ABSTRACT

Type VII collagen is the main component of anchoring fibrils, structures that are integral to basement membrane homeostasis in skin. Mutations in the gene encoding type VII collagen COL7A1 cause recessive dystrophic epidermolysis bullosa (RDEB) an inherited skin blistering condition complicated by frequent aggressive cutaneous squamous cell carcinoma (cSCC). OATP1B3, which is encoded by the gene SLCO1B3, is a member of the OATP (organic anion transporting polypeptide) superfamily responsible for transporting a wide range of endogenous and xenobiotic compounds. OATP1B3 expression is limited to the liver in healthy tissues, but is frequently detected in multiple cancer types and is reported to be associated with differing clinical outcome. The mechanism and functional significance of tumour-specific expression of OATP1B3 has yet to be determined. Here, we identify SLCO1B3 expression in tumour keratinocytes isolated from RDEB and UV-induced cSCC and demonstrate that SLCO1B3 expression and promoter activity are modulated by type VII collagen. We show that reduction of SLCO1B3 expression upon expression of full-length type VII collagen in RDEB cSCC coincides with acquisition of front-to-rear polarity and increased organisation of 3D spheroid cultures. In addition, we show that type VII collagen positively regulates the abundance of markers implicated in cellular polarity, namely ELMO2, PAR3, E-cadherin, B-catenin, ITGA6 and Ln332.

INTRODUCTION

Incidence of non-melanoma skin cancer (NMSC) is highest among all malignancies in Caucasian populations (Toms, 2004; Cancer Research UK, 2012, http://publications.cancerresearchuk.org/cancerstats/statsskin/keyfactsskin.html). Approximately 80% of NMSCs comprise basal cell carcinoma (BCC) and 20% cutaneous squamous cell carcinoma (cSCC) (Alam and Ratner, 2001; Yanofsky et al., 2011). BCC is usually a benign disease and is rarely responsible for skin-cancer-related mortality, whereas cSCC can develop into metastatic disease and contributes to around 1 in 4 skin cancer deaths in the UK (ISD Scotland, http://www.isdscotland.org/isd/183.html). The main risk factor for cSCC in the general population is UV radiation (Fears and Scotto, 1983; Preston and Stern, 1992), but other factors such as HPV infection (McGregor and Proby, 1996; Harwood and Proby, 2002) and immunosuppression in organ transplant patients (Rowe et al., 1992; Veness et al., 1999) have been documented. Inherited skin disease can also predispose to developing cSCC (Ng et al., 2011); in particular, individuals diagnosed with recessive dystrophic epidermolysis bullosa (RDEB) (Reed et al., 1975; Fine et al., 2009). Patients with this disease are affected by deleterious mutations in COL7A1, leading to either defective or absent type VII collagen (Christiano et al., 1993). Type VII collagen is the main component of anchoring fibrils, structures which associate with the basement membrane and are essential for the proper attachment of the dermis to the epidermis (Sakai et al., 1986; Burgeson, 1993). Type VII collagen interacts with proteins localised close to the basement membrane; laminin 332 (Ln332) and type IV collagen (Chen et al., 1997; Rousselle et al., 1997), as well banded collagen I that are localised in the dermis (Villone et al., 2008), providing strong adhesive connection between these two compartments. Over 90% of RDEB patients develop metastatic cSCC by age 55, a figure significantly higher than that in the general population where metastatic disease is relatively rare (Weinstock et al., 1991; Miller and Weinstock, 1994; Fine et al., 2009). The underlying molecular mechanism driving predisposition to develop aggressive cSCC in RDEB is not fully understood (South and O'Toole, 2010) although recent data from our laboratory show that tumour microenvironment, in particular the altered composition of extracellular matrix in RDEB, can promote tumours (Ng et al., 2012). Similarly, despite an alarming rise in the incidence of cSCC in the general population and a relatively low 5-year survival in patients presenting with regional metastasis (Rowe et al., 1992; Czarnecki et al., 1994), little is understood about these tumours with poor prognosis.

We previously identified a few differentially expressed genes (DEGs) using microarray to compare RDEB with UV-induced cSCC in a relatively small number of keratinocyte cultures (n = 9) (Watt et al., 2011). We further investigate those 18 DEGs and identify SLCO1B3, a gene that encodes the liver-specific organic anion transporter polypeptide OATP1B3, as being upregulated in both UV-induced and RDEB cSCC when compared with normal skin. Transporter genes are biologically significant for cell homeostasis and can regulate cellular mechanisms such as metabolism, cell death and motility (Olk et al., 2009; Klaassen and Aleksunes, 2010). OATP1B3 is normally expressed in the liver and is involved in the transport of bile salts, glutathione and other organic anions (Smith et al., 2005a; Hagenbuch and Gui, 2008). In agreement with our data, OATP1B3/SLCO1B3 is frequently upregulated in solid tumours including breast and colon (Lockhart et al., 2008; Maeda et al., 2010; Pressler et al., 2011) and it has been suggested that OATP1B3 has therapeutic potential owing to its ability to transport a broad range of drugs, including methotrexate, paclitaxel, docetaxel, bromsulphalein (BSP) and pitavastatin (Smith et al., 2005b; Hagenbuch and Gui, 2008). OATP1B3 expression has been reported to associate with clinical outcome in a number of studies (Hamada et al., 2008; Lockhart et al., 2008) yet little data exist on the mechanism of expression or functional consequence. Lee and colleagues report that OATP1B3 promotes apoptotic resistance in tumour cells (Lee et al., 2008) and very recently, hypoxia has been shown to positively regulate OATP1B3 expression (Ramachandran et al., 2013).

Here, we investigate the relationship between type VII collagen and tumour expression of SLCO1B3 and demonstrate that type VII collagen, Ln332 and ELMO2 regulate tumour expression of SLCO1B3. We further show that expression of type VII collagen in RDEB cSCC keratinocytes promotes front-to-rear polarity on uncoated plastic and increases organisation of 3D spheroid cultures. These structural changes in response to expression of type VII collagen coincide with an increase in the abundance of markers implicated in cellular polarity, namely ELMO2, PAR3, E-cadherin, B-catenin, ITGA6 and Ln332.

RESULTS

SLCO1B3 is expressed in cSCC in vitro and in vivo

We previously identified only 18 genes to be significantly differentially expressed between cultured RDEB cSCC keratinocytes and UV-induced cSCC keratinocytes using gene expression arrays, demonstrating these two tumour groups are not significantly different in this assay (Watt et al., 2011). Initial follow-up of these 18 genes comparing three RDEB cSCC and three UV cSCC cultures using sybr®green qPCR revealed that 12 genes showed the same trend as array data but only five were significantly dysregulated: SLCO1B3, DUSP23, TJAP1, SNRPN and SNURF (data not shown). From these five genes, SLCO1B3 showed a consistent upregulation in RDEB cSCC keratinocytes in culture compared with normal primary keratinocytes and the majority of UV-induced cSCC keratinocytes, with the exception of SCCT8 (Fig. 1A). However this experiment also identified a significant increase in expression in the majority of UV-induced cSCC keratinocytes when compared with normal keratinocytes. Six of eight cultured cSCC cell lines showed a greater level of expression compared with the hepatocellular carcinoma cell line HepG2 (Fig. 1A).

Fig. 1.

SLCO1B3 is upregulated in cSCC-derived keratinocytes and primary tissue. (A) SLCO1B3 mRNA expression as assessed by Sybr®Green qPCR in six cSCC, three RDEB cSCC, two metastatic cSCC, one normal human keratinocyte (NHK) and one hepatocellular carcinoma (HepG2) culture. The expression levels were calculated using the ΔΔCT method with ACTB as the normaliser and NHK as the calibrator. (B) Sybr®Green qPCR analysis of SLCO1B3 mRNA levels in tissue from 11 RDEB cSCC, three RDEB skin, three cSCC, one normal skin (NHS) and one normal liver. The expression levels were calculated using the ΔΔCT method with ACTB as the normaliser and NHS as the calibrator. Experiment was repeated twice, data from the same experiment (performed in triplicate) are shown as means ± s.d.

Fig. 1.

SLCO1B3 is upregulated in cSCC-derived keratinocytes and primary tissue. (A) SLCO1B3 mRNA expression as assessed by Sybr®Green qPCR in six cSCC, three RDEB cSCC, two metastatic cSCC, one normal human keratinocyte (NHK) and one hepatocellular carcinoma (HepG2) culture. The expression levels were calculated using the ΔΔCT method with ACTB as the normaliser and NHK as the calibrator. (B) Sybr®Green qPCR analysis of SLCO1B3 mRNA levels in tissue from 11 RDEB cSCC, three RDEB skin, three cSCC, one normal skin (NHS) and one normal liver. The expression levels were calculated using the ΔΔCT method with ACTB as the normaliser and NHS as the calibrator. Experiment was repeated twice, data from the same experiment (performed in triplicate) are shown as means ± s.d.

Q-PCR using reverse transcribed mRNA obtained from patient tissues showed high SLCO1B3 expression in RDEB cSCC, and to a lesser extent UV-induced cSCC and RDEB non-SCC skin compared with normal skin, with the highest levels evident in RDEB cSCC tissue (Fig. 1B).

Endogenous OATP1B3 expression is detected in cSCC keratinocytes but only localises to the membrane in vivo

OATP1B3 expression was readily detected in cSCC keratinocytes but not in normal primary keratinocytes in culture (Fig. 2A). Plasma-membrane-bound OATP1B3 was, however, not observed in cultured cSCC keratinocytes using indirect immunofluorescence unless overexpressed by retroviral transduction (Fig. 2B). Membrane-bound protein was detected in organotypic 3D cultures (data not shown) and xenograft tumours (Fig. 2B, right).

