OUP user menu

The regulatory role of natural killer cells in multiple sclerosis

Kazuya Takahashi , Toshimasa Aranami , Masumi Endoh , Sachiko Miyake , Takashi Yamamura
DOI: http://dx.doi.org/10.1093/brain/awh219 1917-1927 First published online: 30 June 2004


Multiple sclerosis is a chronic demyelinating disease of presumed autoimmune pathogenesis. The patients with multiple sclerosis typically shows alternating relapse and remission in the early stage of illness. We previously found that in the majority of multiple sclerosis patients in a state of remission, natural killer (NK) cells contain unusually high frequencies of the cells expressing CD95 (Fas) on their surface (>36.0%). Here we report that in such ‘CD95+ NK-high’ patients, NK cells may actively suppress potentially pathogenic autoimmune T cells that can mediate the inflammatory responses in the CNS. Using peripheral blood mononuclear cells (PBMCs) derived from ‘CD95+ NK-high’ or ‘CD95+ NK-low’ multiple sclerosis in a state of remission, we studied the effect of NK cell depletion on the memory T cell response to myelin basic protein (MBP), a major target antigen of multiple sclerosis. When we stimulated PBMCs of the ‘CD95+ NK-high’ multiple sclerosis after depleting CD56+ NK cells, a significant proportion of CD4+ T cells (1/2000 to 1/200) responded rapidly to MBP by secreting interferon (IFN)-γ, whereas such a rapid T cell response to MBP could not be detected in the presence of NK cells. Nor did we detect the memory response to MBP in the ‘CD95+ NK-low’ multiple sclerosis patients in remission or healthy subjects, regardless of whether NK cells were depleted or not. Depletion of cells expressing CD16, another NK cell marker, also caused IFN-γ secretion from MBP-reactive CD4+ T cells in the PBMCs from ‘CD95+ NK-high’ multiple sclerosis. Moreover, we showed that NK cells from ‘CD95+ NK-high’ multiple sclerosis could inhibit the antigen-driven secretion of IFN-γ by autologous MBP-specific T cell clones in vitro. These results indicate that NK cells may regulate activation of autoimmune memory T cells in an antigen non-specific fashion to maintain the clinical remission in ‘CD95+ NK-high’ multiple sclerosis patients.

  • multiple sclerosis
  • myelin basic protein
  • NK cell
  • NK2
  • T cell–NK cell interaction
  • CBA = cytokine bead array
  • HLA = human leukocyte antigen
  • IFN = interferon
  • IL = interleukin
  • MBP = myelin basic protein
  • MS-rel = multiple sclerosis in relapse
  • MS-rem = multiple sclerosis in remission
  • NK = natural killer
  • NK2 = NK type 2
  • OVA = ovalbumin
  • PBMCs = peripheral blood mononuclear cells
  • PI = propidium iodide
  • PLP = proteolipid protein
  • TCC = T-cell clone
  • TNF = tumour necrosis factor


Multiple sclerosis is a chronic neurological disease the pathology of which is characterized by multiple foci of inflammatory demyelinating lesions accompanying a variable degree of axonal changes (Bjartmar and Trapp, 2001). Regarding the pathogenesis of multiple sclerosis, studies have indicated that autoimmune T cells targeting myelin components play a crucial role in mediating the inflammatory process, particularly in the early stages of relapsing–remitting multiple sclerosis (Steinman, 2001). A number of laboratories have studied the properties of potentially pathogenic autoimmune T cell clones (TCC) reactive to myelin antigens such as myelin basic protein (MBP) and proteolipid protein (PLP), which have been derived from the peripheral blood of multiple sclerosis (Ota et al., 1990; Pette et al., 1990; Martin et al., 1991; Ohashi et al., 1995). The large majority of the TCC are CD4+ and produce T helper type 1 (Th1) cytokines such as interferon (IFN)-γ after recognizing the myelin peptide bound to human leukocyte antigen (HLA)-DR molecules. These results are consistent with the idea that the inflammatory process of multiple sclerosis is triggered by invasion of autoimmune Th1 cells into the CNS, and that exogenous or endogenous factors altering the Th1/Th2 balance may influence the disease activity. The relevance of this postulate is actually supported by clinical observations that Th2-inducing medications, such as copolymer-1, are beneficial for multiple sclerosis (Duda et al., 2000; Neuhaus et al., 2000), and that administration of IFN-γ showed deleterious effects on multiple sclerosis in previous clinical trials (Panitch et al., 1987).

Although there are a number of candidate target antigens for multiple sclerosis, MBP is thought to be a primary target for autoimmune T cells, at least in some patients (Bielekova et al., 2000). It is of note that MBP- or PLP-specific TCC can be established not only from multiple sclerosis, but also from peripheral blood of healthy subjects, which raised the intriguing issue as to how healthy subjects are protected from self-attack by the potentially pathogenic autoimmune Th1 cells. Although much remains to be clarified, studies in the last decade have showed that regulatory cells are involved in prevention of or recovery from autoimmune diseases in rodent (Das et al., 1997; Zhang et al., 1997; Olivares-Villagomez et al., 1998; Sakaguchi et al., 2001). This allows us to speculate that regulatory cells may contribute to protecting healthy subjects from developing autoimmune diseases such as multiple sclerosis, or to prohibiting acute attacks or enhancing the recovery from clinical exacerbations in patients with relapsing–remitting multiple sclerosis.

