Illustrating the interplay between the extracellular matrix and microRNAs

Abstract

The discovery of cell surface receptors that bind to extracellular matrix (ECM) components marked a new era in biological research. Since then there has been an increasing appreciation of the importance of studying cells in the context of their extracellular environment. Cell behaviour is profoundly affected by the ECM, whose synthesis and turnover must be finely balanced in order to maintain normal function and prevent disease. In the last decade, microRNAs (miRNAs) have emerged as key regulators of ECM gene expression. As new technologies for the identification and validation of miRNA targets continue to be developed, a growing body of data supporting the role of miRNAs in regulating the ECM biology has arisen from a variety of cell and animal models along with clinical studies. However, more recent findings suggest an intriguing interplay between the ECM and miRNAs: not only can miRNAs control the composition of the ECM, but also the ECM can affect the expression of specific miRNAs. Here we discuss how miRNAs contribute to the synthesis, maintenance and remodelling of the ECM during development and disease. Furthermore, we bring to light evidence that points to a role for the ECM in regulating miRNA expression and function.

Introduction

Once regarded as an inert scaffold that provided support for cells and mechanical strength to tissues and organs, the extracellular matrix (ECM) is now considered a highly active 3D network of molecules that interacts with cells, thereby regulating or instructing their behaviour. Thus, over the last two decades, a major shift has been observed in matrix biology research, from being focussed on the structure and function of ECMs to being much more oriented towards the signalling cascades generated or modulated by the ECM [reviewed in Hynes (2009); Hubmacher & Apte (2013)]. A highly dynamic structure with specific physical and biochemical properties, the ECM affects the phenotype of all resident cells and regulates numerous cellular events such as adhesion, migration, proliferation, differentiation, survival and immune system signalling that ultimately affect development, homeostasis and tissue repair (Hynes 2009; Rozario & DeSimone 2010; Lu et al. 2011). A fine balance between synthesis and turnover determines the ultimate composition of the ECM and the maintenance of normal function (Visse & Nagase 2003; Verrecchia & Mauviel 2007). Perturbation of this equilibrium results in cellular and tissue changes that can lead to the development or progression of disease, including cancer (Lu et al. 2012). Consequently, the integrity and structure of the ECM must be stringently controlled. Formation, maintenance and remodelling of the ECM are regulated on multiple levels; among them, microRNAs have emerged as key regulators of ECM molecule gene expression.

microRNAs (miRNAs) are a large family of endogenous, evolutionary conserved, non-coding RNAs that are ∼22 nucleotides long and which have been implicated in the regulation of nearly every biological process by modulating gene expression at the post-transcriptional level (Ambros 2004). [Excellent reviews on the biogenesis and expression of miRNAs have been published elsewhere (Bartel 2004, 2009; Lagos-Quintana et al. 2002; Wienholds et al. 2005) and will not be discussed further here]. Genetic ablation of key components of the miRNA machinery in mice results in embryonic lethality or severe developmental defects as a consequence of cell cycle arrest and impaired differentiation, underscoring the importance of miRNAs in development and differentiation (Bernstein et al. 2003; Kanellopoulou et al. 2005; Wang et al. 2007). The loss of even a single miRNA has been shown to severely compromise the ability of the host to respond to various environmental cues and pathologic stress (Rodriguez et al. 2007; Thai et al. 2007). Moreover, deregulated miRNA expression has been shown to play a role in the pathogenesis of a growing list of human diseases, including cancer, cardiovascular and autoimmune disorders (Calin & Croce 2006; van Rooij et al. 2006; O'Connell et al. 2010). However, the mechanisms underlying altered miRNA levels remain largely obscure. Emerging evidence shows that the cellular microenvironment has a significant effect on miRNA expression and function. Here, we will discuss how miRNAs contribute to the formation, maintenance and remodelling of the ECM, and we will explore the evidence that points to a role for the ECM in regulating miRNA expression and function.

Regulation of the ECM by miRNAs

The mechanical properties of tissues and the phenotype of their resident cells are influenced by the ECM. miRNAs represent one regulatory mechanism that controls ECM composition and thus the differentiation and behaviour of the cells that reside in that ECM. miRNAs exert their control over the ECM either directly, by targeting ECM molecule mRNAs, or indirectly, by modulating the expression of genes that regulate the synthesis or the degradation of ECM molecules. A variety of technologies are available for the identification and validation of miRNA targets. High-throughput methods for identifying miRNA targets are typically combined with the overexpression and/or inhibition of the miRNA. These include: (i) expression profiling by microarray, high-throughput sequencing or proteomic approaches such as SILAC (stable isotope labelling by amino acids in cell culture) followed by mass spectrometry analysis (Thomas et al. 2010); (ii) polysome profiling by recovery of ribosomal protected RNA fragments (RPFs) followed by deep sequencing (Guo et al. 2010); and (iii) pull-down assays with components of mRISC, including biotinylated or digoxigenin-labelled miRNA pull-down (Orom & Lund 2010; Hsu & Tsai 2011) and immunoprecipitation (IP) of mRISC proteins (e.g. Ago2) such as ribonucleoprotein IP followed by microarray chip analysis (RIP-ChIP) or high-throughput sequencing (RIP-Seq; Keene et al. 2006; Baroni et al. 2008). Optimization of the IP in order to stabilize the RNA-protein binding has led to the development of high-throughput sequencing of RNAs isolated by cross-linking IP (HITS-CLIP) and, more recently, photoactivatable-ribonucleoside-enhanced cross-linking and IP (PAR-CLIP; Licatalosi et al. 2008; Hafner et al. 2010). Validation of the functionality of the identified miRNA target is then performed in a biological system of choice by overexpressing or inhibiting the miRNA of interest and analysing the effect on the target gene expression. Ideally, changes in both mRNA and protein levels should be evaluated. Finally, in order to confirm a direct regulation of gene expression by a miRNA, reporter gene assays are universally performed (Long & Lahiri 2012). Briefly, the 3′ UTR of the target gene is cloned downstream of a luciferase ORF and is transfected into cells that express the targeting miRNA. Reduced luciferase activity indicates direct regulation of gene expression by the miRNA that should be confirmed by mutating bioinformatically predicted binding sites in the 3′ UTR. Below, we review the work which has identified miRNAs that either directly or indirectly target ECM molecules in development and disease. In Table1, we highlight the methods and the biological system used to demonstrate miRNA-mediated regulation of ECM gene expression.

Table 1.

Major ECM molecules directly and indirectly regulated by miRNAs in development and disease

