doi:10.1006/jmbi.2001.5298availableonlineathttp://www.idealibrary.comon J. Mol. Biol. (2002) 316, 127±137

Tissue-Specific Co-expression and in vitro Heteropolymer Formation of the Two Small Branchiostoma Proteins A3 and B2

AntonKarabinos1,JuÈrgenSchuÈnemann1,DavidA.D.Parry2 andKlausWeber1*

1Max Planck Institute for The two small intermediate ®lament (IF) proteins A3 and B2 of the Biophysical Chemistry cephalochordate Amphioxus were investigated. Blot overlays indicated a Department of heterotypic interaction pattern of the recombinant proteins. While the Am Fassberg 11 individual proteins formed only aggregates, the stoichiometric mixture 37077 Goettingen, formed obligatory heteropolymeric ®laments. Mutant proteins with a single cysteine residue in equivalent positions gave rise to ®laments that 2Institute of Fundamental oxidize to the disul®de-linked heterodimer, which can again form IF. Sciences, Massey University Thus the A3/B2 ®laments, which are expressed in the intestinal Palmerston North, New epithelium, are based on a hetero coiled coil. This keratin-like assembly Zealand process of A3 plus B2 was unexpected, since previous evolutionary tree calculations performed by two laboratories on the various Amphioxus IF proteins identi®ed keratin I and II orthologs but left the A/B group as a separate branch. We discuss obvious evolutionary aspects of the Amphioxus IF multigene family, including the previously made obser- vation that B1, the closest relative of B2, forms homopolymeric IF in vitro and is, like vertebrate type III proteins, expressed in mesodermally derived tissues. # 2002 Elsevier Science Ltd. Keywords: Amphioxus; Branchiostoma; cephalochordate; coiled coil; *Corresponding author developmental expression

Introduction IV spans ®ve neuronal IF proteins. Two large mol- ecular mass proteins (synemin and paranemin), and the two IF proteins of the eye lens (phakinin A variety of epidermal fragility syndromes are and ®lensin) fall outside these subfamilies. Finally, caused by mutations in human keratin genes, type V covers the nuclear lamins. suggesting that an important function of cyto- plasmic intermediate ®laments (IF) is to provide Lamins differ from vertebrate cytoplasmic IF resistanceagainstmechanicalstress.1Mostofthe proteins by an extra 42 residues in the coil 1b sub- close to 60 members of the mammalian multigene domain of the central rod domain and by unique family of IF proteins fall into one of the ®ve major carboxy-terminal domains that harbor a nuclear subgroups.2,3TypeIandtypeIIkeratinsarethe localisation signal and end, in most cases, with a largest subfamilies and give rise to the epithelial CaaXbox.4,5Theshortcoil1bdomain,®rstde®ned keratin ®laments that are based on obligatory het- in vertebrate cytoplasmic IF proteins, extends to eromeric double-stranded coiled coils formed by a thecephalochordatesandurochordates,6±9while type I and a type II keratin. Type III covers four cytoplasmic IF proteins from the protostomia show mesenchymally expressed proteins, which, at least the long coil 1b version of the nuclear lamins and, in part, are able to form homopolymeric IF. Type in most cases, display a lamin homology segment intheirtaildomains.10,11 Recent studies on the lancelet, the cephalochor- Abbreviations used: IF, intermediate ®lament; PVDF, 6,12±14 poly(vinylidene ¯uoride). dateBranchiostoma(Amphioxus), andthe E-mail address of the corresponding author: urochordateStyela8,9documentthattheearlychor- [email protected] dates have among their cytoplasmic IF proteins

