Circulating Adipose Fatty Acid Binding Protein Is a New Link Underlying Obesity-Associated Breast/Mammary Tumor Development

SUMMARY

It is clear that obesity increases the risk of many types of cancer, including breast cancer. However, the underlying molecular mechanisms by which obesity is linked to cancer risk remain to be defined. Herein, we report that circulating adipose fatty acid binding protein (A-FABP) promotes obesity-associated breast cancer development. Using clinical samples we demonstrated that circulating A-FABP levels were significantly increased in obese patients with breast cancer in comparison to those without breast cancer. Circulating A-FABP released by adipose tissue directly targeted mammary tumor cells, enhancing tumor stemness and aggressiveness through activation of the IL-6/STAT3/ALDH1 pathway. Importantly, genetic deletion of A-FABP successfully reduced tumor ALHD1 activation and obesity-associated mammary tumor growth and development in different mouse models. Collectively, these data suggest circulating A-FABP as a new link between obesity and breast cancer risk, thereby providing A-FABP as a potential new therapeutic target for treatment of obesity-associated cancers.

Graphical abstract

INTRODUCTION

In the past several decades, obesity has become an epidemic in the U.S. and worldwide (Finucane et al., 2011; Flegal et al., 2016). Obesity is mostly caused by overnutrition (Swinburn et al., 2011). Excess calories stored in adipose tissues lead to metabolic abnormalities and multiple obesity-related diseases. Besides cardiovascular disease and diabetes, obesity increases the risk of many types of cancer, including endometrial, esophageal and breast cancer (Hoyo et al., 2012; Dougan et al., 2015; Khandekar et al., 2011). For example, the risk of developing breast cancer is 20% to 40% higher in postmenopausal obese women compared to normal weight women (Munsell et al., 2014). Moreover, overweight and obesity were significantly associated with poor prognosis and increased mortality in women with breast cancer (Cleary and Grossmann, 2009; Calle et al., 2003; Cleveland et al., 2007). Although the links between obesity and cancer are increasingly appreciated, the underlying mechanisms by which obesity increases cancer risk remain elusive.

Several cellular and molecular mechanisms have been proposed to explain the obesity/cancer axis, which includes cancer-associated adipocytes (CAA), obesity-related inflammatory cytokines (IL-6 and TNFα), lipids (lysophosphatidic acid and prostaglandins), adipokines (leptin and adiponectin), insulin/insulin-like growth factors (IGFs), and sex hormones (Khandekar et al., 2011; Park et al., 2014; Louie et al., 2013). Although substantiated by significant clinical and experimental data, these mechanisms remain contentious due to the complex multisystem interactions between obesity and cancer (Roberts et al., 2010). For instance, the sex hormone hypothesis may help to explain hormone-sensitive tumors, but obesity is associated with many types of sex hormone-unrelated cancers. Moreover, there is little information on how these mechanisms converge or interact to drive cancer development. Considering the clear epidemiologic association of obesity with many types of cancer (Lauby-Secretan et al., 2016), we speculate that other unrecognized mechanisms are involved in the obesity/cancer associations.

Fatty acid binding proteins (FABPs) consist of a family of cytosolic proteins which are known to coordinate lipid transport and responses (Storch and McDermott, 2009; Furuhashi and Hotamisligil, 2008). FABP members display tightly-regulated patterns of tissue distribution, such as A-FABP in adipose tissues and epidermal FABP (E-FABP) in the skin, suggesting distinct roles for individual FABPs in different tissues and cells. Our recent studies demonstrated that A-FABP and E-FABP exhibit different expression patterns in macrophages. While E-FABP was highly expressed in CD11c+ macrophages and promoted the production of pro-inflammatory cytokines, A-FABP was predominantly expressed in CD11c-macrophages, facilitating saturated fatty acid-induced ceramide synthesis (Zhang et al., 2014; Zhang et al., 2015; Zhang et al., 2017; Zeng et al., 2018). Emerging evidence suggests a remarkable increase of serum levels of A-FABP in obesity (Xu et al., 2007; Hotamisligil and Bernlohr, 2015; Burak et al., 2015), and interestingly, the elevated A-FABP levels are associated with breast and ovarian cancer progression (Nieman et al., 2011; Hancke et al., 2010). Thus, A-FABP may represent a previously unappreciated obesity-derived factor promoting obesity-associated cancer initiation and progression.

It has been shown that cancer cells can remain dormant in vivo for many years before growing into metastatic tumors (Rice, 2012). Understanding the mechanisms by which cancer (stem-like) cells awaken is critical to minimize tumor recurrence. Unlike fluctuating surface markers, aldehyde dehydrogenase 1(ALDH1) has recently been identified as an operative marker for breast cancer stem cells, which provides a critical tool for identifying the activation of these tumor-initiating cells (Balicki, 2007; Douville et al., 2009; Ginestier et al., 2007; Gunjal et al., 2015). Given that STAT3 signaling is preferentially activated in tumor-initiating breast cancer cells, and environmental extrinsic factors (e.g. immune cell-derived IL-22) are able to increase cancer stemness and tumorigenic potential through activation of STAT3 (Kryczek et al., 2014), we hypothesize that obesity-increased A-FABP serves as a new extrinsic factor promoting breast cancer development by enhancing STAT3-mediated cancer stemness.

Here, we designed experiments to address our hypothesis using both human samples and genetically modified mouse models. Our data demonstrated that elevated levels of circulating A-FABP in obese subjects promoted tumor stemness and aggressiveness. Importantly, we compared mammary tumor development and growth in obese transgenic mouse models with or without A-FABP ablation and determined the molecular mechanisms by which A-FABP enhanced mammary/breast tumor stemness through activation of the IL-6/STAT3/ALDH1 pathway. Our findings provide clinical evidence and a biological mechanism demonstrating how circulating A-FABP drives obesity-associated breast cancer.

RESULTS

Obesity is associated with elevated levels of circulating A-FABP

Accumulating evidence suggests that circulating levels of A-FABP are markedly increased in obese mice and humans (Xu et al., 2007; Hotamisligil and Bernlohr, 2015; Burak et al., 2015). To determine the tissue and cellular source of elevated A-FABP in the circulation, we collected and analyzed serum samples from obese women before and whenever possible, at 3, 6, and 12 months after Roux-en-Y gastric bypass obesity surgery (see Table S1 for surgery patient information). Analyses using linear mixed effect models showed that circulating levels of A-FABP were directly associated with body mass index (BMI) for all 84 patients (49 bypass surgery and 35 non-surgery patients) with an estimated slope of 1.25, p<0.001 and ρ=0.351 (Fig. 1A). Analyses stratified by menopausal status demonstrated that A-FABP levels were associated with both pre-menopausal women (n=45) and post-menopausal women (n=39) with p values of 0.015 and 0.001 and ρ values of 0.344 and 0.552, respectively (Fig. 1B, 1C), indicating that the positive association between A-FABP and BMI is independent of menopausal status. Interestingly, when we further analyzed A-FABP levels in patients after obesity surgery, we found that the change in A-FABP levels was positively related to the change in BMI (p=0.02, ρ=0.32), especially in post-menopausal women with an estimated slope of 6.10, p=0.001 and ρ=0.327 (Fig. 1D, 1E). These results extend previous finding by showing that elevated A-FABP levels decrease in response to the loss of adipose tissue, suggesting that circulating levels of A-FABP are derived from adipose tissue in these patients.

Figure 1. Obesity-associated circulating A-FABP levels are elevated in obese women with breast cancer.

(A-C) Correlation analyses between serum levels of A-FABP and BMI of 84 women (A), the sub-samples (n=45) of pre-menopause patients (B), and the sub-samples (n=39) of post-menopause patients (C).

(D-E) Correlation analyses between the change in A-FABP levels and the change in BMI from the entire 35 sample set (D) and sub-samples (n=11) from post-menopausal patients after the Roux-en-Y gastric bypass surgery (E).

(F-G) Female mice weaned at 3–4 weeks were randomly grouped and fed LFD and HFD, respectively. Weight of these mice was shown after feeding LFD or HFD (n=7/group) for 22 weeks (F). Serum levels of A-FABP were measured in these mice by ELISA (G).

(H) Equal numbers of adipocytes and macrophages (10⁵/ml) collected from LFD-fed lean mice and HFD-fed obese mice (n=3/group) were cultured for 24 hour. Concentrations of A-FABP levels in the supernatants were measured by ELISA.

