S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

T CELLS Copyright © 2024 The

Authors, some rights
Serine enrichment in tumors promotes regulatory T cell reserved; exclusive
licensee American
accumulation through sphinganine-­mediated Association for the
Advancement of
regulation of c-­Fos Science. No claim to
original U.S.
Government Works
Sicong Ma1*†, Roger Sandhoff2†, Xiu Luo3†, Fuwei Shang4,5†, Qiaozhen Shi6, Zhaolong Li3,
Jingxia Wu1, Yanan Ming1, Frank Schwarz7, Alaa Madi8,9, Nina Weisshaar10,9, Alessa Mieg10,9,
Marvin Hering10,9, Ferdinand Zettl10,9, Xin Yan8,9, Kerstin Mohr10, Nora ten Bosch10, Zhe Li11,
Gernot Poschet12, Hans-­Reimer Rodewald4, Nina Papavasiliou8, Xi Wang6*, Pu Gao3*,
Guoliang Cui1,10*

CD4+ regulatory T (Treg) cells accumulate in the tumor microenvironment (TME) and suppress the immune system.

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
Whether and how metabolite availability in the TME influences Treg cell differentiation is not understood. Here, we
measured 630 metabolites in the TME and found that serine and palmitic acid, substrates required for the synthe-
sis of sphingolipids, were enriched. A serine-­free diet or a deficiency in Sptlc2, the rate-­limiting enzyme catalyzing
sphingolipid synthesis, suppressed Treg cell accumulation and inhibited tumor growth. Sphinganine, an interme-
diate metabolite in sphingolipid synthesis, physically interacted with the transcription factor c-­Fos. Sphinganine
c-­Fos interactions enhanced the genome-­wide recruitment of c-­Fos to regions near the transcription start sites of
target genes including Pdcd1 (encoding PD-­1), which promoted Pdcd1 transcription and increased inducible Treg
cell differentiation in vitro in a PD-­1–dependent manner. Thus, Sptlc2-­mediated sphingolipid synthesis translates
the extracellular information of metabolite availability into nuclear signals for Treg cell differentiation and limits
antitumor immunity.

INTRODUCTION alterations in nutrient profiles influence antitumor T cell responses

T cell–based immunotherapeutic regimens have substantial prom- (7–12). Thus, a systemic, large-­scale, and unbiased quantification of
ise in cancer treatment, but the efficacy of currently available im- metabolites in the TME is necessary to identify immunosuppressive
munotherapies is limited by the immunosuppressive nature of the metabolites and to reveal potential metabolic pathways as new targets
tumor microenvironment (TME). CD4+ regulatory T (Treg) cells to stimulate immune responses to tumor cells.
express the transcription factor forkhead box P3 (FoxP3) (1–3) and Sphingolipids are not only important structural lipids in the plas-
play an essential role in suppressing antitumor immune responses. ma membrane but also regulate immune cell migration, differentia-
Transient depletion of Treg cells decreases tumor growth (4); there- tion, and function (13–21). Serine and palmitoyl coenzyme A, a
fore, understanding the mechanisms through which Treg cells dif- metabolic derivative from palmitic acid, are condensed by the bio-
ferentiate and accumulate in the TME is important to reverse synthetic enzyme serine palmitoyltransferase (SPT), thus forming
immunosuppression in the TME and to develop more potent im- 3-­keto-­sphinganine (3-­KDS), which is then converted to sphinga-
munotherapies. nine. Sphinganine further fuels the biosynthesis of more complex
Nutrient profiles in the TME are very different from those in sphingolipids (22). Whether and how the sphingolipid synthetic
noncancerous tissues (5, 6). Accumulating evidence suggests that the pathway regulates the accumulation of Treg cells in the TME remains
unclear. (22).
1
In this study, we quantified 630 metabolites in the tumor inter-
Key Laboratory of Immune Response and Immunotherapy, Center for Advanced In-
terdisciplinary Science and Biomedicine of IHM, School of Basic Medical Sciences,
stitial fluid (TIF) and plasma of tumor-­bearing mice. We identified
Division of Life Sciences and Medicine, University of Science and Technology of China, an enrichment of sphingolipids in the TIF, as well as increased ser-
Hefei, 230601, China. 2Lipid Pathobiochemistry Group (A411), 69120 Heidelberg, ine and palmitic acid, the substrates for de novo sphingolipid syn-
Germany. 3CAS Key Laboratory of Infection and Immunity, CAS Center for Excellence thesis. By using a mouse model with Treg cell–specific deficiency in
in Biomacromolecules, National Laboratory of Biomacromolecules, Institute of Bio-
physics, Chinese Academy of Sciences, Beijing 100101, China. 4Cellular Immunology SPT long-­chain base subunit 2 (Sptlc2), and by feeding mice a
(D110), German Cancer Research Center, 69120 Heidelberg, Germany. 5Faculty of serine-­free diet, we found that serine enrichment in the TME pro-
Medicine, Heidelberg University, Heidelberg, Germany. 6State Key Laboratory of moted Treg cell accumulation in the TME in an Sptlc2-­dependent
Reproductive Medicine, Nanjing Medical University, Nanjing, China. 7Core Facility
Antibodies (W170), German Cancer Research Center, 69120 Heidelberg, Germany.
manner. Further analysis revealed that the metabolite sphinganine,
8
Immune Diversity (D150), German Cancer Research Center, 69120 Heidelberg, whose biosynthesis depends on Sptlc2, promoted Treg cell differen-
Germany. 9Faculty of Biosciences, Heidelberg University, Heidelberg, Germany. tiation. Sphinganine physically bound c-­Fos protein in a manner
10
T Cell Metabolism (D192), German Cancer Research Center, 69120 Heidelberg, dependent on three critical amino acids: threonine-­164 (T164),
Germany. 11Division of Pathogenesis of Virus Associated Tumors (F100), German Can-
cer Research Center (DKFZ), 69120 Heidelberg, Germany. 12Metabolomics Core Tech- asparagine-­185 (N185), and leucine-­187 (L187). Sphinganine pro-
nology Platform, Centre for Organismal Studies (COS), Heidelberg University, 69120 moted the recruitment of c-­Fos to the promoter region of the target
Heidelberg, Germany. gene Pdcd1 (encoding PD-­1). PD-­1 was required for the effect of
*Corresponding author. Email: s.​ma@​ihm.​ac.​cn (S.M.); xiwang@​njmu.​edu.​cn (X.W.);
gaopu@​ibp.​ac.​cn (P.G.); g.​cui@​ihm.​ac.​cn (G.C.) sphinganine on increasing FoxP3 expression. Overall, this study
†These authors contributed equally to this work. revealed that serine enrichment promotes Treg cell accumulation in

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 1 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

the TME through sphinganine. Sphinganine functions as a second Tumor-­infiltrating Treg cells express the sphingolipid
messenger molecule that translates information on serine enrich- synthetic enzyme Sptlc2 at high levels, and Sptlc2
ment in extracellular compartments into intracellular cues promot- expression is induced by antigenic stimulation
ing Treg cell differentiation. To study whether Treg cells adapted to the enrichment of substrates for
sphingolipid synthesis in the TME, we monitored the expression of
Sptlc2, a rate-­limiting enzyme catalyzing sphingolipid synthesis in
RESULTS tumor-­infiltrating Treg cells. We implanted C57BL/6 mice with B16
Metabolomics analysis of the TME reveals enrichment of melanoma cells and performed flow cytometry staining to measure
substrates for sphingolipid synthesis Sptlc2 protein levels in tumor-­infiltrating Treg cells and splenic Treg
To comprehensively profile metabolites in the TME, we used a cells. Tumor-­infiltrating Treg cells expressed Sptlc2 protein at signifi-
Biocrates MxP Quant 500 kit to measure the concentrations of 630 cantly higher levels than did splenic Treg cells (Fig. 2, A and B), as
metabolites in the TIF of mice subcutaneously implanted with B16 further confirmed by Western blot analysis (Fig. 2C). These results
melanoma, Yale University Mouse Melanoma Exposed to Radia- suggested that tumor-­infiltrating Treg cells tuned their expression of a
tion (YUMMER), and MC-­38 colon tumors (Fig. 1, A to C), as well sphingolipid synthetic enzyme and adapted to the increased availabil-
as mice with MC-­38 orthotopic tumors (23) or inducible melano- ity of substrates for sphingolipid synthesis in the TME.

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
ma BrafV600E PtenFlox/Flox Tyr-­CreERT2 (also known as Braf/Pten) To examine whether antigenic stimulation or cytokines increased
(fig. S1, A and B) (24). Plasma samples from the same tumor-­ Sptlc2 expression in Treg cells, we cultured splenocytes of naïve mice
bearing mice were included as controls. We performed principal with or without anti-­CD3 and a panel of cytokines. Anti-­CD3 sig-
components analysis (PCA) of the metabolite profiles in TIF and nificantly increased the expression of Sptlc2 in Treg cells (Fig. 2D).
plasma. The metabolite profiles in TIF were markedly different Cytokines, such as interleukin-­2 (IL-­2), IL-­4, and tumor necrosis
from those in the plasma, forming distinct clusters (Fig. 1, A to C, factor–α (TNF-­α), increased Sptlc2 expression to a much lesser ex-
and fig. S1, A and B). The “fatty acids and related” group of me- tent. Non-­Treg CD4+ T cells shared a similar expression pattern of
tabolites includes fatty acids and acylcarnitines. The “sphingolip- Sptlc2 (Fig. 2D). Collectively, these results suggested that antigen po-
ids” group of metabolites includes ceramides, dihydroceramides, tently increased Sptlc2 expression levels in both Treg cells and non-­
hexosylceramides, dihexosylceramides, trihexosylceramides, and Treg CD4+ T cells.
sphingomyelins. The “phospholipids” group of metabolites in- To further study whether the sphingolipid synthesis pathway was
cludes phosphatidylcholines and lysophosphatidylcholines. Com- functional in Treg cells, we treated Treg cells with 13C315N1 tracer–
pared with plasma samples, TIF samples had higher concentrations labeled serine (Fig. 2E). Because non-­Treg CD4+ T cells, similarly to Treg
of amino acids (Fig. 1, A to C, and fig. S1, A and B), a finding cells, increased Sptlc2 expression levels when cultured with different
reminiscent of those from a previous study showing that several stimuli (Fig. 2D), we included non-­Treg CD4+ T cells as a control in
amino acids were enriched in pancreatic tumors (5). Furthermore, this assay. We purified splenic Treg cells and non-­Treg CD4+ T cells
lipids such as fatty acids and sphingolipids were more abundant in from naïve C57BL/6 wild-­type mice. We stimulated Treg cells and
the TIF than in plasma (Fig. 1, A to C), as previously reported (7, non-­Treg CD4+ T cells with anti-­CD3 and anti-­CD28 in the presence
25). Biogenic amines, including spermine, spermidine, putrescine, of 13C315N1 tracer–labeled serine then analyzed the newly synthesized
β-­alanine, histamine, and γ-­aminobutyric acid, were also enriched sphingolipids by mass spectrometry. We reasoned that if the sphingo-
in TIF, whereas serotonin was enriched in plasma (Fig. 1, A to C, lipid synthesis was active in Treg cells, 13C and 15N tracers would be
and table S1). Choline was enriched in TIF in all three tumor mod- incorporated into the downstream metabolites (Fig. 2F). 13C and 15N
els, in agreement with a reported link between choline metabolism tracers were readily detectable in all metabolites in the sphingolipid
and malignant transformation (26). Compared with TIF, plasma synthetic pathway in both Treg cells and non-­Treg CD4+ T cells, and
had higher concentrations of hexoses, a group of metabolites in- antigenic stimulation increased the incorporation of tracers into
cluding glucose, in agreement with previous reports (10, 12, 27). sphingolipids (Fig. 2G). Together, these results suggested that the
Sphingolipid metabolism has been shown to regulate T cell dif- sphingolipid synthetic pathway is functional in Treg cells.
ferentiation and function (19–21). Sphingolipid de novo synthesis
is initiated by the condensation of serine and palmitoyl coenzyme Treg cell–specific ablation of Sptlc2 decreases Treg cell
A, which is derived from its precursor molecule palmitic acid, thus accumulation in tumors and enhances antitumor
forming 3-­KDS, as catalyzed by the rate-­limiting enzyme Sptlc2 T cell responses
and other SPT subunits. Subsequently, 3-­ KDS is converted to To examine whether Sptlc2-­mediated de novo sphingolipid synthesis
sphinganine and more complex sphingolipids (Fig. 1D). The con- is required for Treg cell–mediated immunosuppression in tumors, we
centration of serine was significantly higher in TIF samples than created a mouse strain with a Treg cell–specific deficiency in Sptlc2 by
plasma samples from mice bearing B16, YUMMER, and MC-­38 breeding Foxp3YFP-­Cre mice with Sptlc2Fl/Fl mice (28, 29). Compared
tumors (Fig. 1E). Palmitic acid was also significantly enriched in with their Sptlc2-­sufficient Sptlc2+/+Foxp3YFP-­Cre littermates, Sptlc2-­
B16 TIF and YUMMER TIF compared with plasma and showed a deficient Sptlc2Fl/FlFoxp3YFP-­Cre mice had similar numbers of thymo-
nonsignificant trend of enrichment in MC-­38 TIF (Fig. 1E). The cytes, inguinal lymphocytes, and splenocytes (fig. S2A). Likewise, Treg
MC-­38 orthotopic tumor model and Braf/Pten inducible melano- cell–specific Sptlc2 deficiency had no effects on the numbers of CD4+
ma models also showed similar patterns (fig. S1D). Collectively, or CD8+ T cells or on the expression of surface markers, including
these results suggested that substrates for the sphingolipid synthet- CD62L and CD44 (fig. S2B). However, Treg cell–specific Sptlc2 defi-
ic pathway were enriched in the TME, thus prompting us to further ciency decreased the total number of Treg cells, particularly a subset of
study whether and how tumor-­infiltrating T cells adapted to this CD62L+ Treg cells (30). Furthermore, Sptlc2 deficiency impaired Treg
environmental metabolic change. cell–mediated suppressive functions in coculture assays (fig. S2C).

