The ADAMTS2 metalloproteinase inhibits tumor growth by regulating the innate immune system

Abstract

Background: ADAMTS2 is a metalloproteinase known to be implicated in collagen maturation and regulation of (lymph)angiogenesis. As these properties are likely to alter tumor progression, we aimed to assess the overall impact of ADAMTS2 on cancer development.

Methods and results: Using publicly available human cancer datasets, we found that high expression of ADAMTS2 in primary tumors is associated with poor prognosis across various cancer types. Similar analyses were repeated, but this time using the ratio of ADAMTS2 on COL1A1 expression to take into account potential biases due to the involvement of ADAMTS2 in collagen fibril formation. Remarkably, these data indicate that patients with a high ADAMTS2/COL1A1 ratio exhibit an improved overall survival rate, suggesting that ADAMTS2 may inhibit cancer progression by a mechanism independent of collagen accumulation. This hypothesis was evaluated in vivo using ADAMTS2-KO mice and different tumor models characterized by the absence or presence of tumor collagen accumulation, as in MMTV-PyMT mice which develop spontaneous desmoplastic mammary tumors. In all the models, the growth of primary tumors was strongly increased in ADAMTS2-KO mice versus their wild type counterparts, confirming that ADAMTS2 displays anti-tumor properties. In stark contrast, the spread of lung metastases from mammary tumors was virtually prevented in ADAMTS2-KO mice, showing a dual role of ADAMTS2, either beneficial or detrimental, at different stages of cancer progression. Additional investigations, notably by FACS and single cell sequencing, showed that the effect of ADAMTS2 on primary tumors does not result from a direct effect on cancer cells, but rather from modifications in the intratumor innate immune system which becomes more immunosuppressive in the absence of ADAMTS2.

Conclusion: We have shown that ADAMTS2 suppresses tumor growth by inhibiting the progressive establishment of an immunosuppressive microenvironment. Conversely, its presence allows efficient formation of lung metastases. These data identify ADAMTS2 as a cancer regulator with antagonistic functions, limiting initial progression but promoting efficient metastatic dissemination.

Keywords: ADAMTS; IMMUNITY; Tumor micro-environment.

Fig. 1

ADAMTS2 expression and patient overall survival. The “Kaplan Meier plotter” (Pan-cancer RNA-Seq) and “GEPIA” (a web server for cancer and normal gene expression profiling) were used to evaluate overall survival of cancer patients expressing high (red) or low (blue) levels of ADAMTS2 in the primary tumors (See the Methods section for more details). (A) Data obtained for kidney renal clear cell carcinoma, bladder carcinoma and lung adenocarcinoma are provided as examples and show that high ADAMTS2 expression correlates with reduced overall survival. Such correlation was also found when all cancer types were grouped together in a single cohort (right panel: Pooled TCGA cancers). A table summarizing the data related to individual cancer types is provided as Figure S1A. (B) Similar analyses were repeated, but using the ratio of ADAMTS2 on COL1A1 expression as a normalization for CAFs abundance and activation (see also Figure S1B). This completely reversed the correlations indicating that high ADAMTS2/Col1A1 ratio is of good prognosis. HR: hazard ratio

Fig. 2

ADAMTS2 does not affect cancer cells in vitro but represses tumor growth in vivo. Modified B16 melanoma (A) or LLC (B) cells were left untreated (-Dox) or were incubated with doxycycline (+ Dox; 0.1 µg/ml) for 48 h to induce ADAMTS2 expression. Cells are visualized by phase contrast and fluorescence microscopy (for DsRed staining). For cell proliferation evaluation (C), B16 melanoma cells, either control (Ctrl, absence of Dox) or expressing ADAMTS2 (+ Dox) were imaged during 144 h to quantify cell confluence (n = 16). For the migration assay (D), B16 melanoma cells, either control (Ctrl) or expressing ADAMTS2 due to the presence of Dox (TS2), were transferred in a wound scratch assay-dedicated device and treated with mitomycin (8 µg/ml during 2 h) to prevent mitosis. The inserts were then removed to generate a “cell-free gap” (wound) and wound closure (in %) was evaluated during 9 h (in fresh medium supplemented with 0.5% FBS, and containing or not Dox) (n = 6). E. Diagram summarizing the experimental design used to evaluate the effect of ADAMTS2 on tumor growth. Mice, wild-type or TS2-KO, received (+ Dox) or not (-Dox) doxycycline in their drinking water from 48 h before cell injection (day 0) until the end of the experiment. B16 melanoma cells (with Dox-induced ADAMTS2 expression) and LLC cells (with Dox-induced ADAMTS2 or EGFP expression) were pretreated or not with Dox (0.1 µg/ml) for two days before subcutaneous injection. Mice were sacrificed and tumors were weighed at day 14 for LLC tumors or day 21 for B16 tumors. Dots represent individual tumor weight, the medians and the 25th and 75th quartiles are shown. The experiments for LLC and B16 cells were performed 3 and 2 times respectively and the graphs were generated by pooling tumor weights from each experiment. In the absence of Dox, tumor weights were significantly higher for both B16 cells (F) and CLL cells (G). Notably, Dox-induced ADAMTS2 expression in TS2-KO mice reduced tumor growth to levels observed in WT mice. As a control, Dox-induced EGFP expression (H) did not alter tumor growth. Histological analyses were carried out on subcutaneous tumors (I). These tumors were highly heterogeneous and characterized by the presence of leaky vessels (yellow arrows) or large hemorrhagic areas (yellow dotted lines), areas of necrosis and pyknotic cancer cells, as well as the presence of numerous blood vessels (red arrowheads). However, no differences due to ADAMTS2 expression were observed. **: p < 0.01; ***: p < 0.001; ****: p < 0.0001

