A novel sulfatide analog recognized by human iNKT cells enhances polyfunctional antigen-specific CD8⁺ T cell responses

Abstract

Invariant natural killer T (iNKT) cells are CD1d-restricted T cells that recognize lipid antigens. Studies of analogs of a prototypic iNKT cell antigen, α-galactosylceramide (α-GalCer), have shown that ceramide structure affects the activities of mouse iNKT cells recognizing antigen. However, the impact of antigen structure on human iNKT cell activities is less well studied. This study evaluated the antigenicity of eight novel synthetic sulfatide-based lipids with fluorobenzene at the acyl chain terminus and a variety of sphingoid base structures. Using CD1d-tetramers, we evaluated TCR recognition in ex vivo iNKT cells from healthy donors’ peripheral blood mononuclear cells (PBMCs). One analog, SAP-3, was commonly recognized by iNKT cells from all donors. SAP-3 increased the proportion of CD69⁺CD25⁺ iNKT cells in a CD1d-dependent manner and triggered IL-2 production. Furthermore, SAP-3-activated ex vivo iNKT cells increased the frequency of polyfunctional antigen-specific CD8⁺ T cells (IFN-γ⁺IL-2⁺TNF-α⁺) in PBMC stimulation culture. Importantly, repeated in vitro stimulation with SAP-3 did not induce iNKT-cell functional anergy, unlike α-GalCer; instead, SAP-3 increased IFN-γ-producing iNKT cells. These findings suggest that the structures of sulfatide analogs critically influence their immunogenicity. SAP-3 emerges as a novel human iNKT agonist that enhances antigen-specific CD8⁺ T cell responses and represents a promising vaccine adjuvant candidate.

Introduction

Natural killer T (NKT) cells are unconventional T cells that recognize lipid antigens presented by the monomorphic MHC I-like molecule CD1d[1, 2]. Invariant NKT (iNKT) cells are a subset of NKT cells that express a semi-invariant TCRα chain comprised of Vα24-Jα18 (TRAV10-TRAJ18) in humans. The invariant TCRα chain pairs with a limited repertoire of TCRβ chains, comprised of Vβ11 (TRBV25-1) in humans[3-5]. While a single TCR on conventional T cells can recognize peptides with a limited amino acid variation, the TCR on iNKT cells recognizes a much broader range of lipid antigens. Lipid antigens with different structures induce significantly different cytokine profiles in activated iNKT cells[6, 7]. Studies of analogs of α-galactosylceramide (α-GalCer) (Fig. 1A), a prototypical iNKT cell antigen, have shown that differences in the structures of glycolipid antigens significantly alter the activities of iNKT cells[8]. For instance, α-GalCer triggers the production of both Th1- and Th2-type cytokines by iNKT cells, whereas C20:2 and OCH (a phytosphingosine truncated analog) predominantly elicit Th2-type cytokines[9, 10]. In contrast, α-C-galactosylceramide (α-C-GalCer) and 7DW8-5 predominantly induce Th1-type cytokines[11, 12]. Furthermore, 7DW8-5 (Fig. 1A) has been shown to be significantly more potent at activating both human and mouse iNKT cells than α-GalCer and it has exhibited enhanced efficacy as an adjuvant in malaria and HIV vaccines in various mouse studies[13, 14]. Importantly, 7DW8-5 can markedly improve malaria-specific immune responses in non-human primates when paired with a human malaria vaccine candidate, indicating significant potential for future human immunotherapeutic applications[15]. Thus, α-GalCer analogs provide a strong adjuvant effect on vaccine-driven induction of CD8⁺ T cells. Previous reports on utilizing α-GalCer for vaccine development in humans and mice have confirmed that it is well tolerated and safe. However, high doses of soluble α-GalCer were found to induce long-term functional anergy in murine iNKT cells[16], thereby limiting its utility as an adjuvant. Additionally, structure-function studies of β-ManCer and α-C-GalCer glycolipid antigens for iNKT cells suggest that the structure–function relationships identified from α-GalCer analogs may not always apply to other glycolipid antigens[17, 18]. These limitations underscore the need for novel iNKT cell agonists that possess immunostimulatory activity without driving functional anergy in iNKT cells.

Figure 1. Sulfatide analogs presented by CD1d are recognized by human iNKT TCRs.

(A) Structures of sulfatide analogs, α-GalCer (KRN7000), and 7DW8-5 used in this study. (B) Representative flow cytometry plots from one of three healthy donor PBMC-derived iNKT cells stained with CD1d tetramers loaded with sulfatide analogs. PBMCs were pre-incubated with or without the 6B11 antibody (dark gray rows). iNKT cells were identified as CD19⁻CD3⁺Vα24⁺Vβ11⁺ cells. Numbers indicate the percentage of CD1d tetramer–positive cells among iNKT cells. Sulfatide analog-loaded CD1d tetramer–positive population is shown in red in each plot. (C) FACS-sorted CD19⁻CD3⁺CD1d-tetramer⁺ iNKT cells reactive to α-GalCer or SAP-3 were collected from PBMCs and subjected to scTCR-seq, followed by downstream data analysis. Comparison of the usage of the top 10 CDR3β sequences enriched in SAP-3-reactive iNKT cells compared with the usage in α-GalCer-reactive iNKT cells. (D) Correlation analysis of the usage of CDR3β sequences between SAP-3-reactive and α-GalCer-reactive iNKT cells.

Sulfatides are a class of self-glycosphingolipid antigens containing a β-linked sulfated sugar moiety and are highly enriched in the myelin sheath. In mice, sulfatides (e.g. Fig. 1A) are recognized by type II NKT cells, a subset of NKT cells that do not express the invariant TCR[19-21]. Conversely, in humans, sulfatides have been reported to be recognized by iNKT cells[21-24], although the functions of human iNKT responses to sulfatides remain incompletely understood. These observations suggest that sulfatide-based analogs may provide unique opportunities to modulate human iNKT cell activity through distinct structural determinants.

In this study, we evaluated the antigenicity of novel synthetic sulfatide analogs against human iNKT cells. Significantly, no structure-activity relationship (SAR) studies around human iNKT cell activation by sulfatides have been previously reported. With the potent activation of human iNKT cells by 7DW8-5 and its promising outcomes in primate studies, we elected to focus on structural aspects of its ceramide and combine these with β-3-O-sulfogalactose. In particular, we examined acyl chains capped by fluorobenzene. The relevant carboxylic acids were coupled to C18-phytosphingosine, C18-sphinganine, or C18-sphingosine. The variation in sphingoid base was rationalized by the presence of all of these in either natural sulfatides (which generally have sphingosines) or known stimulatory analogs of α-GalCer. Eight new sulfatide analogs with structures similar to 7DW8-5 were synthesized (Fig. 1A). We investigated the ability of ex vivo iNKT cells from healthy donors’ PBMCs to recognize the analogs through their TCR and found that the analog designated SAP-3 was recognized across all donors. The sulfatide analogs activated iNKT cells in a CD1d-dependent manner. Cytokine analyses revealed donor-specific variability, with the commonly recognized analog SAP-3 inducing IL-2 production. Furthermore, this analog induced a higher frequency of polyfunctional antigen-specific CD8⁺ T cells in human PBMC stimulation culture with an influenza matrix protein (FMP)-derived peptide. Importantly, repeated stimulation with SAP-3 did not induce iNKT-cell functional anergy; moreover, it increased the frequency of IFN-γ-producing iNKT cells. SAP-3, therefore, avoids a key issue associated with α-GalCer-induced activation of iNKT cells. These findings suggest that the novel agonist SAP-3 for human iNKT cells could serve as an effective adjuvant for human vaccine development.

