In Search Of Molecular Dovetails in Mussel Byssus: From The Threads to The Stem

Marine mussels (Mytilus spp.) are superbly successful inhabitants of wind- and wave-swept rocky shores. This success is due in large part to their strong and opportunistic attachment to hard surfaces. Attachment is mediated by a fibrous shock-absorbing structure known as the byssus (Brown, 1952). At their distal ends, byssal fibers or threads are glued to the substratum by adhesive plaques; proximally, the threads merge with a stem that is deeply rooted in the base of the mussel foot. In recent years, much effort has been directed towards developing a molecular explanation for the unusual mechanical properties of byssal threads. They resemble soft rubber at one end and rigid nylon at the other (Smeathers and Vincent, 1979; Bell and Gosline, 1996). The transition from one end to the other is seamless and gradual.

Three protein precursors of byssal threads have now been sequenced and reveal a block copolymer structure. All have the following block-like domains: histidine-rich amino and carboxy termini and a central collagenous domain sandwiched between two flanking domains (Waite et al., 1998). The flanking domain sequences most distinguish the three proteins from one another and determine their axial distribution in the thread. In preCol-P, the protein that prevails in the rubbery proximal end of byssal threads, the flanking domain sequence resembles elastin, a protein rubber (Coyne et al., 1997). In contrast, the major component in the stiff distal end of the thread is preCol-D, with a flanking domain sequence resembling that of spider dragline silk (Qin et al., 1997). There is no available information to suggest how these precursors are assembled in mature byssus, but a recent model suggests that a third protein, preCol-NG, might serve to mediate the graded progression of preCol-P and preCol-D between the proximal and distal ends (Qin and Waite, 1998).

In the present work, we report on the existence of two additional variants of preCol-P in the foot of M. edulis. One of these has a graded distribution in the foot, as previously reported for preCol-P. The other variant is expressed only in the stem gland region and may represent a crucial molecular ‘dovetail’ for joining the proximal threads to the byssal stem. Dovetailing is an important mechanism that allows organisms to mitigate the stresses of joined materials with dramatically different properties; mechanical, electrical, thermal or otherwise. Examples of joined materials are dentine to tooth enamel (Meredith et al., 1996), tendon to bone (Fukuta et al., 1998), spider frame silks to viscid silks (Guerette, 1997), claw or hoof to skin (Kasapi and Gosline, 1997) and hinge ligament to shell (Ono, 1995), to mention a few. The interplay of materials such as these has a significant impact on biological function. Were these materials to be assembled in simple butt-jointed unions, considerable stresses arising from their mismatched properties would lead to structural failure. Molecular dovetailing creates an interfacial region with extensive overlap of the two materials usually mediated by a mutually compatible matrix. In effect, then, this region of overlap is a new material endowed with intermediate properties.

Materials with mismatched properties occur many times in the structure of mussel byssus: (i) at the plaque–substratum interface, where a leathery attachment plaque is bonded to the surface of a stone; (ii) in the stiff thread to plaque fusion; (iii) in the stiff distal to soft proximal thread transition; (iv) at the juncture between thread and stem; and (v) in the attachment between stem and muscle tissue. We are concerned here with the juncture between thread and stem. Given the different chemical compositions (Pujol et al., 1970; Bdolah and Keller, 1976) and mechanical properties (Bell and Gosline, 1996; Rudall, 1955) of the stem and the byssal threads, information about such molecular dovetailing may shed light on the mechanism of load distribution and load transfer between dissimilar biomaterials.

Several full-length clones, approximately 2.8 kb each, of σ-preCol-P cDNA were isolated from a cDNA library constructed from the foot tissue of Mytilus edulis L. as described previously (Coyne et al., 1997). The 5′ and 3′ ends of three clones (designated P22, P33 and P38) that exhibit different restriction patterns were sequenced with the PRISM dye terminator cycle sequencing ready reaction kit (ABI, Foster City, CA, USA) using vector-specific primers as described by Coyne et al. (1997). Approximately 1000 bases were sequenced from each clone.

