Promotion of periostin expression contributes to the migration of Schwann cells

Schwann cell precursors and PNS sensory neurons originate from the neural crest. The neural crest cells that are fated to participate in the formation of dorsal root ganglia (DRG) migrate through the rostral half of each somite and condense in repetitive units on both sides of the spinal cord. These nascent sensory ganglia comprise neural and glial progenitors with heterogeneous developmental potentialities. Upon differentiation of sensory neurons and axonal outgrowth, Schwann cell precursors migrate from the DRG along these projections. Later in development they differentiate into myelinating or non-myelinating Schwann cells that envelope sensory- and motor-neuron axons. The development of peripheral nerve tissue results from a series of coordinated interactions between neurons and Schwann cells (Jessen and Mirsky, 2005). Targeted mutagenesis of different components of the ErbB-signalling complex has demonstrated its importance for Schwann cell development (Britsch et al., 1998; Garratt et al., 2000; Michailov et al., 2004; Riethmacher et al., 1997; Taveggia et al., 2005). Those experiments revealed that, later in development, mutant embryos exhibit a total loss of Schwann cells along their peripheral axons, whereas pre-migratory Schwann cells are present near to the dorsal root ganglia but fail to migrate along the axons. The impaired glial cell migration in these mutants indicates a direct function in migration. This interpretation is supported by the finding that mutations in the neuregulin–ErbB signalling system lead to changes in neural crest migration in mice (Britsch et al., 1998) and in directed Schwann cell migration in zebrafish (Lyons et al., 2005). However, an earlier function in fate specification and/or proliferation that could in turn be a prerequisite for migration cannot be completely ruled out.

To identify components that are potentially involved in the migration process of Schwann cells, we took advantage of the impairment of Schwann cell migration in ErbB3-deficient embryos. We conducted gene expression profiling of cultured embryonic DRG [12.5 days post coitum (dpc)], isolated from wild-type and ErbB3-deficient embryos. Aiming to identify genes expressed by migrating Schwann cells, we have been focusing on genes that are upregulated in wild-type compared with mutant cultures, because the wild-type cultures are enriched for migrating Schwann cell precursors as opposed to those that are immobile. Periostin was one such gene that showed a higher expression in wild-type compared with ErbB3-deficient cultures.

Periostin was originally isolated from osteoblasts and found to be preferentially expressed in the periosteum, hence the name (Horiuchi et al., 1999; Takeshita et al., 1993). Periostin contains an N-terminal secretory signal peptide, followed by a cysteine-rich domain, four internal homologous repeats and a C-terminal hydrophilic domain. The four internal repeats region, the fasciclin domains (Fas), exhibit homology to the axon guidance protein Fasciclin I that is involved in the development of nervous system in invertebrates (Zinn et al., 1988). Interestingly, in invertebrates, it is expressed in neurons and attached to the membrane through a glycosylphosphatidylinositol (GPI) anchor (Elkins et al., 1990), whereas in vertebrates, it is a secreted protein with a much broader expression pattern – including heart, periodontal ligament, fibroblasts and Schwann cells. In primary osteoblasts, periostin expression can be increased through stimulation with TGFβ-1 (Horiuchi et al., 1999). Several studies aiming to locate the protein product have revealed its presence in the ECM surrounding the expressing cells. In the ECM, periostin seems to be able to interact with several other ECM molecules, including collagen and tenascin-C (Kii et al., 2006,, 2010). Furthermore, it has been demonstrated that periostin promotes adhesion and potentiates cancer cell motility by binding to integrins α_νβ₃ and α_νβ₅ (Gillan et al., 2002), and that forced expression of periostin induces invasive activity through epithelial–mesenchymal transition (EMT) in HEK-293 cells (Yan and Shao, 2006). Several studies correlate periostin expression with tumour angiogenesis and metastasis in human cancer (Bao et al., 2004; Li et al., 2007; Malanchi et al., 2011; Puglisi et al., 2008; Sasaki et al., 2001a,,b,, 2003). Additionally, high pre-operative serum levels of periostin correlate with a poor prognosis in individuals with hepatocellular carcinoma after hepatectomy (Lv et al., 2013). Based on these observations, it has been suggested that periostin belongs to the group of matricellular proteins that are characterised as being modulators of cell-matrix interactions and cellular function (Kii et al., 2010; Norris et al., 2008). Interestingly, the existence of a non-secreted intracellular pool of periostin that is able to suppress the invasiveness of cancer cells, possibly by interacting with intracellular signalling molecules such as TAB1 and Tak1, has also been described (Isono et al., 2009; Yoshioka et al., 2002). In our current study, we show that periostin expression is downregulated in ErbB3-deficient DRG. In cultured rat Schwann cells, expression of periostin can be induced by neuregulin-1 (NRG1) and TGFβ-1. Additionally, we demonstrate that periostin deficiency leads to reduced migration of embryonic Schwann cell precursors and that recombinant periostin is able to promote cellular motility.