Whereas regulatory cells constitute various lymphoid populations, substantial evidence supports that natural killer (NK) cells play significant roles in protecting against autoimmune diseases (Zhang et al., 1997; Matsumoto et al., 1998; Smeltz et al., 1999). In fact, it has previously been demonstrated that NK cell depletion augments the severity of a model for multiple sclerosis, experimental autoimmune encephalomyelitis (EAE) (Zhang et al., 1997; Matsumoto et al., 1998), which can be induced by sensitization against CNS myelin component. Given that autoimmune Th1 cells would mediate the pathology of EAE, we propose a possible involvement of NK cells in suppressing autoimmune Th1 cells in multiple sclerosis.

With the hypothesis that NK cells may contribute to maintaining the remission in relapsing–remitting multiple sclerosis, we have previously examined the cytokine production and surface phenotype of NK cells freshly isolated from the peripheral blood mononuclear cells (PBMCs) of multiple sclerosis in remission (MS-rem) or relapse (MS-rel) (Takahashi et al., 2001). The results demonstrate that NK cells in MS-rem (but not MS-rel) are characterized by a remarkable elevation of interleukin (IL)-5 mRNA and a decreased expression of IL-12Rβ2 mRNA, as well as a higher percentage of CD95+ cells among the CD56+ NK cells. These features of the cells are reminiscent of NK type 2 (NK2) cells, which can be induced in vitro in the presence of IL-4 and of anti-IL-12 antibodies (Peritt et al., 1998). The NK2 cells induced from PBMCs of healthy subjects inhibit the generation of IFN-γ-secreting Th1 cells from the PBMCs of the same subjects (Takahashi et al., 2001), leading us to postulate that NK2-like cells detected in MS-rem may play a regulatory role. While the NK2-like features were found to be lost in patients at acute relapsing state, they tended to be restored along with clinical recovery. Obviously, these results do not imply that clinically diagnosed MS-rem represents a homogeneous condition. In fact, the parameters characteristic for NK2-like cells (i.e. up-regulation of IL-5 mRNA and an increased frequency of CD95+ cells) showed a substantial variance in MS-rem, indicating their heterogeneity.

More recently, we have noticed that MS-rem can be divided into two subgroups, ‘CD95+ NK-high’ and ‘CD95+ NK-low’, according to the frequency of CD95+ cells among NK cells. Here, we demonstrate that these two groups significantly differ in the responsiveness to MBP ex vivo in an NK-cell-depleted condition. Namely, NK-deleted PBMCs from ‘CD95+ NK-high’ multiple sclerosis responded rapidly to MBP, as assessed by the frequency of IFN-γ-secreting CD4+ T cells at 8 h after stimulation with MBP, whereas those from the ‘CD95+ NK-low’ or from healthy subjects responded only marginally. Moreover, we showed that NK cells from a ‘CD95+ NK-high’ multiple sclerosis could inhibit the antigen-driven secretion of IFN-γ by MBP-specific TCC established from the same patient. These results demonstrate, for the first time to our knowledge, that NK cell depletion leads to augmentation of memory T cell response to an autoantigen in human, and that an elaborate interplay between NK cells and MBP-specific memory T cells may be involved in the regulation of multiple sclerosis in ‘CD95+ NK-high’ patients.

Material and methods


To clarify the heterogeneity among patients with MS-rem regarding NK cell phenotype, we first examined 30 patients with MS-rem (male/female = 11/19; aged 37.7 ± 11.1 years) for the lymphoid cell expression of CD95. As a control for multiple sclerosis, we examined 26 healthy sex- and age-matched subjects (male/female = 11/15; aged 39.9 ± 12.2 years). Furthermore, for a new cohort of 14 patients with MS-rem (male/female = four/10; aged 39.2 ± 10.7 years) (Table 1) and 14 healthy subjects (male/female = five/nine; aged 35.3 ± 8.0 years), we conducted the cytokine secretion assay as well as flow cytometer analysis for the frequency of CD95+ NK cells. Two of the patients were examined again after a 1-year interval.

View this table:
Table 1

List of the PBMC samples examined for the frequency of memory Th1 cells

Information on patients
PBMC codeAge (years)/sexCD95+ NKfrequencyEDSS#
  • The samples marked with * or ** are derived from the same patients, with an interval of 1 year between samples. The phenotype of both of these patients changed from ‘CD95+ NK-low’ to ‘CD95+ NK-high’. M = male; F = female; EDSS = Expanded Disability Status Scale.

