Transcriptome analysis reveals a role for the endothelial ANP-GC-A signaling in interfering with pre-metastatic niche formation by solid cancers.

Cancer establishes a microenvironment called the pre-metastatic niche in distant organs where disseminated cancer cells can efficiently metastasize. Pre-metastatic niche formation requires various genetic factors. Previous studies suggest that inhibiting a single niche-factor is insufficient to completely block pre-metastatic niche formation especially in human patients. Here we show that the atrial natriuretic peptide (ANP), an endogenous hormone produced by the heart, inhibits pre-metastatic niche formation and metastasis of murine solid cancer models when pharmacologically supplied in vivo. On the basis of a wealth of comprehensive RNA-seq data, we demonstrated that ANP globally suppressed expression of cancer-induced genes including known niche-factors in the lung. The lungs of mice overexpressing GC-A, a receptor for ANP in endothelial cells, were conferred resistance against pre-metastatic niche formation. Importantly, neither ANP administration nor GC-A overexpression had a detrimental effect on lung gene expression in a cancer-free condition. The current study establishes endothelial ANP-GC-A signaling as a therapeutic target to control the pre-metastatic niche.

INTRODUCTION

Solid cancer is an abnormal organ that affects systemic host physiology in various ways [1]. Such alteration is dependent on the host’s physiological conditions, cancer locations, genetic and epigenetic backgrounds of cancers, and so on. Mechanisms and physiological importance of cancer-affected host physiology in vivo is hard to dissect due to its complex nature: both host and cancer states are highly dynamic. Specifically, solid cancer promotes its status in a step-wise manner in vivo, breaking through the basement membranes to disseminate into the blood and/or lymphatic stream, invading distant tissues for colonization, and metastasizing to form secondary cancers [1-4]. Unfortunately, metastatic diseases are still hard to cure. Therefore, understanding how solid cancer modulates host physiology for successful metastasis is essential to devise therapeutic strategies to suppress it.

Primary cancer “prepares” distant organs to be more preferable landing sites for disseminated cancer cells prior to their actual arrival, known as the pre-metastatic niche hypothesis [1-4]. Although the nature of pre-metastatic niches is still incompletely characterized, there are several key factors that define them. Importantly, it has been noted that the pre-metastatic niche and cancer-induced inflammation harbor a close relationship. As an example, Hiratsuka et al. showed that, using transplantation models of mouse melanoma and lung carcinoma, cancer-derived TNFα, TGFβ, and VEGF-A cooperatively recruit CD11b+ myeloid cells, causing inflammation [5, 6]. Cancer-induced inflammation is positively correlated with successful metastases in the lung, and thus considered as the pre-metastatic niche [5]. Moreover, elevated expression of a matrix remodeler MMP9 both in resident endothelial cells and recruited myeloid cells is also associated with metastasis efficiency [6]. These processes require VEGFR-1, a receptor for VEGF-A and Toll-like receptor 4 (TLR4) [5, 6]. Other genetic factors MD-2 (a co-receptor for TLR4), C-C chemokine receptor type 2 (CCR2) and its ligand CCL2 have also been demonstrated to be essential for pre-metastatic niche formation [7]. As another example for metastasis-related chemokine pathways, CXCR2 is an important factor for efficient metastasis in a mouse model of pancreatic ductal adenocarcinoma [8]: CXCR2 inhibition recruits T cells to alter a cancer-associated niche and enhances T cell dependent-suppression on cancer cells [8]. Recent study indicates that cancer-derived exosomal RNAs activate innate immune responses via TLR3 in alveolar epithelial cells to develop the lung pre-metastatic niche [9]. Cancer exosomes also contribute to organ tropism of metastasis [10, 11]. These studies suggest that immune cells, and resident cells such as endothelial and alveolar epithelial cells are likely to be a crucial cell type for pre-metastatic niche formation [4]. In these respects, it is again of note that the pre-metastatic niche and innate immune responses share genetic factors with each other, suggesting an extensive overlap between the pre-metastatic niche and cancer-induced inflammation.

Various clinical trials have been on-going but none have resulted in successfully controlling metastasis by targeting individual niche-factor(s) such as TGFβ and MMPs [12-15]. These indicate that targeting multiple factors together may be essential to efficiently control the pre-metastatic niche. Furthermore, side effects caused by targeting these factors are not negligible, since these factors are also critical for normal biological processes. From this standpoint, directly targeting multiple niche factors may increase the risk for serious side effects.