Results

Sulfatide analogs presented by CD1d are recognized by the human iNKT TCR

In this study, eight new sulfatide analogs were synthesized (Fig. 1A). Each analog has a fluorobenzene at the end of an acyl chain with variable lengths and one of three C18-sphingoid bases. The acyl chain moieties are related or identical to the acyl chain of 7DW8-5[12]. The structure of 7DW8-5 has a fluorobenzene ring at the end of an 11 carbon (C11) acyl chain. The eight novel sulfatide analogs were classified into three groups based on their acyl chain structure. Group 1 analogs, including SAP-1, SASg-1, and SAS-1, have a C13-acyl chain capped with a fluorobenzene. Group 2 analogs, including SAP-2, SASg-2, and SAS-2, have a C9-acyl chain capped with a triple bond connected to a terminal fluorobenzene. Group 3 analogs, including SAP-3 and SASg-3, have a C11-acyl chain capped with a fluorobenzene. The sulfatide analogs were named SAP-1, SAP-2, and SAP-3 for the phytosphingosine series, SASg-1, SASg-2, and SASg-3 for the sphinganine series, and SAS-1 and SAS-2 for the sphingosine series.

First, we examined the recognition of the sulfatide analogs by human iNKT cell TCRs, using human CD1d-tetramers loaded with sulfatide analogs (Fig. 1B, Fig. S1). We stained human PBMC-derived ex vivo iNKT cells from healthy donors. In one donor, iNKT cells (Vα24⁺Vβ11⁺CD3⁺ cells) bound to CD1d-tetramer presenting SAP-1, SASg-1, or SAP-3. Pre-incubation of iNKT cells with 6B11 antibody, a monoclonal antibody against the conserved CDR3 region of the Vα24-Jα18 TCR[25, 26], partially reduced the interaction of positive control CD1d-tetramers (PBS-57/CD1d-, KRN7000/CD1d-, and 7DW8-5/CD1d-tetramers) measured by the reduction of the signal intensity. 6B11 completely abrogated the bindings of sulfatide analog-loaded CD1d-tetramers, demonstrating that these CD1d-presented analogs bind to the TCR of PBMC-derived iNKT cells. Results from three donors using PBMC-derived ex vivo iNKT cells showed that SAP-3 is commonly recognized by iNKT cells. We therefore concluded that SAP-3 is recognized by human iNKT cells.

CDR3β repertoire of SAP-3- and α-GalCer-reactive iNKT cells

Since it has been suggested that CDR3β contributes to the recognition of self-antigens[27], we investigated TCR diversity between SAP-3- and α-GalCer-reactive iNKT cells by single-cell T-cell receptor sequencing. We analyzed a combined dataset of 3829 cells from SAP-3-reactive iNKT cells and 12528 cells from α-GalCer-reactive iNKT cells that express the TCRα-chain gene segments TRAV10 and TRAJ18. Nearly all SAP-3- and α-GalCer-reactive iNKT cells expressed the invariant CDR3α sequence CVVSDRGSTLGRLYF, accounting for 98.6% and 99.7% of cells, respectively (Fig. S2A). For the TCRβ chain, TRBV25-1 was used by 99.7% of cells in both SAP-3- and α-GalCer-reactive iNKT cells (Fig. S2B). Greater diversity was observed in TCRβ J-gene segment usage, with clonotypes utilizing the TRBJ1-1, TRBJ1-2, TRBJ1-3, TRBJ1-4, TRBJ1-5, TRBJ1-6, TRBJ2-1, TRBJ2-2, TRBJ2-3, TRBJ2-4, TRBJ2-5, TRBJ2-6, and TRBJ2-7 gene segments (Fig. S2C). The CDR3β repertoire was highly diverse, with 189 and 295 unique CDR3β sequences detected in SAP-3- and α-GalCer-reactive iNKT cells, respectively. The two groups of iNKT cells exhibited highly similar CDR3β sequence usage, with a strong positive correlation (Pearson's r = 0.89) (Fig. 1C & 1D). Despite the high overall similarity in CDR3β repertoire usage between SAP-3- and α-GalCer-reactive iNKT cells, individual clones differed in relative abundance. S194 was the dominant clone in both the SAP-3 (6%) and α-GalCer (12%) groups. The SAP-3 group showed relatively higher frequencies of clones S90 and S11 (2.7% and 3.2%, respectively), whereas these clonotypes accounted for 1.9% and 1.4% of the repertoire in the α-GalCer group. Together, these findings indicate that, despite the substantial overlap in TCRβ repertoires between SAP-3- and α-GalCer-reactive iNKT cells, the two populations exhibit distinct patterns of clonotype usage.

Insights from molecular dynamics (MD) simulation of SAP-3 interaction with CD1d and iNKT TCR

The preferential enrichment of the S90 and S11 clones among SAP-3-reactive iNKT cells raises the question of whether their differential usage reflects enhanced interactions between SAP-3–CD1d complexes and the specific TCRs. Establishing such a relationship could provide mechanistic insight into the structural determinants of iNKT cell antigen recognition. To investigate this possibility, we performed 50-nanosecond (50-ns) molecular dynamics (MD) simulations of SAP-3–CD1d and α-GalCer–CD1d complexes bound to the three TCRs preferentially enriched in SAP-3-reactive (TCR_S90 and TCR_S11) or α-GalCer-reactive (TCR_S194) iNKT cells to characterize the molecular interactions underlying antigen recognition (Fig. S3A-F, Fig. S4A-F).

In the TCR-glycolipid-CD1d complexes with the three TCRs, SAP-3 and α-GalCer displayed highly conserved binding modes within the CD1d groove, with subtle variations in their hydrogen-bonding networks (Fig. S3A-F). Both glycolipids formed recurrent interactions with CD1d residues ASP80, ASP151, and THR154 and with residues in the invariant TCRα chain, most commonly GLY96, and, depending on the clone and ligand, PHE30 or SER31. Notably, the SO₃-bearing galactose of SAP-3 formed additional hydrogen bonds with ASP151 that were absent in the corresponding α-GalCer complexes. Conversely, α-GalCer showed more consistent interactions with PHE30 or SER31 in the CDR1α loop. Although the overall binding pattern remained consistent across the three TCRs, distinct ligand-specific differences in hydrogen-bonding patterns clearly emerged, underscoring the nuanced yet critical variations in these interactions.

MM/PBSA binding free-energy calculations further supported the formation of stable complexes for both glycolipids. SAP-3 exhibited predicted binding energies of −918.1, −850.0, and −841.2 kJ/mol with TCR_S194, TCR_S90, and TCR_S11, respectively, whereas α-GalCer showed stronger predicted binding energies of −1078.0, −1049.0, and −1050.9 kJ/mol, respectively (Table S1). Together, these findings indicate that SAP-3 and α-GalCer engage the invariant TCR through a largely conserved interaction network, while the SO₃ group of SAP-3 introduces additional stabilizing contacts without substantially altering the overall binding mode.