Written informed consent was obtained from all patients and healthy volunteers and the study was approved by the Ethics Committee of the National Center of Neuroscience (NCNP). All patients fulfilled standard criteria for the diagnosis of relapsing–remitting multiple sclerosis (Poser et al., 1983; McDonald et al., 2001). The clinical status of multiple sclerosis (MS-rem or MS-rel) was operationally determined as described previously (Takahashi et al., 2001). In brief, we selected MS-rem patients for study who had been clinically stable without any immunosuppressive medications for >3 months, and had shown no sign of new lesions as assessed by a recent MRI scan with gadolinium enhancement. None of our patients represented the pure optic-spinal form of multiple sclerosis (Misu et al., 2002), which may be rather unique to Japanese populations.


Anti-CD3-FITC or -ECD, anti-CD4-PC5, anti-CD8-FITC, anti-CD16-Phytoerythrin, and anti-CD56-PC5 or -PE mAbs were purchased from IMMUNOTECH (Marseille, France). Anti-CD57-FITC, anti-CD69-PE, anti-CD94-FITC, anti-CD95-FITC, -Cych or -PE, anti-CD158a-FITC, anti-NKB1-FITC, and anti-HLA-DR-FITC mAbs were purchased from BD PharMingen (San Jose, CA, USA). Human MBP was purified with a modification of previously described methods (Deibler et al., 1972, 1995).

Cell preparation and NK cell deletion

Shortly after drawing peripheral blood, PBMCs were separated by density gradient centrifugation with Ficoll-Hypaque™ PLUS (Amersham Biosciences, Uppsala, Sweden). They were washed three times in phosphate-buffered saline (PBS), and resuspended at 1 × 106 cells/ml in AIM-V culture medium (Invitrogen Corp., Carlsbad, CA, USA) containing 2 mM l-glutamine, 100 U/ml penicillin and 100 µg/ml streptomycin (Life Technologies, Rockville, MD, USA). NK cells were depleted from the PBMCs with either CD56- or CD16-MicroBeads (Miltenyi Biotech, Gradbach, Germany), following the protocol provided by the manufacturer.

T cell clones

CD4+ TCC were generated from a ‘CD95+ NK-high’ multiple sclerosis patient (HLA-DRB1*1502) by repeated selection against human whole MBP with modification of a previously described method (Pette et al., 1990). The TCC proliferated and secreted Th1 cytokines specifically in response to MBP, and the proliferative response and cytokine production was greatly reduced in the presence of antibodies against HLA-DR. The DR-restricted clone cells were grown in AIM-V medium supplemented with 2 mM l-glutamine, 100 U/ml penicillin and 100 µg/ml streptomycin.

T-cell stimulation with MBP

To assess the presence of memory MBP-reactive T cells in the peripheral blood, fresh PBMCs or NK-deleted PBMCs were stimulated for 8 h with 10 µg/ml MBP in 96-well round-bottomed plates at 2 × 105 cells/well, and then analysed for the presence of IFN-γ-secreting cells using the cytokine secretion assay. To evaluate the regulatory function of NK cells from ‘CD95+ NK-high’ multiple sclerosis, resting cells of MBP-specific TCC (2 × 104 cells/well) were stimulated with 10 µg/ml MBP in the presence of X-irradiated (5000 rad) autologous total PBMCs or CD56+ NK-deleted PBMCs (1 × 105 cells/well) for 8 h prior to the cytokine secretion assay, and for 60 h to determine the proliferation of the TCC. To assess cell proliferation, we counted incorporation of [3H]thymidine (1 µCi/well) during the final 12 h with a beta-1205 counter (Pharmacia, Uppsala, Sweden).

Cytokine secretion assay

We used a commercial kit from Miltenyi Biotech to identify T cells secreting IFN-γ. The principle of this assay has been described previously (Manz et al., 1995). Briefly, cells were stained with IFN-γ capture antibody 8 h after stimulation with MBP or ovalbumin (OVA), then washed and cultured again for 45 min. They were stained with PE-conjugated IFN-γ detection antibody, together with anti-human CD3-FITC and -CD4-PC5, then washed and resuspended in PBS containing propidium iodide (PI) (BD PharMingen). Samples were analysed using flow cytometry.

Cytokine bead array

The levels of IL-2, -4, -5, -10, tumour necrosis factor (TNF)-α and IFN-γ in the culture supernatants were measured by cytokine bead array (CBA) (BD PharMingen), in which six bead populations with distinct fluorescence intensities are coated with capture antibodies specific for each cytokine (Cook et al., 2001). The cytokine capture beads were mixed with the PE-conjugated detection antibodies and then incubated with recombinant standards or supernatant samples to form sandwich complexes. After washing the beads, sample data were acquired using the flow cytometer and were analysed with the BD CBA Analysis Software® (BD PharMingen).


An increased frequency of CD95+ NK cells distinguishes a subgroup of multiple sclerosis

As we have reported previously (Takahashi et al., 2001), whereas proportions of CD3 CD56+ NK cells in fresh PBMCs weakly express CD95 on their surface, the frequency of CD95+ NK cells is significantly elevated in MS-rem as compared with healthy subjects or MS-rel. We have further noticed that MS-rem can be divided into two subgroups according to the frequency (%) of CD95+ cells among NK cells (Fig. 1A; see also the left panels in Figs 1B and 2A, showing the distinction between CD95+ and CD95 cells). When we determined the mean +2 SD value for healthy subjects (35.86%) as an upper boundary for healthy subjects, three-quarters of MS-rem had a percentage value higher than this boundary. We defined these patients in remission with a higher frequency of CD95+ cells in NK cells as ‘CD95+ NK-high’ multiple sclerosis, and the rest as ‘CD95+ NK-low’ (Fig. 1A). In contrast to CD56+ NK cells, CD3+CD56 T cells and CD3+CD56+ NK T cells were not different between healthy subjects and multiple sclerosis patients as regards the frequency of CD95+ cells (data not shown), which directed our attention to the analysis of CD56+ NK cells.

