Periostin and TGF-β-induced Protein: Two Peas in a Pod?

Abstract

Periostin (PN) and TGF-β-induced protein (βig-h3) are paralogs that contain a single emilin and four fasciclin-1 modules and are secreted from cells. PN receives attention because of its up-regulation in cancer and degenerative and allergic diseases. βig-h3 is highly enriched in cornea and best known for harboring mutations in humans associated with corneal dystrophies. Both proteins are expressed widely, and many functions, some over-lapping, have been attributed to PN and βig-h3 based on biochemical, cell culture, and whole animal experiments. We attempt to organize this knowledge so as to facilitate research on these interesting and incompletely understood proteins. We focus particularly on whether PN and βig-h3 are modified by vitamin K-dependent γ-glutamyl carboxylation, a question of considerable importance given the profound effects of γ-carboxylation on structure and function of other proteins. We consider the roles of PN and βig-h3 in formation of extracellular matrix and as ligands for integrin receptors. We attempt to reconcile the contradictory results that have arisen concerning the role of PN, which has emerged as a marker of T_H2 immunity, in murine models of allergic asthma. Finally, when possible we compare and contrast the structures and functions of the two proteins.

Introduction

Periostin (PN, human gene POSTN) and TGF-β-induced protein (βig-h3 or TGFBIp, human gene TGFBI) are paralogs that contain a single emilin (EMI) and four fasciclin-1 (FAS1) modules (Fig. 1A). The literature on PN and βig-h3 is extensive and expanding, with eight listings for the two proteins together, 680 listings for PN, and 313 listings on βig-h3 in PubMed (http://www.ncbi.nlm.nih.gov/pubmed) at the time this review was being prepared. PN receives attention because of its up-regulation in cancer and degenerative and allergic diseases. βig-h3 is highly enriched in cornea and best known for harboring mutations in humans that lead to protein precipitation and corneal dystrophies. Both proteins are expressed widely, and many functions, some over-lapping, have been attributed to PN and βig-h3 based on biochemical, cell culture, and whole animal experiments. Our goal is to organize this knowledge so as to facilitate future research on these interesting and incompletely understood proteins. We focus particularly on question of whether PN and βig-h3 are modified by vitamin K-dependent γ-carboxylation, a proverbial “800-pound gorilla” lurking in the field given the profound effects of γ-carboxylation on structure and function of other proteins. We consider the roles of PN and βig-h3 in formation of extracellular matrix (ECM) and as ligands for integrin receptors. Rather than duplicate recent reviews relating PN and βig-h3 to various disease conditions (Conway et al., 2014, Liu et al., 2014, Lakshminarayanan et al., 2014), we concentrate on reconciling the contradictory results that have arisen concerning the role of PN in murine models of allergic asthma, an interest of our group (Johansson et al., 2013). Finally, when possible we compare and contrast the structures and functions of the two proteins.

Fig. 1.

Modular structures of of MFAS, PN, and βig-h3 and identity between mouse and human proteins. (A) The five modules are in blocks, within which are the relative locations of cysteines (C) and asparagines with potential to be glycosylated (N) and a schematic of the proposed pattern of disulfides in the EMI module. Thick lines represent tail sequences present in all variants; thin lines represent sequences that may or may not be present depending on splicing. Exons encoding sequences in the C-terminal tail of human PN are numbered. The site of the Arg-Gly-Asp (RGD) sequence in βig-h3 is indicated. (B) The total numbers of differences and non-conservative differences (in parentheses) between mouse and human proteins are given for each region as a fraction of number of residues that differed divided by total residues. For the PN C-terminus, the longest splice variant has been scored. Only the online version of this figure is in color.

Primary sequences and phylogenetic origins

The EMI module is named for emilin (elastin microfibril interface-located protein), an ECM protein of the C1q/multimerin family that was identified associated with elastin and microfibrils in chickens (Colombatti et al., 2000, Colombatti et al., 2011). The FAS1 module was first described in Drosophila fasciclin-1, which, like PN and βig-h3, has four tandem FAS1 modules (Zinn et al., 1988). Fasciclin-1 is one of three numbered fasciclins that play roles in development of the fly nervous system; the other two fasciclins lack FAS1 modules (Clout et al., 2003). Drosophila fasciclin-1 is distinct from PN and βig-h3 in lacking an EMI module and being tethered to cells by a lipid anchor rather than being secreted (Clout et al., 2003, Zinn et al., 1988). There is, however, a second FAS1-containing Drosophila protein, midline fasciclin (MFAS, gene Mfas) (Hu et al., 1998), which is annotated (http://www.uniprot.org/uniprot/O61457) (http://flybase.org/reports/FBgn0260745.html) as being homologous to PN and βig-h3, comprising a single EMI module and four FAS1 modules, and lacking an obvious site for a lipid tether (Fig. 1A). MFAS is expressed in the tracheal placode and tracheal tubes of Drosophila as well as in the nervous system (Hu et al., 1998).

Sequences of the EMI and FAS1 modules of human PN and βig-h3 and Drosophila MFAS are shown in Fig. 2. Alignment was with T-Coffee (http://www.ebi.ac.uk/Tools/msa/tcoffee/) and adjusted slightly based on features emphasized in a review of EMI modules (Colombatti et al., 2011) or the description of the structure of FAS1-3 and FAS1-4 modules of fasciclin-1 (Clout et al., 2003). The three proteins vary most conspicuously in their N- and C-terminal tails (Fig. 1A). The C-terminal tail of PN is subject to extensive alternative splicing, an attribute that is evolving rapidly in vertebrates in terms of exon count, length, and splicing pattern, whereas the tail of βig-h3 is short, not alternatively spliced, and evolving slowly (Hoersch and Andrade-Navarro, 2010). Examination of the two postn genes in zebrafish defined a sequential 13-amino acid repeat unit in the differentially expressed exons; this unit is well conserved in teleost fish and more obscure in higher vertebrates (Hoersch and Andrade-Navarro, 2010). FlyBase lists 16 different transcripts for MFAS (the UniProt entry for the longest 905-residue conceptual protein is given above). Remarkably, the alternatively spliced exons encode for an N- rather than a C-terminal tail of variable length (Fig. 1).

Fig. 2.

Sequences of EMI and FAS1 modules of MFAS (M), PN (P), and βig-h3 (β) aligned with T-Coffee as described in the text. Numbering of MFAS is based on the longest splice variant. Invariant residues and a nearly invariant phenylalanine are in red and blue, respectively. Blocks of highly conserved sequences surrounding invariant FAS1 module residues are in orange. Cysteine residues are bolded and in blue. Secondary structural elements are positioned based on the structure of FAS1-4 of βig-h3. Sequences proposed to constitute recognitions sites for γ-glutamyl carboxylase are underlined in green for three of the FAS1 modules. Arrows point to the locations of Tyr-His (YH) and Asp-Ile (DI) sequences implicated as being important for integrin binding. Only the online version of this figure is in color.