In order to identify genes that are potentially involved in Schwann cell precursor migration, we performed DNA microarray expression analysis with DRG explant cultures of ErbB3-deficient embryos. In mutant cultures, the migration of Schwann cell precursors along the axons did not occur, yet axonal outgrowth and neuronal survival in these cultures was not impaired (Fig. 1A). At 12.5 dpc, the DRG of ErbB3-deficient and wild-type embryos were isolated. Following 36 h of in vitro culture, the explant cultures were harvested, RNA was isolated and then used to hybridize Affymetrix MicroArray (probe sets MG-U74Av2, MG-U74Bv2 and MG-U74Cv2). After evaluation of the obtained data sets, 176 genes were found to exhibit a significantly higher expression in samples that had been derived from wild-type DRG cultures, in contrast, only nine genes showed increased expression in samples from ErbB3-deficient DRG. To further analyse the data, we grouped the differentially expressed genes into functional categories. The three major groups were transcription-related proteins, ECM proteins and nuclear proteins (Fig. 1B). However, when we focused our analysis on genes that exhibited the most significant increases in expression in wild-type DRG (top 20 of the list), 36% (7 out of 20) could be assigned to the group of the ECM proteins (Fig. 1C).

The most prominent differential expression was detected for the ECM molecule periostin. To verify and further support this finding, we next compared expression levels by using other approaches. A virtual northern blot analysis employing the total RNA of DRG explant cultures derived from wild-type and ErbB3-deficient cultures confirmed the lower expression levels of periostin in ErbB3-deficient cultures (Fig. 1D). Moreover, quantitative real-time PCR (qRT-PCR) analysis revealed a 10.5-fold higher expression of periostin in samples derived from wild type compared with those from mutant cultures (data not shown). As the most prominent difference between wild-type and ErbB3-deficient DRG cultures is the population of migrating Schwann cell precursors, these results point towards expression in these cells and also suggest a potential role of periostin for the migration process of these cells.

We therefore analysed the expression pattern of periostin during embryonic development by using in situ hybridisation of wild-type embryos at different stages. Prominent periostin expression is detectable in a variety of tissues and organs including bone, cartilage, heart and tongue as early as 10.5 dpc. At this stage of development, neural crest cells have already migrated out of the neural tube, and DRG formation has been initiated. Periostin expression was readily detectable in DRG at this early stage (Fig. 2A), and expression within DRG persists throughout embryonic development (Fig. 2B,C). At 12.5 dpc, strong expression of periostin could additionally be detected along axonal projections (Fig. 2D). To further characterise the expression pattern, we made use of a mouse mutant in which the β-galactosidase open reading frame had been fused in frame to the ATG start codon of the periostin gene, thereby replacing most of exon 1 and exon 2, resulting in a frameshift (the generation of these mice will be described elsewhere). Heterozygous animals were phenotypically normal but expressed the β-galactosidase gene within the cells that expressed periostin. Co-immunofluorescence analysis of heterozygous embryos at 12.5 dpc with antibodies specific for β-galactosidase and Sox10, a marker for glial cells in the PNS, confirmed the expression of periostin in the glial cell population (Fig. 3A–C). To examine the expression pattern of periostin in DRG explant cultures, we performed a co-immunofluorescence analysis that employed antibodies against Sox10 and periostin. Sox10-positive cells were found to exhibit robust periostin expression (Fig. 3D–F). Interestingly, even though the majority of the protein is secreted and can be detected in and purified from the supernatant, periostin was also detected within the cells.