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 2 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024

Fig. 1. Substrates for the de novo synthesis of sphingolipids are enriched in the tumor microenvironment. Mice were implanted with B16-­F10 melanoma, YUMMER
melanoma, or MC-­38 colon tumor cells (2 × 105 cells per mouse). TIF and plasma samples were collected for metabolomics analysis. (A to C) Scattered dot plots (top) docu-
menting PCA of different metabolite profiles between the TIF and plasma of mice bearing B16-­F10 melanoma (A), YUMMER melanoma (B), or MC-­38 colon tumors (C). PCA
was performed with MetaboAnalyst (v5.0) online software. Heatmaps (bottom) show the z-­scores of 630 metabolites in the TIF and plasma, quantified with a Biocrates MxP
Quant 500 kit. (D) Illustration of the sphingolipid synthetic pathway. Serine and palmitic acid, two substrates for synthesizing sphingolipids, are highlighted in red. (E) Bar
graphs showing the concentrations of serine and palmitic acid in TIF and plasma collected from mice bearing B16-­F10 melanoma, YUMMER melanoma, or MC-­38 colon tu-
mors. Data are presented as the means ± SD and are pooled from eight (B16-­F10 melanoma), three (YUMMER melanoma), and four (MC-­38 colon tumor) mice. *P < 0.05;
****P < 0.0001. Statistical analysis was performed with unpaired two-­tailed Student’s t tests (both serine and palmitic acid in B16-­F10 melanoma and YUMMER melanoma;
data points were normally distributed) or the Mann-­Whitney U test (both serine and palmitic acid in MC-­38 colon tumors; data points were not normally distributed).

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 3 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024

Fig. 2. reg cells in tumors express the sphingolipid synthetic enzyme Sptlc2 at high levels and increase sphingolipid synthesis after antigenic stimulation. T (A and
B) Flow cytometry contour plots documenting gating strategies (A) and the expression of Sptlc2 protein in FoxP3+ Treg cells in spleens or tumors from C57BL/6 mice bearing
B16-­F10 tumors (B). Bar graphs show cumulative data from two independent experiments with six mice in total. (C) Sptlc2 protein in cell lysates was analyzed with western
blotting. Grp94 was included as a loading control. (D) Splenocytes of naïve C57BL/6 mice were cultured with or without anti-­CD3 or cytokines for 24 hours. Bar graphs show-
ing the expression levels of Sptlc2 protein in Treg cells and non-­Treg CD4+ T cells. Data are means ± SD and are cumulative from three independent experiments, with one
mouse per experiment, n = 3 mice in total. (E) Experimental design used to measure sphingolipid synthetic flux in Treg cells and non-­Treg CD4+ T cells. (F) Illustration of de novo
sphingolipid biosynthesis, with the tracers highlighted in red. (G) Heatmaps showing z-­scores of incorporation of the 13C and 15N tracers into sphingolipids in both Treg cells
and non-­Treg CD4+ T cells that were or were not treated with anti-­CD3 and anti-­CD28. Each row represents the results from a single sample pooled from 10 mice. Three pairs
of cell samples were prepared independently (total n = 30 mice), and the abundance of tracer-­labeled sphingolipids was measured together in a single experiment. *P < 0.05;
**P < 0.01; ****P < 0.0001. Statistical analysis was performed with unpaired two-­tailed Student’s t tests (B; data points were normally distributed) or one-­way analysis of vari-
ance (ANOVA) [(D), a comparison was between the “medium” group and the other groups].

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 4 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

Sptlc2 deficiency showed a modest but not significant trend toward mice. The tumors that developed in the Sptlc2Fl/FlFoxp3YFP-­Cre mice
decreasing the percentages of Treg cells among CD4+ T cells in non- were much smaller than those in littermate controls (Fig. 3, A and B).
lymphoid tissues, such as fat and lung (fig. S3, A and B). These find- Expression of the inhibitory receptors PD-­1 and T cell immunorecep-
ings suggested that, although Sptlc2 deficiency decreased Treg cells tor with Ig and ITIM domains (TIGIT) on CD8+ TILs, CD4+ TILs, and
and affected their suppressive functions in vitro, Sptlc2 expression in tumor-­infiltrating Treg cells was lower in the Sptlc2Fl/FlFoxp3YFP-­Cre mice,
Treg cells was dispensable for the development of thymic T cells and as was the percentage of Treg cells within the population of CD4+ TILs
for the maintenance of systemic homeostasis in vivo under steady-­ (Fig. 3, C to E). Although the percentages of cytokine-­producing cells
state conditions. in TILs between Sptlc2+/+Foxp3YFP-­Cre mice and Sptlc2Fl/FlFoxp3YFP-­Cre
To explore the effects of Treg cell–specific Sptlc2-­deficiency on Treg mice were comparable, the densities of cytokine-­producing CD8+
cell–mediated immunosuppression in tumors, we subcutaneously im- TILs and CD4+ TILs significantly increased under Sptlc2 deficiency
planted B16 melanoma cells into Sptlc2-­deficient Sptlc2Fl/FlFoxp3YFP-­Cre (Fig. 3, C and D). In the tamoxifen-­inducible mouse model, Sptlc2Fl/
mice and Sptlc2-­sufficient Sptlc2+/+Foxp3YFP-­Cre littermate control Fl
Foxp3CreERT2 mice grew much smaller tumors than littermate

Fig. 3. Treg cell–specific ablation of Sptlc2

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
enhances antitumor T cell responses. Sptlc2Fl/
Fl
Foxp3YFP-­Cre (Fl/Fl) or Sptlc2+/+Foxp3YFP-­Cre (+/+)
littermates were implanted with 2 × 105 B16-­F10
melanoma cells. (A) Line graph showing tumor
growth over time in Sptlc2Fl/FlFoxp3YFP-­Cre (n = 13)
mice and Sptlc2+/+Foxp3YFP-­Cre (n = 14) mice (left).
Bar graph showing the areas under the curves
(AUCs) of the tumor growth plots (right). (B) Bar
graph showing tumor weight. (C to E) Flow cy-
tometry plots and bar graphs showing the expres-
sion of the indicated proteins in CD8+ TILs (C),
CD4+ TILs (D), and tumor-­infiltrating Treg cells (E).
Cells were cultured with phorbol 12-­myristate
13-­acetate and ionomycin in the presence of
brefeldin A for 6 hours before flow cytometry
analysis of IFN-­γ and TNF-­α. Tumors smaller than
30 mg were pooled to obtain a sufficient num-
ber of cells for flow cytometry analysis. Thirteen
Sptlc2+/+Foxp3YFP-­Cre samples and eight Sptlc2Fl/
Fl
Foxp3YFP-­Cre samples were used in the flow cytom-
etry analysis of PD-­1, TIGIT, IFN-­γ, and TNF-­α. Data
are cumulative (A to E) from two independent ex-
periments. The end point of both experiments was
day 27 after tumor implantation. (F and G) Sptlc2Fl/
Fl
Foxp3CreERT2 or Sptlc2+/+Foxp3CreERT2 littermates
were implanted with 2 × 105 B16-­F10 melanoma
cells. Mice received tamoxifen (2 mg per mouse,
intraperitoneal injection) on days 15, 17, and 19 af-
ter tumor cell implantation. A line graph (left) shows
tumor growth over time in Sptlc2Fl/FlFoxp3CreERT2
(n = 20) mice and Sptlc2+/+Foxp3CreERT2 (n = 20) mice
(F). A bar graph (right) shows the AUCs of the tumor
growth plots. A bar graph shows tumor weight (G).
The results are presented as means ± SEM. *P < 0.05;
**P < 0.01; ****P < 0.0001. Statistical analysis was
performed with the Mann-­Whitney U test [% PD-­1 in
(D); data points were not normally distributed] or
unpaired two-­tailed Student’s t tests (the rest of the
analysis is in this figure; data points were normally
distributed).

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 5 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

controls (Fig. 3, F and G), further confirming that Sptlc2 is required for diet had lower percentages, numbers, and densities of tumor-­
Treg cell–mediated immunosuppression in tumors. infiltrating Treg cells than Sptlc2+/+Foxp3YFP-­Cre mice fed a chow diet
To examine whether naturally occurring Treg (nTreg) cells or in- (Fig. 4, D to F), thus suggesting that serine enrichment was required
ducible Treg (iTreg) cells mediate the protumor growth phenotypes for Treg cell accumulation in the TME. In contrast, Treg cells in tumors
observed in (Fig. 3A), we adoptively transferred purified Sptlc2-­ from Sptlc2Fl/FlFoxp3YFP-­Cre mice were not affected by serine restric-
deficient or -­sufficient non-­Treg CD4+ T cells to Tcrb−/− mice before tion. These results suggested that serine enrichment-­dependent Treg
implanting B16 tumor cells into these Tcrb−/− mice (fig. S4A). Be- cell accumulation in the TME is dependent primarily on Sptlc2.
cause Tcrb−/− mice lack Treg cells, all the Treg cells detected in these Serine restriction significantly increased the concentrations of
host mice resulted from iTreg cell differentiation from the non-­Treg 1-­deoxy-­dihydroceramide (1-­deoxy-­DH-­Cer) (fig. S6A), in line
CD4+ donor T cells. Sptlc2 deficiency in iTreg cells significantly re- with previous findings (37). The abundance of 1-­deoxy-­methyl-
duced the tumor size and weight (fig. S4, B and C). The percentages dihydroceramide (1-­deoxy-­methyl-­DH-­Cer) was not influenced
of Treg cells among CD4+ TILs were significantly decreased by Sptlc2 by either serine restriction or Sptlc2 deficiency (fig. S6B). The de-
deficiency (fig. S4D), suggesting that Sptlc2 deficiency affected iTreg oxysphingoid bases 1-­deoxy-­sphinganine (1-­deoxy-­SA) and 1-deoxy-
cell differentiation in vivo. methyl-­sphinganine (1-­deoxy-­methyl-­SA) were undetectable. Because
Because Sptlc2 is required for effector T (Teff) cell function (20), we deoxysphingolipids are toxic to tumor cells, an alternative explanation