An extensive analysis of the evolution of PN and βig-h3 in vertebrates has been carried out based on genomic and transcriptomic sequence data (Hoersch and Andrade-Navarro, 2010). Two postn genes but only one tgfbi gene were identified in teleost fish, indicating that either tgfbi was not part of the whole genome duplication event that took place in the common ancestor of the fish or one gene was lost prior to the radiation of teleost species (Hoersch and Andrade-Navarro, 2010). It was speculated that postn and tgfbi arose from a common precursor gene during one of the prior whole genome duplication events that are thought to have occurred at the base of the jawed vertebrates or of all vertebrates (Hoersch and Andrade-Navarro, 2010). Such analysis should be worth re-doing taking into account Drosophila Mfas along with additional relevant whole-genome sequences that have been deposited since the Hoersch/Andrade-Navarro paper was published in 2010.

Comparisons of the human and murine amino acid sequences reveal numerous differences, some conservative, some non-conservative, throughout both PN and βig-h3 (Fig. 1B). The percent identities, 90.9% for PN and 91.3% for βig-h3, are close to the median of 89% identity between many pairs of mouse and human orthologs (Makalowski and Boguski, 1998) and the 92.4% identity found when we compared human and murine fibronectin.

Secondary, tertiary, and quarternary structure

The N-terminal tails of PN and βig-h3 comprise 19 and 22 residues, respectively, and have unknown structure. The EMI modules presumably adopt a global fold that is able to accommodate the extra residues found in MFAS (Fig. 2). More is known about FAS1 modules. The crystal structure was solved for a tandem construct comprising FAS1-3 and FAS1-4 of fasciclin-1 (PDB 1O70 (Clout et al., 2003)) and used in a follow-up paper to propose a model of FAS1-4 of βig-h3 (Clout and Hohenester, 2003). Structures of FAS1-4 of βig-h3 were subsequently determined by crystallography (PDB 2VXP, not published), and NMR (PDB 2LTB (Underhaug et al., 2013)). In addition, there are NMR-determined structures of single FAS1 modules MPB7 of Mycobacterium tuberculosis (Carr et al., 2003) and Fdp of Rhodobacter sphaeroides (Moody and Williamson, 2013). The global fold consists of two sets of three ɑ-helices and a seven-stranded β-wedge (Figs. 2 and 3). The fold appears to be stabilized by a core that contains consensus apolar residues and also three conserved polar residues that participate in network of hydrogen-bonding interactions (Clout et al., 2003). The FAS1-3 and FAS1-4 of fasciclin-1 were found to be arranged in a linear fashion and interact through a polar interface that immobilizes the short segment linking the modules (Clout et al., 2003). Although analyses of the overall shape of full-length βig-h3 by small angle X-ray scattering were complicated by the tendency of the full-length protein to multimerize (Basaiawmoit et al., 2011), results were compatible with a model in which the four tandem FAS1 modules were in a “beads-on-a-string” arrangement, looking rather like a wooly bear caterpillar (Basaiawmoit et al., 2011). It is not known whether the substantial interface between FAS1-3 and FAS1-4 in the crystal structure of fasciclin-1 (Clout et al., 2003) is a good model for interfaces between FAS1 modules of PN or βig-h3. Such a question is important in considerations of structure-function because substantial interfaces would limit the surface available to interact with other molecules, dampen inter-module flexibility, and allow the structure of one module to impact the structures of neighboring modules.

Fig. 3.

Structure of FAS1-4 of βig-h3 as determined by NMR. Cartoon and surface representations of βig-h3 FAS1-4 NMR structure (PDB 2LTB (Underhaug et al., 2013)). Highlighted in orange are the two most conserved regions among the FAS1 modules of human βig-h3 and PN and Drosophila fasciclin 1 and MFAS. Residues that are conserved in all 16 modules, Thr538, Pro542 and His572, are in hot pink. Phe540, which in some modules is replaced by another large hydrophobic residue, is in blue. Sequences proposed to constitute recognitions sites for γ-glutamyl carboxylase are underlined in green. The two views are related by a 90° rotation around a vertical axis and have been chosen to illustrate how the four conserved residues are buried in the core of the FAS1 domain. Only the online version of this figure is in color.

The C-terminal tail of βig-h3 continues 51 residues beyond FAS1-4 and includes the sequence Arg-Gly-Asp (RGD) that has the potential to be recognized by a number of integrin cell adhesion receptors (Fig. 1A). When a construct comprising FAS1-4 plus the first 25 residues of the tail, including the RGD sequence, was examined by NMR, the tail residues were unstructured (Underhaug et al., 2013). The C-terminal tail of PN ranges from 93 to 208 residues depending on splicing. Modeling of secondary structure suggests that much of the PN tail has intrinsic disorder or a propensity to form β-strands that are broken at intervals by bends that contain prolines (Hoersch and Andrade-Navarro, 2010). To our knowledge, this hypothesis has not been tested by spectroscopic techniques sensitive to secondary structure. Binding of PN to tenascin-C, fibronectin, type V collagen, and itself was enhanced when the C-terminal tail was not present (Takayama et al., 2006, Kii et al., 2010), raising the possibility that the tail folds back to interact with the FAS1 and EMI modules and thus modulate the interactions of these modules with other proteins.

Glycosylation and disulfide bonding

PN and βig-h3 have signal sequences of 21 and 23 residues, respectively. The proteins have no obvious membrane tethers and can be presumed to traffic through the endoplasmic reticulum, Golgi apparatus, and secretory vesicles before being secreted into the pericellular space. None of the 40 asparagines of human βig-h3 is in the NxT/S consensus sequence for N-glycosylation, and only one of the 44 in human PN fits this consensus, Asn599 in FAS1-4. This residue has been determined to be N-glycosylated (Liu et al., 2005, Chen et al., 2009). In contrast, eight of the 49 asparagines in MFAS fit the consensus (Fig. 1). The paucity of N-glycosylation sites suggests that folding of PN and βig-h3 in the secretory pathway presents few problems to cells inasmuch as the proteins have evolved so as to not be vetted by glycosylation-based quality control, i.e., monitoring of proper protein folding utilizing calnexin and UDP-glucose glycoprotein glucosyl transferase (Dejgaard et al., 2004). Lack of carbohydrate “handles” to expedite secretory trafficking may explain accumulation of PN and βig-h3 in the Golgi apparatus (Kim et al., 2009a, Maruhashi et al., 2010). Failure to employ glycosylation-based monitoring may also explain why βig-h3 harboring mutations is secreted by corneal cells and deposited to cause corneal dystrophy (Runager et al., 2011) rather than degraded as a misfolded protein in the endoplasmic reticulum.