0022-2836/02/010127±11 $35.00/0 # 2002 Elsevier Science Ltd. 128 Novel Heterotypic IF System in Amphioxus orthologs of the type I to III proteins originally sion numbers of the revised sequences are de®ned in the vertebrates. AJ223575 and AJ223576). The molecular masses We previously cloned 13 cytoplasmic IF proteins calculated from the revised sequences are in good from Amphioxus. Evolutionary tree calculations agreement with those obtained experimentally. indicated three type I and two type II keratins. ThealignmentinFigure1showsthatIFproteins This assignment was con®rmed by the obligatory A3 and B2 essentially lack a C-terminal tail heteropolymeric ®lament formation of the recombi- domain. Only two (B2) and four (A3) residues fol- nant proteins. Any stoichiometric mixture of type I low the well-conserved end of the rod domain. (k1, Y1 and E1) and type II (D1 and E2) proteins Unusually short tail domains are known for the provided IF. In addition, two of the lancelet type I Amphioxus keratin E1 and the human keratin 19 keratins formed chimeric IF when mixed with (6,12andreferencestherein). humankeratin8,atypeIIkeratin.6Threekeratins (k1, Y1 and D1) and protein X1 are the only IF pro- Blot overlays reveal heterotypic interactions teins expressed in the gastrula, which is reminis- between A3 and B2 cent of the expression of two type I keratins (K18 and K19) and one type II keratin (K8) in the mur- The ten currently available recombinant IF pro- inegastrula.15ThenumberoflanceletIFproteins teins of the lancelet were analysed in blot overlays increases at the neurula and early larval stages to using biotinylated A3 and B2 proteins. After exten- seven and 11, respectively. However, 13 different sive washing steps, the biotinylated proteins were proteins are found in the adult. The keratins are detected by the streptavidin-peroxidase system. the major IF proteins in the Amphioxus nerve Figure2(middlepanel)showsthatbiotinylatedA3 cord. Proteins X1 and C2 were found, by immu- binds strongly to B2. Except for a very weak reac- noelectron microscopy, to be integrated into the tion on B1, no other IF protein was detected by the epidermal and nerve cord IF, which are built from A3 probe. The biotinylated B2 protein reacted keratins. The latter two proteins and the C1 protein strongly with the A3 protein and more weakly seemtobelancelet-speci®cIFproteins.13The®ve withtheproteinB1(Figure2,lowerpanel).These remaining IF proteins form the A/B group in the results show a strong heterotypic interaction evolutionarytreepreviouslycalculated(Figure4of between A3 and B2, and indicate a lack of homo- Karabinosetal.6).Sofar,onlytheB1proteinhas typic interactions, i.e. A3 with A3 or B2 with B2, in been characterized. It can form homopolymeric IF the overlay system. These observations raised the in vitro and is expressed in the mesodermally question of whether A3 and B2 could form hetero- derived muscle tails and in coelomic epithelia. polymeric IF. These properties indicate some type III character 12,13 fortheB1protein. Here,weshowthatA3and Equimolar amounts of A3 and B2 form B2, which are co-expressed in the intestinal epi- heteropolymeric IF thelial cells, form obligatory heteropolymeric IF based on a hetero coiled coil consisting of one A3 Puri®ed recombinant proteins A3 and B2 were and one B2 polypeptide chain. dialysed either alone or in equimolar mixture to remove the urea, and the self-assembly products were analysed by electron microscopy after nega- Results tive staining. While the individual proteins yielded Branchiostoma IF proteins A3 and B2 only aggregated material, the mixture of A3 and essentially lack a tail domain B2formedlongIF(Figure3).Thus,A3andB2lack the ability to form homopolymeric IF but partici- Cloning and recombinant expression of A3 and pate in obligatory heteropolymeric IF assembly. B2havebeendescribed.6,7Duringthecurrent study we noted that the apparent molecular The structural unit of A3/B2 filaments is masses of both proteins observed by SDS-PAGE the heterodimer (A3, about 37 kDa; B2, about 39 kDa) were smaller than the masses calculated from the protein To decide whether the A3/B2 ®laments are sequences predicted from the cDNAs (A3, 41,679 based on a A3/B2 heterodimer or the A3 and B2 Da; B2, 49,326 Da). Automated sequencing of the homodimers, we used the approach previously proteins blotted on to a poly(vinylidene ¯uoride) taken to demonstrate the heterodimeric nature of (PVDF) membrane provided the predicted N-term- keratin®laments.9,16Itisbasedontheobservation inal sequences. The experimentally inferred masses that in parallel and unstaggered a helices of a derived by mass spectrometry (A3, 37,278 Da; B2, double-stranded coiled coil, a cysteine residue pre- 39,694 Da) were lower than the values calculated sent in an a or d position of the heptads can form a from the predicted sequences. We therefore sub- disul®de bond crosslink, provided both a helices jected the coding fragments to a detailed sequence have the cysteine residue in equivalent positions. analysisandfoundthattheoriginalanalysis7had Thus we aimed at mutants of A3 and B2 that have overlooked an earlier stop codon. Thus, the open a single cysteine residue, which is located in the reading frames of A3 and B2 were shortened by 39 same position as in the previously studied ker- and 92 codons, respectively (the GenBank acces- atins.9,16Thispositionisexactly40residuesfrom Novel Heterotypic IF System in Amphioxus 129