Fig. 2.

OATP1B3 is endogenously expressed in cSCC-derived keratinocytes. (A) Endogenous OATP1B3 protein expression was detected using immunoblotting in cSCC keratinocytes, but not in normal human keratinocytes (NHK). SLCO1B3-overexpressing cells (OATP1B3_SCCRDEB2) were used as a positive control. EV_SCCRDEB2, empty vector control cells. Adjacent western blot quantification graph shows percentage average relative OATP1B3 protein density (n = 3, means ± s.d.). (B) OATP1B3 protein expression is readily detected in monolayer 2D cultured cells (SCCT8) (left) but is only localised to the plasma membrane when it is overexpressed in 2D culture (middle) or in mouse xenograft tumours (right) using the same SCCT8 cells. Top panel shows immunostaining with an antibody raised against OATP1B3 whereas bottom panel shows DAPI nuclear staining. Scale bars: 10 µm.

Fig. 2.

OATP1B3 is endogenously expressed in cSCC-derived keratinocytes. (A) Endogenous OATP1B3 protein expression was detected using immunoblotting in cSCC keratinocytes, but not in normal human keratinocytes (NHK). SLCO1B3-overexpressing cells (OATP1B3_SCCRDEB2) were used as a positive control. EV_SCCRDEB2, empty vector control cells. Adjacent western blot quantification graph shows percentage average relative OATP1B3 protein density (n = 3, means ± s.d.). (B) OATP1B3 protein expression is readily detected in monolayer 2D cultured cells (SCCT8) (left) but is only localised to the plasma membrane when it is overexpressed in 2D culture (middle) or in mouse xenograft tumours (right) using the same SCCT8 cells. Top panel shows immunostaining with an antibody raised against OATP1B3 whereas bottom panel shows DAPI nuclear staining. Scale bars: 10 µm.

Tissue sections obtained from RDEB cSCC (n = 7), RDEB skin (n = 3) and UV-induced cSCC (n = 6) patients along with normal human liver (positive control) and normal human skin (negative control) were assessed for OATP1B3 expression. Samples from 6/7 RDEB cSCC and 5/6 cSCC showed heterogeneous plasma-membrane-bound protein expression along with nuclear localisation in 3/7 RDEB cSCC and 2/6 cSCC tissues (Fig. 3A, top and Fig. 3B). Membrane-bound protein was found in 1/3 RDEB skin samples (Fig. 3A, bottom). Colocalisation with plakoglobin confirmed membrane localisation, although higher levels of OATP1B3 were readily detected in regions of cSCC tissue where membrane-bound plakoglobin was absent (Fig. 3B).

Fig. 3.

OATP1B3 is overexpressed in cSCC, RDEB cSCC and RDEB patient skin but is absent from normal skin. (A) cSCC tissues and RDEB patient skin analysed for OATP1B3 showed a strong plasma membrane protein expression along with nuclear localisation. Normal liver was used as a positive control and normal skin as a negative control for immunofluorescence. Top panel shows OATP1B3 immunostaining and the lower panel shows DAPI counter stain for all tissues. (B) Confocal images of cSCC tissue showing OATP1B3 in red and plakogloblin in green. The merged image shows OATP1B3 nuclear staining (indicated by yellow arrows) and plasma membrane colocalisation with plakogloblin (indicated by white arrows). Scale bars: 10 µm.

Fig. 3.

OATP1B3 is overexpressed in cSCC, RDEB cSCC and RDEB patient skin but is absent from normal skin. (A) cSCC tissues and RDEB patient skin analysed for OATP1B3 showed a strong plasma membrane protein expression along with nuclear localisation. Normal liver was used as a positive control and normal skin as a negative control for immunofluorescence. Top panel shows OATP1B3 immunostaining and the lower panel shows DAPI counter stain for all tissues. (B) Confocal images of cSCC tissue showing OATP1B3 in red and plakogloblin in green. The merged image shows OATP1B3 nuclear staining (indicated by yellow arrows) and plasma membrane colocalisation with plakogloblin (indicated by white arrows). Scale bars: 10 µm.

SLCO1B3 expression is regulated by type VII collagen

We noted that when wild-type COL7A1 was expressed in RDEB cSCC keratinocytes (SCCRDEB-C7 cells) the level of SLCO1B3 expression was reduced compared with empty vector control cells (SCCRDEB-EV cells) (P = 0.004, Fig. 4A). This observation was confirmed using separate clonal populations of transduced cells, ruling out any insertion effects (Fig. 4A–C). To determine whether type VII collagen had a direct effect on the promoter activity of SLCO1B3, we cloned the SLCO1B3 promoter, previously described by Jung and co-workers (Jung et al., 2002) and transfected a SLCO1B3 promoter-driven luciferase construct into either SCCRDEB3-C7 or SCCRDEB3-EV cells. SLCO1B3 promoter activity was significantly (P = 0.02) reduced in SCCRDEB3 keratinocytes transduced with collagen VII compared with control cells (Fig. 4D).

Fig. 4.

SLCO1B3 expression is negatively regulated by full-length type VII collagen in both RDEB cSCC and UV-induced cSCC keratinocytes. (A) SCCRDEB3 keratinocytes transduced with full length COL7A1 (clones C7-6 and C7-9) show reduced SLCO1B3 mRNA levels compared with the empty vector cells (EV) as determined by Sybr®Green qPCR analysis. The expression levels were calculated using the ΔΔCT method with ACTB as the normaliser and EV as the calibrator. (B) Western blot confirms type VII collagen overexpression and reduced OATP1B3 expression in both SCCRDEB3 C7-6 and C7-9 clones compared with the empty vector control. (C) Western blot quantification shows relative average OATP1B3 protein density from (B) (n = 3, means ± s.d). (D) A dual luciferase gene reporter assay was designed to assess the effect of COL7A1 on SLCO1B3 promoter activity, which was significantly reduced in the cells overexpressing type VII collagen (C7-6), compared with the empty vector (EV) controls. (E) SLCO1B3 expression is significantly reduced when SCCRDEB3 and SCCRDEB2 cells are cultured on matrix derived from cells overexpressing type VII collagen (C7-6) compared with the empty vector control (EV). Experiments performed in triplicate and data are shown as the means ± s.d.; **P<0.005, *P<0.05.

Fig. 4.

SLCO1B3 expression is negatively regulated by full-length type VII collagen in both RDEB cSCC and UV-induced cSCC keratinocytes. (A) SCCRDEB3 keratinocytes transduced with full length COL7A1 (clones C7-6 and C7-9) show reduced SLCO1B3 mRNA levels compared with the empty vector cells (EV) as determined by Sybr®Green qPCR analysis. The expression levels were calculated using the ΔΔCT method with ACTB as the normaliser and EV as the calibrator. (B) Western blot confirms type VII collagen overexpression and reduced OATP1B3 expression in both SCCRDEB3 C7-6 and C7-9 clones compared with the empty vector control. (C) Western blot quantification shows relative average OATP1B3 protein density from (B) (n = 3, means ± s.d). (D) A dual luciferase gene reporter assay was designed to assess the effect of COL7A1 on SLCO1B3 promoter activity, which was significantly reduced in the cells overexpressing type VII collagen (C7-6), compared with the empty vector (EV) controls. (E) SLCO1B3 expression is significantly reduced when SCCRDEB3 and SCCRDEB2 cells are cultured on matrix derived from cells overexpressing type VII collagen (C7-6) compared with the empty vector control (EV). Experiments performed in triplicate and data are shown as the means ± s.d.; **P<0.005, *P<0.05.

Furthermore, SLCO1B3 expression was significantly reduced in two separate patient-derived parental populations of RDEB SCC cells, SCCRDEB3 and SCCRDEB2, when cultured on matrix derived from SCCRDEB3-C7 cells compared with matrix derived from SCCRDEB3-EV vector control cells (Fig. 4E).

We then used siRNA-mediated depletion of COL7A1 to show that reduction of recombinant type VII collagen increases SLCO1B3 expression in a clonal population of SCCRDEB3-C7 cells (Fig. 5A,B). We note that in this experiment the NT control level of SLCO1B3 was greater in SCCRDEB3-EV cells than SCCRDEB3-C7 cells as expected (1.5-fold), yet the impact of siRNA-mediated depletion of COL7A1 resulted in a greater increase in SLCO1B3 levels than seen in the EV population (Fig. 5B). We attribute this observation to the clonal nature of the two cell populations. To confirm the observation that siRNA depletion of COL7A1 increased SLCO1B3 expression in non-clonal populations we next used siRNA depletion of COL7A1 in SCCRDEB4-C7 cells (Fig. 5C) and the UV-induced nonRDEB cSCC keratinocyte population SCCT8 (Fig. 5D), providing a direct link to type VII collagen and SLCO1B3 tumour keratinocyte expression. The same effect was also observed in normal primary keratinocytes (NHK) following knockdown of type VII collagen, although the overall level of expression was significantly lower than in tumour cells (supplementary material Fig. S1).

Fig. 5.

SLCO1B3 mRNA expression is increased following COL7A1 knockdown in both UV-induced and RDEB cSCC keratinocytes. (A) Knockdown of type VII collagen was achieved using three siRNA constructs in clonal populations of SCCRDEB3 C7 and EV cells (SiC7-1, SiC7-2 and SiC7-3) and compared with the non-targeting control (NT). (B) SLCO1B3 mRNA levels showed a significant increase in cells overexpressing type VII collagen following knockdown of COL7A1 compared with the empty vector control cells. The same assay was repeated for (C) SCCRDEB4-C7 and (D) SCCT8 cells.

