Unravelling the T-cell-mediated autoimmune attack on CNS myelin in a new primate EAE model induced with MOG34–56 peptide in incomplete adjuvant


  • S. Anwar Jagessar,

    1. Department of Immunology, Biomedical Primate Research Centre, Rijswijk, The Netherlands
    2. MS Centre ErasMS, Rotterdam, The Netherlands
    Search for more papers by this author
  • Nicole Heijmans,

    1. Department of Immunology, Biomedical Primate Research Centre, Rijswijk, The Netherlands
    Search for more papers by this author
  • Erwin L. A. Blezer,

    1. Image Sciences Institute, University Medical Center Utrecht, Utrecht, The Netherlands
    Search for more papers by this author
  • Jan Bauer,

    1. Department of Neuroimmunology, Center for Brain Research, Medical University of Vienna, Vienna, Austria
    Search for more papers by this author
  • Jeroen H. Blokhuis,

    1. Department of Comparative Genetics and Refinement, Biomedical Primate Research Centre, Rijswijk, The Netherlands
    Search for more papers by this author
  • Jacqueline A. M. Wubben,

    1. Department of Immunology, Biomedical Primate Research Centre, Rijswijk, The Netherlands
    Search for more papers by this author
  • Jan W. Drijfhout,

    1. Department of Immunohematology and Blood Transfusion, Leiden University Medical Center, Leiden, The Netherlands
    Search for more papers by this author
  • Peter J. van den Elsen,

    1. Department of Immunohematology and Blood Transfusion, Leiden University Medical Center, Leiden, The Netherlands
    2. Department of Pathology, VU University Medical Center, Amsterdam, The Netherlands
    Search for more papers by this author
  • Jon D. Laman,

    1. MS Centre ErasMS, Rotterdam, The Netherlands
    2. Department of Immunology, Erasmus MC, University medical Center, Rotterdam, The Netherlands
    Search for more papers by this author
  • Bert A.'t Hart

    Corresponding author
    1. Department of Immunology, Biomedical Primate Research Centre, Rijswijk, The Netherlands
    2. MS Centre ErasMS, Rotterdam, The Netherlands
    3. Department of Medical Physiology, University Medical Centre Groningen, The Netherlands
    • Department of Immunobiology, Biomedical Primate Research Centre, PO Box 3306, 2280 GH Rijswijk, The Netherlands Fax: +31-15-2842600

    Search for more papers by this author


Induction of experimental autoimmune encephalomyelitis (EAE) has been documented in common marmosets using peptide 34–56 from human myelin/oligodendrocyte glycoprotein (MOG34-56) in incomplete Freund's adjuvant (IFA). Here, we report that this EAE model is associated with widespread demyelination of grey and white matter. We performed an in-depth analysis of the specificity, MHC restriction and functions of the activated T cells in the model, which likely cause EAE in an autoantibody-independent manner. T-cell lines isolated from blood and lymphoid organs of animals immunized with MOG34–56 displayed high production of IL-17A and specific lysis of MOG34–56-pulsed EBV B-lymphoblastoid cells as typical hallmarks. Cytotoxicity was directed at the epitope MOG40–48 presented by the non-classical MHC class Ib allele Caja-E, which is orthologue to HLA-E and is expressed in non-inflamed brain. In vivo activated T cells identified by flow cytometry in cultures with MOG34–56, comprised CD4+CD56+ and CD4+CD8+CD56+ T cells. Furthermore, phenotypical analysis showed that CD4+CD8+CD56+ T cells also expressed CD27, but CD16, CD45RO, CD28 and CCR7 were absent. These results show that, in the MOG34–56/IFA marmoset EAE model, a Caja-E-restricted population of autoreactive cytotoxic T cells plays a key role in the process of demyelination in the grey and white matter.


Multiple sclerosis (MS) is a chronic inflammatory disease of the human central nervous system (CNS) of unknown aetiology. The pathological hallmark of MS is the lesion. Lesions are regions of usually focal demyelination of variable size localized in the grey and/or white matter of the brain and spinal cord, formed by a combined cellular and humoral autoimmune attack. CNS-targeting autoimmune reactions are thought to be induced as a response to infection (response-to-infection paradigm) 1, although the pathogen(s) that elicit this pathogenic process in MS has not been identified. We have proposed a response-to-damage paradigm for MS, based on the work in a unique non-human primate model of MS, experimental autoimmune encephalomyelitis (EAE) in common marmosets (Callithrix jacchus), 2. The new concept postulates that autoimmunity in MS patients is caused by a genetically predisposed hyper-response to myelin antigens released from damaged white matter due to an unknown antecedent event. We showed that the most important anti-myelin reactivity for the induction of neurological deficit is mediated by antigen-experienced T-cells specific for peptides 34–56 of myelin/oligodendrocyte glycoprotein (MOG34–56) 3. Subsequently it was demonstrated that these cells could be activated in vivo by immunization of marmosets with MOG34–56 in incomplete Freund's adjuvant (IFA) 4.