Atrial natriuretic peptide (ANP) is a hormone endogenously produced from atrial cells in the heart [16]. Major biological roles of ANP are to promote diuresis, to reduce central blood pressure and to decrease cardiac hypertrophy [17]. ANP functions through binding to its receptor guanylyl cyclase-A (GC-A) [18]. Clinically, ANP is utilized to treat heart failure patients in Japan [19]. In addition, clinical studies demonstrate a beneficial effect of ANP treatment in acute lung injury and postoperative cardiopulmonary complications in lung cancer patients [20, 21]. A retrospective study suggests that ANP treatment may reduce the frequency of post-operative lung cancer-recurrence in lung cancer patients who had a surgical resection [22]. The study also showed that ANP prevents cancer metastasis in the mouse model of melanoma [22]. We proposed that ANP inhibits adhesion of cancer cells to the host endothelium through suppressing inflammation-induced expression of E-selectin [22]. Therefore, inflammation leading to increased adhesive blood vessels may efficiently capture disseminated cancer cells, thereby promoting cancer metastasis. Accordingly, a prospective clinical study has launched to examine whether perioperative ANP administration is able to reduce post-operative cancer recurrence. However, whether these observations are true for other types of cancers and the mechanism underlying how ANP exerts its metastasis-preventing roles—which metastatic process(es) ANP suppresses—have been unanswered. Answering these questions is essential to extend clinical applications of ANP, and to improve quality of life of cancer patients.

Here, using different cancer transplantation models in mice, we report that the ANP-GC-A pathway suppresses pre-metastatic niche formation in the lung of BALB/c females bearing 4T1 breast cancer and of C57BL/6 males bearing Lewis Lung Carcinoma (LLC), suggesting a general role for ANP in metastasis inhibition. Comprehensive gene expression analyses revealed that ANP administration significantly repressed cancer-induced gene expression changes representative of pre-metastatic niche formation in the lung. Overexpression of GC-A in endothelial cells conferred resistance to the lung against pre-metastatic niche formation, thereby providing genetic evidence that endothelial cells mediate ANP-dependent suppression of the pre-metastatic niche. Activation of the ANP-GC-A pathway in cancer-free mice had subtle effects on lung gene expression, indicating that ANP does not affect homeostasis of the lung under a non-pathogenic condition, an observation in line with published clinical data in humans. Altogether, our study establishes the ANP-GC-A pathway as a target for drugging pre-metastatic niche formation by solid cancers.

RESULTS

ANP prevents lung metastasis in a mouse model of 4T1 breast cancer

To quantify the frequency of metastasis in the presence or absence of ANP, we utilized a transplantation model of mouse breast cancer 4T1 that expresses enhanced green fluorescent protein (EGFP) (4T1-EGFP). Cultured 4T1-EGFP cells were subcutaneously injected into the 1st recipient BALB/c mice that were later sacrificed to obtain cancer tissues at 7 days after injection. Obtained cancer tissues were cut into uniform-sized pieces (3 mm × 3 mm × 3 mm), and then transplanted into the mammary pad of 2nd recipients. Cancer size was determined on 10, 17, and 24 days post transplantation (dpt). The mice were sacrificed at 28 dpt, and the number of EGFP positive metastases to the lungs, one of the most frequent targets for metastasis, were examined by a clinical oncologist (T.N). Strikingly, ANP treatment significantly reduced the number of metastatic events (Figure 1A–1B). To extend our knowledge on the metastasis-suppressing activity of ANP beyond the 4T1 model, we performed similar experiments for colon26-EGFP cancer cells, demonstrating that ANP lowered the efficacy of metastasis in the colon26-EGFP model (Supplementary Figure 1A–1B). These results indicated that ANP has an activity to suppress metastasis of various solid cancers.

Figure 1

ANP suppresses lung metastasis of 4T1 breast cancer

(A) Representative images of lung metastasis in vehicle- or ANP-treated 4T1-bearing mice. Mice were sacrificed four weeks after cancer cell transplantation. Scale bars represent 10 mm. (B) Dot plot showing the number of nodules representing lung metastasis of 4T1-EGFP cells in mice grouped as in (A) (10 mice per a group). (C) Primary cancer volume in vehicle- or ANP-treated 4T1-bearing mice on day 10, 17, and 24 after cancer cell transplantation. Data are means ± s.e.m. (10 mice per a group). (D) Primary cancer weight in vehicle- or ANP-treated 4T1-bearing mice on 28 dpt (8 mice for vehicle and 10 for ANP).