(I) Analysis of serum levels of A-FABP in obese (BMI>30, n=69) and non-obese (BMI≤30, n=115) patients without breast cancer by ELISA.

(J) Analysis of serum levels of A-FABP in obese (BMI>30, n=34) and non-obese (BMI≤30, n=67) patients with breast cancer by ELISA.

(K) Comparison of serum levels of A-FABP in non-obese patients with (n=67) or without (n=115) breast cancer.

(L) Comparison of serum levels of A-FABP in obese patients with (n=34) or without (n=69) breast cancer.

Figure F-H was repeated 3 times. Data shown are mean ± SD (*p<0.05, **p<0.01, ****p<0.0001, n.s.: no significance). Also see Figure S1 and Table S1, S2.

Next we used high fat diet (HFD)-induced obese mice to validate our clinical observations. Compared to low fat diet (LFD)-fed lean mice, HFD-fed obese mice exhibited significant increases of body weight and adiposity (p=0.001) (Fig. 1F). Meanwhile, circulating levels of A-FABP were correspondingly elevated in obese mice (31.8 ± 5.8ng/ml for lean mice vs 104.9 ± 8.3ng/ml for obese mice) (p=0.0018) (Fig. 1G). However, circulating levels of other biomarkers, including E-FABP, IL-6, TNFα and IL-10, were not obviously altered in the HFD-induced obese mice (Fig. S1A-S1D). A-FABP is highly expressed in adipocytes and macrophages, the two most common cell populations in the obese adipose tissue. To determine the cellular source of elevated A-FABP in the circulation, we measured A-FABP levels in culture supernatants containing equal numbers of adipocytes and macrophages. Levels of adipocyte-derived A-FABP were 300-fold higher than those from macrophages (Fig. 1H). These results suggest that circulating A-FABP is primarily secreted by adipocytes, not macrophages, in the adipose tissue.

Circulating A-FABP, but not E-FABP, is upregulated in obesity-associated breast cancer patients

Given the fact that obesity is associated with increased risk of breast cancer, it was likely that elevated levels of circulating A-FABP might be a new obesity-associated link to breast cancer. We collected serum prior to diagnostic biopsy from a cohort of 285 women who presented with concerning breast lesions and analyzed A-FABP levels in the samples (see Table S2 for patient information). Consistent with previous studies, A-FABP levels in obese women who underwent biopsy were significantly higher than these in lean women both in the cohort with benign lesion (p<0.05) (Fig. 1I) and in women diagnosed with breast cancer (p<0.0001) (Fig. 1J). However, upon further analysis of A-FABP levels in non-cancer vs. cancer patients, we found for the first time that A-FABP levels were only elevated in obese breast cancer women (31.3 ± 2.61 ng/ml vs 40.8 ± 3.98 ng/ml, p<0.05), but not in lean breast cancer patients (23.7 ± 1.60 ng/ml vs 20.1 ± 2.13 ng/ml, p=0.17) (Fig. 1K, 1L). Analysis stratified by menopausal status showed that A-FABP levels were overall higher in postmenopausal than premenopausal women (Fig S1E), but cancer stages and estrogen receptor status did not affect A-FABP levels in breast cancer patients (Fig S1F, S1G). These data suggest that obesity, especially postmenopausal obesity, but not cancer status, contributes to the elevated levels of circulating A-FABP. By contrast, serum levels of E-FABP, another adipocyte/macrophage-associated FABP member, were similar in different patient groups (Fig. S1H, S1I), further supporting that A-FABP is uniquely related to obesity-associated breast cancer. We also analyzed obesity-associated adipokines and cytokines in women with and without breast cancer using protein arrays. We did not notice any obvious alterations for these proteins, such as adiponectin, IGFBPs (insulin-like growth factor binding proteins), leptin, etc (Fig S1J, S1K). Altogether, our results indicate that circulating levels of A-FABP are abnormally elevated in obesity-associated breast cancer, supporting our hypothesis that circulating A-FABP is a driver of obesity-associated breast cancer development.

Soluble A-FABP promotes tumor cell aggressiveness in vitro

To assess the biological importance of circulating A-FABP in obesity-related breast cancer, we next determined if elevated A-FABP impacted mammary tumor cell behavior. First, we cultured human breast cancer-derived MCF-7 cells and mouse mammary tumor E0771 cells in the presence/absence of soluble A-FABP at the levels found in obese women (100~150ng/ml) and measured if exogenous A-FABP promoted tumor colony formation. Soluble A-FABP increased colony formation of both MCF-7 and E0771 cells as compared to untreated controls (p<0.05). By contrast, similar levels of soluble E-FABP did not have colony-promoting effects (Fig. 2A-2D). Secondly, we determined if soluble A-FABP enhanced tumor cell migration. Using wound healing assays we observed that A-FABP significantly promoted the migration of both MCF-7 and E0771 cells (p<0.01) (Fig. 2E-2H). Thirdly, we demonstrated that A-FABP, but not E-FABP, significantly enhanced tumor cell proliferation (Fig. 2I, 2J). The proliferative effect of A-FABP was blocked in the presence of anti-A-FABP antibody, confirming a specific role of A-FABP in enhancing tumor aggressiveness in vitro. We also analyzed the pro-tumor effect of A-FABP using tumor cells derived from other tissues, such as lung LL/2 carcinoma cells. Consistently, A-FABP, but not E-FABP significantly promoted LL/2 cell proliferation (Fig. S2A, S2B). Finally, we evaluated the effects of A-FABP on tumor cell invasion using a 3-dimensional spheroid assay (Berens et al., 2015). A-FABP treatment significantly increased invaded area and distance of both MCF-7 cells and E0771 cells (Fig. 2K-2N, Fig.S2C,S2D). All these data clearly indicate that exogenous A-FABP increases tumor cell aggressiveness in vitro. Of note, analyses of tumor cell cycle and apoptosis did not show apparent alterations in response to A-FABP treatment (Fig. S2E), suggesting that a unique mechanism underlies the pro-tumor effects of A-FABP.

Figure 2. Soluble A-FABP promotes invasiveness of tumor cells.

(A-B) Colony formation of human MCF-7 breast cancer cells treated with recombinant human A-FABP (100ng/ml) or E-FABP (100ng/ml) for 3 weeks (A). Average numbers of colonies are shown panel B.

(C-D) Colony formation of mouse E0771 mammary tumor cells treated with recombinant mouse A-FABP (100ng/ml) or E-FABP (100ng/ml) for 3 weeks (C). Average numbers of colonies are shown panel D.

(E-F) Migration of MCF-7 cells in response to human A-FABP treatment for 48h (E). Migration distance is shown in panel F.

(G-H) Migration of E0771 cells in response to mouse A-FABP treatment for 24h (G). Migration distance is shown in panel H.

(I-J) Analysis of proliferation of MCF-7 cells (I) and E0771 cells (J) with MTT assays with treatment of A-FABP (100ng/ml), E-FABP (100ng/ml) or A-FABP plus anti-A-FABP neutralized antibody for 72 hours.

(K-L) Analysis of invasiveness of MCF-7 cells treated with or without A-FABP (100ng/ml) for 7 days by spheroid invasion assays. K, representative pictures of MCF-7 cell invasion (red curved line: invaded area, blue circle: spheroid). L, analysis of quantitative invaded area of MCF-7 treated with or without A-FABP using Image J software.

(M-N) Analysis of invasiveness of E0771 cells treated with or without A-FABP (100ng/ml) for 2 days by spheroid invasion assays. M, representative pictures of E0771cell invasion (red curved line: invaded area, blue circle: spheroid). N, analysis of quantitative invaded area of E0771 cells treated with or without A-FABP using Image J software.

All experiments were repeated three times. Data shown are mean ± SD (*, p<0.05, **, p<0.01, ***, p<0.001, ****p<0.0001 as compared to untreated controls). Also see Figure S2.

