Multi-organ Site Metastatic Reactivation Mediated by Non-canonical Discoidin Domain Receptor 1 Signaling

Abstract

Genetic screening identifies the atypical tetraspanin TM4SF1 as a strong mediator of metastatic reactivation of breast cancer. Intriguingly, TM4SF1 couples the collagen receptor tyrosine kinase DDR1 to the cortical adaptor syntenin 2 and, hence, to PKCα. The latter kinase phosphorylates and activates JAK2, leading to the activation of STAT3. This non-canonical mechanism of signaling induces the expression of SOX2 and NANOG; sustains the manifestation of cancer stem cell traits; and drives metastatic reactivation in the lung, bone, and brain. Bioinformatic analyses and pathological studies corroborate the clinical relevance of these findings. We conclude that non-canonical DDR1 signaling enables breast cancer cells to exploit the ubiquitous interstitial matrix component collagen I to undergo metastatic reactivation in multiple target organs.

Copyright © 2016 Elsevier Inc. All rights reserved.

Figures

Figure 1. TM4SF1 Induces Metastatic Reactivation in the Lung and Predicts Breast Cancer Relapse

(A and B) 4T1 cells expressing the bioluminescence reporter TGL (4T1-TGL) were infected with lentiviruses encoding two shRNAs targeting TM4SF1 (sh #1 and #5) or a control shRNA (sh-Co.), analyzed by Q-PCR and immunoblotting (IB) (A, top left), and inoculated intravenously (i.v.) in syngeneic mice. Lung metastasis was measured by bioluminescent imaging (BLI). Normalized photon flux at the indicated times (A, bottom left), representative images (A, right). Error bars, mean ± SD. P values (last time point); Student’s t test. Lung sections of indicated times were stained with antibodies to GFP (tumor cells) and PECAM-1 (endothelial cells) (B, top). White arrows: solitary tumor cells. Number of GFP-positive solitary tumor cells/high power field (hpf) (blue bars) and metastatic lesions/lung section (red bars) at the indicated times (B, bottom). (C–E) Schematic representation of inducible expression experiment (C). 4TO7 cells expressing DOX-inducible TM4SF1 were inoculated i.v. in syngeneic mice. DOX was administrated either immediately after injection (Dox day 0) or 14 days after inoculation of the cells (Dox day 14). Lung metastasis was measured by BLI. Normalized photon flux at the indicated time; error bars, mean ± SE.P values; Student’s t test (D). Representative images (E). (F) Kaplan-Meier analysis of relapse-free survival of ER+ (left) and ER− patients (right) in publicly available breast cancer datasets (source KM Plotter for breast cancer). Patients were divided according to TM4SF1 expression as indicated. HR: Hazard Ratio. (G and H) TMA comprising 147 primary breast tumors of MSKCC’s patients were subjected to immunohistochemistry with anti-TM4SF1 and counterstaining with Hematoxylin (H). Representative images of cases exhibiting varying levels of TM4SF1 (G). Distribution of cumulative staining intensities across all samples (H, left). Patients were divided according to the intensity of TM4SF1 staining as indicated by the red arrow (left) and metastasis-free survival data were subjected to Kaplan-Meier analysis (H, right). (I) Hierarchical clustering of genes concordantly up or downregulated (≥ 1.5 fold) in triplicate samples of TM4SF1-overexpressing 4TO7 cells as compared to control cells. (J) Kaplan-Meier analysis of relapse-free survival in the MSK82, EMC192, EMC286, and NKI295 combined dataset. Patients were divided according to the expression of the 8-gene TM4SF1 signature. Individual genes comprising the signature are listed to the right of the graph. See also Figure S1.