Total RNA was extracted from Mytilus edulis foot tissue using a modified guanidinium thiocyanate (GTC)–phenol extraction procedure (Chomszynski and Sacchi, 1987). The mussel foot was excised and quickly frozen in liquid N₂ and ground to a fine powder. Nucleic acids were extracted with water-saturated phenol:solution D (4 mol l⁻¹ GTC, 0.5 % N-laurylsarcosine, 25 mmol l⁻¹ sodium citrate, pH 7.0; Chomszynski and Sacchi, 1987) (2:1) and heated to 65 °C. The cellular debris was pelleted, and the supernatant was re-extracted twice with a water-saturated phenol:chloroform:isoamyl alcohol mixture (200:144:6) and finally with an equal volume of chloroform. Total RNA was precipitated in ethanol and resuspended in TE buffer (10 mmol l⁻¹ Tris, pH 7.4, 1 mmol l⁻¹ EDTA).

RNA was reverse-transcribed in a 25 μl reaction mixture containing 100 ng of total RNA, 25 mmol l⁻¹ Tris-HCl (pH 8.4), 3 mmol l⁻¹ MgCl₂, 0.4 mmol l⁻¹ of each dNTP, 5.4 ng μl⁻¹ random hexamers (Life Technologies, Inc., Gaithersburg, MD, USA), 10 mmol l⁻¹ dithiothreitol (DTT) and 200 units of SuperScript II reverse transcriptase (Life Technologies). The mixture was incubated at 25 °C for 10 min, followed by 50 min at 42 °C. PreCol-P cDNA products were diluted to 50 μl. Reverse-transcribed cDNA (2 μl) was amplified by polymerase chain reaction (PCR) in a 20 μl reaction mixture containing 0.2 mmol l⁻¹ dNTPs, 1.25 mmol l⁻¹ MgCl₂, 1× PCR buffer (Promega, Madison, WI, USA), 0.5 units of Taq polymerase and 1 μmol l⁻¹ each of preCol-P-specific primers P33.98F.B (5′-GGAATCAAAGTAGTAC-CCTACCACGG-3′) and RACE.116R (5′-CCACCTCCTA-AACCGTTATG-3′). The PCR reaction consisted of 35 cycles of 30 s at 94 °C, 1 min at 62 °C and 1 min at 72 °C, followed by a 5 min extension at 72 °C. Positive control reactions included as template a mixture of equal amounts of plasmids encoding P22, P33 and P38 cDNAs. The PCR products were fractionated in triplicate on an 8 % polyacrylamide gel containing a gradient of 20 % to 40 % urea to denature heteroduplex products.

PCR products were transferred from the polyacrylamide gel to a positively charged nylon membrane (Boehringer Mannheim, Indianapolis, IN, USA) by horizontal transfer (Muyzer et al., 1993). The membranes were then cut into three pieces and hybridized for 16 h at 64 °C in hybridization buffer, which was made up of 5×SSC (1×SSC is 0.15 mol l⁻¹ NaCl, 0.015 mol l⁻¹ sodium citrate), 1 % blocking reagent (Boehringer Mannheim), 0.1 % N-laurylsarcosine, 0.2 % SDS, 10 % formamide, containing 15 pmol ml⁻¹ of digoxigenin (DIG)-dUTP-tailed oligonucleotide P22.VAR (5′-CCAATA-CCGCTTCCGCCATG-3′), P33.VAR (5′-TGGGCAGATGC-ATGTCTTCC-3′) or P38.VAR (5′-CCAATACCGCCATG-TCTTCC-3′). After washing, the membranes were incubated with alkaline phosphatase (AP)-conjugated anti-DIG Fab fragments (Boehringer Mannheim). Hybridization products were localized by detection of the bound antibody using 5-bromo-4-chloro-3-indolylphosphate (BCIP; 15 mg ml⁻¹) in the presence of Nitroblue Tetrazolium (NBT; 0.3 mg ml⁻¹).