A peptide in IFA emulsion is a more common formulation for the in vivo activation of antigen-experienced T-cells than for autoreactive T-cells as bacterial ligands of innate antigen receptors are usually required. It is noteworthy that MOG34–56-specific CD4+ T cells identified in MS patients also display antigen-experienced characteristics 5. The aim of the current study was therefore a more in-depth characterization of the T cells that are activated in this new model, with a focus on specificity, MHC restriction and functional analysis.

Despite the immunological proximity of marmosets and humans, some fundamental differences exist at the level of the major histocompatibility complex, a polymorphic region encoding the molecules that present antigens to T-cell receptors. The MHC class I region of marmosets lacks the genes encoding the equivalents of the human classical MHC class I molecules HLA-A, -B and -C, but a polymorphic Caja-G (multiple alleles) and an oligomorphic Caja-E locus (two alleles) have been identified 6, 7. Hence, we hypothesized that peptides processed from MOG34–56 are presented by the non-classical MHC class Ib molecules Caja-G or Caja-E to antigen-specific cytotoxic T lymphocytes.


Clinical and pathological aspects of EAE induced with equation image/IFA

Five unrelated marmoset monkeys were immunized with MOG34–56 in IFA at 28 day intervals until clinically evident EAE (score ≥2.0) developed. Clinical scores (Fig. 1A) show that 4 out of 5 animals developed clinically evident EAE, whereas in one monkey (Cj1) only mild EAE symptoms were observed. Data of all animals are pooled and presented in a survival curve (Fig. 1B) showing time to EAE score 2.0 and time to necropsy. High-resolution T2-weighted and magnetization transfer ratio images showed in four monkeys the presence of white matter lesions (Fig. 1C), which were characterized with respect to the severity of inflammation and demyelination 8. A detailed histological characterization of the white matter lesions in this model has been published elsewhere 4. While most lesions were located in the white matter, also cortical grey matter lesions were detected in Cj3 and Cj4.

Figure 1.

Clinical course and pathology of MOG34–56/IFA EAE model in marmosets. (A) Five unrelated monkeys were immunized at 28 day intervals (arrowheads) with 100 μg MOG34–56 in IFA. Monkey Cj1 was re-challenged only once because the presence of a short episode of blindness suggesting optic neuritis (EAE score 2.0) showed that the disease was already ongoing. Depicted are clinical scores (solid line, left y-axes) and % body weight loss relative to day 0 (dotted line, right y-axes). (B) Survival curves show the disease free survival (time until the animals developed EAE score 2.0) and survival time until the day of necropsy. PSD, post-sensitization day. (C) Magnetic resonance images of formalin fixed brains from 4 of the 5 monkeys. Left: T2-weighted (T2W) images; middle: magnetization transfer ratio images (MTR); right: WAIR images. White squares circumscribe the same white matter lesion in the three types of scans. White arrows in the WAIR images of Cj3 and Cj4 points to cortical grey matter lesions. (D) Snap-frozen brain sections of Cj4 were immunostained as representative example of T-cell localization within early lesions. The picture shows a small perivascular T-cell infiltrate: CD3 (upper left), CD4 (upper right) and CD8 (bottom left). Laminin (bottom right) shows that infiltrated MNCs are mainly localized in the perivascular space. (400×, bar=50 μm). (E) Formalin-fixed brain section of Cj3 was immunostained for proteolipid protein (PLP), showing widespread demyelination in the white matter (I, 4×, bar=200 μm) and cortical grey matter (III, 25×, bar=500 μm). The square in (III) is enlarged in (IV, 250×, bar=50 μM) demonstrating that activated macrophages are undetectable in grey matter, although macrophages containing degraded PLP products were detected in white matter. Few infiltrated T cells expressing Granzyme B were found in the white matter of the corpus callosum (II, 400×, bar=20 μm). Data from monkey Cj3 is shown as a representative.

T-cell localization within white matter lesion

Serum antibodies binding were detected to MOG34–56, but not to recombinant human myelin/oligodendrocyte glycoprotein (rhMOG) (data not shown). Hence, we hypothesized that lesions are induced by MOG34–56-specific T-cells. To examine the presence of T cells in early brain lesion stages, these were stained for CD3, CD4, CD8 and for laminin, a component of the basal membranes lining the perivascular Virchow–Robin space (Fig. 1D). All animals displayed small to moderately sized perivascular cuffs containing CD3+, CD4+ and CD8+ T cells. CD4+ T cells were more frequently detected than CD8+ cells. Laminin staining confirmed that mononuclear cell (MNC) infiltrates were largely confined to perivascular spaces whereas some cells seemed to have passed the glia limitans. Figure 1E shows a typical large sized lesion in the white matter (Fig. 1EI) and in the grey matter (Fig. 1EIII). Figure 1EII shows that T cells infiltrating the white matter express granzyme B, an established marker of cytotoxic T-cells. Figure 1EIV shows the absence of activated MRP-14 macrophages in grey matter lesions, although macrophages containing myelin degradation products were abundant in the white matter lesion.