Next we asked if ANP exhibits an anti-cancer activity to suppress metastasis. To this end, volume and weight of primary cancer tissues were determined at multiple time points. As shown in Figure 1C and 1D, neither volume nor weight of primary 4T1 cancer cells was significantly affected by ANP treatment. This was also the case for colon26-EGFP cancer cells (Supplementary Figure 1C). We further confirmed that ANP did not largely affect cancer cell viability in vitro (Supplementary Figure 2A). Thus, it appeared unlikely that ANP treatment affects growth and viability 4T1/colon26 cancer cells. To assess the possibility that ANP treatment influenced the nature of primary cancer cells, we examined Epithelial-to-Mesenchymal Transition (EMT) markers (E-cadherin and Vimentin) in breast-4T1 and colon26 cultures in the presence or absence of ANP. Our data showed that ANP had no impact on protein expression of EMT markers in 4T1 and colon26 cancer cells (Supplementary Figure 2B). Moreover, ANP treatment did not strongly disrupt formation of spheroids, indicating that Cancer-Stem-Cell (CSC) property was unaffected by ANP (Supplementary Figure 2C) [23, 24]. Altogether, we concluded that ANP prevents metastasis of 4T1 and colon26 cancers to the lungs, likely without affecting primary cancer cells.

ANP suppresses 4T1-induced gene expression changes representative of pre-metastatic niche formation in the lung

To know how ANP accomplished its anti-metastasis activity, we investigated the effects of ANP on lung gene expression in the presence or absence of 4T1 transplantation via RNA-seq. We treated control and 4T1-bearing mice with vehicle or ANP, and analyzed gene expression changes in the lung at 7 dpt, roughly 2 weeks before visible metastases are detected in our system (Supplementary Figure 3). We found that, in the lung of 4T1-bearing mice compared to those of controls, genes associated with the pre-metastatic niche and inflammation such as S100A8 were elevated (Figure 2A and Supplementary Figure 4). S100A8 is a calcium-binding protein that heterodimerizes with S100A9 and subsequently acts as an endogenous ligand for toll-like receptor-4 [25]. S100A8 and S100A9 are strongly expressed in neutrophils and play a crucial role in the inflammation response as well as in mediating a paracrine loop between cancer cells and neutrophils cells to enhance invasion of cancer cells to the surrounding microenvironment [26]. Thus, increased expression of S100A8/9 is an indicator for pre-metastatic niche formation and inflammation. Gene ontology analyses indeed confirmed that the elevated genes represented “leukocyte migration”, “neutrophil chemotaxis”, and “myeloid leukocyte migration” (Figure 2B). These observations were in accordance with the previous reports that solid cancer induces lung inflammation as the pre-metastatic niche [5-8].

Figure 2

ANP represses 4T1-induced gene expression changes representative of pre-metastatic niche formation in the lung

(A) Scatter plot showing log2 fold changes between the lungs of 4T1-bearing or sham-operated mice. Genes exhibiting more than 2-fold changes are highlighted. Data from two biological replicates are shown (rep1 and rep2). (B) Gene ontology analysis. Top 100 up-regulated genes were subjected to GO analysis and signatures highly significantly enriched in the group are shown. (C) Scatter plot comparing log2 fold changes between 4T1-bearing or sham-operated mice with or without ANP treatment. Genes exhibiting ≥ 3-fold or ≤ 0.3-fold changes in vehicle are highlighted. Data from two biological replicates are averaged. (D) Heatmap of genes exhibiting more than 3-fold increases in the lung of 4T1-bearing mice is shown. Gene expression changes of the indicated genes in ANP-treated sham group are also shown. Validation for (A)-(D) by qPCR and immunohistochemistry is presented in Supplementary Figure 4. (E) Representative images of the lungs stained with hematoxylin-eosin (HE) and anti-Mac3 antibody obtained 7 days after 4T1 cancer-transplantation in each group. Scale bar represents 100 µm. Higher resolution pictures are shown in Supplementary Figure 6. (F) Dot plot showing the number of Mac3–positive cells per a filed in mice grouped as in (E) (9 mice per a group except for 4T1-ANP (n = 8)). p-values were calculated using one-way ANOVA. Representative low-magnification images are shown in Supplementary Figure 5.