miRNA	ECM molecule	Validation method	System	Process	Reference
Direct regulation
Development
miR-17	Fibronectin	WB, LRA	Human U343 glioblastoma cells, miR-7 TG mouse	Organogenesis	Shan et al. (2009)
miR-138	Versican	qPCR, WB, LRA	Zebrafish embryo, mouse NIH3T3 FBs	Heart development	Morton et al. (2008)
miR-143	MMP-9	qPCR, WB, LRA	Primary rat aortic SMCs	SMC differentiation	Wang et al. (2010c)
miR-212/132	MMP-9	LRA	miR-212/132−/− mouse	Mammary gland development	Ucar et al. (2010)
miR-140	ADAMTS5	qPCR	Primary HACs	Chondrogenesis	Miyaki et al. (2009)
		qPCR, IHC, LRA	Primary miR-140−/− mouse chondrocytes		Miyaki et al. (2010)
miR-142-3p	ADAM9	RT-PCR	Primary chicken limb mesenchymal cells	Chondrogenesis	Kim et al. (2011)
miR-378	Nephronectin	LRA	Mouse MC3T3-E1 osteoblasts	Osteogenesis	Lee et al. (2011)
Disease
let-7g	COL1A2	qPCR, WB, LRA	Human HuH1 HCCs	Cancer metastasis	Ji et al. (2010)
miR-29c	COL1A1, COL1A2, COL3A1, COL4A1, COL4A2, COL15A1, laminin γ1	qPCR, LRA	Human HeLa cervical cancer and HepG2 HCC cells, primary NPC biopsies	Cancer metastasis	Sengupta et al. (2008)
miR-143	FNDC3B	RT-PCR, LRA*	Human HepG2 HCC cells, primary hepatic tissue, 21-HBx TG mice	Cancer metastasis	Zhang et al. (2009)
miR-335	Tenascin-C	MA, LRA	Human MDA-MB-231 breast adenocarcinoma cells, breast cancer biopsies	Cancer metastasis	Tavazoie et al. (2008)
miR-31	Integrin α2, α5, αv, β3	qPCR, WB, FC, LRA	Human MDA-MB-231 breast adenocarcinoma cells	Cancer cell invasion and metastasis	Augoff et al. (2011)
miR-124	Integrin β1	qPCR, WB, LRA	Human OSCC- and HCC-derived cell lines	Cancer cell invasion	Hunt et al. (2011)
miR-183	Integrin β1	qPCR, WB, IC, LRA*	HEK 293 cells, HDF and HTM cells	Cancer metastasis	Li et al. (2010)
miR-93	Integrin β8	WB, IC, LRA	Human U87 and U343 glioblastoma cells, primary HUVECs	Tumour growth, angiogenesis	Fang et al. (2011)
miR-373, miR-520c	CD44	qPCR, WB, LRA	Human MCF7 adenocarcinoma cells, primary SCID mouse bone metastases	Cancer cell invasion and metastasis	Huang et al. (2008)
miR-328	CD44	WB, FC, IC, LRA	Human A431 squamous carcinoma cell, COS-7 monkey fibroblasts	Cancer cell invasion and metastasis	Wang et al. (2008a)
miR-34a	CD44	WB, LRA	Human PPC-1 and Du145 prostate carcinoma cells and xenograft tumours	Tumour progression and metastasis	Liu et al. (2011)
miR-143	MMP-13	Ago2 IP, LAMP, WB	Human osteosarcoma biopsies, HOS and 143b osteosarcoma cells	Cancer metastasis	Osaki et al. (2011)
miR-29b	MMP-2, COL1A1, COL3A1, COL4A1	WB, LRA	Human PC3 and Du145 prostate carcinoma cells	Tumour growth	Steele et al. (2010)
miR-29b	MMP-9	qPCR, WB, LRA	Human MDA231 and mouse 4T1 breast cancer cells, primary adenoma and carcinoma cells from MMTV-Py MT mice	Cancer metastasis	Chou et al. (2013)
miR-491-5p	MMP-9	LRA	Human U251 and U87 glioma cells	Cancer cell invasion	Yan et al. (2011)
miR-31	MMP-16, integrin α5	qPCR, WB, LRA	Human MDA-MB-231 breast adenocarcinoma cells	Cancer metastasis	Valastyan et al. (2009)
miR-181b	TIMP-3	qPCR, WB, LRA	Human hepatoma Hep3b and Sk-Hep1 hepatic adenocarcinoma cells	Tumorigenesis	Wang et al. (2010a)
mir-221/222	TIMP-3	qPCR, WB, LRA	Human MEG01 megakaryoblastic leukaemia and NSCLC cells	Tumorigenesis and metastasis	Garofalo et al. (2009)
miR-29b	COL1A1, COL1A2, integrin β1	WB, qPCR	Mouse primary hepatic stellate cells	Liver fibrosis	Sekiya et al. (2011)
miR-29b	COL1A1	qPCR, LRA*	COS monkey fibroblasts, primary mouse CFs	Myocardial fibrosis	van Rooij et al. (2008)
miR-133a	COL1A1	qPCR, LRA	HEK293 cells, rat heart	Myocardial fibrosis	Castoldi et al. (2012)
miR-29b	COL1A2, COL5A2, elastin, MMP-2	WB, qPCR proteomics	Primary mouse CFs	Cardiac fibrosis	Abonnenc et al. (2013)
miR-29a, miR-29b, miR-29c	COL1A1, COL3A1	qPCR, WB, LRA*	Primary human systemic sclerosis fibroblasts, HEK293 cells	Systemic sclerosis	Maurer et al. (2010)
miR-27b	MMP-13	LRA, ELISA	Primary human OA chondrocytes, HeLa cells	OA	Akhtar et al. (2010)
miR-9	MMP-13	ELISA	Primary human OA chondrocytes	OA	Jones et al. (2009)
miR-140	MMP-13	qPCR, WB, LRA	Human IL-1β-stimulated cartilage C28/I2 cells	OA	Liang et al. (2012)
miR-146a	Aggrecan, ADAMTS5, MMP-13, COL2A1	qPCR, WB	Primary human OA chondrocytes	OA	Li et al. (2011b)
miR-155	MMP-1, MMP-3	qPCR, ELISA	Primary human RA SFs	RA	Stanczyk et al. (2008)
miR-203	MMP-1	qPCR, ELISA	Primary human RA SFs	RA	Stanczyk et al. (2011)
miR-146a	Fibronectin	qPCR, ELISA, LRA	Type I and II diabetic rat retina, heart and kidney	Diabetes	Feng et al. (2011)
miRNA	miRNA direct target	Validation method	System	ECM molecule	Process	Reference
Indirect regulation
Development
miR-145	SOX9	qPCR, WB, LRA	HEK293 cells, C3H10T1/2 mesenchymal stem cells	COL2A1, aggrecan, COMP, COL9A2, COL11A1	Chondrogenesis	Yang et al. (2011)
		qPCR, WB, LRA	Primary HACs	COL2A1, aggrecan		Martinez-Sanchez et al. (2012)
miR-675	?	qPCR, WB	Primary HACs	COL2A1	Chondrogenesis	Dudek et al. (2010)
miR-1247	SOX9	qPCR, WB, LRA*	Primary HACs	COL2A1	Chondrogenesis	Martinez-Sanchez and Murphy (2013)
miR-488	?	Gelatin zymography	Primary chicken limb mesenchymal cells	MMP-2	Chondrogenesis	Song et al. (2011)
Disease
miR-340	c-Met	qPCR, WB, LRA*	Human MDA-MB-231 and MCF-7 breast adenocarcinoma cells	MMP-2, MMP-9	Cancer cell invasion and metastasis	Wu et al. (2011)
miR-218	IKK-β	qPCR, WB, LRA	Human U87MG and SNB19 glioma cells	MMP-9	Cancer cell invasion	Song et al. (2010)
miR-10b	HOXD10	WB, LRA	Human U87 and U251 glioma cells	MMP-14	Cancer cell invasion	Sun et al. (2011)
miR-21	?	MA, qPCR, WB, LRA	Human HeLa cervical cancer, A172 and LN229 glioma, MCF-7 breast carcinoma and U2OS osteosarcoma cells	TIMP-3	Cancer cell invasion	Gabriely et al. (2008)
		qPCR, WB	Human CCA biopsies, human CCA-derived CAK-1, TFK1 and HuCCT1 cell lines			Selaru et al. (2009)
miR-27a	?	qPCR	Primary human OA chondrocytes	MMP-13	OA	Tardif et al. (2009)
miR-22	PPARA, BMP7	qPCR, WB, IC, LRA*	Primary human OA chondrocytes	MMP-13	OA	Iliopoulos et al. (2008)
miR-101	SOX9	qPCR, WB, LRA	Primary rat IL-1β-stimulated chondrocytes	Col-2, aggrecan	OA	Dai et al. (2012)
miR-19a/b	TLR2	qPCR, RT-PCR, WB, LRA	Primary human RA SFs	MMP-3	RA	Philippe et al. (2012)
miR-192	Zeb2	qPCR, LRA*	Primary mouse mesangial cells	COL1A2	Diabetic nephropathy	Kato et al. (2007)
miR-216a, miR-217	PTEN	qPCR, WB, LRA	Primary mouse mesangial cells	COL1A2	Diabetic nephropathy	Kato et al. (2009)
miR-377	PAK1, SOD1, SOD2	MA, WB, LRA*	Primary human mesangial cells	Fibronectin	Diabetic nephropathy	Wang et al. (2008b)

WB, Changes in the protein levels of a target gene measured by Western blot after miRNA transfection or inhibition; LRA, luciferase reporter gene assay including wild-type and mutant 3′ UTR; LRA*, luciferase reporter gene assay including only wild-type or wild-type and reversed 3′ UTR; qPCR, changes in the level of target mRNA measured by quantitative PCR after miRNA transfection or inhibition; SMC, smooth muscle cells; HACs, human articular chondrocytes; IHC, changes in the protein levels of a target gene measured by immunohistochemistry after miRNA transfection or inhibition; RT-PCR, changes in the level of target mRNA measured by RT-PCR after miRNA transfection or inhibition; HCC, hepatocellular carcinoma; FNDC3B, fibronectin type III domain containing 3B; MA, microarray analysis: high-throughput measurement of the change in the level of target mRNA after miRNA transfection or inhibition; FC, changes in the cell surface protein levels of a target gene were measured by flow cytometry after miRNA transfection; OSCC, oral squamous cell carcinoma; IC, changes in the protein levels of a target gene were measured by immunocytochemistry after miRNA transfection; HEK293, human embryonic kidney 293; HDF, human diploid fibroblast; HTM, human trabecular meshwork; HUVEC, human umbilical vein endothelial cell; Ago2 IP, anti-Ago2 antibody immunoprecipitation followed by microarray analysis; LAMP, labelled miRNA pull-down assay followed by microarray analysis; HOS, human osteogenic sarcoma; CF, cardiac fibroblast; RA SF, rheumatoid arthritis synovial fibroblast; RA, rheumatoid arthritis; COMP, cartilage oligomeric matrix protein; c-Met, proto-oncogene c-Met or hepatocyte growth factor receptor; HOXD10, homeobox D10; CCA, cholangiocarcinoma; PPARA, peroxisome proliferator-activated receptor alpha; BMP7, bone morphogenetic protein 7; Col-2, type II collagen; TLR2, Toll-like receptor 2; PTEN, phosphatase and tensin homolog; PAK1, p21-activated kinase; SOD, superoxide dismutase.

miRNA-mediated regulation of the ECM during development

As cells divide and differentiate and tissues are formed, cells dynamically communicate with each other and with the ECM via junctions, receptors and soluble factors. This mutual exchange of information between cells and their microenvironment directs organogenesis during development and allows the functional organization of cells into tissues.

Organogenesis

The ECM contributes to the regulation of several processes in organogenesis, including cell adhesion, proliferation, migration, survival and differentiation. For instance, fibronectins are glycoproteins essential for embryonic development as demonstrated by the complete knockout of the fibronectin gene. Fibronectin null mice die around embryonic day 8.5 because of severe defects in mesodermally derived tissues (George et al. 1993). By binding to integrins, fibronectin mediates cell adhesion, proliferation and tissue development (Jiang et al. 2003; Karaulanov et al. 2006; Rifes et al. 2007). Expression of fibronectin and fibronectin type III domain containing 3A (FNDC3A) was found to be downregulated in cells and tissues from transgenic mice expressing miR-17 (Shan et al. 2009), a miRNA that belongs to the six-miRNA cluster miR-17∼92 and has been shown to play important roles in tissue development and cancer (Mendell 2008). miR-17 transgenic mice showed retarded growth rates of the heart, liver, spleen and whole body as a result of the repression of fibronectin and FNDC3A (Shan et al. 2009); however, it is possible that other mRNAs are also targeted by miR-17.

Versican is a chondroitin sulphate proteoglycan, whose expression is prominent in the heart and vascular system (Yao et al. 1994). By virtue of its ability to modulate cell differentiation, migration, proliferation and apoptosis, versican plays a role in cardiomyocyte differentiation and heart development (Henderson & Copp 1998; Wight 2002; Chan et al. 2010). Its expression during development has been shown to be regulated by a number of miRNAs, including miR-138 and miR-143. In the zebrafish heart, miR-138 was shown to coordinate appropriate chamber-specific gene expression patterns. It specifically restricted versican expression to the atrioventricular canal, which separates atrial and ventricular chambers, by repressing its expression in the ventricle in two ways: (i) direct repression of retinoic acid synthesis via aldehyde dehydrogenase-1a2, which would otherwise induce expression of versican, and (ii) direct targeting of versican mRNA (Morton et al. 2008). Furthermore, versican expression was suppressed by myocardin-induced expression of miR-143 during the differentiation of smooth muscle cells, which, as a consequence, lost their ability to migrate. In support of these findings, inhibition of miR-143 expression by PDGF treatment induced versican expression (Wang et al. 2010c).