Fig. 1

Characterization of CD95+ NK cells from ‘CD95+ NK-high’ multiple sclerosis. (A) Multiple sclerosis patients in remission (MS-rem) can be subgrouped into ‘CD95+ NK-high’ and ‘CD95+ NK-low’. Freshly isolated PBMCs from 26 healthy subjects or 30 MS-rem were stained with the combination of anti-CD3-FITC, -CD56-PC5 and -CD95-PE, and evaluated for the frequency of CD95+ cells in the CD56+CD3 NK cell population (the fluorescence intensity for CD95 expression is shown in the histograms in B and Fig. 2A). Note that flow fluorocytometric analysis was completed within 2 h after drawing blood in order to avoid spontaneous death of CD95+ cells. (B) Comparison of CD95+ versus CD95 NK cells in the expression of various surface molecules. We stained the PBMCs from ‘CD95+ NK-high’ patients with the panel of antibodies for surface molecules expressed by NK cells. Red lines represent the histogram gated for CD95 NK cells and blue lines for CD95+ NK cells. Values in red and in blue represent the positive percentage in CD95 and CD95+ cells, respectively. As indicated, CD95+ NK cells did not differ significantly from CD95 NK cells in the staining pattern for each antibody regarding the proportion of positive cells as well as the mean fluorescence intensity. Shown are the results of a representative case.

Fig. 2

Evidence for the role of NK cells in the regulation of MBP-reactive memory T cells in ‘CD95+ NK-high’ multiple sclerosis.(A) IFN-γ secretion assay for NK-cell-deleted PBMCs and freshly isolated PBMCs. Whole PBMCs or PBMCs depleted for CD56+ NK cells [CD56(−) PBMC] from the ‘CD95+ NK-high’ multiple sclerosis (n = 9), ‘CD95+ NK-low’ multiple sclerosis (n = 7) or healthy subjects (n = 14) were stimulated with 10 µg/ml of human MBP for 8 h for the IFN-γ secretion assay. The cells were also stained with anti-CD4-PC5 and -CD3-FITC, and the CD4+CD3+ and PI cells were gated for analysis. Here we show representative results from ‘CD95+ NK-high’ (top), ‘CD95+ NK-low’ (middle) and healthy subjects (bottom). The IFN-γ-secreting CD4+ T cells are shown as red dots; blue dots represent IFN-γ-negative cells. The histograms demonstrate the level of CD95 expression on the fresh CD56+ NK cells from each individual, and the attached values show the frequency of CD95+ cells. (B) Frequency (%) of MBP-reactive memory T cells among CD4+ T cells. By using the cytokine secretion assay, we determined the frequency of IFN-γ-positive cells among CD4+ T cells in each individual after culture with or without MBP. Here we plot the Δ% values [(%) with MBP – (%) without MBP], which represent the frequency of MBP-reactive CD4+ T cells in each subject. Kruskal–Wallis test with Scheffe's F post hoc test was used for statistical analysis. *P < 0.05; **P < 0.02.

Because NK cells from MS-rem were found to express a larger amount IL-5 mRNA, and since they were neither defective in cytolytic function nor reduced in number (Takahashi et al., 2001), we hypothesized that the CD95 expression may reflect an activation state of the NK cells. To test this hypothesis, we compared the CD95+ and CD95 NK populations derived from ‘CD95+ NK-high’ patients by flow cytometry. Histogram plot analysis for the proportion of positive cells and for mean fluorescence intensity showed that the two populations are analogous in the expression of HLA-DR, CD69, CD8, CD16, CD57, CD94, CD158a and NKB1 (Fig. 1B). Whereas HLA-DR and CD69 molecules are regarded as cell activation markers, few populations of CD95+ NK cells from multiple sclerosis or healthy subjects expressed these molecules. These results do not support the idea that the CD95+ NK cells are in a state of activation, nor do they indicate that the CD95+ cells represent a unique subset of monoclonal or oligoclonal origin. It has recently been suggested that CD56bright NK cells may represent a distinct subset (Jacobs et al., 2001). However, we saw no difference in the proportion of CD56bright cells between CD95+ and CD95 NK cells (data not shown).