Different isoforms of periostin varying within the 3′-translated region have been described (Horiuchi et al., 1999; Litvin et al., 2004). Interestingly, exon 17 has not been included in the official periostin cDNA or protein sequences; however, it is present in the genome sequence releases and thus should be included. Exon 17 has been described by Litvin et al. as only being present in the isoform periostin-like factor (PLF) (Litvin et al., 2004), whereas Kyutoku et al. correctly state it to be part of the periostin coding region (Kyutoku et al., 2011). To assess the presence of different periostin isoforms in glial cells of the PNS, we generated cDNA from DRG. Amplification by using gene-specific primers and subsequent sequence analysis revealed the expression of five different isoforms in embryonic glial cells, one of which (isoform IV) was previously unidentified (Fig. 4). We were able to detect a full-length transcript (containing all 23 exons) and several variants where either exon 21 (isoform II) or 17 (isoform III) were absent, or a combination of exons 17 and 20 (isoform IV) or exon 17, 20 and 21 (isoform V) were absent (Fig. 4).

As we identified periostin in our screen as being upregulated in wild-type DRGs compared with ErbB3-deficient DRGs, we next investigated whether the neuregulin–ErbB signalling system has any direct influence on periostin expression levels. For these experiments, we used the S16 Schwann cell line, which closely corresponds to Schwann cells at an early stage of myelination (Toda et al., 1994). These cells express the ErbB receptors and are responsive to NRG1 (data not shown). S16 cells were grown either on laminin or on poly-DL-ornithine (PDO), and left unstimulated or were stimulated with NRG1. In cells grown on PDO, non-secreted periostin localized in the centre of the cytoplasm, surrounding the nucleus (Fig. 5A), whereas in cells grown on laminin, periostin was observed mainly at the protrusions of the cells (Fig. 5C). Interestingly, upon stimulation with NRG1, expression levels appeared to be upregulated and the protein localized at cellular protrusions, whether cells were grown on PDO or laminin (Fig. 5B,D). Thus, the stimulation with NRG1 of S16 Schwann cells that had been grown on PDO resulted in a relocalisation of periostin, reminiscent of the localisation of the protein that was observed when cells were grown on a migration-promoting surface such as laminin.

To assess the impact of NRG1 stimulation on periostin expression levels, we performed a qRT-PCR analysis. S16 cells were cultured until they were semi-confluent and then stimulated with NRG1, TGFβ-1 or a mixture of both factors. TGFβ-1 is a known stimulator of periostin expression in a variety of cell types (Horiuchi et al., 1999; Tai et al., 2005). Stimulation with NRG1 resulted in an almost fivefold increase of periostin expression, whereas stimulation with TGFβ-1 resulted in a sixfold increase compared with that of unstimulated cells (Fig. 5E). Interestingly, we observed an 11-fold increase of periostin transcription levels upon combined stimulation with both NRG1 and TGFβ-1 (Fig. 5E), suggesting that the ErbB3 ligand NRG1 and TGFβ-1 cooperate to upregulate periostin expression in S16 Schwann cells.

Periostin is known to interact with collagen I (Norris et al., 2007), as well as other ECM molecules such as fibronectin, tenascin-C and collagen IV (Kii et al., 2010; Takayama et al., 2006). The relocalisation of periostin in S16 Schwann cells that had been cultured on laminin prompted us to investigate whether periostin and laminin can interact directly. In order to test such a potential interaction, a His₁₀- and c-Myc double-tagged recombinant periostin construct (isoform I, full length) was expressed in HEK-293 cells and purified to homogeneity from cellular supernatants by using Ni²⁺-affinity chromatography, as assessed by using SDS-PAGE and western blotting (Fig. 6A). Solid-phase binding assays demonstrated a robust interaction of recombinant periostin with laminin and a less pronounced interaction with collagen I and fibronectin (Fig. 6B). This interaction was dependent on the presence of Ca²⁺ because without added Ca²⁺, no specific interaction could be detected (data not shown). We were unable to observe any specific interaction with collagen IV (Fig. 6B). To further corroborate this interaction between periostin and laminin, we performed co-immunoprecipitation experiments with periostin-expressing cells and exogenously added laminin. In these experiments we were able to co-precipitate periostin by using laminin-specific antibodies (Fig. 6C). Having demonstrated the potential of the two proteins to physically interact, we determined their location in developing nerves. We were able to co-localise both proteins in developing nerves by employing sections of DRG and spinal nerves from 12.5-dpc embryos (Fig. 7A). Co-localization and the ability to physically interact indicate that there is a functional interplay between the two molecules in vivo.