Collagen deposition in developing tissues and organs is regulated by the orchestrated action of several MMPs and tissue inhibitor of metalloproteinases (TIMPs). MMPs are regulated at least in part by miRNAs. For instance, during mammary gland development, the miR-212/132 family was shown to be specifically expressed in mammary stroma where it directly targeted MMP-9. Ablation of these miRNAs resulted in an increased expression of MMP-9, which accumulated around the ducts, thereby interfering with collagen deposition and impairing ductal outgrowth (Ucar et al. 2010).

Chondrogenesis

During development, mesenchymal cells aggregate and differentiate into chondrocytes. With the exception of chondrocytes in articular cartilage, which are committed to maintain a functional cartilage layer, chondrocytes undergo a series of differentiation processes, including proliferation, hypertrophy, terminal differentiation, mineralization and programmed cell death. Rounds of chondrocyte cell division and ECM secretion into growth plates lead to cartilage growth (Mackie et al. 2011). Throughout chondrogenesis, ECM production and deposition patterns change dramatically, generating mesenchymal, precartilaginous and cartilaginous matrices, all distinct ECMs [reviewed in Singh & Schwarzbauer (2012)]. Emerging evidence suggests that miRNAs contribute to the regulation of the changes in the expression of ECM proteins during chondrogenesis.

Research in this field has focussed largely on miR-140, which was identified as a cartilage-specific miRNA in developing zebrafish (Wienholds et al. 2005) and during murine skeletal development (Tuddenham et al. 2006). In humans, miR-140 expression increased, in parallel with SOX9 and COL2A1, during chondrogenesis in differentiated mesenchymal cells compared to undifferentiated control cells. Targets of miR-140 included the major cartilage aggrecanase ADAMTS5, whose repression led to an increased expression of aggrecan, the most abundant proteoglycan in the cartilage (Miyaki et al. 2009, 2010).

Interestingly, ADAM9, a membrane-anchored protease known to be implicated in a variety of biological processes involving cell–cell and cell–ECM interactions and in the ectodomain shedding of membrane-anchored heparin-binding EGF-like growth factor (Nath et al. 2000; Tao et al. 2010), was reported as a direct target of miR-142-3p during chondrogenic differentiation of chick limb mesenchymal cells. Its expression was regulated in wing and leg mesenchymal cells and was shown to contribute to the modulation of position-dependent chondrogenesis by repressing chondrogenic differentiation of mesenchymal cells (Kim et al. 2011). In this experimental setting, ADAM9 inhibited cell migration, decreased total cell number and increased apoptotic cell death, suggesting that ADAM9 may also mediate cell–cell communication to generate critical survival signals.

Indirect, but nonetheless significant, regulation of ECM protein expression by miRNAs during chondrogenesis has also been reported. One example is miR-145, which was shown to suppress chondrogenic differentiation of murine embryonic mesenchymal cells by directly targeting SOX9, a master positive regulator of chondrogenesis. Its decreased expression at early stage of chondrogenic differentiation was accompanied by an increased expression of COL2A1, aggrecan, cartilage oligomeric matrix protein (COMP), collagen type IX (COL9A2) and collagen type XI (COL11A1). Conversely, overexpression of miR-145 led to decreased levels of these ECM molecules (Yang et al. 2011). In line with these findings, Martinez-Sanchez et al. reported that miR-145 directly regulates SOX9 in human normal articular chondrocytes and its overexpression reduced the levels of SOX9 along with COL2A1 and aggrecan, while increasing the hypertrophic markers RUNX-2 and MMP-13 (Martinez-Sanchez et al. 2012). Other examples of miRNAs that indirectly regulate the ECM during chondrogenesis stem from work conducted by the same group. They reported that miR-675 was tightly regulated by SOX9 during chondrocyte differentiation and upregulated the expression of COL2A1 (Dudek et al. 2010). Recently, they found that miR-1247 expression correlated with the differentiation status of healthy human articular chondrocytes and had a significant impact on COL2A1 production through its direct targeting of SOX9 (Martinez-Sanchez & Murphy 2013).

Thus, a subset of miRNAs that regulate chondrogenesis has been identified. Interestingly, ADAM and ADAMTS metalloproteinases have emerged as direct miRNA targets, while a miscellaneous group of ECM molecules, including aggrecan, MMPs and distinct types of collagen, are indirectly modulated by a number of miRNAs through their regulation of SOX9.

Osteogenesis

In the foetus, most of the skeleton is composed of cartilage. However, as the foetus grows, osteoblasts and osteoclasts replace cartilage cells and ossification begins with the deposition of calcium and phosphate salts. The ECM of the bone differs from that of cartilage not only for the presence of crystalline salts, but also in the composition of the collagen fibres, which are mostly made of collagen type I and some type V, and in the content of proteoglycans, which is less than that in cartilage.

Nephronectin is a secreted, integrin α8β1-binding ECM protein that is expressed in several tissues in the developing mouse embryo, including the long bone (Brandenberger et al. 2001). Kahai et al. used MC3T3-E1 cells, an established in vitro model system for studying bone development, to determine the function of nephronectin in osteogenesis. They found that expression of this ECM protein promotes osteoblast differentiation and bone nodule formation via its epidermal growth factor (EGF)-like repeats (Kahai et al. 2010). They also demonstrated that miR-378 can bind the 3′ UTR of nephronectin mRNA, thereby enhancing nephronectin expression and modulating osteoblast differentiation. Interestingly, overexpression of the 3′ UTR of nephronectin resulted in higher nephronectin glycosylation and secretion and thus in increased rate of osteoblast differentiation. This appeared to be the consequence of competition for miR-378 binding between nephronectin and UDP-N-acetyl-alpha-D-galactosamine:polypeptide N-acetylgalactosaminyltransferase 7 (GalNT7), an enzyme known to glycosylate proteins (Kahai et al. 2009; Lee et al. 2011). In addition to miR-378, other miRNAs such as miR-23a, miR-101a, miR-296-5p, miR-328, miR-340-3p and miR-425 can directly target nephronectin 3′ UTR; however, further studies are required to unravel the precise mechanism by which altered expression of nephronectin by miRNAs regulates osteoblast differentiation.

Several miRNAs have been demonstrated to play an important role in tissue development through the regulation of the expression of ECM proteins that are involved in ECM remodelling. However, neither the spatiotemporal expression nor the function of these miRNAs during distinct stages of development is clear. Furthermore, it is not well defined whether these miRNAs play a role exclusively in the initiation or maintenance of tissue differentiation or both.

miRNA-mediated regulation of the ECM in disease

The ECM plays an active role in the development and progression of disease. In addition to the recognized link between the ECM and cancer, the contribution of the ECM to other diseases such as arthritis, fibrosis and metabolic disorders has also emerged. Here, we discuss how miRNAs regulate the ECM in these diseases.

Cancer

Genetic damage can lead to uncontrolled cell division and growth, the hallmark of cancer. Primary cancer cells can spread from one site to another through the lymphatic system or bloodstream, a process known as metastasis. The ECM, a component of the tumour microenvironment, is a key determinant of oncogenesis, cancer progression and metastasis as it affects cell growth, survival, adhesion, motility and invasion.

(i) Collagen. Increased deposition of fibrillar collagen such as collagen type I enhances ECM rigidity, which promotes the loss of epithelial polarity and tumour progression. Abnormal extracellular levels of collagens and also laminins have been shown to promote invasiveness of many tumour cells in vitro and enhance the incidence of metastases in animal models (Kuratomi et al. 1999, 2002; Koenig et al. 2006; Shintani et al. 2006). Shi et al. reported that miR-301 overexpression promoted human breast cancer cell migration and invasion by downregulating collagen type II alpha 1 (COL2A1) besides enhancing cell proliferation and tumour growth by repressing the tumour suppressors FoxF2, BBC3 and PTEN (Shi et al., 2011).

In metastatic hepatocellular carcinomas, let-7g expression is significantly lower than that in hepatocellular carcinoma without metastasis and it was found to inhibit cell migration and colony formation by regulating the 3′ UTR of COL1A2, which was experimentally validated as a direct target of let-7g. COL1A2 and let-7g levels were inversely correlated in clinical samples and the addition of COL1A2 counteracted the inhibitory effect of let-7g on cell migration. This work suggests that let-7g is able to suppress hepatocellular carcinoma metastasis at least in part through targeting COL1A2 (Ji et al. 2010).

When screening nasopharyngeal carcinomas, Sengupta et al. found that miR-29c was significantly downregulated compared to healthy nasopharyngeal epithelial samples. They further showed that miR29c was able to target multiple collagens, including COL1A1, COL1A2, COL3A1, COL4A1, COL4A2 and COL15A1, and laminin γ1. Lower levels of miR-29c in tumours corresponded to increased levels of these ECM molecule mRNAs, whose expression has been associated with augmented invasiveness and metastatic potential (Sengupta et al. 2008).

(ii) Fibronectin and tenascin-C. Not only collagens and laminins, but also fibronectin levels are dysregulated in tumour cells with high metastatic potential (Urtreger et al. 2006). Fibronectin plays a crucial role in cell adhesion and motility and thus represents an important determinant of metastatic success of cancer cells. Regulation of fibronectin type III domain containing 3B (FNDC3B) by miR-143 was reported in the metastasis of hepatitis B virus-related hepatocellular carcinoma (HBV-HCC) both in mice and in patients. In particular, miR-143, which was shown to be transcribed by NF-kB, was overexpressed in HBV-HCC and directly targeted FNDC3B, leading to the repression of its expression and thus promoting cancer cell invasion and migration and tumour metastasis (Zhang et al. 2009).