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
further investigated its impact at a pan-­T cell level. We implanted for the observed lack of difference in tumor growth between Sptlc2Fl/
B16-­F10 melanoma cells into mice that harbor a T cell–specific Fl
Foxp3YFP-­Cre mice or Sptlc2+/+Foxp3YFP-­Cre littermates fed a serine-­free
deficiency of Sptlc2 (Sptlc2Fl/FlCd4Cre) and their Sptlc2-­sufficient diet (the right tumor growth panel of Fig. 4B) is that serine restric-
(Sptlc2+/+Cd4Cre) littermates. Cd4Cre enables the excision of loxP-­ tion caused accumulation of toxic deoxysphingolipids, and deoxysphin-
flanked (Fl/Fl) DNA sequences in CD4+CD8+ thymocytes and results golipids reduced the sizes of tumor in Sptlc2+/+Foxp3YFP-­Cre mice.
in the deletion of loxP-­flanked Sptlc2 in mature CD4+ and CD8+ As a result, the tumor growth in Sptlc2Fl/FlFoxp3YFP-­Cre mice and
T cells (31, 32). We observed accelerated tumor growth in the Sptlc2Fl/ Sptlc2+/+Foxp3YFP-­Cre littermates fed a serine-­free diet was comparable.
Fl
Cd4Cre mice compared with their Sptlc2+/+Cd4Cre littermate controls To test whether the comparable tumor growth under serine restriction
(fig. S5A). The pan T cell–specific deficiency in Sptlc2 was associated was because of impaired Treg cell–mediated immunosuppression of
with significantly reduced densities of both CD8+ and CD4+ tumor-­ CD8+ T cells or direct toxicity to tumor cells, we depleted CD8+ T cells
infiltrating lymphocytes (TILs) (fig. S5, B and C). This phenotype is (fig. S7A). Sptlc2Flox/FloxFoxp3Cre mice treated with IgG isotype control
reminiscent of a previous study showing that Sptlc2 deficiency affect- exhibited smaller tumors than Sptlc2+/+Foxp3Cre littermates on a chow
ed Teff cell survival and proliferation in viral infections (20). diet (fig. S7B). When fed a serine-­free diet, both groups showed similar
tumor sizes (fig. S7B). Furthermore, both Sptlc2 deficiency and serine
Serine promotes Treg cell accumulation in the TME in a starvation reduced the percentages of Treg cells (fig. S7C), consistent
manner primarily dependent on Sptlc2 with results shown above (Fig. 4, B and D). Upon CD8+ T cell deple-
To determine whether serine enrichment in the TME was required tion, no significant difference in tumor sizes was observed between
for Treg cell accumulation, we next fed mice a serine-­free diet. Because Sptlc2Flox/FloxFoxp3Cre mice and Sptlc2+/+Foxp3Cre littermates (fig. S7B),
glycine is a precursor molecule for serine synthesis (33), we depleted suggesting that CD8+ T cells were primarily responsible for the reduc-
both serine and glycine from the diet, as previously reported (34–36). tion of tumor sizes in Sptlc2Flox/FloxFoxp3Cre mice. When CD8+ T cells
Serine is a substrate in multiple metabolic pathways in T cells beyond were depleted, there was a trend toward smaller tumor size in mice fed
Sptlc2-­mediated sphingolipid biosynthesis, including the folate and a serine-­free diet compared with those fed a chow diet (fig. S7B), sug-
methionine cycles as well as those associated with glutathione biogen- gesting that although CD8+ T cells were primarily responsible for the
esis (35, 36). To examine whether Sptlc2 played a major role in medi- reduction of tumor sizes in the Sptlc2Flox/Flox Foxp3Cre mice, other fac-
ating serine enrichment-­dependent Treg cell accumulation in the tors may also contribute to the tumor control in this experimental
TME, we fed the Sptlc2-­deficient Sptlc2Fl/FlFoxp3YFP-­Cre mice and setting.
Sptlc2-­sufficient Sptlc2+/+Foxp3YFP-­Cre littermates a serine-­free diet or
a regular chow diet for 1 month before implanting mice with B16 Sphinganine promotes mouse and human Treg
melanoma cells (Fig. 4A). When fed a chow diet, Sptlc2Fl/FlFoxp3YFP-­Cre cell differentiation
mice grew significantly smaller tumors than Sptlc2+/+Foxp3YFP-­Cre To explore the mechanisms through which Sptlc2-­mediated sphingo-
mice (Fig. 4, B and C). In contrast, Sptlc2Fl/FlFoxp3YFP-­Cre mice and lipid synthesis promotes Treg cell accumulation in tumors, we hypoth-
Sptlc2+/+Foxp3YFP-­Cre mice grew tumors of similar sizes when fed a esized that the direct and indirect metabolic products of Sptlc2
serine-­free diet (Fig. 4, B and C), suggesting that Sptlc2 was required enhance Treg cell differentiation. To test this hypothesis, we differenti-
for Treg cells to regulate tumor growth only when serine was enriched ated Treg cells in the presence or absence of a panel of metabolites in
in the TME. the sphingolipid synthetic pathway (Fig. 5A). We found that 3-­KDS,
The percentages of Treg cells among CD4+ TILs, the absolute num- the direct product of the Sptlc2-­mediated biochemical reaction, sig-
bers of Treg cells, and the densities of Treg cells in tumors were signifi- nificantly increased the expression of FoxP3 (Fig. 5B). Sphinganine, a
cantly decreased under Sptlc2 deficiency when mice were fed a chow metabolite derived from 3-­KDS, promoted FoxP3 expression more
diet (Fig. 4, D to F). When fed a serine-­free diet, Sptlc2+/+Foxp3YFP-­Cre potently than 3-­KDS. Sphinganine did not significantly influence
mice and Sptlc2Fl/FlFoxp3YFP-­Cre mice had comparable percentages T cell proliferation (fig. S8, A and B) or the phosphorylation of S6
(Fig. 4D), numbers of tumor-­infiltrating Treg cells (Fig. 4E), and densi- protein, a surrogate of the mTOR activity (fig. S9). Sphinganine can be
ties of Treg cells in tumors (Fig. 4F). These results suggested that Sptlc2 converted into dihydroceramide (Fig. 5A) through catalysis by the
was required for Treg cells to accumulate in the TME only when serine enzyme ceramide synthase. The addition of fumonisin B1, an in-
was enriched in the TME. Sptlc2+/+Foxp3YFP-­Cre mice fed a serine-­free hibitor of ceramide synthase (38), did not significantly affect the

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 6 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
Fig. 4. reg cells require Sptlc2 to accumulate in tumors only when serine is enriched in the tumor microenvironment. T (A) Illustration of the experimental design. Sptl-
c2Fl/FlFoxp3YFP-­Cre or Sptlc2+/+Foxp3YFP-­Cre littermate mice were fed a regular chow diet or a serine-­free diet for 4 weeks before being implanted with 2 × 105 B16-­F10 melanoma
cells. Mice were continuously fed a chow diet or a serine-­free diet after melanoma cell implantation until the endpoints. (B to F) Line graphs showing tumor growth in Sptlc2Fl/
Fl
Foxp3YFP-­Cre mice and Sptlc2+/+Foxp3YFP-­Cre littermates fed the indicated diets. Bar graphs showing AUCs of the tumor growth plots (B). Bar graphs showing the tumor weight
(C), the number of Treg cells (E), and the density of Treg cells in tumors (F). Flow cytometry plots showing the percentages of Treg cells among CD4+ TILs (D). Sptlc2+/+Foxp3YFP-­Cre
mice (n = 9) and Sptlc2Fl/FlFoxp3YFP-­Cre littermates (n = 9) were fed a chow diet. Sptlc2+/+Foxp3YFP-­Cre mice (n = 10) and Sptlc2Fl/FlFoxp3YFP-­Cre littermates (n = 9) were fed a serine-­
free diet. One Sptlc2Fl/FlFoxp3YFP-­Cre mouse fed a serine-­free diet was euthanized on day 21 because of tumor bleeding. Therefore, eight mice were used for tumor weight
measurement and Treg cell counting on day 25. Three small tumors from the group of Sptlc2Fl/FlFoxp3YFP-­Cre mice fed a chow diet were combined as one biological sample to
yield sufficient TILs for flow cytometry analysis. Similarly, two small tumors from the group of Sptlc2Fl/FlFoxp3YFP-­Cre mice fed a serine-­free diet were combined as one biological
sample for flow cytometry analysis (D to F). Results are presented as means ± SD (B, C, E, and F). *P < 0.05; **P < 0.01; n.s., not significant. Statistical analysis was performed
with Mann-­Whitney U tests [analysis of AUC of the “chow diet” groups in (B); data points were not normally distributed], unpaired two-­tailed Student’s t tests [analysis of AUC
of the serine-­free diet groups in (B); data points were normally distributed]. The rest of the statistical analysis was performed with a two-­way ANOVA.

sphinganine-­induced expression of FoxP3 protein (Fig. 5C). These To investigate the role of Sptlc2 in Treg cells in inflammatory re-
results suggested that the conversion of sphinganine into dihydrocer- sponses, we used an experimental autoimmune encephalomyelitis
amide was not necessary to promote sphinganine-­mediated Treg cell (EAE) model. We observed that Treg cells required Sptlc2 to suppress
differentiation. We also differentiated Treg cells in the presence of di- autoinflammation. EAE development in the Sptlc2Fl/FlFoxp3YFP-­Cre
hydroceramide and ceramide. We found that these metabolites did mice was accelerated, and the symptoms were more severe (fig. S12A).
not significantly increase FoxP3 expression (Fig. 5B). Sphinganine The increase in EAE clinical scores was accompanied by reduced Treg
enhanced Treg cell differentiation in a dose-­dependent manner cell percentages in the spinal cord (fig. S12B). Furthermore, spinal
(Fig. 5D). Furthermore, sphinganine also suppressed the differentia- cord–infiltrating CD4+ T cells produced more inflammatory cyto-
tion of IL-­17–producing helper T (TH17) cells and enhanced the ex- kines in the Sptlc2Fl/FlFoxp3YFP-­Cre mice than in the control littermates
pression of FoxP3 under conditions promoting polarization toward (fig. S12C). Collectively, these results suggest that Sptlc2 promotes
the TH17 cell lineage (Fig. 5E), accompanied by reduced expression of Treg cell–mediated immunosuppression in autoimmune diseases.
RORγt (fig. S10, A and B). Sphinganine did not influence the stability
of nTreg cells (fig. S11, A and B). Similar to results from experiments Sphinganine increases FoxP3 expression in a manner
performed in murine T cell cultures, sphinganine dose-­dependently dependent on the c-­Fos–PD-­1 signaling axis
enhanced human Treg cell differentiation and suppressed the ex- To identify mechanisms through which sphinganine promotes the
pression of inflammatory cytokine IL-­17 and interferon-­γ (IFN-­γ) expression of FoxP3, we performed single-­cell RNA sequencing
(Fig. 5F). These results suggested that the proximal metabolites down- (scRNA-­seq) to analyze the transcriptional profiles of TCR-­transgenic
stream of Sptlc2, particularly sphinganine, promoted both mouse and OT-­II CD4+ T cells. FoxP3YFP × OT-­II splenocytes were cultured un-
human Treg cell differentiation. der Treg cell polarization conditions in the presence of sphinganine

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 7 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
Fig. 5. Sphinganine increases the expression of FoxP3 and decreases the expression of proinflammatory cytokines in mouse and human T cells. (A) Illustration of
the sphingolipid biosynthesis pathway. The enzyme ceramide synthase is inhibited by the fungal toxin fumonisin B1. (B) Naïve mouse CD4+ T cells were cultured under
conditions promoting Treg cell differentiation in the presence of sphinganine (SA) or its upstream or downstream sphingolipid metabolites. The expression of FoxP3 was
assessed with flow cytometry. Bar graph showing the percentages of FoxP3+CD4+ T cells. The results are cumulative from three independent experiments with one mouse
used in each experiment. (C) Dot plots and a bar graph showing expression of FoxP3 in mouse CD4+ T cells cultured for 4 days under conditions promoting Treg cell polar-
ization in the presence of SA (5 μM) and/or fumonisin B1 (5 μM) as indicated. Data are pooled from two independent experiments with four mice in total. (D and E) Expres-
sion of FoxP3 (D and E) and IL-­17A (E) in mouse CD4+ T cells cultured for 4 days with SA under conditions promoting Treg cell (D) or TH17 cell (E) differentiation. (F) Contour
plots and line graphs showing expression of FoxP3, IL-­17A, and IFN-­γ in human CD4+ T cells cultured for 5 days with SA under conditions promoting Treg cell (top) or TH17
cell (bottom) differentiation. Data are presented as means ± SEM. (B and E) or means ± SD (C, D, and F) and are pooled from three [(D), n = 3 mice; (E), n = 4 mice; (F), n = 3
samples] independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Statistical analysis was performed with one-­way ANOVA.

or dimethyl sulfoxide (DMSO) vehicle control. Sphinganine treat- mRNA after sphinganine treatment than after DMSO treatment
ment increased the percentages of FoxP3 among total CD4+ T cells (Fig. 6B). The expression of Pdcd1 (encoding PD-­1) was up-­regulated
compared with DMSO (Fig. 6A). Because the expression of PD-­1 in by sphinganine (Fig. 6B), in agreement with the flow cytometry anal-
Treg cells was decreased by Spltc2 deficiency (Fig. 3E), we also mea- ysis results (Fig. 6A). Expression of Ccr6, a gene expressed at high
sured PD-­1 expression. Sphinganine potently increased the expres- levels in Treg cells (39), was increased by sphinganine. Sphinganine
sion of PD-­1 in both FoxP3+ CD4+ T cells and FoxP3− CD4+ T cells, also promoted the expression of Nr4a3, which is required for the dif-
suggesting that the effect of sphinganine on PD-­1 expression was not ferentiation of precursor Treg cells to mature Treg cells (40). Sphinga-
limited to Treg cells (Fig. 6A). In line with the flow cytometry analysis nine inhibited the expression of genes encoding cytokines of helper
results, scRNA-­seq revealed more CD4+ T cells expressing Foxp3 T cell lineages, such as IL-­4 and IL-­21, suggesting that sphinganine

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 8 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024

Fig. 6. Sphinganine promotes the expression of FoxP3 protein via the c-­Fos–PD-­1 signaling axis. (A) Flow cytometry plots showing the expression of FoxP3 and
PD-­1 in CD4+ T cells. Naïve FoxP3YFP × OT-­II splenocytes from three mice were cultured under Treg cell polarization conditions for 3 days in the presence of SA (5 μM) or
DMSO. CD4+ T cells were FACS-­purified for scRNA-­seq analysis. (B) tSNE plots showing the expression of the indicated genes in DMSO-­treated cells and SA-­treated cells,
as determined with scRNA-­seq analysis. (C) Dot plots and a bar graph showing expression of PD-­1 in CD4+ T cells cultured for 4 days with or without SA under conditions
promoting Treg cell polarization. (D) Dot plots and bar graph showing the expression of FoxP3 in CD4+ T cells cultured for 4 days with or without SA or PD-­1 blocking an-
tibody under conditions promoting Treg cell polarization. (E and F) Dot plots and bar graphs showing expression of PD-­1 (E) and FoxP3 (F) in CD4+ T cells cultured for 4
days in the presence or absence of SA or the c-­Fos inhibitor T-­5224 under conditions promoting differentiation of Treg cells. Data are representative of three independent
experiments, each with three technical replicates [(C to F), n = 3 mice total]. Data are presented as means ± SD. **P < 0.01; ***P < 0.001; ****P < 0.0001. Statistical
analysis was performed with a two-­way ANOVA.