Notably, ANP suppressed most of 4T1-induced gene expression changes in the lungs of 4T1-bearing mice (Figure 2C–2D). Figure 2C and 2D demonstrated that genes up-regulated in the lung in 4T1-bearing mice were less elevated in the ANP-treated 4T1-bearing mice. This observation was validated extensively using immunohistochemistry and qPCR for S100A8/9, CXCR2, and MMP9 as known niche-factors (Supplementary Figure 4). Immunohistochemistry suggested that the number of cells positive for each marker was increased by 4T1-transplantation, which was mitigated by ANP treatment (Supplementary Figure 4A). Hence, increased expression of S100A8/9 probably reflects infiltration of S100A8/9-positive neutrophils, rather than gene expression changes in lung cells. To further confirm that ANP suppressed the pre-metastatic niche/cancer-induced inflammation, we performed histological analyses with anti-Mac3 antibody, a marker for macrophages [27]. Considerable macrophage infiltration was observed in the lungs of 4T1-bearing mice, which was ameliorated by ANP treatment (Figure 2E–2F and Supplementary Figures 5–6). ANP did not affect expression of 4T1-induced genes in the normal lung (Figure 2D, Supplementary Figure 4 and Supplementary Figure 7A–7B). Instead, ANP treatment in the absence of 4T1 elevated expression of non-inflammatory genes such as Mup3 that did not respond to 4T1 transplantation (Supplementary Figure 7A–7B). Furthermore, we noted that ANP treatment elevated expression of the endogenous ANP gene only in the co-presence of 4T1 cancer cells and exogenous ANP (Supplementary Figure 7C). These collectively suggested that ANP functions against 4T1-induced pre-metastatic niche formation in the lung.

In addition to S100A8/9, CXCR2, and MMP9, we found a number of genes whose cancer-induced up-regulation was buffered by ANP treatment. It is currently unclear whether these other differentially expressed genes are (1) a genuine niche-factor and/or (2) an inflammatory response factor or (3) a factor involved in non-niche/non-inflammation phenomenon in the lung. Clarifying these possibilities awaits further extensive genetic investigation.

ANP suppresses LLC-induced gene expression changes representative of pre-metastatic niche formation in the lung

Next, we wished to know whether ANP is able to suppress pre-metastatic niche formation in the Lewis Lung Carcinoma-EGFP (LLC-EGFP) model using C57BL/6 male mice. It should be noted that LLC-EGFP model is less metastatic in our hands (data not shown). As was the case for breast-4T1 and colon26 models, viability and CSC properties of LLC cells were not strongly affected by ANP treatment (Supplementary Figure 8). We analyzed lung gene expression at 10 dpt, finding that expression of inflammation/pre-metastatic niche markers was elevated in the lung of LLC-bearing mice (Figure 3). Gene set enrichment analyses (GSEA) demonstrated that differentially expressed genes in the lung of the LLC- and 4T1-bearing mice were quite similar to one another (Figure 3A); genes up-regulated in 4T1-bering mice tended to be also up-regulated in LLC-bearing mice, and vice versa. For instance, known niche factors such as S100A8/9 were up-regulated both in the lungs of 4T1- and LLC-bearing mice (Figures 2–3). Similar to the 4T1 model (Figure 2), ANP buffered LLC-induced gene expression changes in the lung (Figure 3B–3C). All genes showing more than a 3-fold increase in the lung of LLC-bearing mice were less elevated in the lung of ANP-treated LLC-bearing mice (Figure 3B–3C). Cancer-induced increases and ANP-dependent decreases for S100A8/9, CXCR2, and MMP9 were confirmed by immunohitochemistry and qPCR (Supplementary Figure 9). Consistent with these, our immuno-histological analysis showed that ANP suppressed infiltration of Mac3 positive cells in the lung (Figure 3D–3E, and Supplementary Figure 10–11) Moreover, the lungs of C57BL/6 males appeared to be largely unresponsive to ANP in the absence of LLC at the mRNA level (Figure 3C and Supplementary Figure 9 and 12). Combined with Figure 2, our data indicate that ANP inhibited pre-metastatic niche formation of two different types of cancers, and that the activity was exerted in a context-dependent manner.

Figure 3

ANP suppresses LLC-induced gene expression changes representative of pre-metastatic niche formation in the lung

(A) Gene set enrichment analysis comparing 4T1- and LLC-induced gene expression changes in the lung. (B) Scatter plot comparing log2 fold changes between LLC-bearing or sham-operated mice with or without ANP treatment. Genes exhibiting ≥ 3-fold or ≤ 0.3-fold changes in vehicle are highlighted. Data from two biological replicates are averaged. (C) Heatmap of genes exhibiting more than 3-fold increases in the lung of LLC-bearing mice is shown. Gene expression changes of the indicated genes in the ANP-treated sham group are also shown. Validation for (A)-(C) by qPCR and immunohistochemistry is presented in Supplementary Figure 9. (D) Representative images of the lungs stained with HE or anti-Mac3 antibody obtained 10 days after LLC cancer-transplantation in each group. Scale bars represent 100 µm. Higher resolution pictures are shown in Supplementary Figure 11. (E) Dot plot showing the number of Mac3–positive cells per a filed in mice grouped as in (D) (9 mice per a group except for LLC-ANP (n = 8)). p-values were calculated using one-way ANOVA. Representative low-magnification images are shown in Supplementary Figure 10.

