Metabolic engineering to facilitate anti-tumor immunity

Summary

Fructose consumption is elevated in western diets, but its impact on anti-tumor immunity is unclear. Fructose is metabolized in the liver and small intestine, where fructose transporters are highly expressed. Most tumors are unable to drive glycolytic flux using fructose, enriching fructose in the tumor microenvironment (TME). Excess fructose in the TME may be utilized by immune cells to enhance effector functions if engineered to express the fructose-specific transporter GLUT5. Here, we show that GLUT5-expressing CD8⁺ T cells, macrophages, and chimeric antigen receptor (CAR) T cells all demonstrate improved effector functions in glucose-limited conditions in vitro. GLUT5-expressing T cells show high fructolytic activity in vitro and higher anti-tumor efficacy in murine syngeneic and human xenograft models in vivo, especially following fructose supplementation. Together, our data demonstrates that metabolic engineering through GLUT5 enables immune cells to efficiently utilize fructose and boosts anti-tumor immunity in the glucose-limited TME.

eTOC

While high fructose intake has been implicated in the progression of metabolic disease and cancer, Schild et al. take advantage of the availability of fructose in the tumor microenvironment. Engineered immune cells can use metabolic flexibility and leverage fructose as a nutrient source to fuel their metabolism and facilitate function in vivo.

Graphical Abstract

Introduction

Immunotherapy has surged as a cancer therapy, but its effect has only been demonstrated in a subset of cancers. Growing evidence suggests metabolic conditions in the tumor microenvironment (TME) are unfavorable for immune cells and dampen their antitumor functions^1,2. Cancer cells often harbor multiple genetic mutations in signaling pathways, including phosphoinositide-3-kinase (PI3K), leading to increased glucose metabolism and a glucose-deprived TME^3,4. The metabolic needs of other tumor resident cells, such as fibroblasts and macrophages may further deplete the TME of nutrients^5–7.

Unfortunately, the effector function of immune cells, particularly effector T (T_eff) cells, is heavily dependent on glycolysis and may be affected by the glucose-deficient TME⁸. T cell activation is accompanied by the upregulation of the glucose transporter GLUT1 and glycolytic enzymes^9–11. Enhanced glycolysis in T cells provides biosynthetic and energetic precursors and supports interferon gamma (IFNγ) production through epigenetic reprogramming^12,13. Recent studies showed that decreased levels of glycolytic intermediates in T_eff cells impair Ca²⁺ signaling¹⁴ and N-linked glycosylation¹⁵, which are critical for T_eff function. It was observed that the high glycolytic rates of melanoma and non-small-cell lung cancer (NSCLC) correlate with resistance to adoptive T cell therapy¹⁶. Recent work showed increased enolase activity in tumor-infiltrating lymphocytes (TILs) during checkpoint blockade, further linking glycolytic flux to potential enhancement of immunotherapy^17,18. If the availability of nutrients for T cells could be increased in this deprived condition, glycolytic flux could potentially be restored.

Fructose consumption, especially high fructose corn syrup, has increased in the recent decades in the United States^19–21. Under homeostatic conditions, fructose is metabolized in the small intestine and liver via cellular import through fructose transporters SLC2A2 (GLUT2) and SLC2A5 (GLUT5)^22–25. It is subsequently converted to fructose-1-phosphate by ketohexokinase (KHK)²⁶ or directly to fructose-6-phosphate by hexokinase²⁷. In a normal physiological context, SLC2A5 is predominantly transcribed in fructolytic tissues like the small intestine (Figure S1A). SLC2A5 expression is generally low (Figure S1B) and fructose uptake slow, in most cancers²⁸ except for some hematopoietic and lymphoid cancers discussed in our previous work²⁵. Therefore, GLUT5-derived carbons typically cannot populate glycolytic intermediates at near the level of glucose. While fructose has been implicated as a potential carbon source for some cancers^29,30, the underlying mechanism remains unclear, especially as most cancer cells do not have the ability to drive glycolytic flux using fructose.

Given this information, we aimed to impart metabolic flexibility on immune cells, which are present under low glucose environments in the TME. Specifically, we engineer T_eff, chimeric antigen receptor (CAR) T cells, and macrophages to utilize fructose as an alternative to glucose for their major carbon source to enhance antitumor functions.

Results

Increased fructose transport facilitates T cell killing in glucose-limited conditions in vitro

We overexpressed retrovirally encoded Glut5 in murine OT-I T cells post activation and did not observe any major protein expression abnormalities in GLUT5 (GT5)-expressing OT-I cells compared to empty vector (EV) control via mass spectrometry (Figures 1A, S1C–D). To monitor mouse GLUT5 expression via flow cytometry, we generated a mouse anti-mouse GLUT5 monoclonal antibody 8f7 (Figure S1E). To evaluate whether the ectopically expressed GLUT5 is active, we incubated EV and GT5 OT-I cells with isotopic ¹³C fructose or glucose and measured metabolism through lactate generation. GT5 cells grown in isotopic fructose were able to generate lactate while EV T cells were not (Figure 1B). EV and GT5 OT-I cells grown in glucose had comparable levels of lactate (Figure 1B). A compilation of data from the human metabolome database (HMDB) suggested that concentrations of glucose are limited when transitioning from circulating blood to the TME³¹, though fructose is abundant in various tumors (Figure 1C)^30,32. TME glucose concentration has been reported to be 0.1–5 mM³³ and was observed by us to be approximately 0.5 mM in B16 murine melanoma tumors (Figure 1C). While absolute TME fructose concentration is unknown, fructose concentration in blood and bone marrow can reach 5 mM under pathogenic conditions³². Therefore, we decided to conduct in vitro experiments under similar physiological concentrations. GT5 T cells cultured in 1mM glucose plus 2.5 mM fructose grew faster than EV T cells (Figure 1D). Undertaking a more unbiased approach, we incubated B16-OVA-Luciferase melanoma cells with EV or GT5 OT-I cells in various fructose concentrations. B16-OVA luminescence was lower in GT5 co-cultures across all fructose concentrations, as compared to control, with the largest difference between 2-3 mM (Figure 1E). This finding supports our usage of fructose at or below 5 mM. Similarly, in co-culture experiments using media with 0.5 mM glucose and 2.5 mM fructose, GT5 OT-I cells killed more B16-OVA cells, quantified through Annexin V staining (Figures 1F, S1F). GT5 OT-I cells also produced higher levels of IFNγ and TNFα cytokines than EV OT-I cells (Figures 1G–H, S1G). To explore whether GT5 engineering alters the epigenome of T cells, we analyzed enzymes and metabolites with potential epigenetic effects in GT5 and EV OT-I cells using mass spectrometry (Figure S1H–I). No significant differences were found in profiled enzyme expression or metabolite levels. Comparable AMP/ATP ratios indicated similar AMPK activation in both GT5 and EV OT-I cells (Figure S1J). Altogether, our findings indicate that GT5 engineering does not cause epigenetic modifications in T cells.

Figure 1: Increased fructose transport facilitates killing in glucose-limited conditions in vitro.

A) Proteomic analysis comparing GT5-expressing and EV-expressing OT-I cells by mass spectrometry (n=3). The horizontal axis indicates log₂ of fold-change of protein abundance (GT5 relative to EV).

B) LC-MS derived measurement of intracellular ¹³C lactate following a 4h incubation with [U-¹³C] 10 mM glucose or fructose. Left: schematic of metabolic labeling of G6P, F6P, pyruvate and lactate starting from [U-¹³C] glucose and fructose. Letter “M” indicates the number of carbon atoms labeled with ¹³C. Right: mean ± SEM of total pools of intracellular lactate in EV or GT5 cells incubated in [U-¹³C] glucose or fructose (n=3, one-way ANOVA).

C) Levels of indicated metabolites in human serum and TME, as compiled from the HMDB³¹ and relevant scientific publications ^{30,32,41–45}.

D) Growth curve of WT mouse CD8⁺ cells cultured in 1 mM glucose plus 2.5 mM fructose media. Mean cell numbers ± SEM shown (n=6, two-tailed Student’s T test per timepoint).

E) Luminescence levels of B16-OVA cells incubated with EV or GT5 OT-I cells in indicated fructose concentrations for 24h; mean fold-change in luminescence relative to B16 grown without T cells ± SEM shown (n=6, two-tailed Student’s T test per fructose concentration point).

F) Annexin V and propidium iodide (PI) staining of B16-OVA cells incubated with EV or GT5 OT-I cells for 25h in 0.5 mM glucose plus 2.5 mM fructose, as measured via intracellular flow cytometry. Mean percentages ± SEM of CD45- Annexin V⁺ PI⁻ cells shown; EV: n=11 replicates, GT5: n=10 replicates; two-tailed Student’s T test.

