T cells from ICI-responsive tumors show greater cholesterol uptake. To study cholesterol uptake and the immunotherapy response in tumor-infiltrating T cells, we utilized cell lines that model TNBC but exhibit different responses to immunotherapy. EO771 cells, one of the only mammary tumor cell lines derived from spontaneous tumorigenesis in C57BL/6J mice, responds to immunotherapy and has been used to test effects of therapeutic agents to improve immunogenicity (14). In contrast, AT-3 cells are derived from transgenic MMTV-PyMT mice and respond poorly to immunotherapies, including anti–PD-1, despite the presence of intratumoral T cells (15, 16). To directly compare effects of anti–PD-1 antibody therapy on tumor growth, we orthotopically implanted either EO771 or AT-3 breast cancer cells with syngeneic immortalized mouse mammary fibroblasts. Three days after implanting cells, we randomly assigned mice to treatment with anti–PD-1 antibody therapy or vehicle. Mice received a total of 4 doses administered every 3 days. We quantified tumor size by caliper measurements and monitored survival until mice reached humane endpoints for tumor burden. Mice with EO771 tumors showed remarkable reductions in tumor growth (P = 0.0001) and significantly increased overall survival in response to anti–PD-1 therapy (P = 0.031) (Figure 1, A–C). By comparison, anti–PD-1 treatment elicited no change in tumor growth (Figure 1, D and E) or survival over control in mice with AT-3 tumors (Figure 1F). These data validate responsive and treatment refractory mouse models of immunotherapy in TNBC, providing contrasting systems to investigate cholesterol as a marker of T cell activation.

E0771 tumors respond to anti–PD-1 immunotherapy, while AT-3 tumors do not. Three days after orthotopically injecting E0771 or AT-3 breast cancer cells plus mouse mammary fibroblasts into syngeneic C57BL/6J mice, we randomly assigned animals to treatment with anti–PD-1 antibody or PBS vehicle every 3 days for 4 doses total. Graphs show mean values ± SEM (symbols) and calculated logistic regression (smooth line) for E0771 (A) or AT-3 (D) tumors (n = 6 control; n = 8 anti–PD-1) treated with anti–PD-1 antibody or PBS. (B and E) Growth of E0771 and AT3 tumor growth, respectively, for individual mice over time. Three tumors from the anti–PD-1 group failed to grow tumors and are overlapped on the x axis of panel B. We analyzed differences in tumor growth data by logistic regression. Survival curves demonstrate that anti–PD-1 treatment significantly prolonged survival for mice with E0771 tumors (C) but not with AT-3 (F), as analyzed by the Mantel-Cox test.

We measured uptake of cholesterol in T cells isolated from EO771 or AT-3 breast tumors and restimulated ex vivo in the presence of fluorescent 3-NBD-cholesterol. Due to the location and orientation of the NBD tag, this fluorescent cholesterol models cholesterol orientation in membranes better than previous fluorescently tagged cholesterol analogs (17). Solid tumors rarely contain naive T cells, so ex vivo tumor-infiltrating T cells were stimulated with low-level CD3ε for 18 hours (18). We assessed uptake of 3-NBD-cholesterol in combination with the activation marker CD69 on T cells using spectral cytometry (Figure 2, A and D). Cholesterol uptake correlated significantly with expression of the activation marker, CD69 (P = 0.0151) (Figure 2, B and C). Additionally, cholesterol uptake increased to a markedly greater extent in CD8+ (P = 0.026) and CD4+ (P = 0.0286) T cells from EO771 tumors when compared with T cells from AT-3 (nonresponsive) tumors (Figure 2, E and F).