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 9 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

promoted Treg cell differentiation at the expense of helper T cell lin- neutral amino acid. Basic amino acids arginine (R) and lysine (K)
eages. Expression of Il7r, which is expressed in Treg cells at low levels were found at corresponding positions in FosB and FosL2, respec-
(41), was decreased by sphinganine. Expression of Batf3, a gene that tively. Similarly, amino acid 185 in c-­Fos is an asparagine (N) residue.
promotes the generation of Teff cells, was decreased by sphinganine, This is also a neutral amino acid and therefore has properties distinct
suggesting that sphinganine shifted the balance of Teff cell versus Treg from those of the acidic glutamate (E) found at this position in the
cell differentiation toward Treg cell lineage differentiation. corresponding sequences of FosB and FosL2. Amino acid 187 of c-­
Previous studies have documented the role of the PD-­1 pathway Fos is leucine (L), which has a nonpolar side chain and consequently
in promoting Treg cell differentiation (42, 43). Sphinganine in- is distinct from glutamine (Q), an amino acid with a polar side chain,
creased the protein levels of PD-­1 in Treg cells in a dose-­dependent found at this position in FosB and FosL2. The other three mis-
manner (Fig. 6C). In line with this observation, Sptlc2 deficiency matched amino acids at positions 174, 178, and 181 were expected to
reduced the expression levels of FoxP3 and PD-­1 (fig. S13, A and B) have a less complex or minimal impact on biochemical properties.
and increased the expression levels of IL-­17A (fig. S13C). To test To test whether amino acids T164, N185, and L187 were required for
the hypothesis that sphinganine promotes Treg cell differentiation the binding between sphinganine and c-­Fos protein, we performed
by increasing PD-­1 expression, we differentiated Treg cells in the MST assays using mutant c-­Fos protein bearing three mutations
presence of sphinganine and anti–PD-­1. PD-­1 blockade abrogated (T164R, N185E, and L187Q). These three mutations abrogated the

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
the sphinganine-­mediated increase in FoxP3 expression (Fig. 6D), binding between sphinganine and c-­Fos protein (Fig. 7A), suggesting
suggesting that PD-­1 was required for the sphinganine-­mediated that these three amino acids were indispensable for the physical in-
increase in FoxP3 expression. teraction between sphinganine and c-­Fos protein.
Pdcd1 is a target gene of the transcription factor c-­Fos, which To study whether sphinganine required binding with c-­Fos pro-
binds DNA at 36 to 44 base pairs upstream of the Pdcd1 transcription tein to enhance Treg cell differentiation, we overexpressed mutant c-­
start site (TSS), thus promoting Pdcd1 transcription (44). Therefore, Fos protein or wild-­type c-­Fos protein in CD4+ T cells and cultured
we hypothesized that sphinganine increased the expression of PD-­1 these T cells under Treg cell polarization conditions in the presence of
in a manner dependent on c-­Fos. To test this hypothesis, we induced sphinganine or DMSO vehicle control (Fig. 7C). The expression lev-
Treg cell differentiation in the presence or absence of T-­5224, a selec- els of PD-­1 and FoxP3 in CD4+ T cells were comparable between
tive inhibitor of c-­Fos DNA binding activity (45). T-­5224 significant- DMSO-­treated CD4+ T cells overexpressing mutant c-­Fos protein
ly inhibited sphinganine-­induced expression of both PD-­1 (Fig. 6E) and those overexpressing wild-­type c-­Fos protein, suggesting that
and FoxP3 (Fig. 6F). These results suggested that sphinganine in- the three amino acid mutations did not affect Treg cell differentiation
duced the expression of FoxP3 in a manner dependent on the under steady-­state conditions (Fig. 7, D and E). In contrast to DMSO-­
c-­Fos–PD-­1 axis. treated CD4+ T cells, sphinganine-­treated CD4+ T cells overexpress-
ing mutant c-­Fos protein expressed lower levels of PD-­1 and FoxP3
Sphinganine enhances the expression of FoxP3 by directly than those overexpressing wild-­type c-­Fos protein, suggesting that
binding c-­Fos sphinganine required binding to c-­Fos protein to increase the expres-
Because small-­molecule compounds can directly bind proteins (46, sion levels of PD-­1 and FoxP3.
47), we hypothesized that sphinganine might physically interact with We next examined the structural mechanisms through which
c-­Fos and thereby influence c-­Fos binding to its target genes, thus sphinganine bound c-­Fos protein and increased PD-­1 expression. Be-
regulating target gene transcription. To test this hypothesis, we per- cause c-­Fos is known to bind its partner protein c-­Jun, NFAT1 (also
formed microscale thermophoresis (MST) binding assays, a well-­ known as NFATc2), and DNA as a quaternary complex (51, 52), we
established method to evaluate interactions between small-­molecule sought to explore how sphinganine might influence the formation of
compounds and proteins (47–49). Sphinganine bound recombinant the c-­Fos-­c-­Jun-­NFAT1-­DNA quaternary complex. We studied a pre-
full-­length c-­Fos protein with a Kd [dissociation constant (binding af- viously reported crystal structure of the quaternary complex com-
finity)] value similar to that of T-­5524 (Fig. 7A). T-­5524 is known to posed of c-­Fos, c-­Jun, NFAT1, and a target DNA fragment (51) and
bind c-­Fos protein (45) and, therefore, was used as a positive control. performed molecular docking to generate a model of the sphinganine-­
Sphinganine did not bind the 6× His peptide tag, a negative control in bound complex (fig. S14A). In this docking model, sphinganine
this assay. Furthermore, sphinganine did not bind to FosB protein, formed interactions with the three critical amino acids of c-­Fos (T164,
which is also a member of the Fos family that shares amino acid se- N185, and L187) and several amino acids of NFAT1, thereby suggest-
quence similarity with c-­Fos. Another member of the Fos family, Fos-­ ing a mechanism in which sphinganine might function as a “molecu-
like 2 (FosL2), interacted weakly with sphinganine but only at very lar glue” stabilizing the interactions between c-­Fos and its partner
high concentrations of sphinganine (Fig. 7A). These findings sug- proteins. Bridging of c-­Fos to its partners by sphinganine presumably
gested that the interaction between sphinganine and c-­Fos protein favored the recruitment of these three transcription factors to the
was relatively specific. Pdcd1 promoter and subsequent activation of Pdcd1 transcription. In
To identify the amino acid residues required for the binding be- a second docking model of the c-­Fos-­DNA complex (fig. S14B),
tween sphinganine and c-­Fos protein, we aligned the amino acid se- sphinganine formed contacts with multiple amino acids from both
quences of the coiled-­coil domains of c-­Fos, FosB, and FosL2. We c-­Fos molecules, including T164, N185, and L187, thus potentially
chose to analyze this domain because it is critical for DNA binding facilitating the formation of the c-­Fos homodimer.
and protein-­protein dimerization interactions (50). An alignment of To investigate the potential influence of the binding between
amino acid sequences revealed six amino acid mismatches. Three sphinganine and c-­Fos protein on Pdcd1 transcription, we performed
mismatched amino acids, which are highlighted within red rectan- chromatin immunoprecipitation (ChIP) analysis. Supplementation
gles, introduced markedly distinct biochemical properties (Fig. 7B). with sphinganine increased the recruitment of c-­Fos to a previously
For example, the threonine (T) residue at position 164 in c-­Fos is a defined binding site within the promoter of Pdcd1 (fig. S15, A and B)

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 10 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
Fig. 7. Sphinganine induces FoxP3 expression in a manner dependent on the binding between sphinganine and c-­Fos protein. (A) Line graphs showing MST analysis of
binding between SA and wild-­type c-­Fos protein, mutant c-­Fos protein (T164R, N185E, and L187Q), and two other members of the Fos family, FosB and FosL2. All proteins were
fused with a 6× histidine (His) tag, which was used for protein purification and fluorescence labeling. The c-­Fos binding inhibitor T-­5224 was a positive control, and 6× His pep-
tide was a negative control. The normalized fluorescence values (Fnorm) determined in the presence of increasing concentrations of SA or T-­5224 are shown. (B) Alignment of the
coiled-­coil domains of c-­Fos, FosB, and FosL2 proteins. The amino acid sequences of these domains are the same between Homo sapiens and Mus musculus. The mismatched
amino acids that introduce markedly distinct biochemical properties are highlighted within red rectangles. The numbers above refer to the position of each amino acid within
the sequence of c-­Fos protein. (C) Illustration of experimental design of (D) and (E). (D and E) Flow cytometry plots and bar graphs showing the percentages of PD-­1 (D) and
FoxP3 (E) in CD4+ T cells retrovirally transduced to express wild-­type or mutant c-­Fos protein. Data are representative of three independent experiments (A) or two independent
experiments with six mice total (D and E). ****P < 0.0001. Data are presented as means ± SD. Statistical analysis was performed with one-­way ANOVA.

(44). Furthermore, sphinganine treatment promoted T-­5524–sensitivebeen reported to partner with c-­Jun and NFAT1 in binding DNA
transcriptional activation of the Pdcd1 promoter (fig. S15C). Collec-
(51, 52). NFAT1 protein does not bind Pdcd1 promoter. NFAT2
tively, these results suggested a model in which sphinganine bound (also known as NFATc1), a protein that shares amino acid sequence
c-­Fos and enhanced its DNA-­binding and transcriptional activity. similarity with NFAT1, has been shown to directly bind the Pdcd1
promoter (53). Therefore, we assessed the global binding patterns
Sphinganine enhances the recruitment of c-­Fos protein to its of NFAT2, together with c-­Fos and c-­Jun on their target genes. We
target genes found an enrichment of the binding events for c-­Fos, c-­Jun, and
To study the influence of sphinganine on the binding between c-­ NFAT2 occurring near TSSs of genes (Fig. 8A). Sphinganine treat-
Fos and its target genes, we performed CUT & RUN sequencing ment significantly promoted c-­Fos binding to 7736 sites and de-
analysis of the genomic DNA binding patterns of c-­Fos. c-­Fos has creased c-­Fos binding to 4911 sites in the genome (Fig. 8B). In

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 11 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

Fig. 8. Sphinganine promotes the recruitment of c-­Fos to

chromatin accessible regions of target genes including
both c-­Fos solo genes and symphony genes. (A) OT-­II cells
were cultured with SA (5 μM) or DMSO under Treg cell differen-
tiation conditions for 60 hours before CUT & RUN sequencing
analysis of the global binding sites for c-­Fos, c-­Jun, and NFAT2.
Averaged density plots (top) and heatmaps (bottom) show the
signal enrichment in the three transcription factors at TSS adja-
cent regions. Each row represents one gene in the heatmaps
for the indicated transcription factors. (B) Volcano plots show-
ing the differential binding of the indicated transcription fac-
tors in SA-­or DMSO-­treated cells. The yellow dots show the
binding sites significantly enriched or under-­represented in
SA-­treated cells. (C) Stacked bar graphs showing the profiles of
genomic regions bound by c-­Fos, c-­Jun, and NFAT2. (D) Heat-
maps showing genes bound by individual, or combinations of,
transcription factors in SA-­or DMSO-­treated cells. (E) Heatmap

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
showing genes differentially bound by the indicated transcrip-
tion factors between SA-­treated cells and DMSO-­treated cells.
Dark red: genes bound by the indicated transcription factors in
only SA-­treated cells; dark blue: genes bound by the indicated
transcription factors in only DMSO-­treated cells; gray: genes
bound by, or not bound by, the indicated transcription factors
in both SA-­treated cells and DMSO-­treated cells. (F) CUT & RUN
sequencing and ATAC sequencing signal profiles across Pdcd1
and Ptpn11 loci in OT-­II cells cultured with SA or DMSO under
Treg cell differentiation conditions. Promoter regions with CUT
& RUN sequencing signal peaks and ATAC sequencing signal
peaks are highlighted in red. (G) The heatmap on the left
shows the ATAC sequencing analysis of chromatin openness
(dark purple and light blue indicate accessible and inaccessible
chromatin regions, respectively) of OT-­II cells cultured with SA
or DMSO under Treg cell differentiation conditions. Each line
indicates one gene in the heatmap. The bar graphs on the
right show the correlation between chromatin accessibility
and binding patterns of c-­Fos. Data are pooled from three mice
each for CUT & RUN sequencing and ATAC sequencing.

contrast, sphinganine had much less of an influence on the num- types of DNA elements bound by c-­Jun or NFAT2 were also not
bers of binding sites for c-­Jun and NFAT2 (Fig. 8B). Although influenced by sphinganine.
sphinganine increased the total amounts of binding sites for c-­Fos, To study the genome-­wide influences of sphinganine on the glob-
sphinganine treatment did not markedly change the types of bind- al binding patterns of c-­Fos, c-­Jun, and NFAT2, we systematically
ing sites for c-­Fos, as demonstrated by comparable percentages of mapped the target genes bound by these three transcription factors
promoters, exons, introns, and intergenic regions between the in DMSO-­treated or sphinganine-­treated cells (Fig. 8D) and further
DMSO-­treated cells and sphinganine-­treated cells (Fig. 8C). The calculated the difference in the pattern of target genes between

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 12 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