Because we discovered periostin in an expression analysis that aimed to identify genes which are potentially involved in Schwann cell migration, we next examined the effect of periostin expression on Schwann cell migration. To this end, we used embryonic DRG cultures from periostin-deficient and wild-type mice and compared Schwann cell migration. Several groups, including ours, have independently generated periostin-deficient animals (Kii et al., 2006; Oka et al., 2007; Rios et al., 2005) (A. Brodarac, Impaired tooth development in periostin deficient mice, PhD thesis, University of Hamburg, 2006). These mice do not show any overt phenotype during embryonic development, and the gross development of the PNS appears normal in the absence of periostin expression. Interestingly, when we analysed DRG cultures, we observed severely reduced migration of Sox10-positive Schwann cells in periostin-deficient cultures compared with that in wild-type cultures, whereas the axon outgrowth was not impaired (Fig. 7B). To characterise this in more detail, we monitored migration in several independent cultures from wild type, and heterozygous and homozygous periostin mutants. Schwann cell migration in cultures from heterozygous mutants was normal, whereas in cultures from periostin-deficient embryos, we observed a 50% reduction in migration (Fig. 8A). Attempts to rescue mutant cultures by either prior coating of the dishes with periostin or by adding it to the medium were not successful (data not shown). We next analysed whether periostin deficiency has any impact on Schwann cell proliferation or apoptosis in these cultures as this would also affect the outcome of our migration analysis. However, proliferation and apoptosis rates in wild-type and periostin-deficient DRG cultures showed no significant differences (Fig. 8B). The finding that neither proliferation nor apoptosis were affected in these cultures clearly suggests that periostin has a direct impact on the migratory capacity of Schwann cell precursors.

Because periostin has been implicated in migration processes in a variety of cell types and different contexts (i.e. cancer, cutaneous wound healing), we wanted to support our observations of DRG cultures by employing another experimental setup. Therefore, periostin- or mock-transfected HEK-293 cells were analysed in a Boyden chamber assay using conditioned medium from these cells or control medium that did not contain any periostin (Fig. 8C). In this experimental setup, conditioned medium containing periostin strongly stimulated the migration of HEK-293 cells. Interestingly, migration was highest when cells were exposed to periostin but did not express it endogenously (Fig. 8D), and the observed migration in this experimental setup was regarded as 100%. Periostin-expressing HEK-293 cells exhibited a slightly lower migration rate (80%). When no periostin was present in the medium, mock-transfected (EGFP) control cells were stationary, whereas periostin-expressing cells were still able to migrate (40%) (Fig. 8D). To assess whether this higher migration rate was accompanied by higher rates of proliferation of these cells, which might influence the migration behaviour, a BrdU labelling experiment was performed. Again, we were unable to detect any changes in the proliferation rates (data not shown). These findings and the results obtained using DRG cultures clearly indicate that periostin can increase the migratory capacity of cells.

Considerable progress has been made in the identification of molecules involved in neural crest cell development and in understanding their impact on processes – including proliferation, differentiation and de-differentiation (Crane and Trainor, 2006; Dupin et al., 2007). However, there is only scarce knowledge of the molecules involved in Schwann cell migration. One important player in neural crest and Schwann cell development is the neuregulin–ErbB signalling system (Newbern and Birchmeier, 2010). DRG explant cultures of ErbB3-deficient embryos reflect the migration impairment of the Schwann cell precursor (Riethmacher et al., 1997) and therefore represent an ideal model for expression analyses of the genes that are involved in migration when compared with those of wild-type cultures. Our microarray analyses revealed a prominent differential expression of ECM molecules, which is independently supported by microarray data that has been collected using isolated neural crest stem cells and Schwann cell precursors (Buchstaller et al., 2004). The highly differential expression of ECM molecules coincides with axonal growth and the on-going process of Schwann cell precursor migration along the axons, suggesting a potential involvement of such ECM components in these processes. Furthermore, ECM molecules and their receptors have been implicated in Schwann cell development and function (Chernousov et al., 2008), as well as migration (Chernousov et al., 2001; Milner et al., 1997). The ECM protein periostin, which exhibited the most significant differential expression in our microarray analysis, is expressed in a variety of tissues during development, including DRG and Schwann cell precursors, as shown by us and others (Kruzynska-Frejtag et al., 2001; Lindsley et al., 2005).