A-FABP treatment enhances mammary tumor stemness

It has been well established that cancer stem cells (CSCs) promote a tumor invasive phenotype and thereby contribute to tumor development (Kryczek et al., 2014; Vermeulen et al., 2012). We speculated that A-FABP may promote tumor aggressiveness by enhancing the stemness of mammary tumors. Using stem cell sphere formation assays, we observed that A-FABP treatment greatly promoted sphere formation of E0771 (Fig. 3A, 3B) and MCF-7 cells (Fig. 3C, 3D), whereas E-FABP treatment had no effect on the sphere formation of E0771 or MCF-7 cells (Fig. S3A-S3D). The sphere-promoting effect of A-FABP was confirmed in BT474 breast cancer cells (Fig. S3E, S3F). Using cells dissociated from primary spheres of E0771 and MCF-7 cells, we further performed the secondary sphere formation and in vitro limiting dilution assays (LDA). Secondary sphere numbers (Fig. S3G, 3H) and sphere-initiating cell frequency (Fig. S3I, S3J, Table S3, S4) in A-FABP treated cells were significantly higher than untreated cells , suggesting that A-FABP treatment enhances the self-renewal of sphere-initiating cells. As CSCs are enriched in the side population (SP) (Harris et al., 2008), we further analyzed the impact of A-FABP on the SP of different tumor cells. A-FABP treatment significantly promoted the SP of E0771 and MCF-7 cells (Fig. 3E-3H). Moreover, A-FABP treatment also enhanced the expression of classical stem cells markers (e.g. OCT3/4) in E0771 and MCF-7 cells (Fig. S3K, S3L).These data demonstrate that obesity-elevated A-FABP can directly promote the stemness of mammary tumor cells.

Figure 3. A-FABP treatment enhances mammary tumor stemness.

(A-B) Sphere formation of E0771 cells treated with or without mouse A-FABP (100ng/ml) for 6 days. Representative pictures of spheres in one high-power field are shown in the panel A. Average numbers of spheres in each well are shown in the panel B.

(C-D) Sphere formation of MCF-7 cells treated with or without human A-FABP (100ng/ml) for 6 days. Representative pictures of spheres in one high-power field are shown in the panel C. Average numbers of spheres in each well are shown in the panel D.

(E-F) Analysis of side population of E0771 cells treated with or without mouse A-FABP (100ng/ml) for 24 hours (E). Average percentage of side population is shown in panel F. Verapamil is the negative control for the experiment.

(G-H) Analysis of side population of MCF-7 cells treated with or without human A-FABP (100ng/ml) for 24 hours (G). Average percentage of side population is shown in panel H. Verapamil is the negative control for the experiment.

(I) A heat-map of breast cancer stem cell related genes in E0771 cell treated with or without mouse 100ng/ml A-FABP.

(J) Analysis of expression of ALDH1a1, a2 and a3 in E0771 cells treated with or without A-FABP (100ng/ml) for 24 hours by real-time PCR.

(K-L) Flow cytometric analysis of ALDH1 activity in E0771 cells treated with 100ng/ml A-FABP for 24 hours by Aldefluor assay (K). Average of ALDH1⁺ tumor cells is shown in panel L.

(M) E0771 cells (1×10⁵) treated with or without 100ng/ml A-FABP for 24h were orthotopically injected in the low fat pads of C57B/6 mice (n=4). Each tumor size was measured at the indicated time points.

(N) The frequency of tumor initiating cells (TIC) of E0771 treated with or without 100ng.ml A-FABP for 24 hours as measured by the in vivo limiting dilution assay. Data shown are mean and 95% confidence interval calculated using the ELDA software.

Experiments were performed with triplicates for three times. Tumor injection was repeated twice. Data are shown as mean ± SD (*p<0.05, **, p<0.01, ***, p<0.001 as compared to untreated controls). Also see Figure S3, Table S3–S6.

To dissect the molecular mechanisms by which A-FABP promotes tumor stemness, we performed Affymetrix microarray using E0771 cells treated with or without A-FABP and focused on analysis of CSC-related genes (Table S5). Interestingly, genes that encode ALDH1 (e.g. ALDH1a3, ALDH1b1), a functional hallmark of breast cancer stem cells, was significantly upregulated in A-FABP-treated cells (Fig. 3I). Upregulation of ALDH1 genes was also confirmed using real-time PCR (Fig. 3J). Functional analysis of ALDH1with the ALDEFLUOR assay further demonstrated that A-FABP treatment enhanced the ALDH1 enzyme activity of E0771 cells (Fig. 3K, 3L). When we injected A-FABP-treated or untreated E0771 cells into mammary fat pads of C57B/6 mice, we observed that A-FABP-conditioned tumors exhibited an accelerated growth rate in vivo (Fig. 3M). Moreover, in vivo limiting dilution assay demonstrated that A-FABP treatment significantly increased the frequency of tumor-initiating cells 4 fold in E0771 cells (Fig. 3N, Table S6). Altogether, our data indicate that circulating A-FABP enhances tumor stemness, leading to increased tumor growth and initiation in vivo.

A-FABP-induced ALDH1 activity is dependent on STAT3 signaling

We next determined the signaling pathway(s) by which soluble A-FABP-induced ALDH1 activation in tumor cells. To determine if soluble FABPs could be directly taken up by mammary tumor cells, we measured A-FABP expression in tumor cells before and after treatment with A-FABP or E-FABP. There were no alterations in protein levels of A-FABP in mammary tumor cells (Fig. S4A, S4B), suggesting neither uptake nor de novo transcription of A-FABP by tumor cells in response to exogenous FABP treatment. Given that mammary tumor (e.g. MCF-7 and E0771) cells responded to exogenous A-FABP treatment by enhanced sphere formation regardless of endogenous A-FABP expression status, we reasoned that exogenous A-FABP might directly bind to tumor cells and initiate signaling cascades for subsequent ALDH1 activation. To this end, we fluorescently labeled A-FABP and analyzed if A-FABP could directly bind to tumor cells by flow cytometric analysis. We observed a clear A-FABP-positive population in MCF-7 and E0771 cells. By contrast, there was no positive staining of either isotype protein or fluorescent-labeled E-FABP under the same condition (Fig. 4A), suggesting a specific binding of A-FABP to cell surface molecules on the tumor cells. Considering the reported interactions of A-FABP with membrane phospholipids during FA transport (Gericke et al., 1997; Liou et al., 2002), we used multiplex lipid beads and demonstrated that A-FABP treatment inhibited the binding of PLCδ to PtdIns(4,5)P2 and PtdIns(3,4,5)P3 (Fig. S4C), suggesting that A-FABP directly interacts with membrane PI(4,5)P2 and PI (3,4,5)P3 initiating signaling events for ALDH1 activation in tumor cells.

Figure 4. A-FABP-induced ALDH1 activity is dependent on STAT3.

(A) Analysis of FABP-binding sub-population in MCF-7 cells and E0771 cells stained with Alexa Fluor 488-conjugated A-FABP or E-FABP by flow cytometry.

(B-C) Detecting the expression of phosphorylated STAT3 and total STAT3 in MCF-7 cells (B) treated with human A-FABP (150ng/ml) at the indicated time points by Western blotting. The relative band density at the 12 hour is shown in panel C.

(D-E) Detecting the expression of phosphorylated STAT3 and total STAT3 in E0771 cells (D) treated with mouse A-FABP (150ng/ml) at the indicated time points by Western blotting. The relative band density at the 12 hour is shown in panel E.

(F) Western blotting for measuring STAT3 expression in MCF-7 cells treated with S3I-201, a specific STAT3 inhibitor, for 12 hours at the indicated concentrations.

(G) Analysis of ALDH1 activity in MCF-7 cells treated with the STAT3 inhibitor S3I-201 for 12 hours by flow cytometry.

(H) Analysis of STAT3 expression in MCF-7 cells transfected with two different sets of STAT3 siRNAs or scrambled oligo controls for 12 hours by Western blotting.

(I-J) Analysis of A-FABP-induced ALDH1 activity by flow cytometry in MCF-7 cells (I) and E0771 cells (J) treated with two different sets of specific STAT3 siRNA or scrambled oligo controls for 12 hours.

(K-L) Flow cytometric analysis of A-FABP-induced ALDH1 activity in the presence of anti-IL-6 neutralizing antibody (20ng/ml) or control IgG at the indicated time points in E0771 cells (K) and MCF-7 cells (L).

(M) Western blotting analysis of phosphorylated STAT3 and total STAT3 in MCF-7 cells treated with human A-FABP (150ng/ml) in the presence or absence of anti-IL-6 neutralizing antibody for 12 hours.

All the experiments are repeated three times. Data shown are mean ± SD (*, p<0.05, **, p<0.01 as compared to untreated or scramble controls). Also see Figure S4.