The endothelial ANP-GC-A signaling prevents pre-metastatic niche formation

ANP acts through a receptor called guanylyl cyclase-A (GC-A) [18]. To investigate whether the pre-metastatic niche-suppressing role of ANP depends on the canonical ANP-GC-A signaling, we utilized mice overexpressing GC-A specifically in endothelial cells (termed EC GC-A-Tg mice) [17, 18, 22]. We previously showed that EC GC-A-Tg is able to mimic pharmacological ANP administration [22, 28] and therefore this model enables us to explore whether ANP attenuates pre-metastatic niche formation via interacting with GC-A in endothelial cells.

Littermate mice with or without GC-A overexpression in endothelial cells were given LLC cancer tissues, and sacrificed at 10 dpt for RNA-seq analyses (Figure 4). Figure 4A demonstrated that the lung of EC GC-A-Tg mice was clearly resistant to LLC-induced gene expression changes compared to the littermate controls (Figure 4A). As an example, expression of S100A8 was much less induced in the lung of EC GC-A-Tg mice (Figure 4B and Supplementary Figure 13). More than 95% of ANP-suppressed genes in the LLC-bearing mice were down-regulated in the lung of LLC-bearing EC GC-A-Tg mice, indicating that ANP inhibited pre-metastatic niche formation through endothelial GC-A (Figure 4C). The effects of EC GC-A-Tg on the niche factors appeared stronger than pharmacological ANP treatment, suggesting that the amount of GC-A proteins was rate-limiting in endothelial ANP-GC-A signaling. These results were further validated by histological analyses and qPCR, showing that EC GC-A-Tg suppressed the pre-metastatic niche (Supplementary Figure 13). Similar to ANP treatment, endothelial GC-A overexpression had only a minor impact on lung gene expression (Figures 4B and Supplementary Figure 14). Indeed, we identified only one statistically significant gene with more than a 3-fold change in the lung of EC GC-A-Tg mice: GC-A (Npr1) itself (up-regulated) (Supplementary Figure 14).

Figure 4

The endothelial GC-A overexpression attenuates pre-metastatic niche formation in the lung

(A) Scatter plot comparing log2 fold changes between LLC-bearing or sham-operated mice in WT and EC GC-A-Tg. Genes exhibiting ≥ 3-fold or ≤ 0.3-fold changes are highlighted. Data from two biological replicates are averaged. (B) Heatmap of genes exhibiting more than 3-fold increases in the lung of LLC-bearing WT mice is shown. Gene expression changes of the indicated genes between WT and EC GC-A-Tg (sham-operated) are also shown. (C) Venn diagram demonstrates extensive overlap between ANP- or GC-A-regulated genes. Validation for (A)-(C) by qPCR and immunohistochemistry is presented in Supplementary Figure 13. (D) Representative images of the lungs stained with HE or anti-Mac3 antibody obtained 10 days after LLC tissue-transplantation in each group. Scale bars represent 100 µm. Higher resolution pictures are shown in Supplementary Figure 16. (E) Dot plot showing the number of Mac3–positive cells per a filed in mice grouped as in (D) (n = 9 for the WT group and n = 8 for the EC GC-A Tg group). p-values were calculated using one-way ANOVA. Representative low-magnification images are shown in Supplementary Figure 15.

Altogether, these results suggest that the endothelial ANP-GC-A signaling suppresses pre-metastatic niche formation but does not affect normal lung/leukocyte homeostasis in the healthy lung (Figures 4D and Supplementary Figure 15–16). Our data (Figures 2–4) were also consistent with previous work demonstrating that ANP does not have a role in reducing the number of leukocytes in vivo [29]. In summary, we have shown that endothelial ANP-GC-A signaling inhibits cancer-induced inflammation in the lung in a context-dependent manner, establishing this pathway as a target for drugging pre-metastatic niches by solid cancers (Figure 5).

Figure 5

Summary of this study

The ANP-GC-A signaling inhibits adhesion of cancer cells in blood vessels [22] and the pre-metastatic niche (this study) to protect different murine cancer models from metastasis. Although the detailed mechanism remains unknown, the ANP-GC-A signaling appears to simultaneously repress multiple cancer-induced signals (TGFβ, inflammation, and so on).