Figure 2. TM4SF1 Enhances Cancer Stem Cell Traits in Response to Collagen I

(A and B) Control and TM4SF1-silenced mammary tumor cells from MMTV-Neu(YD) mice (ErbB2 cells) were cultured in 3D Matrigel. Representative wells (A) and mean number of organoids (± SE) per 100 cells from triplicate samples (B). P values; Student’s t test. (C and D) Control and TM4SF1-silenced ErbB2-TGL cells were admixed with Matrigel and inoculated in the mammary fat pad of NSG mice at the indicated numbers. Mean of primary and secondary tumor volumes at 5 weeks (C; left y axis). Incidence of tumors at 5 weeks (C; right y axis) and tumor latency at indicated cell number (D). P values; Student’s t test. Data point, mean ±SD (E) Control and TM4SF1-silenced ErbB2 cells were subjected to Q-PCR analysis. Error bars, mean ± SD. (F) ErbB2 cells were subjected to tumor sphere assay with or without collagen (Coll) I (50 µg/ml), Coll IV (50 µg/ml), laminin 1 (Lam 1; 10 µg/ml), or fibronectin (Fn; 50 µg/ml) for 10 days. Number of tumor spheres per 103 cells (±SE); P values; Student’s t test. (G) Indicated ErbB2 cells were subjected to tumor sphere assay in medium supplemented with or without 50 µg/ml collagen I. Number of primary (left) and secondary (right) tumor spheres per 103 cells seeded (±SE). P values; Student’s t test. (H) ErbB2 cells were subjected to tumor sphere assay in medium supplemented with or without 50 µg/ml coll I in combination with vehicle (0.25% DMSO), Y-27632 (ROCK1 inhibitor, 10 µM), β-Alaninenitrile (BAPN, 200 µM), or function-blocking antiLox1 antibody (5 µg/ml).Number of tumor spheres formed per 103 cells seeded (±SE). P values; Student’s t test. See also Figure S2.

Figure 3. TM4SF1 Promotes Metastasis to the Bones and Brain

(A) ErbB2-TGL cells were inoculated (3 × 105) intracardially (i.c.) in nude mice. Bone and brain metastases were measured by BLI. Representative images are shown. (B) Representative X-ray radiography of hind legs. Yellow arrows point to osteolytic lesions. (C and D) Hematoxylin and Eosin (H&E) and anti-GFP IHC of adjacent sections from bone (C) and brain metastases (D). (E) The graphs show the cumulative photon flux from both hind legs (top) or the brain (bottom) at day 21 after i.c. injection (± SE). P values, unpaired Student’s t test. (F) Bone (hind limbs) and brain sections of mice were processed for IHC detection of GFP (tumor cells). The graphs show the mean numbers of GFP-positive solitary tumor cells per microscopic field (blue bars) and of metastatic lesions per section (red bars) at end point of the experiment (±SE). P values; Student’s t test. See also Figure S3.

Figure 4. DDR1 Combines with TM4SF1 to Promote Tyrosine Kinase-independent Signaling and Metastasis