Sulfatide analogs activate human iNKT cells in a CD1d-dependent manner

We next investigated the stimulatory properties of the sulfatide analogs against human PBMC-derived ex vivo iNKT cells by stimulating PBMCs with each sulfatide analog. The frequency of CD69⁺CD25⁺ cells increased in a dose-dependent manner, except for SAP-2 (Fig. 2A; SAP-2 shows cytotoxicity (reduced viability) to cultured cells at 3.3 μM, but not 1.1 μM). Because CD69⁺CD25⁺ iNKT cells constitute a relevant readout of antigen-specific, CD1d-dependent activation[28], stimulation assays were performed at 10 μM for all analogs except SAP-2, for which assays were performed at 1 or 1.1 μM.

Figure 2. Sulfatide analogs activate human iNKT cells in a CD1d-dependent manner.

(A) The percentages of CD69⁺CD25⁺ iNKT cells (CD3⁺CD1d tetramer⁺) in healthy donor PBMC-derived ex vivo iNKT cells 4 days after stimulation with various concentrations of lipids. (B) Frequencies of CD69⁺CD25⁺ iNKT cells (CD3⁺ CD1d tetramer⁺) 4 days after stimulation with various sulfatide analogs (1 or 1.1 μM SAP-2, or 10 μM of each sulfatide analog except SAP-2) in six donors. (C) Frequencies of CD69⁺CD25⁺ iNKT cells (CD3⁺CD1d tetramer⁺) 4 days after stimulation with various sulfatide analogs (1 μM SAP-2, or 10 μM of each sulfatide analog except SAP-2) in the presence or absence of anti-CD1d antibody from a representative donor are shown (n=3 independent donors). (D) Cytokine levels in culture supernatant of human PBMCs from six healthy donors stimulated with lipids (1 μM α-GalCer, 1 μM 7DW8-5, 1 μM SAP-2, or 10 μM of each sulfatide analog except SAP-2). Each symbol represents one donor (n=6). Signals below the detection limit for each cytokine are set to zero. (E) Human PBMCs from eleven healthy donors were cultured with 10 μM SAP-3 or vehicle control (0.5% Tween 20 PBS) in the presence or absence of anti-CD1d antibody. Each symbol represents one donor (n=11). (F) Human PBMCs from eight healthy donors were cultured with 1 μM α-GalCer or 10 μM SAP-3 in the presence or absence of anti-CD25 antibody. IL-2 levels in culture supernatants from 4-day culture. Each dot represents one donor (n=8). (G) Human PBMCs were stimulated with 1 μM α-GalCer or 10 μM SAP-3 and then restimulated with the same stimulus for intracellular staining. Fresh uncultured PBMCs, used as controls, were directly stimulated with PMA/ionomycin cocktail. Cytokine production (IFN-γ, IL-2, and IL-4) by iNKT cells was quantified using flow cytometry. The bar graphs show the percentage of iNKT cells producing each cytokine under the indicated conditions in individual donors (n=3). iNKT cells were defined as CD3⁺Vα24⁺Vβ11⁺ iNKT cells. (B, D, E, F) Data are presented as box and whisker plots showing median (line within box), interquartile range (box), and minimum to maximum values (whiskers). Statistical significance was calculated using Wilcoxon matched-pairs signed rank test. *p < 0.05, **p < 0.01. (H and I) FACS-sorted CD19⁻CD3⁺CD1d-tetramer⁺iNKT cells reactive to SAP3 or α-GalCer were collected from PBMCs and subjected to scRNA-seq, followed by downstream data analysis. (H) UMAP showing the iNKT cell clusters identified by unsupervised clustering. (I) Distribution of the cell clusters across samples (SAP-3 and α-GalCer).

Of the eight sulfatide analogs, all except SAP-2 and SAS-2, activated ex vivo iNKT cells, as evidenced by increased frequencies of CD69⁺CD25⁺ iNKT cells (Fig. 2B). Additionally, all except SAP-2, SAS-2, and SASg-3 increased the frequencies of CD69⁺CD25⁻ iNKT cells (Fig. S5A). The CD69⁻CD25⁺ subset did not show a meaningful increase above the unstimulated control across all eight analogs (Fig. S5A). CD1d blockade abrogated the increase in CD69⁺CD25⁺ iNKT cell populations (Fig. 2C), indicating antigen-specific, CD1d-dependent activation. Taken together, the sulfatide analogs activate human ex vivo iNKT cells in a CD1d-dependent manner.

SAP-3 induces IL-2 production by PBMC-derived iNKT cells

Next, we measured the production of 12 cytokines in culture supernatants collected from sulfatide analog-stimulated PBMCs (Fig. 2D). SAP-3 was the only compound that significantly induced IL-2 production, whereas α-GalCer did not. The induction of IL-2 production was CD1d-dependent (Fig. 2E). Taken together, SAP-3 stimulates human ex vivo iNKT cells in a CD1d-dependent manner. SAS-2 significantly reduced the production of 8 cytokines (IL-2, IFN-γ, TNF-α, IL-5, IL-9, IL-10, IL-13, and IL-17F) compared to the control group. To determine whether IL-2 was produced but subsequently consumed by activated iNKT cells via the IL-2 receptor, we blocked CD25, IL-2Rα. Under CD25 blockade, α-GalCer induced significantly higher levels of IL-2 compared to the unstimulated control. These findings suggest that the low IL-2 concentrations observed following α-GalCer stimulation are likely due to consumption of IL-2 by activated iNKT cells themselves (Fig. 2F). To assess and compare the potency of SAP-3 and α-GalCer to activate iNKT cells, ex vivo iNKT cells were stimulated with increasing concentrations of lipid antigens (Fig. S5B). SAP-3 induced a dose-dependent increase in the percentage of CD69⁺ iNKT cells. Following α-GalCer treatment, TCR downregulation compromised clear gating of the iNKT cell population, precluding assessment of CD69 expression in these cells. For IL-2 production, considerable variability was observed among technical replicates. SAP-3 induced a concentration-dependent increase in IL-2 production, whereas α-GalCer did not (Fig. 2D). However, no clear concentration-dependent trend was observed when the IL-2 receptor (CD25) blocking antibody was added. This may reflect a contribution from IL-2 production by non-NKT cells in the culture. In addition, to examine the phenotype of ex vivo iNKT cells expanded in response to SAP-3 or α-GalCer stimulation, the production of IL-4, IFN-γ, and IL-2 in iNKT cells was analyzed after a week of stimulation with either SAP-3 or α-GalCer followed by restimulation with the same lipid antigen, using intracellular cytokine staining. (Fig. 2G). Fresh ex vivo iNKT cells stimulated with PMA and ionomycin were used as a control. SAP-3-stimulated iNKT cells had a higher proportion of IL-2-producing cells compared with α-GalCer-stimulated iNKT cells. Among the three donor samples we examined, SAP-3 increased the proportion of IL-2-producing cells in two samples, even compared with fresh iNKT cells stimulated with PMA+ionomycin. This finding is consistent with the increased levels of IL-2 detected in the culture supernatant upon SAP-3 stimulation (Fig. 2D), suggesting that SAP-3 skews the cytokine response of iNKT cells toward IL-2 production. In addition, a subset of IL-2-producing cells also co-produced IFN-γ and IL-4, indicating that the SAP-3-induced IL-2-producing population includes multifunctional cells.