To identify possible signaling pathways activated by extracellular A-FABP, we measured epigenetic regulators by focusing on DNA methylation and histone modifications due to their roles in cancer stem cell reprogramming (Munoz et al., 2012). We did not observe any consistent alterations in either DNA methyltransferases (DNMTs) or histone methylation of lysine residues in MCF-7 and E0771 cells in response to A-FABP treatment (Fig. S4D-S4G). Nor did we see the activation of ERK, AKT or mTOR signaling in A-FABP-treated mammary tumor cells (Fig. S4H, S4I). However, we noticed that A-FABP consistently induced phosphorylation of STAT3 in both human MCF-7 cells and mouse E0771 cells (Fig. 4B-4E). Given that STAT3 is activated in ALDH1+ breast cancer stem cells (Thakur et al., 2015; Wei et al., 2014; Lin et al., 2013), we reasoned that A-FABP might enhance ALDH1 activity via STAT3 signaling in mammary tumors. We first inhibited STAT3 activation in MCF-7 cells using S3I-201, a specific chemical inhibitor of STAT3 (Fig. 4F), and showed that chemical inhibition of STAT3 significantly reduced ALDH1 activity (Fig. 4G). We then transfected MCF-7 cells with small interfering RNAs (siRNAs) which specifically targeted STAT3 (Fig. 4H), and further confirmed that STAT3 inhibition by siRNAs reduced A-FABP-induced ALDH1 activity (Fig. 4I). Consistent with these observations, inhibition of STAT3 with siRNAs in E0771 cells also abrogated A-FABP-induced ALDH1 activity (Fig. 4J). Thus, STAT3 signaling is critical for A-FABP-induced ALDH1 activity.

To further dissect how A-FABP induces STAT3 activation, we measured STAT3-activating cytokines in mammary tumor cells in response to A-FABP treatment. We found that soluble A-FABP promoted marked IL-6 production in E0771 and MCF-7 cells (Fig. S4J, S4K), implying that IL-6 may be responsible for exogenous A-FABP-induced STAT3 activation in tumor cells. To this end, we measured A-FABP-induced ALDH1 activity in the presence or absence of IL-6 neutralizing antibody. Indeed, upregulation of ALDH1 activity by A-FABP was blocked by the anti-IL-6 antibody in both E0771 and MCF-7 cells (Fig. 4K, 4L). Moreover, A-FABP-induced STAT3 phosphorylation was also blocked by the IL-6 neutralizing antibody (Fig. 4M). Collectively, our data indicate that exogenous A-FABP induces tumor stemness through activation of STAT3 signaling in mammary tumor cells.

A-FABP expression is critical for tumor ALDH1 activity in vivo

As A-FABP induced ALDH1 expression and activity in mammary tumor cells in vitro, we next determined whether A-FABP was essential for ALDH1 activation in vivo. By analysis of publicly accessible databases with human breast cancer patients, we found that patient overall survival was associated with the expression levels of ALDH1 (the higher ALDH1 expression, the lower survival rate of the patients) (Fig. S5A, S5B), suggesting a critical role of ALDH1 in the survival of breast cancer patients. To determine the A-FABP/ALDH1 associations in an animal model of breast cancer, we injected E0771 tumor cells into the mammary fat pad of C57BL/6 mice and measured circulating levels of FABPs and tumor ALDH1 activity at different time points. As tumor cells mobilize host lipid metabolism facilitating their growth (Huang et al., 2016), we noticed that circulating A-FABP was increased in the early stage (0–10 days), but declined in the late stage (21 days) of tumor growth. In contrast, circulating E-FABP levels remained low levels during the process (Fig. 5A). As implanted E0771 tumor cells did not secrete A-FABP to the circulation (Fig. S5C), these data suggest that adipose tissue-derived A-FABP is mobilized to support tumor growth. Accordingly, tumor ALDH1 activity was enhanced and then decreased in a similar pattern as circulating levels of A-FABP (Fig. 5B), substantiating our speculation of a positive association of A-FABP, but not E-FABP, with ALDH1 activity in vivo. To assess a causal relationship between A-FABP-induced ALDH1 activation and tumor growth in vivo, we measured ALDH1 activity and tumor growth using mice with genetically ablated A-FABP (A-FABP-/- mice). Tumors collected from A-FABP-/- mice exhibited a significant reduction of ALDH1 activity as compared to WT mice (Fig. 5C, 5D), and tumor growth was much slower (p<0.01) in the absence of A-FABP (Fig. 5E). Of note, A-FABP deficiency had no impact on circulating levels of E-FABP (Fig. 5F). Moreover, using BMS309403, an A-FABP specific inhibitor (Yan et al., 2018), we confirmed that A-FABP inhibition suppressed tumor growth and tumor weight with decreased IL-6 and tumor ALDH1 activity (Fig. 5G-5J). When we further analyzed the influence of E-FABP on mammary tumor growth using E-FABP-/- mice, we demonstrated that E-FABP deficiency did not impact serum levels of A-FABP (Fig. S5D), nor did it slow tumor growth in vivo (Fig. S5E). Taken together, these results indicate that circulating A-FABP, but not E-FABP, is critical in inducing tumor ALDH1 activity and tumor growth in vivo.

Figure 5. A-FABP expression is critical for tumor ALDH1 activity in vivo.

(A) Measurement of the serum levels of A-FABP and E-FABP on day 0, day 10 and day 21 post tumor implantation in C57B/6 mice (n=4) by ELISA.

(B) Analysis of ALDH1 activity in E0771 cells before and after E0771 tumor injection in vivo for 10 and 21 days by flow cytometry.

(C-D) Detection of the ALDH1 activity in E0771 tumors on day 12 after tumor implantation in WT and A-FABP-/- mice (n=3/group) (G). Statistical analysis of ALDH1 activity is shown in panel H.

(E) Measurement of tumor volume in WT and A-FABP-/- mice (n=4/group) post orthotopical injection of E0771 cells (5×10⁵) at the indicated time points.

(F) Analysis of the serum levels of A-FABP and E-FABP in WT and A-FABP-/- mice on day 21 post E0771 tumor implantation (n=4/group) by ELISA.

(G) Measurement of the mammary E0771 tumor volume in WT mice treated with BMS309403 (5mg/kg, i.p.) or control DMSO (n=5/group) at the indicated time points.

(H) Tumor weight of mice treated with DMSO or BMS309403 post E0771 implantation (5×10⁵) for 21 days.

(I) Measurement of the serum levels of IL-6 in BMS309403-treated or untreated mice after E0771 tumor implantation for 21 days by ELISA.

(J) Analysis of ALDH1 activity in tumors collected from BMS309403-treated or untreated mice after E0771 tumor implantation for 21 days by flow cytometry.

Data shown as mean ± SD are representative of three experiments (*p<0.05, **p<0.01 as compared to either day 0 controls or WT controls). Also see Figure S5.

A-FABP deficiency uncouples obesity and transplanted mammary tumor growth in HFD-induced obese mice

As obesity increases breast cancer risk, we next fed C57BL/6 mice either a LFD or a HFD for 5–6 months and orthotopically injected the same numbers of E0771 mammary tumor cells in these lean and obese mice. Tumor growth in obese mice was much faster than in lean mice, and the tumor weight in obese mice was 4-fold higher than lean mice (Fig. S6A, S6B). Accelerated tumor growth in obese mice has been attributed to the excess energy supply in these mice. Our initial observations were consistent with this notion as tumor growth markedly lowered circulating levels of free fatty acids (FFAs) in both lean and obese mice (Fig. S6C). However, when we compared body weight and tumor growth using HFD-fed obese WT and A-FABP-/- mice, we noticed that A-FABP-/- mice gained more weight than WT mice consuming the same HFD (Fig. 6A), and serum levels of FFAs were the same between obese A-FABP-/- and WT mice (Fig. 6B). If more energy supply was the actual cause of accelerated tumor growth in obese mice, we would expect enhanced tumor growth in A-FABP-/- mice vs. WT mice. In fact, tumor growth in obese A-FABP-/- mice was significantly reduced as compared to obese WT mice (Fig. 6C), and tumor weight was 4-fold lower when A-FABP was absent in these obese mice (Fig. 6D). The observation that A-FABP deficiency uncoupled obesity and mammary tumor growth in vivo suggests that A-FABP represents a previously unappreciated link underlying obesity-associated tumor progression.

Figure 6. A-FABP deficiency uncouples obesity and mammary tumor growth in high fat diet-fed mice.

(A) Monitoring body weight of WT and A-FABP-/- mice on the HFD for 22 weeks (n=7/group).

(B) Analysis of the serum FFA levels in obese WT and A-FABP-/- mice after 22 weeks on HFD by ELISA (n=4/group).