We are not aware of experiments to determine the oxidation states or pairings of the cysteines/cystines in PN or βig-h3. As depicted in Fig. 1, the six cysteines in the EMI modules of PN and βig-h3 are annotated in UniProt as forming disulfides in the pattern 1–5, 2–3, 4–6 based on “Prosite-ProRule:PRU00384” ((http://www.uniprot.org/uniprot/Q15063, (http://www.uniprot.org/uniprot/Q15582). We cannot find evidence to support the proposed pairing. The other FAS1 modules of PN and βig-h3 have five conserved cysteines, one in FAS1-1 and two each in FAS1-2 and FAS1-3 (Figs. 1 and 2). When the positions of the cysteines are mapped on the global fold, one finds that the two in FAS1-3 are in close proximity on opposing β-strands, seemingly in perfect proximity to form an intra-module disulfide; the two in FAS1-2 are in proximity, but would require some adjustments of the module to form an intra-module disulfide; and the single cysteine in FAS1-1 has the potential to be on the surface of the module but could be buried at the module-module interface. Based on such observations, it has been concluded that disulfide linkages of βig-h3 to types VI and XII collagens, as has been found in nuchal ligament (Hanssen et al., 2003) and cornea (Runager et al., 2013), respectively, are likely mediated by the unpaired cysteine in FAS1-1. This cysteine also has the potential mediate formation of disulfide-linked homo-dimers. The number and locations of cysteines in FAS1-2, FAS1-3, and FAS1-4 of MFAS are different from that of PN and βig-h3, with three cysteines rather than none in FAS1-4 (Figs. 1 and 2).

Possible γ-carboxylation

The FAS1-1, FAS1-2, and FAS1-4 modules of PN and βig-h3 have sequences that are similar to those that mediate docking of γ-glutamyl carboxylase and vitamin K-dependent γ- carboxylation of nearby glutamate residues (Coutu et al., 2008). This feature came to the fore when proteins secreted by cultured mouse mesenchymal stem cells or human A549 adenocarcinoma cells were separated by two-dimensional isoelectric focusing/electrophoresis in sodium dodecyl sulfate followed by immuno-blotting with monoclonal antibody that recognizes γ-carboxyl-glutamic acid (Gla) independent of protein context (Coutu et al., 2008). Upon excision of anti-Gla-reactive spots, trypsinization, and analysis of tryptic peptides by mass spectrometry, peptide sequence matches with PN and βig-h3 were found (Coutu et al., 2008). Based on the locations of the putative γ-glutamyl carboxylase recognition sequences, it was concluded that multiple glutamates are γ-carboxylated in the three modules harboring the sequences. Thus, UniProt annotates PN (http://www.uniprot.org/uniprot/Q15063) and βig-h3 (http://www.uniprot.org/uniprot/Q15582) as containing up to 24 and 29 Gla residues, respectively, which would make these two proteins by far the most extensively γ-carboxylated proteins in the proteome. The possibility that such γ-carboxylation of PN and βig-h3 occurs is important. γ-carboxylation has profound effects on Ca⁺⁺-binding and structure-function of proteins subject to the modification, e.g., vitamin-K dependent blood coagulation factors, matrix Gla-protein, and osteocalcin (Furie and Furie, 1988, Stenflo, 1999). Studies of PN and βig-h3 are commonly done using recombinant proteins from bacteria or from insect or mammalian cells cultured in the absence of vitamin K, expression systems which would lack the ability to γ-carboxylate. If γ-carboxylated PN and βig-h3 are the functional forms of these protein, published and ongoing biochemical and cell biological studies, including those discussed throughout this review, would be irrelevant.

It is difficult to reconcile the structure of FAS1 modules with what is known about γ-carboxylation of well-established substrates. γ-glutamyl carboxylase works in concert with vitamin K oxido-reductase (VKOR) and vitamin K (Furie and Furie, 1988). For most γ-carboxylated peptides or proteins, substrate recognition and modification by the γ-glutamyl carboxylase are bipartite. Primary binding is to the recognition sequence in an N-terminal propeptide whereas modification occurs processively on multiple glutamyl residues on adjacent C-terminal peptide or protein; the propeptide is then removed by a furin-type protease as the peptide or protein matures in the endoplasmic reticulum and proximal Golgi apparatus (Furie and Furie, 1988, Bristol et al., 1996, Bush et al., 1999, Stenflo, 1999, Rishavy and Berkner, 2012). Although matrix Gla-protein provides an example of a recognition sequence in a small protein that is not removed prior to secretion (Price et al., 1987), there is no precedent for γ-glutamyl carboxylase working through multiple recognition sequences in a complex modular protein. In addition, although the prevailing view is that γ-carboxylation is determined by a chemical surface with a topology that is complementary to the surface of the γ-glutamyl carboxylase (Bush et al., 1999), arguments for such a surface existing on FAS1 modules are tenuous. The 16 residues (FxxxxxAxxxLxxxxR, underlined in Figs. 2 and 3) identified as constituting putative recognition sequences in PN and βig-h3 based on homology with propeptide recognition sequences (Coutu et al., 2008) overlap with the most conserved sequence of FAS1 modules (Figs. 2 and 3). Side-chains of three of the four residues that were emphasized in the comparison to known vertebrate γ-carboxylase recognition sequences (Coutu et al., 2008) are located on the inner faces of the β1-strand and α4-helix as shown in Fig. 3. Several other side-chains also are oriented towards the interior of the module.

No evidence was found by mass spectrometry for γ-carboxylation of corneal βig-h3 (Andersen et al., 2004). The molecular mass determined by MALDI mass spectrometry of a large fragment of porcine βig-h3 that encompassed the FAS1 modules was 68,092 Da, close to the predicted mass of 68,079 Da. In addition, tryptic peptides generated from porcine or human βig-h3 were devoid of post-translational modifications when analyzed by electrospray ionization mass spectrometry. Our directed attempts to demonstrate γ-carboxylation of PN and βig-h3 have been unsuccessful (Annis et al., 2015). First, we found that PN extracted from lungs of patients with idiopathic pulmonary fibrosis is not γ-carboxylated as assessed by its pI in two-dimensional electrophoresis, which was 7.0 to >8.0 as predicted for splice variants of unmodified PN, and by lack of reactivity with antibody that recognizes Gla residues irrespective of context (Brown et al., 2000). Second, we failed, using the anti-Gla antibody or mass spectrometry, to detect any γ-carboxylation of PN secreted by cells optimized for production of γ-carboxylated blood coagulation Factor VII, which is considered to be one of the more difficult of the vitamin K-dependent blood coagulation factors to produce as an active recombinant protein.