G) Intracellular IFNγ of GT5 or EV OT-I cells after incubation with B16-OVA cells in 0.5 mM glucose plus 2.5 mM fructose, as measured via flow cytometry. Mean percentages ± SEM of IFNγ⁺ cells shown (n=4, two-tailed Student’s T test).

H) Same as in G) but for TNFα.

I) Cytotoxicity assay of control (EV) 19BBz CAR T cells or GT5 19BBz CAR T cells against SKOV3-CD19 cells, under 0.5 mM glucose plus 2.5 mM fructose, using differing effector to target (E:T) ratios (n=3, two tailed Student’s T test per E:T ratio).

J) Same as in I) but targeting Raji cells.

K) Same as in I) but targeting Nalm6 cells.

*, p < 0.05; **, p < 0.01; ***, p < 0.001.

See also Figures S1 and S2.

To bring further translational relevance to this study, we overexpressed GLUT5 in human CAR T cells. We designed a P2A-linked bicistronic vector termed GT5 19BBz that expresses GLUT5 and an anti-CD19 CAR (αCD19-scFv/4-1BB/CD3 ζ /P2A/GLUT5) (Figure S2A). We observed robust human GLUT5 expression in human GT5 19BBz CAR T cells (Figure S2B). To confirm that GT5 engineering does not alter the immunophenotype or impede the effector function of GT5 19BBz CAR T cells, we compared transduction efficiency, CD4⁺/CD8⁺ ratio, T cell differentiation marker, activation marker and exhaustion marker expression, as well as pro-inflammatory cytokine secretion of 19BBz and GT5 19BBz CAR T cells (Figure S2C–E). None of the investigated parameters showed significant differences between 19BBz CAR T cells and GT5 19BBz CAR T cells. When cultured in 1 mM glucose 2.5 mM fructose RPMI, both OKT3/IL-2 stimulated and non-stimulated GT5 19BBz CAR T cells grew significantly better than 19BBz CAR T cells (Figure S2F). We next assayed the cytotoxic activity of GT5 19BBz CAR T cells in same conditions as above. GT5 cells outperformed 19BBz EV cells in two liquid tumor cell lines, Nalm6 and Raji across most effector to target cell (E:T) ratios (Figure 1J–K). Interestingly, the more resistant ovarian cancer cell line transduced to express CD19 (SKOV3-CD19), was more preferentially killed by GT5 CAR T cells at higher (E:T) ratios (Figure 1I). Cytotoxic assays performed under standard high glucose conditions did not show a significant difference in killing between GT5 and EV T cells (Figure S2G). These findings are notable because solid tumors often have low intratumoral glucose. Engineering CAR T cells to use fructose as an alternative carbon fuel may therefore help cell-based immunotherapies overcome this metabolic barrier in the TME.

GT5 T cells are highly glycolytic in vitro

Since we observed that GT5 OT-I cells produced relatively more lactate than EV cells, we wanted to quantify lactate produced into media under high fructose, high glucose, or moderate fructose plus low glucose conditions. We incubated EV or GT5 cells in either [2-¹³C] glucose or fructose, or in the mixture of glucose and fructose, with only one being isotopically enriched (Figure 2A). Then, we detected [2-¹³C] lactate produced into media through nuclear magnetic resonance (NMR) spectroscopy. Interestingly, GT5 cells grown in high fructose only produced as much lactate as EV cells grown under high glucose conditions (Figure 2B). This signifies that GT5 cells can import and metabolize fructose efficiently through glycolysis, possibly as efficiently as either EV or GT5 cells grown under high glucose conditions. We also measured if GT5 cells can use fructose in the presence of low glucose. When incubated with 1 mM glucose plus 5 mM [2-¹³C] fructose, GT5 cells still produced more lactate than EV cells (Figure 2C). While GT5 cells produced less lactate in 5 mM fructose plus 1 mM glucose than in 10 mM fructose alone, they still produced more than EV cells, signifying that glycolytic flux from fructose is not blocked by the presence of glucose, but further enhanced (Figure 2B–C). T cell receptor and CD28 engagement result in PI3K activation and subsequent increases in GLUT1 at the cell surface, glucose import, and glycolysis^34–36. Since PI3K inhibition results in decreased glucose import and glycolysis³⁴, we wanted to determine if fructose-mediated glycolysis in GT5 T cells is independent of such inhibition. Indeed, co-incubation of EV and GT5 OT-I cells with [2-¹³C] glucose and PI3K inhibitor LY294002 resulted in decreased lactate production, while co-incubation of GT5 cells with [2-¹³C] fructose and PI3Ki resulted in a negligible blunting of lactate production (Figure S3A). To further investigate the relationship between PI3K-AKT signaling and GT5 metabolic engineering, we cultured EV and GT5 OT-I cells in either 10 mM glucose or fructose. Our results indicated that AKT phosphorylation, a process downstream of PI3K activation, was effectively inhibited by LY294002 (Figure S3B). Additionally, the ratio of phosphorylated AKT to total AKT remained consistent between GT5 and EV cells in both glucose and fructose conditions, even with PI3K inhibition (Figure S3B). These findings indicate that GLUT5-mediated fructose uptake in T cells is independent of PI3K signaling.

Figure 2: GT5 T cells are highly glycolytic in vitro.

A) Schematic for the experiments below.

B) NMR spectroscopy analysis of extracellular [2-¹³C] lactate, produced by OT-I cells following a 4h incubation with [2-¹³C] glucose or [2-¹³C] fructose alone. Peak areas were used for measuring concentrations. Mean lactate concentrations ± SEM shown (n=3, two-tailed Student’s T test).

C) NMR spectroscopy analysis of extracellular ¹³C lactate, produced by OT-I cells, following a 4h incubation with [2-¹³C] glucose plus unlabeled fructose or with [2-¹³C] fructose plus unlabeled glucose. Analysis as in B). (n=3, two-tailed Student’s T test).

D) Heatmap of LC-MS derived fractional enrichment of indicated intracellular metabolites in OT-I cells following a 4h incubation with [U-¹³C] glucose plus unlabeled fructose or with [U-¹³C] fructose plus unlabeled glucose; bold text indicates isotopic. Gradient of blue shading represents the magnitude of fractional enrichment.

E) Fractional enrichment of representative metabolites as in D). n=3, one-way ANOVA.

F) Same as in B) but in 19BBz CAR T cells. n=3, two-tailed Student’s T test.

G) Same as in C) but in 19BBz CAR T cells. n=3, two-tailed Student’s T test.

*, p < 0.05; **, p < 0.01; ***, p < 0.001.

See also Figure S3.

To fully appreciate the metabolic mechanism that allows these cells to proliferate and kill in low glucose conditions, we further profiled the metabolism of EV and GT5 OT-I cells in the 1 mM glucose plus 5 mM fructose condition, using uniformly labeled ¹³C sugars (Figures 2D, S3C). Similar to what we detected in media, GT5 cells produced more lactate than EV cells when fructose was labeled. EV cells produced more lactate in labeled glucose than GT5 cells, which differs from the media NMR measurements, though the sum of the derived sources yields equivalent intracellular lactate (Figure 2D–E). This is likely because the pool size of lactate in the cell turns over rapidly, thus the true flux is observed in the pool size exported. Curiously, central metabolites of glycolysis such as phosphoenolpyruvate (PEP), 3-phosphoglycerate (3PG) and pyruvate were higher across the board in GT5 labeled fructose cells than EV cells in labeled fructose or glucose (Figure 2D–E, S3D). The citric acid cycle (TCA) metabolites, such as citrate and malate, and pentose phosphate pathway (PPP) metabolites, such as ribulose-5P and sedoheptulose-7P, mirrored the pattern of lactate (Figure 2D–E, S3D). Fructose can feed directly into glycolysis downstream of glucose-6P (G6P), at fructose 6P, and would have to go in reverse to populate the PPP (Figure S3C). This may be why glycolytic flux downstream of G6P is elevated in GT5 cells in both labeled fructose and glucose, while PPP flux is not (Figure 2D–E, S3D). Altogether, this suggests that GT5 cells can efficiently restore glycolytic flux from fructose under low-glucose conditions, as well as recover their PPP and TCA. Further, this supports the notion that these intermediates are key to the functional output of the cell.