Analysis of T-cell reactivity against equation image

End-stage proliferation

At necropsy MNCs were isolated from blood, spleen, axillary (A), inguinal (I), lumbar (L) and cervical (C) lymph nodes (LNs) and tested for proliferation against a panel of overlapping MOG peptides and rhMOG. Proliferation was only detected against the immunizing MOG34–56 peptide (Fig. 2A).

Figure 2.

Analysis of T cells after ex vivo stimulation. MNCs were isolated from spleen, ALNs, ILNs, LLNs and CLNs. (A) Proliferation of T cells against MOG34–56 was determined by 3H-thymidine incorporation during final 18 h culture. Data are expressed as stimulation index (SI) relative to unstimulated cells. SI≥2.0 is considered positive. (B) T-cell lines against MOG34–56 were generated from spleen MNCs and phenotyped for CD3, CD4, CD8 and CD56 expression. The cells were further characterized on basis of CD27, CD28, CD45RO, and CCR7 expression. The T-cell lines were gated for CD3+ (55±17.3% of vital cells); remaining 45% were CD3 low or negative. The CD3+ T cells were 26±5.3% CD4+, 23±5.8% CD8+ and 36±8.2% CD4+CD8+. Percentages (mean±SD) are given of four pooled T-cell lines. Data from one (Cj1) out of the four T-cell lines are shown; staining patterns were analysed in two independent experiments. Numbers in each quadrant represent the percentage of each subpopulation. (C) IL-17A production was measured in PBMCs after 48 h stimulation with MOG34–56 or no antigen stimulation. Production of IFN-γ and IL-10 was below the detection level and is therefore not shown. PSD, post-sensitization day. *p<0.05, Mann–Whitney U test, compared with data of not stimulated cells. The results are pooled of five animals and reported as mean±SEM. Experiment is performed only once since we had small volume of culture supernatant.

Cross-sectional comparison of proliferation in blood and lymphoid organs

Figure 2A summarizes the reactivity profiles against MOG34–56 of all 5 monkeys. In blood MNC proliferation against MOG34–56 was around the cut-off level of stimulation index (SI) 2.0. Variable proliferation was observed in ALNs, ILNs and LLNs. Only in one animal, Cj1, positive response was detected in the CLN. CLN from monkey Cj3 was not tested because too few cells could be isolated.

Cell lines

T-cell lines against MOG34–56 were generated from spleen and ALN assuming that these bulk T-cells reflect the profile of in vivo activated T cells. These cells were phenotyped using monoclonal antibodies with confirmed cross-reactivity with marmosets. Based on CD4 and CD8 expression three subsets of CD3+ T cells were discerned: CD4 and CD8 single-positive populations and a CD4/CD8 double-positive population (Fig. 2B). The three subsets were characterized for surface expression of additional markers. The main proliferating CD3+ subset was previously found to be double positive for CD4 and CD8 and also expressed CD56, but not CD16 4. The CD4+CD8+CD56+ subset (right column of Fig. 2B) was negative for CD45RO, CD28 and CCR7 but positive for CD27. Note that these markers are differently expressed in the CD3+CD8+ subset.

Longitudinal cytokine profiles

Blood MNCs collected at 14 days interval were cultured with or without MOG34–56. The small blood volume available per time point (1 mL) precludes large-scale analyses. We therefore chose to determine besides proliferation levels of three selected cytokines in culture supernatants: IL-17A (Th17), IFN-γ (Th1) and IL-10 (anti-inflammatory). Cytokine profile presented in Fig. 2C shows that more IL-17A was produced when MNCs were stimulated with MOG34–56 versus no antigen stimulation. Data are only shown of IL-17A as IFN-γ and IL-10 production was below the detection level (see Supporting Information Fig. 2).

Target cell specificity of the cytolytic T cells

The cytolytic activity of MOG34–56-specific T-cell lines from spleen and ALN was tested using EBV B-LCL as target cells. Figure 3A shows the results from two cell lines, which lysed autologous B-LCL pulsed with the specific peptide MOG34–56. Figure 3B shows that the peptide does not need to be presented by an autologous B-LCL as the T cells also recognized and killed MOG34–56-pulsed B-LCL from an unrelated donor (allogeneic B-LCL). This suggests that the cytotoxic T cells recognize MOG34–56-derived peptide by invariant MHC molecules. The relatively high background lysis is unexplained. We assume that this is due to background T-cell immunity against EBV-like viruses with which marmosets are naturally infected 9.

Figure 3.