To determine whether SAP-3-reactive iNKT cells have a distinct transcriptomic profile from α-GalCer-reactive iNKT cells, we performed paired single-cell RNA sequencing (scRNA-seq) and scTCR-seq on FACS-sorted iNKT cells stained with human CD1d-tetramer loaded with either SAP-3 or α-GalCer (Fig. 2H, I, Fig. S5C). Unsupervised clustering identified five iNKT cell clusters (Fig. 2H). Cluster 0 comprised CD4⁺, CD8⁺, or double-negative iNKT cells and represented the largest subset (84% in SAP-3 and 85% in α-GalCer). This cluster also expressed genes associated with cytotoxic lymphocytes, including NKG7, KLRB1, KLRG1, GNLY, and PRF1 (Fig. S5C). Cluster 1 was predominantly composed of CD4⁺ iNKT cells and exhibited a naïve/early activation-like population with high expression of LEF1, XCL1, and COTL1. LEF1 is a hallmark of stem-like/precursor human iNKT cells and is associated with maintenance of developmental potential and lymphoid homing. COTL1 has been implicated in actin cytoskeleton remodeling and immune cell activation, and XCL1 is a chemokine associated with early effector responses. Cluster 2 also contained CD4⁺ cells and showed high expression of many genes in cluster 1. However, cluster 2 showed lower LEF1 expression, suggesting that this cluster may represent CD4⁺ iNKT cells in a transient state. Cluster 3 is a double-negative subset and showed chemokine-producing capacity, with high expression of the chemokines CCL4 and CCL5. Cluster 4 is a double-negative cell subset and showed expression of genes associated with cell motility (MYADM) and migration (TAGLN2 & RGCC). The iNKT subsets identified in this study showed substantial overlap with previously described transcriptionally defined human PBMC iNKT subsets[29]. There was no significant difference in the distribution of iNKT cell subsets between the SAP-3 and α-GalCer-reactive groups (Fig. 2I). Therefore, enrichment of IL-2-producing iNKT cells after stimulation with SAP-3 was not due to preferential activation of iNKT cells that were already producing IL-2 but rather to the expansion of IL-2-producing cells upon the stimulation.

Sulfatide analogs activate ex vivo iNKT cells from glioblastoma patient PBMCs

In cancer patients, iNKT cells have been reported to lose the ability to produce Th1-type cytokines upon α-GalCer stimulation[30, 31]. Although IL-2 is not a “Th1-type” cytokine, we next examined cancer patient samples to determine whether sulfatide analogs would exert similar effects, particularly whether SAP-3 could function as a promising novel adjuvant to enhance anti-tumor immune responses. Furthermore, Coombs et al. reported that several sulfatides highly expressed in glioblastoma stem cells are recognized by and activate iNKT cells[24]. Building on this observation, we focused on the reactivity of PBMC-derived ex vivo iNKT cells from two patients with glioblastoma in response to the novel sulfatide analogs SAS-2 and SAP-3. Patient-derived ex vivo iNKT cells showed consistently elevated frequencies of CD69⁻CD25⁺ cells under the unstimulated (control) condition, compared with healthy donors, with both patient donors ≥10% and healthy donors ≤3% (Fig. 3A, Fig. S5A). Especially in Patient 1, the unstimulated sample showed a higher CD69⁻CD25⁺ frequency than the sulfatide-stimulated sample. On the other hand, the frequencies of CD69⁺CD25⁻ and CD69⁺CD25⁺ cells increased upon stimulation with SAP-3 in both patients (Fig. 3A). Stimulation with SAS-2 reduced the production of multiple cytokines compared to controls (Fig. 3B). In contrast, SAP-3 induced IL-2 production in both patient samples, although the magnitude of induction was weaker in Patient #2, consistent with the response observed in healthy donors (Fig. 3B), suggesting that SAP-3 effectively activates iNKT cells in glioblastoma patients as well.

Figgure 3. Sulfatide analogs activate iNKT cells from glioblastoma patient PBMCs.

PBMCs from two glioblastoma patients were stimulated with 1 μM α-GalCer, 10 μM SAS-2 or 10 μM SAP-3. (A) Frequencies of CD69⁺CD25⁻, CD69⁻ CD25⁺, and CD69⁺CD25⁺ iNKT cells (CD3⁺CD1d tetramer⁺) were assessed 4 days after 4 stimulation. (B) Cytokine levels in culture supernatants were analyzed. Signals below the detection limit for each cytokine are set to zero.

SAP-3-induced iNKT cell activation supports the induction of antigen-specific polyfunctional CD8⁺ T cells

After observing that SAP-3 induced IL-2 production, we hypothesized that SAP-3 could enhance the activation and proliferation of T cells. To test this hypothesis, we stimulated human PBMCs with the FMP-derived HLA-A2-restricted epitope, with or without SAP-3 or α-GalCer, and compared the frequency of FMP-specific CD8⁺ T cells and cytokine production[32]. Combining SAP-3 or α-GalCer with FMP did not alter the frequency of FMP-specific CD8⁺ T cells producing cytokines (IFN-γ, IL-2, or TNF-α) compared to FMP alone (Fig. 4A, Fig. S6A). However, the combination treatment increased the proportion of polyfunctional CD8⁺ T cells, producing three cytokines, IFN-γ, IL-2, and TNF-α (Fig. 4B and C, Fig. S6B). Blocking CD1d during the initial stimulation period reduced the number of polyfunctional CD8⁺ T cells, demonstrating that the effect of SAP-3 is CD1d-dependent (Fig. S6C). Notably, CD1d blockade drastically reduced the proportion of cytokine-producing cells (from 5.20% to 3.16% of CD8⁺ T cells) in culture with FMP and SAP-3. These data suggest that the sulfatide analog SAP-3 in conjunction with an antigen improves the quality of antigen-specific CD8⁺ T cell responses, likely mediated by iNKT cell activation.

Figure 4. SAP-3 improves the quality of antigen-specific CD8⁺ T cell responses.

Human PBMCs from five healthy donors were cultured with the FMP-derived epitope with or without 10 μM SAP-3, 100 nM α-GalCer, or IL-2. (A) The frequency of FMP-specific CD8⁺ T cells producing cytokines was analyzed. Each dot represents one donor (n = 5). (B) Cytokine production (IFN-γ, IL-2, and TNF-α) by FMP-specific CD8⁺ T cells was quantified using flow cytometry. The percentage of FMP-specific CD8⁺ T cells producing each cytokine is shown. Each dot represents one donor (n = 5). Data are presented as box and whisker plots showing median (line within box), interquartile range (box), and minimum to maximum values (whiskers). Statistical significance was calculated using Wilcoxon matched pairs signed rank test. *p < 0.05. (C) Pie charts illustrate the median percentage of influenza-specific CD8⁺ T cells producing each cytokine from five healthy donor PBMCs under each condition.

iNKT cell interaction with APCs through recognition of lipid antigen/CD1d complexes activates APCs. We next examined the effects of SAP-3 recognition by iNKT cells on APC populations (B cells, monocytes/macrophages, and DCs) in PBMCs and compared them with the effects of α-GalCer (Fig. S7). Although we tested only four healthy donor PBMC samples, SAP-3 significantly increased PD-L1 expression on B cells (CD19⁺CD3⁻) and increased HLA-ABC expression on monocytes/macrophages (CD3⁻CD19⁻CD11b⁺HLA-DR⁺). SAP-3 also showed a strong trend toward inducing CD40, CD86, and CD83 on B cells. However, there was no induction of activation markers in DCs (CD3⁻CD19⁻CD11b⁻CD11c⁺HLA-DR⁺). α-GalCer on the other hand significantly increased HLA-ABC and PD-L1 expression on B cells and HLA-ABC on monocytes/macrophages. α-GalCer also induced CD86 on B cells, and CD40 and PD-L1 on monocytes/macrophages, and CD86 on DCs with a strong trend. Thus, iNKT cell interaction with APCs through recognition of SAP-3 induced modest changes in APCs and could not explain SAP-3's ability to increase the polyfunctionality of antigen-specific CD8⁺ T cells.