(A) Pathway and interaction data are from HPRD, Reactome, NCI-Nature Pathway Interaction Database, and the MSKCC Cancer Cell Map, as derived from Pathway Commons. Black lines: direct and specific interactions; white spheres: interactions validated by co-immunoprecipitation (IP). (B) 293FT cells were co-transfected with Myc-TM4SF1, together with Flag-DDR1, or empty vector and IP with anti-Flag. Total lysates (left) and Flag IP (right) were subjected to IB with the indicated antibodies. Asterisks: glycosylated TM4SF1; arrow: TM4SF1 core protein. (C) 293FT cells were co-transfected with Myc-TM4SF1 together with Flag-DDR1, Flag-TrkA-DDR1 (extracellular domain chimera), Flag-DDR1 609* (intracellular domain truncation), Flag-DDR1 529* (intracellular domain truncation), or empty vector and were IP with anti-Flag. Flag IP and total lysates were subjected to IB with the indicated antibodies. (D) ErbB2 cells were treated with or without 10 µg/ml coll I for 30 min, stained with anti-DDR1, and counterstained with the plasma membrane dye FM-4 64 (PM) and DAPI. Representative images (top panels). Myc-TM4SF1 and HA-DDR1 co-transfected 293FT cells were treated with or without 10 µg/ml collagen I for 30 min and stained with anti-HA and anti-TM4SF1 followed by counterstaining with DAPI. Representative images of co-transfected cells (bottom panels). (E) ErbB2 cells were treated with or without 10 µg/ml collagen I for 30 min and stained with anti-DDR1 antibody and DAPI. Representative images (top) and surface fluorescence intensity plots (bottom). The insets show high magnification images of a portion of the plasma membrane adjacent to the boundary of the nucleus demarcated by white dotted lines (bottom). (F) Control and TM4SF1-silenced MDA-MB231 cells were treated with or without 10 µg/ml collagen I for 30 min. When indicated, cells were pretreated with 10 µg/ml anti-DDR1 or anti-ITGB1 (integrin β1) function-blocking monoclonal antibodies. Number of DDR1 clusters per microscopic field. Mean values (± SD); P values; Student’s t test. (G) ErbB2 cells transfected with control or TM4SF1-specific siRNA SMARTpools were treated with 10 µg/ml coll I for the indicated times, and IP with anti-DDR1. Immunoprecipitates were subjected to IB with the indicated antibodies (H) ErbB2 cells were treated with or without 10 µg/ml collagen I for 30 min, alone or in combination with 100 nM Ponatinib or Imatinib, and then subjected to staining with anti-DDR1 followed by DAPI. Number of DDR1 clusters per microscopic field (± SD). P values; Student’s t test. (I and J) Control (sh-co) or DDR1-silenced (sh #6 and #7) (I) and control (vector) or DDR1-P529*-overexpressing (J)ErbB2-TGL cells were inoculated i.v. in nude mice. Lung metastasis was measured by BLI. Normalized photon flux at the indicated times (left (± SD) and representative images (right); P values; Student’s t test. See also Figure S4.

Figure 5. Non-canonical DDR1 Signaling Promotes PKCα-dependent Activation of JAK2 and STAT3

(A) 4TO7-TGL cells were transduced with wild type or mutant (C202G) TM4SF1 or with empty vector, subjected to IB as indicated (left top), and inoculated i.v. in syngeneic mice. Lung metastasis was measured by BLI. Normalized photon flux at the indicated times (left bottom; ±SD) and representative images (right). P values; Student’s t test. (B) Control (sh-co) or syntenin 2-silenced (sh #1 and #3) ErbB2-TGL cells were subjected to IB as indicated (left top), and inoculated i.v. in nude mice. Lung metastasis was measured by BLI. Normalized photon flux at the indicated times (left bottom) and representative images (right). Error bars, mean ± SD. P values Student’s t test. (C) Control or TM4SF1 specific siRNA (SMART pool) transfected ErbB2 cells were treated with 50 µg/ml collagen I for the indicated times. Total lysates were subjected to IB with the indicated antibodies. (D–F) ErbB2 cells transfected with control or TM4SF1-specific siRNA SMARTpools were treated with coll I, alone or in combination with GÖ6976 (50 nM), Bisindolylmaleimide I (Bis I, 300 nM), Fedratinib (1 µM), AZD1480 (100 nM), or Imatinib (100 nM), lysed, and subjected to a PKC kinase assay (D, E) or immunoblotting (F). Mean values (± SD); P values Student’s t test (D, E). (G) ErbB2 cells were treated with 100 nM PMA for the indicated times and subjected to IB. (H) Purified recombinant PKCα and JAK2 were admixed in vitro, alone or in combination with PMA (PKC activator, 50 nM) or GÖ6976 (PKC inhibitor, 100 nM), as indicated, and subjected to in vitro kinase assay with radioactive ATP. (I) JAK2-deficient γ2A cells were transfected with wild type (WT) or mutant forms of Flag-JAK2. Total lysates were subjected to in vitro kinase assay with cold ATP and GST-STAT3as substrates. Samples were subjected to IB as indicated. Note these cells possess elevated levels of endogenous phosphorylated PKC. Wild type JAK2, but none of phosphorylation site mutants, induces phosphorylation of GST-STAT3 at Y705. An endogenous activity induces phosphorylation of STAT3 at S727, serving as an internal control. (J) Model of non-canonical DDR1 signaling. Kinase-dependent pathways are shaded grey. See also Figure S5.