CD56+ NK cell depletion induces the rapid activation of MBP-reactive memory T cells in PBMCs from ‘CD95+ NK-high’ multiple sclerosis

We have previously shown that the CD95+ NK cells found in multiple sclerosis patients resemble the NK cells that can be induced in culture in the presence of IL-4 and anti-IL-12 mAb [referred to as ‘NK2-like cells’ according to the definition by Peritt et al. (1998)]. We also found that Peritt's NK2 cells induced in vitro inhibited the induction of IFN-γ-secreting T cells from peripheral T cells after stimulation with phorbol myristate acetate and ionomycin (Takahashi et al., 2001). Based on these observations, we speculated that NK cells might prohibit Th1 cell activation in the remission of multiple sclerosis in an antigen-non-specific manner, and contribute to maintaining the remission. However, it remained an open question as to whether the NK2-like cells found in MS-rem would indeed regulate pathogenic autoimmune T cells in vivo. To investigate functions of NK cells in MS-rem, we evaluated the effect of NK cell depletion on the peripheral T cell response to MBP, a major target antigen of multiple sclerosis (Bielekova et al., 2000). In brief, we depleted CD56+ cells from the PBMCs with a magnetic sorter, and then stimulated the NK-deleted populations as well as whole PBMCs with MBP in vitro for 8–24 h. Subsequently, we detected the antigen-responsive T cells based on the secretion of IFN-γ (Manz et al., 1995). The preparatory experiments revealed that 8 h of stimulation provides an optimal condition yielding a low background (0–0.03%). This novel assay enables us to selectively detect memory-type Th1 cells that can respond rapidly to antigen, whereas previous assays that depend on long-term cultures (Pette et al., 1990; Martin et al., 1992) evaluate not only memory but also naive T cells. Of note, there is a general consensus that peripheral blood of multiple sclerosis patients contains MBP-reactive T cells that are activated and/or differentiated into memory T cells (Allegretta et al., 1990; Martin et al., 1992; Zhang et al., 1994; Lovett-Racke et al., 1998; Scholz et al., 1998).

We examined 16 PBMC samples from 14 MS-rem patients (nine samples from ‘CD95+ NK-high’, and seven from ‘CD95+ NK-low’) and 14 healthy subjects (see Table 1). When freshly isolated PBMCs were stimulated with MBP before NK cell depletion, four MS-rem and five healthy subjects samples showed a marginal response to MBP (0.01–0.03% increase of IFN-γ-positive cells among CD4+ T cells). We did not find any significant response to MBP with the other PBMC samples. In contrast, when cells were stimulated with MBP after deleting CD56+ NK cells, a significant response with a stimulatory index >3 was detected in seven of the nine ‘CD95+ NK-high’ samples, and a marginal response was detected in two (Fig. 2A and B). Of note, none of the NK-deleted samples from the ‘CD95+ NK-low’ patients and healthy subjects showed a definitive response to MBP. The difference for the ‘CD95+ NK-high’ versus the ‘CD95+ NK-low’ or healthy subjects was statistically significant (Fig. 2B). These ex vivo experiments have revealed that the ‘CD95+ NK-high’ patients may possess a higher number of T cells that can rapidly respond to MBP (MBP-specific memory T cells), compared with ‘CD95+ NK-low’ MS-rem or healthy subjects. In other words, they provide strong evidence for clonal expansion of memory autopathogenic T cells in the ‘CD95+ NK-high’ patients. However, as we could demonstrate an increase of the memory autoimmune T cells only after depleting NK cells, we interpreted that the potentially hazardous autoimmune T cells are being controlled by counter-regulatory NK cells in the ‘CD95+ NK-high’ patients. Of note, previous studies relying on alternative assays have revealed the presence of MBP-reactive T cells with activated and/or memory phenotypes at similar high frequencies in not all, but a major portion, of multiple sclerosis patients (Allegretta et al., 1990; Zhang et al., 1994; Bieganowska et al., 1997; Lovett-Racke et al., 1998; Scholz et al., 1998; Illés et al., 1999).

We conducted the same assay with a foreign antigen OVA in three of the ‘CD95+ NK-high’ (PBMC codes #3, #4 and #5 in Table 1) and one of the ‘CD95+ NK-low’ samples (#6). However, OVA-reactive T cells could not be detected in any sample of the fresh or NK-deleted PBMCs (data not shown). Because NK cells cannot discriminate T cells with different antigen specificities, the negative response to OVA in the four multiple sclerosis patients was interpreted to mean that they do not possess clonally expanded memory T cells reactive to OVA.

Depletion of CD16+ NK cells also allows detection of MBP-reactive memory T cells in PBMCs from ‘CD95+ NK-high’ multiple sclerosis

Although we used anti-CD56 magnetic beads to deplete NK cells in the above experiments, the method would also deplete CD3+CD56+ NK T cells that may possibly play a role in the regulation of autoimmunity. To evaluate the possible contribution of CD3+CD56+ NK T cells, we next depleted NK cells from PBMCs from two ‘CD95+ NK-high’ patients on the basis of their expression of CD16. We found that after treatment with CD16-MicroBeads, almost all of CD56+ NK cells are deleted, but CD56+CD3+ NKT cells remain largely untouched (Fig. 3A). However, like CD56+-cell-deleted PBMCs, the CD16+-cell-deleted PBMCs responded to MBP, as assessed by the induction of IFN-γ-secreting CD4+ T cells (Fig. 3B). The responses found in the two patients were considered significant with regard to both percentage increase of IFN-γ-secreting cells (0.08% and 0.04%) and the stimulatory index (9.0 and 5.0) obtained after MBP stimulation. This result indicates that responsible cells to regulate autoimmune T cells in ‘CD95+ NK-high’ multiple sclerosis are not CD56+CD3+ NK T cells but NK cells.

