence that in Xenopus, gp96 generates both a potent CD8 NK-like anti-tumor response that does not depend on the source from which it is purified and a more specific anti-tumor CD8+ T cell response that requires tumor-derived gp96.

3.2 gp96 generates MHC class Ia-unrestricted anti-tumor cytotoxic CD8+ T cell effectors

Although the peptide-binding characteristics of gp96 are not as well understood as they are for hsp70, detailed studies in mice have indicated that it is the tumor peptides associated with gp96, rather than gp96 per se, that are responsible for conferring specificity of the anti-tumor immune response. No mutations resulting in immunogenic variants of gp96 were detected by sequencing the tumor-derived gp96 gene 34. Gp96 purified from either non-tumor tissues or from another tumor 9, 35 provide no or poor protection. In a few cases, specific antigenic peptides have been directly isolated and sequenced from purified gp96 3638. Further evidence of gp96′s ability to generate T cell responses against antigenic peptides comes from in vitro reconstitution experiments in mice where synthetic antigenic peptides are loaded on purified gp96. This approach has shown that gp96 functions as a cross-presenting molecule that is efficiently internalized by APC through receptor-mediated endocytosis. This cross-presentation allows the chaperoned antigenic peptides to be channeled into the APC MHC class Ia presentation pathway and, ultimately, to specifically activate CD8+ T cells 2. We have shown that synthetic antigenic peptides can be complexed in vitro to Xenopus gp96 and subsequently be represented by murine APC to activate T cells as efficiently as with mouse gp96 18.

We took advantage of the LG-6 Xenopus clone which is MHC identical with, but different from, LG-15 (the clone from which the 15/0 tumor is derived 39, 40), by multiple minor H-loci. Using this system, we have shown that Xenopus gp96 generates typical MHC-restricted CTL in vitro against LG-15 minor H-Ag 19. When LG-6 CTL were primed by LG-15 liver-derived gp96, they killed only MHC class I+ LG-15 lymphoblasts. In contrast, when LG-6 CD8+ effectors were primed with 15/0-derived gp96, they killed both normal MHC class I+ LG-15 lymphoblasts and MHC class I 15/0 targets. Several lines of evidence indicate that this killing is mediated by bonafide CD8+ T cells rather than by some kind of NK cells. In Xenopus, the CD8 marker primarily characterizes a subset of αβ T cells that, unlike NK cells, co-express a pan-T cell marker (XT-1), high levels of CD5, and CD45 28, 40, 41; this cell susbset is not detectable in frogs that had been thymectomized at early developmental stages prior to the migration of stem cells into the thymus 42. The anti-tumor CD8+ effectors assayed were more than 95% pure and did not express the 1F8 NK-associated cell marker. Furthermore, unlike typical NK-like killing, this gp96-mediated CD8+ T cell response presented extended specificity. It required both immunization with tumor-derived gp96 (i.e. immunization with either gp96 purified from minor H-Ag-disparate normal tissue or heat-denatured gp96 were ineffective) and in vitro restimulation with irradiated tumor cells or tumor-derived gp96. Moreover, these CD8+ T cell effectors did not kill either cognate LG-6 MHC class I+ lymphoblast targets or MHC class Ia NK-sensitive B3B7 tumor targets derived from the fully allogeneic MHC homozygous F strain of Xenopus. Although all these observation are consistent with the idea that anti-tumor CD8+ effectors generated by gp96 are bona fide T cells, direct evidence that the recognition involves the TCR cannot be obtained, owing to the absence of Xenopus-specific anti-TCR mAb.

Our results indicate that in vitro restimulation is critical for generating anti-tumor cytotoxic CD8+ T cell effectors, presumably because it expends the population of primed precursors as it is the case for bona fide CTL in mammals 31 and in Xenopus19, 32. MHC class Ia-unrestricted killing of the 15/0 tumor was better when 15/0 tumor rather than normal LG-15 stimulators were used, but in vitro stimulation by itself was not sufficient to generate such cytotoxicity. Interestingly, CD8+ T cell effectors generated by immunization with 15/0-derived gp96 also displayed killing activity against normal MHC class I+ LG-15 but not LG-6 or outbred lymphoblasts. This suggests that the 15/0 lymphoid tumor line (derived from a LG-15 spontaneous thymic tumor) shares some LG-15 minor H-Ag with normal LG-15 lymphocytes and that gp96 purified from the 15/0 tumor chaperones these shared LG-15 minor H-Ag and elicits a CD8+ T cell response.

The mechanism responsible for the MHC class I cell surface deficiency of 15/0 tumor cells is unknown. However, the complete absence of the MHC class I heavy chain at the protein and mRNA levels (undetectable even by reverse transcription PCR; 2022) argues that there is no defect in the assembly in the endoplasmatic reticulum or in the Ag presentation pathway, such as TAP deficiency 43. Although 15/0 cells do not express classical MHC class Ia mRNA, they do express non-classical MHC class Ib mRNA 20, 22. Whereas there is only one MHC class Ia gene per genome in Xenopus, there are at least 20 different non-classical MHC class Ib genes 22, 44 that segregate far apart from the MHC class Ia gene 45. Preliminary studies (Goyos and Robert, unpublished) indicate that 15/0 cells express a limited number of MHC class Ib genes (fewer than five subfamilies). Since β2-microglobulin mRNA is also expressed in 15/0 tumor cells (Horton and Robert, unpublished data), surface expression of MHC class Ib molecules may occur. As we have previously proposed 18, 25, expression of Xenopus MHC class Ib molecules expressed by 15/0 tumor cells could present a particular endogenous tumor peptide ligand (i.e. N-formylated) and become targets for a specific effector cell population that was previously primed by representation of the gp96-peptide complexes by APC.