DISCUSSION

Since metastasis is a leading cause of cancer-related mortality [1-3], drugging metastasis is an important challenge in the field. Targeting the pre-metastatic niche, an important factor that enables disseminated cancer cells to colonize distant organs, will be useful for developing anti-metastasis therapies. Here, through comprehensive characterization of gene expression in the lung, a major target of cancer metastasis, we show that activation of the endothelial ANP-GC-A signaling is suppressive on pre-metastatic niche formation, protecting our mouse cancer models from pre-metastatic niche formation in the lung and lung metastasis. Although it is clear that ANP-GC-A signaling is potent in suppressing the pre-metastatic niche, at this point, we do not exclude the possibility that ANP inhibits metastasis also via a non-niche pathway. As discussed in more detail below and summarized in Figure 5, our study is unique and significant in that we have shown that (i) the ANP-GC-A signaling prevents cancer metastasis in two different models (Figure 1 and Supplementary Figure 1), (ii) a role for ANP in globally inhibiting up-regulation of pre-metastatic niche factors is context-specific: no hazardous effects on lung gene expression in the absence of cancer (Figures 2–4), (iii) the host ANP-GC-A signaling in endothelial cells is genetically essential for antagonizing pre-metastatic niche formation in the lung (Figure 4). With respect to these points, ANP is currently under a large-scale prospective clinical trial testing its role in improving post-operative prognosis of lung cancer patients in Japan led by multiple clinical teams including us.

Targeting metastasis has been technically challenging simply because metastasis is a dynamic and multi-step phenomenon, and thus treatment needs to be timely. As Steeg pointed out [3], in most cancer patients when diagnosed, cancer cells may have already been disseminated. In this case, targeting the “seeding” step would result in no benefits. In contrast, targeting the pre-metastatic niche appears to be more promising, as it may prevent further arrival and colonization of disseminated cancer cells. In addition, there are extensive overlaps between the pre-metastatic niche and the metastatic niche (the lung microenvironment after the arrival of disseminated cancer cells) [2, 3]. According to these, the (pre-)metastatic niche could be a continuous requirement for cancer metastasis, leading to an idea that the (pre-)metastatic niche is a good therapeutic window in anti-metastasis therapy. Dormancy is also a strong risk factor for recurrent metastasis, especially for breast and prostate cancers [1-3]. It is a recognized phenomenon that a colonized cancer cell somehow waits for more than years to form metastatic foci [1-3]. To keep dormant cancer cells silent, maintaining the body’s resistance to secondary cancer growth is essential. For this purpose, it is desirable to develop an anti-metastasis drug that works in a context dependent manner (i.e. do nothing in the absence of cancer). ANP may meet the above-described criterion as an anti-metastasis reagent.

The previous clinical studies have tried targeting the pre-metastatic niche by directly drugging niche-factors such as TGFb and MMPs [12-15]. However, it turned out that targeting single factors was insufficient to obtain satisfactory clinical outcomes. Such clinical evidences underscore the need to control the pre-metastatic niche using multiple drugs to target niche-factors simultaneously. However, given the numerous important biological functions of these niche factors in a healthy state, simultaneous inhibition of these factors can lead to serious side effects. In fact, undesirable effects by targeting MMPs have already been described [12]. In contrast, utilizing ANP to pharmacologically activate the ANP-GC-A pathway is unlikely to result in toxicity to organisms since activation of the ANP-GC-A pathway did not have an impact on lung gene expression in the absence of cancer cells (Figures 2–4). In addition, ANP has been used mainly for heart failure patients for decades in Japan without serious adverse effects [20, 21, 28, 30], suggesting that clinical use of ANP has been established to be relatively safe. Unveiling how the ANP-GC-A signaling exerts its function in inhibiting the pre-metastatic niche formation will provide an opportunity to better understand the context-dependency of targeting metastasis within this pathway. Furthermore, we, and others, have already shown that ANP buffers unwanted toxicity caused by cisplatin, an anti-cancer drug, and lipopolysaccharide (LPS)-induced inflammation [29, 31]. Although the underlying mechanisms are still under investigation, these observations indicate that ANP is potentially a general anti-inflammatory reagent. Overall, we suggest that ANP is a unique hormone that exhibits its “stress-buffering” activity specifically under pathological conditions and consequently suggests the ANP-GC-A pathway as a promising drug target.

Either ANP administration or GC-A overexpression potently buffered pre-metastatic niche formation without affecting expression of GC-A (Figures 2–4). Considering this fact, the amount and/or concentration of ANP is most likely in excess to GC-A proteins in endothelial cells. This could justify screening for small molecules that increase expression of GC-A proteins. Finding an agonist for the ANP-GC-A pathway may also be useful. We at this point do not exclude the possibility that ANP administration/GC-A overexpression changes the distribution of ANP to elevate the net activity of the endothelial ANP-GC-A signaling. The increase of GC-A proteins by itself could alter the quality of the signaling too.