Fig. 3

Depletion of CD16+ cells also allows detection of MBP-specific memory T cells in ‘CD95+ NK-high’ multiple sclerosis.(A) Changes in the frequency of CD56+ NK cells and CD56+ NKT cells after deleting CD16+ cells. Using CD16 microbeads, we deleted CD16+ cells from PBMCs from two ‘CD95+ NK-high’ patients and from two healthy subjects. The cells were stained with anti-human CD3-FITC and anti-CD56-PC5 to check the proportion of CD56+ NK cells and CD56+ T cells before and after CD16+ cell depletion (upper versus lower panels). Shown are the results of a representative pair of multiple sclerosis and healthy subjects.(B) CD16+-cell-depleted PBMCs from ‘CD95+ NK-high’ multiple sclerosis responded rapidly to MBP. Using the same PBMC samples (CD16+ or CD16), we conducted the IFN-γ secretion assay as described in Fig. 2A. This figure shows the result of the representative pair of multiple sclerosis patients and healthy subjects.

Unfortunately, it remains unclear whether only CD95+ NK cells play a regulatory role in ‘CD95+ NK-high’ multiple sclerosis or whether CD95 cells could also exhibit regulatory functions in the patients. We attempted to compare directly the function of CD95+ and CD95 populations. However, isolation of CD95+ NK cells with a cell sorter invariably induced cell activation as revealed by the expression of various activation markers. Furthermore, the isolated cells tended to die rapidly, probably due to CD95 ligation by the antibody (data not shown).

NK cells from ‘CD95+ NK-high’ multiple sclerosis inhibit IFN-γ production byMBP-reactive T cell clones

To analyse how the NK cells from ‘CD95+ NK-high’ multiple sclerosis efficiently control autoimmune T cell responses, we established three MBP-specific TCC from a ‘CD95+ NK-high’ patient. These TCC proliferated and secreted IFN-γ, TNF-α, IL-2 and IL-5 in response to MBP presented by irradiated, fresh autologous PBMCs. Using the proliferation response and cytokine secretion by the TCC as read-out, we compared the whole PBMCs and the NK cell-deleted PBMCs for the ability to present whole MBP to the autologous TCC. We found that the whole PBMCs did not differ from the NK-deleted PBMCs in the ability to induce MBP-driven proliferation of TCC (Fig. 4A). However, the proportion of IFN-γ-secreting T cells among the TCC increased significantly when the NK cell-depleted PBMCs were used as antigen presenting cells (APC) (Fig. 4B). We also noticed a significant elevation of IFN-γ in the culture supernatant along with the increase of IFN-γ-secreting T cells (Fig. 4C). However, neither TNF-α nor IL-2 production was enhanced by NK cell depletion. These results support the view that NK cells from ‘CD95+ NK-high’ multiple sclerosis regulate autoimmune T cells by inhibiting the T cell production of IFN-γ.

Fig. 4

Depletion of NK cells augments the antigen-presenting potential of PBMCs from ‘CD95+ NK-high’ multiple sclerosis. (A) Effect of NK-cell deletion on the proliferation of MBP-specific TCC. We established three MBP-specific TCC from a ‘CD95+ NK-high’ patient,and evaluated the proliferative response of the clone cells to MBP (10 µg/ml) in the presence of fresh autologous PBMCs [+ PBMC] or NK-deleted PBMCs [+ CD56 (−) PBMC]. This is a representative result of three TCC, which yielded essentially the same results. Data represent mean ±SD of quadruplicate cultures. (B) Effect of NK cell deletion on IFN-γ secretion by the MBP-specific TCC. MBP-specific TCC were cultured with or without MBP for 8 h in the presence of autologous PBMCs (upper panels) or of the autologous PBMCs depleted for CD56+ NK cells (lower panels). We then conducted the cytokine secretion assay to detect IFN-γ-positive cells. Red dots indicate IFN-γ-secreting cells among CD4+CD3+PI cells; blue dots represent IFN-γ-negative CD4+CD3+PI T cells. The values (%) represent the frequency of IFN-γ-secreting cells among CD4+CD3+PI cells. We conducted the assay with three TCC, which yielded essentially the same results. FS = forward scatter. (C) Effect of NK cell depletion on cytokine release by TCC into culture medium. The TCC were stimulated with MBP for 48 h in the presence of autologous PBMCs or NK-depleted PBMCs. Then we measured the concentrations of IFN-γ, TNF-α, IL-10, IL-5, IL-4 and IL-2 in the supernatants, using ELISA and CBA. Both assays yielded essentially the same results, and here we show the result of a CBA assay. Data represent mean ± SD. The Mann–Whitney U-test was used for statistical analysis. *P < 0.05. We conducted the assay with three TCC, which yielded essentially the same results.