Tenascin-C is an adhesion modulatory glycoprotein that is highly expressed in the microenvironment of most solid tumours. It is involved in tumour initiation and progression by affecting malignant transformation, proliferation, angiogenesis, metastasis and escape from tumour immunosurveillance (Orend & Chiquet-Ehrismann 2006). Breast cancer cell metastasis and migration was shown to be suppressed by miR-335, which directly targeted the 3′ UTR of tenascin-C, along with the transcription factor SOX4 (Tavazoie et al. 2008). No further miRNAs have been reported that target tenascin-C or other members of the tenascin family.

(iii) ECM receptors: integrins and CD44. Alterations in integrin expression have a significant impact on cancer progression and metastasis (Cabodi et al. 2010). Integrins are a large family of ECM receptors involved in cell–cell and cell–matrix interactions. As such, integrins play prominent roles in intracellular signalling, cytoskeleton organization as well as cell adhesion and migration (Humphries et al. 2006). Aberrant expression of miRNAs has been implicated in the dysregulation of integrin expression and function, promoting the development, progression and metastatic potential of cancer [reviewed in Valastyan & Weinberg (2011)]. Augoff et al. identified miR-31 as a master regulator of integrins that targeted multiple α subunit partners (α2, α5, αv) of β1 and β3 integrins. Expression of miR-31 in cancer cells significantly repressed these integrin subunits both at the mRNA and at the protein levels, inhibiting cell spreading on specific target ECM ligands of these affected integrins (Augoff et al. 2011). By targeting this cohort of pro-metastatic integrins, miR-31 acts as a metastasis suppressor gene. Among these integrin subunits, β1 integrin was found to be regulated by a number of other miRNAs and a comprehensive summary of these studies can be found in Table1.

Higher levels of β8 integrin are associated with cell death in tumour mass and in human glioblastoma. Downregulation of β8 integrin by miR-93 in human glioma cells was shown to promote tumour growth and angiogenesis. Strikingly, when miR-93-overexpressing glioma cells were co-cultured with endothelial cells, they supported endothelial cell spreading, growth, migration and tube formation. Furthermore, in vivo studies revealed that miR-93-expressing cells induced blood vessel formation, allowing blood vessels to extend to tumour tissues in high densities. Angiogenesis promoted by miR-93 in turn facilitated cell survival, resulting in enhanced tumour growth (Fang et al. 2011).

Another key ECM receptor involved in cell migration, adhesion and cell–cell and cell–matrix interaction is CD44, which can interact with a variety of ECM ligands, including hyaluronan, collagens, osteopontin and versican. A role for CD44 and its variant isoforms has been well established in cancer (Gunthert et al. 1991; Wielenga et al. 1993; Marhaba & Zoller 2004; Afify et al. 2009), and evidence for miRNA-mediated regulation of CD44 expression is emerging. However, it is not clear which specific isoforms of CD44 are targeted by distinct miRNAs. Huang et al. subjected a non-metastatic, human breast cancer cell line to a transwell migration assay following transduction with a miRNA expression library. They found that miR-373 and miR-520c, which share similar seed sequence, stimulated cancer cell migration both in vitro and in vivo through direct suppression of CD44. Furthermore, levels of miR-373 inversely correlated with those of CD44 in clinical breast cancer metastasis samples (Huang et al. 2008). The 3′ UTR of CD44 mRNA was also found to be a target of miR-328. The expression of this miRNA in a human squamous carcinoma cell line resulted in reduced cell adhesion, aggregation and migration, which was mechanistically explained by the direct suppression of CD44 (Wang et al. 2008a). To determine whether miRNAs regulate CD44 expression in cancer, Liu et al. took a different approach. They purified CD44⁺ prostate stem cells from xenografts and primary cancers and screened them by qRT-PCR for the expression of 324 miRNAs. They found that CD44 was a direct and functional target of miR-34a, a miRNA regulated by p53, which was underexpressed in CD44⁺ prostate stem cells. While enforced expression of miR-34a in CD44⁺ cells inhibited clonogenic expansion and metastasis, inhibition of miR-34a in CD44⁻ cells promoted cancer development and metastasis. In vivo, systemic delivery of miR-34a inhibited prostate cancer metastasis and extended survival of mice with cancer (Liu et al. 2011). This work revealed miR-34a as a key negative regulator of CD44⁺ prostate stem cells.

(iv) Matrix metalloproteinases. Matrix metalloproteinases (MMPs) accomplish ECM turnover and degradation, and their expression in several human cancers has been associated with cancer cell invasion and metastasis. MMP-13 expression was shown to be under the control of miR-143 in osteosarcoma metastasis in the lung. Downregulation of miR-143 correlated with the lung metastasis of human osteosarcoma cells by promoting cellular invasion at least partially via MMP-13 upregulation (Osaki et al. 2011). Pro-MMP-2, a gelatinase also known as type IV collagenase, was found to be repressed by miR-29b in prostate cancer cells (Steele et al. 2010). MMP-2, along with MMP-9, was shown to be indirectly regulated by miR-340 in a number of human breast cell lines. miR-340 targeted the oncogene c-Met, which promoted cell migration and invasion through the modulation of the expression of MMP-2 and MMP-9, both of which can degrade collagen type IV and mediate tumour cell invasion (Wu et al. 2011).

Upregulation of MMP-9 has been associated with the pathogenesis and progression of gliomas (Kondraganti et al. 2000). Furthermore, the transcription factor NF-kB has been shown to induce the expression of MMP-9 by binding to its promoter (Farina et al. 1999). In line with this, Song et al. found that miR-218 inhibited the invasive ability of glioma cells by directly targeting IKK-β. Suppression of the NF-kB pathway, in turn, resulted in downregulation of MMP-9 (Song et al., 2010). On the other hand, Yan et al. showed that the 3′ UTR of MMP-9 was directly targeted by a distinct miRNA, miR-491-5p, which suppressed glioma cell invasion (Yan et al., 2011). Recently, MMP-9 expression in breast cancer has been demonstrated to be modulated by miR-29b, which was shown to bind to a specific sequence motif in the 3′ UTR of MMP-9 mRNA (Chou et al. 2013).

MMP-14 and MMP-16, two members of the membrane-type metalloproteinase (MT-MMP) family, were also shown to promote cancer cell invasion in glioma and breast cancer respectively. On the one hand, MMP-14 expression was regulated by miR-10b through direct suppression of homeobox D10 (HOXD10), which further regulated RhoC, a pro-metastasis gene belonging to the Ras homolog gene family (Sun et al. 2011). On the other hand, MMP-16 expression was regulated by miR-31, which directly targeted the 3′ UTR of MMP-16, along with that of integrin α5, frizzled3, radixin and RhoA. Thus, by repressing a subset of metastasis-promoting genes, miR-31 inhibited several steps of metastasis, including local invasion, extravasation, survival at a distant site and metastatic colonization (Valastyan et al. 2009).

(v) Tissue inhibitor of matrix metalloproteinases. Finally, tissue inhibitor of matrix metalloproteinases (TIMPs), specific endogenous MMP inhibitors, have also been reported to be regulated by miRNAs in cancer. Research in this field has focussed mainly on TIMP-3, which inhibits angiogenesis and tumour cell infiltration and induces apoptosis (Baker et al. 1999; Qi et al. 2003). TIMP3 mRNA and protein levels were found to be modulated by miR-21 expression in glioma cells. However, luciferase miRNA target reporter assays showed that TIMP3 is not a direct target of miR-21, suggesting that TIMP3 is likely to be a downstream effector of miR-21 (Gabriely et al. 2008). In contrast, other studies demonstrated that the 3′ UTR of TIMP3 mRNA can be directly targeted by different miRNAs in cancer. For example, in hepatocellular carcinoma, TIMP3 was directly targeted by miR-181b. Repression of TIMP3 expression led to enhanced MMP-2 and MMP-9 activity and promoted growth, clonogenic survival, migration and invasion of cancer cells that could be reversed by overexpressing TIMP3 (Wang et al. 2010a). Further examples of direct regulation of TIMP3 by miRNAs are listed in Table1.

Fibrosis

Fibrosis is a physiologically occurring process that involves the formation of fibrous connective tissue in response to injury. However, excessive deposition of fibrous tissue is the hallmark of several fibrotic diseases.

Liver fibrosis is a common feature of chronic liver injury. Hepatic stellate cells (HSCs) represent the primary cell population that contributes to fibrinogenesis by accumulating excessive ECM proteins including collagen. Normally quiescent, HSCs become activated upon liver injury and differentiate into myofibroblast-like cells, which are proliferative cells that express α-smooth muscle actin (α-SMA) and secrete profibrogenic mediators and ECM proteins (Friedman 2000). Recently, a link between fibrosis and miRNAs has also been established (Thum et al. 2008; Bauersachs 2010; Bowen et al. 2013). Sekiya and colleagues investigated the roles of miR-29b, a component of the miR-29 family, in the activation of murine primary HSCs. They found that miR-29b expression was downregulated during HSCs activation in primary culture, whereas transfection of a miR-29b precursor markedly reduced the expression of α-SMA, profibrogenic genes as well as COL1A1 and COL1A2 mRNAs (Sekiya et al. 2011).

Beyond liver fibrosis, miR-29b has been shown to play an important role also in myocardial fibrosis by targeting genes involved in the fibrotic process, including COL1A1, the main collagen type expressed in the heart (van Rooij et al. 2008). More recently, miR-133a has been shown to play a protective role against myocardial fibrosis (Matkovich et al. 2010) and Castoldi et al. demonstrated that miR-133a directly targets COL1A1 in myocardial fibrosis caused by angiotensin II-dependent hypertension in rats (Castoldi et al., 2012). In the presence of myocardial fibrosis in this experimental model, expression of both miR-29b and miR-133a was downregulated, while COL1A1 expression was upregulated, suggesting a regulatory mechanism whereby downregulation of these miRNAs may contribute to heart injury leading to myocardial fibrosis as a result of increased COL1A1 expression (Castoldi et al. 2012). This year, a comprehensive proteomic screening, which compared the effects of miR-29b and miR-30c on ECM secretion by cardiac fibroblasts, was published. Mouse cardiac fibroblasts were transfected with pre-/anti-miR-29b and -miR-30c, and their conditioned medium was analysed by mass spectrometry. miR-29b targeted several proteins involved in fibrosis, including multiple collagens and MMPs, in addition to predicted targets of miR-29b, including insulin-like growth factor 1. miR-29b also attenuated the cardiac fibroblast response to TGF-β. In contrast, miR-30c had little effect on ECM production. However, both miRNAs affected cardiac myocytes. Upon transfection with pre-miR-29b, the conditioned medium of cardiac fibroblasts lost its ability to support adhesion of rat ventricular myocytes, whereas pre-miR-30c conditioned medium had a prohypertrophic effect (Abonnenc et al. 2013). These data suggest that miR-29b not only regulates ECM protein deposition but also contributes to the development of cardiac hypertrophy by targeting key growth factors and inflammatory mediators.

