Tumor metastasis to lymph nodes requires YAP-dependent metabolic adaptation

See allHide authors and affiliations

Science  08 Feb 2019:
Vol. 363, Issue 6427, pp. 644-649
DOI: 10.1126/science.aav0173

Fueling lymph node metastases

Metastatic cells can migrate from a primary tumor to distant organs through two routes: They can enter the bloodstream directly, or they can enter a lymph node adjacent to the primary tumor. Little is known about the biological mechanisms that allow tumor cells to survive and grow within lymph nodes. Studying mouse models, Lee et al. found that tumor cells adapt to the lymph node microenvironment by shifting their metabolism toward fatty acid oxidation. This occurs through activation of a signaling pathway driven by the yes-associated protein (YAP) transcription factor. Importantly, inhibition of fatty acid oxidation or YAP signaling suppressed lymph node metastasis in the mice.

Science, this issue p. 644


In cancer patients, metastasis of tumors to sentinel lymph nodes (LNs) predicts disease progression and often guides treatment decisions. The mechanisms underlying tumor LN metastasis are poorly understood. By using comparative transcriptomics and metabolomics analyses of primary and LN-metastatic tumors in mice, we found that LN metastasis requires that tumor cells undergo a metabolic shift toward fatty acid oxidation (FAO). Transcriptional coactivator yes-associated protein (YAP) is selectively activated in LN-metastatic tumors, leading to the up-regulation of genes in the FAO signaling pathway. Pharmacological inhibition of FAO or genetic ablation of YAP suppressed LN metastasis in mice. Several bioactive bile acids accumulated to high levels in the metastatic LNs, and these bile acids activated YAP in tumor cells, likely through the nuclear vitamin D receptor. Inhibition of FAO or YAP may merit exploration as a potential therapeutic strategy for mitigating tumor metastasis to LNs.

Metastasis to a sentinel lymph node (LN) predicts subsequent metastasis to other organs and mortality of cancer patients (1, 2). Most previous studies of tumor metastasis have focused on distant metastasis (3) rather than the mechanism by which tumor cells survive and grow within the LNs. With mounting evidence that the LNs are a foothold for further tumor dissemination (46), elucidating the mechanisms underlying LN metastasis is of paramount importance.

To investigate these mechanisms, we studied mouse tumor models. We performed comparative transcriptomics analysis of three sequential stages of LN metastasis by using sorted green fluorescent protein–labeled (GFP+) B16F10 melanoma cells from primary tumors and micrometastatic and macrometastatic tumor-draining LNs (Fig. 1A and fig. S1, A and B). Heat map and principal components analyses of RNA-sequencing (RNA-seq) data revealed global changes in the transcriptome during tumor progression to LN metastasis (Fig. 1B and fig. S1, C and D). The top up-regulated gene sets in the LN micro- and macrometastatic tumors were related to aspects of lipid biology, such as bile acid metabolism, adipogenesis, fatty acid metabolism, cholesterol homeostasis, and oxidative phosphorylation (Fig. 1, C to F). Pathway analysis of these up-regulated genes indicated that the LN-metastatic tumors stimulated the fatty acid oxidation (FAO) (fig. S1,E and F) and peroxisome proliferator–activated receptor–α signaling pathways (7) (fig. S1, G and H).

Fig. 1 LN-metastatic tumors undergo transcriptomic changes toward increased lipid metabolism.

(A) Immunofluorescence (IF) images showing naïve popliteal LNs (pLNs) and GFP+ tumors in micro- and macrometastatic pLNs. Insets show gross images of each LN. Scale bars, 1 mm. (B) Heat map of the RNA-seq data from primary (PT), micrometastatic (Micro), and macrometastatic (Macro) tumors (n = 4 samples for each group). (C to F) LN-metastatic tumors showing transcriptional induction of lipid metabolism. Shown are gene set enrichment analyses [(C) and (D)] and heat maps for the top 30 up-regulated genes for fatty acid metabolism [(E) and (F)] in micro- and macrometastatic tumors compared with primary tumors. Functions related to lipid biology are underlined in red in (C) and (D). TNFA, tumor necrosis factor α; NFKB, nuclear factor κB; UV, ultraviolet; IL2, interleukin-2; NES, normalized enrichment score; FDR, false discovery rate.