Because some under-γ-carboxylated blood coagulation factors are secreted poorly by some cells (Wu et al., 1996, Wu et al., 1997), we wondered whether the reported accumulation of PN and βig-h3 in the Golgi apparatus of cultured cells (Kim et al., 2009a, Maruhashi et al., 2010) is due to failure to γ-carboxylate these proteins because of absence of vitamin K in culture medium. However, we were unable to clear PN from its patchy perinuclear distribution within MG-63 human osteosarcoma cells by adding vitamin K to medium for 24 hr (Fig. 4).

Fig. 4.

Immuno-staining of PN in MG63 human osteosarcoma cells. Cells were incubated for 18 h on coverslips in medium containing 2% calf serum without or with 10 μg/ml Vitamin K and then fixed with formaldehyde, rendered permeable with 0.5% Triton, and stained with rabbit anti-PN followed by Alexa-Fluor-555-conjugated goat anti-rabbit-IgG antibody. Images were obtained with an Olympus BX60 fluorescence microscope using a 100X/1.30 oil immersion objective lens. Arrows point to PN in a patchy peri-nuclear pattern that was the same without and with vitamin K.

Why our results (Annis et al., 2015) differed from those of Coutu, et al. (Coutu et al., 2008) is not known. We used the same anti-Gla monoclonal antibody that, along with acidic shift in two-dimensional gel electrophoresis, provided the most persuasive evidence for γ-carboxylation in of PN secreted from cultured mouse or human mesenchymal stromal cells or βig-h3 secreted by adenocarcinoma cells (Coutu et al., 2008). It is particularly puzzling why peptides from PN and βig-h3 could be generated from protein focusing at pIs well below those of the unmodified forms of these proteins (Coutu et al., 2008) whereas we detected no PN in this region of two-dimensional gels. Possibilities are that PN or βig-h3 secreted from mesenchymal stem or adenocarcinoma cells are lacking N- or C-terminal basic residues as a result of proteolysis (Karring et al., 2010) or are phosphorylated or carry other acidic modifications. PhosphoSitePlus (www.phosphosite.org) lists ten residues in PN and seven in βig-h3 that have been identified as being phosphorylated in various phosphoproteomic screens.

In conclusion, we feel that it is unlikely that PN and βig-h3 are subject to extensive γ-carboxylation. Additional information may become available that suggests otherwise. However, for the purposes of the discussion below of various activities of PN and βig-h3, we assume that γ-carboxylation of these proteins does not take place.

Effects of knocking out MFAS, PN, or βig-h3 in Drosophila and zebrafish

Two Drosophila lines bearing chromosomal deficiencies overlapping the cytogenetic locus of Mfas were crossed, yielding flies that lacked Mfas transcripts; these flies did not have severe nervous system or tracheal abnormalities (Hu et al., 1998). A closer look at commissural axons that depend on proper formation of midline cells revealed mild defects in some of the Mfas-deleted embryos. As with crosses of mutants of fas1 and the abl non-receptor tyrosine kinase gene, the axon scaffold was more abnormal when flies lacking Mfas were crossed with flies mutant in abl. Use of chromosomal deficiencies to study gene function is less than ideal because other genes in the interval may be missing in addition to Mfas. However, these results combined with the failure of prior saturation mutagenesis screens to identify lethal mutations of Mfas were taken as evidence that presence of MFAS is not required for viability of flies under laboratory conditions (Hu et al., 1998). Additional fly lines with more targeted deletions of Mfas to study gene interactions, e.g., with mutants of fas1, have not been developed. In addition, little is known about roles of MFAS in adult flies. Perhaps most interesting in light of the pro-oncogenic phenotype of βig-h3-deficient mice described below, Mfas was identified as a possible “anti-oncogene” when transcript profiling of Drosophila hematocyte-like cells was used to search for effectors of hematopoietic tumor formation (Bina et al., 2010)

PN mRNA in zebrafish embryos was found in the rostral part of each somite in the early segmentation stage in a striped pattern (Kudo et al., 2004). At the end of segmentation, PN protein localized to the densely collagenous transverse myoseptum and the myotome-epidermis boundary. Antisense morpholino oligonucleotide to one of the PN mRNAs caused defects in myosepta formation, delay in the differentiation of myofibers, and a disordered connection between myofibrils and myoseptum (Kudo et al., 2004). Zebrafish βig-h3 tagged with green fluorescent protein and driven by a muscle-specific promoter was also found to accumulate at the myosepta (Kim and Ingham, 2009). Although attachment of muscle fibres to the myosepta was established and maintained normally in embryos receiving antisense morpholino oligonucleotides to βig-h3 mRNA, muscle fibres were reduced in growth and manifested disrupted myofibril bundles (Kim and Ingham, 2009). These studies, therefore, implicate PN and βig-h3 in myoseptum formation and differentiation of muscle fibers and hint at possible redundant function. It would be of interest to co-localize the two proteins and learn the phenotypes resulting from morpholinos directed to both PN mRNAs and to βig-h3 as well as both PN mRNAs.

Effects of knocking out PN or βig-h3 in mice

Several lines of PN-null mice have been studied. To generate peri^lacZ mice, the translation start site and first exon were replaced with a lacZ reporter gene (Rios et al., 2005). The distribution of β-galactosidase activity in heterozygotes and homozygotes recapitulated the pattern of embryonic endogenous PN mRNA expression. Null embryos were viable in utero, and animals appeared grossly normal at birth. Postnatally, ~15% died before weaning, and the rest had retarded growth. Skeletal analysis revealed that trabecular bone in adult homozygous skeletons was sparse, but overall bone growth was unaffected. By three months, a periodontal disease-like phenotype was noted associated with a severe incisor enamel defect. Placing the peri^lacZ mice on a soft diet alleviated mechanical strain on the periodontal ligament and resulted in partial rescue of the periodontal and enamel phenotypes. In addition, peri^lacZ mice exhibited defects in the adult periosteum, cartilage, and cardiac valves (Rios et al., 2005). Postn^−/− null mice generated by excision of the first exon without replacement with a reporter gene were viable and appeared normal on birth (Shimazaki et al., 2008, Kii et al., 2006). The Postn^−/− mice showed an eruption disturbance of incisors that did not require special chow and developed periosteal abnormalities confined to the proximal tibia at three months (Kii et al., 2010). Independently generated Postn ^−/− null mice lacking the exons 4–10 encoding the FAS1-1, FAS1-2, and FAS1-3 modules (Oka et al., 2007) had lower body weights and slightly smaller hearts than wild-type littermates, but otherwise appeared relatively normal. Both strains of Postn ^−/− mice were more prone to ventricular rupture after a myocardial infarction; surviving deficient mice had less cardiac fibrosis, better ventricular performance, and less inflammatory cell recruitment (Oka et al., 2007, Shimazaki et al., 2008). Postn ^−/− mice also demonstrated less cardiac fibrosis and hypertrophy following long-term pressure overload. These and other observations of peri^lacZ and Postn^−/− mice indicate that the ability to create a PN-rich environment, although dispensable for development, is an important determinant of tissue responses to insult and injury (Conway et al., 2014, Liu et al., 2014).