Fig. 5.

SLCO1B3 mRNA expression is increased following COL7A1 knockdown in both UV-induced and RDEB cSCC keratinocytes. (A) Knockdown of type VII collagen was achieved using three siRNA constructs in clonal populations of SCCRDEB3 C7 and EV cells (SiC7-1, SiC7-2 and SiC7-3) and compared with the non-targeting control (NT). (B) SLCO1B3 mRNA levels showed a significant increase in cells overexpressing type VII collagen following knockdown of COL7A1 compared with the empty vector control cells. The same assay was repeated for (C) SCCRDEB4-C7 and (D) SCCT8 cells.

Expression of the polarity related protein ELMO2 is regulated by type VII collagen

OATP1B3 is a transmembrane transporter that is expressed in the basolateral component of the plasma membrane. Type VII collagen is a basement membrane protein that directly binds Ln332, which itself binds integrin receptors. It has been previously reported that expression of cell surface proteins can be regulated by interactions with basement membrane components that influence cellular polarity (Li et al., 2003; Daley et al., 2012). We therefore postulated that although OATP1B3 was not observed to be asymmetrically localised in tumour keratinocytes adjacent to extracellular matrix (Fig. 3), the absence of type VII collagen might regulate its expression by altering cellular polarity.

Other than basement-membrane-associated proteins in basal keratinocytes or tight junction components in upper stratified keratinocytes, asymmetrically distributed apical basal polarity markers have yet to be defined in the epidermis (Niessen et al., 2012). Better described is the redistribution of proteins involved in the assembly and disassembly of focal adhesions during the acquisition of front-to-rear polarity in cultured single cells (Hamill et al., 2009; Ho and Dagnino, 2012b). Front-to-rear polarity in cultured primary murine keratinocytes in response to Ln332 is mediated in part by a focal adhesion complex containing the protein engulfment and motility 2 (ELMO2) (Ho et al., 2009; Ho and Dagnino, 2012b). We therefore investigated whether a relationship existed between ELMO2, type VII collagen and OATP1B3.

Overexpression of type VII collagen resulted in upregulation of ELMO2 expression in SCCRDEB3-C7 cells (Fig. 6A). Concurrently, siRNA-mediated depletion of COL7A1 in four independent tumor cell populations resulted in a reduction in ELMO2 expression; SCCRDEB3-C7 and SCCRDEB4-C7 cells as well as cSCC cells that endogenously express type VII collagen (SCCT8 and SCCT1) (Fig. 6B). ELMO2 protein levels corresponded with a reduction in transcript (Fig. 6C,D). Furthermore, ELMO2 expression increased in both SCCRDEB3 and SCCRDEB2 keratinocytes that were cultured on matrix derived from SCCRDEB3-C7 cells (Fig. 6E,F). These data demonstrate that type VII collagen positively regulates ELMO2 expression in cSCC keratinocytes.

Fig. 6.

Type VII collagen positively regulates ELMO2 expression in cSCC keratinocytes. (A) Type VII collagen re-expression in SCCRDEB3 keratinocytes (C7-6) results in an increase in ELMO2 expression. The expression levels were calculated using the ΔΔCT method with GAPDH as the normaliser and empty vector control (EV) as the calibrator. (B) Knockdown of type VII collagen was achieved using a siRNA pool (SiC7 pool) and compared with non-targeting control (NT). ELMO2 mRNA levels showed a significant decrease in both UV-induced cSCC and RDEB cSCC keratinocytes overexpressing type VII collagen following knockdown of COL7A1 compared with the empty vector control cells. (C) Western blot confirms an increase in ELMO2 expression in SCCRDEB3 keratinocytes following COL7A1 expression and a reduction in ELMO2 expression in UV-induced cSCC keratinocytes (SCCT1) following knockdown of COL7A1. Adjacent western blot quantification graphs show percentage average relative ELMO2 protein density (n = 3, means ± s.d). (D) Immunofluorescence images confirm the loss of ELMO2 expression following knockdown of type VII collagen in SCCT8. (E) ELMO2 expression was significantly increased when SCCRDEB3 and SCCRDEB2 cells were cultured on matrix derived from cells overexpressing type VII collagen (C7-6) compared with the empty vector control (EV). (F) Immunofluorescence images confirming an increase in ELMO2 expression when SCCRDEB3 cells were cultured on matrix derived from C7-6 cells compared with EV. Experiments were performed in triplicate and the data are shown as the means ± s.d.; **P<0.005, *P<0.05. Scale bars: 10 µm.

Fig. 6.

Type VII collagen positively regulates ELMO2 expression in cSCC keratinocytes. (A) Type VII collagen re-expression in SCCRDEB3 keratinocytes (C7-6) results in an increase in ELMO2 expression. The expression levels were calculated using the ΔΔCT method with GAPDH as the normaliser and empty vector control (EV) as the calibrator. (B) Knockdown of type VII collagen was achieved using a siRNA pool (SiC7 pool) and compared with non-targeting control (NT). ELMO2 mRNA levels showed a significant decrease in both UV-induced cSCC and RDEB cSCC keratinocytes overexpressing type VII collagen following knockdown of COL7A1 compared with the empty vector control cells. (C) Western blot confirms an increase in ELMO2 expression in SCCRDEB3 keratinocytes following COL7A1 expression and a reduction in ELMO2 expression in UV-induced cSCC keratinocytes (SCCT1) following knockdown of COL7A1. Adjacent western blot quantification graphs show percentage average relative ELMO2 protein density (n = 3, means ± s.d). (D) Immunofluorescence images confirm the loss of ELMO2 expression following knockdown of type VII collagen in SCCT8. (E) ELMO2 expression was significantly increased when SCCRDEB3 and SCCRDEB2 cells were cultured on matrix derived from cells overexpressing type VII collagen (C7-6) compared with the empty vector control (EV). (F) Immunofluorescence images confirming an increase in ELMO2 expression when SCCRDEB3 cells were cultured on matrix derived from C7-6 cells compared with EV. Experiments were performed in triplicate and the data are shown as the means ± s.d.; **P<0.005, *P<0.05. Scale bars: 10 µm.

We next investigated the link between ELMO2 and SLCO1B3 expression and showed that overexpression of ELMO2 resulted in a reduction in SLCO1B3 mRNA levels in SCCT8 and SCCRDEB2 (Fig. 7A,B), whereas a reduction in ELMO2 mRNA resulted in an increase in SLCO1B3 expression in SCCT8 (Fig. 7C,D) and non-SCC RDEB keratinocytes (supplementary material Fig. S2A–C), with a significant increase in SLCO1B3 observed in normal primary keratinocytes after ELMO2 reduction (supplementary material Fig. S2D–F), thereby demonstrating that similar to COL7A1, ELMO2 can also negatively regulate SLCO1B3.

Fig. 7.

ELMO2 and Ln332 regulate SLCO1B3 expression in cSCC keratinocytes. (A) ELMO2 was transiently overexpressed in SCCT8 and SCCRDEB2 cells. Inset western blot confirms ELMO2 overexpression in SCCRDEB2 cells. ELMO2, ELMO2-overexpressing cells; EV, empty vector control (B) SLCO1B3 expression was reduced in both SCCT8 and SCCRDEB2 keratinocytes following ELMO2 overexpression, (C) ELMO2 knockdown was achieved in SCCT8 using a pool of three siRNA constructs (SiEL2 pool). (D) SLCO1B3 mRNA levels were increased following ELMO2 depletion compared with the non-targeting control (NT). (E) Western blot showing a reduction of the gamma 2 chain of Ln332 expression in SCCRDEB3 cells following knockdown using a pool of three siRNA constructs after 48 and 72 hours. SLCO1B3 expression was increased in Ln332 knockdown cells (SiLn332-48 hr/72 hr) compared with the non-targeting control (NT). Experiment was performed in triplicate and data are represented as the means ± s.d.; *P<0.05.

Fig. 7.

ELMO2 and Ln332 regulate SLCO1B3 expression in cSCC keratinocytes. (A) ELMO2 was transiently overexpressed in SCCT8 and SCCRDEB2 cells. Inset western blot confirms ELMO2 overexpression in SCCRDEB2 cells. ELMO2, ELMO2-overexpressing cells; EV, empty vector control (B) SLCO1B3 expression was reduced in both SCCT8 and SCCRDEB2 keratinocytes following ELMO2 overexpression, (C) ELMO2 knockdown was achieved in SCCT8 using a pool of three siRNA constructs (SiEL2 pool). (D) SLCO1B3 mRNA levels were increased following ELMO2 depletion compared with the non-targeting control (NT). (E) Western blot showing a reduction of the gamma 2 chain of Ln332 expression in SCCRDEB3 cells following knockdown using a pool of three siRNA constructs after 48 and 72 hours. SLCO1B3 expression was increased in Ln332 knockdown cells (SiLn332-48 hr/72 hr) compared with the non-targeting control (NT). Experiment was performed in triplicate and data are represented as the means ± s.d.; *P<0.05.

