To evaluate the level of lipids in DCs we used the lipophilic fluorescent dye Bodipy 493/503. CD11c⁺ DCs isolated from spleens of mice bearing EL-4 tumor displayed higher fluorescence than their control counterparts (Fig. 1a). Using flow cytometry we detected two populations of CD11c⁺ DCs in tumor-bearing (TB) mice. One population termed DCs with normal lipid content (DC-NL) representing 30–50% of cells had a similar level of fluorescence to that in control DCs. The other population (50–70% of cells) referred to as DCs with high lipid content (DC-HL) had fluorescence higher than that in control DCs (Fig. 1b). The level of lipids in spleen DCs was elevated in all 4 tested tumor models in C56BL/6 and BALB/c mice as compared with DCs from control mice (Fig. 1c) and it increased with tumor progression (Fig. 1d). Increased lipid levels were observed in CD11c⁺CD8⁺ DCs and CD11c⁺CD11b⁺B220⁻ conventional DCs (cDC) but not in CD11c⁺CD11b⁻B220⁺ plasmacytoid DCs (pDC) (Fig. 1e). DCs isolated from tumors had higher levels of lipids than DCs from blood and lymph nodes (LN) (Fig. 1f). DCs from the LN, bone marrow (BM), and blood from TB mice had higher levels of lipids than DCs from control mice (Fig. 1g). The lipid content was equally elevated in DCs from tumor-draining and distant LNs (Fig. 1h).

To evaluate the nature of these lipids, CD11c⁺ DCs were isolated from spleens of naïve and TB mice. Total lipids were extracted and analyzed by high performance thin layer chromatography (HPTLC), and electrospray ionization mass spectroscopy (ESI-MS). DCs from TB mice showed dramatically higher levels of triglycerides (triacylglycerol, TAG) than cells from naïve mice. In contrast, no differences were found in serum TAG level between naive and TB mice (Fig. 1i). Most of the molecular species of TAG were elevated in DCs from TB mice as compared with their control counterparts (Supplementary Fig. 2a) and no differences were found in the profile of major molecular species of TAG between serum samples (data not shown). In contrast to TAG no changes were observed in the levels of cholesteryl-esters and phospholipids (Supplementary Fig. 2b,c).

Lipid levels were evaluated in DCs from 12 individuals with head and neck cancer (HNC). Peripheral blood (PB), draining LN, and tumor tissues were collected during surgery. DCs were identified as Lineage⁻CD4⁺ cells15 (Fig. 2a). DCs from PB of persons with cancer showed significantly higher levels of lipids than PB DCs from the healthy individuals. The lipid level in DCs from LN and tumors was considerably higher than that in DCs from PB (Fig. 2b). Similar findings were obtained in the cohorts of patients with non-small lung cancer and renal cell cancer (Fig. 2c). In separate experiments we found that LN DCs from persons with HNC contained significantly higher level of lipids than LN DCs from individuals with benign tumors (Fig. 2d). Thus, these data strongly corroborate those obtained from mouse tumor models.

To study the mechanisms of lipid accumulation in DCs these cells were generated in vitro from BM progenitors (BMP) using GM-CSF and IL-4. Short 24 h incubation of DCs with tumor explant supernatants (TES) resulted in more than 3-fold increase in the lipid level (Fig. 3a). Analysis of lipid content by ESI-MS revealed that DCs generated in the presence of TES had almost 6-fold higher level of TAG than DCs generated in control culture medium (CM) (Fig. 3b). There was an increase in the level of all molecular species of TAG but no changes were observed in the level of phospholipids and cholesteryl-esters in these DCs (Supplementary Fig. 3). The total level of TAG in CT-26 TES was > 4 times higher than in CM (Fig. 3b). Despite variability in TAG level in TES obtained from different tumor cell lines they all caused comparable up-regulation of lipids in DCs (Supplementary Fig. 4). To directly evaluate fatty acid uptake, DCs cultured for 24 h with TES were incubated with fluorescently labeled palmitic fatty acid for 30 min at 37°C. DCs pre-incubated with TES had dramatically higher uptake of fatty acid than DCs incubated in control medium (Fig. 3c). Overnight incubation of DCs from spleens of TB mice in serum-free medium eliminated elevated level of lipids in these cells, while not affecting the level of lipids in control DCs (data not shown). These data suggested that increased uptake of lipids may be the major mechanism of lipid accumulation by DCs.