Mytilus edulis feet were excised and quickly frozen on aluminum blocks pre-cooled to —80 °C. Each foot was sectioned laterally into nine equal portions. Total RNA was extracted from each section individually and reverse-transcribed into cDNA, as described above. The presence of amplifiable cDNA from each section was verified by PCR using universal primers specific to eukaryotic rRNA, Euk A and Euk B (Medlin et al., 1988).

PreCol-P variants were amplified in separate PCR reactions, as described above, using 1 μmol l⁻¹ each of the forward primer P33.98F.B and a variant-specific reverse primer, P22.VAR.B (5′-TACCACCAATACCGCTTCCG-3′), P33.VAR or P38.VAR. Positive PCR control reactions included 0.5 ng of plasmid templates encoding P22, P33 or P38 cDNAs. Positive RT control reactions included total RNA extracted from M. edulis foot tissue. Negative RT control reactions included RNA template, but omitted the reverse transcriptase from the RT reaction.

Stem glands of M. edulis were dissected, fixed and embedded in methacrylate as described in Warren et al. (1998). Acetone-de-embedded tissue sections (2 μm thick) were hybridized in individual wells for 16 h at 55 °C in hybridization buffer (2×SSC, 0.01 % Tween 20, 1×Denhardt’s, 0.5 mg ml⁻¹ sheared herring sperm DNA, 0.1 μg ml⁻¹ poly A) containing 15 pmol ml⁻¹ of DIG-labeled oligonucleotides P22.VAR, P33.VAR or P38.VAR. Hybridization products were localized as described by Warren et al. (1998), strategy no. 5. Briefly, hybridized tissue sections were incubated with biotin-SP-conjugated anti-DIG antibodies (Jackson Immunoresearch Laboratories, West Grove, PA, USA), followed by incubation with streptavidin–horseradish peroxidase (SA-HRP; NEN Life Science Products, Boston, MA, USA). The HRP catalyzed the enzymatic deposition of biotinyl-tyramide using the TSA-indirect kit (NEN Life Sciences). Hybridization products were visualized after incubation with SA–fluorescein (SA-FITC; NEN Life Sciences) using a confocal laser scanning microscope (Zeiss LSM 510) equipped with argon and helium–neon ion lasers. The 488 nm line of the argon and 543 nm line of the helium–neon lasers were used as the excitation wavelengths.

Sequencing revealed the presence of non-overlapping, 21 base pair (bp) and 54 bp deletions between positions 170 and 245 in the cDNA sequences of two of the clones analyzed, P22 and P33 (Fig. 1). A third clone from the cDNA library, P38, includes the entire sequence, without either deletion. Since these same deletions are also found within the cDNA sequences of clones generated by rapid amplification of cDNA ends (RACE) (Coyne et al., 1997), it is unlikely that the deletions are a cloning artifact.

The 21 bp and 54 bp deletions are adjacent to each other and appear in the N-terminal histidine-rich domain of the translated amino acid sequence of σ-preCol-P. PreCol-P variant P33 and the full-length sequence, P38, each include a seven-residue amino acid sequence, Phe-Arg-Asn-Gly-Arg-His-Gly, that is absent from the deduced amino acid sequence for variant P22. PreCol-P variant P22 and P38 include a translated amino acid segment missing from variant P33 that is characteristic of the histidine-rich and elastic domains of preCol-P: (Gly-Gly-X) clusters followed by a Ser/His/Ala-rich segment (Fig. 1).

To confirm that these variations are real and to identify other possible variants, RNA isolated from M. edulis foot tissue was subjected to RT-PCR using preCol-P-specific primers that flank the region of interest. PCR amplification resulted in three products, 194, 227 and 248 bp in length, that correspond to the amplification product sizes of preCol-P variants P33, P22 and P38, respectively. Because of the almost identical sequences of the variants, heteroduplexes were formed between the PCR products, making it difficult to verify the number of variants on a non-denaturing agarose gel. For this reason, the PCR products were fractionated on a denaturing gradient polyacrylamide gel. Heteroduplexes that formed during the PCR reaction were denatured and easily separated, while homoduplex PCR products remained intact (Fig. 2). The identity of each PCR product was confirmed by transferring them to a solid membrane and probing the membranes with variant-specific oligonucleotide probes (Fig. 2). Although the sequences of the probes overlapped with each other or with sequences shared by all three variants, comparisons with control PCR reactions indicate that the probes were specific to their respective preCol-P variant and that cross-hybridization was minimal.