Specific cytotoxicity of T-cell lines. MOG34–56-specific T-cell lines (E: effector cells) isolated from the spleens of two EAE affected marmoset were tested for specific cytotoxicity against peptide-pulsed EBV B-LCLs (T: target cells) in a 51Chromium release cytotoxicity assay. (A) Autologous B-LCLs were tested without (dotted line) or with (solid lines) MOG34–56 peptide. Specific lysis of target cells is expressed as % killing. (B) Allogeneic B-LCLs pulsed with MOG34–56 were killed by MOG34–56 specific T-cell lines from unrelated marmosets. Data are expressed as % killing compared to negative control. (C) Autologous B-LCLs were pulsed with a panel of 9-mer overlapping MOG peptides derived from MOG34–56. Cytotoxicity was tested at the same three E:T ratios as in (A, B), but only data obtained at 16:1 ratio are shown. Negative control comprised non-pulsed target cells or target cells pulsed with an irrelevant protein (OVA). Baseline shows background killing (dotted line). Data shown are one representative of two independent experiments with similar results.

In the next experiment we determined the epitope in the MOG34–56 peptide that is presented by the B-LCL to cytotoxic T cells. To this end, 51Chromium-labelled B-LCL were pulsed with 9-mer synthetic MOG peptides overlapping by six amino acids and next exposed to autologous MOG34–56 specific effector T cells. Figure 3C shows enhanced cytotoxicity towards B-LCL pulsed with MOG40–48 and MOG46–54, indicating that these are candidate CTL epitopes in MOG34–56.

MHC-Caja class I expression in the marmoset brain

Classical MHC class Ia molecules equivalent to HLA-A, HLA-B and HLA-C are not encoded in the MHC genomic region of the marmoset. However, equivalents of the non-classical MHC class Ib molecules, HLA-E and HLA-G are expressed 6, 7, 10. According to the consensus nomenclature of primate MHC systems 11, these are indicated with the acronym Caja i.e. Caja-E and Caja-G. Whereas Caja-G is polymorphic and seems to function as a classical MHC Ia molecule, Caja-E is oligomorphic containing only 2 alleles, differing only at amino acid position 107 outside the peptide binding cleft. Expression of transcripts for both MHC class I genes in marmoset brain has been published 11 and is confirmed in Fig. 4. The figure showed the expression of Caja-E transcripts in EBV-induced B-LCL for comparison. Thus, we hypothesize that the cytotoxic activity of MOG34–56-induced T cells is directed against MOG34–56 epitope presented via Caja-E molecules.

Figure 4.

MHC class I gene transcripts in marmoset brain and B-LCLs. Specific primer sets for Caja-G and Caja-E (see Materials and methods section) were used to verify, using RT-PCR, that these non-classical MHC molecules are expressed in the brain and by the B-cell lines. The housekeeping gene G3PDH was used as internal control. Arrow indicates Caja-E transcript in the brain. bp, base pairs. Data shown are representative of two independent experiments.

Mapping of HLA-E-binding peptides on the MOG extracellular domain

We have used a published algorithm 12 (Fig. 5A) to identify candidate HLA-E-binding motifs within previously identified T-cell epitopes in the human and marmoset MOG extracellular domains aa 1–125 (Fig. 5A) 13 (Fig. 5B). As can be seen in Fig. 5C the peptides MOG1–22, MOG94–116 and the overlapping peptides MOG64–86/74–96, which encompass dominant T-cell epitopes for rhMOG-immunized marmosets 14, contain putative HLA-E-binding sequences. Remarkably, an HLA-E binding sequence was not detected in MOG34–56, although according to an algorithm published by Stevens et al. 15 peptides 42–50 may bind to HLA-E as position 2 might be less important for binding.

Figure 5.

Delineation of putative HLA-E-binding peptides in rhMOG. (A) Extracellular domain of the human and marmoset MOG sequence, aa 1–120. (B) Algorithm for delineation of HLA-E-binding peptides in red and underlined 12. Position 2 and 9 are the main HLA-E contact positions. Positions 3, 6, and 7 (in blue) are minor residues. (C) Previously identified dominant T-cell epitopes (left) in the rhMOG-induced EAE model contain putative HLA-E binding sequences (right).

The four putative HLA-E binding sequences were synthesized as 9-mer MOG peptides, 13–21, 42–50, 75–83 and 105–113 and tested for binding to HLA-E transfectants of K562 cells using thermostable expression at 37°C as read-out. In a direct binding assay (Fig. 6A) only HLA-E binding of MOG13–21 and the positive control peptide (HCV core peptides 35–44) was observed 16.

Figure 6.