(C) Measurement of the tumor volume in obese WT and A-FABP-/- mice (n=7/group) post orthotopical injection of E0771 cells (5×10⁵) at the indicated time points.

(D) Tumor weight of obese WT and A-FABP-/- mice (n=7/group) post implantation of E0771 cells (5×10⁵) for 21 days.

(E) Measurement of the serum levels of IL-6 in obese WT (n=7) and A-FABP-/- mice (n=6) before and after E0771 tumor injection for 21 days by ELISA.

(F) Analysis of ALDH1 activity in tumors collected from obese WT and A-FABP-/- mice (n=7) by flow cytometry.

(G-H) Analysis of 100ng/ml A-FABP-induced ALDH1 activity in E0771 cells treated with or without palmitic acid (PA, 100µM) (G) or oleic acid (OA, 100µM) (H) for 12 hours by flow cytometry.

(I-J) Flow cytometric staining of A-FABP-binding sub-populations in E0771 cells treated with or without saturated PA or unsaturated OA at the indicated concentration (I). Mean values of A-FABP-binding sub-populations are shown in panel J.

(K) Real-time PCR analysis of IL-6 production in E0771 tumor cells treated with either PA (100 µM) plus A-FABP (100ng/ml) or OA (100 µM) plus A-FABP (100ng/ml) for 8 hours.

Obese tumor model studies were repeated twice. Other in vitro experiments were repeated two or three times. Data shown are mean ± SD (*p<0.05, **p<0.01 as compared to A-FABP-/- group or untreated controls, respectively). Also see Figure S6.

In determining how A-FABP mediated obesity-associated mammary tumor growth, we first compared obesity-associated circulating factors before and after tumor implantation in obese WT and A- FABP-/- mice. IL-6, but not estradiol, E-FABP, TNFα and IL-1β, was significantly reduced in A-FABP-/- tumor-bearing mice (Fig. 6E, Fig. S6D-S6G), suggesting that A-FABP-mediated IL-6 signaling promotes mammary tumor growth in vivo. We next measured ALDH1 activity in tumors from obese WT and A-FABP-/- mice and observed more ALDH1+ tumor stem cells in WT (4.36 ± 1.18%) than A-FABP-/- mice (1.39 ± 0.64%) (Fig. 6F). As A-FABP bound both saturated and unsaturated FAs, which were elevated in HFD-induced obese mice (Fig. S6C), we wondered if upregulated FAs in obese mice might cooperate with A-FABP to induce ALDH1 activity. We measured A-FABP-induced tumor ALDH1 activity in the absence or presence of either saturated palmitic acid (PA) or unsaturated oleic acid (OA). Intriguingly, PA, but not OA, cooperated with A-FABP in promoting ALDH1 activity in mammary E0771 cells (Fig. 6G, 6H). Further analysis revealed that treatment with PA promoted more E0771 cells (13.4%) binding to fluorescent-labelled A-FABP than treatment with BSA (6.56%) or OA (4.47%) (Fig. 6I, 6J). Accordingly, E0771 cells treated with PA and A-FABP produced more IL-6 than cells treated with OA and A-FABP (Fig. 6K). All these data suggest that elevated saturated FAs in obesity cooperate with circulating A-FABP in promoting obesity-associated mammary growth by enhancing tumor cell stemness in vivo.

Deficiency of A-FABP diminishes primary mammary tumor development in transgenic obese mice

MMTV-TGFα transgenic mice slowly develop mammary tumors in the second year of life, representing a good model of postmenopausal human breast cancer (Cleary et al., 2010; Halter et al., 1992). Our previous studies have shown that a HFD increases mammary tumor development in these mice (Dogan et al., 2007). To corroborate the critical role of A-FABP in driving obesity-associated mammary tumor development, we crossed A-FABP-/- mice with MMTV-TGFα mice to generate MMTV-TGFα/A-FABP+/+ (WT) and MMTV-TGFα/A-FABP-/- (KO) mice and observed if A-FABP deficiency uncoupled HFD-induced obesity and mammary tumor development in these mice. A-FABP deficiency did not impact HFD-induced obesity (Fig. 7A), nor did it affect normal mammary development in MMTV-TGFα transgenic mice (Fig. S7A, S7B). However, during the 14 month period on the HFD, 57.1% of obese WT mice developed palpable mammary tumors while none of obese A-FABP-/- mice exhibited tumors except occasionally observed ductal hyperplasia (Fig. 7B-7D). In line with the E0771 syngeneic mouse models, we did not notice a significant difference in FFAs, estradiol and IL-1β comparing obese WT and KO mice (Fig. S7C-S7E). Upon further analysis of STAT3 phosphorylation in primary mammary tumors, we demonstrated that STAT3 was activated in tumor cells, specifically in CD44+ tumor stem cells (Fig. 7E). Consistently, percentage of ALDH1+ cells were significantly elevated in WT vs KO mammary tissue (Fig. 7F, 7G), corroborating the critical role of STAT3/ALDH1 signaling in A-FABP-driving mammary tumor development in vivo. The significantly positive association between serum IL-6 levels and ALDH1 activity was observed in these transgenic mice (Figure 7H), which was consistent with the analysis of human breast cancer databases (Fig. 7I, Fig. S7F, S7G). All these results suggest a critical role of IL-6/STAT3/ALDH1axis in mammary/breast tumor development. In contrast to our findings in obese mice, lean MMTV-TGFα/WT and KO mice also developed mammary tumors at the incidence of 28.6% and 11.1%, respectively (Fig. S7H-S7J). Notably, while obesity increased mammary tumor incidence in WT mice, deficiency of A-FABP uncoupled obesity-associated mammary tumor development in the KO mice. In addition, we did not notice significant differences in STAT3 phosphorylation or ALDH1activity in lean WT and KO mice (Fig. S7K, S7L). Altogether, our data support that circulating A-FABP functions as a driver of obesity-increased mammary tumor development which acts through the IL-6/STAT3/ALDH1 axis.

Figure 7. Deficiency of A-FABP abrogates mammary tumor development in obese MMTV-TGFα transgenic mice.

(A) Monitoring body weight of MMTV-TGFα/WT mice (n=14) and MMTV-TGFα/A-FABP-/- mice (n=15) on the HFD for 14 months.

(B) Representative pictures of mammary tumor development in MMTV-TGFα/WT and MMTV-TGFα/A-FABP-/- mice.

(C) Mammary tumor incidence in MMTV-TGFα/WT and MMTV-TGFα/A-FABP-/- mice after 14 months on the HFD.

(D) Analysis of mammary tumor development in mammary tissues of MMTV-TGFα/WT and MMTV-TGFα/A-FABP-/- mice by hematoxylin and eosin (H&E) staining. The bar represents 100µm.

(E) Confocal microscopy analysis of phospho-STAT3 (red) in CD44+ (green) primary mammary tumor cells (DAPI for nuclei, blue) of obese MMTV-TGFα/WT mice. The bar represents 10µm.

(F-G) Flow cytometric analysis of ALDH1 activity in mammary tissues of MMTV-TGFα/WT and MMTV-TGFα/A-FABP-/- mice by Aldefluor assay (F). Average of ALDH1+ tumor cells is shown in panel G.

(H) Spearman correlation analysis of serum IL-6 levels and tumor ALDH1 activity in obese MMTV-TGFα transgenic mice.

(I) Spearman correlation analysis of the mRNA expression of IL-6 and ALDH1a3 in breast cancer patients.

Data shown are mean ± SD (*, p<0.05 as compared to KO controls). Also see Figure S7.

DISCUSSION

Despite strong evidence linking obesity to cancer development, the molecular mechanisms underlying this association have been unclear. In the current study, we provide novel insights into how obesity drives mammary/breast cancer development. Specifically, we demonstrate that: 1) Obesity is associated with elevated levels of circulating A-FABP in healthy women and even more so in women with breast cancer; 2) Circulating A-FABP enhances the aggressiveness of mammary tumor cells in both in vitro and in mouse models; 3) A-FABP activates the tumor IL-6/STAT3/ALDH1 pathway leading to enhanced stemness of tumor cells; 4) Genetic deletion of A-FABP diminishes obesity-induced tumor development and growth in both transgenic and syngeneic mouse models. Thus, A-FABP may represent a new link between obesity and increased breast cancer risk.

