Forty years ago, the tremendous complexity of extracellular matrix (ECM) was still largely an uncharted area, mainly because many of its components could only be solubilized by denaturing agents. Known were just five types of collagens, elastin, a couple of proteoglycans, and a few ECM glycoproteins, among them fibronectin, thrombospondin-1, and laminin-111 (1). The best studied was fibronectin (2), which became notorious for promoting specific cell adhesion to collagens. In parallel, the search for yet undetected large ECM glycoproteins continued. In 1981-82, Carter (3) observed several novel glycoproteins in human fibroblast ECM extracts. Among them, "GP250" was shown to be distinct from fibronectin but resisted isolation. However, between 1983-85 several research groups independently discovered and characterized a similar ECM glycoprotein that later became known as tenascin-C (see below). Its subunits were comparable in size to fibronectin but instead of dimers formed large (>10⁶ kDa) disulfide-linked oligomers. In the following paragraphs, the history and context of the individual discoveries of tenascin-C is briefly recounted. I then describe how a combination of methods available at the time lead to a detailed structural model of tenascin-C. This was the basis for trying to assign functions to different parts of the molecule. For reasons outlined below, this turned out to be more difficult than for fibronectin.

A crude model of tenascin-C was already laid out in the first publications. The molecule consisted of several similar or identical 220 kDa subunits that were linked together at one end by disulfide bridges (6). EM images of rotary-shadowed molecules revealed an intriguing hexabrachion structure (Figure 1). From opposite sides of a central globule, two triplets of arms were emanating that showed distinct features. Each arm had a thin proximal rod fused to a thicker, flexible middle region and ended in a distal globular domain (7, 18, 19). From expression cloning using the new antibodies against tenascin/cytotactin, its first cDNA sequences were published in 1988 (20, 21). Fortunately, the tenascin-C cDNA contained sequence repeats coding for some small protein modules for which X-ray structures were already available. These were (from N- to C-terminus) an alpha-helical coiled coil domain suitable for oligomerization, epidermal growth factor (EGF)-like modules, a series of fibronectin type III (FN3) domains, and a single globular domain related to the fibrinogen gamma-chain (Figure 1A).

It was tempting to correlate the tenascin-C cDNA sequence with the structural features of hexabrachion particles as observed by EM. The N-terminal coiled-coil domain with adjacent cysteines could link two triplets of subunits at the central globe via disulfide bridges. The EGF-like repeats might correspond to the proximal rod domain of each hexabrachion arm, the stretch of FN3 domains to the flexible middle part, and the fibrinogen-like domain to the distal globule (Figure 1A). To prove this hypothesis for avian tenascin-C, two approaches were used. First, a library of monoclonal antibodies was generated against the purified molecule (18). Individual antibodies were tested for their binding to different recombinant tenascin-C fragments that were derived from various parts of the sequence (22). Simultaneously, individual mAbs were incubated with intact tenascin-C, and the mixture was examined in the EM after rotary shadowing. IgG molecules attached to hexabrachion particles could easily be identified and their binding sites mapped from such images (Figure 1C). For example, mAb M1 reacted with a recombinant chick tenascin-C fragment containing the N-terminal EGF-like repeats, and mAb Tn68 with a fragment comprising the C-terminal FN3 repeats (22). On EM images, mAb M1 bound to the proximal thin rod region of tenascin-C particles (Figure 1C), and mAb Tn68 to the flexible middle region close to the distal globe. Similarly, mAbs against other parts of the molecule were mapped by EM on intact tenascin-C.

Later, a complementary approach derived from analyzing recombinant deletion variants of full-length tenascin-C by EM. Fischer et al. (19) expressed and purified chick tenascin-C subunits that lacked specific parts of the intact molecule, but still assembled correctly into hexabrachion particles. Reactivity of such tenascin-C variants with antibodies matched with the specific deletions. A variant missing the EGF-like domains reacted with all tenascin-C specific mAbs except mAb M1, and when examined in the EM, it lacked the proximal rod-like part of the hexabrachion arms (Figure 1D). A FN3 deletion variant did not react with mAb Tn68, and missed the flexible middle region from the hexabrachion structure. Similarly, when the fibrinogen-like domain was deleted, hexabrachions without the distal globe of the arms were observed by EM. By these means, it was possible to generate a precise model of the hexameric chick tenascin-C molecule (19, 22) (Figure 1). Knowing the X-ray structures of the various modules from those of the original proteins, on paper the model of 1.3x10⁶ kDa hexameric tenascin-C could be extrapolated essentially to atomic resolution.

Elucidation of the structure of a complex protein is rightly considered the basis for understanding its interactions and functions. In the ECM field in the early 1980s, fibronectin was the role model. Once its general structure was known, various defined functions could be assigned to distinct parts of the molecule, among them cell/integrin, gelatin/collagen, and heparin/glycosaminoglycan binding sites. In fibronectin, all these sites fully retained their individual functions as small proteolytic or recombinant fragments. Fibronectin appeared like an interface with independent functional units plugged in, enabling it to connect cells with ECM (23). However, the example of fibronectin spoiled us. For other large ECM proteins, the structure-function relationships turned out to be much more complex, and tenascin-C is a puzzling example. It has been classified as an adhesion-modulating “matricellular” protein with seemingly context-dependent activities. On the one hand, it is a poor adhesion molecule for many cells and notorious for inhibiting spreading of fibroblasts on fibronectin (18). On the other hand, it binds proteoglycans and promotes neurite outgrowth [for review, see (24)]. However, it proved difficult to assign these functions to specific parts of the molecule. Partially conflicting results were reported from using small recombinant tenascin-C fragments (22, 25, 26). In a reciprocal approach, Fischer et al. (19) instead analyzed avian tenascin-C deletion mutants for function. In this case, tenascin-C lacking the C-terminal fibrinogen-like domain completely lost its ability to inhibit fibroblast spreading. However, deletion of other regions (primarily adjacent FN3 repeats) diminished this activity as well. Similarly, the neurite-promoting activity of full-length tenascin-C disappeared completely by removing just the fibrinogen globe. However, when the FN3 repeats were omitted in addition, a strong neurite-promoting activity became unmasked in the EGF-like region, which was not seen with intact tenascin-C and did not depend on the fibrinogen domain. Thus in contrast to fibronectin, distinct activities of tenascin-C appear to depend on strong crosstalks between its various domains (19). Such interactions can also affect functions (such as glycosaminoglycan binding) that are observed with certain fragments but hidden in the full-length molecule (27). Intact tenascin-C is clearly much more than just the sum of its parts, which complicates the analysis of structure-function relationships. This should be taken into account when studying the more recently discovered roles of tenascins in the immune system (28), which are the topic of this special issue.

MC designed the article, assembled the literature, drafted and edited the manuscript, and prepared the figure.

Own work on tenascin-C was supported by a postdoctoral fellowship of the Bay Foundation, and by a career development award (START fellowship #9060) and research grants #9474, #31007, #45952, #55551, #67029, and #107515 from the Swiss National Science Foundation.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.