To determine if GT5 metabolic engineering produced the same effect in 19BBz CAR T cells, we first monitored extracellular [2-¹³C] lactate production from 10 mM [2-¹³C] glucose or 10 mM [2-¹³C] fructose via NMR analogous to our previous OT-I experiments. We observed robust lactate secretion into the media in GT5 cells grown in high fructose, comparable to that of cells grown in high glucose (Figure 2F). Next, we assessed how GT5 19BBz CAR T cells perform in low glucose media, supplemented with moderate fructose. GT5 cells grown in 5 mM [2-¹³C] fructose plus 1 mM non-isotopic glucose produced comparable [2-¹³C] lactate to GT5 cells grown in 5 mM non-isotopic fructose + 1 mM [2-¹³C] glucose (Figure 2G). We noted that GT5 cells produced lactate from fructose even in the presence of glucose, indicating that they can use this non-canonical fuel even when presented with glucose (Figure 2G). We also profiled donor CD8⁺ T cells incubated with uniformly labeled ¹³C glucose or fructose. Lactate production and the levels of key metabolites from glycolysis, TCA, and PPP were significantly lower when cells were incubated with fructose compared to glucose (Figure S3E), suggesting that primary donor CD8⁺ T cells cannot efficiently metabolize fructose without GT5 engineering. This finding supports the advantage of GT5 cells in utilizing fructose as an additional carbon source to drive metabolism in glucose deficient TME, thus enhancing their functionality and persistence in such challenging metabolic environments.

GT5 macrophages are more efficient at tumor cell elimination in vitro and in vivo

Recently, tumor-associated macrophages (TAMs) have become an attractive cancer immunotherapy target, for instance targeting the “don’t eat me” signal CD47 to induce phagocytic clearance of cancer cells³⁷. Since TAMs also experience glucose scarcity, we posited that GT5-engineered macrophages might use fructose to support phagocytic clearance of cancer cells. To test this, we engineered primary macrophages with EV or GT5 (Figure S4A) then co-cultured them with breast cancer cells (MDA-MB-231) and an antibody targeting CD47 to test CD47 antibody-mediated phagocytosis (Figures S4B). Both EV and GT5 macrophages performed antibody-mediated phagocytosis of cancer cells at equivalent levels when cultured in glucose-containing media (Figure S4C–D). Surprisingly, although EV-expressing macrophages were unable to survive in fructose-containing media, GT5-expressing macrophages cultured in the presence of fructose performed significantly more antibody-mediated phagocytosis of cancer cells than both EV and GT5 macrophages on glucose (Figure S4C–D).

Given our findings in vitro, we next tested if GT5-expressing macrophages can better control cancer growth in vivo. To this end, we used two mouse models of orthotopic breast cancer: the checkpoint blockade-sensitive E0771 model (Figure S4E) and the checkpoint blockade-insensitive 4T1 model (Figure S4G). In conjunction with our in vitro results, GT5 engineering in macrophages led to significant control of tumor progression in E0771 tumors treated with anti-CD47 antibody compared to EV macrophages incapable of controlling E0771 growth (Figure S4F). Strikingly, GT5 macrophages were also capable of controlling tumor progression in 60% of mice bearing 4T1 tumors treated with anti-CD47 antibody whereas EV macrophages were incapable of controlling 4T1 growth (Figure S4H–I). Collectively, these findings suggest that immune cell anti-cancer activity may benefit more broadly from GT5 engineering.

GT5 T cells and GT5 CAR T cells have better capacity to kill in vivo

Given that GT5 engineering recovers key pathways in vitro, we aimed to assess the in vivo killing efficiency. We established syngeneic B16-OVA melanoma engraftments in Rag1^−/− mice (Figure 3A) and on day 5 post-engraftment mice were treated with either no T cells, control EV OT-I or GT5 OT-I T cells and assessed by magnetic resonance imaging (MRI) (Figure 3A–B). Robust infiltration of GT5 OT-I cells into B16 tumors (Figure S5A) and significant reduction in tumor growth were observed in GT5-T-cell-treated mice (Figure 3B–D). To specifically measure the metabolic pools of GT5 T cells, we engrafted B16-OVA melanomas in Rag1^−/− mice and pulled down CD8⁺ T cells at day 10 for metabolic analysis (Figure 3E). Just as we found in vitro, T cells exhibited significantly increased pool sizes of key metabolites in glycolysis and TCA, verifying the metabolic engineering mediated fructose utilization (Figures 3F–H, S5B). We next sought to elevate fructose in the TME to try to stimulate GT5 OT-I activity even further (Figure S5C). In vehicle injections we again observed enhanced tumor killing by GT5 OT-I cells, and pharmacological delivery of fructose through intraperitoneal (I.P.) injections further boosted their anti-tumor activity (Figures S5D–G). We noted that we were able to increase the fructose levels in the TME about 7-fold (Figure S5G). Next, we explored the combination of GT5 T cells with immune checkpoint inhibition (ICI) (Figure 4A). As ICI, we used a co-treatment of αPD-1 and αCTLA-4 antibodies, which is currently the standard of care for melanoma (Figure 4A). The combination of GT5 and ICI resulted in both superior tumor control and significantly increased overall survival (Figure 4B–D). To further show the physiologic relevance of GT5 engineering, we adoptively transferred EV and GT5 OT-I cells into mice engrafted with B16-OVA melanoma tumors placed on a high fructose chow diet (Figure S5I). Again, we observed smaller tumors in GT5 OT-I treated mice (Figure S5J), which was expected since mice on a high fructose chow diet exhibited an approximately 5-fold increase in fructose levels in the TME compared to those on a normal chow diet (Figure S5K). The increased fructose availability in the TME can recruit more GT5 T cells, thereby enhancing tumor control. Moreover, expression of key exhaustion markers PD-1 and TIM-3 was also reduced in GT5 OT-I cells assayed from these tumors (Figure S5L). According to these findings, GT5 OT-I adoptive transfer is synergistic with increased TME fructose but also with other immunotherapies like ICI. Finally, we assessed the anti-tumor efficacy of human GT5 19BBz CAR T cells in vivo using an antigen engineered human ovarian cancer xenograft model (SKOV3-CD19). 12-days post tumor engraftment, 19BBz or GT5 19BBz CAR T were engrafted and mice received supplementary I.P. injections of either fructose or vehicle 5 times per week (Figure 4E). With or without additional fructose injections, GT5 19BBz CAR T cells significantly outperformed 19BBz CAR T cells (Figure 4F–G). Collectively, our findings suggest that GT5 engineering improves the anti-tumor efficacy of adoptive T cell therapies in vivo.

Figure 3: GT5 T cells can recapitulate fructolytic metabolism in vivo.

A) Schematic of B).

B) Adoptive transfer experiment of EV or GT5 OT-I cells into B16-OVA bearing Rag1^−/− mice. Around day 5 when tumors were palpable, transduced OT-I cells were engrafted. Tumor volumes were measured twice a week until endpoint; representative MRI images of tumor-bearing mice shown.

C) Same as in B) but showing individual tumor volumes for indicated groups (vehicle: n=7, EV: n=7, GT5: n=6, two-way ANOVA).

D) Same as in B) but showing mean tumor weights ± SEM (Vehicle: n=7, EV: n=7, GT5: n=6, one-way ANOVA).

E) Schematic of the experiments below.

F) Magnetic pull down of CD8⁺ OT-I cells from B16-OVA tumors treated as in B) and profiling of lactate via LC-MS; mean ± SEM of total pools of lactate shown (n=4, two-tailed Student’s T test).

G) Same as in F) but for glutamate.

H) Same as in F) but for aspartate.

*, p < 0.05; **, p < 0.01; ***, p < 0.001.

See also Figures S4 and S5.

Figure 4: GT5 T cells and GT5 CAR T cells have better capacity to kill in vivo.

A) Schematic of B).

B) Adoptive transfer experiment as in Figure 3B but into immunocompetent mice treated with αPD-1 and αCTLA-4 blocking therapy; mean tumor volumes ± SEM shown (n=5, two-way ANOVA).

C) Same as in B) but showing individual tumor volumes for indicated groups, two-way ANOVA.

D) Same as in B) but showing survival plot of indicated groups, log-rank test.

E) Schematic of the experiment in F).

F) Adoptive transfer experiment of either 19BBz CAR T or GT5 19BBz cells in SKOV3-CD19 bearing NSG mice treated either with I.P. vehicle (PBS) 5 times per week; individual tumor volumes of indicated groups shown (n=5, two-way ANOVA).

G) Same as in F) but with I.P. 4g/kg fructose 5 times per week; individual tumor volumes of indicated groups shown (n=5, two-way ANOVA).

*, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001.

See also Figures S4 and S5.