Repeated SAP-3 stimulation did not induce iNKT-cell functional anergy, unlike α-GalCer

Since repeated stimulation with α-GalCer is known to induce iNKT cell anergy, potentially limiting its use as a vaccine adjuvant, we examined whether SAP-3 has a similar effect. Human PBMCs were cultured for 14 days and stimulated weekly with α-GalCer or SAP-3. On day 14, these PBMCs or unstimulated PBMCs were restimulated for 6 h with α-GalCer or vehicle, followed by intracellular cytokine staining. Fresh ex vivo iNKT cells that were not cultured prior to the assay produced both IFN-γ and IL-4 in response to α-GalCer (Fig. 5). In contrast, iNKT cells stimulated twice with α-GalCer prior to the assay showed reduced IFN-γ production, whereas those stimulated twice with SAP-3 showed enhanced production of IFN-γ and IL-4 against a second α-GalCer stimulation (Fig. 5). Altogether, these results suggest that SAP-3 does not induce functional anergy in human iNKT cells and could be a useful vaccine adjuvant.

Figure 5. Repeated administration of SAP-3 did not induce iNKT cell functional anergy.

Human PBMCs were stimulated twice with a one-week interval using 1 μM α-GalCer or 10 μM SAP-3 and then restimulated with α-GalCer or vehicle for intracellular staining. Fresh uncultured PBMCs, used as controls, were directly stimulated with 1 μM α-GalCer or vehicle. Shown are representative data from two independent experiments. (A) The frequency of iNKT cells was analyzed by flow cytometry. Numbers indicate the percentage of Vα24⁺Vβ11⁺ cells among CD3⁺ cells. (B) Cytokine production (IFN-γ and IL-4) by iNKT cells. Data are shown for CD3⁺Vα24⁺Vβ11⁺ iNKT cells. Numbers indicate the percentage of cells within each quadrant.

Discussion

In this study, we aimed to uncover the relationship between the structures of sulfatide analogs and their immunogenic functions against human iNKT cells. Although sulfatide is abundant in the central nervous system and selected sulfatide species can be recognized by human iNKT cells[21-24], the responses of human iNKT cells to sulfatide remain poorly understood. Therefore, we adopted a 7DW8-5–like scaffold because it has a significantly greater potency than α-GalCer[14] and synthesized eight non-natural sulfatide analogs that preserve the sulfated β-galactose headgroup while systematically varying the ceramide moiety. Among these, one sulfatide analog, SAP-3, was consistently recognized by and activated iNKT cells from all donors in tetramer staining and stimulation assays, respectively. Interestingly, several other analogs also activated iNKT cells, as evidenced by CD69 and CD25 upregulation, even though CD1d-tetramers loaded with these sulfatides did not bind to iNKT cells. It is possible that complexes of CD1d loaded with these sulfatide analogs do not have high affinity to maintain the interaction long enough to be detected by flow cytometry. This observation aligns with prior work suggesting that tetramer binding alone may underestimate the functional potential of glycolipid antigens[33]. These results demonstrate that subtle modifications of the sphingoid base and fatty acid chain in sulfatides can enhance their capacity to activate human iNKT cells. Notably, even when tetramer binding was limited, the immunogenic potential was demonstrated in an assessment of antigenicity, as evidenced by the upregulation of surface activation markers such as CD69 and CD25. This highlights the importance of incorporating functional assays, rather than relying solely on tetramer binding, when evaluating iNKT-cell agonists.

In a subset of donors, binding of the CD1d tetramer loaded with SAP-1 or SASg-1 persisted after 6B11 (Fig. S1), suggesting that tetramer binding may involve TCR-independent interactions. In one of these donors, SAP-1 or SASg-1 induced expression of CD69 and CD25, and this expression was reduced by CD1d blockade (Fig. 2C), indicating that the initial activation induced by SAP-1 or SASg-1 is mediated, at least in part, through TCR signaling. However, IL-2 production varied substantially among donors and did not reach statistical significance compared with unstimulated controls. These findings suggest that SAP-1 or SASg-1 stimulation is sufficient to induce early activation but may be insufficient to consistently elicit effector responses accompanied by cytokine production. Therefore, evaluation of antigen-specific responses based solely on CD1d tetramer loaded with SAP-1 or SASg-1 binding may not fully reflect functional cellular responses, highlighting the importance of combining tetramer analysis with functional assessments, including activation marker expression and cytokine production.

It was recently reported that C24:2 sulfatide is processed by antigen-presenting cells and converted into β-GalCer C24:2, which can activate mouse iNKT cells and contribute to antitumor immune responses [33]. Therefore, it is possible that the novel sulfatide SAP-3 may also be converted into a β-GalCer-like molecule by APCs, leading to iNKT cell activation and subsequent IL-2 production. C24:2 presented by recombinant CD1d did not stimulate mouse iNKT cells but did stimulate type II NKT cells. In contrast, in the current study, SAP-3-loaded CD1d-tetramers bound to human iNKT cells. Although we cannot rule out the possibility that a product of antigen processing of SAP-3 activates human iNKT cells, our data demonstrate that SAP-3/CD1d complexes bind to the iNKT TCR, as confirmed by single-cell TCR VDJ sequencing.

The sulfatide analog panel tested in this study (Fig. 1A) was designed to include three types of acyl chain structures and three types of sphingoid base chains, allowing us to evaluate each structure's contribution to the analogs' antigenicity. A previous study on sulfatide analogs with mouse iNKT cells indicated that structures containing phytosphingosine were more reactive to iNKT cells than those with sphingosine[21]. Therefore, we aimed to determine whether a similar effect could be observed with our sulfatide analogs against human iNKT cells. Among the phytosphingosine-containing analogs SAP-1, SAP-2, and SAP-3 with different acyl chains, SAP-3 was the only structure showing better antigenicity, as evidenced by IL-2 induction. Among the sphinganine-containing analogs SASg-1, SASg-2, and SASg-3 with different acyl chains, there was no clear difference in reactivity against iNKT cells between SASg-3, which shares the same acyl chain as SAP-3, and the other two. Taken together, when the sphingoid base was the same among analogs, no clear hierarchy based on acyl chain structure was observed in their reactivity against iNKT cells. Overall, these data suggest that the combination of sphingoid base and acyl chain critically affects the antigenicity of sulfatide analogs against human iNKT cells.