Most studies evaluating the molecular mechanisms driving obesity-associated cancers have focused on metabolic and endocrine factors including leptin/adiponectin and insulin/IGFs (Khandekar et al., 2011; van Kruijsdijk et al., 2009; Park et al., 2010). However, the complexity of the obesity/cancer associations suggests that other general molecular links are involved. To explore additional potential links between obesity and breast cancer, A-FABP attracted our attention due to a number of independent observations: 1) As a lipid chaperone, A-FABP is highly expressed in adipocytes and macrophages, both of which are predominantly involved in the pathogenesis of obesity; 2) Once considered simply as sites of energy storage, adipocytes are now recognized as endocrine cells that play an important role in breast cancer invasion (Dirat et al., 2010; Tan et al., 2011; Wang et al., 2012). Similarly, traditionally viewed as an intracellular protein, it is now clear that A-FABP can be released into the circulation and associated with adiposity (Xu et al., 2007); 3) Emerging evidence also showed that circulating levels of A-FABP were directly associated with breast cancer progression in obese patients (Hancke et al., 2010). Moreover, our recent data demonstrated that A-FABP was expressed in a specific subset of tumor-associated macrophages (TAMs), facilitating the pro-tumor functions of TAMs (Hao et al., 2018). Given that A-FABP occupies a central place in integrating metabolic and inflammatory pathways, we hypothesized that A-FABP represents a novel molecular mechanism that links obesity to breast cancer.

To identify the source of circulating A-FABP, we demonstrated that A-FABP levels were directly associated with baseline BMI and with change in BMI of obese patients undergoing Roux-en-Y bypass surgery for morbid obesity. Cellular experiments confirmed that circulating A-FABP was primarily produced by adipocytes during obesity. E-FABP was also expressed in adipocytes, although to a lesser degree. We did not observe significant differences in the circulating levels of E-FABP in obese patients with or without breast cancer, suggesting distinct roles for A-FABP and E-FABP in obesity-associated cancer. Despite A-FABP and E-FABP having similar lipid binding profiles, these two proteins escorted their ligands into different metabolic pathways (Tan et al., 2002; Lee et al., 2015). Data generated in our laboratory and by others have shown that E-FABP expression in a wide range of immune cells, including CD11c+ macrophages and T cells, is essential for maintenance/performance of its intracellular functions, while A-FABP is elevated in response to external factors (e.g. excess energy intake, tumors) and is secreted for extracellular activities (Zhang et al., 2018; Zeng et al., 2018; Cao et al., 2013; Hertzel et al., 2006; Li et al., 2009). These observations added new insights into our understanding regarding the unique role of individual FABPs during obesity.

It is commonly believed that FABPs function through their hydrophobic ligands. For example, rapid growth and metastasis of ovarian cancer cells in the omentum has been attributed to the energy (fatty acids) carried by A-FABP (Nieman et al., 2011), but whether A-FABP itself signals to promote cancer progression remains unknown. Obesity is associated with elevated levels of circulating and tissue fatty acids (Frayn et al., 1996; Boden, 2011). If elevated fatty acids can directly promote tumor growth, we would assume more tumor burden in mice with higher levels of FFAs. However, this was not the case when A-FABP was genetically ablated in mouse models. A-FABP-/- mice are more obese, exhibit equivalent FFA levels in the circulation and higher levels of lipids in different organs than WT littermates when fed a HFD (Baar et al., 2005). However, tumor development and growth in obese A-FABP-/- mice was significantly reduced as compared to obese WT mice. These observations unequivocally demonstrate that A-FABP is an essential mediator linking obesity-associated tumor development. Moreover, when we exogenously included PA or OA, the two most abundant fatty acids in vivo (Freigang et al., 2013), in cultured mammary tumor cells, we found that PA synergistically promoted tumor cell stemness when combined with A-FABP. In sharp contrast, OA had minimal impacts on A-FABP-induced tumor stemness. Despite the fact that A-FABP binds both PA and OA (Lee et al., 2015), the distinct effects of these FAs indicate that A-FABP does not simply transport metabolic energy to facilitate tumor progression. Rather, A-FABP appears to function as a signaling molecule to drive obesity-associated breast cancer development.

Like other cancer risk factors, obesity does not cause every obese patient to get cancer. Despite the genetic diversity, we noticed that soluble A-FABP only binds to a small percentage of tumor cells when cultured in vitro. Intriguingly, when environmental factors (e.g. saturated FAs) were present, they increased the “visibility” of tumor cells to exogenous A-FABP. It has been shown that A-FABP harbors strong positive potential across the helix-turn-helix motif and effectively binds negatively charged membrane phospholipids (Gericke et al., 1997; LiCata and Bernlohr, 1998; Liou et al., 2002). Consistent with this, we showed that A-FABP directly interacted with tumor cell membrane lipids (e.g. PI(4,5)P2 and PI(3,4,5)P3) in lipid binding assays. It is very likely that elevation of saturated FAs either changes the lipid composition or exposes negatively charged lipids in the cellular plasma membrane, which can be recognized by exogenous A-FABP. Thus, intact epithelial cell membranes in healthy obese women are not targeted by circulating A-FABP. However, if environmental factors (e.g. elevated saturated FAs) compromise the integrity of epithelial cell membrane by exposing membrane phospholipids, circulating A-FABP will recognize these cells by binding to the exposed lipids and initiate tumor-promoting signals in these obese patients.

In exploring the tumor-promoting signaling initiated by exogenous A-FABP binding, we demonstrated a consistent activation of the oncogenic STAT3 signaling in both human and mouse tumor cells. Interestingly, STAT3-activating cytokines (e.g. IL-6) were remarkably increased in tumor cells in response to A-FABP treatment, suggesting that A-FABP binds to cell membrane phospholipids initiating tumor-promoting IL-6 signaling. These results are supported by our previous studies showing that A-FABP induces NFκB and IL-6 signaling in other cells (Yan et al., 2017; Makowski et al., 2005). Other pathways, including PIP2/protein kinase C, PIP3/tyrosine kinases, may also be involved in A-FABP-induced STAT3 activation (Jain et al., 1999; Guryanova et al., 2011). In further dissecting the cancer-promoting mechanisms of STAT3, we showed that inhibition of STAT3 activation either by chemical inhibitors or by siRNAs diminished A-FABP-induced ALDH1 enzyme activity in tumor cells. Breast cancer cells expression of ALDH1 or surface marker CD44⁺CD24⁻ have been shown to display cancer stem cell properties (Ginestier et al., 2007; Al-Hajj et al., 2003), but due to the fluctuating phenotype and discrepancy of surface markers in tumor cells, high ALDH1 activity is characterized as a hallmark of cancer stem cells (Douville et al., 2009; Gunjal et al., 2015). The observations of STAT3-dependent activation of ALDH1 suggest the STAT3/ALDH1 axis as a novel pathway in enhancing breast cancer stemness. Moreover, the reduced activation of IL-6/STAT3/ALDH1 signaling and diminished mammary tumor development and growth observed in both xenograft and MMTV-TGFα transgenic A-FABP-/- mice also support a critical role of this axis in A-FABP-mediated pro-tumor effect. Of note, although estrogens are presumed to contribute to obesity-related breast cancer development, A-FABP deficiency has no impact on estrogen levels in both lean and obese mice. Thus, the current study provides mechanistic insights into how circulating A-FABP binds membrane signaling lipids and initiates tumor-promoting signaling through the IL-6/STAT3/ALDH1 pathway.

Considering that current dogmas underling obesity-cancer associations are somewhat scattered, our findings describing A-FABP as a link between obesity and cancer provide a specific mechanism around which many of these notions can coalesce. First, obesity is associated with expanded volume of adipose tissue and abnormal lipid levels which induce upregulation and release of A-FABP from adipocytes into the circulation. Second, the role of A-FABP in insulin resistance, inflammatory cytokines and sex dimorphism has been well established. Importantly, the current study demonstrates that circulating A-FABP directly binds membrane-compromised cells, enhances tumor stemness and tumorigenicity. Thus, A-FABP bridges tumor-associated stromal cells to tumor stem cells and integrates adipokines to tumor-promoting STAT3/ALDH1 signaling, thereby representing a new link between obesity and cancer. Our study not only demonstrates a novel mechanism of action of A-FABP, but also provides a conceptual advantage in this field. Therefore, targeting A-FABP may emerge as a novel strategy in treatment of obesity-linked breast and other types of cancer.