Discussion

We engineered three different immunotherapy approaches, OT-I cells, CAR T cells and macrophages to use a privileged fuel source – fructose. This type of metabolic engineering can overcome barriers in treating solid tumors with cell-based therapies, as solid tumors tend to be scarce in nutrients like glucose and amino acids. Our approach supports a previously undescribed paradigm in cell based metabolic engineering, demonstrating how transport dynamics can efficiently divert carbons into key pathways and enhance cell effector function. Drawing inspiration from cancer cells’ metabolic adaptations, introducing rapid fructose-specific transport, without increased KHK activity, a metabolic state that shunts carbons directly into glycolysis is created in the cell ^25,27. While reports have suggested the K_m of hexokinase for fructose is high (1.9mM³⁸), this clearly does not rate-limit the flux in an activated immune cell. Notably, GT5 immune cells maintain robust antitumor activity across multiple tumor models even when circulating fructose is not exogenously elevated, indicating sufficient endogenous fructose levels in the TME. Since most U.S. residents already consume elevated fructose, due to high fructose corn syrup in western diets, they may not need pharmacological intervention to raise fructose. Some studies have reported that cancer patients have elevated fructose in their blood^32,39, further supporting the clinical implications of the GT5 engineering approach.

Remarkably, we found that checkpoint inhibition synergizes better with GT5 T cell adoptive transfer than with pharmacological fructose delivery, likely because fructose is already present in the TME at levels adequate to drive GLUT5 mediated flux into glycolysis. Moreover, PD-1 and CTLA-4 engagements dampen glucose import and utilization^34,40, so checkpoint inhibition releases the breaks on glucose import and glycolysis, thereby allowing increased fructose uptake through GLUT5 to further boost glycolytic flux.

Since solid tumors generally have nutrient and oxygen depleted microenvironments, our metabolic engineering system is easily translatable to other cell-based therapies, or other fuel sources. Other cell-based therapies undercut by low glucose concentrations might also gain a metabolic turbo boost from GLUT5 through fructose. Fructose may not be the only extracellular metabolite that can be non-canonically utilized; if a certain cancer cannot metabolize a particular metabolite, immune cells can be reprogrammed to use that metabolite for an increased anti-tumor effect.

Limitations of the Study

Further research is needed to fully characterize the mechanism of action in T cells and macrophages to better understand what governs maximal response. Additionally, future work is needed to validate the in vivo findings in humans to assess efficacy and safety. Further large population studies are also needed to determine to what degree fructose is elevated in the TME and the blood of cancer patients across various diets.

Resource Availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be made available by the lead contact, Kayvan R. Keshari (rahimikk@mskcc.org).

Materials availability

All unique/stable reagents generated in this study are available from the lead contact with a completed materials transfer agreement.

STAR Methods

Experimental model and study participant details

Mice

OT-I mice were a gift from Dr. Andrea Schietinger. 8 to 12 weeks old male and female OT-I mice were sacrificed with spleens collected for all OT-I experiments. Glut5 KO mice were made in house. 8 to 10 weeks old female Glut5 KO mice were used to generate the anti-mouse GLUT5 antibody. C57BL/6J, CD45.1 (B6.SJL-Ptprc^a Pepc^b/BoyJ), Rag1^−/− (B6.129S7-Rag1^tm1Mom/J), BALB/cJ, and NSG (NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ) mice were purchased from The Jackson Laboratory. 8 weeks old female C57BL/6J or BALB/cJ mice were used in EO771 and 4T1 in vivo experiments. 8 to 10 weeks old female C57BL/6J mice were used to measure intratumoral fructose concentration. 8 to 10 weeks old female Rag1^−/− or CD45.1 mice were used in OT-I engraftment experiments. 6 to 8 weeks old female NSG mice were used in CAR T cell in vivo experiments. Mice were bred and maintained at Memorial Sloan Kettering Cancer Center, and all experiments were conducted in agreement with MSKCC Institutional Animal Care and Use Committee (IACUC) guidelines.

Cell Lines

B16-OVA cells were purchased from MilliporeSigma (SCC420). Raji, NALM6, 4T1, E0771 and MDA-MB-231 cells were purchased from ATCC. SKVO3 (CD19⁺), H29 and 293Galv9 cells were a gift from Dr. Renier Brentjens, MSKCC. HEK293T cells were purchased from ATCC (CRL-3216 ). B16 OVA luciferase cells were produced in house. PlatE packaging cells were a gift from Dr. Andrea Schietinger, MSKCC. NIH3T3 cells were provided by the MSKCC antibody core. Sp2/0 hybridoma cells were purchased from ATCC. Cell lines were periodically tested for mycoplasma and grown under standard culture conditions at 37°C and 5% CO₂.

Human peripheral blood mononuclear cells (PBMCs)

Blood from healthy human donors was taken after written informed consent on MSK institutional review board-approved protocols and in accordance with IRB 00009377. Human PBMCs were isolated from whole blood by centrifugation in lymphocyte separation medium and red blood cells were lysed using ACK lysis buffer. Human PBMCs were grown under standard culture conditions at 37°C and 5% CO₂.

Method details

Cell culture

All media was obtained from MSKCC media core. Human cancer cell lines and human T cells were maintained in RPMI-1640 10% fetal bovine serum (FBS), 2 mM L-glutamine, 100 IU/mL penicillin and 100 μg/mL streptomycin (Penstrep). Mouse cancer cell lines, HEK293T and NIH 3T3 cells were grown in standard high glucose DME supplemented with 10% FBS and Penstrep, same as above. Where glucose and fructose concentrations were modified, glucose free DME or RPMI were used and was supplemented with dialyzed FBS (HyClone # SH3007903). Non-isotopic glucose and fructose were purchased from Sigma. Human T cells were maintained in RPMI-1640 medium supplemented with FBS, glutamine and Penstrep, same as above, and additionally, 200 IU/mL IL-2 (MSKCC pharmacy) during spinfection. All retroviral producer cell lines were maintained in DMEM medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 IU/ml penicillin and 100 ug/ml streptomycin. All mouse T cells and mouse splenocytes were cultured in RPMI with 10% heat inactivated fetal bovine serum, 2 mM L-glutamine, 100 IU/ml penicillin and 100 ug/ml streptomycin, 10 mM HEPES, NEAAs, 1 mM sodium pyruvate, 50 IU/ml mouse IL-2 (BioLegend # 575404) and 50 μM 2-Mercaptoethanol (BME) (Thermo Fischer # 21985023).

Luciferase cell lines

Raji, Nalm6 and SKVO3 cell lines were transduced with eGFP and firefly luciferase using 293Galv9 cells encoding the GFP/Luc gene (generously provided by the R Brentjens Lab, MSKCC) as previously described⁴⁶.

GT5 cloning

Untagged mouse Glut5 cDNA was purchased from Origene (MC204180) was PCR cloned into pMIG II (PMSCV-IRES-GFP II) or pMSCV-IRES-mCherry vectors purchased from Addgene (52107 and 52114) using Q5 Hot Start High-Fidelity DNA Polymerase (NEB). Restriction sites used were EcoRI and XhoI. To engineer the human GT5 19BBz CAR T cells, the human GLUT5 gene was cloned into an SFG gamma retroviral vector encoding a CD19-targeted second-generation CAR with 4-1BB costimulatory domain (generously provided by R Brentjens lab, MSKCC)⁴⁶. Final constructs were purchased from Genscript.

Generation of stable overexpressing lines

NIH3T3 and HEK293T stable cell lines for monoclonal antibody generation and testing were made through lentiviral infection, as previously described⁴⁷. Mouse and human Lenti GLUT5 flag/myc or GFP C-terminally tagged constructs were purchased from Origene. Lentivirus was produced by co-transfection of HEK293T cells with plasmids encoding psPAX2 (Addgene plasmid 12260), and pMD2.G (Addgene plasmid 12259) using X-tremeGENE HP (Roche) in accordance with the manufacturer’s protocol. For transduction, polybrene was used at the concentration of 8 μg/m and obtained from SCBT. For selection, puromycin was used at the concentration of 2 μg/mL and obtained from InvivoGen.

Glut5 transduction of mouse T cells

Splenocytes were isolated from mechanically dissociated mouse spleens. After red blood cells were subjected to lysis (RBCL), CD8⁺ cells were isolated from the splenocyte suspension using EasySep^™ Mouse CD8⁺ T Cell Isolation Kit (STEMCELL) and activated for 24 h using Dynabeads Mouse T-Activator CD3/CD28 for T Cell Expansion and Activation (Thermo Fischer). In other cases, whole splenocytes post RBCL were placed on 1 μg/ml OVA peptide (InvivoGen) to specifically activate OT-I for 24h. In both cases, mouse T cells were cultured in RPMI in the presence of 50 IU/ml of mouse IL-2 (BioLegend # 575404) and 50 μM BME (Thermo Fischer # 21985023). Retrovirus for pMIGII empty vector (EV) or Glut5 (GT5) (Addgene) was produced in PlatE cells, using TransIT-LT1 Transfection Reagent (Mirus Bio) according to manufacturer’s instructions. Spinfections were performed for two days in a row, 24h after the initial activation using RetroNectin (Takara Bio) as previously described⁴⁸.