Independently generated targeted disruptions of Tgfbi in mice, when homozygous, were found to be associated with alterations in cartilage and bone (Yu et al., 2012, Lee et al., 2015), predisposition to spontaneous tumor formation (Zhang et al., 2009) and diabetes (Han et al., 2014), and protection from the pro-inflammatory response to septic shock (Bae et al., 2014). The skeletal changes were attributed to altered regulation of cartilage turnover (Lee et al., 2015) and, as with targeted disruption of Postn (Kii et al., 2010), altered periosteal formation of cortical bone (Yu et al., 2012). The effects on tumor formation and islet survival and function were traced to alterations in up-regulation of cyclin D1 (Zhang et al., 2009) and AKT and mTORC1 signaling pathways (Han et al., 2014). Severe sepsis in mice and humans was demonstrated to be accompanied by increase in concentration of serum βig-h3 (Bae et al., 2014). βig-h3 was found in platelets, as has been previously shown by immuno-blotting (Kim et al., 2009b) and mass spectrometry (Burkhart et al., 2012), and in endothelial cells; both were considered possible sources of the increased serum levels (Bae et al., 2014). These observations suggest that, as for PN, βig-h3 is important for ongoing cellular health in maturing and mature mice.

Differences in the phenotypes that result from disruptions of Postn or Tgfbi indicate that the two proteins have functions in mice that, although generally similar, are specific to one or the other protein. The possibility of redundancy has not been addressed systematically. Transcripts for both proteins were demonstrated during valvuloseptal morphogenesis in the developing mouse heart in patterns that are partially overlapping (Lindsley et al., 2005). Usually, however, observations have been directed towards one protein and not the other. Crosses that express neither protein have not been studied to learn if the individual knock-out phenotypes are modified. For instance, do mice lacking both proteins have more severe bone and valve abnormalities, greater disposition to tumor formation and diabetes, or different responses to sepsis than mice with individual deficiencies?

Roles of PN and βig-h3 in ECM formation

PN and βig-h3 are present in the human circulation, with plasma concentrations estimated to be ~16 ng/ml (~2 nM) for PN and ~140 ng/ml (~17 nM) for βig-h3 by mass spectrometry and spectral counting (http://www.plasmaproteomedatabase.org/ and (Farrah et al., 2011)). Labeled βig-h3 has been reported to circulate for only 20 min when infused into mice (Son et al., 2013). Both PN and βig-h3 can be visualized in ECM by immuno-histology of tissues undergoing allergic, reparative, or degenerative processes (Takayama et al., 2006, Conway et al., 2014, Liu et al., 2014) or immuno-histologic or mass spectrometric analysis of tumors (Soikkeli et al., 2010, Nummela et al., 2012, Naba et al., 2014). By immuno-electron microscopy, PN has been localized to deposits associated with cellular processes and collagen bundles of mouse periodontal ligament (Suzuki et al., 2004) and collagen fibers of the mouse atria-ventricular valve (Norris et al., 2007). βig-h3 has been localized by immuno-electron microscopy to type VI collagen microfibrils associated with elastin and collagen bundles of several tissues (Gibson et al., 1997).

PN and βig-h3 are considered to be members of the large class of ECM components called matricellular proteins, which are defined by loose functional criteria rather than sequence homology. Matricellular proteins may be only transient constituents of ECM, do not contribute to structural integrity of ECM, and, importantly, provide information to cells interacting with ECM (Murphy-Ullrich and Sage, 2014). In addition, knock-out of a matricellular protein in mice may have little or no effect on development, belying the major effects that the absence or presence of the protein may have on cultured cells in vitro (Murphy-Ullrich and Sage, 2014). Matricellular proteins are multi-modular and multi-functional and tend to interact with one another and multiple cell surface components. Consequently, important information about matricellular proteins—routes by which the proteins are deposited in ECM, half-lives once deposited, mechanisms of clearance, and cellular receptors mediating information transfer—can be difficult to come by and may not be straightforward. For matricellular proteins that are found in the circulation, it is not known whether these represent proteins that fail to deposit, are released from deposits, or are secreted for the specific purpose of maintaining a “normal” concentration of the protein in blood plasma. Like Drosophila SPARC, a matricellular protein that functions as an apparent chaperone for deposition of type IV collagen by hemocytes (Martinek et al., 2008), PN and βig-h3 may have important functions in the cellular secretory apparatus as well as in the pericellular space (Kudo, 2011). Despite the loose definition of a matricellular protein, grouping the proteins together is proving to have increasing heuristic value in evaluating the roles of these proteins in diverse cellular processes and diseases (Murphy-Ullrich and Sage, 2014).

PN and βig-h3 are found in fibrillar networks associated with metastatic tumor outgrowths; this same ECM contains tenascin-C and fibronectin (Soikkeli et al., 2010, Nummela et al., 2012, Naba et al., 2014). By enzyme-linked absorbent assays, PN has been shown to bind to tenascin-C, fibronectin, type V collagen, and itself (Takayama et al., 2006). Binding was enhanced when the C-terminal tail of PN was not present (Takayama et al., 2006, Kii et al., 2010). Investigations motivated by finding similar patterns of periostitis and irregular deposition of type I collagen in tibia of mice lacking PN or tenascin-C (Kii et al., 2010, Maruhashi et al., 2010), however, revealed an interplay among PN, tenascin-C, and fibronectin that goes beyond simple co-deposition in ECM. Fibronectin and type I collagen deposited by fibroblasts cultured from the PN- or tenascin-C-knock-out mice had a distinctly different meshwork architecture, with loss of branching and fibril bundling (Kii et al., 2010). Tenascin-C co-localized with fibronectin in ECM of wildtype cells expressing PN but was largely missing from the abnormal ECM of cells lacking PN. Thus, PN seems to influence ECM architecture by controlling the deposition of tenascin-C. Remarkably, dual immuno-fluorescence co-localized PN and tenascin-C in the Golgi apparatus rather than in the ECM (Kii et al., 2010), suggesting that an event occurring in the Golgi apparatus somehow influences ECM formation outside of the cell (Fig. 5).

Fig. 5.

Interactions of PN and βig-h3 inside and outside cells. The cartoon depicts collagen (COL), βig-h3 (β), fibronectin (FN), PN, tenascin-C (TN), LOX, and BMP1 being led into endoplasmic reticulum (ER) and trafficking through endoplasmic reticulum and the Golgi apparatus during secretion. Components close to one another have been localized together by immuno-fluorescence microscopy along with a Golgi marker or surmised to interact during secretion. Fibrils of fibronectin and collagen are shown outside the cell along with tenascin-C, which patterns the fibrils, and BMP-1, which activates LOX. LOX, in turn, cross-links collagen. The relationship of the two fibril systems is purposely left vague, and COL could be type I, VI, or XII collagen as described in the text. Only the online version of this figure is in color.