Systemic sclerosis is a rare, chronic disorder characterized by scarring in the skin, joints and internal organs. Normal tissue is progressively replaced by collagen-rich ECM (Maurer et al. 2010), resulting in fibrosis of the skin and internal organs. This seems to be caused by the transition of quiescent fibroblasts to activated myofibroblasts, which overproduce collagens, including types I, III and V, collagen-modifying enzymes and other ECM components (Asano 2010). The expression of various ECM proteins, mainly collagen type I, is upregulated in fibroblasts in patients with systemic sclerosis (Jinnin 2010). The miRNA-29 family has been shown to target COL1A1 and COL3A1 and has been found to be strongly downregulated in skin biopsy and fibroblast samples from patients with systemic sclerosis (Maurer et al. 2010).

Arthritis

Osteoarthritis (OA) is the most common form of arthritis, and it is referred to as degenerative joint disease. Articular cartilage destruction occurs as a result of an imbalance between the synthesis and the degradation of its major ECM components, COL2 and the proteoglycan aggrecan. Matrix-degrading enzymes (e.g. MMPs and ADAMTS) play important roles in this degenerative process. Production and maintenance of the ECM in cartilage is mediated by chondrocytes, the only cell type in cartilage, and is highly regulated by transcription factors, such as SOX9, and growth factors, such as TGF-β and bone morphogenetic proteins (BMPs; Loeser et al. 2012). miRNAs have also emerged as regulators of gene expression in chondrocytes during OA (Miyaki & Asahara 2012).

Induced during chondrogenesis, the expression of miR-140 was shown to be reduced in OA. This resulted in higher levels of ADAMTS5 and, in turn, lower levels of aggrecan in the cartilage (Miyaki et al. 2009; Tardif et al. 2009). In line with these observations, miR-140-null mice display early onset OA-like disease and greater susceptibility to surgically induced OA (Miyaki et al. 2010). In the same year, a study was published on primary OA chondrocytes that were stimulated with IL-1β, a cytokine that strongly induces the synthesis of ECM-degrading enzymes in OA, and screened for their miRNA expression profile compared to control samples. Among dysregulated miRNAs, miR-27b expression was significantly downregulated and was found to directly regulate the expression of the collagenase MMP-13 (Akhtar et al. 2010). Two screens of miRNA expression in normal vs. OA cartilage have been reported by independent groups. Iliopoulos et al. integrated miRNA and proteomic data with miRNA gene-target prediction algorithms generating a potential ‘interactome’ network consisting of 11 miRNAs and 58 proteins linked by 414 potential functional associations. Experimental validation revealed that miR-22 regulated MMP-13 and IL-1β expression by directly targeting PPARA and BMP7 and its inhibition blocked inflammatory and catabolic changes in OA chondrocytes (Iliopoulos et al., 2008). In a second screen, Jones et al. showed that miR-9 expression was dysregulated in OA samples and its overexpression and inhibition modulated MMP-13 protein levels in primary human OA chondrocytes (Jones et al., 2009). However, a direct regulation of MMP-13 by miR-9 was not demonstrated. Later, MMP-13 expression was shown to be directly regulated by miR-140 in vitro (Liang et al. 2012). Furthermore, miR-146a was also shown to be involved in cartilage degeneration in OA (Yamasaki et al. 2009) and to repress the expression of several ECM molecules, including aggrecan, ADAMTS5, MMP-13 and COL2A1 (Li et al. 2011b). Finally, a recent study showed that miR-101 was involved in IL-1β-induced downregulation of COL2 and aggrecan expression, probably through the repression of its target gene SOX9. In support to this, inhibition of miR-101 prevented IL-1β-induced chondrocyte ECM degradation (Dai et al. 2012). Notably, among these miRNAs, miR-9 and miR-146a play a key role in the regulation of inflammation by directly targeting the p50 subunit of NF-kB and IL-1R-associated kinase 1 (IRAK1) and TNFR-associated factor 6 (TRAF6) respectively (Taganov et al. 2006; Bazzoni et al. 2009). Considering that OA is emerging as an inflammatory disease and that the NF-kB signalling pathway is a major catabolic pathway in OA joints, the involvement of these miRNAs in OA comes as no surprise. However, it raises the question of whether other joint tissues are also a source of these miRNAs in addition to cartilage. Moreover, given that these miRNAs may play a protective role in OA, further work is required to exactly determine at which stage of initiation or progression of OA their expression is dysregulated.

Rheumatoid arthritis (RA) is a form of inflammatory arthritis and an autoimmune disease. Main features of RA include synovial inflammation and destruction of joint cartilage and bone mediated by persistent synthesis of proinflammatory cytokines and matrix-destructive enzymes. As in the case of OA, miRNAs have been shown to play a role in the pathogenesis of RA by regulating the expression of genes involved in the synthesis and breakdown of the ECM (Baxter et al. 2012). Among the most studied miRNAs in RA, miR-155 and miR-203 were all found to be expressed at high levels in synovial tissues and synovial fibroblasts isolated from patients with RA, compared to healthy controls (Stanczyk et al. 2008, 2011). Stanczyk et al. demonstrated the involvement of miR-155 and miR-203 in the regulation of the RA synovial fibroblast (RASF) phenotype. Upregulation of miR-155 in RASFs was found to indirectly repress the levels of the collagenase MMP-1 and the stromelysin MMP-3, which can degrade fibronectin, laminin and proteoglycan core proteins in addition to collagen types IV, V, IX and X and reduce the tissue damage (Stanczyk et al. 2008). In contrast, upregulation of miR-203 in RASFs elevates levels of MMP-1 and IL-6, promoting tissue inflammation (Stanczyk et al. 2011). Finally, Philippe et al. reported that the secretion of MMP-3 was strongly downregulated in RASFs that were transfected with either mimic of miR-19a/b. The regulatory function exerted by miR19a/b on MMP-3 resulted from the direct suppression of Toll-like receptor 2 (TLR2; Philippe et al. 2012), whose activation leads to the induction of MMPs (Medzhitov & Janeway 2002).

Metabolic disorders

Diabetes is one group of metabolic diseases characterized by hyperglycaemia that can be due to insufficient insulin production or aberrant cellular response to the insulin that is produced. Metabolic alterations in hyperglycaemia lead to the activation of gene programmes that compromise the functionality of endothelial cells as well as other tissues affected by chronic diabetes complications, including the retina, kidney and heart. For instance, excessive production of ECM proteins such as fibronectin is a common feature of all diabetes complications (Roy et al. 1996; Chen et al. 2003). Increased fibronectin production in diabetes is known to occur through histone acetylator p300 (Kaur et al. 2006). However, a recent study uncovered a novel molecular mechanism in which miR-146a contributes to the regulation of fibronectin expression in diabetic animals. Precisely, glucose downregulated miR-146a expression resulting in increased p300 and fibronectin expression levels in endothelial cells as well as retina, kidney and heart from animals with type 1 and type 2 diabetes. In line with this, administration of a miR-146a mimic restored retinal miR-146a expression and decreased fibronectin levels in diabetes (Feng et al. 2011).

Diabetic nephropathy, a progressive kidney disease, is a major complication of longstanding diabetes. It is characterized by glomerular mesangial expansion, hypertrophy and accumulation of ECM proteins such as collagen and fibronectin in the kidney (Mason & Wahab 2003). In diabetic kidneys, the phosphatidylinositol 3-kinase (PI3K)–Akt pathway is activated by TGF-β1 and plays important roles in fibrosis, hypertrophy and cell survival in glomerular mesangial cells (MMCs; Runyan et al. 2004; Yi et al. 2005). In TGF-β1-treated mouse MMCs and diabetic mouse glomeruli, expression of miR-192 has been shown to be upregulated and to increase COL1A2 expression by downregulating Zeb2, an E-box repressor (Kato et al. 2007). Later, the same group found that this circuit was amplified by other miRNAs and reported that TGF-β1 activates Akt in MMCs by inducing the transcription of miR-216a and miR-217, which target PTEN, an inhibitor of Akt activation. Akt activation by these miRNAs in MMCs resulted into biological effects similar to those of activation by TGF-β1, including ECM gene expression, cell survival and hypertrophy, all features of diabetic nephropathy (Kato et al. 2009). A different approach was taken by Wang et al., who investigated the role of miRNAs in the regulation of ECM protein expression in mesangial cell exposed to high glucose rather than TGF-β. Under these conditions in vitro as well as in murine diabetic nephropathy models in vivo, they found that miR-377 was upregulated relative to controls as it was fibronectin. They further showed that miR-377 did not directly target fibronectin, but suppressed translation of p21-activated kinase (PAK1) and superoxide dismutases 1 and 2 (SOD1 and SOD2) leading to oxidative stress and Smad activation, which ultimately enhanced fibronectin protein synthesis (Wang et al. 2008b).

Notably, miR-29 and miR-31 stand out among the many miRNAs that are reported to regulate the ECM in disease. The miR-29 family, which consists of 3 members that share the same seed sequence (miR-29a, miR29b and miR-29c), suppresses the expression of major ECM genes, including six different types of collagen, laminin γ1 and MMP-2, functioning as a master regulator of the ECM. Conversely, miR-31 regulates the expression of multiple α subunit partners (α2, α5, αv) of β1 and β3 integrins, acting as a master orchestrator of ECM receptors. Remarkably, miR-29 and miR-31 coordinate the expression of a network of genes involved in modulating multiple processes in cancer and fibrosis, including cell adhesion and motility, invasion, proteolysis and ECM composition, organization and stiffness. These pleiotropic effects should be taken into strong consideration prior to developing therapeutic strategies aimed at modulating the expression of such miRNAs to treat a disease.