In a previous report, the combination of a malaria vaccine and α-GalCer was shown to enhance immunogenicity and efficacy of the vaccine, with the adjuvant effect of α-GalCer primarily attributed to IFN-γ production[14]. The present study comprehensively analyzed 12 cytokines in the culture supernatants of PBMC-derived ex vivo iNKT cells stimulated with sulfatide analogs, revealing that SAP-3 induced IL-2 but not IFN-γ. This IL-2-dominant cytokine profile was distinct from that of α-GalCer, which typically induces a broader range of cytokines, including IFN-γ. Importantly, SAP-3 increased the polyfunctionality of antigen-specific CD8⁺ T cells in peptide-stimulated PBMC cultures, similar to α-GalCer. This finding underscores the potential importance of IL-2-mediated pathways in improving the quality of antigen-specific CD8⁺ T cell responses. On the other hand, administration of IL-2 alone increased the frequency of FMP-specific CD8⁺ T cells but did not significantly enhance polyfunctionality of CD8⁺ T cells. This suggests that the effect of IL-2 in enhancing polyfunctionality is likely dependent on additional co-stimulatory signals provided by iNKT cells. When iNKT cells are activated by SAP-3, they not only promote IL-2 secretion but also may express additional co-stimulatory signals that may enhance the quality of antigen-specific CD8⁺ T cell responses.

In this study, blockade of the IL-2 receptor markedly increased the concentration of IL-2 in the culture supernatant (Fig. 2F), suggesting the presence of a IL-2 consuming cell population within PBMCs. Furthermore, activation of IL-2 producing iNKT cells was dependent on the concentration of SAP-3 (Fig. S5B), indicating that IL-2 production itself increased in a concentration-dependent manner. However, no apparent SAP-3 dose dependency was observed in IL-2 levels in the culture supernatant (Fig. S5B). This finding suggests that the differences in IL-2 production may have been offset by a IL-2 consuming mechanism induced by SAP-3 stimulation, resulting in the apparent absence of concentration dependency in the measured supernatant IL-2 levels.

Notably, SAP-3 did not induce iNKT-cell anergy upon repeated administration. SAP-3 preserved recall cytokine responses, maintaining IL-4 and IFN-γ production and increasing the frequency of cytokine-producing iNKT cells, unlike α-GalCer. Because the lack of anergy induction is essential for the potential clinical use of adjuvants that require multiple or repeated dosing, these properties underscore the translational potential of SAP-3. These findings suggest that SAP-3 may function as a vaccine adjuvant with a unique mechanism of action, activating iNKT cells to enhance immune responses through IL-2-mediated pathways and possibly additional co-stimulatory signals, while remaining compatible with repeated dosing.

SAP-3 also activated ex vivo iNKT cells derived from PBMCs of glioblastoma patients, consistent with findings in healthy donors. Although limited by the small sample size (n=2), the data suggest that iNKT cells from glioblastoma patients may have a higher baseline frequency of CD69⁻CD25⁺ cells than those from healthy donors. The observed variability in cytokine responses across patient samples likely reflects greater biological heterogeneity and differences in immune cell activation states. However, the small number of available patient samples precluded further validation of this variability. Nevertheless, responses to SAP-3 were comparable to those of healthy donors, suggesting that SAP-3 may serve as a promising novel adjuvant to enhance anti-tumor immune responses, including potential applications in glioblastoma immunotherapy.

In summary, we identified SAP-3 as a lead sulfatide analog with potent vaccine adjuvant activity, capable of activating human iNKT cells via an IL-2-dependent mechanism. These findings highlight its potential for further preclinical and clinical development.

Data Limitations and Perspectives

Although our findings provide clear evidence that SAP-3 activates human iNKT cells and enhances antigen-specific CD8⁺ T cell responses, several limitations should be considered. This study is limited to ex vivo assays using human PBMCs, and the immunological effects observed may not fully translate to in vivo environments. Ex vivo systems lack the complexity of whole organisms, including interactions with tissue-resident cells and metabolic factors that could influence immune responses. Furthermore, the concentration and duration of sulfatide analog exposure may not accurately reflect physiological conditions. To address these limitations, further studies using humanized mouse models or early-phase clinical trials will be essential to assess the therapeutic potential of SAP-3 in vivo.

Materials and methods

Cells and reagents

Human PBMCs were isolated from buffy coats obtained from healthy adult donors as byproducts of allogeneic blood donation through the National Institutes of Health (NIH) Blood Bank. Blood donors provided written informed consent for the use of their blood for research purposes, and blood samples were de-identified prior to distribution. Glioblastoma patients’ PBMCs were obtained as part of a clinical trial at the Neuro-Oncology Branch at the National Cancer Institute [NCT03018288] with written informed consent. The use of PBMC samples from the patients was approved by the NCI Internal Review Board (IRB ID: IRB002420). PBMCs were prepared by density centrifugation on Ficoll-Paque (Cytiva). α-GalCer (KRN 7000) and 7DW8-5 were obtained from Diagnocine. The details of sulfatide analogs synthesis are described in the supplemental material.

CD1d-Tetramer staining

CD1d monomers (NIH Tetramer Core Facility) were incubated overnight at 37°C with α-GalCer, 7DW8-5, or the sulfatide analogs (dissolved in 0.5% Tween 20) and then tetramerized using APC-conjugated streptavidin (Biolegend). PBMCs were incubated with Fc-block (Biolegend), followed by staining with CD1d-tetramers, anti-CD3, anti-CD19 (Biolegend), anti-Vα24 (Milteny Biotec), anti-Vβ11(Beckman Coulter), and 7-aminoactinomycin D (7AAD; Biolegend). In some experiments, the cells were first incubated with anti-human TCR Vα24-Jα18 (iNKT cell) antibody (clone: 6B11) (Biolegend) prior to the staining with other antibodies and CD1d-tetramers. Samples were run on Aurora (Cytek) or MACSQuant (Miltenyi Biotec) and analyzed using FlowJo (Tree Star).

Single-cell RNA sequencing and TCR sequencing

PBMCs were incubated with Fc-block (Biolegend), followed by staining with CD1d-tetramers, anti-CD3, anti-CD19 (Biolegend), anti-Vα24, anti-Vβ11(Beckman Coulter), and DAPI (4',6-diamidino-2-phenylindole). After staining, target cell populations were isolated by fluorescence-activated cell sorting (FACS). The sorted cells were subjected to single-cell gene expression and T-cell receptor (TCR) library construction using the Chromium GEM-X Single Cell 5’ Reagent Kits v3 (10x Genomics) according to the manufacturer’s protocol. Libraries were sequenced on an Illumina platform.

Single-cell sequencing data analysis

Single-cell RNA sequencing (scRNA-seq) gene expression matrices and paired T cell receptor (TCR) sequencing data were generated using the 10x Genomics platform and analyzed with the Seurat package (version 5) [34]. The transcripts from the single-cell sequencing data were aligned to the human reference data (GEX: GRCh38-2024-A, VDJ:vdj_GRCh38_alts_ensembl-7.1.0) using CellRanger [35]. Bonafide iNKT cells were identified based on expression of the TRAV10 and TRAJ18 gene segments in the TCR-seq data. Corresponding cell barcodes from TCR-seq data were matched to the gene expression data, and only TCR-confirmed iNKT cells were retained for downstream analyses.

As a first step, quality-control metrics including the number of detected genes, total unique molecular identifier (UMI) counts, mitochondrial transcript percentage, and ribosomal transcript percentage were analyzed. Cells with >10% mitochondrial transcripts or >4,000 detected genes were excluded. To avoid clustering driven by clonotype-specific TCR expression, we removed TCR gene segments from the gene expression data prior to downstream analyses. Cell-cycle scores were calculated based on the gene set described by Tirosh et al. [36]. Data were normalized, and the 2000 most variable genes were scaled. Sequence depth variation and cell-cycle effects were regressed. Principal component analysis (PCA) was performed on the scaled expression matrix, and batch correction was performed with Harmony integration[37]. The Harmony-corrected embeddings were used for unsupervised clustering.