Type VII collagen can regulate SLCO1B3 independent of ELMO2

To confirm whether expression of SLCO1B3 in cSCC was dependent on expression of either type VII collagen or ELMO2 alone, we investigated SLCO1B3 expression following knockdown of ELMO2 in RDEB cSCC cells overexpressing type VII collagen (RDEBSCC3-C7) and control cells (RDEBSCC3-EV). SLCO1B3 expression significantly increased in RDEBSCC3-EV cells following ELMO2 knockdown, demonstrating that ELMO2 can regulate SLCO1B3 in the absence of type VII collagen (supplementary material Fig. S3A,B). In RDEBSCC3-C7 cells, no significant difference in SLCO1B3 expression was observed following ELMO2 knockdown (supplementary material Fig. 3A,B), showing that depletion of ELMO2 had no effect on SLCO1B3 expression. In addition, we show that ELMO2 does not regulate type VII collagen, as observed in SCCT8 (supplementary material Fig. 3C–E) or normal primary keratinocytes (supplementary material Fig. S2A–C) following knockdown of ELMO2. We also investigated Ln332, a binding partner of type VII collagen, which is also known to regulate polarity in epithelial cells (Kligys et al., 2007). In RDEB cSCC keratinocytes, siRNA-mediated knockdown of Ln332 resulted in an increase in SLCO1B3 message levels (Fig. 7E,F). This effect was independent of ELMO2 expression (supplementary material Fig. S3F). Collectively, these data show that SLCO1B3 is regulated independently by proteins with known roles in epidermal polarity (ELMO2 and Ln332) and that SLCO1B3 expression is controlled through a single mechanism governed by either of these proteins or type VII collagen alone.

Type VII collagen regulates front-to-rear keratinocyte polarity in 2D culture and regulates structural organisation in 3D spheroid culture

Ln332 and ELMO2 have previously been shown to play a role in epidermal polarity (Li et al., 2003; Ho and Dagnino, 2012b). To investigate whether type VII collagen could also regulate keratinocyte front-to-rear polarity, we seeded RDEBSCC C7 cells at low density and assessed lamellipodia formation at 5 hours as described previously (Hamill et al., 2009) (Fig. 8A). Type VII collagen expression marginally increased the number of attached cells that spread onto uncoated plastic (Fig. 8B) but strikingly increased the number of lamellipodia observed (Fig. 8C). Cells displaying front-to-rear polarity were frequently observed in RDEBSCC C7 cells, but were completely absent in vector control cells after 5 hours (Fig. 8A). The same experiment was repeated in RDEBSCC3 cells following transient overexpression of ELMO2 to confirm the role of ELMO2 in RDEB cSCC polarity. ELMO2-overexpressing cells (ELMO2) showed a significant increase in the number of polarised cells compared with control cells (EV) (supplementary material Fig. S4).

Fig. 8.

Overexpression of type VII collagen imparts front-to-rear polarity and increased expression of polarity-associated proteins in cSCC. (A) Confocal images of individual cells labelled with actin (green) showing retrovirally transduced keratinocytes overexpressing type VII collagen (C7-6) form lamellipodia (indicated by white arrows) and establish front-to-rear polarity earlier in comparison to control cells (EV). (B,C) The adjacent graphs show the percentage of spread cells and lamellipodia formation per spread cell in C7-6 and EV. Experiment was performed in triplicate and data are represented as means ± s.e.m. *P<0.005. (D) C7-6 3D mono-cultured spheroids show increased structural organisation compared to EV cells. PAR3 expression at the plasma membrane is reduced in EV spheroids compared with C7-6 spheroids. (E) Western blot analysis from lysates extracted from C7-6 spheroids show increased Ln332, PAR3, ITAG6 and ELMO2 protein expression compared with EV spheroids. Adjacent western blot quantification graph gives the percentage average relative density for the proteins shown (n = 3; means ± s.d). (F) Fibroblasts overexpressing type VII collagen (C7) impart increased structural organisation when co-cultured with SCCRDEB3 keratinocytes compared with EV fibroblast. 3D culture, H&E and DAPI images show the difference in structural organisation between C7 and EV cells. E-cadherin (green) and β-catenin (red) show reduced plasma membrane localisation and expression in EV cells compared with C7 cells. The adjacent bar graphs show image quantification comparing fluorescence levels for β-catenin and E-cadherin in EV and C7 co-culture spheroids. (G) Western blot showing E-cadherin and β-catenin protein expression in RDEBSCC3 co-culture spheroids shown in F. Scale bars: 10 µm.

Fig. 8.

Overexpression of type VII collagen imparts front-to-rear polarity and increased expression of polarity-associated proteins in cSCC. (A) Confocal images of individual cells labelled with actin (green) showing retrovirally transduced keratinocytes overexpressing type VII collagen (C7-6) form lamellipodia (indicated by white arrows) and establish front-to-rear polarity earlier in comparison to control cells (EV). (B,C) The adjacent graphs show the percentage of spread cells and lamellipodia formation per spread cell in C7-6 and EV. Experiment was performed in triplicate and data are represented as means ± s.e.m. *P<0.005. (D) C7-6 3D mono-cultured spheroids show increased structural organisation compared to EV cells. PAR3 expression at the plasma membrane is reduced in EV spheroids compared with C7-6 spheroids. (E) Western blot analysis from lysates extracted from C7-6 spheroids show increased Ln332, PAR3, ITAG6 and ELMO2 protein expression compared with EV spheroids. Adjacent western blot quantification graph gives the percentage average relative density for the proteins shown (n = 3; means ± s.d). (F) Fibroblasts overexpressing type VII collagen (C7) impart increased structural organisation when co-cultured with SCCRDEB3 keratinocytes compared with EV fibroblast. 3D culture, H&E and DAPI images show the difference in structural organisation between C7 and EV cells. E-cadherin (green) and β-catenin (red) show reduced plasma membrane localisation and expression in EV cells compared with C7 cells. The adjacent bar graphs show image quantification comparing fluorescence levels for β-catenin and E-cadherin in EV and C7 co-culture spheroids. (G) Western blot showing E-cadherin and β-catenin protein expression in RDEBSCC3 co-culture spheroids shown in F. Scale bars: 10 µm.

To investigate whether type VII collagen can influence structural organisation, 3D spheroid cultures were established with SCCRDEB3-C7 cells and compared with the SCCRDEB3-EV. SCCRDEB3-C7 spheroids showed increased structural organisation determined by morphology (Fig. 8D) and overall increased levels of proteins previously associated with cellular polarity, namely PAR3, ELMO2, integrin alpha 6 and Ln332 when compared with SCCRDEB3-EV spheroids (Fig. 8D,E). Next, we assessed the effect of stromal-derived type VII collagen on RDEB SCC cells. To do this, we co-cultured parental SCCRDEB3 cells with fibroblasts either expressing recombinant type VII collagen or vector control in a spheroid assay over 7 days. After this time we observed that mesenchyme delivery of type VII collagen had a more profound effect on RDEBSCC keratinocytes in spheroid culture than on RDEBSCC keratinocytes cultured alone in the presence or absence of type VII collagen. Spheroid cultures of RDEBSCC cells with fibroblast-delivered type VII collagen exhibited more distinct separation of cellular populations and a clear increase in distribution to the membrane of β-catenin and E-cadherin (Fig. 8F). Western blot analyses showed no overall changes in E-cadherin and β-catenin protein levels (Fig. 8G), suggesting that the differences observed in solvent-fixed embedded spheroids (Fig. 8F) are a result of membrane localisation.

DISCUSSION

cSCC is the most common cause of NMSC mortality and represents a significant clinical complication in high risk groups, particularly patients with RDEB. UV exposure resulting in DNA damage is a well-established driver of cSCC (Armstrong and Kricker, 2001). By contrast, RDEB-associated cSCC disease aggression is regarded as UV independent, and although tumour initiation is still not fully understood (Fine et al., 2009), recent work from our laboratory demonstrates the role of microenvironment in RDEB tumour progression (Ng et al., 2012). Here, our study began by interrogating a small number of changes identified through expression array analysis comparing UV cSCC keratinocytes with RDEB cSCC keratinocytes in culture. In agreement with multiple previous studies, few differences were confirmed between these two tumour types (Arbiser et al., 2004; Mallipeddi et al., 2004). In fact, our primary target, SLCO1B3 is aberrantly expressed in a panel of cSCC from both UV induced and RDEB induced cSCC cell lines and tissue (Fig. 1), again in agreement with previous studies identifying SLCO1B3 expression in cancer (Maeda et al., 2010; Pressler et al., 2011; Svoboda et al., 2011). However, by pursuing the molecule responsible for RDEB we show that type VII collagen regulates the expression of SLCO1B3 in both RDEB cSCC and UV-induced cSCC (Fig. 3). Further to this, we demonstrate that front-to-rear polarity and organisation of 3D spheroid culture coincides with changes in expression of SLCO1B3 in cSCC keratinocytes when type VII collagen levels are altered (Fig. 8).