We then investigated why DCs from tumor-bearing hosts have higher lipid uptake than in controls. Scavenger receptors (SR) play an important role in intracellular transport of lipids. Currently little is known about changes in their expression on DCs from tumor-bearing hosts. Therefore, we evaluated the level of four major SR (Macrophage scavenger receptor, Msr1 (CD204), Cd36 (Srb), Cd68, and Marco). EL-4 TES caused significant up-regulation of the expression of Msr1 in DCs (Fig. 3d). This was associated with increased expression of this receptor on the surface of DCs (Fig. 3e). A similar effect was observed with TES from CT-26 and B16F10 tumors (data not shown). Msr1 was the only SR significantly elevated in DCs from TB mice (Fig. 3f). cDC from TB mice had 3-fold higher levels of Msr1 than their control counterparts. In contrast, no up-regulation of Sra was observed in pDC (Fig. 3g). No differences were found in the expression of Msr1 between DCs from naïve mice and DC-NL from TB mice. However, DC-HL expressed substantially higher levels of Msr1 (Fig. 3h). Similar results were obtained in experiments in vitro (Supplementary Fig. 5a). Elevated Msr1 expression was detected in DCs from all tested sites of TB mice and the level of Msr1 in DCs isolated from tumor was 20–50-fold higher than in other sites (Supplementary Fig. 5b). Thus, the pattern of Msr1 expression was similar to that of lipids and suggested that up-regulation of this receptor might be involved in lipid accumulation in DCs.

To directly test this hypothesis we used fucoidan, a soluble ligand which prevents binding of natural ligands to SRs. Fucoidan did not affect DC viability but completely abrogated the effect of CT-26 TES on lipid accumulation (data not shown). Pre-incubation of DCs with neutralizing antibodies to Msr1 completely abrogated the effect of TES on lipid accumulation, whereas Cd36-specific antibody had little effect (Fig. 3i). We generated DCs from BMP of wild-type and Msr1^−/− C57BL/6 mice (CD45.2⁺). These DCs were transferred i.v. into sub-lethally irradiated congenic (CD45.1⁺) tumor-free or EL-4 TB recipients. Two days later the level of lipids was evaluated in CD45.2⁺ donor’s cells isolated from recipient spleens. Donor’s DCs in TB recipients demonstrated significantly higher levels of lipids than cells in tumor-free recipients. In contrast, DCs generated from Msr1^−/− mice failed to accumulate lipid after transfer to TB recipients (Fig. 3j). Analysis of TAG content by ESI-MS showed that in contrast to wild-type DCs Msr1 deficient DCs did not accumulate TAG during the culture with CT-26 TES (Supplementary Fig. 6). Taken together these data indicate that up-regulation of Msr1 plays a major role in accumulation of lipids in DCs in cancer.

To investigate whether lipid accumulation in DCs has functional consequences, DC-HL and DC-NL from spleens of control and CT26 TB mice were sorted using the gates set around lipid levels in control DCs (Fig. 4a) and then used for stimulation of allogeneic T cells. DCs from control mice and DC-NL from TB mice had similar effects on allogeneic T-cell proliferation, whereas DC-HL from TB mice had substantially lower stimulatory activity (Fig. 4b). Similar results were obtained with in vitro generated DCs cultured for 48 h with TES (Fig. 4c).

Msr1-specific antibody abrogated the negative effects of TES on DC function (Fig. 4d). TES inhibited ability of DCs generated from wild-type BMP to stimulate allogeneic T cells, whereas it did not affect this function of DCs generated from Msr1^−/− BMP (Fig. 4e).

DC-NL and DC-HL isolated from EL-4 TB mice equally stimulated responses of transgenic CD4⁺ or CD8⁺ T cells to OVA-derived specific peptides (Fig. 4f,g). Different results were obtained when DCs were loaded overnight with soluble protein (OVA). CD11c⁺ DCs from TB mice had reduced ability to stimulate proliferation of OVA-specific CD4⁺ T cells (Fig. 4h). DC-NL from TB mice and control DCs showed similar activity in stimulation of OVA-specific CD4⁺ (Fig. 4i) and CD8⁺ (Fig. 4j) T cells, whereas DC-HL induced a substantially lower response. Thus DC-HL had substantially reduced capacity to stimulate T cells when it required processing of antigen.

We compared the phenotype of CD11c⁺ DCs with normal and high lipid content from TB mice. We found no differences in the proportion of DCs expressing markers of macrophages or granulocytes (F4/80 and Gr-1) as well as in the level of expression of these markers between DC-NL and DC-HL (Supplementary Fig. 7a). Although DCs from TB mice had significantly lower levels of MHC class I, II, and co-stimulatory molecules (CD40, CD86, and CD80) than DCs from naive mice, no differences between DC-NL and DC-HL were found (Supplementary Fig. 7b).