T cells from ICI-responsive versus -nonresponsive tumors have greater uptake of fluorescent cholesterol. T cells were isolated from ICI-responsive EO771 or ICI-nonresponsive AT-3 tumors and left unstimulated (us) or restimulated on anti-CD3e–coated (stim) dishes for 18 hours in the presence of 3-NBD–labeled cholesterol. We determined uptake of labeled cholesterol in activated, CD69+ CD8+ (A) or CD4+ (B) T cells by flow cytometry. CD8+ (C) and CD4+ (D) T cells from EO771 tumors show significantly greater percentages of CD69+ cells with uptake of 3-NBD cholesterol than from ICI-nonresponsive AT-3 tumors. Stimulated CD8+ (E) and CD4+ (F) T cells from EO771 tumors also exhibited significantly higher mean fluorescence intensity (MFI) for cholesterol uptake. Data are combined from 2 experiments. *P < 0.05, **P < 0.01 by 2-tailed Student’s t test (C and D) or nonparametric Mann-Whitney test (E and F).

To assess uptake of cholesterol in T cells in vivo, we injected tumor-bearing mice with a fluorescent BODIPY–labeled cholesterol. BODIPY-labeled cholesterol effectively partitions with intracellular cholesterol, remains stable in vivo, and exhibits greater fluorescence than the 3-NBD label (19). Twenty-four hours after injecting BODIPY-cholesterol, we collected intratumoral T cells for flow cytometry. T cells from responsive EO771 tumors showed a higher uptake of labeled cholesterol when compared with AT-3 tumors, as seen in histograms (Figure 3, A and B) and normalized BODIPY mean fluorescence intensity (MFI) in both CD8+ (P < 0.0001) and CD4+ (P = 0.0009) T cells (Figure 3, C and D). Prior studies reported increased expression of PD-1 on T cells that take up more cholesterol. Indeed, PD-1 expression on T cells increased significantly on CD8+ T cells from EO771 tumors with higher cholesterol uptake (P = 0.004) (Figure 3E). Modulation of PD-1 did not occur on CD4+ T cells (Figure 3F). Together, these experiments demonstrate increased uptake of cholesterol in T cells from tumors that respond to immunotherapy.

T cells in ICI-responsive EO771 tumors show greater uptake of fluorescent cholesterol in vivo. We injected C57BL/6J mice intraperitoneally with cholesterol labeled with BODIPY and euthanized animals 24 hours later to collect and dissociate tumors for flow cytometry (n = 5 each for EO771 and AT-3). Plots for (A) CD8+ and (B) CD4+ T cells show accumulation of BODIPY-cholesterol in cells from individual tumors from EO771 and AT-3 tumors relative to vehicle only or isotype antibody control. (C) CD8+ and (D) CD4+ T cells in EO771 tumors showed significantly higher fold change accumulation of fluorescent cholesterol relative to FMO control. (E) CD8+, but not (F) CD4+, T cells in EO771 tumors also expressed higher levels of PD-1. **P < 0.01,***P < 0.001, ****P < 0.0001 for differences between means using nonparametric Mann-Whitney tests (C and D; n = 5 mice per group), while differences between T cell population percentages were assessed using 2-tailed Student’s t test (E and F).

Activated T cells increase cholesterol uptake. Cholesterol is required for cell division processes, and its presence in the plasma membrane of T cells stabilizes TCR nanoclusters to enhance overall activation (20). Cholesterol uptake as related to cell status in T cells remains a controversial topic. One study found that hypercholesteremia in T cells reflected proliferation of T cells (21). In contrast, elevated cholesterol in T cells also reportedly reflects exhaustion (7). To further investigate to what extent levels of cholesterol change with activation, we treated naive mouse T cells with increasing amounts of anti-CD3ε and proportional increases in anti-CD28 antibodies (3 times the anti-CD3ε concentration). Using an in vitro assay to quantify total cholesterol, free cholesterol, and cholesterol esters, we established that activation increased total cholesterol concentration in T cells, with differences between 0 μg/mL and 1 μg/mL or 10 μg/mL of anti-CD3e (Supplemental Figure 1 and Supplemental Figure 2A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.175320DS1). The concentration of cholesterol esters did not change with increasing activation in T cells after 24 hours (Supplemental Figure 2A). While this assay demonstrates that total cholesterol increased in T cells after TCR activation, these data do not distinguish between synthesis and uptake of cholesterol.