Polarity is essential for the maintenance of tissue homeostasis and can either be defined as the distribution of apical and basal membrane domains within the cell or the positioning of a cell within the tissue itself, referred to as planar cell polarity (PCP). Directional transport, asymmetric stem cell division, oriented cell division and sensing signals from the neighbouring microenvironment are a few examples of cellular processes that require the maintenance of cell and tissue polarity (Lechler and Fuchs, 2005; Martin-Belmonte and Mostov, 2008; Poulson and Lechler, 2010; Martin-Belmonte and Perez-Moreno, 2012). Previously, three main protein complexes have been implicated in maintenance of polarity in simple epithelia, Par6–Par3–aPKC (Munro, 2006; Goldstein and Macara, 2007), the Crumbs–Stardust–PATJ complex and the Scribble–Lgl–Dlg complex (Martin-Belmonte and Mostov, 2008; Martín-Belmonte et al., 2008). However, the role of these protein complexes in maintenance of stratified epithelia polarity is not clear (Muroyama and Lechler, 2012; Niessen et al., 2012) and recent work has shown that the engulfment and cell motility protein ELMO2 positively regulates front-to-rear keratinocyte polarity by forming complexes with ILK in the presence of EGF (Ho et al., 2009; Ho and Dagnino, 2012a; Ho and Dagnino, 2012b). The data presented here identify for the first time a link between ELMO2 and type VII collagen, demonstrating that, like ELMO2, type VII collagen can regulate front-to-rear polarity (Fig. 8A). We go on to show that, like type VII collagen, ELMO2 negatively regulates SLCO1B3 (Fig. 7 and supplementary material Fig. S2) and this effect is independent of the regulation of ELMO2 by type VII collagen (supplementary material Fig. S3). We also show that Ln332 regulates SLCO1B3 independent of ELMO2 (Fig. 7E,F). Together, these data suggest that SLCO1B3 expression is not modulated by a common mechanism underlying an interaction between type VII collagen and ELMO2, but instead suggests that SLCO1B3 expression is modulated by cellular polarity.

Type VII collagen and Ln332 are secreted proteins important for epidermal keratinocyte attachment (Ogawa et al., 2007). Type VII collagen binds Ln332, a key component of hemidesmosomes and focal adhesions (Chen et al., 1999). Numerous components of these adhesion complexes have been implicated in regulating epithelial polarity (Cox et al., 2001; Martin-Belmonte and Perez-Moreno, 2012) and indeed, ELMO2 is shown to regulate front-to-rear polarity in the presence of Ln332 (Ho et al., 2009; Ho and Dagnino, 2012a). Similarly, integrin α6β4 regulates keratinocyte attachment through Ln332 and has a well-established role in directional migration (Jones et al., 1991; Dowling et al., 1996) because integrin activation of focal adhesion kinases is required for leading edge formation and directional movement (Lim et al., 2010). Integrin α6β4 also interacts with BPAG1e, promoting keratinocyte polarity through modulation of RAC1 (Hamill et al., 2009). We have shown that in the absence of type VII collagen ITGA6 and Ln332 expression is significantly reduced in 3D spheroid cultures (Fig. 8), suggesting a common mechanism, although further work is necessary to confirm this.

E-cadherin and β-catenin have described roles in numerous polarity pathways: APC (den Elzen et al., 2009), SCRIB (Navarro et al., 2005), PAR3 (Le Borgne et al., 2002) and Hippo (Kim et al., 2011); all of which are characterised by a downregulation in cancer progression (Muroyama and Lechler, 2012; Niessen et al., 2012). Our data showing that type VII collagen regulates the expression of membrane-bound E-cadherin in 3D spheroid culture is in agreement with a previous study of siRNA depletion of COL7A1 cSCC showing reduced E-cadherin in 3D organotypic culture (Martins et al., 2009). Further work will be necessary to determine the precise role that individual complexes play in type-VII-collagen-mediated polarity and organisation. Toward this end, we show here a reduction of PAR3 in RDEB cSCC spheroid culture and it is interesting to note that altered PAR3 distribution has been observed as a result of reduced E-cadherin expression (Le Borgne et al., 2002) and that deletion of PAR3 in murine epidermis does not alter normal skin development or function but does favour keratoacanthoma formation. Consequently, wild-type PAR3 mice show increased SCC formation on a background of RAS mutations, suggesting a dual role of PAR3 in skin carcinogenesis (Iden et al., 2012).

Reorganisation of polarity can have a profound effect on proteins located at the cell surface. OATP1B3 is a transmembrane transporter that localises in the basolateral domain of the plasma membrane in normal liver and the data presented here strongly support a change in polarity as an explanation for SLCO1B3 expression on the background of altered tumour cell behaviour. In addition to the identification that type VII collagen regulates OATP1B3 expression, we confirm loss of plasma-membrane-bound OATP1B3 expression in 2D cultured cells, a phenomenon also documented in primary hepatocytes (Ulvestad et al., 2011). Two recent studies have identified an N-terminal-truncated isoform of OATP1B3, OATP1B3 V1 (lacking the first 28 amino acids), as being the predominant OATP1B3 expressed in colon, lung and pancreatic cancer (Nagai et al., 2012; Thakkar et al., 2013). Thakkar and colleagues demonstrate that OATP1B3 V1 localises predominantly to the cytoplasm in culture (Thakkar et al., 2013). Our data show that membrane-localised OATP1B3 is absent in cultured SCCT8 cells, but readily detected when these cells are grown in vivo (Fig. 2C). Therefore, a more detailed analysis of OATP1B3 isoforms expressed in cSCC under these varied conditions is warranted. Because it has been suggested that OATP substrates and inhibitors can be used either to enable uptake of xenobiotics for the specific targeting of cancer cells or by inhibiting OATP-mediated transport of hormones to prevent resistance to existing therapies (Maeda et al., 2010; Svoboda et al., 2011; Obaidat et al., 2012), it will be of interest to understand the relationship between transport and isoform expression.

In summary, we show that tumour expression of SLCO1B3 in cSCC can be regulated by type VII collagen and ELMO2 and that expression of type VII collagen in RDEB SCC keratinocytes increases front-to-rear polarity and structural organisation in spheroid culture coinciding with dramatic reorganisation of the localisation of β-catenin and E-cadherin proteins.

MATERIALS AND METHODS

All human samples were collected after informed, written consent and in accordance with Helsinki guidelines. All animals were used in accordance with UK Home Office regulations.

Keratinocyte culture and xenograft tumour material

Keratinocyte isolation, organotypic culture and xenograft tumour growth were achieved as previously described (Nyström et al., 2005; Watt et al., 2011).

Sybr®Green qPCR validation

cDNA was prepared from RNA isolated using RNA-Bee (AMS Biotechnology, Abingdon-on-Thames, UK) and the QuantiTect reverse transcription kit (Qiagen, Manchester, UK). The 2× Sybr®Green master mix (Qiagen, Manchester, UK) along with the SLCO1B3 forward (5′-GTCCAGTCATTGGCTTTGCA-3′) and reverse (5′-CAACCCAACGAGAGTCCTTAGG-3′) primers, ELMO2 forward (5′-GGGACATGGTTTCAATCACC-3′) and reverse (5′-TGGAGGTGTGAGATGAGCTG-3′) and previously published COL7A1 primers (Wang et al., 2006) were used to quantify mRNA levels. GAPDH primers (Forward 5′-TTCCGGGAAACTGTGGCGTGA-3′ and reverse 5′-ACGGAAGGCCATGCCAG-3′) and previously published ACTB primers (Ehrenfeld et al., 2011) were used to normalise expression data. Raw data were analysed with the Rotor-Gene Q qPCR software (Qiagen, Manchester, UK) using the ΔΔCT method.

Retroviral transduction

The open reading frame of SLCO1B3 and COL7A1 were cloned from an IMAGE cDNA clone (Open Biosystems, Thermo Scientific, Huntsville, AL) or normal keratinocyte cDNA, respectively, and inserted into the pBabe-puro vector using standard molecular biology techniques. RDEB cSCC keratinocytes were retrovirally transduced using the Phoenix retroviral packaging system (Kinsella and Nolan, 1996). All transduced cells were selected using 2 µg/ml puromycin. SCCRDEB3 cells were then clonally selected to generate a homogenous population expressing a high level of type VII collagen. SCCRDEB4 were also transduced as described but without clonal selection.

Immunoblotting and Immunofluorescence

Protein lysates from 2 days post confluent cells were extracted using the radioimmunoprecipitation assay (RIPA) buffer and quantified using the bicinchoninic acid (BCA) assay (ThermoFisher Scientific, Waltham, MA). 4–15% Mini-PROTEAN TGX gels (Bio-Rad, Hertfordshire, UK) were used for electrophoresis.

Formalin-fixed paraffin-embedded (FFPE), fresh frozen tissue sections and cSCC primary keratinocytes were used. Briefly, FFPE sections were rehydrated using 100%, 95% and 70% ethanol and then treated with 1% citric acid buffer (pH 6) at 100°C. FFPE, fresh frozen sections or keratinocytes cultured on coverslips were stained with the primary antibody for 1 hour followed by a 40 minute treatment with the secondary antibody and counterstained with DAPI (Invitrogen, Paisley, UK). Slides were imaged using the Zeiss Axioskop 2 fluorescent microscope (Carl Zeiss Microscopy) and analysed using the Axiovision software or the Leica TCS SP5 confocal microscope and the Leica Application Suite Advanced fluorescence Lite software (Leica Microsystems, Wetzlar, Germany).