Figure 1. Alignment of the amino acid sequences of the ®ve related Branchiostoma IF proteins A1, A2, A3, B1 and B2. The typical structural organisation consisting of head, rod and tail domains is indicated. Arrowheads pointing down or up mark the start and end of the four subdomains of the helical rod domain (coils 1a, b, 2a and 2b), which are connected by three non-helical linkers (L1, L12 and L2). Asterisks (*) between the rod domains of A3 and B2 mark the a and d positions of the heptad repeat pattern with the stutter in coil 2b (for details on IF structure, see Fuchs&Weber2andParry&Steinert3).Dashesareusedtooptimizethesequencealignment.Boldlettershighlight similarities in head or tail domains and indicate residues in the rod domains that are identical in all ®ve proteins. Sequences of two synthetic peptides used for production are boxed. The capital letters X and Z shown above the sequences indicate positions that were changed by site-speci®c mutagenesis. The cysteine codons TGC of the A3 and B2 sequences were changed to the serine codons AGC (marked X). The methionine codons ATG of the A3 and B2 Cys mutants were changed to the cysteine codons TGC (see Results for details). The resulting mutant A3 and B2 proteins contained a single cysteine residue (position Z). The GenBank accession numbers of the A3 and B2 sequences are AJ223575 and AJ223576, respectively.

the end of the rod and involves a heptad d position cing from a protein blot onto a membrane. (Figure1,markedZ).Inthe®rstroundofsite- Figure5(b)showsthatthedisul®de-linkedA3/B2 speci®c mutagenesis, we changed the single heterodimer forms very long IF in the absence of cysteine residue of A3 and B2 (marked X in reducing agent. Thus, the A3/B2 ®laments are Figure1)toserine.Inthesecondstep,themethion- built from double-stranded hetero coiled coils. This ineresiduemarkedZinFigure1waschangedto result is consistent with the strong heterotypic cysteine. The mutations were con®rmed by DNA interaction pattern of A3 and B2 proteins in the sequencing and the mutant recombinant proteins blotoverlays(Figure2). were puri®ed. Filaments were formed by the A3 and B2 mutant proteins in the presence of 2-mercaptoethanol. Sub- Tissue-specific co-expression of B2 and A sequent extensive dialysis removed the reducing proteins in the adult and the early larvae agent and allowed cystine formation. One and To determine expression patterns of the A3 and two-dimensional gels run under non reducing and B2 proteins, we raised two peptide , reducingconditions(Figure4)identi®edasmajor whichwereantigenaf®nity-puri®ed.Figure6 components the A3/B2 disul®de-linked heterodi- shows the results of immunoblotting on ten recom- mer as well as the A3 and B2 monomers. Dimers binant lancelet IF proteins. The B2 antibody recog- ofA3andB2werenotobserved(Figure4(b)).The nized only B2, while A3 antibody detected only oxidized ®laments were dissolved in 8 M urea A3. The A3 antibody, however, detects also A1 without reducing agent. Ion-exchange chromatog- (Figure6(b))andA2(datanotshown).Thiswas raphy on Mono Q using the same solvent yielded expected from the high level of sequence homology the puri®ed heterodimer with a small amount of of three A proteins. For instance, the sequences of contaminatingmonomericB2(Figure5(a)).The A1 and A2 corresponding to the A3 peptide used assignment of the A3/B2 heterodimer and the B2 as antigen differ only by a single valine to isoleu- contaminant was veri®ed by automated sequen- cineexchange(Figure1).Wethereforerefertothis 130 Novel Heterotypic IF System in Amphioxus

Figure 3. Electron micrograph of Branchiostoma IF assembled in vitro from equal amounts of recombinant proteins A3 and B2. The scale bar represents 0.2 mm.

ing myomers and the epidermis (marked with arrowsinFigure7(e)-(h)).Intheearly(onetotwo days) B. ¯oridae larva, the intestinal cells were the only structures stained by the two antibodies (Figure8).