Collectively, these studies reveal an additional level of control exerted by the cells on their surrounding ECM. On the one hand, they illustrate how the orchestrated action of many different miRNAs controls the balance in ECM composition that determines normal tissue development and function. On the other hand, these studies clearly show that alterations in the expression of these miRNAs contribute to a dysregulated and disorganized ECM phenotype, which can cause or exacerbate disease. This naturally leads to the question of whether there exists a mutual control between miRNAs and the ECM: can the ECM also influence the expression and function of miRNAs inside its resident cells?

Regulation of miRNAs by the ECM

While it has been well established that miRNAs can regulate the expression of ECM molecules, it is just beginning to become apparent that the expression and function of miRNAs can be influenced by the ECM. Below, we review the impact of single ECM components as well as the ECM as a global entity on miRNA expression and function. We also discuss recent evidence that points to a role for non-coding regions of ECM molecule mRNAs in modulating miRNA function. Key details of these studies are summarized in Table2.

Table 2.

miRNAs regulated by ECM molecules

ECM molecule	miRNA	System	Regulatory mechanism	Process	Reference
Inflammation
Decorin	miR-21	LPS-stimulated mouse peritoneal macrophages; decorin−/− septic mice; tumour xenografts overexpressing decorin	Post-transcriptional; through the inhibition of TGF-β1-induced processing of miR-21	Inflammation and cancer	Merline et al. (2011)
Tenascin-C	miR-155	LPS-stimulated BMDMs; tenascin-C−/− septic mice	Post-transcriptional	Inflammation	Piccinini and Midwood (2012)
Cancer
Integrin α6β4	miR-29a	Human MDA-MB-435 and SUM159 breast carcinoma cells	? Requires integrin α6β4 cytoplasmic tail	Cancer cell invasion	Gerson et al. (2012)
HA/CD44	miR-10b	Human MDA-MB-435 breast carcinoma cells	Transcriptional; through binding of Twist to E-box elements in the miR-10b promoter	Cancer cell invasion and metastasis	Bourguignon et al. (2010)
HA/CD44	miR-21	Human HSC-3 oral squamous carcinoma cells	Transcriptional; through binding of Nanog-Stat-3 complex to the mir-21 promoter	Cancer cell survival and chemoresistance	Bourguignon et al. (2012a)
HA/CD44	miR-21	Human MCF-7 breast adenocarcinoma cells	Post-transcriptional; through Nanog association with RNA helicase p68 and Drosha	Cancer cell survival and chemoresistance	Bourguignon et al. (2009)
HA/CD44v3	miR-302a/302b	Human HSC-3 squamous carcinoma cells of mouth; primary tumours from patients with HNSCC	Transcriptional; through binding of Oct4–Sox2–Nanog complexes to miR-302 promoter	Oncogenesis	Bourguignon et al. (2012b)
3-D vs. 2-D cultures
Matrigel (BD Biosciences) ± Col-1	miR-21	3-D culture of mouse mK-ras-LE lung epithelial cancer cells, human A549 lung adenocarcinoma and MCF-7 breast adenocarcinoma cells	Post-transcriptional	Tumour initiation and progression	Li et al. (2011a)
Matrigel (BD Biosciences)	miR-17, miR-92b, miR-21, miR-200a	3-D culture of mouse mK-ras-LE lung epithelial cells, human A549 lung adenocarcinoma cells, mouse A549LC lung adenocarcinoma cells	?	Acinar morphogenesis of lung adenocarcinoma in 3-D culture	Li et al. (2012)
Matrigel (BD Biosciences) ± Col-1	miR-141, miR-429, miR-200c, let-7c	3-D culture of human MCF-7 and MDA-MB-231 breast adenocarcinoma cells	?	Morphogenesis of human breast cancer cells in 3-D culture	Nguyen et al. (2012)
Matrigel (BD Biosciences)	miR-1290, miR-210, miR-29b, miR-32	3-D culture of human SW480 and SW620 colon adenocarcinoma cells, HT-29 colon colorectal adenocarcinoma cells, A549 lung adenocarcinoma cells, MDA-231 breast adenocarcinoma cells	?	Cancer cell growth and invasion	Price et al. (2012)
3-D Col-1 gel	let-7a, let-7d, let-7g	3-D culture of primary human PDAC cells	Post-transcriptional; through Col-1-induced TGF-β1 signalling and induction of MMP-14 expression and ERK1/2 activation	Cancer progression	Dangi-Garimella et al. (2011)
ECM molecule 3′ UTR
Versican 3′ UTR	miR-199a-3p	Versican 3′ UTR TG mice, versican 3′ UTR-transfected human U343 and U87 astrocytoma cells	Binding of versican 3′ UTR to endogenous miR-199a-3p enhances translation of versican and fibronectin mRNA	Cell, tissue and organ adhesion	Lee et al. (2009)
	miR-199a-3p, miR-144, miR-136	Versican 3′ UTR TG mice, versican 3′ UTR-transfected mouse 4T1 breast carcinoma cells	Binding of versican 3′ UTR to endogenous miR-199a-3p, miR-144 and miR-136 enhances the translation of versican, Rb1 and PTEN mRNA	Cancer cell proliferation and tumour growth	Lee et al. (2010)
	miR-199a-3p, miR-144, miR-133a, miR-431	Versican 3 ‘UTR TG mice, versican 3′ UTR-transfected human HepG2 liver cancer cells	Binding of versican 3′ UTR to endogenous miR-199a-3p, miR-144, miR-133a and miR-431 enhances the translation of versican isoforms V0 and V1, CD34 and fibronectin mRNA	Cancer cell proliferation and invasion	Fang et al. (2013)
CD44 3′ UTR	miR-216a, miR-330, miR-608	CD44 3′ UTR-transfected human MT-1 breast cancer cells and rat YPEN-1 endothelial cells	Binding of CD44 3′ UTR to endogenous miR-216a, miR-330 and miR-608 enhances translation of CD44 and CDC42 mRNA	Cancer cell proliferation, tumour growth and angiogenesis	Jeyapalan et al. (2011)
	miR-328, miR-512-3p, miR-491, miR-671	CD44 3′ UTR-transfected human MDA-MB-231 breast cancer cells	Binding of CD44 3′ UTR to endogenous miR-328, miR-512-3p, miR-491 and miR-671 enhances translation of COL1A1 and FN1	Cancer cell invasion and metastasis	Rutnam and Yang (2012)
Nephronectin 3′ UTR	miR-378	Nephronectin 3′ UTR-transfected mouse MC3T3-E1 osteoblast progenitor cells	Binding of nephronectin 3′ UTR to endogenous miR-378 represses EGFR and ERK phosphorylation and induces GSK3β-mediated signalling pathway through increased expression of β-catenin	Osteogenesis	(Lee et al. 2011)

TGF-β1, transforming growth factor β1; LPS, lipopolysaccharide; BMDM, bone marrow-derived macrophage; HNSCC, head and neck squamous cell carcinoma; Col-1, type I collagen; PDAC, pancreatic ductal adenocarcinoma; ERK, extracellular signal-regulated kinase; Rb1, retinoblastoma-associated protein; PTEN, phosphatase and tensin homolog; PAK1, p21-activated kinase; CDC42, cell division control protein 42; FN1, fibronectin type 1; EGFR, epidermal growth factor receptor; GSK3β, glycogen synthase kinase 3 beta.

Regulation of miRNAs by the ECM in inflammation

miRNAs are involved in the regulation of innate and adaptive immune responses as well as inflammatory networks in various cell and tissue types [reviewed in Xiao & Rajewsky (2009); O'Connell et al. (2010)]. Recent research that implicates a role for the ECM in regulating miRNAs in inflammation is summarized below.

Decorin, a small leucine-rich proteoglycan of the ECM, has been implicated in cancer, where it inhibits primary tumour growth and metastasis (Goldoni & Iozzo 2008), as well as in various inflammatory processes (Xaus et al. 2001; Comalada et al. 2003). A novel role for decorin as a regulator of miR-21 abundance was recently identified in pathogen-mediated inflammation in the context of sepsis and in sterile inflammation in the context of tumour growth (Merline et al. 2011). TGF-β and Smad signalling can promote the processing of the primary transcript of miR-21 (pri-miR-21), leading to increased mature miR-21 levels (Davis et al. 2008). Merline et al. found that soluble decorin inhibited TGF-β signalling, which resulted in decreased abundance of miR-21 and increased levels of its target PDCD4, a proinflammatory tumour suppressor (Merline et al. 2011). However, the mechanism underlying decorin-mediated inhibition of TGF-β signalling, and thus miR-21 biosynthesis, is as yet unclear. Moreover, it is not known which part of decorin, the protein core or the glycosaminoglycan (GAG) chain, mediates this function.

In the context of pathogen-mediated inflammation, we recently found that the glycoprotein tenascin-C regulates lipopolysaccharide (LPS)-induced miR-155 expression, enabling translation of specific cytokine subsets (Piccinini & Midwood 2012). Historically, tenascin-C was reported to be generated exclusively upon tissue injury in order to stimulate the immune response in sterile conditions. We found that it is also induced in response to bacterial components and is required to generate an effective immune response to bacterial LPS in vivo in septic mice (Piccinini & Midwood 2012). LPS stimulation of Toll-like receptor 4 (TLR4) induces the expression of miR-155, which enhances the proinflammatory response through the inhibition of negative regulators of TLR signalling (Lu et al. 2009; O'Connell et al. 2009) and promotion of TNF-α production (Tili et al. 2007; Bala et al. 2011). We showed that tenascin-C upregulated TNF-α production in macrophages upon LPS activation by enhancing the expression of miR-155. Moreover, we demonstrated that tenascin-C regulates miR-155 at the post-transcriptional level (Piccinini & Midwood 2012). However, further work is required to dissect the exact molecular mechanism underlying tenascin-C-mediated regulation of miR-155 expression.