Antibodies

The following antibodies were used for immunofluorescence and western blot analysis. Primary antibodies used: OATP1B3, H-52 (0.4 µg/ml; Santa Cruz Biotechnology, Dallas, TX); type VII collagen, 234192 (Calbiochem, Darmstadt, Germany); ELMO2, either the E-16 clone (0.8 µg/ml, Santa Cruz) or HPA018811 (0.4 µg/ml; Sigma-Aldrich); integrin alpha 6, 3750 (Cell Signalling, Danvers, MA); E-cadherin: (2.5 µg/ml; 610101, BD Transduction Laboratories™, San Jose, CA); Laminin 332, γ chain MAB19562 (1 µg/ml; Calbiochem); β-catenin, SC-7199 (4 µg/ml; Santa Cruz); PAR3, 07-330 (0.4 µg/ml, Calbiochem); actin, SC-58625 (0.2 µg/ml; Santa Cruz); plakogloblin, 61005 (2.5 µg/ml; Progen Biotechnik, Heidelburg, Germany). Secondary antibodies were: Alexa Fluor 594® rabbit (2.5 µg/ml; Invitrogen) or Alexa Fluor 488® mouse (2.5 µg/ml, Invitrogen)

Luciferase gene reporter assay

A luciferase gene reporter construct was created by inserting the SLCO1B3 promoter (1 kb upstream of the 5′UTR region of the gene: 11375105–11376214 of accession number: NW_001838052.1) (Jung et al., 2002) into the pGL3 basic vector using standard molecular biology techniques. SCCRDEB3 cells transduced with COL7A1 ORF or empty vector control were transfected with the SLCO1B3 promoter construct using Lipofectamine 2000 (Invitrogen) according to the manufacturer's recommendations. Cells were lysed and harvested after 18 hours and luciferase activity was recorded using the Dual-Luciferase® reporter assay system (Promega, Madison, WI) and a Modulus II plate reader (Promega). Data were normalised for transfection efficiency using Renilla and empty vector control cells were used as a calibrator.

siRNA knockdown of COL7A1, LAMC2 and ELMO2

Three siRNA constructs against COL7A1 (SASI_Hs01_00155170/71/73, Sigma-Aldrich), LAMC2 (SASI_Hs01_00136952/53/54, Sigma-Aldrich), ELMO2 (SASI_Hs01_00078453/54/55, Sigma-Aldrich) and a custom designed non-targeting control (sense: GCUGCUAGACGUGUUAUGA[dT][dT], antisense: UCAUAACACGUCUAGCAGC[dT][dT], Sigma-Aldrich) were transfected using Lipofectamine 2000 (Invitrogen). RNA was extracted 24, 48 and 72 hours post transfection and analysed for COL7A1, SLCO1B3, LAMC2 and ELMO2 expression using Sybr®Green qPCR.

ELMO2 overexpression

An ELMO2 cDNA construct (MHS1010-57517, Thermo Scientific) and the empty vector control was transfected into cSCC keratinocytes using Lipofectamine 2000 (Invitrogen). RNA and protein was harvested 17 hours and 24 hours post transfection and assessed for ELMO2 overexpression using Sybr®Green qPCR.

Cell-derived matrix

SCCRDEB3-C7 and SCCRDEB3-EV cells were cultured in a six-well plate until 2 days post confluent. Cells were washed with cold PBS and lysed with 1 ml 0.5% sodium deoxycholate for 10 minutes on ice. Lysed cells were removed and wells were washed with dH2O and air dried. SCCRDEB2 and SCCRDEB3 parental cells were seeded on either collagen VII containing matrix or empty vector control matrix for 3 days before RNA and protein samples were harvested.

Cell plating lamellipodia scoring

500,000 SCCRDEB3-C7 and SCCRDEB-EV cells were seeded in a 10 cm dish. Cells were imaged 5 hours post seeding. Twenty frames were scored blind for each cell type and data was analysed from three independent experiments. Clear cellular protrusions were counted for lamellipodia scoring and the difference in cell diameter was observed in order to accurately score for cell spreading. A Leica TCS SP5 confocal microscope was used to image cells after staining with β-actin and analysed using the Leica Application Suite Advanced fluorescence Lite software (Leica Microsystems, Wetzlar, Germany).

Spheroid monoculture

SCCRDEB3 type VII collagen overexpressing (SCCRDEB3-C7) and the empty vector (SCCRDEB3-EV) cells were cultured for 72 hours in a hanging-drop non-adherent culture. Briefly, 6000 cells were seeded on the lid of a 10 cm culture dish and allowed to form spheroids. Each spheroid was washed with PBS and harvested using a syringe and placed in 4% paraformaldehyde overnight. Spheroids were washed with PBS next day and placed at 4°C and embedded in paraffin. Alternatively, spheroids were washed with PBS and directly frozen in OCT. Spheroids were serially sectioned for immunofluorescence.

Spheroid co-culture

Fibroblasts were transduced with COL7A1 or empty vector controls. 3000 fibroblasts were seeded in 20 µL medium on a sterile 10 cm cell culture dish lid and incubated for four days at 37°C. On day four, 10 µl of culture medium was removed from each droplet. 6000 keratinocytes were seeded in 10 µl of medium onto each droplet of previously formed fibroblasts spheroids and cultured for 3 days. Spheroids were washed with PBS and embedded in OCT for serial sectioning and staining. In parallel, and to visualise the separate populations, cells were labelled with PKH67 green fluorescent cell linker kit (fibroblasts) or PKH26 red fluorescent cell linker kit (keratinocytes) according to the manufacturer's protocol (Sigma-Aldrich). Labelled spheroids were visualised using a Carl Zeiss Axiovert25 CFL microscope.

Statistical analysis

Student's two tailed t-tests were used to determine statistical significance of luciferase gene reporter assays and COL7A1 siRNA knockdown data.

Acknowledgements

We would like to thank Karl Gledhill for technical assistance.

Author contributions

A.P.S. initiated and supervised the work. A.P.S., J.H.S.D., C.L.C., C.P., S.A.W. and Y.Z.L. conducted experiments. All authors participated in data interpretation and manuscript preparation.

Funding

This work was funded by the European Research Council (ERC) [grant number 250170]; DEBRA UK and DEBRA International – funded by DEBRA Austria. I.M.L. is funded by a strategic grant from the Wellcome Trust. Deposited in PMC for release after 6 months.