Investigations exploring why lack of PN results in decreased stability and lysine-based cross-linking of type I collagen (Norris et al., 2007, Shimazaki et al., 2008, Kii et al., 2010) added to this story. Bone morphogenetic protein 1 (BMP1), like tenascin-C, was found to co-localize with PN in the Golgi apparatus and with fibronectin in the ECM (Maruhashi et al., 2010). BMP1 is a tolloid-class metalloproteinase that cleaves lysine oxidase (LOX) as well as a number of collagens, other ECM components, and cytokines into functional forms (Hopkins et al., 2007). LOX catalyzes the oxidative deamination of peptidyl lysine residues in collagen molecules to aldehydes that spontaneously condense to form covalent cross-links. In cell cultures lacking PN, less active LOX and covalent cross-links were present (Maruhashi et al., 2010). Both BMP1 (Huang et al., 2009) and LOX (Fogelgren et al., 2005) are known to bind to fibronectin in an interaction that enhances the activation of LOX by BMP1. The bottom line seems to be that Golgi PN orchestrates delivery of BMP1 to ECM and thereby enhances LOX activation and introduction of cross-links into collagen (Fig. 5).

Mutation of Trp65 to Ala (murine numbering, Trp63 in human PN) in the EMI module destroyed the ability of PN to support the delivery of either tenascin-C (Kii et al., 2010) or BMP1 (Maruhashi et al., 2010) to ECM and effects on ECM architecture and collagen cross-linking. Fibronectin interacts with PN via the EMI module, tenascin-C and BMP1 interact with PN via the FAS1 modules, and the W65A mutation causes loss of the interaction with fibronectin (Kii et al., 2010, Maruhashi et al., 2010). Thus, the effects of the W65A mutation are strong evidence that PN acts through fibronectin. Such action could involve “passing” tenascin-C or BMP1 to fibronectin in the Golgi or chaperoning tenascin-C or BMP1 from Golgi to fibronectin-containing ECM in the pericellular space or some other event that is at present obscure.

βig-h3 was found to be bound by reducible, presumably disulfide linkages to type VI collagen in nuchal tendon microfibrils (Hanssen et al., 2003) and type XII collagen, a “fibril-associated collagen with interrupted triple helices” (FACIT) collagen associated with striated collagen bundles in the cornea (Runager et al., 2013). It has been suggested that the disulfide link between βig-h3 and type VI collagen may take place intracellularly (Hanssen et al., 2003). PN and βig-h3 co-localize to the trans-Golgi network of corneal fibroblasts and epithelial cells, probably via the EMI modules (Kim et al., 2009a). Thus, the two paralogs may co-operate in secretion of the collagens to which βig-h3 is linked.

There are also potential roles of PN and βig-h3 on collagen deposition in the extracellular space beyond the effects of PN on LOX activation. When mixed with polymerizing type I collagen, soluble PN increased the elastic modulus of the gel (Sidhu et al., 2010). βig-h3 has been shown to bind to biglycan and decorin, both small leucine-rich proteoglycans, and thereby influence the aggregation of type VI collagen (Reinboth et al., 2006).

Finally, it should be noted that effects of PN in the secretory pathway have been suggested to extend to other than ECM molecules. Lack of association of PN with Notch1 precursor inside of the cell may account for suppression of Notch1 signaling (Tkatchenko et al., 2009) and expression (Tanabe et al., 2010) in cells lacking PN. Further, it has been observed that endogenously expressed PN preferentially activates intracellular kinase pathways when compared to PN added to culture medium (Ghatak et al., 2014).

Roles of PN and βig-h3 in cell adhesion

Most of the attention on activities of PN or βig-h3 in the pericellular space has focused on stimulation of cell adhesion, migration, and “outside-in” signaling via interaction with transmembrane integrin receptors (e.g. (Son et al., 2013, Ghatak et al., 2014)). In humans, there are 18 ɑ integrin subunits and eight β integrin subunits that pair into 24 different heterodimeric ɑβ receptors (Humphries et al., 2006, Byron et al., 2009). Integrins can be classified by the most commonly paired subunits (the ɑV subunit pairs with five different β subunits, the β1 subunit pairs with 12 different ɑ subunits, the β2 subunit pairs with four different ɑ subunits), whether there is preferential recognition of RGD or Leu-Asp-Val (LDV) or another sequence in a ligand, and whether ligand recognition is by the von Willebrand factor A module in the β subunit (βA) or the homologous ɑA module that is present in nine of the ɑ subunits (Humphries et al., 2006). A given cell expresses multiple different integrins, and a given ligand may be recognized by several different integrins. A given integrin, however, must be in an activated conformation that allows an aspartate or glutamate residue of the ligand to coordinate a cation embedded in the βA or ɑA module (Humphries et al., 2006). Access to the cation is allosterically controlled by binding of cytoplasmic proteins to integrin intracellular tails and transmitted through transmembrane domains, so-called “inside-out signaling” (Humphries et al., 2006, Ginsberg, 2014). Thus, in addition to selecting what repertoire of integrins is expressed, a cell at any one time precisely controls what members of that repertoire are able to interact with the multiple potential ligands deposited in ECM or on the surfaces of other cells.

A classical method for sorting through integrin-ligand interactions is to adsorb purified potential ligand to the surface of a culture plate, quantify adhesion of cells to the ligand-conditioned surface, and inhibit specific adhesion to the ligand by pre-incubation of cells with monoclonal antibodies that block functions of different integrins (Byron et al., 2009). An example of such experiments is shown in Fig. 6, in which we asked if human neutrophils respond to PN the same way that eosinophils do (Johansson et al., 2013). Neutrophils without or with stimulation with granulocyte-monocyte colony stimulating factor (GM-CSF) were placed on a surface to which PN was adsorbed followed by bovine serum to block any non-occupied sites of protein adsorption. The control surface was coated with serum alone. To study migration, the surface was also coated with latex beads, which are cleared as the cells migrate. Adhesion and migration on PN were enhanced by stimulation with GM-CSF and blocked by antibodies to αMβ2 (Fig. 6). As with eosinophils (Johansson et al., 2013), migration was maximum at an intermediate concentration of adsorbed PN.

Fig. 6.