Consistent with this finding, our metabolomics analysis revealed higher levels of fatty acid species in naïve LNs than in the footpads (the primary tumor implantation site) of healthy control mice (fig. S2A). In addition, the LN-metastatic tumors showed greater accumulation of fatty acids than the primary tumors (Fig. 2A), suggesting that LN-metastatic tumors preferentially oxidize fatty acids as fuel in the lipid-rich LN niche. To further investigate this metabolic adaptation of LN-metastatic tumors, we implemented an in vivo enrichment selection method to obtain a highly metastatic subpopulation of B16F10 cells that thrive within the LN parenchyma. At 3 to 4 weeks after the implantation of B16F10 cells into the footpad, LN-metastatic cells were isolated from the LNs, expanded in culture, and re-inoculated into the footpad (fig. S2B). After three rounds of this in vivo selection, the LN metastasis–prone adapted B16F10 cells exhibited higher metastatic activity toward LNs than the parental B16F10 cells (fig. S2C). Moreover, the adapted cells preferentially depended on FAO over glucose or glutamine oxidation as a major pathway for energy production (Fig. 2B and fig. S2D), and they exhibited enhanced FAO capacity when the growth medium was supplemented with fatty acid (fig. S2E). The LN-metastasized tumor exhibited an FAO rate about four times that in the primary tumor or lung-metastasized tumor; this increased rate is comparable to that in brown adipose tissue (fig. S3, A to C, and Fig. 2C).

Fig. 2 Enhanced FAO is required for LN metastasis.

(A) Comparison of in vivo [14C]oleic acid accumulation between primary tumors (PTs) and LN-metastatic tumors (LMTs) by the measurement of the radioactive [14C]oleic acid count per minute (cpm) per milligram of the tumor tissue in the B16F10 melanoma footpad implantation model (n = 6 tumors for each). (B) Metastasis-adapted cells (fig. S2B) favor FAO in vitro. The oxygen consumption rates (OCRs) are compared between the adapted and parental cells for glucose, glutamine, and FAO pathways (n = 8 samples for glucose and fatty acid; n = 7 samples for glutamine). (C) Comparison of 14CO2 production rates in primary tumors, lung-metastatic tumors (lung-MTs), LN-metastatic tumors, and brown adipose tissue (BAT) measured by ex vivo FAO assay with [14C]palmitic acid (n = 4 samples for each group). (D) Treatment schedule for the FAO inhibitor etomoxir in the B16F10 melanoma footpad implantation model. (E and F) Comparison of the gross appearances of metastatic pLNs and various metastasis parameters between PBS- and etomoxir-treated groups (n = 12 samples for each group). Scale bar, 5 mm. AU, arbitrary units. **P < 0.01, ***P < 0.001, and ****P < 0.0001. Error bars are presented as mean ± SEM.

To address the importance of FAO in LN metastasis, we used the mouse B16F10 melanoma footpad implantation model (Fig. 2D). We locally administered etomoxir, a clinically approved inhibitor of FAO, to the anterolateral side of the animal’s leg. Although etomoxir treatment did not affect the size of the primary tumor or the animal’s body weight, it markedly suppressed LN metastasis without affecting the size of the tumor-draining LN (Fig. 2, D to F; fig. S3, D and E; and fig. S4, A to D). It also markedly suppressed LN metastasis after surgical removal of the primary B16F10 melanoma (fig. S4, E to G). Etomoxir treatment also suppressed LN metastasis in two mouse models of breast cancer, the MMTV-PyMT genetic model and the orthotopic 4T1 model (fig. S4, H to M). Moreover, etomoxir treatment suppressed tumor growth when B16F10 cells or 4T1 cells were directly implanted into the LNs (fig. S5, A to E). However, systemic etomoxir treatment did not affect blood-borne metastasis to the lungs in the systemic B16F10 or 4T1 cell injection model (fig. S5, F to K). These findings indicate that the metabolic shift to FAO in tumor cells is required for the cells’ metastatic growth in the tumor-draining LNs.

To identify the molecular events that trigger the metabolic conversion in the LN-metastasized tumors, we revisited our RNA-seq data (Fig. 1). The knockdown of only one gene among the oncogenic signaling genes induced in the metastatic tumors (fig. S6A)—the gene encoding the transcriptional coactivator yes-associated protein (YAP)—significantly reduced FAO in the metastasis-adapted cells (Fig. 3A and fig. S6B). YAP knockdown did not affect glucose or glutamine oxidation (Fig. 3B). Notably, the expression of YAP target genes was induced in the LN-metastatic tumors (Fig. 3C and fig. S6C). Moreover, the retrovirus-mediated overexpression of hyperactive YAP [YAP-5SA (8)] led to a 2.4-fold increase in the level of FAO in cultured B16F10 cells and a 2.3-fold increase in the growth of B16F10 melanoma in LNs (fig. S6, D and E). These results indicate that YAP activation is a key molecular event that mediates FAO activation in LN-metastatic tumors.