Binding of MOG peptides to HLA-E-expressing K562 cells. (A) K562 cells transfected with HLA-Eg were incubated o/n at 26°C with or without a dose titration of the indicated peptides. Thermostable expression of HLA-E at 37°C was visualized by anti-MHC class I antibody (W6/32) staining. Data are expressed as mean fluorescence intensity (MFI) and are corrected for background binding. Specificity controls in this assay were K562/HLA-Eg cells incubated without peptide and K562 cells transfected with mock plasmid (pCDNA3). (B) HLA-E or mock-transfected K562 cell lines were incubated for 4 h at 26°C with a dose titration of the indicated peptides. Subsequently 5 μM of a fluorescently labelled detector peptide was added and incubated overnight at 26°C. In the left panel, HLA-G leader peptide was used as detector peptide. In the right panel, HCVcore35–44 was used as detector peptide. Binding of detector peptides was measured by flow cytometry and data are expressed in MFI. (C) From the same experiment as described in (B), MHC class I expression was measured by staining with W6/32. One representative experiment of three is shown.

We hypothesized that binding of the three other MOG peptides to HLA-E might be too weak to form thermostable complexes at 37°C. This possibility was tested in a more sensitive indirect binding assay. As a positive control we used unlabelled HCV core 35–44 peptide. Figure 6B shows that at the highest dose (300 μM), two peptides, MOG42–50 and HCV core peptide, induced increased binding of the HLA-G leader indicator peptide. Increased binding of the HCVcore35–44 detector peptide was not observed. Staining with W6/32 monoclonal antibody confirmed slightly elevated expression of HLA-E at the highest concentration of MOG42–50 (Fig. 6C). The internal control of this sensitive indirect assay showed complete blocking of binding of fluorescent HCVcore35–44 detector peptide by unlabelled HCVcore35–44, whereas binding of the HLA-G leader detector peptide was unaffected.

In conclusion, the extracellular domain of MOG contains several putative binding peptides localized within previously identified T-cell epitopes 14.

Caja-E binding of putative epitopes within equation image

Just like HLA-E, Caja-E has two known alleles, Caja-E*01 and Caja–E*02, which differ at position 107 of the α chain by a G or T respectively, which is outside the peptide-binding cleft. Whether these two alleles have similar functional differences as their human counterparts 17 or differ in the binding of viral peptides 18 is not known.

Based on the encouraging HLA-E-binding data we cloned the Caja-E*01 allele from EBV-induced B-LCL for expression in K562 cells. Successful transfection was verified by PCR and surface expression of the gene product at 26°C was confirmed by staining with W6/32 mAb (Fig. 7A).

Figure 7.

Binding of MOG peptides with Caja-E on K562 cells. Caja-E*01 was cloned from B-LCLs and subsequently transfected into K562 cells. (A) Transfection of Caja-E was confirmed by W6/32 staining. High expression of Caja-E molecules was observed at 26°C (black line), whereas these molecules desintegrated at 37°C (grey line). (B) Caja-E-transfected K562 cells were incubated o/n at 26°C with a dose titration (0–100 μM) of overlapping 9-mer peptides derived from the MOG34–56 sequence. Thermostable expression of Caja-E at 37°C was confirmed by W6/32 staining. One representative experiment out of 4 is shown.

The presence of Caja-E binding epitopes in MOG34–56 was analysed using the direct binding assay of 9-mer MOG peptides overlapping by 6 (Table 1). Figure 7B shows binding of MOG40–48 being the specific epitope of the MOG34–56-induced cytotoxic T-lymphocytes.

Table 1. Synthetic peptides based on the human MOG sequence
PeptideAmino acid sequence
  • a)

    a) In the marmoset and mouse MOG sequences, position 42 is occupied by S, whereas in human MOG this position is occupied by P.

  • b)

    b) Peptides labelled with the fluorescent dye fluorescein.

MOG 40–48      YRPPFSRVV
MOG 42–50        SPFSRVVHLa)
MOG 43–51        PFSRVVHLY
MOG 46–54          RVVHLYRNG
MOG 49–57            HLYRNGKDQ
MOG 52–60              RNGKDQDGD


Marmosets immunized with MOG34–56 in IFA developed EAE associated with widespread demyelination of CNS white matter 4. Demyelination in this model seems directly induced by MOG34–56-induced T cells independent of MOG protein-binding autoantibodies, which were undetectable in the model. To unravel the underlying mechanism, five unrelated marmosets were immunized with MOG34–56 in IFA inducing progressive EAE associated with extensive demyelination of brain white matter and cortical grey matter (Fig. 1D). This situation illustrates the pathological similarity of the marmoset EAE model with MS 19.

Using immunohistochemistry CD4+ and CD8+ infiltrated T cells were detected within and around perivascular spaces of early-stage lesions. MOG34–56-specific T-cell lines generated from secondary lymphoid organs were characterized to examine their nature and possible origin (Fig. 2B). Using flow cytometry, in vivo activated MOG34–56 reactive T cells were defined in three subsets, namely CD3+CD4+, CD3+CD8+ and CD3+CD4+CD8+. Of particular interest is the co-expression of CD56, which is a marker of NK cells that defines CD4+ and CD8+ T cells with cytotoxic activity towards oligodendrocytes in MS 20, 21. The CD3+CD4+CD56+ and CD3+CD4+CD8+CD56+ populations were for a small proportion CD45RO+ and expressed CD27+, but were negative for CD28 or CCR7. The phenotype of these T cells resembles Natural Killer–Cytotoxic T-lymphocytes (NK-CTL) 22. The CD3+CD8+CD56+ T cells had a somewhat different phenotype as only a relatively small proportion expressed CD27. It is presently unclear which of these three phenotypes mediates IL-17A production or peptide-specific cytolytic activity by the MOG34–56 reactive T cells. Attempts to elucidate this have failed thus far as the T cells kill the EBV-induced B-LCL, which were used as APC for generating lines and clones.