Figure 6. Non-Canonical DDR1 Signaling Promotes Self-renewal in Vitro and Lung Metastasis in Vivo

(A) ErbB2 cells were treated with 50 µg/ml coll I alone or combination with GÖ6976 (50 nM), Bis I (300 nM), Fedratinib (1 µM), Stattic (2.5 µM), BP-1-102 (20 µM), and Ponatinib (100 nM) or they were transduced with DDR1 sh-RNAs (#6 and #7) before treatment with coll I. The graph shows the number of tumor spheres formed per 103 cells seeded (± SE). P values, unpaired Student’s t test. (B) ErbB2 cells transfected with control or TM4SF1-specific siRNA SMARTpools were subjected to Q-PCR analysis (± SE). (C) ErbB2 cells were treated with DMSO or GÖ6976 (50 nM) (top), Fedratinib (1 µM) (middle), and Stattic (2.5 µM) (bottom), and subjected to Q-PCR analysis (± SE). (D and E) ErbB2 cells were transfected with control or Sox2- (si-Sox2), Nanog- (si-Nanog), or Oct4-targeting (si-Oct4) siRNA SMARTpools and subjected to tumor sphere assay with 50 µg/ml coll I. Number of tumor spheres formed per 103 cells seeded (± SE) (D); representative images of tumor spheres (E). P values, unpaired Student’s t test. (F and G) 4T1-TGL cells were transduced with dominant negative (DN) PKCα, PKCβ II, PKCδ, or PKCζ, or empty vector and inoculated i.v. in syngeneic mice. Lung metastasis was measured by BLI. Normalized photon flux at the indicated times for DN PKCα or PKCβ II, PKCδ, or PKCζ relative to empty vector (±SD). P values; Student’s t test (F). Representative images at week 5 (G). (H and I) Control and STAT3-silenced (#1 and #5) 4T1-TGL cells were inoculated intravenously into syngeneic mice. Lung metastasis was measured by bioluminescent imaging. Normalized photon flux at the indicated times (±SD). P values; Student’s t test (H). Representative images at week 5 (I). (J) Control and DDR1- (#6 and 7) or syntenin 2-silenced (#1 and 3) ErbB2-TGL cells were inoculated i.c. in immunodeficient mice. Bone and brain metastases were detected by bioluminescent imaging. The panels show representative images. See also Figure S6.

Figure 7. Fibrillar Collagen I and Activated STAT3 Define Metastatic Niches

(A) Representative images of preneoplastic and MIN lesions (top panels) and adenocarcinomas (bottom panels) from 12 and 20 week-old MMTV-Neu(YD) mice subjected to IHC with the indicated antibodies followed by Hematoxylin or to Masson trichrome staining. Red arrow: normal duct comprising quiescent cells; purple arrow: pre-neoplastic lesion comprising hyperproliferative cells; green arrow: MIN lesion that has undergone luminal filling (inset). (B) Lung sections from mice that had been injected intravenously with ErbB2-TGL cells 14 days earlier were subjected to immunofluorescent staining with antibodies to Neu, collagen I, and Ki-67 followed by counterstaining with DAPI. Solitary tumor cells (Neu+) were categorized as closely apposed to coll I (+) or not (−) and as Ki67+ or Ki-67−. Percentage of Ki-67+ cells among solitary tumor cells in contact or not with coll I (±SE). P values unpaired Student’s t test (left). Representative images of single cells from the predominant categories (right). (C–E) TMAs comprising primary breast tumors and paired lung metastatic lesions from MSKCC patients were subjected to IHC with indicated antibodies. Level of DDR1 (left), TM4SF1 (middle), and P-STAT3 (right) staining in primary breast tumors and lung metastatic lesions (C). Comparison of P-STAT3 staining in 14 matched primary tumors and metastases (D, left); representative images from one case (D, right). Data are mean (± SE and ± SD), P values unpaired student’s t test. Patients (n=61) were divided based on the level of P-STAT3 staining in their lung metastases and their metastasis-free survival examined by Kaplan-Meier analysis (E). P: primary breast tumors; M: lung metastatic lesion. See also Figure S7.