References

Alam
M.
,
Ratner
D.
(
2001
).
Cutaneous squamous-cell carcinoma.
N. Engl. J. Med.
344
,
975
983
.
Arbiser
J. L.
,
Fan
C. Y.
,
Su
X.
,
Van Emburgh
B. O.
,
Cerimele
F.
,
Miller
M. S.
,
Harvell
J.
,
Marinkovich
M. P.
(
2004
).
Involvement of p53 and p16 tumor suppressor genes in recessive dystrophic epidermolysis bullosa-associated squamous cell carcinoma.
J. Invest. Dermatol.
123
,
788
790
.
Armstrong
B. K.
,
Kricker
A.
(
2001
).
The epidemiology of UV induced skin cancer.
J. Photochem. Photobiol. B
63
,
8
18
.
Burgeson
R. E.
(
1993
).
Type VII collagen, anchoring fibrils, and epidermolysis bullosa.
J. Invest. Dermatol.
101
,
252
255
.
Chen
M.
,
Marinkovich
M. P.
,
Veis
A.
,
Cai
X.
,
Rao
C. N.
,
O'Toole
E. A.
,
Woodley
D. T.
(
1997
).
Interactions of the amino-terminal noncollagenous (NC1) domain of type VII collagen with extracellular matrix components. A potential role in epidermal-dermal adherence in human skin.
J. Biol. Chem.
272
,
14516
14522
.
Chen
M.
,
Marinkovich
M. P.
,
Jones
J. C.
,
O'Toole
E. A.
,
Li
Y. Y.
,
Woodley
D. T.
(
1999
).
NC1 domain of type VII collagen binds to the beta3 chain of laminin 5 via a unique subdomain within the fibronectin-like repeats.
J. Invest. Dermatol.
112
,
177
183
.
Christiano
A. M.
,
Greenspan
D. S.
,
Hoffman
G. G.
,
Zhang
X.
,
Tamai
Y.
,
Lin
A. N.
,
Dietz
H. C.
,
Hovnanian
A.
,
Uitto
J.
(
1993
).
A missense mutation in type VII collagen in two affected siblings with recessive dystrophic epidermolysis bullosa.
Nat. Genet.
4
,
62
66
.
Cox
E. A.
,
Sastry
S. K.
,
Huttenlocher
A.
(
2001
).
Integrin-mediated adhesion regulates cell polarity and membrane protrusion through the Rho family of GTPases.
Mol. Biol. Cell
12
,
265
277
.
Czarnecki
D.
,
Staples
M.
,
Mar
A.
,
Giles
G.
,
Meehan
C.
(
1994
).
Metastases from squamous cell carcinoma of the skin in southern Australia.
Dermatology
189
,
52
54
.
Daley
W. P.
,
Gervais
E. M.
,
Centanni
S. W.
,
Gulfo
K. M.
,
Nelson
D. A.
,
Larsen
M.
(
2012
).
ROCK1-directed basement membrane positioning coordinates epithelial tissue polarity.
Development
139
,
411
422
.
den Elzen
N.
,
Buttery
C. V.
,
Maddugoda
M. P.
,
Ren
G.
,
Yap
A. S.
(
2009
).
Cadherin adhesion receptors orient the mitotic spindle during symmetric cell division in mammalian epithelia.
Mol. Biol. Cell
20
,
3740
3750
.
Dowling
J.
,
Yu
Q. C.
,
Fuchs
E.
(
1996
).
Beta4 integrin is required for hemidesmosome formation, cell adhesion and cell survival.
J. Cell Biol.
134
,
559
572
.
Ehrenfeld
P.
,
Conejeros
I.
,
Pavicic
M. F.
,
Matus
C. E.
,
Gonzalez
C. B.
,
Quest
A. F.
,
Bhoola
K. D.
,
Poblete
M. T.
,
Burgos
R. A.
,
Figueroa
C. D.
(
2011
).
Activation of kinin B1 receptor increases the release of metalloproteases-2 and -9 from both estrogen-sensitive and -insensitive breast cancer cells.
Cancer Lett.
301
,
106
118
.
Fears
T. R.
,
Scotto
J.
(
1983
).
Estimating increases in skin cancer morbidity due to increases in ultraviolet radiation exposure.
Cancer Invest.
1
,
119
126
.
Fine
J. D.
,
Johnson
L. B.
,
Weiner
M.
,
Li
K. P.
,
Suchindran
C.
(
2009
).
Epidermolysis bullosa and the risk of life-threatening cancers: the National EB Registry experience, 1986-2006.
J. Am. Acad. Dermatol.
60
,
203
211
.
Goldstein
B.
,
Macara
I. G.
(
2007
).
The PAR proteins: fundamental players in animal cell polarization.
Dev. Cell
13
,
609
622
.
Hagenbuch
B.
,
Gui
C.
(
2008
).
Xenobiotic transporters of the human organic anion transporting polypeptides (OATP) family.
Xenobiotica
38
,
778
801
.
Hamada
A.
,
Sissung
T.
,
Price
D. K.
,
Danesi
R.
,
Chau
C. H.
,
Sharifi
N.
,
Venzon
D.
,
Maeda
K.
,
Nagao
K.
,
Sparreboom
A.
 et al. (
2008
).
Effect of SLCO1B3 haplotype on testosterone transport and clinical outcome in caucasian patients with androgen-independent prostatic cancer.
Clin. Cancer Res.
14
,
3312
3318
.
Hamill
K. J.
,
Hopkinson
S. B.
,
DeBiase
P.
,
Jones
J. C.
(
2009
).
BPAG1e maintains keratinocyte polarity through beta4 integrin-mediated modulation of Rac1 and cofilin activities.
Mol. Biol. Cell
20
,
2954
2962
.
Harwood
C. A.
,
Proby
C. M.
(
2002
).
Human papillomaviruses and non-melanoma skin cancer.
Curr. Opin. Infect. Dis.
15
,
101
114
.
Ho
E.
,
Dagnino
L.
(
2012a
).
Emerging role of ILK and ELMO2 in the integration of adhesion and migration pathways.
Cell Adh. Migr.
6
,
168
172
.
Ho
E.
,
Dagnino
L.
(
2012b
).
Epidermal growth factor induction of front-rear polarity and migration in keratinocytes is mediated by integrin-linked kinase and ELMO2.
Mol. Biol. Cell
23
,
492
502
.
Ho
E.
,
Irvine
T.
,
Vilk
G. J.
,
Lajoie
G.
,
Ravichandran
K. S.
,
D'Souza
S. J.
,
Dagnino
L.
(
2009
).
Integrin-linked kinase interactions with ELMO2 modulate cell polarity.
Mol. Biol. Cell
20
,
3033
3043
.
Iden
S.
,
van Riel
W. E.
,
Schäfer
R.
,
Song
J. Y.
,
Hirose
T.
,
Ohno
S.
,
Collard
J. G.
(
2012
).
Tumor type-dependent function of the par3 polarity protein in skin tumorigenesis.
Cancer Cell
22
,
389
403
.
Jones
J. C.
,
Kurpakus
M. A.
,
Cooper
H. M.
,
Quaranta
V.
(
1991
).
A function for the integrin alpha 6 beta 4 in the hemidesmosome.
Cell Regul.
2
,
427
438
.
Jung
D.
,
Podvinec
M.
,
Meyer
U. A.
,
Mangelsdorf
D. J.
,
Fried
M.
,
Meier
P. J.
,
Kullak-Ublick
G. A.
(
2002
).
Human organic anion transporting polypeptide 8 promoter is transactivated by the farnesoid X receptor/bile acid receptor.
Gastroenterology
122
,
1954
1966
.
Kim
N. G.
,
Koh
E.
,
Chen
X.
,
Gumbiner
B. M.
(
2011
).
E-cadherin mediates contact inhibition of proliferation through Hippo signaling-pathway components.
Proc. Natl. Acad. Sci. USA
108
,
11930
11935
.
Kinsella
T. M.
,
Nolan
G. P.
(
1996
).
Episomal vectors rapidly and stably produce high-titer recombinant retrovirus.
Hum. Gene Ther.
7
,
1405
1413
.
Klaassen
C. D.
,
Aleksunes
L. M.
(
2010
).
Xenobiotic, bile acid, and cholesterol transporters: function and regulation.
Pharmacol. Rev.
62
,
1
96
.
Kligys
K.
,
Claiborne
J. N.
,
DeBiase
P. J.
,
Hopkinson
S. B.
,
Wu
Y.
,
Mizuno
K.
,
Jones
J. C.
(
2007
).
The slingshot family of phosphatases mediates Rac1 regulation of cofilin phosphorylation, laminin-332 organization, and motility behavior of keratinocytes.
J. Biol. Chem.
282
,
32520
32528
.
Le Borgne
R.
,
Bellaïche
Y.
,
Schweisguth
F.
(
2002
).
Drosophila E-cadherin regulates the orientation of asymmetric cell division in the sensory organ lineage.
Curr. Biol.
12
,
95
104
.
Lee
W.
,
Belkhiri
A.
,
Lockhart
A. C.
,
Merchant
N.
,
Glaeser
H.
,
Harris
E. I.
,
Washington
M. K.
,
Brunt
E. M.
,
Zaika
A.
,
Kim
R. B.
,
El-Rifai
W.
(
2008
).
Overexpression of OATP1B3 confers apoptotic resistance in colon cancer.
Cancer Res
68(24)
,
10315
10323
.
Lechler
T.
,
Fuchs
E.
(
2005
).
Asymmetric cell divisions promote stratification and differentiation of mammalian skin.
Nature
437
,
275
280
.
Li
S.
,
Edgar
D.
,
Fässler
R.
,
Wadsworth
W.
,
Yurchenco
P. D.
(
2003
).
The role of laminin in embryonic cell polarization and tissue organization.
Dev. Cell
4
,
613
624
.
Lim
S. T.
,
Chen
X. L.
,
Tomar
A.
,
Miller
N. L.
,
Yoo
J.
,
Schlaepfer
D. D.
(
2010
).
Knock-in mutation reveals an essential role for focal adhesion kinase activity in blood vessel morphogenesis and cell motility-polarity but not cell proliferation.
J. Biol. Chem.
285
,
21526
21536
.
Lockhart
A. C.
,
Harris
E.
,
Lafleur
B. J.
,
Merchant
N. B.
,
Washington
M. K.
,
Resnick
M. B.
,
Yeatman
T. J.
,
Lee
W.
(
2008
).
Organic anion transporting polypeptide 1B3 (OATP1B3) is overexpressed in colorectal tumors and is a predictor of clinical outcome.
Clin. Exp. Gastroenterol
1
,
1
7
.
Maeda
T.
,
Irokawa
M.
,
Arakawa
H.
,
Kuraoka
E.
,
Nozawa
T.
,
Tateoka
R.
,
Itoh
Y.
,
Nakanishi
T.
,
Tamai
I.
(
2010
).
Uptake transporter organic anion transporting polypeptide 1B3 contributes to the growth of estrogen-dependent breast cancer.
J. Steroid Biochem. Mol. Biol.
122
,
180
185
.
Mallipeddi
R.
,
Wessagowit
V.
,
South
A. P.
,
Robson
A. M.
,
Orchard
G. E.
,
Eady
R. A.
,
McGrath
J. A.
(
2004
).
Reduced expression of insulin-like growth factor-binding protein-3 (IGFBP-3) in Squamous cell carcinoma complicating recessive dystrophic epidermolysis bullosa.