Fig. 3 YAP activation is critical for enhanced FAO and tumor LN metastasis.

(A) Comparison of the [14C]palmitic acid oxidation rates in the metastasis-adapted B16F10 cells transfected with small interfering RNAs (siRNAs) targeting the indicated genes (n = 4 samples for each group). siCtrl, control siRNA; siMYC, siEGFR, siKRAS, and siAKT, siRNAs for MYC, epidermal growth factor receptor, KRAS, and AKT, respectively. (B) Comparison of oxygen consumption rates (OCRs) in the adapted cells transfected with control or YAP siRNA (n = 7 samples for each group). (C) Gene set enrichment analysis (left) and heat map (right) of YAP target genes for micrometastatic tumors (Micro) and primary tumors (PTs). FDR, false discovery rate. (D to F) IF images showing metastatic pLNs (D) and highly magnified tumor margins [(E) and (F)]. YAP (green) is localized in the nuclei at the invasive front (arrowheads in b′), but it is localized in the cytosol in the intratumoral area (a′). Scale bars, 100 μm for (D) and (E) and 25 μm for (F). PDGFRβ is an LN stromal cell marker. DAPI, 4′,6-diamidino-2-phenylindole. (G and H) IF images showing low-magnification (G) and high-magnification (H) metastatic axillary LNs with nuclear YAP localization at the invasive front (arrowheads in b′′) in the MMTV-PyMT spontaneous breast cancer mouse. Scale bars, 25 μm. CK8 is a breast cancer cell marker. (I) Quantification of tumor cells with nuclear YAP in both tumor models (n = 13 and 9 samples for B16F10 and MMTV-PyMT, respectively). (J) IF images of primary tumors of implanted B16F10 melanoma and MMTV-PyMT breast cancer showing cytosolic YAP. Scale bars, 25 μm. (K) Schedule depicting inducible knockdown of YAP (YAP-iKD) in the B16F10 melanoma footpad implantation model. The doxycycline diet was started at 2 weeks after the implantation of B16F10 cells into footpads, and pLNs were harvested 10 days after the doxycycline diet. (L and M) Gross appearance of metastatic pLNs and comparisons of metastatic tumor areas and pLN sizes between control (Ctrl-iKD) and YAP-iKD groups (n = 7 samples for each group). Scale bar, 5 mm. AU, arbitrary units. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Error bars are presented as mean ± SEM.

Given that nucleus-localized YAP is viewed as the activated form of YAP (9), we analyzed the localization pattern of YAP in the metastatic LNs. In the tumor cells at the invasive front of LN-metastatic melanoma, YAP was localized predominantly in the nucleus (Fig. 3, D to F and I, top). A similar pattern of YAP localization was found in the tumor cells at the invasive front of LN-metastatic tumors in the MMTV-PyMT breast cancer model (Fig. 3, G, H, and I, bottom). By contrast, in the primary tumors and lung-metastatic tumors of these two cancer models, YAP was localized predominantly in the cytoplasm (Fig. 3J and fig. S6, F to I). Moreover, the migrating tumor cells in the lymphatic vessels also exhibited cytoplasmic localization of YAP (fig. S7, A and B). Thus, YAP activation was limited to the tumor cells that had successfully arrived at the LNs.

To determine whether the LN microenvironment triggers YAP activation, we directly implanted B16F10 cells into the inguinal LNs. At 3 days after implantation, YAP was localized predominantly in the nucleus of the melanoma cells at the invasive front (fig. S7, C and D). We next examined the effect of YAP depletion on LN metastasis by using a system for the doxycycline-inducible knockdown of YAP (YAP-iKD) (fig. S7E). To ensure that YAP was depleted after metastasis, we started mice on a doxycycline diet 2 weeks after the implantation of the melanoma cells (Fig. 3K). YAP depletion strongly suppressed LN metastasis without significantly changing the size of the primary melanoma or the size of the tumor-draining LNs (Fig. 3, K to M, and fig. S7, F to H). YAP depletion after the removal of the primary melanoma also markedly suppressed metastatic tumor growth in LNs (fig. S8, A to C). YAP depletion delayed the growth of B16F10 melanoma and 4T1 breast cancer cells implanted directly into the LNs but had no substantial effect on blood-borne metastasis to the lung in the systemic B16F10 or 4T1 cell injection model (fig. S8, D to O). Together, these findings indicate that YAP activation plays a critical role in efficient LN metastasis.