Molecular docking and dynamics simulation

Molecular docking of the ligand to the T cell receptor-CD1d complex was performed using AutoDock Vina [38]. The co-crystal structure of α-GalCer with a shorter di-unsaturated acyl chain (C20:2) in the TCR-CD1d complex (PDB ID: 3VWJ) was used as a reference [39]. The protein and ligand system was prepared and parameterized for docking with AutoDock Tools (version 4.2) following standard protocols [40, 41]. Briefly, polar hydrogen atoms were added to the protein, and Kollman partial charges were added. In the ligand preparation, structures were generated and energy-minimized using Maestro (Schrödinger Suite, Version 14.4, Release 2025-2). Following this, all hydrogen atoms were added, and Gasteiger partial charges were added to the ligand. Up to 10 binding poses were generated with Autodock Vina, and one binding pose was selected based on a low binding energy value for further analysis. These docking procedures were validated by docking the co-crystal ligand onto the receptor.

The molecular dynamics (MD) simulation of protein-ligand interaction was performed with GROMACS (version 2024.4-gcc11.3) [42, 43] using an established protocol [44]. The ligand was parameterized with the Generalized Amber Force Field 2 (GAFF2) using the Acpype interface [45]. The protein-complex system was solvated with the Transferable Intermolecular Potential 3P (TIP3) solvent model. Ions were added to neutralize the system. Prior to running the MD simulation, the system was energy-minimized and subsequently equilibrated at 298 K and 1 bar. MD simulations were initiated, and snapshots of the system were taken at 50 ns, and the protein-ligand interaction was analyzed. The results of the simulation were analyzed using GROMACS, Visual Molecular Dynamics (VMD), [46] and Pymol (www.pymol.org).

Binding free energy calculations

The Molecular Mechanics/Poisson-Boltzmann Surface Area (MM/PBSA) based binding free energy of the protein-ligand complex was calculated using g_mmpbsa [47]. Free energies were calculated based on snapshots taken between 40 ns and 50 ns. Explicit solvent was excluded. The default parameters for the polar solvation calculations were used. Non-polar solvation free energy was estimated using a solvent-accessible surface area (SASA) based model with a surface tension coefficient of 0.022 kJ mol⁻¹ Å⁻²and probe radius of 1.4 Å, and an offset constant of 3.84928 kJ mol⁻¹. All calculations were conducted at 298 K.

All in silico docking, simulation, and binding energy calculations were performed on the National Institutes of Health (NIH) high-performance computing platform Biowulf.

PBMC stimulation assays

PBMCs from healthy donors or patients were cultured with 1 μM α-GalCer, 1 μM 7DW8-5, 1 or 1.1 μM SAP-2, or 10 μM of each sulfatide analog except SAP-2. Culture supernatants were harvested after 4 days, and cytokines in the supernatants were measured using LEGENDplex^™ Multi-Analyte Flow Assay Kits (Biolegend). PBMCs were harvested and incubated with Fc-block, followed by staining with CD1d tetramers loaded with PBS57 (NIH Tetramer Core facility), anti-CD3, anti-CD4, anti-CD8, anti-CD19, anti-B220, anti-CD69, anti-CD25 (Biolegend), anti-Vα24 (Milteny Biotec or Beckman Coulter), anti-Vβ11(Beckman Coulter), and 7AAD. iNKT cell activation was evaluated by examining CD69 and CD25 expression on iNKT cells. For CD1d blocking assays, cells were incubated in the presence of anti-CD1d Ab (1-5 μg/mL) (Biolegend).

To assess the cytokine production of iNKT cells expanded by SAP-3 or α-GalCer stimulation, human PBMCs from healthy donors (2.5×10⁶ cells/ml) were stimulated with either 1 μM α-GalCer or 10 μM SAP-3. After seven days, the cells were restimulated with the same lipid antigen in the presence of brefeldin A (Biolegend). Fresh PBMCs from the same donor, used as controls, were directly stimulated with PMA/ionomycin cocktail (TONBO biosciences) in the presence of brefeldin A. The cells were harvested and incubated with Fc-block, followed by staining with Live/Dead Aqua viability dye (Thermo Fisher Scientific), anti-CD3 (Biolegend), anti-Vα24 (Beckman Coulter), and anti-Vβ11 (Beckman Coulter). Following fixation and permeabilization, cells were stained intracellularly with antibodies against IFN-γ, IL-2 and IL-4 (Biolegend). iNKT cells were defined as CD3⁺Vα24⁺Vβ11⁺, and cytokine production was analyzed by flow cytometry.

CD25-blocking assays

PBMCs from healthy donors (2.5x10⁶ cells/ml) were cultured with 1 μM α-GalCer or 10 μM SAP-3 in the presence of an anti-CD25 antibody (10 μg/mL). Culture supernatants were harvested after 4 days, and IL-2 in the supernatant was measured by LEGENDplex^™ (Biolegend).

FMP-specific cytotoxic T lymphocytes stimulation assay

PBMCs were stimulated with FMP-derived HLA-A2-restricted epitope (sequence: GILGFVFTL, 10 μM, kindly provided by Jay A. Berzofsky) with or without 10 μM SAP-3, 100 nM α-GalCer, or IL-2. Seven days later, a fraction of cells was collected and used for flow cytometry to enumerate the number of FMP-specific CD8⁺ T cells. For the enumeration, the cells were incubated with Fc-block, followed by staining with HLA-A 0201-tetramers loaded with FMP-derived epitope (NIH Tetramer Core Facility), anti-CD3 (Biolegend), anti-CD8 (BD Biosciences), and 7AAD. The remaining fraction of the cells was incubated with PBMCs, depleted for T cells by magnetic sorting using the TCRα/β T Cell Depletion Kit (Miltenyi Biotec), and 10 μM of FMP-derived epitope in the presence of brefeldin A (Biolegend). The cells were incubated with Fc-block, followed by staining with Live/Dead aqua (Thermo Fisher Scientific), anti-CD3 (Biolegend), and anti-CD8 (BD Biosciences). The cells were then fixed, permeabilized, and stained with antibodies against IFN-γ, TNFα, and IL-2 (Biolegend).

Assessing the changes in APCs upon SAP-3 stimulation in PBMCs

Human PBMCs from healthy donors were plated in 24-well plates at 5 × 10⁶ in 2 mL of culture media per well and incubated at 37 °C for 24h with vehicle control (5 μL per well of 0.5% Tween20 PBS) or 10 μM SAP-3 or 1μM α-GalCer. After incubation, cells were stained with Zombie UV^™ Fixable Viability Dye (BioLegend) followed by incubation with human TruStain FcX^™ (Fc Receptor Blocking Solution, BioLegend) and staining with antibodies against CD3, CD19, CD11c, CD70, CD83, PD-L1 (BioLegend), CD11b, HLA-ABC, CD40, CD80, CD86 (BD Bioscience), and HLA-DR (Invitrogen) for flow cytometry analysis.

iNKT cell anergy assays after repeated α-GalCer or SAP-3 administration

Human PBMCs from healthy donors (2.5×10⁶ cells/ml) were first cultured with either 1 μM α-GalCer or 10 μM SAP-3. After 7 days, CD19⁺ cells isolated from the PBMCs of the same donor were irradiated (30 Gy) and used as APCs to restimulate the cultured cells with the same glycolipid antigen used at the start of the culture. Seven days later, freshly isolated CD19⁺ cells (used as APCs) were added again, and the cells were restimulated with α-GalCer or vehicle control in the presence of brefeldin A (Biolegend). The cells were incubated with Fc-block, followed by staining with Live/Dead Aqua viability dye (Thermo Fisher Scientific), anti-CD3 (Biolegend), anti-Vα24 (Beckman Coulter), and anti-Vβ11 (Beckman Coulter). Following fixation and permeabilization, cells were stained intracellularly with antibodies against IFN-γ and IL-4 (Biolegend). iNKT cells were defined as CD3⁺Vα24⁺Vβ11⁺, and cytokine production was analyzed by flow cytometry.