Endothelial cells are an important lineage of cells for the pre-metastatic niche. Many types of cancer cells produce VEGF-A to modify endothelial cells, affecting vascular integrity and leakiness [1, 5, 6, 32]. VEGF-A overexpression is problematic for instance in that disrupted vascular integrity prevents efficient delivery of anti-cancer drugs into cancer tissues. Bevacizmab, a monoclonal antibody for VEGF-A, is an FDA-approved drug for multiple types of cancers [3, 33]. Importantly, ANP is able to antagonize VEGF-A-dependent vascular leakage [34], which could be an up-stream mechanism that the ANP-GC-A signaling inhibits pre-metastatic niche formation (Figure 5).

A set of genetic requirements for pre-metastatic niche formation is known: mice deficient for TLR4, MMP9, or CXCR2, are conferred resistance to cancer metastasis [5, 6, 8]. Neutralizing S100A9 or SAA3 also resulted in decreased metastasis in the lung [5]. However, since these genes are not endothelial-specific, the contribution of these genes in endothelial cells to pre-metastatic niche formation still remains unclear. A recent study demonstrated that endothelial cell-specific overexpression of receptor activity modifying protein 2 (RAMP2), a receptor for a hormone adrenomedullin, improves vascular integrity and is protective against metastasis of sarcoma and melanoma in mice [35]. We also previously showed that ANP inhibits LPS-induced inflammation through down-regulation of E-selectin, resulting in reduced metastasis of melanoma cells [22]. These findings, combined with our present study highlight the close relationship between cancer metastasis and endothelial cells as a pre-metastatic niche.

There are remaining questions to be addressed in the future. Does ANP affect gene expression in inflammatory cells? Does ANP affect, rather than gene expression, behavior (e.g. migration) of inflammatory cells? Which cell type is the most important in explaining a role for ANP in suppressing cancer metastasis (endothelial cells and/or inflammatory cells and/or others?)? Answering these questions is essential to establish ANP as an anti-metastasis drug. We expect that targeting altered host physiology could be an effective and generalizable way to fight cancers and to improve patient’s quality of life.

MATERIALS AND METHODS

Mice

Six-week–old female BALB/c and male C57BL/6 mice were purchased from Japan SLC, Inc. (Hamamatsu, Japan). Endothelium-specific (Tie2)-GC-A-transgenic mice have been described previously [22]. The information about mice strain and treatment were blinded to the investigators. No animals were excluded from the analysis. All animal experiments were performed according to the protocol approved by the Animal Care Ethics Committee of the National Cerebral and Cardiovascular Center Research Institute.

Cell culture

A mouse breast cancer cell line 4T1-EGFP and a lung cancer cell line LLC-EGFP were obtained from Wako (Osaka, Japan). A mouse colon cancer cell line, colon26-EGFP, was kindly provided by Dr. M.M. Taketo and Dr. M. Sonoshita (Kyoto University) [36]. All these cell lines were maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin streptomycin. All cells were tested for mycoplasma contamination.

Cancer transplantation and ANP treatment

Cultured 4T1-EGFP cells were subcutaneously injected into the right flank of 1st recipient BALB/c mice (1 × 106 cells/mouse). The mice were sacrificed at 7 days after transplantation and cancer tissues were collected. Obtained cancer tissues were cut into uniform-sized pieces (3 mm × 3 mm × 3 mm), and then transplanted into mammary fat pad of 2nd recipient BALB/c mice. Cultured colon26-EGFP cells were transplanted into the rectal smooth muscle of BALB/c mice (1×106 cells/mouse, orthotopic transplantation model) according to the published protocols [36]. Cultured LLC-EGFP cells were subcutaneously injected into the right flank of 1st recipient C57BL/6 mice (1 × 106 cells/mouse) that were later sacrificed to obtain cancer tissues at 7 days after transplantation. Obtained cancer tissues were cut into uniform-sized pieces (4 mm × 4 mm × 4 mm), and then transplanted into the right flank of 2nd recipient C57BL/6 mice. In all experiments, sham-operated mice received identical surgery and saline injection without cancer transplantation. It should be noted that in our hands, LLC cancer cells grow rapidly, causing organismal death even before visible metastasis occurs. Thus, regarding the LLC-related experiments, we focused on pre-metastatic niche formation.

Four weeks after cancer cell transplantation, the mice were euthanized for evaluating pulmonary metastases. The number of nodules reflecting pulmonary metastasis of 4T1-EGFP or colon26-EGFP cells was counted by using images obtained with a fluorescent microscope (OV100, Olympus, Tokyo, Japan). To evaluate the effects of ANP on pre-metastatic niche, sham-operated or cancer-transplanted mice were euthanized for histological and gene expression analyses on the lung at day 7 or 10 after 4T1 or LLC transplantation, respectively.