To selectively measure uptake of cholesterol during T cell activation, we leveraged a PET radiotracer analog of cholesterol, eFNP-59, developed by our research group. Our group has validated eFNP-59 as an imaging probe for cholesterol trafficking and uptake in humans (10). Because we added eFNP-59 to the culture medium, any accumulated radiotracer in T cells represents uptake, rather than de novo synthesis, during activation. We incubated activated T cells with eFNP-59 according to the diagram in Supplemental Figure 2B. After washing cells to remove extracellular radiotracer, we used autoradiography with a phosphor imaging screen to measure uptake by T cells. Phosphor imaging allows scalability of samples (directly in 96-well plates) while maintaining sensitivity with a large linear range of detection. After 1 hour of incubation, uptake of eFNP-59 increased progressively with higher concentrations of activating antibodies. Each T cell treatment displayed significant differences between levels of anti-CD3 stimulation with unique EC50 values when tested against all data points in logistical nonlinear regression (P < 0.0001) (Supplemental Figure 2C and Supplemental Figure 3).

To extend these findings to humans, we treated activated human T cells from peripheral blood with eFNP-59, as outlined in Supplemental Figure 2B. We found that human T cells also take up eFNP-59 when activated, but they were less sensitive to exogenous cholesterol uptake when compared with mice (P < 0.0001) (Supplemental Figure 2D). To study further the impact of cholesterol trafficking on the human immune system, we analyzed data from the Immune Cell Atlas, originating from Martin et al. and hosted by the Broad Institute (22). These data comprise an annotated set of immune cells from an uninflamed lamina propria, an excellent tissue site for studying varied immune responses without the presence of disease. From this data set, we utilized the T cell clusters previously annotated by gene expression (Supplemental Figure 2E). Genes responsible for uptake and trafficking of cholesterol displayed increased relative expression in the cycling cluster compared with other clusters (Supplemental Figure 2F). These included genes specifically annotated for cholesterol uptake and trafficking, including scavenger receptors like SCP2 and SCARB1, endosome-associated proteins such as STARD3NL, NPC1, NPC2, and transcription factor SREBF2 (23–27). Additional upregulated genes, COMMD1, WASH1, ANXA2, and LAMTOR1, function in cholesterol uptake and trafficking as well as other cellular processes (28–31). Increased expression of genes related to cholesterol uptake and trafficking in activated human T cells supports translatability of these findings.

Uptake of eFNP-59 in T cells is elevated with activation and immunotherapy in mouse TNBC. To assess the relationship of eFNP-59 uptake in activated versus exhausted T cells, we injected mice with EO771 cells and treated animals with anti–PD-1 or vehicle, following the protocol in Figure 1. We euthanized mice 4 days after the final injection. Following isolation of TILs from the tumor, we reactivated these cells with a moderate stimulus (2 μg/mL anti-CD3), a strong exhausting stimulus (10 μg/mL anti-CD3), or control stimulus for 1 day. We then incubated cells with eFNP-59, quantified radioactivity normalized to numbers of cells, and stained/fixed cells for downstream analyses (Figure 4A). Radioactivity in each group increased based on dose of anti-CD3 with cells from mice treated with anti–PD-1, exhibiting greater uptake than mice treated with vehicle control for each stimulus (Figure 4B). When comparing each treatment group versus amount of added eFNP-59, the 1000-nCi dose showed best separation for further analyses (Figure 4B). We identified a linear relationship between uptake of eFNP-59 in T cells for the 1000-nCi dose versus levels of the activation marker CD69, with significant separation between treatment groups (P < 0.0001) (Figure 4C). A multiparametric display also demonstrated that exhaustion markers TIM3 and LAG3 also increased as radioactivity and activation signal increased (Figure 4D). To better compare uptake of eFNP-59 in activated versus exhausted T cells, we plotted triple-positive exhausted T cells (PD-1+LAG3+TIM3+) against radioactivity (Figure 4E). This relationship showed a more logarithmic plateau in which moderate and strong activation signals produced similar measures of exhausted triple-positive T cells when compared with uptake of eFNP-59 (Figure 4E). We also compared double-positive T cells (cells with any combination of 2 positive markers for PD-1, LAG3, or TIM3 that are less exhausted but activated) with uptake of eFNP-59. Populations of double-positive cells and uptake of eFNP-59 increased to a greater extent with increasing stimulus (Figure 4F). Overall, we found that cholesterol uptake maintained a more linear relationship to T cell activation than exhaustion, and anti–PD-1 immunotherapy improved overall uptake of eFNP-59 in functioning activated T cells.