GT5 19BBz retrovirus production

Retroviral produces cell lines were generated as previously described⁴⁶. Briefly, to engineer retroviral producer cells lines, H29 cells were transiently transfected with the GT5 19BBz SFG plasmid using Lipofectamine 3000. Stable GT5 19BBz retroviral producer cell lines were engineered by transfecting 293Galv9 cells with the supernatant of H29 cells.

CAR T cell production

Human CAR T cells were produced as previously described⁴⁶. Briefly, to engineer CAR T cells, human peripheral blood mononuclear cells (PBMCs) were isolated from healthy donors in compliance with all relevant ethical regulation and in accordance with IRB 00009377. PBMCs were separated from whole blood by centrifugation in lymphocyte separation medium and red blood cells were lysed using ACK lysis buffer. PBMCs were stimulated for 2 days with 200IU/ml IL-2 and 50ng/mL anti-CD3 (OKT3) in RPMI1640 containing 10% fetal bovine serum, 2mM L-glutamine, 100IU/ml penicillin and 100ug/ml streptomycin. The medium of stable producer cell lines was changed from supplemented DMEM to supplemented RPMI a day before transduction. Stimulated T cells were transduced twice with GT5 19BBz by spinoculation on Retronectin (Takara Bio) coated plates with supernatant of stable retroviral producer cells. CAR T cells were cultured for 3 days in supplemented RPMI as described above and 200 IU/ml IL-2 was supplemented every other day. Transduction efficiency was determined 3 days post transduction using flow cytometry and CAR T cells were used immediately for subsequent analyses unless otherwise indicated.

CAR T expansion assays

At day 0, 1M 19BBz and GT5 19BBz CAR T cells were seeded in triplicates in 24 well plates at a concentration of 1M/ml in either complete RPMI-1640 or ‘high fructose-low glucose’ RPMI with 2.5mM fructose (Sigma Aldrich #F0127) and 1mM glucose (Sigma-Aldrich G8270). Complete 2.5 mM fructose 1mM glucose RPMI-1640 was made as described above. A second identical plate was stimulated by spiking in CD3/CD28 antibody scaffolds (ImmunoCult, STEMCELL technologies # 10991) according to the manufacturer’s instructions. At day 1-3, cell counts were measured via Trypan Blue exclusion (Trypan Blue Solution 0.4 %, Invitrogen # T10282) using a Countess 3FL automated cell counter (Invitrogen # A49866).

CAR T killing assays

20 M SKOV3 cells (engineered to be CD19⁺) were seeded in each well of a 96 well-plate on day −1 to allow adherence, half of a plate in high glucose RPMI, the other in 2.5 mM fructose + 0.5 mM glucose RPMI. Suspension cell lines (Raji GFP fLuc and Nalm6 GFP fLuc) were seeded similarly in at d0 at a concentration of 50 M per well. 19BBz and GT5 19BBz CAR T cells were engineered as previously described and seeded in log₂ Effector to Target cell ratios (E:T) from 4:1 to 0.0625:1 and 0:1. The Raji GFP fLuc and Nalm6 GFP fLuc plates were incubated for 24 h, the SKOV3 plates were incubated for 48h at 37C with 5% CO₂. After incubation, the plates with Raji GFP fLuc and Nalm6 GFP fLuc were incubated with 0.5 mg/ml luciferin for 10 min at RT before analysis using a luminescence plate reader. At day 2, the SKOV plates were washed twice carefully with PBS to remove residual T cells, and the viability of the target cells was determined using the Promega Cell titer glow luminescent viability kit, according to manufacturer’s recommendations before measuring luminescence with a plate reader.

Western blotting

Cells were lysed with RIPA buffer (Thermo Fisher Cat #89900) with phosphatase/protease cocktail (Thermo Fisher Cat # 78440). Lysates were passed through a syringe to maximize lysis. Lysates were not boiled nor thawed/re-refrozen more than once. For GLUT5 immunoblotting, lysates were quantified using BCA protein assay (Thermo Fisher # 23227) and resolved on 4-12% Bis-Tris gels (Thermo Fisher # NP0322BOX) under denaturing conditions. Proteins were blotted onto a nitrocellulose membrane (Thermo Fisher # LC2009) and probed with anti-GLUT5 (SCBT # sc-271055), anti-Actin (CST # 8457S) and anti-Vinculin (Sigma-Aldrich # V9264) antibodies. SCBT GLUT5 antibody was used at a dilution of 1:50. Membranes were visualized using anti-mouse or anti-rabbit HRP-conjugated secondary antibodies (Cytiva # NA931 and NA934), ECL reagent (Cytiva # RPN2209) and films (Fisher Scientific # NC9556985). For other proteins, lysates were quantified using DC protein assay according to manufacturer’s protocol and resolved on 4-12% Bis-Tris gels (Thermo Fisher # NP0321BOX) under denaturing conditions. Proteins were blotted onto a PVDF membrane (Thermo Fisher # 88520) and probed with anti-phospho-Akt (CST # 4060), anti-Akt (CST # 4685), and anti-Vinculin (Sigma-Aldrich # V9264) antibodies. Anti-phospho-Akt antibody was used at a dilution of 1:2000, anti-Akt 1:1000, and anti-Vinculin 1:500. Membranes were visualized using anti-mouse or anti-rabbit HRP-conjugated secondary antibodies same as above. Signals were visualized using Immobilon Forte Western HRP substrate (MilliporeSigma # WBLUF0500). Densitometry analysis was performed with ImageJ.

B16-OVA tumor model

Approximately 1 x 10⁵ B16-OVA cells in Matrigel were injected into the flanks of Rag1^−/− (KO) or CD45.1 mice. 5-7 days after B16-OVA implantation when tumors were palpable, 2 million of transduced EV or GT5 CD45.2 OT-I cells were injected into mice through the tail vein. One day prior to adoptive transfer, CD45.1 mice were treated I.P. with cyclophosphamide at 60 mg/kg. In some instances, mice were immunized with 10 μg of OVA peptide (InvivoGen) with Poly I:C (InvivoGen) after adoptive transfer. For checkpoint inhibitor studies, mice were injected I.P. with either isotype controls or 250 μg anti-PD-1 inhibitor and 100 μg anti-CTLA-4 inhibitor antibodies (Bio X Cell). Injections started one day post-OT-I cell transfer (day 6) and repeated on days 9 and 12. For injected fructose studies, I.P. 4 g/kg fructose or vehicle injections began 1 day prior to OT-I transfer and continued 5 times a week throughout the duration of the study. For fructose diet studies, mice were placed on high fructose chow (Inotiv TD.89247) two weeks prior to B16-OVA implantation. The mice were kept on the chow throughout the entire study. Tumor sizes were measured with calipers or with Bruker 3 T magnetic resonance imaging (MRI).

B16-OVA co-culture with OT-I cells

B16 OVA-luciferase cells were plated in standard media 24h prior to addition of EV or GT OT-I cells at 2:1 effector to target ratio in glucose free media, supplemented with different concentrations of fructose ranging from 0 - 5 mM. After 24h in culture, media and T cells were carefully removed and B16 luciferase activity was measured using Steady-Glo Luciferase Assay System (Promega), according to the manufacturer’s instructions. Apoptosis and cytokine measurements: B16-OVA cells were allowed to attach overnight under standard culture conditions before addition of OT-I cells. EV or GT5 OT-I cells were added at 2:1 effector to target ratio in 0.5 mM glucose plus 2.5 mM fructose RPMI supplemented with dialyzed serum. After 24 h of co-culture, OT-I cells were collected for cytokine analysis via flow cytometry, and B16 cells were collected for Annexin V staining using Dead Cell Apoptosis Kit with Annexin V (Thermo Scientific). Prior to staining for flow cytometry, OT-I cells were restimulated using Cell Activation Cocktail (BioLegend) for 2h.