It is generally held that relapse of multiple sclerosis represents the destructive CNS inflammation triggered by recently activated autoimmune T cells. In other words, pathogenic autoimmunity is apparently active during clinical relapse, which can be objectively defined by clinical status as well as MRI findings. In contrast, remission of multiple sclerosis, which is chiefly determined by exclusion of active inflammation in the CNS, may probably cover a wider range of disease states. The present results show that multiple sclerosis patients in remission can be divided at least into two subgroups, ‘CD95+ NK-high’ and ‘CD95+ NK-low’, based on the frequency of CD95+ cells among NK cells. Furthermore, our functional analysis combining NK cell deletion and stimulation with MBP has indicated that the two subgroups differ significantly with regard to the responsiveness of the MBP-specific memory T cells to MBP in the absence of NK cells. Namely, after deleting CD56+ NK cells, we saw a rapid induction of IFN-γ-secreting, anti-MBP T cells in ‘CD95+ NK-high’ multiple sclerosis, whereas such a rapid response to MBP was not seen in ‘CD95+ NK-low’ multiple sclerosis or healthy subjects. This result is in harmony with the previous results that clonally expanded MBP-specific T cells can be detected in a majority of multiple sclerosis patients (Zhang et al., 1994; Smeltz et al., 1999), and indicates that patients with an increased number of the autoimmune T cells may have the ‘CD95+ NK-high’ phenotype during remission. Thus, the frequency of CD95+ NK cells correlates with the frequency of MBP-reactive memory T cells and may serve as a useful marker to evaluate the immunological status of multiple sclerosis during remission.

The role of NK cells in the regulation of MBP-specific T cells was further strengthened by the demonstration that deletion of CD16+ cells also enabled detection of memory MBP-specific T cells. Because we confirmed that depletion of the CD16+ cells would greatly reduce the number of NK cells but did not significantly reduce CD56+CD3+ NK T cells, the role of the NK T cells in the regulation was excluded.

We have previously described that the ‘CD95+ NK-low’ phenotype could also be seen in multiple sclerosis patients during relapse. However, the ‘CD95+ NK-low’ phenotype in MS-rel was not persistent, but the ‘CD95+ NK-high’ phenotype could be regained in a month or so along with clinical recovery. This fact raised the possibility that ‘CD95+ NK-low’ MS-rem may represent an active state of multiple sclerosis, contrary to our speculation. To evaluate this possibility, we examined three patients with MS-rem for the ‘CD95+ NK-high/low’ phenotype every 4–6 weeks, and found that they maintained the ‘CD95+ NK-low’ phenotype for longer than several months (data not shown). This is in a striking contrast to the transient appearance of the ‘CD95+ NK-low’ phenotype during relapse. Together with the clinical observations that these patients were in a very stable condition with minimal neurological disability, we estimate the disease condition in ‘CD95+ NK-low’ MS-rem to be truly inactive and distinct from MS-rel.

It is of note that IFN-γ-secreting T cells could be identified as early as 8 h after stimulation with MBP in the absence of NK cells. This result implies that the NK cells should interact with the autoimmune T cells shortly after antigen stimulation to regulate very early T cell response. To account for such a rapid regulation by NK cells, we speculate that the regulatory NK cells may detect the subtle change of the autoimmune T cells during the early stage of activation. At present, very little is known about the molecular basis of T cell–NK cell interaction. However, it is obvious that NK cells must interact with T cells in an antigen-non-specific fashion, as they do not express highly variable receptors like T cell antigen receptors. Our results indicate that attempts to identify the ligand and receptors involved in T cell–NK interactions are very rewarding.

It is currently speculated that activation of autoimmune T cells could occur in response to microbial proteins whose sequence has a significant homology to the self-peptide (Steinman, 2001). We predict that the increased MBP-reactive Th1 cells in the ‘CD95+ NK-high’ patients will most likely respond to microbial peptides mimicking MBP from time to time. However, counter-regulatory NK cells would maintain the clinical silence by actively suppressing activation of the autoimmune T cells that might lead to destructive CNS inflammation (Fig. 5). We then imagine that the clinical silence in the ‘CD95+ NK-high’ patients could readily be disrupted when NK cells are numerically or functionally altered by exogenous or endogenous factors independent of multiple sclerosis (Wu et al., 2000). In contrast, the clinical remission in ‘CD95+ NK-low’ multiple sclerosis appears to be stable, as they are expected to possess much lower numbers of MBP-specific memory T cells, which does not necessitate the active engagement of regulatory NK cells. If these premises hold true, we may consider that the ‘CD95+ NK-high’ patients are at a greater risk than ‘CD95+ NK-low’ of developing relapses when exposed to potentially dangerous microbes that have cross-reactive epitopes. To describe the immunological status in ‘CD95+ NK-high’, which seems to be more active than the ‘CD95+ NK-low’, it might be appropriate to use the term ‘smouldering’ state rather than ‘remission’.