eFNP-59 uptake in T cells correlates directly with activation, differing from exhaustion. (A) Mice were inoculated with E0771 and treated with and without anti–PD-1 immunotherapy. T cells were extracted, activated, and activation status compared to cholesterol uptake as in the diagram. (B) Cholesterol uptake was first compared between activation and immunotherapy groups with different amounts of eFNP-59. Using the data from the 1000-nCi treatment, (C) T cell activation marker CD69 was compared to normalized activity in a scatter plot and (D) multiparametric plot also assessing activation signal with pseudocolor scale, estimation of triple-positive T cells with bubble scale, and information on immunotherapy status (samples above or below dashed line). (E) Normalized activity (cholesterol uptake) was directly compared to triple-positive T cells and then again as (F) a multiparametric plot with pseudocolor plot for double-positive T cells, bubble size for anti-CD3 stimulus, and immunotherapy status (above or below dashed line). ****P < 0.0001 by least squares fit.

To test uptake of eFNP-59 in T cells in vivo and use of this radiotracer to predict and monitor response to immunotherapy, we established mice with EO771 or AT-3 breast tumors. After tumors reached approximately 7 mm in diameter, we randomly assigned mice to treatment with anti–PD-1 or vehicle. We hypothesized that anti–PD-1 treatment would further alter cholesterol uptake, with reinvigoration of endogenous T cells after a single dose of ICI. Four days after treatment, we intravenously injected mice with 100 μCi of eFNP-59. We chose to analyze cholesterol uptake 4 days after anti–PD-1 treatment to allow the ICI treatment to impact T cell activation/function before measuring cholesterol uptake. We quantified uptake of eFNP-59 in total dissociated splenocytes, and in T cells recovered from dissociated tumors. Splenocytes from mice treated with vehicle only accumulated comparable amounts of eFNP-59 for both the EO771 and AT-3 groups (P = 0.0039) (Figure 5A). Relative to vehicle only, anti–PD-1 treatment modestly increased radiotracer uptake in splenocytes from both groups, although amounts differed significantly only for mice with AT-3 tumors (P = 0.0039). Isolated T cells from the vehicle-treated EO771 tumors accumulated more eFNP-59 than vehicle-treated AT-3 tumors, reproducing results with fluorescent cholesterol analogs (Figure 5B). Furthermore, T cells recovered from anti–PD-1–treated EO771 tumors had significantly greater uptake of eFNP-59 than anti–PD-1–treated AT-3 tumors (P = 0.01) (Figure 5B). These experiments demonstrate that (a) baseline uptake of eFNP-59 before ICI treatment correlates with response to therapy in these models, and (b) change in uptake of this radioactive cholesterol analog reflects T cell response to ICI.