One possible explanation for these results could be a decreased uptake of soluble proteins by DC-HL. To test this possibility DCs generated from BMP in the presence of TES were incubated with fluorescently labeled OVA-APC. DC-NL had the same levels of OVA uptake as control DCs, whereas DC-HL showed significantly higher levels of OVA-APC (Fig. 5a). Similar results were obtained in vivo after injection of EL-4 TB mice i.p. with OVA-APC (Fig. 5b). Thus, DC-HL internalized more soluble protein than DC-NL.

To assess the ability of DCs to process antigen we used BSA-DQ, a self-quenching conjugate of bovine serum albumin that exhibits fluorescence upon proteolytic degradation. To normalize for increased protein uptake by DC-HL we used the ratio of OVA-FITC and BSA-DQ fluorescence after loading DCs with these proteins. DCs isolated from TB mice or cultured with TES had substantially lower BSA-DQ/OVA-FITC ratio than DCs from control mice or cells cultured in control medium (Fig. 5c). This effect was observed at all tested time points (Fig. 5d). To verify these observations we used 25-d1.16 antibody that recognizes the OVA derived epitope bound to H2K^b (pMHC). CD11c⁺ DCs isolated from the spleens of EL-4 TB or control mice were loaded with OVA overnight and then stained with 25–D1.16 antibody. The binding of this antibody to DCs from TB mice was lower than to DCs from control mice (Fig. 5e). DCs generated from BMP in the presence of TES were loaded with OVA and then stained with Bodipy and pMHC antibody. DC-HL had considerably lower binding of the antibody than DC-NL (Fig. 5f).

We then addressed whether DC-NL and DC-HL differ in the presentation of tumor-associated antigen. Draining LN were isolated from mice with EG-7 tumors that express OVA and the level of pMHC in LN DCs was evaluated. DC-HL had lower expression of pMHC than DC-NL from the same mice (Fig. 5g). DC-HL and DC-NL cells were sorted from LN of EG-7 TB mice and cultured with OVA-specific OT-II T cells. DC-HL induced a lower response of OT-II T cells than DC-NL (Fig. 5h). In contrast, no difference in the T-cell response between DC-HL and DC-NL was seen when DCs were loaded with specific peptide prior to incubation with T-cells (Fig. 5i). Taken together these data indicate that DC-HL may have defects in processing of tumor-associated proteins.

To identify the potential role of lipid accumulation in abnormal DC function in cancer we tried to regulate fatty acid levels using an inhibitor of acetyl-CoA carboxylase (ACC), 5-(tetradecycloxy)-2-furoic acid (TOFA)16. Since TAG undergo rapid degradation in the cells, the maintenance of their high level requires active fatty acid synthesis. We hypothesized that when synthesis is blocked, then cells will not be able to sustain high level of TAG even if lipid uptake is increased. When HPCs were cultured with TES in the presence of TOFA, it prevented the accumulation of lipids in DCs (Fig. 6a) without affecting the viability or causing up-regulation of co-stimulatory molecules or MHC class II (data not shown). Treatment with TOFA significantly improved the ability of DCs generated in the presence of TES to stimulate allogeneic T cells (Fig. 6b). To test the effect of TOFA in vivo, EL-4 TB mice were treated for 2 weeks with TOFA filled osmotic pumps. At the selected dose (1μg/hr) TOFA did not affect the number of DCs and did not cause changes in the expression of MHC class II or co-stimulatory molecules (data not shown) but it reduced the presence of lipid-laden DCs (Fig. 6c) and restored the ability of DCs to stimulate allogeneic T cells (Fig. 6d).

To test the possibility that TOFA might improve antitumor response EL-4 TB mice were immunized with DCs transduced with Ad-survivin (Ad-surv)17. Vaccine or TOFA alone had little effect on growth of established tumor. Combination of TOFA with vaccine resulted in significant (p=0.041) decrease in tumor growth (Fig. 6e). In a different experimental system, C57BL/6 mice were injected s.c. with B16F10 melanoma cells and then vaccinated with the peptide derived from melanoma specific protein - tyrosinase 1. TOFA alone did not affect tumor growth. Vaccine alone caused a temporary decrease in tumor growth, but the combination of TOFA with vaccine provided for significantly (p<0.01) more potent antitumor effects (Fig. 6f).

Several cohorts of human subjects have been analyzed. All individuals signed University of South Florida IRB approved consent forms. The description of cohorts and processing of samples is provided in supplementary materials.