We next examined the distribution of the preCol-P variants in the foot tissue. RT-PCR was performed on total RNA extracted from nine sections of Mytilus edulis foot tissue (Fig. 3) using primers specific to each variant. The primers were designed from overlapping segments of the cDNA sequence such that they bridged the deleted portions of the other two variants (Fig. 1). They are identical to the probes used above except for the P22.VAR.B primer. The sequence of this primer is shifted to the 3′ end of the sequence by five bases with respect to probe P22.VAR to minimize nonspecific amplification.

The results indicate a graded distribution of preCol-P transcripts for variants P33 and P38, with the greatest concentration at the base of the foot (Fig. 3A, top and middle panels). This distribution is typical of both the protein preCol-P and its RNA transcripts in the foot tissue. Although we did not attempt to quantify our results, visual inspection of Fig. 3A, top and middle panels, suggests that variant P33 may be expressed in greater amounts than P38. The greater amplification for variant P33 may have resulted from PCR bias because of differences in base composition. A comparison of positive control amplification products, however, demonstrates that differences in amplification due to PCR bias are minimal. RT-PCR products with variant-P22-specific primers reveal that this variant of preCol-P is expressed strongly, but only at the very base of the foot (Fig. 3A, bottom panel). Fig. 3B confirms the presence of amplifiable cDNA from each foot section.

The strong expression of preCol-P variant P22 at the base of the foot suggests that the P22 variant may be a component of the byssus stem. In situ hybridization experiments with catalyzed reporter deposition (CARD) amplification confirmed that preCol-P variant P22 is expressed in the musculo-glandular region at the base of the stem (Fig. 4A). For orientation, Fig. 4A–D shows the parallel ribbons of the tissue septae protruding from the base of the stem generator in the musculo-glandular region. The tissue septae are oriented parallel to the mid-sagittal plane of the foot. The red propidium iodide counterstain (specific for nucleic acids) is clearly visible in the septae and at the base of the stem generator. The green fluorescent labeling located primarily in the musculo-glandular region at the base of the stem generator of sections probed with P22-specific probes positively identified the presence of the preCol-P variant P22 transcripts within the stem gland (Fig. 4A). Hybridization experiments performed on sections from the same tissue block with oligonucleotide probes specific to variants P33 and P38 indicate that these variants are expressed at either very low levels or not at all in the stem gland (Fig. 4B,C). Some signal from transcripts of variant P22 with probes P33.VAR and P38.VAR may be expected because of overlap of the probe sequences with the sequence of P22. The no-probe control demonstrates negligible background staining (Fig. 4D).

A protein gradient of a preCol-P-derived fragment (Col-P) was demonstrated in byssal threads and culminated with highest concentrations at the proximal end of the thread (Qin and Waite, 1995). The graded expression of both preCol-P and its corresponding mRNA in the foot confirmed the gradient concept and suggested that gradients are established in the collagen gland prior to secretion (Qin and Waite, 1998; Qin et al., 1997). It is tempting to attribute the extensibility and flexibility of the proximal portion of byssal threads to the elastin-like sequences flanking the collagen domain in each σ-chain of preCol-P (Coyne et al., 1997). The validity of this model, however, depends critically upon the macromolecular assembly of preCols, about which little is known at present. The picture is further complicated by the detection of two additional variants, P22 and P33, in this study. These variants may be generated by alternative splicing of the full-length transcript, P38, which corresponds to the original sequence reported by Coyne et al. (1997). In all likelihood, the variants have a functional significance that could be manifested in different ways, including the formation of preCol-P heterotrimers or tissue specific localization.