Regulation of miRNAs by the ECM in cancer

miRNAs, as oncogenes and oncosuppressor genes, regulate the expression of their target genes to affect the functional status of cancer and stromal cells and the tumour microenvironment. Dysregulated miRNA expression in the epithelial as well as the stromal compartment of tumours indicates that miRNAs are important drivers of tumorigenesis and metastasis (Chou & Werb 2012; Melo & Kalluri 2013). Understanding how the ECM molecules of the tumour microenvironment regulate miRNA expression will further elucidate how miRNAs affect tumour biology and may offer potential therapeutic opportunities. This research direction has been taken by a number of groups and has focussed on integrins, CD44 and its ligand hyaluronan (HA).

The first report that integrins can regulate the expression of miRNAs was published in 2012 by Gerson et al., who investigated the role of integrin α6β4 (referred to as β4 integrin) in the context of breast cancer. β4 integrin functions as an adhesion receptor for laminins that promotes carcinoma invasion (Lipscomb & Mercurio 2005). This group found that β4 integrin could decrease miR-29a expression and further showed that the cytoplasmic tail of β4 integrin was required for the repression of miR-29a. This miRNA targets secreted protein acidic and rich in cysteine (SPARC), a glycoprotein that plays an important role in ECM remodelling and invasion. Thus, its repression by β4 integrin facilitated invasion (Gerson et al. 2012). This novel mode of regulation of miR-29a would enable SPARC expression to be modulated rapidly in response to microenvironmental cues sensed by β4 integrin.

Focussing on CD44, Bourguignon et al. discovered a new HA/CD44-mediated signalling mechanism that regulates Twist-associated miR-10b production. Specifically, they found that HA binding to CD44 promotes the nuclear translocation and transcriptional activation of Twist, a stem cell-related transcription factor. They further showed that miR-10b is controlled by an upstream promoter containing Twist binding sites and, by performing chromatin immunoprecipitation (ChIP) assays, they demonstrated that the induction of miR-10b expression in breast cancer cells is HA/CD44 specific and Twist dependent. This event led to the downregulation of the tumour suppressor protein HOXD10, RhoGTPase-ROK activation and tumour cell invasion (Bourguignon et al. 2010). These events are essential for the acquisition of metastatic properties by human breast cancer cells.

The same group also published a series of works, which together revealed a HA/CD44-mediated Nanog signalling and miR-21 production leading to oncogenesis and chemoresistance in cancer cells. In 2009, they showed that HA/CD44 interaction promoted Nanog association with the RNA helicase p68 and Drosha, leading to the processing and thus production of mature miR-21 in breast cancer cells (Bourguignon et al. 2009). Three years later, they reported that HA binding to tumour cell surface promoted CD44 association with the stem cell marker Nanog, which was then translocated into the nucleus and, in complex with Stat-3, interacted with an upstream/enhancer region, containing Stat-3-binding sites, of the miR-21 promoter, resulting in miR-21 gene expression and mature miR-21 production in head and neck squamous carcinoma cells. Upregulated miR-21 expression led to lower levels of the tumour suppressor protein PDCD4, upregulation of inhibitors of the apoptosis family of proteins (IAPs) and chemoresistance (Bourguignon et al. 2012a). These findings show a functional link between the HA/CD44-mediated Nanog signalling pathway and miR-21 production and function during oncogenesis.

Finally, this group found that a highly tumorigenic human mouth squamous carcinoma cell line expressed high levels of a CD44 variant isoform (CD44v3). They showed that HA binding to CD44v3 promoted CD44v3 association with the stem cell markers Oct4, Sox2 and Nanog as a complex. This signalling complex was found to translocate into the nucleus and interact with the promoter region of miR-302a/miR-302b. The resulting mature miR-302 cluster then repressed several epigenetic regulators that induced global DNA demethylation and upregulated a number of survival proteins, leading to self-renewal, clone formation, survival and chemoresistance (Bourguignon et al. 2012b).

The cross-talk between miRNAs and various intracellular signalling pathways that are activated upon HA-CD44 interaction has been thoroughly investigated and partially explains how these ECM components promote tumour progression. However, CD44 can interact with other ECM ligands, including collagens, osteopontin and MMPs, and participates in cellular processes other than cancer metastasis, including lymphocyte activation and haematopoiesis. It would be interesting to know whether CD44 can regulate other miRNAs in these distinct contexts. Moreover, further work is necessary to continue to unravel the mechanism(s) by which other ECM molecules regulate miRNA biogenesis.

In addition to single ECM components, the ECM as a whole with its biomechanical properties, including stiffness, influences cell behaviour. In the next section, we introduce a line of research that uses 3-D cultures, which recapitulate the physiological matrix, to investigate the impact of the ECM on miRNA expression profiles.

ECM composition and rigidity: 3-D vs. 2-D cultures

Based on the knowledge that both the ECM composition/rigidity and miRNAs play established roles in tumour initiation and progression (McAllister & Weinberg 2010), Li et al. investigated whether collagen type I deposition in the ECM and expression of miRNAs are linked by using organotypic 3-D culture. In brief, cells were embedded within Matrigel, which resembles basement membrane matrix (Matrigel 3-D; Hebner et al. 2008). Matrigel 3-D culture of lung and mammary epithelial cells formed acini, an in vitro equivalent of apical–basolateral polarity that consists of a polarized monolayer of epithelial cells facing a central lumen. Interestingly, addition of collagen type I disrupted acini and upregulated the expression of miR-21, an oncogenic miRNA. Collagen-mediated regulation of miR-21 expression was post-transcriptional and occurred at the processing step from pre-miR-21 to mature miR-21. Indeed, collagen upregulated only miR-21 and not its precursors. Moreover, ectopical expression of pre-miR-21 failed to produce mature miR-21. A similar post-transcriptional upregulation of miR-21 correlated with deposition of collagen type I in a murine model of lung fibrosis (Li et al. 2011a). This study linked altered ECM composition and rigidity with the expression of oncogenic miRNAs; however, further work is required to define the mechanism for post-transcriptional regulation of miR-21 by collagen type I at the molecular level.

Two follow-up papers were published a year later by the same group. They investigated the expression and function of miRNAs in Matrigel 3-D culture in order to elucidate the cell biology pertinent to tumorigenesis, such as epithelial cell polarity and cross-talk between the tumour cells and their microenvironment, which cannot be established using 2-D cultures (Li et al. 2012; Nguyen et al. 2012). This approach was used in both studies, which used cells from different types of tumours. In their latest work, they compared the expression profile of 700 miRNAs in 3-D and 2-D cultures of two lung adenocarcinoma cell lines. Striking differences emerged in miRNA profiles between the two types of cultures. As opposed to 2-D culture, the 3-D culture-specific miRNA profile displayed high expression of tumour-suppressive miRNAs (i.e. miR-200 family) and low expression of oncogenic miRNAs (i.e. miR-17–92 cluster and miR-21). Moreover, the expression pattern of miR-17, miR-21 and miR-200a in 3-D culture correlated with the expression of their targets and acinar morphogenesis (Li et al. 2012).

A similar approach was adopted by an independent research group that performed miRNA profiling of two colon cancer cell lines by microarray analysis and identified significant differential expression of miRNAs between cells cultured in Matrigel and those cultured on plastic (Price et al. 2012).

Interestingly, a study has uncovered the existence of feedback mechanisms that allow ECM molecules, which are downstream targets of a specific miRNA, to regulate the expression of that miRNA. For instance, type I collagen is a downstream target of let-7a and let-7g (Ji et al. 2010; Makino et al. 2013) and a feedback mechanism allows collagen to regulate let-7 expression in pancreatic cancer cells as well (Dangi-Garimella et al. 2011). When these cells were grown in 3-D collagen gels, the expression of mature let-7 was greatly reduced, whereas the expression of the let-7 precursor was not affected. The 3-D cultures displayed also increased TGF-β signalling and increased expression of MMP-14. Furthermore, overexpression of MMP-14 in cells embedded in 3D collagen gels or grown in vivo in mice repressed let-7 levels, an effect that could not be observed in 2-D tissue culture plastic (Dangi-Garimella et al. 2011). Besides identifying an additional mechanism by which fibrosis may contribute to pancreatic cancer progression, this study points out an interplay between two key ECM molecules, MMP-14 and its substrate collagen type I, in modulating miRNA expression.

Together, these results highlight how the cell microenvironment affects not only tumour biology, but also miRNA expression. Thus, 3-D culture may be considered as an essential experimental approach for a comprehensive understanding of the miRNA biology in cancer and other diseases that feature dysregulated matrix deposition. Indeed, the gene expression signature of 3-D culture of breast cancer cells has been shown to hold prognostic values for patients with breast cancer (Martin et al. 2008). However, in future, it would be important to establish whether miRNA expression signatures in 3-D vs. 2-D cultures are affected exclusively by the stiffness of Matrigel or also by its components. In fact, Matrigel has a heterogeneous composition and contains not only collagen, laminin and entactin, but also growth factors and small amounts of other proteins. Moreover, it would be of great interest to identify the molecular mechanism by which 3-D culture affects miRNA expression.

3′ UTR of ECM molecule mRNAs: endogenous decoys for miRNAs?

Emerging evidence shows that the 3′ UTR of ECM molecule mRNAs can regulate miRNA levels and function (Lee et al. 2009, 2010, 2011; Jeyapalan et al. 2011; Rutnam & Yang 2012). It has been shown that a number of mRNAs containing miRNA-binding sites, rather than being just passive targets of miRNAs, can titrate miRNAs, thereby regulating their availability (Seitz 2009; Poliseno et al. 2010; Salmena et al. 2011; Sumazin et al. 2011; Tay et al. 2011). In particular, Salmena et al. proposed that 3′ UTRs of mRNAs, besides acting as cis regulatory elements that alter the stability of their own transcripts, may also act in trans to modulate gene expression through miRNA binding (Salmena et al., 2011). This suggests that 3′ UTR of mRNAs can act as ‘endogenous decoys’ for miRNAs. The recent identification of human, mouse and fly 3′ UTRs expressed separately from the associated protein-coding sequences to which they are normally linked supports this hypothesis (Mercer et al. 2011). Furthermore, this hypothesis is also supported by the fact that miRNA effectiveness is influenced by the concentration of its target mRNAs (Arvey et al. 2010). The analysis of the ability of 3′ UTRs of mRNAs to regulate miRNA levels and function has been extended to ECM molecule mRNAs, including versican, CD44 and nephronectin.