T cells in ICI-responsive tumors had greater uptake of eFNP-59. When tumors reached approximately 70 mm2, we treated mice with anti–PD-1 antibody or control for 4 days. Mice were then injected with 100 μCi of eFNP-59 followed by T cell isolation protocols. (A) Graph shows uptake of eFNP-59 per microgram of spleen tissue measured by scintigraphy. Symbols show individual mice with annotations for mean values and standard deviations. (B) We isolated CD4+ and CD8+ tumor-infiltrating lymphocytes (TILs) by positive selection with immunomagnetic beads and determined accumulation of eFNP-59 normalized to total cell protein with a BCA assay. Representative data from 2 experimental replicates, with statistical comparisons by 1-way ANOVA with Dunn’s multiple-comparison test. *P < 0.05; **P < 0.01.

Cycling T cells in human TNBC tumors maintain elevated expression of cholesterol trafficking genes. To translate these discoveries to human breast cancer, we queried publicly available data sets and analyzed single-cell RNA sequencing (scRNA-seq) data from human TNBC (32). Louvain clusters of TNBC patient cells without additional characterization or bias reported 4 major T cell states (Figure 6A). These states included resting/naive T cells, highly activated/cycling T cells, transitional effectors (toward exhaustion), and exhausted effectors. We annotated T cell states based on highly expressed genes in each Louvain cluster: (a) cycling T cells are defined by the high expression of genes related to microtubule reorganization and mitosis, including upregulation of STNM1 and MKI67; (b) transitional state cytotoxic T cells based on high expression of GZMK and moderate levels of PDCD1 and HAVRC2; (c) dysfunctional T cells classified by high expression of PDCD1, HAVCR2, and CXCL13 plus expression of CSF1; and (d) resting memory T cells defined by high expression of CXCR4 and ribosomal proteins (33–35). Cell counts of clusters from each patient revealed that patients with large T effector populations (whether transitional or exhausted) also carried a smaller population of actively proliferating T cells (Figure 6B). We determined and annotated T cell states of each cluster using top gene expression of the Louvain clusters (Figure 4B). Cycling T cells displayed the greatest expression of genes directly and indirectly involved in cholesterol uptake and trafficking (Figure 6C). Furthermore, genes involved in cholesterol distribution (Figure 6D), endosomal transport (Figure 6, E and F), receptor recycling (Figure 6G), receptor expression (Figure 6H), and transcription factor control (Figure 6I) were modulated in the cycling T cell populations when compared with other clusters (Supplemental Table 1 for all comparisons). When analyzed using Enrichr, we found upregulation families of genes related to “Regulation of Cholesterol Biosynthesis By SREBF” and “Activation Of Gene Expression By SREBF” in the cycling T cell cluster (Supplemental Table 2). Overall, these data show some patients with TNBC develop tumors that produce a population of cycling T cells (often accompanied by an expanded population of effector T cells; Figure 6B), with features of gene expression indicative of cholesterol uptake and trafficking. Together, these data suggest that increased uptake of cholesterol demarks activated, antitumor T cells in human TNBC.

Cycling T cell populations in patients with TNBC upregulate genes related to cholesterol metabolism. (A) Reanalysis of single-cell RNA sequencing data (29) with Cellenics software displays Louvain clusters of annotated T cell states, including custom cell sets derived from these clusters: T memory/resting, highly activated, T effector/transitional state, and T effector/dysfunctional. (B) Plot shows proportions of the absolute counts across for various T cell subsets across TNBC patients (n = 8429 cells from 9 tumors). (C) Clustered averaged gene expression data reveal upregulation of relevant genes involved in uptake and intracellular trafficking of cholesterol in cycling T cells. (D–I) Violin plots of normalized expression of specific genes: (D) ANXA2 (cholesterol distribution to the plasma membrane), (E) LAMTOR1 (endosomal transport), (F) STARD3NL (endosomal transport), (G) COMMD1 (LDLR recycling), (H) LDLR (cholesterol uptake), and (I) SREBF2 (positive regulation of cholesterol uptake and synthesis). The ⱡ symbol indicates P < 10–20 between the cycling cluster and other T cell clusters, as determined by Welch’s t test.