Fig. 3

Spontaneous MMTV-PyMT mammary tumors develop more rapidly, but generate fewer lung metastases, in S2-KO mice. (A) Diagram summarizing the experimental design used to evaluate the effect of ADAMTS2 on the formation of spontaneous mammary tumors in MMTV-PyMT mice expressing (WT) or not (TS2-KO) ADAMTS2. Mice were followed individually and sacrificed once the ethical endpoint for tumor size was reached. (B) The histology of these tumors was highly heterogeneous, with areas of necrosis, numerous blood vessels and large regions containing packed cancer cells. No difference could be seen between tumors from WT and TS2-KO mice. (C) The age at sacrifice (when reaching the ethical endpoint) was used to compare the rate of tumor growth between the two genotypes. Age at sacrifice was highly significantly reduced for TS2-KO mice. Statistical analyses were performed using Log-rank Mantel-Cox test (P value < 0.0001) and Gehan-Breslow-Wilcoxon test (P value < 0.0001). (D) The mean number of metastatic foci in PyMT-WT and PyMT-TS2-KO mice was determined after HE staining of lung sections. (E) Illustrations of metastases in the lung of WT and TS2-KO mice (HE staining). A representative section stained for the presence of the MMTV-PyMT antigen is provided as Figure S3 and confirms the metastatic nature of the cell clusters identified by HE staining. ****: p < 0.0001

Fig. 4

ADAMTS2 represses tumor growth independently of its aminoprocollagen activity. (A) MMTV-PyMT cancer cells, established from MMTV-PyMT-WT or MMTV-PyMT-TS2-KO mice, were characterized in culture where they displayed a similar epithelial phenotype, with cells proliferating in clusters (phase contrast microscopy, left panels) and strongly expressing epithelial markers (such as keratin) but not mesenchymal markers (such as vimentin). (B) Experiments were set up to compare the tumors formed in WT and TS2-KO mice by subcutaneous injection of PyMT-WT and PyMT-TS2-KO cancer cells. (C) Tumor weights were higher in TS2-KO than in WT controls, and this increase was similar for MMTV-PyMT WT and TS2-KO cancer cells. (D) RT-qPCR evaluation of relative ADAMTS2 expression in subcutaneous tumors generated from PyMT-WT and PyMT-TS2-KO cancer cells in WT and TS2-KO mice. (E) Western blot analysis of the processing of type I collagen extracted from subcutaneous tumor generated by PyMT-WT and PyMT-TS2-KO cancer cells in WT and TS2-KO mice. pNα1: alpha 1 type I chain retaining its aminopropeptide. α1: fully processed alpha 1 type I chain. pNα2: alpha 2 type I chain retaining its aminopropeptide. α2: fully processed alpha 2 type I chain. (F) Histological sections of subcutaneous tumors formed by PyMT-TS2-KO cancer cells in WT and TS2-KO mice. Staining was performed with H&E for general histology; Sirius red for fibrillar collagen organisation; antibodies to keratin or vimentin to identify cells with an epithelial or mesenchymal phenotype, respectively. (G) Immunofluorescence to identify blood vessels (CD31 staining of endothelial cells) and proliferating cells (KI67). Nuclei are stained in blue with Dapi. Quantifications are provided on the right panels