Starting on the same day of cancer transplantation, ANP (0.5 μg/kg/min, Peptide Institute Inc, Osaka, Japan) in 0.9% saline or saline alone (vehicle) was continuously injected into sham-operated or cancer-transplanted mice with the aid of osmotic pumps (Alzet Model 1002 or 2004, DURECT Corporation, Cupertino, CA) implanted subcutaneously in the upper back of the mice [22].

RNA isolation, cDNA synthesis, and quantitative PCR (qPCR)

Total RNAs were prepared by using Trizol reagent (Thermo Fisher Scientific, MA, USA). The obtained supernatants containing total RNAs were further purified with RNeasy mini kit (Qiagen, Venlo, Netherlands) according to the manufacture’s instruction. 0.1–2 µg of total RNAs were then reverse transcribed with SuperScript III first-strand synthesis system (Thermo Fisher Scientific) according to the manufacture's instruction. The obtained cDNAs were 5- or 10-fold-diluted and subjected into qPCR experiments by using LightCycler480 Instrument II system and SYBR Green Master Mix (Roche, Basel, Switzerland). The obtained data were analyzed using the delta-Ct method. 36B4 was used as internal control. Primers used are indicated in Supplementary Table 1.

Western blot

Western blot analysis was performed as previously described [37]. Anti-Vimentin (ab92547) was purchased from abcam. Anti-β actin antibody (A2228) was provided from Sigma-Aldrich. Anti-β catenin antibody (610182) was purchased from BD.

Spheroid formation

Spheroid formation experiments were performed as previously described [23, 24].

RNA-seq and Bioinformatic analysis

RNA-seq analyses were performed as described previously [38, 39]. The obtained gene list with Reads per million per a kilobase (RPKM) scores were shown in Supplementary Tables 2 and 3. To identify differentially expressed genes (DEGs), we focused on the well-annotated protein-coding genes. RPKM scores from two replicates were averaged, and the ratio were calculated. In this calculation, 1 was added to all averaged RPKM scores to ignore the scores below “1”, and to make analyses more stringent. The obtained ratios were used to sort genes to find DEGs, followed by statistical analyses and qPCR validation. Data visualization was done mostly using ggplot2 (http://ggplot2.org/). Used in-house R scripts are all available upon request. Gene set enrichment analyses (GSEA) were performed following the instructions (http://software.broadinstitute.org/gsea/index.jsp) [40, 41]. Gene ontology analyses were done with http://geneontology.org/page/go-enrichment-analysis.

Histological analysis

The left lung was fixed by intratracheal instillation of 4 % paraformaldehyde for histological analysis, and subsequently embedded in paraffin. Paraffin sections were stained with hematoxylin–eosin and anti-Mac3 antibody. For Mac3 staining, tissue sections were deparaffinized, and endogenous peroxidase was blocked with 3 % H2O2 for 30 min. After each step, the tissue sections were rinsed twice in phosphate-buffered saline (PBS) for 5 min. The deparaffinized tissue sections were incubated with Protein Block (Dako Cytomation, Glostrup, Denmark) for 15 min. Anti-Mac3 antibody was diluted in an antibody diluent buffer (dilution 1:500; BioLegend, San Diego, CA, USA) and applied overnight at 4 °C. For other staining, anti-S100A8 antibody (ab92331 1:500), Anti-S100A9 antibody (ab105472 1:500), Anti-MMP9 antibody (ab38898 1:400) and Anti-CXCR2 antibody (ab217314 1:500) were purchased from abcam. After incubation with primary antibodies, the slides were incubated with biotinylated rabbit anti-rat IgG for 60 min, followed by incubation with peroxidase-conjugated avidin–biotin complex (Vectastain ABC kit; Vector Laboratories, Burlingame, CA) for 30 min. Antigen–antibody complexes were visualized with 0.5 % diaminobenzidine (DakoCytomation) and 0.3 % hydrogen peroxide, and then counterstained with hematoxylin. The numbers of Mac3-positive cells were counted directly under the microscope (FSX100, Olympus).

Statistical analysis

Between-group comparisons were performed using the unpaired Student’s t-test when the data exhibited normal distribution, and Wilcoxon rank-sum test when the data did not pass the normal distribution test. For multiple-group comparisons, one-way ANOVA, followed by the post-hoc Fisher’s least significant difference test, was used. Statistical significance was assumed if p < 0.05.

Data availability statement

The authors declare that all data supporting the findings of this study are available within the article and its Supplementary Information files or from the corresponding author upon reasonable request. RNA-seq data published in the present study have been deposited under the accession number of DRA005198.