P33 and P38 both have a similar graded distribution in foot sections as revealed by RT-PCR analysis. Thus, these are likely to be co-occurring variants, although whether as heterotrimers or as different homotrimers cannot be deduced from the present data. Variant P33 lacks an 18-residue sequence that in P38 encodes two palindromic peptides [SSHAHAHSS and GGIGGIGGG]. The histidine-rich sequence is believed to be a site where metal-mediated intermolecular contacts occur (Qin and Waite, 1998). The other sequence is elastin-like in its content of Gly and Ile. The consequences of reducing the metal-binding and elastic functionalities in this variant are intriguing but not known.

In contrast, variant P22 is a clear case of tissue-specific expression in the musculo-glandular region of the stem generator at the base of the foot (Tamarin, 1976). Given this, one is inclined to expect a dramatically different sequence in P22, but this is not the case. P22 lacks only the sequence FRNGRHG present in P38 between residues 55 and 61. The absence of any similarity to known structural proteins suggests that this may be a binding sequence for another protein in the proximal thread. That the stem should require specific structural proteins is less surprising than the essentially preCol-P-like sequence of variant P-22. The amino acid composition of the stem is strikingly different from that of the threads. In particular, with Gly and Hyp levels reduced by more than 50 % in the stem (Pujol et al., 1970; Bdolah and Keller, 1976), a much smaller role for collagen-containing proteins is suggested. Moreover, a β-keratin-like structure was deduced from low-angle X-ray diffractograms of Mytilus stem (Rudall, 1955). In the same study, Rudall reported an ultimate strain of 0.9 for stem. This is less than half the strain of 2–3 reported for proximal threads (Smeathers and Vincent, 1979; Bell and Gosline, 1996) and reflects a less extensible material.

Ostensibly, the mechanical properties of the proximal thread and stem are poorly matched, and some kind of dovetailing is required to relax discontinuities between the materials. This is morphologically evident. The byssal stem is a large complicated structure that consists of a laminated core and a sheath of overlapping rings or cuffs (Fig. 5A). The core is rooted to the byssal retractor muscles and formed from precursors packed into laminar sheets that are concentrically molded to one another and extruded from the stem gland. As the laminae progressively emerge from the stem gland, they become recognizable as ring structures after being coated by an accessory gland secretion (Brown, 1952). It is to these ring structures that the proximal portions of the threads are fused. Brown (1952) scrutinized fiber anisotropy in the rings and threads using polarized light and concluded that the rings contain two fiber orientations that are perpendicular to one another. In the distal portion of each ring, fibers run parallel to the ring edge. Perpendicular to these is a second set belonging to the root laminae. Both contribute to fusion with the proximal thread (Fig. 5B). What has not been clear is how far these fibers extend into the proximal thread and from which precursors they are derived.

The present results suggest that the stem generator produces a stem-specific variant of preCol-P. Given its near-identity with preCol-P from the proximal portion of the byssal thread, one must assume that it serves as a molecular ‘dovetail’ between the thread and the stem, and that, like its thread counterparts, it will impart extensibility and elastic recoil to the ring portion of the stem. Extensibility in the rings would allow these structures to absorb a large amount of energy, while the elastic recoil would tend to impose a self-righting orientation on mussels that are disoriented by turbulent water flow (Dolmer and Svane, 1994). In other words, when mussels attach, they generally select a shell orientation that minimizes drag to the prevailing ebb and flow. When flows become unpredictable, mussels are often rotated about the stem axis, thereby creating a torsional strain that will be self-righting by elastic recoil upon relaxation.

Extensive engineering studies have established that any pair of macro-or microscopically bonded materials with mismatched mechanical, thermal or electronic properties will fail unless their ability to relax interfacial stresses is enhanced (Rabin et al., 1995). A common strategy for reducing interfacial stresses involves molecular dovetailing or functional gradients. This strategy is also evident in biological structures, although few are understood in any molecular detail. The mechanical differences between the highly extensible proximal portion of byssal threads and the stiffer stem are greatest at the rings. Here, the stem generator secretes a preCol-P variant to dovetail with the thread. We anticipate that other dovetailing elements are incorporated into the threads to interact specifically with the tensile elements of the stem.

Support for this research was provided from grant 2 RO1 DE10042-07A1 from the National Institute of Dental and Craniofacial Research to J.H.W.