Neutrophil adhesion and motility on PN. (A) Adhesion of blood neutrophils incubated at 10⁵ cells/ml for 1 hr in the presence or absence of GM-CSF, 10 ng/ml, in wells of microtiter plates. Wells were coated with PN, 10 μg/ml, and post-coated with fetal bovine serum or coated with fetal bovine serum alone. Adhesion was assayed by cell content of myeloperoxidase and is expressed as percentage of input cells (Stark et al., 1996). (B) Migration of blood neutrophils at 10⁴ cells/ml in the presence or absence of GM-CSF, 10 ng/ml. Monolayers of polystyrene beads were made in wells coated with PN at different concentrations and post-coated with fetal bovine serum (diagram below). Migration was determined by morphometric analysis of percentage of the bead coating cleared after 20 hr, as described (Johansson et al., 2013). (C) Migration of blood neutrophils preincubated for 5 min with blocking monoclonal antibody to integrin subunits or isotype control (mIgG1) before application to wells coated with PN, 5 μg/ml, and a monolayer of beads as in (B). GM-CSF, 10 ng/ml, was present, and percentage of bead coating cleared after 20 hr was determined. The cartoons underneath panels A and B/C depict the experimental set-ups with neutrophils in red and latex beads in green. Only the online version of this figure is in color.

PN and βig-h3 are able to interact, depending on the cell type, with a remarkably diverse set of integrins: ɑV-containing/RGD-recognizing/non-ɑA-containing integrins ɑVβ3 and ɑVβ5 (Gillan et al., 2002, Kim et al., 2002, Nam et al., 2003, Bao et al., 2004, Orecchia et al., 2011, Son et al., 2013), β1-containing/laminin-recognizing/non-ɑA-containing integrin ɑ3β1(Kim et al., 2000), β1-containing/collagen-recognizing/ɑA-containing integrin ɑ1β1 (Ohno et al., 1999), β4-containing/laminin-recognizing/non-ɑA-containing integrin ɑ6β4 (Kim et al., 2003, Baril et al., 2007), and β2-containing/LDV-recognizing/ɑA-containing integrin ɑMβ2 (Kim and Kim, 2008, Johansson et al., 2013). In addition, platelets interact with βig-h3 by RGD-recognizing ɑ5β1 and ɑIIbβ3 (Kim et al., 2009b). βig-h3 contains the RGD sequence in its C-terminal tail, but this sequence when tested did not seem to play an essential role in integrin recognition (Ohno et al., 1999, Kim et al., 2000, Kim et al., 2009b). Instead, FAS1 modules of βig-h3 in their tandem array or individually are adhesive (Kim et al., 2000, Kim et al., 2002, Nam et al., 2003, Son et al., 2013). For adhesion to the FAS1-4 module of βig-h3 by ɑ3β1, mutation of the conserved Asp-Ile (DI) sequence in the β6 strand (Fig. 2) resulted in loss of activity (Kim et al., 2000). Adhesion mediated by ɑvβ3 and ɑvβ5, in contrast, was lost upon mutation of the conserved Tyr-His (YH) sequence in the β2 strand (Fig. 2) (Kim et al., 2002, Nam et al., 2003). The epitope for the OC-20 mouse monoclonal anti-PN IgM antibody, which inhibits ɑVβ3-mediated adhesion to PN, mapped to a YH-containing sequence in FAS1-2 on PN, further implicating the YH motif as being important for adhesive activity (Orecchia et al., 2011). However, a compelling argument has been made based on structural analyses of FAS1 modules (Clout and Hohenester, 2003) that the DI or YH motifs cannot interact directly with integrins according to the paradigms established by the RGD sequence of fibronectin or Ile-Asp-Ser (IDS) sequence of vascular cell adhesion molecule. The two-residue motifs are not in surface loops like the RGD (Leahy et al., 1996) or IDS (Wang et al., 1995) sequences and instead participate in intra-module interactions that are intrinsic to the global fold (Clout and Hohenester, 2003), raising the possibility that mutating the motifs corrupts protein structure sufficiently to abrogate integrin binding to surface sites distal from the motifs rather than changing residues that directly engage the integrin.

The literature suggests, therefore, that the FAS1 modules of PN and βig-h3 interact with integrins via yet-to-be-defined surface features that are displayed by the modules in their native state. The interaction with ɑvβ3 and ɑvβ5 is optimal when the complete set of tandem FAS1 modules of βig-h3 is present as opposed to a single module or a construct comprising four copies of the same module and is further enhanced by inclusion of the RGD-containing C-terminal tail (Son et al., 2013). Possible explanations for such results are cooperative interactions of integrins with binding sites in multiple FAS1 modules and requirement for module-module interactions among adjacent FAS1 modules in order for a conformationally appropriate binding surface to be present on one of the FAS1 modules. The surface features must be somewhat different in PN and βig-h3 as evidenced by PN-binding DNA aptamers (PNDA) selected to bind to PN and inhibit interactions of PN with ɑvβ3 and ɑvβ5 (Lee et al., 2013). The best of the aptamers, PNDA-3, was found to require FAS1-3 for binding and to bind to PN with 30-fold higher affinity than to βig-h3. Mapping the binding sites on PN and βig-h3 and the many integrin partners likely will require attention to the effects of mutagenesis on PN or βig-h3 structure as well as on adhesive function. Such studies would be facilitated by an expanded set of reagents like OC-20 and PNDA-3 that recognize specific features of PN and βig-h3 and can be used to define which sites interact with which integrins.

Roles of PN and βig-h3 in asthma

Interleukin-4 (IL-4) and IL-13, cytokines released by T-helper 2 (T_H2) cells, are powerful up-regulators of PN synthesis and deposition in bronchi of humans with asthma and mice subjected to allergic challenge of the airway (Takayama et al., 2006, Woodruff et al., 2007, Sidhu et al., 2010, Conway et al., 2014). IL-13 has been shown to act upon bronchial epithelial cells to cause basal secretion of PN (Sidhu et al., 2010). In addition, forced expression of PN by epithelial cells resulted in an epithelial to mesenchymal transition associated with increased expression of TGF-β and type I collagen (Sidhu et al., 2010). Finally, epithelial cell-derived TGF-β also was shown to increase type I collagen synthesis by lung fibroblasts (Sidhu et al., 2010). Such results indicate a scenario in which epithelial cells direct formation of sub-epithelial “TH2-conditioned ECM” rich in PN. In experiments such as those shown in Fig. 6 for neutrophils, PN supported ɑMβ2 integrin-mediated adhesion and migration of IL-5-stimulated eosinophils; migration was maximal at intermediate coating densities (Johansson et al., 2013). These findings suggest that a gradient of deposited PN may guide eosinophils to subepithelial regions of high PN density in the asthmatic airway A number of recent studies have demonstrated that an increased concentration of PN in serum, presumably “spill-over” from the increased synthesis of PN in the airway, is a biomarker of T_H2-associated airway inflammation predictive of airway eosinophilia and responses to inhaled corticosteroids and lebrikizamab, a humanized monoclonal antibody that binds and inhibits IL-13 (Parulekar et al., 2014). Thus, PN can be viewed as a central player in the T_H2-conditioned ECM in asthmatic airways by orchestrating the movement of eosinophils and other white cells and remodeling of other ECM molecules.