Calculation of interchain ionic interaction between the A and B proteins Figure 2. Blot overlay search for interactions of the The ionic interactions that could occur in prin- biotin-labeled A3 and B2 proteins with blotted immobi- ciple between the chains that constitute a coiled lized Branchiostoma IF proteins. Equal amounts of puri- coilmoleculearereadilycalculated.18,19Suchinter- ®ed recombinant lancelet proteins A3, B1, B2, C2, D1, actions are known to play an important role in spe- E1, E2, k1, Y1 and X1 were separated by SDS/10 % PAGE and stained with Coomassie brilliant blue (upper panel) or transferred to nitrocellulose membranes. The middle and lower panels show the blots after incubation and staining with biotinylated A3 and B2 proteins, respectively (marked as A3-biotin and B2-biotin on the left). The biotinylated A3 (middle panel) strongly dec- orates the B2 protein and a proteolyzed derivative. There is a very weak reaction on B1 (marked by the asterisk (*)) but all other proteins were negative. The biotinylated B2 protein (lower panel) recognizes the pro- tein A3, and B1 plus some aggregate of it. All other pro- teins were negative with biotinylated B2. Marker proteins (M) and an approximate molecular mass stan- Figure 4. Identi®cation of cysteine-linked heterodi- dard (in kDa) are given at the right of the upper panel. mers of mutant recombinant A3 and B2 proteins. Fila- ments produced in the presence of 2-mercaptoethanol in ®lament buffer were dialysed extensively against the same buffer in the absence of reducing agent to allow oxidation. Filaments were harvested by centrifugation. antibody as the A-speci®c antibody. Used in An aliquot was solubilized in 8 M urea and analysed by non-reducing and reducing SDS-PAGE (lanes R and NR immuno¯uorescence microscopy on frozen sections in (a)) and by two-dimensional gel-electrophoresis of adult Branchiostoma ¯oridae, the antibodies under non-reducing conditions in the ®rst dimension demonstrated co-expression of A and B2 proteins and (b) non-reducing or (c) reducing conditions in the intheintestine(Figure7(a)-(d)).Thestainingwas second dimension. Positions of the A3 monomer, the B2 mostly concentrated into the apical part of the monomer and the A3/B2 cross-linked dimer are marked simplecolumnarepitheliumoftheintestine.17In on the right as A3, B2 and A3/B2, respectively. Note addition, both antibodies provided a regular stain- the high yield of the disul®de-linked heterodimer A3/ ing of small areas surrounded by the two connect- B2 and the absence of homodimers of A3 and B2. Novel Heterotypic IF System in Amphioxus 131

Figure 5. Puri®cation of disul®de-linked A3/B2 het- erodimer and its ability to form IF in vitro. The disul- ®de-linked heterodimer was puri®ed from the harvested oxidized®laments(seeFigure4).Filamentsweredis- solved in 8 M urea buffer without reducing agent and the protein was subjected to ion-exchange chromatog- raphy on Mono Q, which removed the A3 and B2 monomers. An aliquot of the Mono Q-puri®ed heterodi- mer was analysed by non-reducing and reducing SDS- PAGE (lanes R and NR in (a), respectively) and was found to contain a small amount of monomeric B2. The nature of all polypeptides was con®rmed by automated sequencing after blotting onto a PVDF membrane. Mar- ker proteins (M) and an approximate molecular mass standard (in kDa) are given at the left. (b) The ability of the disul®de-linked A3/B2 heterodimer to form long IF in the absence of reducing agent. The scale bar rep- resents 0.2 mm.