Flow cytometry staining and antibodies

To quantitate EV GFP or GT5 GFP expression, OT-I cells were stained with DAPI (Thermo Fischer # 62248) and subsequently the GFP signal was detected on flow cytometer after gating on DAPI negative cells. To evaluate GT5 expression, OT-I cells were incubated in spent media from the hybridoma clone 8f7, followed by staining in Alexa Fluor 647 secondary antibody (Jackson Immunoresearch 115-605-071). For the in vitro assays, prior to performing the Annexin V protocol, B16 cells were stained extracellularly with Pacific Blue anti-mouse CD45 (BioLegend # 103126) to exclude any T cells. OT-I cells from the same assay were fixed in eBioscience Intracellular Fixation & Permeabilization Buffer Set (Thermo Fischer # 88-8824-00) according to their protocol. Prior to fixation, OT-I cells were stained with Zombie Violet Fixable Viability Dye (BioLegend # 423113). OT-I cells were then stained with PE-Cy7 anti-mouse TNFα (BioLegend # 506324) and Brilliant Violent 785 anti-mouse IFNγ (BioLegend # 505837) antibodies. To perform staining on mouse tumors, tumors were digested with DNase I (Roche # 10104159001) and Liberase TL (Roche # 05401020001). Tumor cell suspensions were subjected to viability dye staining (same as above) and Fc block (BD # 553142). OT-I cells were CD45.2 and the tumor hosts were CD45.1: to analyze the CD45.2 OT-I cells, tumor suspensions were stained with FITC CD45.2 (BioLegend # 109806), BV650 CD8 (BioLegend # 100741), PE TIM-3 (BioLegend # 119706) and PerCP-Cy5.5 PD-1 (BioLegend # 109120). 19BBz CAR T cells or GT5 19BBz CAR T cells were immunophenotyped by staining for CAR expression with 19E3-2D5-3F10 A647 conjugate or PE conjugate (anti-idiotype for 19BBz, generously provided by the Sadelain lab, MSKCC), BV785 anti-human CD3 (BioLegend # 317330), FITC anti-human CD4 (BioLegend # 317407), APC-Cy7 anti-human CD8 (BioLegend # 344714), and Viakrome 808 Fixable Viability dye (Beckman Coulter #C36628), by staining for T cell activation markers with BUV395 anti-human CD25 (BD Biosciences #563551), PE-eflour610 anti-human CD69 (ThermoFisher Scientific #61-0699-42), PerCP-Cy5.5 anti human CD71 (BioLegend # 334114) and APC anti-human CD95 (BioLegend # 305612), by staining for T cell exhaustion markers with BV510 anti-human TIM-3 (BioLegend #345030), BV785 anti-human CTLA-4 (BioLegend # 369624), PE-Cy7 anti-human LAG3 (BioLegend # 369208), APC anti-human PD-1 (BioLegend # 621610), and by staining for T cell differentiation markers with BV510 anti-human CD45RO (BioLegend #304246), BV785 anti-human CD45RA (BioLegend #304140), PE-Cy7 anti-human CD62L (BioLegend #304822) and APC anti-human CCR7 (BioLegend # 353214). 19BBz CARs and GT5 19BBz CARs of the same experiment were permeabilized using the BD Cytofix/Cytoperm Fixation/Permeabilization Kit (BD Biosciences # 554714) according to the manufacturer’s instructions. Fixed and permeabilized CAR T cells were stained with BV412 anti-human IFNγ (BioLegend # 506538), PE-Cy7 anti-human TNFα (BioLegend # 502930) and APC anti-human IL-2 (BioLegend # 500310). Flow cytometry was performed using BD Fortessa and Beckman Coulter Cytoflex LX and data was analyzed with FlowJo 10.10.0 (Tree Star).

Monoclonal GT5 antibody production

Glut5 knockout mice, were immunized subcutaneously with 10 million GLUT5-GFP NIH3T3 stable expressing cells in Titermax, TiterMax USA (Norcross, GA), and 10 million GLUT5-GFP NIH3T3 in PBS on opposing flanks. The mice were immunized twice more subcutaneously with 10 million GLUT5-flag/myc expressing NIH3T3 cells per immunization, at three weeks and 8 weeks after the initial immunization. Sera of the mice were assayed for immunogenicity via dot blot and flow cytometry against HEK293T cells, transiently expressing mouse GLUT5 (HEK293T cells were kept from virus generation and used for testing). A final intraperitoneal immunization of 5 million GLUT5 expressing NIH3T3 cells was performed 19 weeks after the initial immunization. Four days post the final immunization splenocytes were isolated manually with frosted microscope slides (Fisher Scientific # 12-550-343) and then fused with Sp2/0 hybridoma cells (ATCC) by electrofusion. The electrofusion product was either frozen or immediately cultured in a semi-solid medium containing fluorescent anti-mouse IgG antibody (Jackson ImmunoResearch # 155-545-071). Single colonies were picked and placed into culture with the aid of a robotic instrument (ClonePix 2, Molecular Devices, San Diego, CA). Supernatant from the monoclonal cultures were assessed by flow cytometry and western blot for mouse GLUT5 specificity.

NMR quantification of extracellular lactate

1 x 10⁶/mL OT-I or CAR T cells were plated in either 10 μM [2-¹³C] fructose or 10 μM [2-¹³C] glucose. In the combination experiment, OT-I cells were plated in either 1 μM [2-¹³C] glucose plus unlabeled fructose or 1 μM unlabeled glucose plus 5 μM [2-¹³C] fructose. After 4h incubation, media was collected from cells and 10 mM ¹³C Urea standard was added to each media sample. The PI3K inhibitor experiment was conducted in a similar fashion but with a 4h co-incubation of isotope media and PI3K inhibitor LY294002. ¹³C-NMR spectra were acquired in a 600MHz NMR system (AVANCE III, Bruker). The acquired spectra were analyzed with MestReNova software. All isotopes were acquired from Sigma.

Measurement of metabolites with LC-MS

At least 4 million of OT-I cells were plated in either 10 μM [U-¹³C] fructose or 10 μM [U-¹³C] glucose. In the combination experiment, OT-I cells were plated in either 1 μM [U-¹³C] glucose plus unlabeled fructose or 1 μM unlabeled glucose plus 5 μM [U-¹³C] fructose. After 4h incubation, cells were pelleted and washed with PBS. Cell pellets were lysed in 80% methanol. For blood serum and tumor experiments, mice were injected with either PBS vehicle or 4 g/kg [U-¹³C] fructose. After 1h post injection, mice were euthanized and sera and tumors were collected. The samples were snap frozen for storage and thawed prior to methanol extraction. For isolation of OT-I cells from tumors for metabolomics, mice were injected with 4 g/kg [U-¹³C] fructose. 1h post injection, mice were euthanized and tumors were collected for dissociation. Tumors were dissociated using the GentleMACS system and the Mouse Tumor Dissociation kit (Milnenyi), according to the manufacturer’s instructions. CD8⁺ T cells were isolated from tumors using the mouse CD8 TIL Microbeads kit (Miltenyi), according to the manufacturer’s instructions. Isolated CD8⁺ cells were counted, pelleted and extracted with 80% methanol. All cell and tissue extracts were dried down and then re-dissolved in water. Targeted LC/MS analyses were performed on a Q Exactive Orbitrap mass spectrometer (Thermo Scientific) coupled to a Vanquish UPLC system (Thermo Scientific). The Q Exactive operated in polarity-switching mode. A Sequant ZIC-pHILIC column (2.1 mm i.d. × 150 mm, particle size of 5 μm, Millipore Sigma) was used for separation of metabolites. A 2.1 × 20 mm guard column with the same packing material was used for protection of the analytical column. Flow rate was set at 150 μL/min. Buffers consisted of 100% acetonitrile for mobile phase A, and 0.1% NH4OH/20 mM CH3COONH4 in water for mobile phase B. The chromatographic gradient ran from 85% to 30% A in 20 min followed by a wash with 30% A and re-equilibration at 85% A. The raw data was processed using El-MAVEN (v0.12.0). Metabolites and their 13C isotopologues were identified on the basis of exact mass within 5 ppm and standard retention times. Absolute quantification of glucose and fructose was based on standard curves.