We next sought to identify the mechanism by which YAP is activated in LN-metastatic tumors. Although hypoxia and cell proliferation are each associated with YAP activation (10, 11), we found no definitive association between either of these factors and the nuclear localization of YAP (fig. S9, A and B). Given that nuclear YAP localization is restricted to the invasive front in the metastatic LNs, we hypothesized that certain signaling ligands present in the LN microenvironment activate YAP. We found that the abundance of several bile acid species was markedly elevated in the LN-metastatic melanoma (Fig. 4A). These bile acids were detected exclusively in the LN-metastatic melanoma and not in normal tissues (footpad tissue and naïve LNs); by contrast, cholesterol—the precursor molecule that gives rise to bile acids—was present at similar levels in all of these tissues (fig. S9C). Moreover, the systemic lymph of mice bearing LN-metastatic tumors contained higher levels of bile acids than that of healthy control mice (fig. S9D). RNA-seq data also revealed highly induced gene signatures of bile acid metabolism in the LN-metastatic tumors compared with primary tumors (Fig. 4B and fig. S9E).

Fig. 4 LN-metastatic tumors produce bile acids that can activate YAP, and YAP activation is correlated with melanoma LN metastasis and patient survival.

(A) Metabolomics analysis revealing a relative abundance of bile acids in primary tumors and LN-metastatic tumors in the B16F10 melanoma footpad implantation model and footpad tissue and naïve LNs of control mice (n = 3 samples for each group). (B) Gene set enrichment analysis (left) and heat map (right) for the genes related to bile acid metabolism in micrometastatic tumors and primary tumors. (C to E) Immunoblot analyses of phosphorylated and total YAP in adapted B16F10 cells after treatment with the indicated concentrations of TDCA (C) or cholesterol [(D) and (E)]. In (D), methyl-β-cyclodextrin (MβCD) was used as a solubilizer for cholesterol (see the blots in fig. S10A). In (E), adapted B16F10 cells were transfected with two different (#1 and #2) CYP7A1 siRNAs and then treated with cholesterol for 60 min. See fig. S9F, S10B, and S10C for quantifications. (F) Images showing YAP in the metastatic sentinel LN from a melanoma patient. YAP is localized mainly in nuclei in the invasive front (right) but not in the intratumoral region (left). The brown color of the metastatic melanoma indicates melanin pigments. Scale bars, 100 μm. (G) Kaplan-Meier survival curves of melanoma patients with nuclear or cytoplasmic YAP at the invasive front of LN-metastatic tumors (n = 8 and 13 patients, respectively). P = 0.0037 by a log-rank test. CA, cholic acid; DCA, deoxycholic acid; TCA, taurocholic acid; MCA #1 and MCA #2, two different muricholic acid (MCA) isomers; FDR, false discovery rate; pYAP, phosphorylated YAP. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Error bars are presented as mean ± SEM.

In addition to their well-established role in dietary fat digestion, bile acids can act as signaling molecules to activate YAP (1216). Thus, we examined whether the bile acids elevated in the LN-metastatic tumors can activate YAP. Treatment with the bile acid taurodeoxycholic acid (TDCA) induced YAP dephosphorylation (activation) in the cultured adapted B16F10 cells within 30 min (Fig. 4C and fig. S9, F to H). The subcutaneous administration of TDCA promoted the growth of LN-metastatic melanoma by ~2.5-fold compared with the control vehicle, phosphate-buffered saline (PBS) (fig. S9, I to L). Treatment with cholesterol also dephosphorylated YAP in the cultured adapted B16F10 cells but with slower kinetics (Fig. 4D and fig. S10, A and B), and this effect was blocked by the depletion of CYP7A1, a key rate-limiting enzyme for the conversion of cholesterol into bile acids (Fig. 4E and fig. S10, C and D). We found that the cultured adapted B16F10 cells generated several putative bile acids from cholesterol in a CYP7A1-dependent manner (fig. S10, E and F). We also observed a marked suppression of tumor growth, together with decreased nuclear YAP localization, in CYP7A1-depleted B16F10 cells compared with control B16F10 cells in the LNs (fig. S10, G to K). These findings suggest that the LN-metastatic tumor itself produces bile acids that can activate YAP in an autocrine manner and stimulate further growth of the LN-metastatic tumor; however, this hypothesis requires further investigation. We identified nuclear vitamin D receptor (VDR) as the bile acid–activated receptor that likely mediates the bile acid–induced YAP activation and tumor growth in LNs (fig. S11, A to G). On the basis of these collective data, we postulate that LN-metastatic tumors accumulate bile acids that activate YAP mainly via VDR, leading to FAO activation and successful adaptation to the LN microenvironment.