DMSO-­treated or sphinganine-­treated cells by subtracting the bind- produce CCL5 and recruit Treg cells (55). Treg cells require the Fas/
ing events in DMSO-­treated cells from those in sphinganine-­treated FasL pathway to inhibit Teff cell proliferation (56). Sphinganine in-
cells (Fig. 8E). Compared with DMSO-­treated cells, sphinganine-­ creased the chromatin accessibility of these three genes, indicating
treated cells gained 4736 and lost 658 target genes bound by c-­Fos. that sphinganine confers these genes a “ready-­to-­be induced” status,
Sphinganine treatment influenced fewer target genes bound by c-­Jun potentially promoting the transcription of Ccl3, Ccl5, and Fasl in Treg
or NFAT2 than c-­Fos, as demonstrated by smaller blocks of genes in cells to regulate Treg cell migration and suppressive function. Genes in
dark red or in dark blue in the groups of c-­Jun or NFAT2 than c-­Fos group B include Tbx21, Il4, and Il21 (fig. S16). Tbx21 encodes protein
(Fig. 8E). T-­bet. Il4 and Il21 encode cytokines IL-­4 and IL-­21. Sphinganine de-
Sphinganine markedly rewired the genome-­wide patterns of target creased the chromatin accessibility of these marker genes of helper T
genes of c-­Fos. We categorized genes into several clusters on the basis cells, indicating that sphinganine may reduce the transcription of
of the individual or combinational binding by the three transcription Tbx21, Il4, and Il21 in Treg cells to maintain the identity of Treg cells at
factors (Fig. 8D). For example, c-­Fos “solo genes” were those bound the expense of helper T cell lineages. These genes are not direct target
only by c-­Fos and not by c-­Jun or NFAT2. We observed more c-­Fos genes of c-­Fos protein, suggesting that sphinganine likely regulates
solo genes than c-­Jun solo genes or NFAT2 solo genes. One example their transcription through c-­Fos–independent mechanisms. We did
of c-­Fos solo genes was Pdcd1. c-­Fos bound the Pdcd1 locus near its not observe a clear pattern of correlation between c-­Fos recruitment

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
TSS, an accessible chromatin region, as documented by a peak in and chromatin accessibility of cluster 2 or 3 genes. Collectively, these
an assay for transposase-­accessible chromatin (ATAC) sequencing results suggest that sphinganine only modestly changed the epigenetic
(Fig. 8F). “Symphony genes” were those bound by all three transcrip- landscape of T cells. The recruitment of c-­Fos to its target genes, as
tion factors in SA-­treated cells. Most symphony genes were concomi- triggered by sphinganine treatment, was uncoupled from its modest
tantly bound by all three transcription factors in both DMSO-­treated effects on the epigenetic regulation of T cells.
cells and SA-­treated cells, and sphinganine further enhanced the
binding of c-­Fos to target genes such as Ptpn11 (encoding SHP-­2 pro-
tein, a component of the PD-­1 signaling pathway) (Fig. 8D). The DISCUSSION
shared binding site for c-­Fos, c-­Jun, and NFAT2 on Ptpn11 was also in In this study, we identified that serine enrichment in the TME pro-
an accessible chromatin region (Fig. 8F). Our computer docking motes Treg cell accumulation in an Sptlc2-­dependent manner (fig. S17).
models were useful for understanding both modes of sphinganine-­ The TME harbors characteristics of altered nutrient availability and
mediated regulation of c-­Fos binding to the c-­Fos solo genes and to Treg cell–mediated immunosuppression (4–6), and our findings pro-
the symphony genes. In the case of c-­Fos solo genes, sphinganine was vide an example of how extracellular nutrient availability shapes intra-
estimated to stabilize the c-­Fos homodimer. In the case of symphony cellular signal transduction and Treg cell differentiation. In contrast to
genes, sphinganine was predicted to function as a molecular glue other studies showing that metabolites in the TME, such as lactic acid,
strengthening the physical interactions between c-­Fos and partner regulate cellular energy production (57, 58), this study highlights that
proteins (fig. S14). Collectively, sphinganine promoted c-­Fos recruit- sphinganine functions as a messenger molecule that translates external
ment to accessible chromatin regions of c-­Fos solo genes and sym- metabolic cues into internal transcriptional information in Treg cells by
phony genes, and sphinganine had minimal effects on recruiting physically interacting with c-­ Fos and strengthening genome-­ wide
c-­Jun and NFAT2 to target genes (Fig. 8B). c-­Fos–driven transcription of target genes.
To understand the effects of sphinganine on the chromatin acces- Serine enrichment has been observed in the cerebrospinal fluid of
sibility of target genes bound by c-­Fos, c-­Jun, and NFAT2, we ana- human patients with brain tumors (59), and fatty acids are enriched
lyzed the CUT & RUN sequencing results and ATAC sequencing in human breast TIF (60). In combination with these studies, our
results side by side and correlated genes in clusters 1, 2, and 3 in the findings suggest that enrichment of serine and fatty acids occurs in
CUT & RUN sequencing with genes in groups A, B, C, and D in both human and mouse tumors and may represent a conserved pat-
ATAC sequencing (Fig. 8G). Sphinganine treatment significantly in- tern of nutrient availability. A limitation of this current study is that
creased or decreased the chromatin accessibility of 3526 or 3306 we used mouse TIF to perform metabolomics analysis, and further
genes, respectively. We identified 3114 or 23652 genes that were con- research using TIF from human patients with cancer is required. An
tinually inaccessible or accessible regardless of sphinganine treatment. unanswered question pertains to the mechanism through which ser-
The chromatin accessibility of most genes was not influenced by ine and fatty acids are enriched in tumors. One possible explanation
sphinganine, suggesting that sphinganine did not potently change the is that tumor cells are metabolically hyperactive and consume elevat-
epigenetic landscape of T cells. Sphinganine-­induced recruitment of ed amounts of serine and fatty acids. Therefore, tumor cells may
c-­Fos to its solo genes occurred mainly in the continually accessible “suck” these nutrients from the circulation, enriching serine and fatty
chromatin regions, as demonstrated by the significant enrichment of acids in the TME. A second explanation is that the unique properties
cluster 1 genes in group D (Fig. 8G). In contrast, the target genes of of the TME enhance the enrichment of these nutrients, as demon-
c-­Fos in sphinganine-­treated cells were not significantly enriched in strated by hypoxia-­induced serine accumulation (61) and tumor
chromatin regions that were opened or closed by sphinganine treat- cell–released esterified fatty acids (62, 63).
ment, as documented by the under-­representation of cluster 1 genes Serine metabolism has been shown to regulate T cell differentia-
in groups A and B. These results indicate that sphinganine-­caused tion and proliferation. For example, dietary restriction of serine
changes in the epigenetic landscape and sphinganine-­induced re- inhibits antibacterial CD8+ T cell responses (35). Furthermore, ser-
cruitment of c-­Fos to target genes were not closely coupled with each ine is a substrate for glutathione synthesis, and genetic deficiency in
other. Genes in group A, such as Ccl3, Ccl5, and Fasl (fig. S16), play the glutathione synthetic enzyme glutamate-­cysteine ligase causes
crucial roles in Treg cell migration and function. Treg cell–derived excessive accumulation of serine in Treg cells, thereby impairing
CCL3 is required to suppress autoinflammation (54). Tumor cells Treg cell suppressive function (36). Our study reveals that the

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 13 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

serine-­sphingolipid synthetic pathway is required for Treg cell ac- sphingolipid biosynthesis in Teff cells and Treg cells serves as a “notifi-
cumulation in the TME. These results do not contradict a previous cation” of the metabolic activation of Teff cells to Treg cells, and this
study which found that excessive accumulation of serine inside Treg metabolic synchronization ensures timely Treg cell–mediated im-
cells, as mediated by glutamate-­cysteine ligase deficiency, inhibits mune tolerance.
Treg cell–mediated immunosuppression (36). Our study showed The requirements of Sptlc2-­mediated sphingolipid biosynthesis
that extracellular serine enrichment in tumors and Sptlc2-­mediated for both Teff cell–mediated effector functions and Treg cell–mediated
sphingolipid synthesis in Treg cells promoted Treg cell differentia- immunosuppressive functions create challenges in attempting to
tion, whereas the previous study had indicated that Treg cells re- use a pharmacological approach to inhibiting Sptlc2 in Treg cells to
quire a glutathione synthetic enzyme to consume serine in the unleash antitumor Teff cell function because the pan inhibition of
glutathione synthetic pathway and to maintain Treg cell suppressive Sptlc2 also suppresses Teff cell metabolic fitness and function. Our
function (36). The distinct roles of serine in Treg cell differentiation results suggest that one suitable strategy to selectively edit this
correlate with the different genotypes of Treg cells and the different sphingolipid biosynthetic pathway in Treg cells is to enhance rather
tissues investigated, i.e., wild-­type Treg cells in tumors in our than inhibit sphingolipid biosynthesis. For example, sphinganine
study versus glutamate-­cysteine ligase–deficient Treg cells in the supplementation selectively promoted Treg cell formation and func-
spleen in the previous study. Together, our results and previously tion and decreased inflammatory cytokine production by Teff cells.

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
reported findings suggest that serine metabolism influences Treg Therefore, whereas both Treg cells and Teff cells required endoge-
cell differentiation and immunosuppressive function in a context-­ nous sphingolipid biosynthesis, Treg cells were more sensitive than
dependent manner. Teff cells to exogenously supplied sphinganine. These observations
Our findings in combination with previous studies suggest that suggested that the ways in which immune cells acquire metabolites
Sptlc2-­mediated de novo sphingolipid synthesis may either support (i.e., endogenous synthesis versus exogenous acquisition) might di-
or suppress immune responses in Teff cells or Treg cells. For example, a rectly influence cell differentiation. This observation echoes results
previous study from our laboratory reported that TCR stimulation– from previous studies documenting the different effects of lipid
induced sphingolipid biosynthesis is required for the metabolic fitness synthesis versus lipid uptake on CD4+ T cell lineage differentiation
of CD8+ Teff cells as well the development of their antiviral function (72) and macrophage-­mediated antiviral responses (73). Briefly,
(20). The results of the current study revealed that Sptlc2 and sphinga- TH17 cells depend on de novo fatty acid synthesis to produce phos-
nine are critical for the development of Treg cell–mediated immuno- pholipids, whereas Treg cells acquire exogenous fatty acids to sus-
suppressive functions. Together, the findings from these two studies tain phospholipid production (72). Type I interferon inhibits de
suggest that Sptlc2-­mediated sphingolipid biosynthesis promotes the novo cholesterol synthesis and increases cholesterol import in mac-
differentiation and function of both Teff cells and Treg cells, supporting rophages, and inhibiting de novo cholesterol synthesis up-­regulates
necessary Teff cell responses and preventing uncontrolled immune re- production of type I interferon (73). Together, previous studies and
actions. These seemingly contradictory observations are reminiscent our results suggest that shifting between endogenous synthesis of
of the distinct roles played by the lineage commitment factors T-­bet, lipids versus exogenous acquisition of lipids is a promising strategy
GATA3, IRF4, STAT3, and JunB in Teff cells and Treg cells. These tran- to modulate immune responses in favor of disease treatment.
scription factors promote Teff cell lineage commitment and activation We showed that sphinganine directly interacted with the coiled-­
and are also expressed in Treg cells. Shared expression of these tran- coil domain of the c-­Fos protein and increased genome-­wide recruit-
scription factors in both Teff cells and Treg cells presumably “notifies” ment of c-­Fos to its cognate binding site within the promoter regions
Treg cells of Teff cell lineage differentiation and activation and is es- of target genes, including Pdcd1. These results suggested that sphinga-
sential for the development of Treg cell–mediated immune tolerance nine is not only an important intermediate that fuels the biosynthesis
(64–71). The results of 13C215N1 tracer–labeling experiments revealed of more complex sphingolipids but also directly engages a critical in-
that TCR stimulation induced sphingolipid biosynthesis concurrently tracellular signal transduction pathway and influences gene transcrip-
in both Teff cells and Treg cells. Sptlc2-­mediated metabolic synchroni- tion and T cell differentiation. These results echo findings from
zation between Treg cells and Teff cells might function as a mode of previous studies showing that sphingosine-­1-­phosphate physically
intercellular communication, in which the real-­time information as- interacts with histone deacetylases and inhibits histone deacetylation
sociated with Teff cell activation status is transmitted to Treg cells and (46). A limitation of this current study is that we were not able to ob-
provides Treg cells with the means to sense metabolic activation of tain the experimental structure of the c-­Fos–c-­Jun–NFAT2–DNA
neighboring Teff cells. Several hours or more may be required for na- complex bound to sphinganine. Because NFAT2 shares amino acid
ïve T cells to respond to environmental signals (e.g., cytokine stimula- sequence similarity with NFAT1, we used a previously reported crys-
tion) and to express Teff cell lineage–determining transcription tal structure of the c-­Fos–c-­Jun–NFAT1–DNA complex (51). A re-
factors. In contrast, the tracer-­labeled lipid biosynthetic flux in T cells maining task is to identify the experimental structure, which would
is detectable within 30 min of the initiation of tracer labeling (20). provide more accurate experimental evidence than the docking mod-
Hence, this activation-­sensing mechanism can ensure appropriate els used in the current study. These future structural studies will be
immune responses even before Treg cells can identify the specific lin- essential for understanding the two modes of c-­Fos binding its target
eage commitment pathways of Teff cells. These findings suggest that genes, i.e., forming of a c-­Fos/c-­Fos homodimer versus forming a c-­
synchronization of sphingolipid biosynthesis may be among the earli- Fos/partner protein heterocomplex in sphinganine-­treated T cells.
est alarm signals enabling Treg cells to sense emerging Teff cell re- Previous studies have shown that c-­Fos plays multifaceted roles in
sponses. This activation-­sensing mechanism complements the Treg cell differentiation and function. For example, adenosine pro-
well-­documented transcription factor–dependent immunoregulatory motes the expression of FoxP3 via mechanisms that are dependent
mechanisms associated with Teff cell lineage commitment. Our results on c-­Fos and on other members of the activator protein 1 family of
suggest a working model in which Sptlc2-­mediated synchronized transcription factors (74). In contrast, the c-­Fos–c-­Jun heterodimer