Figure 6. Immunoblot analysis of Branchiostoma cifying chain assembly and, in particular, the rela- recombinant proteins. (a) Equal amounts of puri®ed tive chain stagger and polarity. For IF proteins, the proteins A3, B1, B3, C2, D1, E1, E2, k1, Y1 and X1 were preferred calculated arrangement has been without separated by SDS/10 % PAGE and stained with Coo- exception for the chains to lie parallel with one massie brilliant blue (upper panel) or transferred to another and in axial register. Experimental obser- nitrocellulose membranes. Blots were incubated with vationshavecon®rmedthesecalculations.3Inthe af®nity-puri®ed antibodies to Branchiostoma A3 and B2 case of Branchiostoma, the interchain ionic inter- proteins as indicated on the left. Both antibodies recog- actions have been determined for a total of 15 nised exclusively the corresponding protein and no cross-reactivity with other IF proteins was observed. chain combinations involving three members of the Marker proteins (M) and an approximate molecular A family and the two members of the B family. In mass standard (in kDa) are given at the left of the all cases, the preferred arrangement is for the upper panel. (b) Bacterial lysates containing equal chains to lie parallel with one another and to be in amounts of the recombinant A1 and A3 GST-fusion pro- axial register. All of the A/B heterodimers score teins (marked by asterisk (*)) were separated by SDS/ more highly than the A/A homodimers and A/A 10 % PAGE and stained with Coomassie brilliant blue heterodimers(Table1).Thisisnotsurprising,since (left panel) or immunoblotted with af®nity-puri®ed anti- keratin heterodimers are the norm. However, the body A3. The antibody strongly stained the A1 and the (joint) highest calculated scores are for the B1 A3 GST-fusion protein bands (right panel). Marker pro- teins (M) and an approximate molecular mass standard homodimer and the B1/B2 heterodimer (score (in kDa) are given at the left. ‡16).ThisisconsistentwiththeobservationofB1 homodimers12butsuggests,intheory,thatB1/B2 heterodimers might also be viable. As regards A3, its preferred partners are B1 and B2 (score ‡14). Overall, there is good but not perfect correlation Designation of chain type of the 13 between the experimental observations and the Branchiostoma IF proteins theoretical calculations. However, X-ray crystallo- graphic studies on synthetic polypeptides that Each of the 13 Branchiostoma chains (A1, A2, adopt a coiled coil conformation indicate that only A3, B1, B2, C1, C2, D1, E1, E2, k1, X1 and Y1) was some of the possible interchain ionic interactions compared with the consensus sequences unique to are actually made in vivo. The calculations must thetypeIandtypeIIkeratinchains(Figure9;20 thus be seen as indicative rather than de®nitive. but updated in line with the additional data now 132 Novel Heterotypic IF System in Amphioxus

Table 1. Calculation of interchain ionic interactions of individual rod homo- and heterodimers lying parallel and in axial register

IF chain A1 A2 A3 B1 B2 A1 8 7 9 13 12 A2 7 10 8 12 12 A3 9 8 10 14 14 B1 13 12 14 16 16 B2 12 12 14 16 12 The numbers show potential ionic interactions calculated for charged residue pairs in positions 2e0-1g, 1;g0-2e, 2a0-1g, 1g0-2a, 1e0-1d,and1d0-1eofthetwochains.Attractiveinteractionsbetweenoppositelychargedresidueswerescored‡1andinteractions betweenresiduesthatareeitherbothpositivelyorbothnegativelychargedwerescored1.Allotherinteractionswerescored0.21

available).Ineightcases,theanalysis(Table2)was however, were more dif®cult to type unequivocally such that the chain type became immediately (A3 and C1). For the A3 chain, the data pertaining apparent (type I, B1, B2, C2, E1, k1 and Y1; type II, to the entire sequence slightly favoured type I D1 and E2). In three instances, the designation of designation. However, the analysis relating to the chain type, while less de®nitive, clearly favoured residues in the non-a and non-d positions ®rmly one option (type II, A1, A2 and X1). Two chains, favoured the chain being type II (see Materials and Methods). For this reason, A3 was speci®ed a type II chain. The situation was even more dif®cult for the C1 chain, where the data only slightly favoured type II designation on the basis of the consensus sequence in all positions of the heptad repeat but proved effectively indeterminate in selecting chain type using the non-a and non-d position data. The decision to designate C1 a type II chain on the basis of sequence homology was thus a marginal one.

Evolution of the Branchiostoma IF proteins Bootstrap maximum parsimony analysis of the rod sequences from all 13 Branchiostoma IF pro- teins, various IF proteins representing vertebrate

Figure 7. Tissue-speci®c co-expression of A and B2 IF proteins in adult Branchiostoma. Frozen cross-sections were stained with Hoechst 33258 (DNA staining; (a), (c), (e) and (g)) to facilitate the identi®cation of cells and tis- Figure 8. Tissue-speci®c co-expression of A and B2 IF sues. Indirect immuno¯uorescence microscopy of paral- proteins in the Branchiostoma early larva. Frozen longi- lel sections labeled with anti-A and anti-B2 antibodies as tudinal sections were labeled with (b) A and (d) B2 anti- indicated on the right. (b) and (d) The terminal web bodies and (a) and (c) simultaneously with Hoechst (tw) of the intestinal cells (int) is stained by both anti- 33258 as indicated. (b) and (d) Expression of A and B2 bodies. Arrows in (f) and (h) indicate the area between proteins in the terminal web (tw) of the intestinal simple two myomeres (m) and the epidermis (ep), which shows epithelia (int) is indicated. The epidermis (ep) and other signi®cant expression of both proteins. The scale bar body parts were negative. The scale bar represents represents 20 mm. 20 mm. Novel Heterotypic IF System in Amphioxus 133