Female Balb/c or C57BL/6 mice were obtained from the National Cancer Institute. Tumor cell lines included CT-26 (on Balb/c mice) and MC38 (on C57BL/6 mice) colon carcinomas, B16F10 melanoma, EL4 lymphoma (both on C57BL/6 mice), and DA3 breast carcinoma (on Balb/c mice) described elsewhere26. All cells were maintained in Dulbecco’s modified eagle’s medium (Invitrogen Corp.) plus 10% fetal bovine serum (FBS, Sigma-Aldrich) at 37°C, 5% CO₂. Tumors were injected subcutaneously (s.c.) at 5×10⁵ cells per mouse and allowed to grow 3–4 weeks or until tumors reached 2 cm in diameter. OT-I TCR-transgenic mice (C57Bl/6-Tg(TCRaTCRb)1100mjb) have a T cell receptor that recognizes OVA residues 257–264 in the context of H2K^b. OT-II transgenic mice (C57BL/6-Tg(TcraTcrb)425Cbn/J) express the mouse T cell receptor that is specific for OVA_323–339 in the context of I-A^b. These mice, as well as Msr1^−/− mice (B6.Cg-Msr1^tm1Csk/J), were obtained from Jackson Laboratories. Mice were housed and euthanized according to University of South Florida IACUC guidelines.

The following peptides were used in the studies: OT-I specific peptide – SIINFEKL, OT-II specific peptide – ISQAVHAAHAEINEAGR, control non-specific peptide – RAHYNIVTF. All peptides were obtained from the American Peptide Company (Vista, CA).

DCs were generated from BM progenitor cells using culture with 10 ng/mL recombinant mouse GM-CSF (Invitrogen Corp), and 10 ng/mL interleukin 4 (IL-4) (R&D Systems) as described previously27.

Tumor explants were prepared from freshly isolated subcutaneous tumors (see supplementary material).

Cells were incubated with fluorescently labeled antibodies directed against cell surface markers. All antibodies were purchased from either BD Pharmingen or Serotec. 25-d1.16 antibody was obtained from eBioscience. Cell surface labeling was performed on ice for 30 min. Cells were then washed and re-suspended in 500 μL of Bodipy 493/503 at 0.5μg/mL in PBS. Cells were stained for 15 min at room temperature before the analysis. All experiments using Bodipy 493/503 staining were performed on LSRII or FACSAria (BD). For cell sorts, the gates defining normal lipid levels for discrimination of DC-NL vs. DC-HL in TB hosts were set using the fluorescence of control DCs as the background. A low pressure sort was performed using the BD FacsAria cell sorter with a 100 μm nozzle.

See supplementary material.

DCs were incubated with fluorescently labeled (Bodipy FL) palmitic fatty acid (Molecular Probes) at 37°C for 15–30 min. Controls consisted of unloaded DCs or DCs loaded on ice. 5×10⁶ DCs were re-suspended in 500μL of 10μg/mL fatty acid.

T cell proliferation was measured using a mixed leukocyte reaction or antigen specific proliferation. DCs were isolated as described above. T cells were isolated using R&D Systems Mouse T cell Enrichment columns, according to the manufacturer’s protocol (R&D Systems). T cells were plated at 10⁵ T cells per well. DCs and T cells were mixed at different ratios. Antigen specific T cell proliferation was assessed with either OT-I or OT-II transgenic T cells. Isolated DCs (5×10⁶ cells/mL) were incubated with 10 μg/mL of specific or control peptides for 90 minutes. Peptide loaded DCs were then washed and mixed with appropriate transgenic T cells at different ratios. Cells were incubated for 72 h. ³[H]-thymidine was then added at 1μCi per 200uL of cells per well for an additional 18 hours of incubation followed by cell harvesting and radioactivity count using a liquid scintillation counter.

C57BL/6 mice were injected s.c. with 5×10⁵ EL-4 or 3×10⁵ B16F10 tumor cells. Four-five days later EL-4 TB mice were immunized with s.c. injection of 5×10⁵ DCs transduced Ad-surv and vaccination was repeated twice with weekly interval. B16F10 TB mice were vaccinated twice weekly i.v. with a mixture of 150 μg optimized Trp1_455–463/9M (TAPDNLGYM) peptide and two Toll-like receptor ligands (TLR-L) [50 μg poly-IC, 100 μg CpG-1826]. CpG-1826 was prepared by the Mayo Clinic Molecular Core Facility. Poly-IC (Hiltonol, a clinical grade stabilized formulation containing poly-L-lysine and carboxymethyl cellulose) was provided by Dr. Andres Salazar (Oncovir, Inc., Washington, DC).

PCR was performed as described earlier 28 and details are provided in supplementary material.

Statistical analysis was performed using 2-tailed Mann-Whitney or Wilcoxon non-parametric tests with significance determined at p<0.05. The geometric mean is reported for all fluorescence results. Tumor measurements were analyzed using two-way ANOVA test with Bonferroni post-test. All statistical analyses were performed using GraphPad PRISM 5 software (GraphPad Software, Inc. La Jolla, CA).