J. Invest. Dermatol.
122
,
1302
1309
.
Martin-Belmonte
F.
,
Mostov
K.
(
2008
).
Regulation of cell polarity during epithelial morphogenesis.
Curr. Opin. Cell Biol.
20
,
227
234
.
Martin-Belmonte
F.
,
Perez-Moreno
M.
(
2012
).
Epithelial cell polarity, stem cells and cancer.
Nat. Rev. Cancer
12
,
23
38
.
Martín-Belmonte
F.
,
Yu
W.
,
Rodríguez-Fraticelli
A. E.
,
Ewald
A. J.
,
Werb
Z.
,
Alonso
M. A.
,
Mostov
K.
(
2008
).
Cell-polarity dynamics controls the mechanism of lumen formation in epithelial morphogenesis.
Curr. Biol.
18
,
507
513
.
Martins
V. L.
,
Vyas
J. J.
,
Chen
M.
,
Purdie
K.
,
Mein
C. A.
,
South
A. P.
,
Storey
A.
,
McGrath
J. A.
,
O'Toole
E. A.
(
2009
).
Increased invasive behaviour in cutaneous squamous cell carcinoma with loss of basement-membrane type VII collagen.
J. Cell Sci.
122
,
1788
1799
.
McGregor
J. M.
,
Proby
C. M.
(
1996
).
The role of papillomaviruses in human non-melanoma skin cancer.
Cancer Surv.
26
,
219
236
.
Miller
D. L.
,
Weinstock
M. A.
(
1994
).
Nonmelanoma skin cancer in the United States: incidence.
J. Am. Acad. Dermatol.
30
,
774
778
.
Munro
E. M.
(
2006
).
PAR proteins and the cytoskeleton: a marriage of equals.
Curr. Opin. Cell Biol.
18
,
86
94
.
Muroyama
A.
,
Lechler
T.
(
2012
).
Polarity and stratification of the epidermis.
Semin. Cell Dev. Biol.
23
,
890
896
.
Nagai
M.
,
Furihata
T.
,
Matsumoto
S.
,
Ishii
S.
,
Motohashi
S.
,
Yoshino
I.
,
Ugajin
M.
,
Miyajima
A.
,
Matsumoto
S.
,
Chiba
K.
(
2012
).
Identification of a new organic anion transporting polypeptide 1B3 mRNA isoform primarily expressed in human cancerous tissues and cells.
Biochem. Biophys. Res. Commun.
418
,
818
823
.
Navarro
C.
,
Nola
S.
,
Audebert
S.
,
Santoni
M. J.
,
Arsanto
J. P.
,
Ginestier
C.
,
Marchetto
S.
,
Jacquemier
J.
,
Isnardon
D.
,
Le Bivic
A.
 et al. (
2005
).
Junctional recruitment of mammalian Scribble relies on E-cadherin engagement.
Oncogene
24
,
4330
4339
.
Ng
Y-Z.
,
Dayal
J. H. S.
,
South
A. P.
(
2011
).
Genetic predisposition to cutaneous squamous Cell carcinoma.
In
Skin Cancers – Risk Factors, Prevention and Therapy
La Porta
C
, ed
53
86
.
Rijeka, Croatia
:
InTech
.
Ng
Y. Z.
,
Pourreyron
C.
,
Salas-Alanis
J. C.
,
Dayal
J. H.
,
Cepeda-Valdes
R.
,
Yan
W.
,
Wright
S.
,
Chen
M.
,
Fine
J. D.
,
Hogg
F. J.
 et al. (
2012
).
Fibroblast-derived dermal matrix drives development of aggressive cutaneous squamous cell carcinoma in patients with recessive dystrophic epidermolysis bullosa.
Cancer Res.
72
,
3522
3534
.
Niessen
M. T.
,
Iden
S.
,
Niessen
C. M.
(
2012
).
The in vivo function of mammalian cell and tissue polarity regulators – how to shape and maintain the epidermal barrier.
J. Cell Sci.
125
,
3501
3510
.
Nyström
M. L.
,
Thomas
G. J.
,
Stone
M.
,
Mackenzie
I. C.
,
Hart
I. R.
,
Marshall
J. F.
(
2005
).
Development of a quantitative method to analyse tumour cell invasion in organotypic culture.
J. Pathol.
205
,
468
475
.
Obaidat
A.
,
Roth
M.
,
Hagenbuch
B.
(
2012
).
The expression and function of organic anion transporting polypeptides in normal tissues and in cancer.
Annu. Rev. Pharmacol. Toxicol.
52
,
135
151
.
Ogawa
T.
,
Tsubota
Y.
,
Hashimoto
J.
,
Kariya
Y.
,
Miyazaki
K.
(
2007
).
The short arm of laminin gamma2 chain of laminin-5 (laminin-332) binds syndecan-1 and regulates cellular adhesion and migration by suppressing phosphorylation of integrin beta4 chain.
Mol. Biol. Cell
18
,
1621
1633
.
Olk
S.
,
Zoidl
G.
,
Dermietzel
R.
(
2009
).
Connexins, cell motility, and the cytoskeleton.
Cell Motil. Cytoskeleton
66
,
1000
1016
.
Poulson
N. D.
,
Lechler
T.
(
2010
).
Robust control of mitotic spindle orientation in the developing epidermis.
J. Cell Biol.
191
,
915
922
.
Pressler
H.
,
Sissung
T. M.
,
Venzon
D.
,
Price
D. K.
,
Figg
W. D.
(
2011
).
Expression of OATP family members in hormone-related cancers: potential markers of progression.
PLoS ONE
6
,
e20372
.
Preston
D. S.
,
Stern
R. S.
(
1992
).
Nonmelanoma cancers of the skin.
N. Engl. J. Med.
327
,
1649
1662
.
Ramachandran
A.
,
Betts
G.
,
Bhana
S.
,
Helme
G.
,
Blick
C.
,
Moller-Levet
C.
,
Saunders
E.
,
Valentine
H.
,
Pepper
S.
,
Miller
C. J.
 et al. (
2013
).
An in vivo hypoxia metagene identifies the novel hypoxia inducible factor target gene SLCO1B3.
Eur. J. Cancer
49
,
1741
1751
.
Reed
W. B.
,
Roenigk
H.
 Jr
,
Dorner
W.
 Jr
,
Welsh
O.
,
Martin
F. J.
(
1975
).
Epidermal neoplasms with epidermolysis bullosa dystrophica with the first report of carcinoma with the acquired type.
Arch. Dermatol. Res.
253
,
1
14
.
Rousselle
P.
,
Keene
D. R.
,
Ruggiero
F.
,
Champliaud
M. F.
,
Rest
M.
,
Burgeson
R. E.
(
1997
).
Laminin 5 binds the NC-1 domain of type VII collagen.
J. Cell Biol.
138
,
719
728
.
Rowe
D. E.
,
Carroll
R. J.
,
Day
C. L.
 Jr
(
1992
).
Prognostic factors for local recurrence, metastasis, and survival rates in squamous cell carcinoma of the skin, ear, and lip. Implications for treatment modality selection.
J. Am. Acad. Dermatol.
26
,
976
990
.
Sakai
L. Y.
,
Keene
D. R.
,
Morris
N. P.
,
Burgeson
R. E.
(
1986
).
Type VII collagen is a major structural component of anchoring fibrils.
J. Cell Biol.
103
,
1577
1586
.
Smith
N. F.
,
Figg
W. D.
,
Sparreboom
A.
(
2005a
).
Role of the liver-specific transporters OATP1B1 and OATP1B3 in governing drug elimination.
Expert Opin. Drug Metab. Toxicol.
1
,
429
445
.
Smith
N. F.
,
Acharya
M. R.
,
Desai
N.
,
Figg
W. D.
,
Sparreboom
A.
(
2005b
).
Identification of OATP1B3 as a high-affinity hepatocellular transporter of paclitaxel.
Cancer Biol. Ther.
4
,
815
818
.
South
A. P.
,
O'Toole
E. A.
(
2010
).
Understanding the pathogenesis of recessive dystrophic epidermolysis bullosa squamous cell carcinoma.
Dermatol. Clin.
28
,
171
178
.
Svoboda
M.
,
Wlcek
K.
,
Taferner
B.
,
Hering
S.
,
Stieger
B.
,
Tong
D.
,
Zeillinger
R.
,
Thalhammer
T.
,
Jäger
W.
(
2011
).
Expression of organic anion-transporting polypeptides 1B1 and 1B3 in ovarian cancer cells: relevance for paclitaxel transport.
Biomed. Pharmacother.
65
,
417
426
.
Thakkar
N.
,
Kim
K.
,
Jang
E. R.
,
Han
S.
,
Kim
K.
,
Kim
D.
,
Merchant
N.
,
Lockhart
A. C.
,
Lee
W.
(
2013
).
A cancer-specific variant of the SLCO1B3 gene encodes a novel human organic anion transporting polypeptide 1B3 (OATP1B3) localized mainly in the cytoplasm of colon and pancreatic cancer cells.
Mol. Pharm.
10
,
406
416
.
Toms
J.
(
2004
).
CancerStats Monograph
London
:
Cancer Research UK
.
Ulvestad
M.
,
Björquist
P.
,
Molden
E.
,
Asberg
A.
,
Andersson
T. B.
(
2011
).
OATP1B1/1B3 activity in plated primary human hepatocytes over time in culture.
Biochem. Pharmacol.
82
,
1219
1226
.
Veness
M. J.
,
Quinn
D. I.
,
Ong
C. S.
,
Keogh
A. M.
,
Macdonald
P. S.
,
Cooper
S. G.
,
Morgan
G. W.
(
1999
).
Aggressive cutaneous malignancies following cardiothoracic transplantation: the Australian experience.
Cancer
85
,
1758
1764
.
Villone
D.
,
Fritsch
A.
,
Koch
M.
,
Bruckner-Tuderman
L.
,
Hansen
U.
,
Bruckner
P.
(
2008
).
Supramolecular interactions in the dermo-epidermal junction zone: anchoring fibril-collagen VII tightly binds to banded collagen fibrils.
J. Biol. Chem.
283
,
24506
24513
.
Wang
T. W.
,
Sun
J. S.
,
Huang
Y. C.
,
Wu
H. C.
,
Chen
L. T.
,
Lin
F. H.
(
2006
).
Skin basement membrane and extracellular matrix proteins characterization and quantification by real time RT-PCR.
Biomaterials
27
,
5059
5068
.
Watt
S. A.
,
Pourreyron
C.
,
Purdie
K.
,
Hogan
C.
,
Cole
C. L.
,
Foster
N.
,
Pratt
N.
,
Bourdon
J. C.
,
Appleyard
V.
,
Murray
K.
 et al. (
2011
).
Integrative mRNA profiling comparing cultured primary cells with clinical samples reveals PLK1 and C20orf20 as therapeutic targets in cutaneous squamous cell carcinoma.
Oncogene
30
,
4666
4677
.
Weinstock
M. A.
,
Bogaars
H. A.
,
Ashley
M.
,
Litle
V.
,
Bilodeau
E.
,
Kimmel
S.
(
1991
).
Nonmelanoma skin cancer mortality. A population-based study.
Arch. Dermatol.
127
,
1194
1197
.
Yanofsky
V. R.
,
Mercer
S. E.
,
Phelps
R. G.
(
2011
).
Histopathological variants of cutaneous squamous cell carcinoma: a review.
J. Skin Cancer
2011
,
210813
.

Competing interests

The authors declare no competing interests.

Supplementary information