Mass spectrometry and analysis

The cell pellet was lysed in urea buffer containing 8 M urea, 200 mM EPPS (pH 8.5), phosphatase inhibitor cocktails 2 and 3 (Sigma) and cOmpleteTM Mini EDTA-free protease inhibitor (Roche). The lysate was sonicated in a water bath sonication for 5 minutes, followed by addition of benzonase to a final concentration of 50 units/ml. The samples were centrifuged at 14,000 x g for 10 minutes at 11°C, and the supernatant was transferred to a new Eppendorf tube. Protein concentration was measured using a PierceTM bicinchoninic acid (BCA) assay. Proteins were reduced with tris (2-carboxyethyl) phosphine (TCEP) to a final concentration of 5 mM for 15 minutes at room temperature (RT), then alkylated with iodoacetamide (IAA) to a final concentration of 10 mM for 30 minutes at RT in the dark. The reaction was quenched with dithiothreitol (DTT) to a final concentration of 1 mM for 15 minutes at RT. Following chloroform/methanol precipitation, the pellet was resuspended in 50 ul of 200 mM EPPS (pH 8.5) and sonicated in water bath for 5 minutes. Lys-C (Wako) and trypsin (Promega) were added at an enzyme-to-protein ratio of 1:100, and the samples were digested overnight at 37°C with shaking at 1000 rpm. The next morning, anhydrous acetonitrile (ACN) and TMTpro 16plex (Thermo Fisher Scientific) were added to each sample, and the peptides were labeled according to the manufacturer’s instructions. A label check was performed, and the reaction was quenched with 50% hydroxylamine to a final concentration of 0.3% for 15 minutes at RT. The labeled samples were combined at a 1:1 ratio across samples based on the values from the label check, dried down in a SpeedVac, and reconstituted in 1 mL of 1% trifluoroacetic acid (TFA). Peptides were desalted using Sep-Pak cartridges (Waters), which were conditioned with 100% methanol (MeOH), followed by 70% ACN/1%TFA, and 5% ACN/1% TFA twice. After sample loading, the cartridges were washed twice with 5% ACN/1% formic acid (FA), and the peptides were eluted twice with 70% ACN/1% FA and vacuum centrifuged to dryness. The sample was reconstituted in 300 ul 0.1% TFA and fractionated using the Pierce High pH Reversed-Phase Peptide Fractionation Kit (Thermo Fisher Scientific) following the manufacturer’s instructions. The twelve fractions were dried using a SpeedVac and reconstituted in 15 ul 0.1% FA.

Mass spectrometry data analysis

The peptides were loaded onto a 50 cm column (Thermo Fisher ES903) and separated by reversed-phase chromatography using a gradient from 1% to 30% Buffer B over 220 minutes, followed by an increase to 90% Buffer B over 20 minutes. Buffer A consisted of 0.1% FA in HPLC grade water; Buffer B consisted of 90% ACN and 0.1% FA, at a flow rate of 300 nl/min using an EASY-nLC 1200 liquid chromatography system (Thermo Fisher Scientific). Mass spectrometry data were acquired on a Fusion Lumos mass spectrometer (Thermo Fisher Scientific) using an SPS MS3 method. The data were acquired in a data-dependent acquisition mode. Full MS spectra were acquired in the Orbitrap over a range of 375-1500 m/z at a resolution of 120K, with an AGC target of 400.000 and maximum injection time of 50 msec. MS2 was performed using CID using a collision energy of 32%, a maximum injection time of 50 msec, an isolation window of 0.7 m/z, and a AGC target of 10.000. Following the acquisition of each MS2 spectrum, a synchronous precursor selection (SPS) MS3 scan was collected for the top 10 most intense ions in the MS2 spectrum. The MS3 spectra were acquired in the Orbitrap with a maximum injection time of 150 msec, a scan range of 100-1000 m/z, and a resolution of 50K.

TMT data analysis

Raw data files were processed using Proteome Discoverer (PD) version 2.4.1.15 (Thermo Scientific). For each of the TMT experiments, raw files from all fractions were merged and searched with the SEQUEST HT search engine with a Mouse UniProt protein database downloaded on 2021/12/13 (92,249 entries). Methionine oxidation was set as variable modification, while cysteine carbamidomethylation, TMTpro (K), and TMTpro (N-term) were specified as fixed modifications. The precursor and fragment mass tolerances were set to 10 ppm and 0.6 Da, respectively. A maximum of two trypsin missed cleavages was allowed. Searches used a reversed sequence decoy strategy to control peptide false discovery rate (FDR) and 1% FDR was set as the threshold for identification.

Macrophage culturing

To model tumor hypoxia, we adopted a previous protocol to culture macrophages in continuous 1% O₂ using the BioSpherix Xvivo X3 system⁴⁹. Primary macrophage progenitors immortalized with a modified estrogen receptor-Hoxb8 fusion (ER-Hoxb8) were generated from female C57BL/6 and Balb/c mice as previously reported⁵⁰ and transduced with EV or GT5 retrovirus. Progenitors were maintained in RPMI 1640 containing 5% heat-inactivated fetal bovine serum, 5% P885L (GM-CSF producing)-conditioned media, and 0.5μM b-estradiol (Sigma) in standard oxygen (~21% O₂). To differentiate into primary macrophages, progenitors were first washed three times with cold PBS to remove b-estradiol. Progenitors were then cultured in a-MEM containing 5% heat-inactivated fetal bovine serum, 1% Penicillin-Streptomycin-Glutamine (100X), and 10% L929 (M-CSF producing)-conditioned media. On day 6 of differentiation, mature macrophages were replated in non-TC treated plates and cultured in hypoxia (1% O₂) for 7 days prior to use in efferocytosis assays. Media was replenished every other day and cells were incubated at 37°C.

CD47 antibody-mediated phagocytosis

Live MDA-MB-231 breast cancer cells were stained with 50μM TAMRA-SE (ThermoFisher) in serum-free HBSS for 45min. Then, cells were washed by incubating in serum-containing assay media for an additional 25min. Labeled MDA-MB-231 cells were co-cultured with macrophages at a 1:1 phagocyte to target ratio together with 10μg/ml anti-CD47 antibody (Clone B6H12, ThermoFisher) for 2h. All phagocytosis assays included four technical replicates for each condition and were performed three independent times (three biological replicates). Cancer cells were subsequently removed via three washes with cold assay media and macrophages were harvested using a cell scraper (Biotium) and assessed by flow cytometry.

Orthotopic breast cancer model

Female, 7-week-old C57BL/6J or Balb/c mice were obtained from Jackson Laboratories and allowed to acclimate for one week. Orthotopic mammary fat pad implantation of breast cancer cells was performed as follows: mice were injected with 5x10⁵ strain-specific EV- or GT5-expressing macrophages together with 5x10⁵ E0771 or 4T1 resuspended in a 50/50 mix of PBS and Matrigel (40μl of each) into the mammary fat pad. Sterile tweezers were used to lift the right forth nipple and a 26G syringe needle was used to implant cell suspensions directly into the mammary fat pad. Treatment with anti-CD47 antibody (MIAP301, BioXCell, 50mg/mouse) began six days after inoculation and continued for five consecutive days, then switched to treatment every other day until the end of the experiment. Body weights and tumor size [length (L) and width (W)] were measured using calipers, and tumor volume (vol) was calculated as [vol = (L x W²)/2].

SKOV3-CD19 human xenograft model

Female, 6 – 8 weeks old NSG mice were obtained from Jackson Laboratories and allowed to acclimate for a week. 2 x 10⁶ SKOV3-CD19 cells in Matrigel were injected s.c. into the flanks of NSG mice. 12 days after tumor engraftment when tumors were palpable, 2 x 10⁶ of transduced EV or GT5 19BBz CAR T cells were injected retro-orbitally. On the same day I.P. 4 g/kg fructose or vehicle (PBS) injections began and continued 5 times a week throughout the duration of the study. Body weights and tumor size [length (L) and width (W)] were measured using calipers, and tumor volume (vol) was calculated as [vol = (L x W²)/2].

Images and schematics

In figure schematics were made with Biorender.

Quantification and statistical analyses

Data analyses were performed using GraphPad Prism. A two-tailed paired Student’s t test was used to determine significance when two conditions were compared. For experiments with more than two conditions a one or two-way ANOVA was used. For survival data a log-rank test was used. In all cases values of p < 0.05 were considered significant. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001. Data are shown as mean ± SEM (standard error of the mean).

