Our data indicated that ATF3 in the host facilitates breast cancer metastasis, at least in part, by its function in macrophages/myeloid cells. Although ATF3 was shown to play a role in macrophages (38
), the previous studies differ from our report in 2 important aspects. First, they studied the roles of ATF3 in noncancer, acute stress contexts: septic shock and post–septic shock infection. Second, they used the whole-body Atf3
KO mice, without using CKO mice to address the roles of ATF3 in macrophages/myeloid cells. Thus, our present study is the first to use CKO mice and demonstrate a functional role of ATF3 in macrophages/myeloid cells to promote metastasis. Importantly, Atf3
is an adaptive response gene and is induced in macrophages by a variety of stress signals relevant to the tumor microenvironment, such as hypoxia, cytokines (IL-4, TGF-β, IL-6, and TNF-α), and coculture with cancer cells (Supplemental Figure 4 and reviewed in ref. 18
). Thus, our present work linked macrophage/myeloid stress response to metastasis.
Epidemiological findings support the concept that chronic stress, such as obesity and psychological distress, is a risk factor for cancer (42
). Recently, studies using mouse models provided experimental support for a potential causal relation between chronic stress and increased cancer risk. These include increased melanoma development accompanying high-fat diet (44
), increased breast tumor size in transgenic mice with diabetic symptoms (45
), and increased mammary carcinoma metastasis in mice with restriction stress (daily restraint of movement for 2 hours; ref. 46
). Importantly, in the restriction stress model, the stress treatment increased macrophage infiltration in the tumor, and GW2580 (a CSF1R inhibitor) reduced macrophage infiltration and dampened restriction stress–enhanced metastasis, supporting a role of macrophages in this process (46
). Since Atf3
is a stress-inducible gene, it will be interesting to determine whether ATF3 in macrophages plays a role in any stress-enhanced cancer progression.
Although various genes have been shown to be important for macrophages/myeloid cells in the breast cancer context, few have been documented to be necessary in vivo using the CKO approach, including Ets2
, and Hif1a
differs from them in a key aspect, in that it is an immediate-early gene induced by a wide spectrum of stress signals. Immediate-early genes encode transcription factors and are known to modulate genes encoding transcription factors, which in turn regulate downstream genes, leading to a cascade of changes in transcriptional programs. Thus, ATF3 has far-reaching effects in orchestrating the cellular adaptation to stress signals. Our data suggest that ATF3 in macrophages promotes M2 skew, since Il12
mRNA (M1) was lower in WT than KO cells, but Arg1
mRNA (M2) was higher (Figure D). We note that the promoters of both genes contain consensus ATF/CRE sites (38
), potential binding sites for ATF3. Since ATF3 can either activate or suppress gene expression depending on its dimerization with other basic region leucine zipper proteins (18
), it is possible that ATF3 directly upregulates Arg1
, but downregulates Il12
. Our microarray data indicate that ATF3 modulates the expression of genes involved in cell-cell communication, a key component for the function of the immune network. Supplemental Figure 9 shows the cytokines involved in communications among various immune cells, many of which were regulated by ATF3 (directly or indirectly) in macrophages. Thus, ATF3 can affect the ability of macrophages to communicate with other immune cells. Our data also indicate that ATF3 affects the ability of macrophages to communicate with cancer epithelial cells in vitro, as evidenced by 2 coculture assays (3D invasiveness and transendothelial migration; Figures and ). Macrophages are well documented as affecting cancer cell invasion, intravasation, and extravasation (2
). Our present study identified Atf3
, an adaptive response gene that encodes a transcription factor and is induced by various stress signals, as a regulator of these macrophage bioactivities. Mechanistically, Mmp9
is a functionally important target of ATF3 for macrophages to stimulate cancer cell invasion and transendothelial migration in vitro. However, we do not imply that Mmp9
is the only functionally important target; more analyses are required to address the potential importance of other ATF3 downstream genes. Taken together, our results are consistent with the model that, as an immediate-early gene, Atf3
in macrophages links various stress signals in the cancer environment to alterations in transcriptional programs. This results in changes in macrophage bioactivities to facilitate metastasis (Figure ).
Model by which ATF3, a hub of the cellular adaptive response network, links various signals in the tumor microenvironment to alterations in macrophage transcriptional programs and bioactivities.
Our data did not show statistically significant differences in macrophage recruitment in the WT and KO primary tumors (Figure B). Preliminary analysis of CD45+
hematopoietic cells or CD11b+
myeloid-derived suppressor cells also showed no difference in primary tumors, lung, spleen, or bone marrow (Supplemental Figure 3, D and E). Analyses of VEGFr1+
cells, a cell type important for premetastatic niche formation (5
), showed no difference between WT and KO lung or bone marrow (Supplemental Figure 3F). Thus, our studies did not show obvious differences in cell recruitment. However, we do not rule out the possibility that our assays were not sensitive enough to detect subtle differences or that ATF3 in the host may affect the trafficking of some unexamined or unidentified cell types.
Although we showed that the macrophage/myeloid is a key cell type for ATF3 function, we do not exclude the potential roles of ATF3 in other cell types. This is because Atf3
can be induced by stress signals in many cell types (18
) and because various host cells other than macrophages/myeloid cells — such as T cells, B cells, fibroblasts, adipocytes, and endothelial cells — have been shown to play important roles in cancer development (1
). Further investigation is required to address these issues. We note that it is not clear when macrophage/myeloid starts to express ATF3
during human breast cancer development; further investigation using a large number of samples at different stages of breast cancer development is also required.
Finally, the ATF3-117 and ATF3-60 genes derived from mouse TAMs could stratify the breast cancer patients in the NKI and Stockholm datasets into low- and high-risk groups (Figure B and Supplemental Figure 7C). Importantly, Figure D and Table indicate that the expression of ATF3
in the mononuclear cells of human breast tumors correlated with worse outcome. Thus, our data from mouse models and human samples corroborate each other, suggesting that ATF3 may be used as a prognostic marker for metastatic disease. We note that ATF3
expression in the cancer epithelial cells did not correlate with clinical outcomes. This may be because the study was underpowered, or because the expression of ATF3
in the breast cancer epithelium is not a predictor of outcomes. However, this does not mean that ATF3
expression in the cancer epithelium is not functionally important. As shown previously, ATF3 in breast cancer cells plays an oncogenic role in both in vitro and in vivo models (21
). Thus, the expression of ATF3
in mammary epithelium has functional importance, even though it does not correlate with clinical outcome.
In conclusion, our studies demonstrated that ATF3 — a hub of the cellular adaptive response network — is induced in both the cancer epithelia and the mononuclear cells in the stroma. Our data support a model whereby ATF3 induction in the mononuclear cells (presumably reflecting a reactive stroma) alters their gene expression and bioactivities, contributing to host-enhanced metastasis. Importantly, this expression of ATF3 predicted poor clinical outcome in a cohort of patients, suggesting that dampening ATF3 expression in the host may be a potential therapeutic approach.