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 14 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

complex competes with FoxP3 for binding NFAT1 consensus bind- Human samples
ing sites. The outcome of this competition determines the transcrip- Buffy coats from healthy donor blood samples were provided by the
tion of genes involved in acute T cell activation versus those that Blutspendezentrale Karlsruhe (Karlsruhe Blood Bank) after approval
promote Treg cell suppressive functions (75). The current study re- by the ethics committee of DKFZ and Heidelberg University. In-
vealed that sphinganine treatment enhanced the recruitment of c-­ formed consent was obtained before the analysis. CD4+ T cells were
Fos to the Pdcd1 promoter, increased PD-­1 expression, and promoted purified from isolated peripheral blood mononuclear cells for use in
FoxP3 expression in a c-­Fos– and PD-­1–dependent manner. Collec- in vitro differentiation experiments.
tively, those studies together with our results suggest that the role of
c-­Fos in Treg cell differentiation is context-­dependent. Tumor cell implantation
In summary, we found that serine enrichment in the TME pro- B16-­F10 cells, YUMMER cells, and MC-­38 cells were cultured in
motes Treg cell accumulation through the Sptlc2-­mediated sphingo- Dulbecco’s modified Eagle’s medium supplemented with 10% fetal
lipid synthetic pathway. Furthermore, we identified that sphinganine bovine serum, glutamine, penicillin, and streptomycin. Mice were
is an immunosuppressive messenger molecule that promotes the dif- shaved, and tumor cells were injected subcutaneously (2 × 105 cells
ferentiation and function of mouse and human Treg cells and might per mouse) into the flanks. Tumors were measured every 2 to 3 days
have clinical relevance for treating autoimmune diseases. This study with calipers. In some experiments, mice were fed a serine/glycine-­

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
focuses on the serine-­sphingolipid metabolic pathway, which is one free diet (catalog no. 827030, Special Diets Services) for 1 month be-
of several pathways of serine metabolism. The current study does not fore implantation of B16-­ F10 melanoma cells in the mice. The
clarify whether other serine metabolic pathways regulate Treg cell dif- tumor-­bearing mice continued to be kept on the same diets as those
ferentiation in the TME. Further work is warranted to investigate the before tumor implantation.
allocation of serine in each metabolic pathway and to evaluate the
individual contribution of each pathway in regulating Treg cell dif- Antibodies
ferentiation in the TME. Human antibodies were obtained from BioLegend for detecting the
following antigens: CD4 (catalog no. 357416), FoxP3 (catalog no.
320108), IL-­17A (catalog no. 512314), and IFN-­γ (catalog no. 502528).
MATERIALS AND METHODS Mouse antibodies were for detecting the following antigens: FoxP3
Study design (catalog no. 17-­5773-­82, eBioscience), CD4 (catalog no. 552051, BD
The objective of this study was to investigate whether and how me- Bioscience), CD8a (catalog no. 100734, BioLegend), CD44 (catalog
tabolites in the TME influence Treg cell differentiation. We performed nos. 103028 and 103040, BioLegend), CD62L (catalog no. 104418,
metabolomics analysis of the TME and found that serine and palmitic BioLegend), PD-­1 (catalog no. 135216, BioLegend), TIGIT (catalog
acid, two substrates for sphingolipid synthesis, were enriched in the no. 142104, BioLegend), IL-­17A (catalog no. 506904, BioLegend),
TME. The enzyme Sptlc2 catalyzes the rate-­limiting step of sphingo- TNF-­α (catalog no. 506306, BioLegend), and IFN-­γ (catalog no.
lipid synthesis. We bred a mouse model with Treg cell–specific defi- 505826, BioLegend). Anti-­ CD3 (catalog no. 100223, BioLegend),
ciency in the gene Sptlc2. We found that Sptlc2 deficiency suppressed anti-­CD28 (catalog no. 102116, BioLegend), anti-­PD-­1 (catalog no.
Treg cell accumulation in the TME and inhibited tumor growth. The 135247, BioLegend), and anti–IFN-­γ (catalog no. 505834, BioLegend)
pro-­Treg cell role depends on serine accumulation in the TME, as antibodies were used for murine cell culture. Anti-­CD3 (catalog no.
demonstrated by the observation that mice with Treg cells deficient or 317315, BioLegend), anti-­CD28 (catalog no. 302923, BioLegend), and
sufficient in Sptlc2 grew tumors of similar sizes when fed a serine-­free anti–IFN-­γ (catalog no. 506532, BioLegend) antibodies were used for
diet. By testing a panel of sphingolipids, we found that sphinganine, human cell culture.
an intermediate metabolite in sphingolipid synthesis, promoted in-
ducible Treg cell differentiation in vitro in a c-­Fos–PD-­1–dependent
manner. We indicated the numbers of mice in each experimental Supplementary Materials
The PDF file includes:
group in figure legends. We determined sample sizes based on our Supplementary Materials and Methods
previous studies in the laboratory. We did not exclude data from Figs. S1 to S17
the analysis. References (77–84)

Other Supplementary Material for this manuscript includes the following:

Mice Data files S1 and S2
Sptlc2Fl/Fl and Foxp3YFP-­Cre mice on the C57BL/6 background were MDAR Reproducibility Checklist
provided by X. Jiang (SUNY Downstate Medical Center, New York)
and A. Rudensky (Memorial Sloan Kettering Cancer Center, New REFERENCES AND NOTES
York), respectively (28, 29). Cd4Cre and OT-­II mice on the C57BL/6 1. S. Hori, T. Nomura, S. Sakaguchi, Control of regulatory T cell development by the
background were purchased from the Jackson Laboratory (31, 32, transcription factor Foxp3. Science 299, 1057–1061 (2003).
2. J. D. Fontenot, M. A. Gavin, A. Y. Rudensky, Foxp3 programs the development and
76). Mice were maintained in the specific pathogen–free facility at function of CD4+CD25+ regulatory T cells. Nat. Immunol. 4, 330–336 (2003).
Deutsches Krebsforschungszentrum (DKFZ) and the Institute of 3. R. Khattri, T. Cox, S. A. Yasayko, F. Ramsdell, An essential role for Scurfin in CD4+CD25+ T
Health and Medicine (IHM). Age-­and sex-­matched littermates (5 to regulatory cells. Nat. Immunol. 4, 337–342 (2003).
10 weeks old) were used as controls in all experiments. In rare cases, 4. K. Klages, C. T. Mayer, K. Lahl, C. Loddenkemper, M. W. L. Teng, S. F. Ngiow, M. J. Smyth,
A. Hamann, J. Huehn, T. Sparwasser, Selective depletion of Foxp3+ regulatory T cells
mice with fighting wounds were excluded from the sampling. All
improves effective therapeutic vaccination against established melanoma. Cancer Res.
studies were performed following DKFZ and IHM regulations after 70, 7788–7799 (2010).
approval by the German regional council at the Regierungspräsidi- 5. M. R. Sullivan, L. V. Danai, C. A. Lewis, S. H. Chan, D. Y. Gui, T. Kunchok, E. A. Dennstedt,
um Karlsruhe and the IHM Animal Care and Use Committee. M. G. Vander Heiden, A. Muir, Quantification of microenvironmental metabolites in

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 15 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

murine cancers reveals determinants of tumor nutrient availability. eLife 8, e44235 M. Herlyn, G. J. Freeman, J. D. Rabinowitz, H. C. J. Ertl, Enhancing CD8+ T cell fatty acid
(2019). catabolism within a metabolically challenging tumor microenvironment increases the
6. E. A. Burgess, B. Sylven, Glucose, lactate, and lactic dehydrogenase activity in normal efficacy of melanoma immunotherapy. Cancer Cell 32, 377–391.e9 (2017).
interstitial fluid and that of solid mouse tumors. Cancer Res. 22, 581–588 (1962). 26. K. Glunde, Z. M. Bhujwalla, S. M. Ronen, Choline metabolism in malignant transformation.
7. S. Xu, O. Chaudhary, P. Rodriguez-­Morales, X. Sun, D. Chen, R. Zappasodi, Z. Xu, Nat. Rev. Cancer 11, 835–848 (2011).
A. F. M. Pinto, A. Williams, I. Schulze, Y. Farsakoglu, S. K. Varanasi, J. S. Low, W. Tang, 27. B. I. Reinfeld, M. Z. Madden, M. M. Wolf, A. Chytil, J. E. Bader, A. R. Patterson, A. Sugiura,
H. Wang, B. McDonald, V. Tripple, M. Downes, R. M. Evans, N. A. Abumrad, T. Merghoub, A. S. Cohen, A. Ali, B. T. Do, A. Muir, C. A. Lewis, R. A. Hongo, K. L. Young, R. E. Brown,
J. D. Wolchok, M. N. Shokhirev, P. C. Ho, J. L. Witztum, B. Emu, G. Cui, S. M. Kaech, Uptake of V. M. Todd, T. Huffstater, A. Abraham, R. T. O’Neil, M. H. Wilson, F. Xin, M. N. Tantawy,
oxidized lipids by the scavenger receptor CD36 promotes lipid peroxidation and W. D. Merryman, R. W. Johnson, C. S. Williams, E. F. Mason, F. M. Mason, K. E. Beckermann,
dysfunction in CD8+ T cells in tumors. Immunity 54, 1561–1577.e7 (2021). M. G. Vander Heiden, H. C. Manning, J. C. Rathmell, W. K. Rathmell, Cell-­programmed
8. L. Bunse, S. Pusch, T. Bunse, F. Sahm, K. Sanghvi, M. Friedrich, D. Alansary, J. K. Sonner, nutrient partitioning in the tumour microenvironment. Nature 593, 282–288 (2021).
E. Green, K. Deumelandt, M. Kilian, C. Neftel, S. Uhlig, T. Kessler, A. von Landenberg, 28. Z. Li, Y. Li, M. Chakraborty, Y. Fan, H. H. Bui, D. A. Peake, M. S. Kuo, X. Xiao, G. Cao,
A. S. Berghoff, K. Marsh, M. Steadman, D. Zhu, B. Nicolay, B. Wiestler, M. O. Breckwoldt, X. C. Jiang, Liver-­specific deficiency of serine palmitoyltransferase subunit 2 decreases
R. Al-­Ali, S. Karcher-­Bausch, M. Bozza, I. Oezen, M. Kramer, J. Meyer, A. Habel, J. Eisel, plasma sphingomyelin and increases apolipoprotein E levels. J. Biol. Chem. 284,
G. Poschet, M. Weller, M. Preusser, M. Nadji-­Ohl, N. Thon, M. C. Burger, P. N. Harter, 27010–27019 (2009).
M. Ratliff, R. Harbottle, A. Benner, D. Schrimpf, J. Okun, C. Herold-­Mende, S. Turcan, 29. Y. P. Rubtsov, J. P. Rasmussen, E. Y. Chi, J. Fontenot, L. Castelli, X. Ye, P. Treuting, L. Siewe,
S. Kaulfuss, H. Hess-­Stumpp, K. Bieback, D. P. Cahill, K. H. Plate, D. Hänggi, M. Dorsch, A. Roers, W. R. Henderson Jr., W. Muller, A. Y. Rudensky, Regulatory T cell-­derived
M. L. Suvà, B. A. Niemeyer, A. von Deimling, W. Wick, M. Platten, Suppression of antitumor T interleukin-­10 limits inflammation at environmental interfaces. Immunity 28, 546–558