Figure 9. Consensus sequences using the one-letter code for type I and type II keratin intermediate ®lament chains for rod domain seg- ments 1a, 1b, 2a, L2 and 2b. These are based on the earlier work by Conway&Parry20butareupdated in line with the additional data now available. Where there is no clear consensus or where three or more different types of residue are seen in a particular position, then that position is indicated by a dot. The symbol X implies that the same consensus residue is observed in both the type I and type II chains, and as it is not a dis- tinguishing feature of either chain type it plays no further part in des- ignation of chain type.

type I to type IV and several additional invert- We6,12andLuke&Holland14reportedsimilar ebrate IF proteins resulted in two equally parsimo- relationships within the Branchiostoma sequences, niousphylograms(Figure10(a)).Themajority-rule and between Branchiostoma and vertebrate consensus of these trees clearly demonstrates the sequences using the unweighted pair-group meth- presence of the type I (E1, k1 and Y1) and type II od with arithmetic mean (UPGMA) and the boot- (D1 and E2) keratin orthologues in Branchiostoma. strap neighbor-joining method, respectively. Thus, This supports a direct common ancestry of all the A, B, C and X1 molecules seem to have no these sequences and the vertebrate keratins from orthologueineithervertebratesorurochordates.9 archetypal type I and type II genes. On the other This raises the possibility that they evolved only hand, the A and B genes, and the C and X1 genes after the divergence of the cephalochordate lineage. of Branchiostoma appear on the separate A/B and On the other hand, since these sequences possess C/X branches, of which only the former had sometypeIortypeIIcharacter(Table2),aposs- strongbootstrapsupport(99%;Figure10(a)). ible common ancestry with the ®ve lancelet kera-

Table 2. Theoretical designation of chain type in the 13 Branchiostoma IF proteins

Accession Non-aand Calculated IF Chain Species numberAlla non-da chain type Evidence from experiments A1 B. floridae AJ223574 1.8 3.9 II Probably like A3 A2 B. floridae AJ223582 3.3 5.3 II Probably like A3 A3B.floridaeAJ223575‡0.74IIHeteropolymericIFwithB2(thisstudy) B1B.lanceolatumX64522‡12.7‡7.9IFormshomopoly-mericIFb B2 B. floridae AJ223576 ‡6.8 ‡6.5 I Heteropolymeric IF with A3 (this study) C1 B. floridae AJ223577 1.9 ‡0.1 II Not determined C2B.lanceolatumAJ223578‡5.1‡5.8IPresentinkeratinIFc D1B.lanceolatumAJ22357919.712.8IITypeIId E1B.lanceolatumAJ010294‡20.7‡16.9ITypeId E2B.lanceolatumAJ01029315.713.6IITypeIId k1B.lanceolatumAJ245426‡12‡3.8ITypeId X1B.lanceolatumAJ2454271.65.2IIPresentinkeratinIFc Y1B.lanceolatumAJ245428‡10.2‡2.4ITypeId a The numbers of identities between the Branchiostoma chains and the type I and type II consensus sequences were calculated both for the entire sequence (All) and for those residues falling in the non-a and non-d positions of the heptad repeat (Non-a and non-d). b SeeinKarabinosetal.12 c Immunoelectromicroscopicalresults(Karabinosetal.6 d Obligatoryheteropolymeric®lamentassembly,E1andY1formIFwithhumankeratin8,atypeIIkeratin(Karabinosetal.13). Note that there is only one disagreement between the calculated chain type and the experimental evidence. B1 although calculated as type I chain is able to form homopolymeric IF in vivo (see the text). 134 Novel Heterotypic IF System in Amphioxus