Fig. 5

Single cell sequencing analysis of the different cell populations present in subcutaneous tumors formed in WT and TS2-KO mice. (A) Tumors were induced by subcutaneous injection of PyMT-TS2-KO cancer cells in WT and TS2-KO mice. After 14 days, tumors were collected and digested. Cells were then sorted into CD45⁺ and CD45⁻ cells (by magnetic-activated cell sorting: MACS) before sequencing. (B) Cell clustering mostly identified macrophages and PyMT cancer cells. (C) Their re-clustering was then carried out to identify potential differences between tumor formed in WT and TS2-KO mice (Tables S10 and S11). In the macrophages, an increase in population 3 (characterized by polarization towards a more M2-type immunosuppressive phenotype, see Figure S5) was observed, probably at the expense of the populations 0 and 1

Fig. 6

ADAMTS2 deficiency leads to increased immunosuppression in early-stage tumors. (A) MMTV-PyMT cells were subcutaneously injected in WT and TS2-KO mice, and tumors were collected 7 days later. (B) Using this “early” model, tumor growth was still increased in TS2-KO mice. Tumors were then digested and the immune cell profile characterized by FACS (C-P). (C) Relative abundance (%) of CD45⁺ cells within live cells. (D) Relative abundance of macrophages (CD45⁺ CD11b⁺ F4/80^high) within CD45⁺ immune cells. (E) Expression (Mean Fluorescence Intensity, MFI) of Arginase 1 (Arg1) in macrophages. (F) Relative abundance (%) of dendritic cells within CD45⁺ immune cells. Proportion of (G) M1-like (CD206^low CD11c^high) and (H) M2-like (CD206^high CD11c^low) macrophages in the macrophage population. Illustration of the gating strategy of M1-like and M2-like macrophages in WT (I) and TS2-KO (J) tumors. (K) Relative abundance of T cells within CD45⁺ cells and relative abundance of CD3⁺T cells expressing IL-10 (L) and the related gating strategy (M, N). Quantification of the expression (Mean Fluorescence Intensity, MFI) of CD14 in M-MDSC (O) and G-MDSC (P), accompanied by representative profiles obtained from tumors formed in WT (green) and TS2-KO (red) mice

Fig. 7

ADAMTS2 reduces tumor growth through a mechanism involving the innate immune system. (A) Two days before subcutaneous injection, mice (Nude, NOD-Scid or NSG) received or not doxycycline in their drinking water (+/- Dox). Similarly, cells engineered (TetON) to express ADAMTS2 (B, C, D) or EGFP (E) were treated or not with doxycycline. HEK and LLC cells were injected in B and C-E, respectively. Experiments were stopped and tumors were weighed at 21 days (B) or 14 days (C-E) after subcutaneous injection. Squares represent individual tumor weights, and Box plots report medians and (upper and lower) quartiles. (B) and (D) graphs result from the pooling of two distinct experiments. Dox: doxycycline. **: P < 0.01 and *: P < 0.05

Fig. 8

ADAMTS2 is an enzyme with antagonistic functions for cancer patients, being beneficial when expressed in primary tumors, but deleterious regarding metastasis formation and growth. In the tumor microenvironment (TME), ADAMTS2 is mainly produced by cancer-associated fibroblasts (CAFs) and macrophages (TAM). Its absence in Adamts2^−/− mice (TS2-KO) leads to a more immunosuppressive TME, as evidenced by the increased number of M2-like macrophages, which ultimately results in increased tumor growth. This suggests that ADAMTS2, through cleavage of as yet unidentified substrate(s), may support/extend an inflammatory TME and, thereby, limit tumor growth. Interestingly, the absence of ADAMTS2 strongly suppresses lung metastasis. Since our data do not show changes in the phenotype of cancer cells within tumors, the strong reduction in metastasis in Adamts2-/- mice could result from mechanisms affecting “circulating” cancer cells in the blood or lymphatic system. It is worth noting that lymph node “fibroblastic reticular cells” are among the body’s cells that produce the most ADAMTS2 and that CD169/Siglec1-positive macrophages in lymph nodes have been shown to protect against metastasis [37]. Finally, the sharp reduction in the number of metastases observed in TS2-KO mice could also be the result of a reduction in the proliferation of cancer cells having reached the lungs. Alveolar macrophages and lung mesenchymal cells express Adamts2. Alveolar macrophages (which are increased in TS2-KO mice) can induce the “dormancy” of invading cancer cells by secreting TGFβ2 which then interacts with TGFβ-R3 present on the surface of cancer cells [38]. We previously showed that TGFβ-R3 and other TGFβ pathway factors are substrates of ADAMTS2. Therefore, their cleavage and inactivation in the lung could reduce dormancy and lead to increased metastatic burden in wild-type compared to TS2-KO mice. Hypotheses still requiring confirmation are indicated by question marks (?)