So far, several isoforms of periostin with varying domain compositions in the C-terminus, generated through alternative splicing, have been reported (Horiuchi et al., 1999; Litvin et al., 2004; Ozdemir et al., 2014). Using RNA that was isolated from DRG cultures, we could identify five different isoforms, including an isoform that has not been described previously. The exact function and physiological relevance of the different isoforms remain elusive. However, because periostin has been implicated in a variety of different, partially even conflicting, functions (Bao et al., 2004; Kim et al., 2005; Shao et al., 2004), alternative splicing might offer an explanation for these findings. Similarly, the different isoforms present in the precursor and migrating Schwann cells might have varying impacts on cellular behaviour – e.g. motility. Further experiments are required to investigate how different compositions of the C-terminus influence the ability to interact with other ECM molecules or receptors. In this context, it will also be interesting to determine the expression levels of the different splice variants.

In order to identify the biological functions of periostin, several labs have independently generated periostin-deficient mice. Surprisingly, despite the robust expression during embryonic development, mutants show apparently normal development and are viable, indicating that there is a backup mechanism for periostin function in its absence. The most apparent phenotype is impairment of the eruption of incisors, which first presents itself shortly after weaning (A. Brodarac, Impaired tooth development in periostin deficient mice, PhD thesis, University of Hamburg, 2006) and causes secondary defects such as malnutrition and dwarfism (Kii et al., 2006; Rios et al., 2005). More recent studies involving different challenging strategies have revealed subtle defects in skin, tendons, heart valves, and changes in regeneration dynamics following myocardial infarction (Norris et al., 2008,, 2007; Oka et al., 2007; Shimazaki and Kudo, 2008; Snider et al., 2008). Interestingly, subtle phenotypes during normal development and impairment of tissue remodelling and repair are hallmarks of mice carrying targeted mutations in other matricellular proteins (Bornstein and Sage, 2002).

Although Schwann cell development is normal in periostin-deficient animals, in this paper we were able to show that cultured DRG of periostin-deficient animals could not sustain Schwann cell migration at levels comparable to those of wild-type DRG. During development, there is most likely to be an alternative mechanism operating in the absence of periostin that is not present in our DRG culture. This experimental setup shows that periostin can be involved in cell migration. Our Boyden chamber assay further showed that the concentration and localisation of periostin is also important for its migratory function. We therefore conclude that periostin can influence the migratory behaviour of Schwann cells and also other cells, but in the absence of periostin, other proteins – i.e. murine TGFBI – can compensate for its function during embryonic development. Interestingly, our preliminary data indicate that the absence of periostin leads to delayed and reduced regeneration after peripheral nerve injury (E.S.-R. and D.R., unpublished data). We show here that paracrine expression alone enhances migration compared with simultaneous paracrine and autocrine expression (Fig. 8D). However, if no paracrine expression is present, cells expressing periostin in an autocrine fashion are more likely to migrate than cells that lack periostin. The migratory function is also interesting with respect to expression of periostin in tumour development and metastasis. The ability of tumour cells to migrate and invade other tissues is an important prerequisite for the development of metastasis (Hanahan and Weinberg, 2000). Papers with conflicting results have been published showing that upregulation as well as downregulation of periostin can lead to a more severe tumour phenotype and to enhanced metastasis (Kim et al., 2005; Malanchi et al., 2011; Morra and Moch, 2011). This leads us to propose a model in which the periostin expression levels of the tumour itself, the periostin expression levels of the surrounding tissue and the periostin levels in the serum cooperatively influence the metastatic potential of a tumour.

We showed that the expression levels of periostin are not only influenced by TGFβ-1, as previously shown (Horiuchi et al., 1999), but also by NRG1. Additionally, the interaction of the two ligands results in a higher expression level than either ligand on its own (Fig. 5E). Experiments using different conditions have shown that the neuregulin signalling and TGFβ-1 can cooperate to influence the invasiveness of cancer cells (He et al., 2011), and also in peripheral nerve regeneration (Ribeiro-Resende et al., 2009). Both of these studies suggest the existence of cooperative, synergistic effects for TGFβ-1 and neuregulin signalling. From our results, it is very possible that these effects are mediated through periostin. In recent experiments, it has been shown that Wnt5a can also upregulate periostin expression but seems to exert this upregulation solely through augmentation of TGFβ-1 levels (Hasegawa et al., 2015). The connection between neuregulins and the expression level of periostin, as well as our preliminary data on nerve regeneration, are exciting as a recent paper shows that re-expression of neuregulins after injury to Schwann cells is highly important for regeneration and remyelination (Stassart et al., 2013).

Our results show that periostin exists in a variety of isoforms and that its expression levels are influenced cooperatively by different ligands of receptor kinases. Furthermore, periostin can interact with different ECM molecules, and these interactions are most likely to affect the localisation of periostin. Therefore, this study suggests that periostin is a protein that is involved in cell motility, the exact function of which can be fine-tuned through expression levels, interaction partners and also possibly different isoforms.