Statistical analysis

Wilcoxon matched pairs signed rank test was performed using Prism 10 (GraphPad Prism 10.4.2). Paired t-tests were performed in R (v4.6.1).

Supplementary Material

Supplementary Material

Supplementary Figure S1. Sulfatide analogs presented by CD1d are recognized by human iNKT TCRs.

Representative flow cytometry plots of PBMC-derived iNKT cells from healthy donors stained with CD1d tetramers loaded with sulfatide analogs. PBMCs were pre-incubated with or without the 6B11 antibody. CD1d tetramers loaded with α-GalCer or 7DW8-5 were used as positive controls, and vehicle-loaded CD1d tetramers were used as negative controls. iNKT cells were identified as CD19⁻CD3⁺Vα24⁺Vβ11⁺ T cells. Numbers indicate the percentage of sulfatide analog-loaded CD1d tetramer–positive cells among iNKT cells. The positivity gate was set using PBS-loaded CD1d tetramers. Data from Donor 2 (top) and Donor 3 (bottom) are shown. Figure 2 shows representative data from Donor 1.

Supplementary Figure S2. Single-cell TCR sequencing analysis of SAP-3-reactive and α-GalCer-reactive iNKT cells.

FACS-sorted CD19⁻CD3⁺CD1d-tetramer⁺iNKT cells reactive to SAP3 or α-GalCer were collected from PBMCs and subjected to scTCR-seq, followed by downstream data analysis. Pie chart showing the comparison of the usage of (A) CDR3α sequence, (B) V-gene segment and (C) J-gene segment of the β-chain.

Supplementary Figure S3. Protein-ligand interaction of SAP-3 and α-GalCer from 50ns MD simulations.

(A) SAP-3-TCR_s194-CD1d (B) α-GalCer-TCR_s194-CD1d (C) SAP-3-TCR_s90-CD1d (D) α-GalCer-TCR_s90-CD1d (E) SAP-3-TCR_s11-CD1d (F) α-GalCer-TCR_s11-CD1d. In the protein-ligand complex, the ligand is shown in cyan stick representation; the TCR CDR1α, CDR2α, and CDR3α are shown as salmon cartoons, and CDR3β is shown as light blue cartoons. CD1d (α-Helix 1 and α-Helix 2) are shown as gray cartoons. H-bond interactions are shown as magenta dotted lines.

Supplementary Figure S4. Root mean square deviation (RMSD) of the protein backbone and ligand during the 50ns simulation of the protein-ligand complexes.

(A) SAP-3-TCR_s194-CD1d (B) α-GalCer-TCR_s194-CD1d (C) SAP-3-TCR_s90-CD1d (D) α-GalCer-TCR_s90-CD1d (E) SAP-3-TCR_s11-CD1d (F) α-GalCer-TCR_s11-CD1d. RMSD of the ligand and protein complex (TCR and CD1d) are shown in red and black colors.

Supplementary Figure S5. Sulfatide analogs induce CD69 and CD25 expression on human ex vivo iNKT cells.

(A) Frequencies of CD69⁺CD25⁻ and CD69⁻CD25⁺ iNKT cells (CD3⁺ CD1d tetramer⁺) after 4 days of stimulation with various sulfatide analogs (1 μM SAP-2, or 10 μM of each sulfatide analog except SAP-2) in six donors. Data are presented as box-and-whisker plots showing median (line within box), interquartile range (box), and minimum to maximum values (whiskers). Statistical analysis was conducted using the Wilcoxon matched pairs signed rank test. *p < 0.05. (B) (Left) The frequencies of CD69⁺ iNKT cells (CD3⁺CD1d tetramer⁺) in healthy donor PBMC-derived ex vivo iNKT cells 4 days after stimulation with increasing concentrations of SAP-3 in the presence or absence of anti-CD25 antibody. (Right) Concentrations of IL-2 in culture supernatants after a 4 day-stimulation with increasing concentrations of SAP-3 or α-GalCer in the presence or absence of anti-CD25 antibody. Data are shown as the mean of technical triplicates. Upper and lower lines indicate the maximum and minimum values, respectively. (C) Heat map showing the top 10 differentially expressed genes in clusters 0-4 at a resolution of 0.4.

Supplementary Figure S6. Sulfatide analog SAP-3 improves the quality of antigen-specific CD8⁺ T cell responses.

Human PBMCs from five healthy donors were cultured with the FMP-derived peptide with or without 10 μM SAP-3, 100 nM α-GalCer, or IL-2. (A) The frequency of influenza-specific CD8⁺ T cells was analyzed in four out of the five donors. (B) Cytokine production (IFN-γ, IL-2, and TNF-α) by influenza-specific CD8⁺ T cells was quantified using flow cytometry. The pie charts show the proportion of influenza-specific CD8⁺ T cells producing each cytokine under each condition, for individual donors (n = 5). (C) Anti-CD1d was added during the initial one-week stimulation culture, and cytokine producing cells by FMP-specific CD8⁺ T cells were enumerated by intracellular cytokine staining using flow cytometry. Graphs show the number of cells obtained from culture.

Supplementary Figure S7. Effects of NKT cell activation induced by SAP-3 on APC.

Human PBMCs from four healthy donors were cultured with either vehicle, α-GalCer or SAP-3 for 24 hours. The expression of activation markers for APCs was assessed by flow cytometry. Raw single-donor values for the indicated activation markers, stratified by population (rows: B cells (CD3⁻CD19⁺Live cells), MonoMac (CD3⁻CD19⁻CD11b⁺HLA-DR⁺), DC (CD3⁻CD19⁻CD11b⁻HLA-DR⁺CD11c⁺) and stimulation condition (Unstim, aGC (1 μM α-GalCer), 10 μM SAP-3) at 24h. CD40, CD70, CD80, CD83, CD86, and PD-L1 are shown as % Positive (Freq. of Parent, capped at 100%); HLA-ABC is shown as Median Fluorescence Intensity (Geometric Mean, ×10³). Boxes show median and interquartile range; whiskers extend to the minimum/maximum values; dots represent individual donors (n = 4). P-values above each panel compare aGC or SAP-3 to Unstim by paired t-test.

Abbreviations:

iNKT cells

invariant Natural killer T cells

α-GalCer

α-galactosylceramide

FMP

influenza matrix protein

MM/PBSA

Molecular Mechanics/Poisson-Boltzmann Surface Area

Data availability

All data will be made available upon request to the corresponding author.

The output data from the Cell Ranger pipeline and the Seurat processed data used in this study are available on Zenodo (DOI: 10.5281/zenodo.21048458). Raw data of the scRNA-seq and scTCR-seq may be provided upon reasonable request.