In a first study, Lee et al. reported that versican 3′ UTR could antagonize miR-199a-3p levels and function, thereby inducing cell, tissue and organ adhesion. They found that the expression of versican 3′ UTR induced internal organ adhesion in vivo, in transgenic mice expressing a construct harbouring the 3′ UTR of versican and in vitro, in a human astrocytoma cell line. They then demonstrated that the overexpression of versican 3′ UTR interacted with endogenous miR-199a-3p, freeing versican and fibronectin mRNA for translation (Lee et al. 2009). In a second study, they showed that overexpression of versican 3′ UTR reduced breast cancer cell proliferation and tumour growth by decreasing miR-199a-3p levels as well as those of miR-134 and miR-136, which repressed the translation of the tumour suppressor PTEN (Lee et al. 2010). These observations were later confirmed by others using different in vivo and in vitro systems (Fang et al. 2013). These studies suggested that, upon arrival in the cytoplasm, miRNA activities can be modulated by 3′ UTR of versican mRNAs.

The 3′ UTR of CD44 was demonstrated by Jeyapalan et al. to antagonize cytoplasmic miRNAs in a human breast cancer cell line, inhibiting proliferation, colony formation and tumour growth. Furthermore, it modulated angiogenesis of a rat endothelial cell line. These activities were explained by the interaction of the CD44 3′ UTR with multiple miRNAs, including miR-216a, miR-330 and miR-608 (Jeyapalan et al. 2011). A year later, Rutnam et al. reported that overexpression of the CD44 3′ UTR in a human breast carcinoma cell line resulted in enhanced cell motility, invasion and cell adhesion. Furthermore, they found that expression of the CD44 3′ UTR enhanced metastasis in vivo. They hypothesized that the CD44 3′ UTR would interact with multiple miRNAs that target ECM properties. Computational analysis indicated that miRNAs that interact with the CD44 3′ UTR also have binding sites in other matrix-encoding mRNA 3′ UTRs, including COL1A1, which is repressed by miR-328, and fibronectin type 1, which is repressed by miR-512-3p, miR-491 and miR-671. Protein analysis demonstrated that expression of CD44, COL1A1 and fibronectin type 1 was synergistically upregulated in vitro and in vivo upon transfection of the CD44 3′ UTR (Rutnam & Yang 2012).

Finally, overexpression of a fragment containing the 3′ UTR of nephronectin in the osteoblast progenitor cells MC3T3-E1 was shown to promote cell differentiation. The authors observed that β-catenin and GSK3b were upregulated in the 3′ UTR-transfected cells, contributing to the increased cell differentiation, through reduction in EGFR and ERK phosphorylation. Finally, the authors showed that these events were due to the binding of the 3′ UTR of nephronectin to miR-378 (Lee et al. 2011), a miRNA that regulates nephronectin expression (Kahai et al. 2009).

Together, these studies show that the 3′ UTR of ECM molecule mRNAs serves as competitor for miRNA binding and subsequently inhibits miRNA functions, by preventing target mRNA repression.

The overexpression of 3′ UTR of ECM or other molecule mRNAs is an undoubtedly useful system for studying the mutual regulation of miRNAs and mRNAs both in vitro and in vivo. It may even be a potential approach in the development of gene therapy. However, when screening candidate ‘endogenous decoys’ for miRNAs, a number of criteria should be taken into account. Firstly, the 3′ UTR of mRNAs and the miRNAs must have the same expression pattern and subcellular localization. Secondly, the extent of the effect of a 3′ UTR of mRNAs on miRNAs should correlate with its expression levels, the number of binding sites that it contains and the degree of complementarity between mRNA and miRNA at the binding sites.

Conclusions and future prospects

There is a growing body of evidence suggesting that a mutual and coordinated regulation between miRNAs and the ECM is in place to help maintain equilibrium between the cell and its environment (Figure1). Dysregulation of any of the two components can severely impair development or lead to pathological consequences. Although still in its infancy, the investigation of the mutual regulation of cellular miRNAs and their surrounding ECM may be key to the development of strategies to prevent or treat diseases that feature abnormal miRNA levels or activity and/or ECM composition and architecture.

Figure 1.

Mechanisms of regulation of miRNAs by the ECM. The ECM can regulate miRNAs on several levels. Firstly, it can induce signalling pathways that lead to the nuclear translocation of transcription factors that drive miRNA transcription (a). Secondly, the ECM can interfere with signalling pathways that regulate Drosha-mediated processing of miRNA precursors (b). Thirdly, ECM molecules can modulate miRNA activity through their 3′ UTRs, which bind to miRNAs preventing target mRNA regulation (c). Finally, the ECM may stabilize miRNAs that are secreted into the extracellular microenvironment (d).

In addition to residing and operating inside their mother cell, miRNAs can be secreted and circulate in the bloodstream (Creemers et al. 2012). Secreted miRNAs are remarkably stable and have been shown to be delivered to recipient cells where they also participate in the regulation of physiological and pathological processes (Valadi et al. 2007; Chen et al. 2012). Extracellular miRNAs can be enclosed in microvesicles or bound to RNA-binding proteins (e.g. high-density lipoproteins) and several studies indicate that their internalization by recipient cells depends on specific cell surface-targeting motifs on microvesicles and receptors on recipient cell surface (Cocucci et al. 2009; Simons & Raposo 2009; Vickers et al. 2011). It has been reported that miRNAs can also be secreted into the extracellular microenvironment by fission or budding of the cell membrane (Muralidharan-Chari et al. 2010). However, the precise cellular release mechanisms of miRNAs remain largely obscure. Furthermore, it has been suggested that, in the ECM, the vast majority of miRNAs are bound to proteins, whereas a low percentage resides in microvesicles or apoptotic bodies (Wang et al. 2010b; Vickers et al. 2011). This raises a number of interesting questions: can the ECM store miRNAs and thus function as an instantly available source of ready-to-use miRNAs? Does the changing composition of the ECM across tissues limit or determine to which target site and cell type miRNAs can be delivered? Does the dysregulation or disorganization of the ECM affect miRNA secretion and delivery to target cells? The improvement of currently available techniques and the development of new ones to detect miRNAs extracellularly is a priority if we want to answer questions like these.

The main molecular components of the ECM have all been found to be regulated by at least one miRNA. Even though the majority of the ECM genes have been validated as direct miRNA targets, there are plenty of examples of ECM molecules that are indirectly regulated by miRNAs. In many of these cases, the direct miRNA target is a transcription factor (e.g. SOX9), a growth factor (e.g. BMP7) or a signalling molecule (e.g. IKK-β), all of which can regulate or induce the synthesis of ECM components. However, in many instances, the direct miRNA target responsible for the regulation of the ECM remains unknown. Furthermore, a small percentage of studies reported here have not confirmed a direct miRNA-dependent regulation of gene expression by means of reporter gene assays or have carried out these assays without seeding sequence mutagenesis, a method useful to test the specificity of the miRNA–target interaction. Future work will generate data to fill these gaps.

As miRNAs can regulate many ECM molecule genes, the ECM seems also to be able to influence the expression of several miRNAs. One such molecule that emerged from this review is HA in the context of cancer. By interacting with CD44, HA can activate a number of intracellular signalling pathways that ultimately regulate the biogenesis of multiple miRNAs. Several mechanisms of action by which ECM molecules can regulate miRNA expression have emerged from this study. Clearly, the ECM can affect miRNA biogenesis at the transcriptional as well as post-transcriptional level. In particular, ECM molecules can induce the nuclear translocation of transcription factors and their binding to specific sites in a promoter allowing miRNA transcription or can interfere with Drosha- and/or Dicer-mediated processing of miRNA precursors in the nucleus and cytoplasm respectively.

By definition, matrix molecules are confined and act extracellularly. However, an alternative mechanism by which the ECM can regulate the expression and activity of miRNAs in the cytoplasm has been proposed whereby the ECM molecule binds to miRNAs through its 3′ UTR. By acting as ‘endogenous decoys’ for miRNAs, the 3′ UTR of ECM molecule mRNAs can modulate miRNA availability and prevent the regulation of miRNA target genes.

Interestingly, not only individual ECM components, but also the ECM as a functional unit can influence miRNA expression profiles, probably through physical cues such as stiffness and elasticity. Therefore, 3-D culture of adherent cells may recapitulate miRNA expression and function more accurately than culture on plastic, as it takes into account the cellular context and environment. The results obtained so far with this model highlight the potential networking of miRNA regulatory pathways to achieve a coordinated cellular outcome and suggest that miRNAs are regulated by the cellular context and environment. This cell culture system may reveal more physiological miRNA, gene and protein expression signatures, which in future might be used as prognostic or diagnostic markers. However, caution should be taken as accurate standardization of 3-D culture has as yet not been achieved.

Efforts have been made to explore the effect of the ECM on miRNAs in disease. However, it is not known whether the ECM can influence miRNA biogenesis and function during development and other physiological processes. The generation of conditional-null animals for miRNAs and ECM molecules would provide useful tools to address this fundamental question.

As more and more regulatory mechanisms and stimuli, including environmental cues, which can induce or repress miRNA expression, are identified, a major challenge will be elucidating how these factors cooperate to orchestrate miRNA expression and whether there exist potential hierarchies among them. More global, side-by-side miRNA, mRNA and protein analyses will help to distinguish between cause and effect for specific patterns on gene expression changes.

Together, this new knowledge not only expands our understanding of how the ECM and miRNAs are regulated and, in turn, influence cell function, but could also serve as groundwork for the future development of new drug targets.