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
cell immunity by the oncometabolite (R)-­2-­hydroxyglutarate. Nat. Med. 24, 1192–1203 (2018). (2008).
9. S. K. Vodnala, R. Eil, R. J. Kishton, M. Sukumar, T. N. Yamamoto, N. H. Ha, P. H. Lee, M. Shin, 30. K. S. Smigiel, E. Richards, S. Srivastava, K. R. Thomas, J. C. Dudda, K. D. Klonowski,
S. J. Patel, Z. Yu, D. C. Palmer, M. J. Kruhlak, X. Liu, J. W. Locasale, J. Huang, D. J. Campbell, CCR7 provides localized access to IL-­2 and defines homeostatically
R. Roychoudhuri, T. Finkel, C. A. Klebanoff, N. P. Restifo, T cell stemness and dysfunction in distinct regulatory T cell subsets. J. Exp. Med. 211, 121–136 (2014).
tumors are triggered by a common mechanism. Science 363, eaau0135 (2019). 31. S. Sawada, J. D. Scarborough, N. Killeen, D. R. Littman, A lineage-­specific transcriptional
10. P. C. Ho, J. D. Bihuniak, A. N. Macintyre, M. Staron, X. Liu, R. Amezquita, Y. C. Tsui, G. Cui, silencer regulates CD4 gene expression during T lymphocyte development. Cell 77,
G. Micevic, J. C. Perales, S. H. Kleinstein, E. D. Abel, K. L. Insogna, S. Feske, J. W. Locasale, 917–929 (1994).
M. W. Bosenberg, J. C. Rathmell, S. M. Kaech, Phosphoenolpyruvate is a metabolic 32. P. P. Lee, D. R. Fitzpatrick, C. Beard, H. K. Jessup, S. Lehar, K. W. Makar, M. Perez-­Melgosa,
checkpoint of Anti-­tumor T cell responses. Cell 162, 1217–1228 (2015). M. T. Sweetser, M. S. Schlissel, S. Nguyen, S. R. Cherry, J. H. Tsai, S. M. Tucker, W. M. Weaver,
11. C.-­H. Chang, J. Qiu, D. O’Sullivan, M. D. Buck, T. Noguchi, J. D. Curtis, Q. Chen, M. Gindin, A. Kelso, R. Jaenisch, C. B. Wilson, A critical role for Dnmt1 and DNA methylation in T cell
M. M. Gubin, G. J. W. van der Windt, E. Tonc, R. D. Schreiber, E. J. Pearce, E. L. Pearce, development, function, and survival. Immunity 15, 763–774 (2001).
Metabolic competition in the tumor microenvironment is a driver of cancer progression. 33. N. Alexander, D. M. Greenberg, Studies on the biosynthesis of serine. J. Biol. Chem. 214,
Cell 162, 1229–1241 (2015). 821–837 (1955).
12. P. M. Gullino, S. H. Clark, F. H. Grantham, The interstitial fluid of solid tumors. Cancer Res. 34. O. D. K. Maddocks, C. R. Berkers, S. M. Mason, L. Zheng, K. Blyth, E. Gottlieb, K. H. Vousden,
24, 780–794 (1964). Serine starvation induces stress and p53-­dependent metabolic remodelling in cancer
13. G. Liu, S. Burns, G. Huang, K. Boyd, R. L. Proia, R. A. Flavell, H. Chi, The receptor S1P1 cells. Nature 493, 542–546 (2013).
overrides regulatory T cell-­mediated immune suppression through Akt-­mTOR. Nat. 35. E. H. Ma, G. Bantug, T. Griss, S. Condotta, R. M. Johnson, B. Samborska, N. Mainolfi, V. Suri,
Immunol. 10, 769–777 (2009). H. Guak, M. L. Balmer, M. J. Verway, T. C. Raissi, H. Tsui, G. Boukhaled,
14. G. Liu, K. Yang, S. Burns, S. Shrestha, H. Chi, The S1P1-­mTOR axis directs the reciprocal S. Henriques da Costa, C. Frezza, C. M. Krawczyk, A. Friedman, M. Manfredi, M. J. Richer,
differentiation of TH1 and Treg cells. Nat. Immunol. 11, 1047–1056 (2010). C. Hess, R. G. Jones, Serine is an essential metabolite for effector T cell expansion. Cell
15. J. Rivera, R. L. Proia, A. Olivera, The alliance of sphingosine-­1-­phosphate and its receptors Metab. 25, 345–357 (2017).
in immunity. Nat. Rev. Immunol. 8, 753–763 (2008). 36. H. Kurniawan, D. G. Franchina, L. Guerra, L. Bonetti, L. S. Baguet, M. Grusdat, L. Schlicker,
16. L. Kappos, E. W. Radue, P. O’Connor, C. Polman, R. Hohlfeld, P. Calabresi, K. Selmaj, O. Hunewald, C. Dostert, M. P. Merz, C. Binsfeld, G. S. Duncan, S. Farinelle,
C. Agoropoulou, M. Leyk, L. Zhang-­Auberson, P. Burtin; FREEDOMS Study Group, A Y. Nonnenmacher, J. Haight, D. Das Gupta, A. Ewen, R. Taskesen, R. Halder, Y. Chen,
placebo-­controlled trial of oral fingolimod in relapsing multiple sclerosis. N. Engl. J. Med. C. Jäger, M. Ollert, P. Wilmes, V. Vasiliou, I. S. Harris, C. B. Knobbe-­Thomsen, J. D. Turner,
362, 387–401 (2010). T. W. Mak, M. Lohoff, J. Meiser, K. Hiller, D. Brenner, Glutathione restricts serine
17. M. Matloubian, C. G. Lo, G. Cinamon, M. J. Lesneski, Y. Xu, V. Brinkmann, M. L. Allende, metabolism to preserve regulatory T cell function. Cell Metab. 31, 920–936.e7 (2020).
R. L. Proia, J. G. Cyster, Lymphocyte egress from thymus and peripheral lymphoid organs 37. T. Muthusamy, T. Cordes, M. K. Handzlik, L. You, E. W. Lim, J. Gengatharan, A. F. M. Pinto,
is dependent on S1P receptor 1. Nature 427, 355–360 (2004). M. G. Badur, M. J. Kolar, M. Wallace, A. Saghatelian, C. M. Metallo, Serine restriction alters
18. F. Sallusto, C. Nicolo, R. De Maria, S. Corinti, R. Testi, Ceramide inhibits antigen uptake and sphingolipid diversity to constrain tumour growth. Nature 586, 790–795 (2020).
presentation by dendritic cells. J. Exp. Med. 184, 2411–2416 (1996). 38. E. Wang, W. P. Norred, C. W. Bacon, R. T. Riley, A. H. Merrill Jr., Inhibition of sphingolipid
19. J. Herz, J. Pardo, H. Kashkar, M. Schramm, E. Kuzmenkina, E. Bos, K. Wiegmann, R. Wallich, biosynthesis by fumonisins. Implications for diseases associated with Fusarium
P. J. Peters, S. Herzig, E. Schmelzer, M. Kronke, M. M. Simon, O. Utermohlen, Acid moniliforme. J. Biol. Chem. 266, 14486–14490 (1991).
sphingomyelinase is a key regulator of cytotoxic granule secretion by primary T 39. T. Yamazaki, X. O. Yang, Y. Chung, A. Fukunaga, R. Nurieva, B. Pappu, N. Martin-­Orozco,
lymphocytes. Nat. Immunol. 10, 761–768 (2009). H. S. Kang, L. Ma, A. D. Panopoulos, S. Craig, S. S. Watowich, A. M. Jetten, Q. Tian, C. Dong,
20. J. Wu, S. Ma, R. Sandhoff, Y. Ming, A. Hotz-­Wagenblatt, V. Timmerman, N. Bonello-­Palot, ccr6 regulates the migration of inflammatory and regulatory T cells. J. Immunol. 181,
B. Schlotter-­Weigel, M. Auer-­Grumbach, P. Seeman, W. N. Löscher, M. Reindl, F. Weiss, 8391–8401 (2008).
E. Mah, N. Weisshaar, A. Madi, K. Mohr, T. Schlimbach, R. M. H. Velasco Cárdenas, 40. T. Sekiya, S. Hibino, K. Saeki, M. Kanamori, S. Takaki, A. Yoshimura, Nr4a receptors regulate
J. Koeppel, F. Grünschläger, L. Müller, M. Baumeister, B. Brügger, M. Schmitt, G. Wabnitz, development and death of labile treg precursors to prevent generation of pathogenic
Y. Samstag, G. Cui, Loss of neurological disease HSAN-­I-­associated gene SPTLC2 impairs self-­reactive cells. Cell Rep. 24, 1627–1638.e6 (2018).
CD8+ T cell responses to infection by inhibiting T cell metabolic fitness. Immunity 50, 41. W. Liu, A. L. Putnam, Z. Xu-­Yu, G. L. Szot, M. R. Lee, S. Zhu, P. A. Gottlieb, P. Kapranov,
1218–1231.e5 (2019). T. R. Gingeras, B. F. de St Groth, C. Clayberger, D. M. Soper, S. F. Ziegler, J. A. Bluestone,
21. S. A. Apostolidis, N. Rodríguez-­Rodríguez, A. Suárez-­Fueyo, N. Dioufa, E. Ozcan, CD127 expression inversely correlates with FoxP3 and suppressive function of human
J. C. Crispín, M. G. Tsokos, G. C. Tsokos, Phosphatase PP2A is requisite for the function of CD4+ T reg cells. J. Exp. Med. 203, 1701–1711 (2006).
regulatory T cells. Nat. Immunol. 17, 556–564 (2016). 42. L. M. Francisco, V. H. Salinas, K. E. Brown, V. K. Vanguri, G. J. Freeman, V. K. Kuchroo,
22. C. R. Gault, L. M. Obeid, Y. A. Hannun, An overview of sphingolipid metabolism: From A. H. Sharpe, PD-­L1 regulates the development, maintenance, and function of induced
synthesis to breakdown. Adv. Exp. Med. Biol. 688, 1–23 (2010). regulatory T cells. J. Exp. Med. 206, 3015–3029 (2009).
23. J. D. Greenlee, M. R. King, A syngeneic MC38 orthotopic mouse model of colorectal 43. S. Amarnath, C. W. Mangus, J. C. Wang, F. Wei, A. He, V. Kapoor, J. E. Foley, P. R. Massey,
cancer metastasis. Biol. Methods Protoc. 7, bpac024 (2022). T. C. Felizardo, J. L. Riley, B. L. Levine, C. H. June, J. A. Medin, D. H. Fowler, The PDL1-­PD1
24. D. Dankort, D. P. Curley, R. A. Cartlidge, B. Nelson, A. N. Karnezis, W. E. Damsky Jr., M. J. You, axis converts human TH1 cells into regulatory T cells. Sci. Transl. Med. 3, 111ra120
R. A. DePinho, M. McMahon, M. Bosenberg, BrafV600E cooperates with Pten loss to induce (2011).
metastatic melanoma. Nat. Genet. 41, 544–552 (2009). 44. G. Xiao, A. Deng, H. Liu, G. Ge, X. Liu, Activator protein 1 suppresses antitumor T-­cell
25. Y. Zhang, R. Kurupati, L. Liu, X. Y. Zhou, G. Zhang, A. Hudaihed, F. Filisio, W. Giles-­Davis, function via the induction of programmed death 1. Proc. Natl. Acad. Sci. U.S.A. 109,
X. Xu, G. C. Karakousis, L. M. Schuchter, W. Xu, R. Amaravadi, M. Xiao, N. Sadek, C. Krepler, 15419–15424 (2012).

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 16 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

45. Y. Aikawa, K. Morimoto, T. Yamamoto, H. Chaki, A. Hashiramoto, H. Narita, S. Hirono, 68. A. Chaudhry, D. Rudra, P. Treuting, R. M. Samstein, Y. Liang, A. Kas, A. Y. Rudensky, CD4+
S. Shiozawa, Treatment of arthritis with a selective inhibitor of c-­Fos/activator protein-­1. regulatory T cells control TH17 responses in a stat3-­dependent manner. Science 326,
Nat. Biotechnol. 26, 817–823 (2008). 986–991 (2009).
46. N. C. Hait, J. Allegood, M. Maceyka, G. M. Strub, K. B. Harikumar, S. K. Singh, C. Luo, 69. M. A. Koch, G. Tucker-­Heard, N. R. Perdue, J. R. Killebrew, K. B. Urdahl, D. J. Campbell, The
R. Marmorstein, T. Kordula, S. Milstien, S. Spiegel, Regulation of histone acetylation in the transcription factor T-­bet controls regulatory T cell homeostasis and function during type
nucleus by sphingosine-­1-­phosphate. Science 325, 1254–1257 (2009). 1 inflammation. Nat. Immunol. 10, 595–602 (2009).
47. S. Hang, D. Paik, L. Yao, E. Kim, J. Trinath, J. Lu, S. Ha, B. N. Nelson, S. P. Kelly, L. Wu, 70. J. Wu, S. Ma, A. Hotz-­Wagenblatt, P. Angel, K. Mohr, T. Schlimbach, M. Schmitt, G. Cui,
Y. Zheng, R. S. Longman, F. Rastinejad, A. S. Devlin, M. R. Krout, M. A. Fischbach, Regulatory T cells sense effector T-­cell activation through synchronized JunB expression.
D. R. Littman, J. R. Huh, Bile acid metabolites control TH17 and Treg cell differentiation. FEBS Lett. 593, 1020–1029 (2019).
Nature 576, 143–148 (2019). 71. S.-­i. Koizumi, D. Sasaki, T.-­H. Hsieh, N. Taira, N. Arakaki, S. Yamasaki, K. Wang, S. Sarkar,
48. R. Wild, R. Gerasimaite, J.-­Y. Jung, V. Truffault, I. Pavlovic, A. Schmidt, A. Saiardi, H. J. Jessen, H. Shirahata, M. Miyagi, H. Ishikawa, JunB regulates homeostasis and suppressive
Y. Poirier, M. Hothorn, A. Mayer, Control of eukaryotic phosphate homeostasis by inositol functions of effector regulatory T cells. Nat. Commun. 9, 5344 (2018).
polyphosphate sensor domains. Science 352, 986–990 (2016). 72. L. Berod, C. Friedrich, A. Nandan, J. Freitag, S. Hagemann, K. Harmrolfs, A. Sandouk,
49. M. E. Welsch, A. Kaplan, J. M. Chambers, M. E. Stokes, P. H. Bos, A. Zask, Y. Zhang, C. Hesse, C. N. Castro, H. Bahre, S. K. Tschirner, N. Gorinski, M. Gohmert, C. T. Mayer,
M. Sanchez-­Martin, M. A. Badgley, C. S. Huang, T. H. Tran, H. Akkiraju, L. M. Brown, J. Huehn, E. Ponimaskin, W. R. Abraham, R. Muller, M. Lochner, T. Sparwasser, De novo
R. Nandakumar, S. Cremers, W. S. Yang, L. Tong, K. P. Olive, A. Ferrando, B. R. Stockwell, fatty acid synthesis controls the fate between regulatory T and T helper 17 cells. Nat. Med.
Multivalent small-­molecule Pan-­RAS inhibitors. Cell 168, 878–889.e29 (2017). 20, 1327–1333 (2014).
50. C. R. Vinson, P. B. Sigler, S. L. McKnight, Scissors-­grip model for DNA recognition by a 73. A. G. York, K. J. Williams, J. P. Argus, Q. D. Zhou, G. Brar, L. Vergnes, E. E. Gray, A. Zhen,