Testing such a conjecture in murine models of allergic airway inflammation has yielded distinctly different results for Postn^−/− mice lacking exons 4–10 (Oka et al., 2007) (available through Jackson Laboratory and referred to in some publications as 129-Postn^tm1Jmol/J mice, www.jax.org) and peri^lacZ mice expressing lacZ at the Postn locus (Rios et al., 2005). The proportion of eosinophils in bronchoalveolar lavage (BAL) and the eosinophil count in the esophageal mucosa after intranasal challenge with allergenic extracts of Aspergilllus fumigatus were decreased in Postn^−/− mice as compared to wild-type (Blanchard et al., 2008). Reciprocally, eosinophils in blood were increased after challenge compared to control. Postn^tm1Jmol mice exposed to Dermatophagoides pteronyssimus (house dust mite) extract by intranasal challenge were found to have decreased BAL eosinophils compared to wild-type mice along with decreased BAL neutrophils, monocytes, and lymphocytes (Bentley et al., 2014). Treatment of wild-type mice with OC-20 anti-PN mAb also decreased BAL eosinophils and lymphocytes. Further, Postn^tm1Jmol knockout or OC-20-treated wild-type mice had decreased airway responsiveness, mucous metaplasia, and lung tissue expression of IL-13 (Bentley et al., 2014). Similarly, Postn^tm1Jmol mice challenged intra-nasally had decreased eosinophil count in the nasal mucosa, thinner subepithelial tissue, and decreased subepithelial collagen I staining in this tissue (Hur et al., 2012). In contrast, peri^lacZ mice sensitized and challenged with ovalbumin had no significant differences in BAL cell counts compared to wild-type mice ((Sehra et al., 2011) It should be noted that the description of the mice is confusing in this paper. Their origin has been confirmed in correspondence with one of the paper’s authors.). Further, challenged peri^lacZ mice had increased airway resistance, mucus production by goblet cells, and production of IL-4, IL-5, and IL-13 by lymph node cultures compared to wild-type littermates, suggesting that PN is part of a negative feedback loop regulating Th2 responses and allergic inflammation. Likewise, after intranasal challenge with A. fumigatus antigen, peri^lacZ mice had no difference in BAL cell counts, increased airway responsiveness and serum IgE concentration, and decreased TGFβ1 transcript and protein in lung compared to wild-type controls, although there was no difference in epithelial mucin (Gordon et al., 2012).

Thus, three studies comparing Postn^−/− or Postn^tm1Jmol mice, null in PN due to excision of six exons, to wildtype littermates indicate that the PN present in wildtype mice promotes allergic inflammation by stimulating recruitment of eosinophils and other leukocytes and generation of Th2-related cytokines leading to airway hyper-responsiveness. In contrast, two studies comparing peri^lacZ mice, expressing lacZ at the Postn locus, to wildtype controls indicate that PN in wildtype mice suppresses T_H2 responses, airway remodeling, and airway responsiveness and that the absence or presence of PN does not influence leukocyte recruitment. Possible explanations put forward to explain these discrepancies include differences in backcrossing of mice or mode of allergen administration or that peri^lacZ mice experience a subtle change in airway wall development that increases airway collapsibility (Gordon et al., 2012, Bentley et al., 2014). Another possibility is that effects are due to one or more of the altered transcripts and proteins with the potential to be generated from the altered Postn loci being produced in greater amounts when bronchial epithelial cell are exposed to IL-4 and IL-13. The balance between T_H1 and T_H2 immune responses resulting from insults to epithelium is finely balanced, as if on a razor edge, by interactions among various types of lymphocytes, dendritic cells, monocytes, eosinophils, basophils, and mast cells, all working through cytokines and other regulatory molecules (Wynn, 2015). Many and perhaps all of these cells may be able to adhere to and migrate on PN in subepithelial matrix. It is easy to envision the T_H1/T_H2 balance being tipped in different directions by the manipulations of the Postn locus that generated the Postn^−/− and Postn^tm1Jmol and the peri^lacZ mice. Thus, the most important message may be that all five studies describe major changes in airway inflammation upon disruption of Postn and not that the patterns of inflammation were different. Jackson Laboratory is in the process of phenotyping a B6N(Cg)-Postn^{tm1.1(KOMP)V/Cg}/J mouse generated by the Knockout Mouse Phenotyping Program (http://jaxmice.jax.org/strain/024186.html). The “ZEN-UB1 Velocigne” cassette was inserted into the gene, replacing all Postn coding exons and intervening sequences, and then removed after selection of targeted embryonic stem cells, leaving behind no transgenes. Studies of these mice may be useful in resolving the present inconsistencies about roles of PN in allergic airway inflammation particularly if crossed with mice harboring other alterations that bias the balance between T_H1 and T_H2 responses.

Future directions

This review leaves many questions about PN and βig-h3 unanswered. What is the core structure of the EMI module? Is there connectivity among EMI and FAS1 modules such that the structure of one influences the structures of neighboring modules? What are the structures of the tails? How does the PN C-terminal tail inhibit interactions of EMI and FAS1 modules with ECM partners? Do different splice variants of the tail inhibit interactions differently? What constitute the binding sites in PN and βig-h3 for other ECM molecules? Do different integrins recognize different binding sites? What are the dynamics of secretion? Especially, what accounts for the apparent accumulation of PN and βig-h3 in the Golgi apparatus? Are tenascin-C, BMP-1, and collagens chaperoned by PN and βig-h3 into ECM, and if so, how does that happen? What are the roles in barrier immunity, tissue repair, and over-exuberant or chronic T_H2 responses of PN that is hyper-expressed as part of the T_H2 response? Is βig-h3, which, as its name implies, was identified early on as a TGF-β-responsive gene in adenocarcinoma cells (Skonier et al., 1992) and bronchial smooth muscle cells (Billings et al., 2000), deposited alongside PN and a co-determinant of functions of T_H2-conditioned ECM? More generally, what functions of PN and βig-h3 are redundant and what are specific? Can models of tissue insult and injury be established in Drosophila, with a single PN/βig-h3 homolog, and zebrafish, with two PNs and one βig-h3, to allow exploration of function after tissue insult and stress more definitively and quickly than with murine models of disease? The PN/βig-h3 literature is extremely rich, and this review hardly touches on PN and βig-h3 in the context of cancer, inflammatory diseases other than asthma, tissue regeneration, and corneal dystrophies. There is much research to be done on PN and βig-h3 that will be important and have wide impact in diverse areas of biomedical science.