Lastly, we investigated the LN-metastatic tumors in 21 patients with melanoma (table S1). About half of the metastatic LNs dissected from patients with melanoma showed a predominantly nuclear YAP localization pattern, specifically at the invasive front (Fig. 4F and fig. S12A). By contrast, primary melanomas from the same patients did not exhibit nuclear YAP localization (fig. S12B). Notably, nuclear YAP localization in the metastatic LNs of melanoma patients correlated with a reduction in distant metastasis–free survival (Fig. 4G). This pattern was independent of patient age or the thickness of the primary lesion (fig. S12C). YAP activation is therefore a key biomarker that distinguishes LN-metastatic tumors with a high risk for metastasis to distant organs in humans.

In this study, by using comparative systematic analysis of LN-metastatic and primary tumors, we discovered that tumors undergo a metabolic shift toward FAO during LN metastasis. We identified YAP, potentially activated by accumulated bile acids, as a crucial driver for tumor LN metastasis through selective stimulation of FAO. The metabolic shift of tumors to aerobic glycolysis (the Warburg effect) is a well-established hallmark of cancer (17). However, different tumors may rely on distinct metabolic pathways, such as glutaminolysis, branched-chain amino acid oxidation, and creatine usage, depending on their microenvironments (1820). This flexible fuel choice may allow tumor cells to survive in metastatic sites containing different nutrients (2125). We found that tumor cells undergoing metastasis to tumor-draining LNs adapt to the LN microenvironment through YAP-dependent adjustment to the FAO pathway, using the plentiful fatty acids within the node microenvironment as fuel. Through our metabolomics and transcriptomics analyses, we unexpectedly identified bile acids as potential molecular triggers of YAP activation in metastatic LNs. Thus, the special microenvironment of the LN, which is enriched with fatty acids as a fuel source and which has an abundance of bile acids, may simultaneously facilitate tumor adaptation and metabolic conversion (fig. S13).

For a variety of mouse and human tumors, the LNs are a preferred route for distant metastasis (5, 6, 2628). Selective inhibition of LN metastasis may prevent or reduce the dissemination of tumor cells to distant organs. Our findings with mouse models suggest that therapies targeting FAO and YAP may be one way to suppress LN metastasis and subsequent distant metastasis. Whether these findings will translate to cancer patients remains to be investigated.

Supplementary Materials

Materials and Methods

Figs. S1 to S13

Table S1

References (2940)

References and Notes

Acknowledgments: We thank S. Y. Choi, T. W. Noh, S. P. Jang, J. H. Song, and D.-S. Lim (KAIST) and Y. Kim (Asan Medical Center) for discussions and experimental support; J. Shin and S. Lee (KAIST) for discussions and statistical support; and S. Seo, J. Bae, and H. T. Kim (Institute for Basic Science) for their technical assistance. Funding: This study was supported by the Institute for Basic Science, funded by the Ministry of Science and ICT, Republic of Korea (IBS-R025-D1 to G.Y.K.). Author contributions: C.-k.L. designed and performed the experiments, analyzed and interpreted the data, and wrote the manuscript; S.-h.J. contributed to the majority of the in vivo experiments; C.J. supervised and analyzed metabolism studies and participated in manuscript preparation; H.B. contributed to metabolism experiments; Y.H.K. contributed to in vitro experiments; I.P. contributed to in vivo experiments and participated in manuscript preparation; S.K.K. analyzed human melanoma–metastatic LNs; G.Y.K. supervised the project, oversaw data analyses and interpretation, and wrote the manuscript. Competing interests: The authors declare no competing interests. Data and materials availability: RNA-seq data have been deposited in the ArrayExpress database at EMBL-EBI under accession number E-MTAB-7621.
View Abstract

Stay Connected to Science

Navigate This Article