Next, we determined the fine-specificity and MHC restriction of the cytotoxic T cells. To distinguish between the two possible options for MHC class I restricted CTL activity, i.e. via polymorphic Caja-G or oligomorphic Caja-E molecules, we first tested whether a certain T-cell line could lyse B-LCL from unrelated monkeys. This was indeed the case indicating that MHC class I molecules presenting the peptide epitope to the CTL are shared by B-LCL from unrelated monkeys, making Caja-E the most likely candidate. Further analysis supported this assumption as the 9-mer peptide, MOG40–48, which was identified as the specific epitope of the cytolytic cells, was the only MOG34–56 peptide that stably bound Caja-E. It has been shown in mice that the MOG sequence 40–48 is critical for EAE induction, and is required as the minimal epitope for stimulating MOG35–55-specific T cells 23. These findings support the conclusion that the cytotoxic activity of MOG34–56 induced T cells in the marmoset targets MOG40–48/Caja-E complexes (Fig. 8).

Figure 8.

Proposed mechanism of demyelination in the MOG34–56/IFA marmoset EAE model. MOG40–48 peptide presented by Caja-E molecules expressed on the myelin sheath or oligodendracyte (ODC) is recognized by cytotoxic T cells. IL-17A was the only cytokine that was produced in clearly detectable amounts. Based on various similarities with anti-CMV effector memory CTLs found in humans, we propose to call the autoreactive T cells NK-CTLs 36. According to previously reported experiments 4, cytotoxicity was dependent on the exocytosis of cytotoxic granules. Recognition of the MOG40–48:Caja-E complex by NK-CTLs results in injury to the ODCs or myelin sheath.

Several characteristics of the autoreactive T cells suspected for mediating demyelination in this new model render them capable to induce lesions. First, they share preferential production of IL-17A with Th17 cells, which infiltrate non-inflamed brain via the chorioid plexus 24. Indeed, the first site where in the marmoset EAE model CNS inflammation can be detected on contrast-enhanced T1-weighted magnetic resonance images is around the dorsal horns of the lateral ventricles where the chorioid plexus is located (data not shown). Second, antigen-specific cytolytic of oligodendrocyte (ODC) by MOG34–56 activated T cells could be the cause of demyelination 20. However, the model lacks an obvious source of IFN-γ that is needed to induce classical MHC class I molecules on ODC 25. The autoreactive T-cells themselves do not produce (detectable amounts of) IFN-γ and IFA does not contain innate antigen receptor ligands that could induce this cytokine via activation of myeloid cells are absent in IFA 4.

It has been well established that cells with low expression of classical MHC class Ia molecules, such as ODC, are exposed to cytolysis by NK cells. One of the mechanisms to avoid NK cells attack is HLA-E expression. When engaged with certain QdM peptides such as in the HLA-G leader, HLA-E relays inhibitory signals to NK cells via NKG2A/CD94. This system is used by herpesviruses, such as cytomegalovirus (CMV) for immune evasion 26. However, HLA-E restricted CD3+CD8+CD56+CD27CCR7 T cells are present in the human repertoire which can kill CMV-infected cells via the recognition of viral peptides (e.g. UL40) presented by HLA-E 22.

It is unclear at this stage to which extent this situation also occurs in our marmosets. Similar to the rhesus macaque, the common marmoset is naturally infected with herpesviruses related to human CMV 27. We observed that CD8+ T cells specific for MOG34–56 and MOG40–48 in the rhesus monkey crossreact with an almost identical peptide from the immunodominant UL86 antigen of human CMV 28. Both peptides bind to Caja-E. However, this could not be proven in the marmosets, as specific reagents for testing the CMV status in marmosets are not available.

In conclusion, we postulate that the T cells that induce demyelination in the MOG34–56/IFA EAE model may be recruited from the repertoire of effector memory cells that controls latent CMV infection.

Materials and methods

Antigens and peptides

Synthetic MOG peptides used for immunization and for in vitro experiments were purchased from Cambridge Research Biochemicals Limited (Cleveland, UK) or were produced in house (JWD, Leiden University Medical Center, Leiden, The Netherlands). The amino acid sequences of all peptides used in this study are summarized in Table 1. rhMOG, residues 1–125, purified from Escherichia coli was produced in the BPRC laboratory as previously described 29, 30.