Generation, maintenance and genotyping of mice heterozygous for a deletion in ErbB3 has been described previously (Riethmacher et al., 1997), whereas generation and genotyping of periostin-deficient mice will be described elsewhere (A. Brodarac, Impaired tooth development in periostin deficient mice, PhD thesis, University of Hamburg, 2006). All animals in Hamburg were cared for according to the ‘Guide for the Care and Use of Laboratory Animals’ prepared by the National Academy of Sciences and published by the National Institutes of Health (NIH; publication 86-23 revised 1985), and experiments involving animals bred in Southampton were performed under a Home Office Licence and in accordance with the Animals (Scientific Procedures) Act, 1986, UK.

Heterozygous mutant animals were mated, and pregnant females euthanized at 12.5 dpc. The embryonic DRG were dissected and grown in a defined medium that has been used previously for embryonic mouse neurons (Davies, 1998). Genotyping was performed using standard techniques with the following primers – ErbB3 forward-1 5′-GGGTGTCTGAGTCTTTGAAGCTGGAG-3′, reverse-1 5′-ACCTGTAATCTCCCGACTGTCCTGAA-3′, reverse-2 5′-CGAATTCGCCAATGACAAGACGCTGG-3′; Periostin forward-1 5′-TGTGGTGGCAGAGACAGAAG-3′, reverse-1 5′-CTCCCCTGTGGTGCATTTAC-3′, reverse-2 5′-GTTGCACCACAGATGAAACG-3′.

S16 cells were maintained in Dulbecco's modified Eagle's medium that had been supplemented with heat-inactivated 10% fetal calf serum. All cell culture reagents were purchased from Invitrogen unless otherwise specified. TGFβ-1 from porcine platelets as well as collagen I, collagen IV, collagen V and fibronectin were purchased from Sigma. NRG1 was purchased from Neomarkers, and Laminin-1 from Roche Applied Science. The anti-periostin antibody was an affinity purified polyclonal antibody from rabbit directed against the first 347 amino acids of the protein without the leader sequence. The monoclonal Sox10 antibody was provided by Michael Wegner (University of Erlangen, Germany).

DRG explant cultures were cultivated for 36 h, and total RNA was extracted using Trizol (Invitrogen). Reverse transcription was performed with Superscript II (Invitrogen) and oligo(dT) primer. Transcribed cDNA was amplified using periostin-specific primers F15 and R23 (forward primer 5′-CATCCACGTCGTGGACAAACTCCTC-3′, reverse primer 5′-TTGGCTTCTGTTGGTTGTCA-3′) and purified by using agarose gel electrophoresis. The amplicons were isolated and then subcloned into pGemT easy vectors (Promega); the constructs were subsequently sequenced.

DRG explant cultures were cultivated for 36 h, and total RNA was isolated using Trizol (Invitrogen), according to the manufacturer's instructions. RNA was reverse transcribed using a T7-promoter-tagged polyT primer [5′-GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGG-(dT)₂₄-3′], and antisense RNA was produced using T7 RNA polymerase. Antisense RNA was examined by using gel electrophoresis. Equivalent amounts of RNA from five age-matched embryos of the same genotype were pooled to generate a microarray probe. Probes were biotin-labelled in a second round of T7 amplification, as previously described (Luo et al., 1999). Profiles of ErbB3-deficient mice were compared with the profiles of wild-type littermates of the same genetic background. Microarray profiling was performed in triplicate from independent preparations of embryonic RNA using Affymetrix MG-U74Av2, MG-U74Bv2 and MG-U74Cv2 GeneChips (36,000 sequences). Data were normalized, and comparative analyses were performed using the statistical algorithm of Affymetrix Microarray Suite (MAS) 5.0 software following the manufacturer's guidelines. Ranking and filtering of genes was performed using Microsoft Excel.

Total RNA from DRG cultures was transcribed with oligo-dT primer by reverse transcriptase and amplified using the SMART™ technology (Clontech) following the manufacturer's guidelines. Thereafter, Southern blots were performed using the first 582 nucleotides of the periostin cDNA or full-length GAPDH labelled with [³²P]-CTP.