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Glut5 (E-2)	SCBT	sc-271055
Actin	Cell Signaling Technology	8457S
Vinculin	Sigma-Aldrich	V9264-200UL
Phospho-Akt (Ser473)	Cell Signaling Technology	4060
Akt (pan)	Cell Signaling Technology	4685
Monoclonal mouse anti-Glut5	MSKCC Antibody Core	N/A
InVivoMab Anti-Mouse PD-1	Bio X Cell	BE0146
InVivoMab mouse IgG2b isotype control	Bio X Cell	BE0086
InVivoMab anti-mouse CTLA-4 (CD152)	Bio X Cell	BE0164
InVivoMab rat IgG2a isotype control	Bio X Cell	BE0089
InVivoMAb anti-mouse CD47 (IAP)	Bio X Cell	BE0270
Human anti-CD3 (OKT-3)	Miltenyi	130-093-387
Alexa Fluor 647 secondary antibody	Jackson Immunoresearch	115-605-071
Alexa Fluor 488 secondary antibody	Jackson Immunoresearch	155-545-071
Pacific Blue anti-mouse CD45	BioLegend	103126
PE-Cy7 anti-mouse TNFα	BioLegend	506324
Brilliant Violent 785 anti-mouse IFNγ	BioLegend	505837
FITC anti-mouse CD45.2	BioLegend	109806
BV650 anti-mouse CD8	BioLegend	100741
PE anti-mouse TIM-3	BioLegend	119706
PerCP-Cy5.5 anti-mouse PD-1	BioLegend	109120
anti-CD47 antibody (Clone B6H12)	Thermo Fisher	14-0479-82
BV785 anti-human CD3	BioLegend	317330
FITC anti-human CD4	BioLegend	317407
APC-Cy7 anti-human CD8	BioLegend	344714
BUV395 anti-human CD25	BD Biosciences	563551
PE-eflour610 anti-human CD69	ThermoFisher Scientific	61-0699-42
PerCP-Cy5.5 anti-human CD71	BioLegend	334114
APC anti-human CD95	BioLegend	305612
BV510 anti-human TIM-3	BioLegend	345030
BV785 anti-human CTLA-4	BioLegend	369624
PE-Cy7 anti-human LAG-3	BioLegend	369208
APC anti-human PD-1	BioLegend	621620
BV510 anti-human CD45RO	BioLegend	304246
BV785 anti-human CD45RA	BioLegend	304140
PE-Cy7 anti-human CD62L	BioLegend	304822
APC anti-human CCR7	BioLegend	353214
BV412 anti-human IFNγ	BioLegend	506538
PE-Cy7 anti-human TNFα	BioLegend	502930
APC anti-human IL-2	BIoLegend	500310
A647 anti-19bbz idiotype antibody 19E3	MSK Antibody and Bioresource Core Facility
PE anti-19bbz idiotype antibody 19E3	MSK Antibody and Bioresource Core Facility
Biological samples
Healthy human adult blood	In house at MSKCC
Chemicals, peptides, and recombinant proteins
Human IL-2	MSKCC pharmacy
Mouse IL-2	BioLegend	575404
Dialyzed FBS	HyClone	SH3007903
2-Mercaptoethanol (BME)	Thermo Fisher	21985023
Polybrene	SCBT	sc-134220
Puromycin	Invivogen	ant-pr-1
OVA peptide	Invivogen	vac-sin
RetroNectin	Takara Bio	T100B
D-Luciferin	Goldbio	LUCK-100
Poly I:C	Invivogen	vac-pic
DNase I	Roche	10104159001
Liberase TL	Roche	05401020001
Titermax	Titermax
D-(−)-Fructose, ≥99%	Sigma-Aldrich	F0127
D-Fructose-U13C6	Sigma-Aldrich	587621
D-[2-13C]fructose	Omicron Biochemicals	FRU-003
D-(+)-Glucose, ≥99.5%	Sigma-Aldrich	G8270
D-Glucose-U13C6	Sigma-Aldrich	464058
D-[2-13C]glucose	Omicron Biochemicals	T24007
PI3K inhibitor LY294002	Selleck Chemicals	S1105
5(6)-TAMRA, SE	Thermo Fisher	C1171
ImmunoCult Human CD3/CD28 T Cell Activator	STEMCELL Technologies	10991
Critical commercial assays
Q5 Hot Start High-Fidelity DNA Polymerase	NEB	M0493S
X-tremeGENE HP	Roche	06366244001
EasySep™ Mouse CD8+ T Cell Isolation Kit	STEMCELL Technologies	19853
Dynabeads™ Mouse T-Activator CD3/CD28 for T-Cell Expansion and Activation	Thermo Fisher	11456D
TransIT-LT1 Transfection Reagent	Mirus Bio	MIR 2300
Lipofectamine 3000	Thermo Fisher	L3000001
CellTiter-Glo	Promega	G7570
RIPA	Thermo Fisher	89900
Halt™ Protease and Phosphatase Inhibitor Cocktail (100X)	Life Technologies	78440
BCA protein assay	Thermo Fisher	23227
DC Protein Assay Reagent A	Bio-Rad	5000113
DC Protein Assay Reagent B	Bio-Rad	5000114
4-12% Bis-Tris gels	Thermo Fisher	NP0322BOX
4-12% Bis-Tris gels	Thermo Fisher	NP0321BOX
Nitrocellulose membranes	Thermo Fisher	LC2009
PVDF transfer membrane, 0.2 um	Thermo Fisher	88520
Anti-mouse HRP-conjugated secondary	Cytiva	NA931
Anti-rabbit HRP-conjugated secondary	Cytiva	NA934
ECL reagent	Cytiva	RPN2209
Immobilon® Forte Western HRP Substrate	MilliporeSigma	WBLUF0500
Restore™ PLUS Western Blot Stripping Buffer	Thermo Fisher	46430
Denville Scientific BLUE BIO FILM	Thermo Fisher	NC9556985
High fructose chow	Inotiv	TD.89247
Steady-Glo Luciferase Assay System	Promega	E2520
Dead Cell Apoptosis Kit with Annexin V	Thermo Fisher	V13241
Cell Activation Cocktail	Biolegend	423303
DAPI	Thermo Fisher	62248
Intracellular Fixation & Permeabilization Buffer Set	Thermo Fisher	88-8824-00
Zombie Violet Fixable Viability Dye	BioLegend	423113
Fc block	BD	553142
Mouse Tumor Dissociation kit	Miltenyi	130-096-730
CD8 TIL Microbeads kit	Miltenyi	130-116-478
Viakrome 808 Fixable Viability Dye	Beckman Coulter	C36628
BD Cytofix/Cytoperm Fixation Permeabilization Kit	BD Biosciences	554717
Experimental models: Cell lines
B16-OVA	MilliporeSigma	SCC420
RAJI	ATCC	CCL-86
NALM6	ATCC	CRL-3273
4T1	ATCC	CRL-2539
HEK293T	ATCC	CRL-3216
E0771	CH3 Biosystems	94A001
MDA-MB-231	ATCC	HTB-26
Sp2/0	ATCC	CRL-1581
PlatE	Dr. Schietinger, MSKCC
SKVO3 (CD19+)	Dr. Brentjens, MSKCC
H29	Dr. Brentjens, MSKCC
293Galv9	Dr. Brentjens, MSKCC
NIH3T3 cells	MSKCC Antibody Core
Experimental models: Organisms/strains
Mice: C57BL/6J	The Jackson Laboratory	RRID:IMSR_JAX:00 0664
Mice: CD45.1 (B6.SJL-Ptprca Pepcb/BoyJ)	The Jackson Laboratory	RRID:IMSR_JAX:00 2014
Mice: Rag1−/− (B6.129S7-Rag1tm1Mom/J)	The Jackson Laboratory	RRID:IMSR_JAX:00 2216
Mice: NSG (NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ)	The Jackson Laboratory	RRID:IMSR_JAX:00 5557
Mice: BALB/c	The Jackson Laboratory	RRID:IMSR_JAX:00 0651
Mice: OT-I	Dr. Schietinger, MSKCC
Mice: Glut5 KO	In house at MSKCC
Oligonucleotides
Forward primer to clone mouse Glut5 into pMSCV (EcoRI site) TAAGCAGAATTCATGGAGGAAAAACATCAAGAGGAGACAGG	This paper
Reverse primer to clone mouse Glut5 into pMSCV (XhoI site) TGCTTACTCGAGTCATTACTACTGCTCCAAGATGGCGTGTGG	This paper
Recombinant DNA
Mouse Glut5 cDNA	Origene	MC204180
pMIG II (PMSCV-IRES-GFP II)	Addgene	52107
pMSCV-IRES-mCherry	Addgene	52114
pMIG II Glut5	This paper
pMSCV-IRES-mCherry Glut5	This paper
pLenti-C-Myc-DDK-P2A-Puro human GLUT5	Origene	RC200418L3
pLenti-C-mGFP-P2A-Puro human GLUT5	Origene	RC200418L4
pLenti-C-Myc-DDK-P2A-Puro mouse Glut5	Origene	MR208045L3
pLenti-C-mGFP-P2A-Puro mouse Glut5	Origene	MR208045L4
CD19BBz CAR construct	Dr. Brentjens, MSKCC
CD19BBz GLUT5	This paper, made by Genscript
psPAX	Addgene	12260
pMD2.G	Addgene	12259
Software and algorithms
FlowJo	Tree Star
Prism	GraphPad
MestReNova	Mestrelab
El-MAVEN (v0.12.0)	Elucidata
Biorender	Biorender.com
ImageJ	imagej.net

Research Highlights.

• Engineering fructose transport in T cells facilitates function

• Engineered cells show metabolic flexibility using fructose in the presence of glucose

• Fructose metabolic engineering extends beyond T cells to other immune cells

• Function can be further enhanced by both pharmacologic and dietary fructose in vivo

Data and code availability

All underlying data and additional information required to reanalyze the data reported in this work is available from the lead contact upon request. This paper does not report original code.