Induction of EAE and post-mortem examination

EAE was induced with 100 μg of MOG34–56 emulsified in IFA. As previously described the inoculum was injected into the dorsal skin at the inguinal and axillary regions 4. All animals were daily monitored for neurological symptoms using a standard scoring system, as described before 4, 14. Briefly, 0=no clinical signs; 0.5=apathy, loss of appetite, altered walking pattern without ataxia; 1=lethargy, anorexia, loss of tail tonus, tremor; 2=ataxia, optic disease; 2.5=paraparesis or monoparesis, sensory loss; 3=paraplegia or hemiplegia; 4=quadriplegia; 5=spontaneous death due to EAE. The clinical endpoint for each monkey was EAE score 3. Monkeys reaching this score were first deeply sedated by intramuscular injection of Alfaxan (10 mg/kg) (Vétoquinol S.A., Magny-Vernois, France). After collection of the maximum venous blood volume, they were euthanized by infusion of sodium pentobarbital (Euthesate®; Apharmo, Duiven, The Netherlands).

At necropsy the brain and spinal cord were removed. Secondary lymphoid organs aseptically removed for preparation of MNCs were: ALNs, ILNs, LLNs, CLNs and spleen.

According to the Dutch law all experimental procedures were reviewed and approved by institute's animal experiment committee.

Magnetic resonance imaging

High-contrast post-mortem magnetic resonance images (MRIs) were recorded of one cerebral hemisphere on a 9.4 T horizontal bore NMR spectrometer (Varian, Palo Alto, CA, USA), equipped with a quadrature coil (RAPID, Biomedical, Rimpar, Germany) as previously described 31. In addition, an inversion recovery experiment was performed in which the signal from the white matter was suppressed, white matter attenuated inversion recovery (WAIR), enabling the visualization of grey matter lesions (fast spin echo, echo train length=8, echo spacing=8.93 ms, effective TE=17.86 ms; TR=4050 ms; inversion time for white matter suppression was empirically determined per sample and was approximately 350–375 ms).

T-cell proliferation

MNCs were prepared every 2 wk from heparinized blood (PBMCs) or at necropsy from spleen and lymph nodes; ALNs, ILNs, LLNs and CLNs. PBMCs and MNCs were tested for proliferation against a panel of overlapping MOG peptides (10 μg/mL) and rhMOG (10 μg/mL). 32. Proliferation is expressed as SI, being 3H-thymidin incorporation in the stimulated versus unstimulated cultures. SI values above 2.0 were considered positive.

Flow cytometry

T-cell lines against MOG34–56 were generated from spleen and ALNs as previously described 14. Part of the cells were phenotyped with monoclonal antibodies that were pre-selected for cross-reaction with the marmoset 33: anti-CD4 clone MT310 (Dako, Glostrup, Denmark), anti-CD8 clone LT8 (Serotec, Düsseldorf, Germany), anti-CD3 clone SP34-2, anti-CD56 clone NCAM16-2, anti-CD16 clone 3G8, CD27 clone M-T721 (BD Biosciences, San Diego, CA, USA), CD28 clone CD28.2, CD45RO clone UCHL1, CCR7 clone 150503 (R&D systems, Oxon, UK). For detection of HLA-E or Caja-E expression the anti-MHC class I antibody clone W6/32 (DAKO) was used. Flow cytometric analysis was performed on an FACS LSRII flow cytometer using FACSDiva software 5.0 (BD Biosciences) or FACSCalibur using CellQuest.

Cytotoxicity assay

Specific T-cell lines against MOG34–56 generated from spleen and ALNs were used as effector cells in a cytotoxicity assay. Autologous or allogeneic 51Chromium-labelled B-LCLs were pulsed for at least 1 h at 37°C with MOG peptides and used as target cells as previously described 4, 14.

Cytokine profiling

Supernatants of PBMCs were collected after 48 h stimulation with rhMOG or a panel of overlapping MOG peptides. Supernatants were assayed for the presence of cytokines using commercial ELISA kits following the manufacturer's instruction or as described previously 4.

RNA extraction and RT-PCR

Total RNA was extracted from PBMCs and snap-frozen brain material using the RNeasy mini kit from Qiagen (Hilden, Germany). Isolated RNA was used for reverse transcriptase-PCR on an aliquot (3 μL) of each sample for 25 cycles with Caja-E and Caja-G-specific primers: Caja-E; FW 5′-GCCAGGGACACCGCACAGAG-3′, RV 5′-AGAAACCCAGGGCCCAGCAT-3′. Caja-G; FW 5′-GCAAGCTTATGACGGTCATGGCTCCCCGAA-3′, RV 5′-CAAGCCGTGAGAGACACATCAGAGCCCTG-3′. As an internal control the housekeeping gene G3PDH was used; FW 5′-TGAAGGTCGGAGTCAACGGATTTGGT-3′. RV 5′ CATGTGGGCCATGAGGTCCACCAC-3′. See manufacture of AccesQuick™ RT-PCR System (Promega, Madison, WI, USA) for RT-PCR conditions. PCR products were analysed by electrophoresis on a 1.2% agarose gel. Bands of interest were purified according to the Qiaquick gel extraction kit (Qiagen, Hilden, Germany) and sequenced on an ABI 3130xl genetic analyser (Applied Biosystems, Foster City, CA, USA) as described 34.