Fig. 5

The role of NK cells in ‘CD95+ NK-high’ multiple sclerosis. As described in the text, the ‘CD95+ NK-high’ patients are characterized by a concurrent increase of memory autoimmune T cells and CD95+ NK cells. In the sense that memory autoimmune T cells cannot be detected in other patients in remission (‘CD95+ NK-low’) even after NK cell depletion, we describe the immunological status of the ‘CD95+ NK-high’ as a ‘smouldering’ state rather than ‘remission’. Given that T cell recognition is much more promiscuous than previously anticipated, we imagine that autoimmune T cells in the ‘CD95+ NK-high’ patients would respond to exogenous self-mimicking peptides from time to time. However, our results indicate that the CD95+ NK cells could detect the early sign of T cell activation and then interact with autoimmune T cells to prohibit their full activation. Once this delicate control by NK cells is disrupted, the autoimmune T cells could be fully activated in response to the self-mimicking peptides. The fully activated T cells may be controlled by other regulatory cells such as CD4+CD25+ T cells (Sakaguchi et al., 2001) or B7-1+CD4+ T cells (Kipp et al., 2000). However, it is difficult to predict how efficiently the regulatory T cells may control the activated autoimmune T cells in individual cases.

After determining the presence of the ‘CD95+ NK-high’ and ‘CD95+ NK-low’ phenotypes in the patients with MS-rem, an important question might be whether the ‘CD95+ NK-high/-low’ phenotype correlates with some clinical parameters or disease course. We speculated that ‘CD95+ NK-low’ might be clinically less active than ‘CD95+ NK-high’, when evaluated retrospectively. However, it might take time and would require a large number of patients to verify this postulate, taking the heterogeneity and chronic nature of the illness into consideration. Furthermore, it is of note that the ‘CD95+ NK-high’ or ‘-low’ phenotype appears to be interchangeable. For example, two of the patients who were examined for the memory T cell frequency showed the ‘CD95+ NK-low’ phenotype in the first examination, but were found to have the ‘CD95+ NK-high’ phenotype when examined 1 year later (Table 1). The phenotype switch in these patients was associated with an increase in the frequency of MBP-reactive memory T cells. We speculate that activity of multiple sclerosis may have been increased in these patients during the 1-year interval, although it is too early to draw any conclusions from the analysis of two patients.

Conversely, we have recently seen an opposing phenotype switch (from the ‘CD95+ NK-high’ to ‘CD95+ NK-low’) in two other patients. The frequency of CD95+ cells among NK cells was >46.0% in both cases in the initial examinations, but the latest test showed normal values (27.4% and 10.0%). Although the patients appeared to be in the state of remission at the last examination, they developed serious signs of acute exacerbation 2 days later. As stated above, a transient switch from ‘CD95+ NK-high’ to ‘CD95+ NK-low’ could occur during relapse. Therefore, we speculate that the phenotype switch from ‘high’ to ‘low’ may be triggered by the very early events leading to clinical relapse. However, it is also possible that the reduction of the CD95+ NK cells might have been triggered by multiple sclerosis-independent factors, such as infection or stress, and that this led to the occurrence of the relapse in these patients. This speculation is supported by the fact that a number of physiological conditions can alter NK cell number and/or function, and that CD95+ NK cells tend do die more rapidly in culture than CD95 NK cells (our unpublished data). In future, it will be worthwhile to examine more systematically whether the phenotype switch may be the earliest marker to detect occurrence of relapse.

As Japanese neurologists have traditionally stressed that multiple sclerosis in Japan might be quite unique in immunopathology, it is theoretically possible that the regulatory function of CD95+ NK cells reflects the uniqueness of Japanese multiple sclerosis and that the T cell–NK cell interaction is not operative in Caucasian multiple sclerosis. However, recent studies suggest that the frequency of pure optic-spinal form of multiple sclerosis linked with Japanese patients (Misu et al., 2002) is drastically declining, possibly due to change in lifestyle or environmental factors in Japan (Yamamura, 2002; Houzen et al., 2003). Reflecting this fact, the patients randomly recruited in this study did not have optic-spinal multiple sclerosis, and all had brain lesions similar to those found in Western multiple sclerosis. We therefore speculate that our experimental results will be reproduced in Caucasian patients in the future.

In summary, we have revealed that multiple sclerosis patients in remission have either ‘CD95+ NK-high’ or ‘CD95+ NK-low’ phenotype, and that ‘CD95+ NK-high’ patients have a higher frequency of memory autoimmune T cells and have more active multiple sclerosis than ‘CD95+ NK-low’ patients. Our ex vivo assay has demonstrated that ‘CD95+ NK-high’ patients possess NK cells that actively inhibit activation of memory autoimmune T cells. In the sense that clinical silence depends on the functional regulatory NK cells, the condition of ‘CD95+ NK-high’ is thought to be so unstable, as could be expressed by the term ‘smouldering’. As such, evaluation of the NK cell functions and phenotypes in multiple sclerosis gives us a new insight into the autoimmune pathogenesis of multiple sclerosis, encouraging further efforts to clarify the NK cell–T cell interactions.


We wish to thank Christian Rochford for critical reading of this manuscript. This work was supported by grants from the Ministry of Health, Labor and Welfare of Japan and from the Organization for Pharmaceutical Safety and Research.


View Abstract