3.3 gp96-mediated stimulation of anti-tumor NK-like cytotoxicity in Xenopus

In mammals, gp96 perse (i.e. independent of complexed peptides) stimulates APC to produce inflammatory cytokines (e.g. IL-1β, TNF-α; 4), chemokines (e.g. MCP-1, RANTES; 46), and increased expression of costimulatory surface molecules (e.g. CD86; 46). This immunological property may be the basis of the so-called “non-specific effect” obtained in priming/challenge experiments in which gp96 derived from normal (i.e. non-tumor) tissues affects the challenge of tumors by delaying their appearance and partially inhibiting their growth in mice 9, 10 and frogs 18.

In vivo blocking-antibody treatments suggest that in addition to CD8+ T cells 11, 12, 47, NK cells 10 are critical for hsp-mediated protective immunity to tumors. In fact, the involvement of NK and/or other cell types in hsp-mediated anti-tumor responses may have been underestimated by anti-CD8 mAb depletion, considering that some NK cells as well as NK-T, NK-CTL and some DC also express CD8 molecules 48. Results obtained with a mouse tumor engineered to secrete gp96 indicate that gp96-Ag cross-presentation may require a perforin-dependent positive feedback loop between NK and DC for both sustained NK activation and clonal CTL expansion 13. It has also been proposed that enhanced secretion of IL-12 following the interaction of hsp and APC may induce this NK anti-tumor response 1. Finally, a particularly strong peptide-independent anti-tumor effect in mice has been obtained by immunization with fibroblasts that secrete a truncated recombinant gp96 lacking the canonical peptide-binding motif 15.

Convergent evidence supports the ability of gp96 in Xenopus to stimulate anti-tumor NK effectors. The stimulation of an NK response by gp96 is revealed by the partial inhibition of killing of 15/0 targets by pre-incubating the effectors with the anti-NK mAb 1F8. Preliminary results suggest that the gp96-mediated anti-tumor response is abrogated in vivo (no delay in tumor appearance) by anti-NK mAb treatment 49. More definitive evidence comes from the fact that gp96 purified either from tumor or normal tissue generates strong CD8 cytotoxic effectors that kill MHC class Ia tumor cells (15/0 and B3B7) but not MHC class I+ allotargets.

An issue that has become important with respect to the peptide-independent immunostimulatory properties of gp96 is the possible synergistic involvement of LPS 14, 50. The fact that Xenopus is poorly responsive to LPS in vivo and in vitro23, 24 strongly suggests that the NK anti-tumor response reported in this study results from a direct effect of gp96. These findings could have implications for the ancient origins of immune surveillance in the absence of classical MHC class Ia molecules.

4 Materials and methods

4.1 Animals and tumor lines

LG-15 and LG-6 Xenopus clones 16 were obtained from our breeding colony. The 15/0 lymphoid tumor line is derived from a spontaneously arisen thymic tumor in an LG-15 frog 20, and the B3B7 thymic tumor cell line is derived from the F strain (f/f MHC haplotype). Outbred Xenopus were purchased from Xenopus I (Ann Arbor, MI).

4.2 Purification of Xenopus gp96

gp96 was purified as described 35 by 50–70% ammonium sulfate fractionation, concanavalin A-Sepharose and DEAE chromatography. Purity was assessed by SDS-PAGE and silver staining and by Western blotting.

4.3 Flow cytometry

Splenocytes (105 cells) were stained as described 19 with Xenopus anti-CD8 (AM22), anti-CD5 (2B1), pan anti-T cell (XT-1), anti-IgM (10A9), or isotype-matched control mAb, followed by fluorescein-labeled goat anti-mouse Ig [F(ab’)2; preadsorbed twice on Xenopus erythrocytes].

4.4 Immunization and skin grafting

Xenopus were injected two or three times (at 2-week intervals) with purified gp96 in 0.3 ml amphibian PBS (APBS). Outbred frogs were grafted with skin from MHC-disparate LG-15 skin according to published methods 21.

4.5 In vitro killing assay

Splenocytes from naive or immunized LG-6 frogs were harvested and restimulated (5×106 cells/ml) in vitro for 6 days with irradiated LG-15 splenocytes (2:1 ratio) in medium containing 0.25% Xenopus normal serum (medium supplementation with Xenopus serum considerably improves the survival of cytotoxic effectors; 20, 30). Cultures were further supplemented at day 3 with an IL-2-like culture rich supernatant 32. Cells were incubated with anti-CD8 mAb (IgM isotype) for 30 min on ice, and antibody-coated cells were positively selected using MACS microbeads (Miltenyi Biotec, Auburn, CA) coupled to mouse-specific anti-µ, following the manufacturer's instructions. Killing activity was determined by the “JAM” assay 51. Targets were labeled for 2 h at 26°C with 5 mCi/ml [3H]thymidine (2-day-old LG-15 normal splenic PHA-induced lymphoblast targets were labeled for 20 h), washed, incubated for 4 h at different effector-to-target cell (E/T) ratios and harvested with a 96-well harvester (Betaplate Wallac); thymidine uptake was determined by scintillation spectrometry.


We would like to thank David Albright for the expert animal husbandry and Gregory Maniero for his critical reading of the manuscript. This research was supported by RO1 grants AI-44011 and CA-76312 from the NIH and MCB-0136536 grant from the NSF.


  1. 1


  2. 2


  3. 3


  4. 4


  5. 5


  6. 6


  7. 7