Caja-E transfection into K562

K562 cells, a human erythroleukemia cell line which lacks MHC class I and II, was transfected with Caja-E*01 (FW 5′-CACCATGTAGCCCCGAAGCCTCCTCTTGCTG-3′, RV 5′-TCAGACTTTACAACCTGTGAGAGACAC-3′) according to the manufacturers instruction using the pcDNA™3.1 directional TOPO® expression kit (Invitrogen, Carlsbad, CA, USA). Briefly, 1 μL PCR product was blunt-end ligated into the pcDNA™3.1/V5-His-TOPO® vector and incubated for 5 min at RT. As a positive control vector pcDNA™3.1D/V5-Hisd/lacZ was used. Subsequently, the construct was transduced into One Shot® TOP10 chemically competent E. coli at 42°C for 30 s. The transduced cells were plated on LB selection plates containing 50 μg/mL ampicillin and incubated overnight (o/n) at 37°C. The following day, a minimum of 24 clones were scaled up in LB+ampicillin medium o/n at 37°C. Plasmid DNA was extracted from E. coli and purified based on the alkaline lysis method (QIAprep spin miniprep kit). To confirm that Caja-E was cloned in the correct orientation, the construct was sequenced (ABI 3130xl genetic analyser). Constructs containing correctly inserted genes were scaled up to large quantity and plasmid DNA was isolated using the QIAGEN EndoFree plasmid maxi kit. Plasmids were transfected into K562 cells by electroporation using the AMAXA Nucleofector kit V for cell lines (Lonza Cologne AG, Cologne, Germany) and transfected cells were grown in IMDM medium (Gibco, Glasgow, UK) in the presence of 200 μg/mL G418 (Gibco). After one day the cells were incubated with anti-MHC class I antibody (W6/32) conjugated to PE, and positive selection was performed with anti-PE MACS separation beads (Miltenyi Biotec, Bergisch Gladbach, Germany). Limiting dilution assay (1 cell/well in 96-well plate) was performed to select the best-transformed cells. Expression of Caja-E on the K562 cells was confirmed with anti-MHC class I antibody, (W6/32; DAKO) and anti-HLA-E clone MEM/07 (Abcam, Cambridge, UK).

K562 cells transfected with HLA-E*01033 (HLA-Eg) and the mock-transfectants, used as the negative control (pcDNA3), were kindly provided by Prof. Dr. H. Weiss (Department of Biology II, Ludwig-Maximilians-Universität München, Germany).

HLA-E/Caja-E binding assays

Direct assay

K562 Caja-E or HLA-E transfectants were incubated overnight at 26°C with a selected panel of peptides (see Table 1). Then the cells were incubated for 1 h at 37°C to allow degradation of unoccupied MHC class I molecules. Residual expression of MHC class I molecules was visualized by staining with PE-labelled MHC class I antibody, clone W6/32 (DAKO). Binding of W6/32 was measured by FACS LSRII or FACSCalibur flow cytometer.

Indirect assay

K562 Caja-E or HLA-E transfectants were incubated for 4 h at 26°C with a dose range of unlabelled peptides (3 μm–300 μm). Subsequently, a fixed concentration (5 μM) of fluorescently labelled indicator peptides with known binding capacity to HLA-E (HLA-G leader 12 or HCV35–44 16 was added to the cells and incubated overnight at 26°C. Subsequently, cells were incubated for 1 h at 37°C to allow degradation of unoccupied MHC class I molecules. Residual expression of MHC class I molecules was tested by measuring the amount of bound fluorescently labelled peptide and with PE-labelled W6/32 staining.

Immunohistochemistry and histology

Cryosections of 6 μm thick were made from snap-frozen brains and immuno-labelled for specific markers as described previously. Slides were stained for CD3 (clone sp34-2, BD Biosciences), CD4 (home-made mix of OKT4, OKT4a, RIV6, RIV7 and MT-310), CD8-biotin (clone LT8, Serotec, Düsseldorf, Germany) and laminin (DAKO). Formalin-fixed brains were used for histological examination as described 32, 35.

Statistical analysis

Where possible data were analysed using the Mann–Whitney U test; p-values <0.05 are considered to be significant.


The authors thank the Animal Science Department for biotechnical assistance and the Department Comparative Genetics & Refinement (BPRC) for help with molecular techniques. This work was supported by an internal grant of the Biomedical Primate Research Centre and by the Netherlands Organization for Scientific Research (NWO) funding the 9.4 T horizontal bore NMR spectrometer.

Conflict of interest: The authors declare no financial or commercial conflict of interest.