Embryos were isolated at 10–16.5 dpc from staged pregnancies and fixed in 4% paraformaldehyde (PFA; in PBS) for 4–16 h. Samples were then equilibrated in 30% sucrose, embedded in OCT (Miles), frozen and sectioned using a cryostat. In situ hybridisation was performed as described previously (Sonnenberg et al., 1991). The first 582 nucleotides of periostin were used as periostin-specific probe.

For immunohistochemistry, embryos and tissues were fixed in 4% PFA in PBS at 4°C overnight, cryoprotected in 20% sucrose in PBS for 12 h at 4°C, embedded in OCT compound (Miles) and sectioned using a cryostat (thickness of 10 μm). Sections were rinsed three times with PBS, blocked for 30 min with PBS containing 0.1% Triton X-100 and 0.2% bovine serum albumin (BSA) and incubated overnight with primary antibodies at 4°C. After being washed three times with PBS (each wash lasting 5 min), the sections were incubated with the appropriate secondary antibodies conjugated to Alexa 466 (Molecular Probes) or Cy3 (Jackson Laboratories; Chemicon) for 1 h and then washed. Primary mouse immunoglobulin G1 antibodies were detected with Zenon technology (Molecular Probes). After the sections had been rinsed with PBS and nuclei had been counterstained with 4′,6′-diamidino-2-phenylindole (DAPI; 0.001 mg/ml of PBS), sections were examined with a Zeiss Axioplan 2 microscope, and images were taken with a Zeiss AxioCam digital camera (Brockschnieder et al., 2004). In the case of double-labelling experiments, slides were fixed in 4% PFA (in PBS) for 30 min after the final wash. Then the second primary antibody was added, and the procedure described above was followed through again. The following antibodies were used in this study: anti-Sox10 (monoclonal antibody from Michael Wegner's Lab); anti-periostin (Abcam); anti-β-galactosidase (Cappel); anti-laminin (Lab Vision Corporation).

Total RNA from DRG cultures was transcribed with oligo-dT primer 5′-GACCACGCGTATCGATGTCGAC-(dT)₁₆-3′ by reverse transcriptase. Periostin cDNA was amplified by using gene specific primers 5′-GGCTGAAGATGGTTCCTCTC-3′ and a primer binding to the overhang of the oligo-dT primer (5′-GACCAGGCGTATCGATGTCGAC-3′). For recombinant protein expression, the cDNA was amplified with primers 5′-GATCGCTAGCATGGTTCCTCTCCTG-3′ and 5′-GATCGCGGCCGCCTGAGAACGGCCTTC-3′ providing NheI and NotI restriction sites, respectively. The PCR product was purified by using agarose gel electrophoresis and inserted into pcDNA5/FRT (Invitrogen) providing a 10×His and a c-Myc tag at the 3′-end. In order to generate stable lines, Flp-In HEK-293 cells (Invitrogen) were transfected with the expression vector using jet PEI (Q BIO gene) following the manufacturer's guidelines. This cell line enables the rapid generation of stable expression from a defined locus. The protein was purified from the supernatant by using immobilized metal ion affinity chromatography using Ni-NTA-matrix (Qiagen) following the manufacturer's guidelines and was then dialyzed against PBS.

Total RNA for qRT-PCR analyses was extracted from S16 cells that had been cultured for 24 h on laminin-coated plates in Opti-Mem medium and subsequently stimulated with 10 ng/ml NRG1 or 1 ng/ml TGFβ-1 for 24 h. After treatment with RNase-free DNase for 30 min at 37°C to remove genomic DNA, the quantity and quality of the RNA was determined by using the RNA 6000 Nano LabChip Kit (Agilent). The cDNA was synthesised from 3 µg of total RNA using 1 µg random primers with Superscript II (Invitrogen) and spiked with the amylose extender1 (ae I) gene from wheat. qRT-PCR was performed using the ABI PRISM 7700 (ABI). The efficiency of cDNA synthesis was evaluated by amplifying the ae I cDNA (ae I forward 5′-ACGCGATGCCCTCCATAG-3′, ae I reverse 5′-AATTCGAGACCGGATCCT-3′). All reactions contained 15 ng cDNA, 300 nM periostin forward primer 5′-TTCCGAGAGATCATCCAACC-3′, 50 nM periostin reverse primer 5′-AGGCTGAGGAAGATGCTGAA-3′ (product size 82 bp), 18S 60 nM forward primer 5′-CGGCTACCACATCCAAGAA-3′, 18S 60 nM reverse primer 5′-GCTGGAATTACCGCGGCT-3′ (product size 182 bp) and QPCR Mastermix for SYBR^® Green I (Eurogentec) in a final volume of 20 µl. Primer concentrations and conditions were optimised in preliminary experiments. The thermal profile started with an initial step of 50°C for 2 min, followed by 10 min 95°C to activate the polymerase and subsequently 40 cycles each comprising 15 s 95°C and 1 min 65°C. In each analysis, all samples were run in parallel. The control gene selected to normalize each sample for RNA content was 18S ribosomal RNA. The fold difference was analysed by using the standard deviation calculations using the relative standard curve method. The paired sample t-test was used to evaluate the significance of differences in expression in paired groups.