LIMITATIONS OF STUDY

Given the complexity of breast cancer and the breath of obesity/cancer associations, circulating A-FABP is unlikely to be the sole unifying “one fits all” mechanism underlying obesity-associated cancers. There are several potential caveats of the study. For example, circulating A-FABP levels may be affected by other factors, including medication usage (e.g. statins) for other comorbidities (e.g. cardiovascular diseases), but these factors have not been considered in the current study due to the unavailable information. As circulating A-FABP activates STAT3/ALDH1 pathway to enhance breast tumor stemness, we don’t exclude that other A-FABP-mediated pathways (e.g. MAPK) may be involved for breast cancer progression (Guaita-Esteruelas et al., 2017). Of note, obesity can increase circulating A-FABP levels higher than 100ng/ml in both mouse and human samples. We chose the dose of A-FABP at 100 or 150ng/ml in our studies, but other doses higher than physiological ranges (20–30ng/ml) are also expected to get similar results. In addition, we noticed that A-FABP-binding mammary tumor cells are not correlated with ALDH1 positivity in these cells during flow cytometric analysis (data not shown), suggesting a crosstalk between different subsets of tumor cells in response to elevated levels of circulating A-FABP during obesity, which warrants further investigation in follow-up studies.

STAR METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for reagents may be directed to and will be fulfilled by the Lead Contact: Bing Li (b.li@louisville.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Human samples

The serum samples from patients with or without breast cancer were collected in a double-blind fashion. All patients were given informed consent under an Institutional Review Board-approved protocol (IRB No.15.0069). The human samples were classified into breast cancer and non-cancer groups according to their clinical diagnosis. Based on the body mass index (BMI), they were separated into non-obese (BMI≤30) or obese (BMI>30) groups. Details of patient information, and human sample adipokine array and prognostic survival analysis are shown in the supplemental tables and figures.

Mice

The mouse studies were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Louisville (IACUC No. 15113). Wild type, A-FABP-/- and E-FABP-/- mice, MMTV-TGFα/WT and MMTV-TGFα/A-FABP-/- mice (all C57BL/6 background) were bred and housed in the animal facility at the University of Louisville. Female mice were used in our experiments as breast cancer predominantly occurs in women. For syngeneic mouse models, wild type, A-FABP-/- and E-FABP-/- mice (C57BL/6 background) were bred and house in the animal facility in the University of Louisville. All the mice were fed with either chow diet, low fat diet (LFD, 10% fat) or high fat diet (HFD, 60% fat) (Research Diets, Inc.) for more than 6 months. For transgenic mouse models, MMTV-TGFα mice that overexpress human TGFα were crossed with A-FABP-/- mice to generate MMTV-TGFα/WT and MMTV-TGFα/A-FABP-/- mice. Starting from at least 4 weeks old, MMTV-TGFα/WT and MMTV-TGFα/A-FABP-/- female mice were fed on custom LFD (10% fat, Research Diets, Inc.) and HFD (45% fat, Research Diets, Inc.), respectively. Mouse body weight as well as breast tumor development was monitored biweekly. Mice without skin lesions were euthanized around 14 months on the HFD when mammary tumors were developed.

Cell culture experiments

Primary adipocytes and macrophages were isolated from mouse adipose tissues, and cultured for less than 48 hours for in vitro analysis. All tumor cell lines used in this study were purchased from the ATCC or CH3 BioSystems. In vitro tumor cell culture was less than 6 months for all experiments. As serum levels of A-FABP in obese human and HFD-induced obese mice ranged between 50–150ng/ml, we treated tumor cells in vitro with 100–150ng/ml of recombinant mouse or human A-FABP. See detailed methods of tumor cell treated with FABPs and FAs, and assays for analyzing tumor invasive phenotype in the supplementary information.

METHODS DETAILS

Analysis of ALDH1 activity by ALDEFLUOR Assay

ALDEFLUOR kit (Cat. #01700, Stem Cell Technologies) was used to detect ALDH1 enzymatic activity in mammary tumor cell lines and in primary tumors. Briefly, primary tumors were dispersed into single cell suspensions, stained with 2µl Aldefluor reagent per 0.5×10⁶ cells followed by incubation at 37°C for 45min and quantified by flow cytometry. In each experiment, the specific ALDH1 inhibitor diethylaminobenzaldehyde (DEAB) was used as a negative control for gating on flow cytometric plots.

Inhibition of STAT3 with siRNAs or chemical inhibitors

For endogenous STAT3 knockdown assays, tumor cells (e.g. MCF-7 cells and E0771 cells) were transfected with designated 50nM siRNA oligos (purchased from Integrated DNA technologies and Sigma-Aldrich) in combination of Oligofectamine and Opti-MEM I reduced serum medium (Cat. #51985034) for 12 hours and then replaced with the RP10 medium. For chemical inhibition of STAT3, S3I-201 or NSC74859, specific STAT3 inhibitors, were added in the culture medium at doses from 0 to 25µM for 12 hours to inhibit STAT3 activation in tumor cells.

Quantitative Real-time PCR

Cells were collected for RNA isolation using a PureLink RNA Mini Kit (Cat. #12183025) from Ambion (Life technologies). Complementary DNA synthesis was performed with a QuantiTect Reverse Transcription Kit (Cat. #205314) from Qiagen. Quantitative PCR was performed with Power SYBR Green PCR Master Mix (Cat. #4368708) using a StepOnePlusTM Real-time PCR System (Applied Biosystems). Detailed primer and siRNA sequences are shown in Supplementary Table S7.

Flow Cytometric Analysis

Recombinant A-FABP or E-FABP was labeled using an Alexa Fluor 488 or Alexa Fluor 647 microscale protein labeling kit (Cat. #A30006, #A30009,ThermoFisher Scientific) according to the manufacturer’s instruction. Tumor cells were surface stained for 30 min at 4°C with fluorescent-labeled A-FABP or E-FABP for analysis of tumor/FABP binding. For the cell apoptosis assay cells were stained with Annexin V and 7-AAD. Samples were acquired on a BD LSR Fortessa. Data were analyzed by DIVA software (BD, Biosciences).

Western Blotting

Western blotting was performed to measure A-FABP-induced signaling inside tumor cells. Protein concentration was determined using a BCA assay (Cat. #23225, Thermo Scientific). GAPDH or β-actin was used as a loading control. An Image Quant TL system was used to determine the relative protein quantification.

Microarray analysis

A total of 1×10⁶ E0771 cells were treated with 100ng/ml recombinant mouse A-FABP for 24 hours. Total RNA was extracted from treated E0771 cells and controls using PureLink RNA Mini Kit as mentioned above and subjected to mouse mRNA Gene expression analysis by Affymetrix microarray. Microarray analysis was performed using Partek Genomics Suite 6.6 and transcripts were normalized on a gene level using RMA as normalization and background correction method.

Lipid-Protein Interaction

To detect lipid-A-FABP interactions, we used FlowPIPs which are microparticles coated with different types of membrane lipids that are commonly used to determine lipid-protein interactions in many studies (Nair et al., 2016; Nishikimi et al., 2009). Multiplex Lipids Beads (Cat # P-FB01A and P-FB01B) were purchased from Echelon Biosciences. Briefly, we prepared 10µl of the FlowPIPs beads and mixed with or without A-FABP (20ng) in buffer I with the final volume of 200µl for 1 hour at RT. After washing, 200µl of 0.1µg/ml of recombinant GST-tagged PLCδ protein was added to the mixed beads for 1 hour. The mixtures were then incubated with the secondary anti-GST antibody (conjugated with Alexa Fluor 488 at the dilution of 1:1000 in Buffer I) for 1 hour after washing with buffer II for 3 times. Flow cytometry was used for the analysis of lipid-A-FABP interactions using FITC and PE channels.

Measurement of FFAs and cytokines by ELISA

ELISA kits for mouse A-FABP (Cat. #CY-8077) and E-FABP (Cat. #CY-8056) were from MBL. ELISA kits for mouse IL-1β (Cat. #432601), IL-6 (Cat. #431301), IL-10 (Cat. #431414) and TNFα (Cat. #430901) were from Biolegend. Estradiol ELISA kit was from Biovision (Cat.#K3830). ELISA kits for human A-FABP (Cat. #A05181) and E-FABP (Cat. #MBS2501304) were from Bertin Corp and MyBiosource, respectively. EnzyChromTM FFA assay kit (Cat. #EFFA-100) for FFA measurement was from BioAssay Systems. All cytokines, soluble proteins and FFAs in mouse or human serum were measured according to the manufacturer’s instruction.