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024
family of leucine zipper proteins. Science 246, 911–916 (1989). N. C. Wu, D. H. Yamada, C. R. Cunningham, E. J. Tarling, M. Q. Wilks, D. Casero, D. H. Gray,
51. L. Chen, J. N. M. Glover, P. G. Hogan, A. Rao, S. C. Harrison, Structure of the DNA-­binding A. K. Yu, E. S. Wang, D. G. Brooks, R. Sun, S. G. Kitchen, T. T. Wu, K. Reue, D. B. Stetson,
domains from NFAT, Fos and Jun bound specifically to DNA. Nature 392, 42–48 (1998). S. J. Bensinger, Limiting cholesterol biosynthetic flux spontaneously engages Type I IFN
52. P. Zhou, L. J. Sun, V. Dötsch, G. Wagner, G. L. Verdine, Solution structure of the core signaling. Cell 163, 1716–1729 (2015).
NFATC1/DNA complex. Cell 92, 687–696 (1998). 74. R. Bao, J. Hou, Y. Li, J. Bian, X. Deng, X. Zhu, T. Yang, Adenosine promotes Foxp3 expression
53. K. J. Oestreich, H. Yoon, R. Ahmed, J. M. Boss, NFATc1 regulates PD-­1 expression upon T in Treg cells in sepsis model by activating JNK/AP-­1 pathway. Am. J. Transl. Res. 8,
cell activation. J. Immunol. 181, 4832–4839 (2008). 2284–2292 (2016).
54. S. J. Patterson, A. M. Pesenacker, A. Y. Wang, J. Gillies, M. Mojibian, K. Morishita, R. Tan, 75. Y. Wu, M. Borde, V. Heissmeyer, M. Feuerer, A. D. Lapan, J. C. Stroud, D. L. Bates, L. Guo,
T. J. Kieffer, C. B. Verchere, C. Panagiotopoulos, M. K. Levings, T regulatory cell chemokine A. Han, S. F. Ziegler, D. Mathis, C. Benoist, L. Chen, A. Rao, FOXP3 controls regulatory T cell
production mediates pathogenic T cell attraction and suppression. J. Clin. Invest. 126, function through cooperation with NFAT. Cell 126, 375–387 (2006).
1039–1051 (2016). 76. M. J. Barnden, J. Allison, W. R. Heath, F. R. Carbone, Defective TCR expression in transgenic
55. M. C. B. Tan, P. S. Goedegebuure, B. A. Belt, B. Flaherty, N. Sankpal, W. E. Gillanders, mice constructed using cDNA-­based α‐ and β‐chain genes under the control of
T. J. Eberlein, C.-­S. Hsieh, D. C. Linehan, Disruption of CCR5-­dependent homing of heterologous regulatory elements. Immunol. Cell Biol. 76, 34–40 (1998).
regulatory T cells inhibits tumor growth in a murine model of pancreatic cancer. J. 77. G. Cui, M. M. Staron, S. M. Gray, P.-­C. Ho, R. A. Amezquita, J. Wu, S. M. Kaech, IL-­7-­induced
Immunol. 182, 1746–1755 (2009). glycerol transport and TAG synthesis promotes memory CD8+ T Cell longevity. Cell 161,
56. D. Baatar, P. Olkhanud, K. Sumitomo, D. Taub, R. Gress, A. Biragyn, Human peripheral 750–761 (2015).
blood t regulatory cells (Tregs), functionally primed CCR4+ Tregs and unprimed 78. Y. Hao, S. Hao, E. Andersen-­Nissen, W. M. Mauck III, S. Zheng, A. Butler, M. J. Lee, A. J. Wilk,
CCR4− Tregs, regulate effector T cells using FasL. J. Immunol. 178, 4891–4900 (2007). C. Darby, M. Zager, P. Hoffman, M. Stoeckius, E. Papalexi, E. P. Mimitou, J. Jain,
57. M. J. Watson, P. D. A. Vignali, S. J. Mullett, A. E. Overacre-­Delgoffe, R. M. Peralta, A. Srivastava, T. Stuart, L. M. Fleming, B. Yeung, A. J. Rogers, J. M. McElrath, C. A. Blish,
S. Grebinoski, A. V. Menk, N. L. Rittenhouse, K. DePeaux, R. D. Whetstone, D. A. A. Vignali, R. Gottardo, P. Smibert, R. Satija, Integrated analysis of multimodal single-­cell data. Cell
T. W. Hand, A. C. Poholek, B. M. Morrison, J. D. Rothstein, S. G. Wendell, G. M. Delgoffe, 184, 3573–3587.e29 (2021).
Metabolic support of tumour-­infiltrating regulatory T cells by lactic acid. Nature 591, 79. M. Dodt, J. T. Roehr, R. Ahmed, C. Dieterich, FLEXBAR-­flexible barcode and adapter
645–651 (2021). processing for next-­generation sequencing platforms. Biology 1, 895–905 (2012).
58. A. Angelin, L. Gil-­de-­Gómez, S. Dahiya, J. Jiao, L. Guo, M. H. Levine, Z. Wang, W. J. Quinn III, 80. B. Langmead, S. L. Salzberg, Fast gapped-­read alignment with Bowtie 2. Nat. Methods 9,
P. K. Kopinski, L. Wang, T. Akimova, Y. Liu, T. R. Bhatti, R. Han, B. L. Laskin, J. A. Baur, 357–359 (2012).
I. A. Blair, D. C. Wallace, W. W. Hancock, U. H. Beier, Foxp3 reprograms T Cell metabolism to 81. Y. Zhang, T. Liu, C. A. Meyer, J. Eeckhoute, D. S. Johnson, B. E. Bernstein, C. Nusbaum,
function in low-­glucose, high-­lactate environments. Cell Metab. 25, 1282–1293.e7 (2017). R. M. Myers, M. Brown, W. Li, X. S. Liu, Model-­based analysis of ChIP-­Seq (MACS). Genome
59. B. Reichl, L. Niederstaetter, T. Boegl, B. Neuditschko, A. Bileck, J. Gojo, W. Buchberger, Biol. 9, R137 (2008).
A. Peyrl, C. Gerner, Determination of a tumor-­promoting microenvironment in recurrent 82. S. Heinz, C. Benner, N. Spann, E. Bertolino, Y. C. Lin, P. Laslo, J. X. Cheng, C. Murre, H. Singh,
medulloblastoma: A multi-­omics study of cerebrospinal fluid. Cancers 12, 1350 (2020). C. K. Glass, Simple combinations of lineage-­determining transcription factors prime
60. Y. Gu, X. Niu, L. Yin, Y. Wang, Y. Yang, X. Yang, Q. Zhang, H. Ji, Enhancing fatty acid cis-­regulatory elements required for macrophage and B cell identities. Mol. Cell 38,
catabolism of macrophages within aberrant breast cancer tumor microenvironment can 576–589 (2010).
re-­establish antitumor function. Front. Cell Dev. Biol. 9, 665869 (2021). 83. F. Ramírez, D. P. Ryan, B. Grüning, V. Bhardwaj, F. Kilpert, A. S. Richter, S. Heyne, F. Dündar,
61. A. Graf, L. Trofimova, A. Ksenofontov, L. Baratova, V. Bunik, Hypoxic adaptation of T. Manke, deepTools2: A next generation web server for deep-­sequencing data analysis.
mitochondrial metabolism in rat cerebellum decreases in pregnancy. Cells 9, 139 Nucleic Acids Res. 44, W160–W165 (2016).
(2020). 84. A. R. Quinlan, I. M. Hall, BEDTools: A flexible suite of utilities for comparing genomic
62. G. V. Richieri, A. M. Kleinfeld, Free fatty acids are produced in and secreted from target features. Bioinformatics 26, 841–842 (2010).
cells very early in cytotoxic T lymphocyte-­mediated killing. J. Immunol. 147, 2809–2815
(1991). Acknowledgments: We thank X. Jiang (SUNY Downstate Medical Center, New York) and
63. A. M. Kleinfeld, C. Okada, Free fatty acid release from human breast cancer tissue inhibits A. Rudensky (Memorial Sloan Kettering Cancer Center, New York) for providing Sptlc2Fl/Fl mice
cytotoxic T-­lymphocyte-­mediated killing. J. Lipid Res. 46, 1983–1990 (2005). and the Foxp3YFP-­Cre mice. We also thank R. Locksley (UCSF) and D. Zemmour (University of
64. A. G. Levine, A. Mendoza, S. Hemmers, B. Moltedo, R. E. Niec, M. Schizas, B. E. Hoyos, Chicago) for providing the sequence of IRES-­YFP-­Cre, which was used for the scRNA-­seq
E. V. Putintseva, A. Chaudhry, S. Dikiy, S. Fujisawa, D. M. Chudakov, P. M. Treuting, analysis of Foxp3YFP-­Cre cells. J. Boss (Emory University, Atlanta) provided the Pdcd1-­luciferase
A. Y. Rudensky, Stability and function of regulatory T cells expressing the transcription reporter plasmid. We thank M. Bosenberg (Yale University, New Haven) for providing YUMMER
factor T-­bet. Nature 546, 421–425 (2017). cells. We thank B. von Tümpling-­Radosta (Lipid Pathobiochemistry at DKFZ Heidelberg) for
65. Y. Wang, M. A. Su, Y. Y. Wan, An essential role of the transcription factor GATA-­3 for the excellent technical assistance. Funding: S.M. is supported by a Chinese Scholar Council
function of regulatory T Cells. Immunity 35, 337–348 (2011). fellowship and the National Natural Science Foundation of China (#82303187). Y.M. is
66. Y. Zheng, A. Chaudhry, A. Kas, P. deRoos, J. M. Kim, T. T. Chu, L. Corcoran, P. Treuting, supported by the Huatuo Postdoctoral Research fellowship. A. Madi is supported by the
U. Klein, A. Y. Rudensky, Regulatory T-­cell suppressor program co-­opts transcription Helmholtz International Graduate School. J.W. and G.C. are supported by funding from the
factor IRF4 to control TH2 responses. Nature 458, 351–356 (2009). Institute of Health and Medicine, Hefei Comprehensive National Science Center. N.P. and G.C.
67. E. Cretney, A. Xin, W. Shi, M. Minnich, F. Masson, M. Miasari, G. T. Belz, G. K. Smyth, are supported by the German Research Foundation (DFG, #CU375/5-­2) and the Hector
M. Busslinger, S. L. Nutt, A. Kallies, The transcription factors Blimp-­1 and IRF4 jointly Foundation (#M20102). G.C. is also supported by a CRI Lloyd J. Old STAR Award (#3914). P.G. is
control the differentiation and function of effector regulatory T cells. Nat. Immunol. 12, supported by the National Natural Science Foundation of China (#32130057 and #31922037).
304–311 (2011). X.W. is supported by the National Natural Science Foundation of China (#32170742), the Start

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 17 of 18

S c i e n c e Imm u n o l o g y | R e s e ar c h A r t i c l e

Fund for Specially Appointed Professors of Jiangsu Province, and the Start Fund for High-­level competing interests. Data and materials availability: The GEO accession number for the
Talents of Nanjing Medical University (#NMUR2020009). Author contributions: G.C., S.M., P.G., RNA, ATAC, and CUT & RUN sequencing data is GSE213756. All data needed to evaluate the
X.W., and N.P. designed the experiments. S.M. performed most of the biological experiments. conclusions in the paper are present in the paper and/or the Supplementary Materials.
R.S. extracted lipids, performed the lipid mass spectrometry experiments, and analyzed the Additional data are available from authors upon request. Underlying tabulated data for all
results. F.S. and H.-­R.R. assisted in the experimental design, library sequencing, and data figures can be found in data file S2.
analysis. Q.S. and X.W. performed the bioinformatics analysis. X.L. assisted in the MST assay.
Zhaolong Li performed the docking assay. J.W., Y.M., F.S., A. Madi, N.W., A. Mieg, M.H., X.Y., K.M.,
N.t.B., Zhe Li, F.Z., and G.P. helped with the experiments. S.M. and G.C. analyzed the biological Submitted 27 January 2023
data. G.C. wrote the manuscript with input from other authors. Competing interests: S.M. and Resubmitted 23 November 2023
G.C. hold a patent application related to this study [EP 19180971.4 (18 June 2019)]. G.C. Accepted 15 March 2024
receives research funding from Bayer AG and Boehringer Ingelheim. The funding has no direct Published 19 April 2024
relevance to the findings presented in this study. The other authors declare that they have no 10.1126/sciimmunol.adg8817

Downloaded from https://www.science.org at Institute Pasteur of Shanghai, CAS on May 12, 2024

Ma et al., Sci. Immunol. 9, eadg8817 (2024) 19 April 2024 18 of 18