Cells were washed twice with ice-cold PBS and lysed with a buffer containing 50 mM Tris-HCl (pH 7.4), 150 NaCl, 1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride and a protease inhibitor cocktail (Complete mini, Roche). Protein extracts or purified protein were separated by using SDS-PAGE, transferred to PVDF membrane (Millipore), blocked with 5% milk powder in PBS and incubated with the indicated antibodies overnight. Immunoreactive bands were visualized by using chemiluminescence (Millipore). For immunoprecipitation, the HEK-293 cells were transfected with a plasmid expressing periostin or β-galactosidase. Laminin was added to the cell culture medium and, at 24 h after transfection, the supernatant was collected and precleared with protein-G/A–Sepharose (Santa Cruz Biotechnology) for 2 h at 4°C. The pre-cleared supernatant was incubated with anti-laminin antibody (1:100) (Clone A5, NeoMarkers) overnight. The immunocomplex was extensively washed, the samples (30 µg) were subjected to immunoblotting, and periostin was detected with anti-c-Myc antibody (clone 9E10, Sigma).

96-well plates (Nunc) were coated overnight with 100 µg/ml of collagen I, collagen IV, collagen V, fibronectin, bovine serum albumin fraction V (Sigma) and laminin (Roche). After blocking with 5% milk powder in 100 mM Tris-HCl pH 7.5, 150 mM NaCl, 200 ng/ml of recombinantly expressed and purified periostin was incubated for 5 h in 100 mM Tris-HCl pH 7.5, 150 mM NaCl or 100 mM Tris-HCl pH 7.5, 150 mM NaCl, 2 mM CaCl₂. Periostin was detected using an anti-c-Myc horseradish peroxidase (HRP) conjugate. The immunoreactivity was visualized by using ABTS as a substrate and measured at 405 nm.

For the migration assay, HEK-293 cells were transfected for 36 h with plasmid encoding periostin or β-galactosidase. The conditioned medium was collected and filtered. Additionally, HEK-293 cells were transfected for 24 h with plasmids encoding periostin or EGFP. 1×10⁵ of these cells were then transferred onto 8-µm pore polycarbonate filter (Nunc). The lower chamber of Transwell inserts was loaded with the conditioned medium. After 24 h of incubation, the top cells on the Transwell membrane were removed using Q-tips. The cells trapped by the membrane were fixed and stained with Diff-Quick (Düdingen, Switzerland). Cell numbers within the entire membrane area were determined.

For the analyses of Schwann cell migration, DRG explant cultures were cultured for 72 h and were then analysed by immunofluorescence using an antibody against Sox10 and the Neurofilament 200 antibody (Sigma) to visualise Schwann cells and axons, respectively. All migrating Sox10-positive cells per DRG were counted.

The DRG explant cultures were cultured for 48 h and incubated for 1 h with 10 µM BrdU, washed with PBS and fixed for 20 min with Carnoy's fixative (methanol and acetic acid, 3:1) at −20°C. Subsequently, the DRG cultures were incubated for 15 min with 2.4 M HCl at 37°C. To identify incorporated BrdU, thus proliferating cells, an antibody against BrdU (Roche) was used.

The DRG were cultured for 48 h and fixed with 1% PFA (in PBS) for 10 min at room temperature. To identify apoptotic cells, the ApopTag Apoptosis detection kit (Q-BIOgene) was used following the manufacturer's guidelines.

In all experiments, appropriate statistical analysis (ANOVA, t-test) was performed using GraphPad Prism 4 or MiniTab (Release 15) software. Data are shown as mean values with standard errors of the mean.

We thank Andreja Brodarac (Center for Molecular Neurobiology, University of Hamburg, Hamburg, Germany) for her involvement in the generation of the periostin-deficient mouse strain. We thank Michael Wegner for the Sox10 antibody.