Human adipokine array

Human adipokine array kit (Cat. #ARY024, R&D systems) was used to detect up to 58 human adipokines in serum from breast cancer patients and normal subjects. Antibody-loaded membranes were incubated with 80µl serum and analyzed according to the manufacturer’s instruction at the same condition. The images were captured and acquired using Azure C300 Ultimate Western Blot Imaging System. Data was transferred by Image J software, analyzed by subtracting the average background signal from each spot and finally compared to the average of reference spots on different arrays to determine the relative levels of interested proteins.

Prognostic survival analyses

For analyzing the associations of ALDH1, A-FABP and E-FABP with the overall survival of breast cancer patients, we performed the analyses using PROGgene, an online tool to investigate prognostic implications of interesting genes in cancers. We used the GSE19783-GPL6480 breast cancer datasets and created the overall survival curve by providing the protein name, cancer type, survival measurements, bifurcate gene expression, etc.

Tumor challenges

To monitor the tumor progression in lean and obese mice, 5×10⁵ E0771 cells were orthotopically implanted into the mammary fat pad of lean or obese WT and A-FABP-/- mice, respectively. For A-FABP inhibition, A-FABP selective inhibitor BMS309403 was given at 5mg/kg intraperitoneally 12 hours prior E0771cells implantation, followed by injections at 3-day intervals until mice were sacrificed (Yan et al., 2018). Tumors were measured with calipers and the volume was calculated by the formula 0.5× (large diameter) × (small diameter)². For the in vivo E0771 limiting dilution assay (LDA), E0771 cells were treated with 100ng/ml mouse A-FABP and control PBS for 24 hours. A-FABP-treated or untreated E0771 cells were orthotopically injected in WT mice at the indicated cell numbers from 500,000 to 50,000 per mouse. ELDA software (http://bioinf.wehi.edu.au/software/elda/index.html) was used to calculate the cancer initiating cell frequency by scoring the number of tumors formed out of the number of injection sites (Hu and Smyth, 2009). For transgenic mouse models, tumor and mammary tissue were collected for hematoxylin and eosin (H&E) staining and histopathology analysis. Serum and lymphoid organs were collected for biological analyses. To detect the ALDH1 activity in vivo, single cell from solid tumors or mammary tissues were prepared by digestion in an enzyme mixture (0.5 mg/mL collagenase type 2, 0.2 mg/mL hyaluronidase, and 0.02 mg/mL DNase in RPMI-1640) at 37°C for 45 minutes. The separated cells were washed for Aldefluor assay.

Tumor colony formation assays

To analyze soluble FABP-induced tumor invasiveness in vitro, various types of murine or human tumor cell lines were treated with mouse or human FABPs, respectively. For the colony formation assay, 2×10³ tumor cells (e.g. E0771, MCF-7, BT474 cells) were seeded into 6-well plates and treated with 100ng/ml A-FABP or E-FABP to achieve a range of 10–100 colonies. After 3 weeks of incubation for colony formation, the plates were washed with PBS and the colonies were fixed using 10% neutral buffered formalin solution for 30min followed staining with 0.01% (w/v) crystal violet for 1 hr. Excess crystal violet in the plates was washed with PBS and colonies containing more than 50 individual cells were counted under a microscopy.

Tumor migration and proliferation assays

For the wound-healing migration assay, tumor cells were grown to confluence in a 6-well plate. The linear wound of cellular monolayer was created by scratching confluent cell monolayer using a 200µl plastic pipette tip. The scratched cell monolayer was washed by PBS to remove debris. After incubation at 37°C for 24–48 hours, the migration of the cells towards the wound was photographed under a light microscopy. Image J was used to determine the migration distance. For the proliferation assay, E0771 and MCF-7 cells were seeded into 96-well plates and cultured with 100ng/ml recombinant A-FABP or E-FABP protein for 48 hours. Cells were then added 10µl MTT solution (5mg/ml) and incubated at 37°C for 4 hours. The OD values were measured at 490 nm using a microtiter plate reader.

Tumor invasion assay

Tumor cell invasion was assessed using a three-dimensional (3D) spheroid invasion assay (Berens et al., 2015). E0771 and MCF-7 cells formed spheres in hanging drops of culture medium on the lid of cell culture dishes (approximate 500 cells per drop). After 48 hours, spheres from the lid were aliquoted into the same volume, mixed with rat tail type I collagen (final concentration is 1.7mg/ml) and embedded in wells to generate a 3D culture system. Mouse and human recombinant A-FABP (100ng/ml) were added in the cell culture medium of E0771 and MCF-7, respectively. Invasion was concluded at 48 hours for E0771 cells and 7 days for MCF-7 cells. Quantitative analyses were determined by measuring the maximal invasive distance (longest invasive distance-spheroid radium) and invaded area (total invaded area-spheroid area) using the Image J software.

Analysis of tumor side population (SP) and sphere formation

To acquire the SP and non-SP cells, the experiments were performed using the method as previously described (Golebiewska et al., 2011). Cells were stained at 1×10⁶ cells/ml in pre-warmed medium with Hoechst33342 (5 µg/ml) in the presence or absence of the ABC transporter inhibitor Verapamil-HCL (100 mM) for 90 min at 37°C. Cells were analyzed using UV laser on BD LSR Fortessa with Indo-1-violet and Indo-1-blue which was compatible with the typical filter set consisting of a blue BP (450/50) and a red BP (670/14), respectively, and a dichroic mirror to separate the blue and red filters (575LP). For the sphere assay, 4×10³ single cell suspension was plated into Costar Ultra-low attachment surface plate (Cat. #29443–030, VWR) with MammoCult basal medium with or without A-FABP (100ng/ml). Spheres (>50µm) were counted after 6 days of incubation. For the second sphere formation assay, cells from the primary spheres were treated with pre-warmed Trypsin-EDTA to acquire single-cell suspension. Cells were either plated in 24-well ultra-low adherence plates, or seeded into a 96-well plate in a 2-fold serial dilution from 1000 to 8 cells per well, with or without A-FABP treatment (100ng/ml) for sphere formation. The number of positive cells was identified and used to calculate the frequency of sphere-forming cells using the ELDA software (Hu and Smyth, 2009).

Confocal microscopy

Mammary tissues obtained from MMTV-TGFα/WT and KO mice were snap-frozen in cryo-embedding media OCT (Sakura Finetechnical Co., Ltd) and cut to 6µM sections for staining. After fixation and permeabilization, the sections were blocked with 5% BSA and stained with phospho-STAT3 antibody (Cell Signaling Technology, Cat# 9145S) overnight at 4°C. After 20 min of staining with second antibody (APC-labeled goat anti-rabbit IgG, Jackson Immunoresearch, Cat# 111–136-144) and anti-CD44 antibody, the sections were stained with 0.2µm DAPI (Invitrogen) for nuclei. Confocal analysis was done with Nikon Eclipse TE2000 confocal microscope.

Tumor cells treated with FAs

For the observation of synergistic effects of A-FABP and FAs, various saturated and unsaturated FAs were purchased from Nu-Chek Prep and conjugated with endotoxin free-BSA according to our previous studies (Zhang et al., 2014; Zhang et al., 2017). Various FAs were added in the culture medium to observe if they induced tumor cell membrane lipid exposure and A-FABP recognition and ALDH1activation by flow cytometric analyses.

QUANTIFICATION AND STATISTICAL ANALYSIS

Linear mixed effect models were used to examine the relationship between A-FABP and BMI or change in BMI over time. Two-sided t test was applied to compare the measurements of two groups. However, if the data was not normally distributed, the Mann-Whitney test (two-sided) was applied for group comparisons. For microarray analysis, one-way ANOVA was set up to compare A-FABP-treated and untreated groups. P values less than 0.05 were considered statistically significant.

DATA AVAILABILITY

Microarray data were submitted to the Gene Expression Omnibus (accession number GSE102975).

Highlights.

1. Circulating levels of A-FABP are elevated in obesity-associated breast cancer

2. Soluble A-FABP promotes mammary tumor stemness and aggressiveness

3. A-FABP-induced ALDH1 activity is dependent on IL-6/STAT3 signaling

4. A-FABP deficiency uncouples HFD-induced obesity and mammary tumor development

eTOC Blurb.

Hao et al. identified that circulating A-FABP elevated in obesity promoted mammary tumor stemness and development through IL-6/STAT3/ALDH1 axis. Thus, circulating A-FABP represents a novel mechanism linking obesity and the increased risk of breast cancer.