Figure 10. (a) Phylogenetic analysis of Branchiostoma and other IF rod sequences from invertebrates and ver- tebrates.Majority-ruleconsensusphylogramofweightedbootstrap(1000replications)maximumparsimonyanalysis22 of individual IF sequences. The bootstrap values are shown above the internal nodes. The ®ve subfamilies I-V of ver- tebrate IF proteins, the A/B and C/X subfamilies of Amphioxus IF proteins as well as the protostomic IF branch are indicated. (b) Hypothetical scheme for the evolution of the 13 Branchiostoma IF genes. The indicated relation of C and X1 to the type II gene cluster is interpreted as provisional result (see Results). This scheme suggests that the moreorlesssigni®cantkeratincharactersoftheAandBgenes(seeaboveandTable2)hasprobablybeentheresult of an independent convergent evolution (see the text). tins is indicated. In addition, immunoelectron references therein). Using phylogenetic tree analysis microscopy on tissue sections shows that C2 and we previously showed that ®ve of the 13 lancelet X1occuralongIFbuiltfromkeratinorthologues.13 cytoplasmic IF proteins are orthologues of ver- They may therefore re¯ect some kind of keratin- tebrate keratins. Three type I (E1, k1 and Y1) and related or keratin-derived molecules that have two type II (D1 and E2) keratin orthologues clus- evolved more rapidly than the keratin orthologues. tered with the corresponding human or vertebrate Based on these considerations, we propose a typeIandtypeIIkeratins.6,12,14Moreimportantly, hypothetical scheme of the evolution of IF protein the corresponding ®ve recombinant proteins inBranchiostoma(Figure10(b)).Thesinglearche- showed the obligatory heteropolymeric IF assembly typal ancestor of both keratin groups (typeI/typeII based on a hetero coiled coil containing a keratin gene) was duplicated into typeI/typeII and A/B type I and a keratin type II polypeptide. Indepen- genes. This was followed by additional rounds of dent evidence for this assignment comes from the gene duplications leading to the three type I genes chimeric ®lament formation observed for two lance- (E1, k1 and Y1), the ®ve type II genes (D1, E2, C1, let type I keratins with human keratin 8, a type II C2 and X1), the three A genes (A1, A2 and A3) keratin. Just as with mammalian keratins, any lan- and the two B genes (B1 and B2). The scheme, celet type I protein formed IF when mixed with a which is in line with the present and the previous lancelettypeIIprotein.6,12Keratinexpressioninthe phylogenetic analyses (see above), indicates that lancelet occurs in a tissue-speci®c manner in epider- the more or less signi®cant keratin character of mal and internal epithelia and, in contrast to ver- theAandBchains(seeaboveandTable2)was 6,13 probably the result of independent convergent tebrates,inthenervecord. evolution. Here, we identi®ed a second obligatory hetero- polymeric IF system of Branchiostoma. It is based on A3 as a representative of the three closely Discussion related A proteins (sequence identity about 90 %) and the B2 protein. Recombinant A3 plus B2 pro- The cephalochordate Branchiostoma (Amphioxus) teins form long, obligatory heteropolymeric IF, is thought to be a key organism for understanding while each individual protein yielded only aggre- theoriginandevolutionofthechordates(14and gated material. Mutant proteins with a single Novel Heterotypic IF System in Amphioxus 135 cysteine residue located to the same heptad d pos- tion of the two types of bona ®de keratins in ition in coil 2b formed IF, which oxidized readily. Amphioxus and (b) that the gene duplication event Gel electrophoresis under non-reducing conditions of the B-gene ancestor may have been followed by identi®ed the disul®de-containing heterodimer the rapid acquisition of functional differences A3/B2, and the two monomers A3 and B2. No within a remaining context of similar B1 and B2 crosslinked homodimers were observed. Since the sequences. Finally, we note, that the current puri®ed A3/B2 heterodimer with the cystine cross- (Figure10(a))andallpreviousphylogenetictrees link formed readily IF the A3/B2 ®laments are (see above) indicated the lack of a sequence ortho- based on a double-stranded hetero coiled coil. The logue of vertebrate type III IF proteins within the A3/B2 keratin-like IF differ from the true keratin currently known Branchiostoma IF proteins. Such ®laments described above by a strict preference at proteinsare,however,presentinurochordates,8,9 the dimeric stage, which is seen in gel overlay and can therefore be expected for cephalochor- assays. While normal keratins show a broad prom- dates. iscuity,inthatanytypeIbindstoanytypeIIpro- tein(formammaliankeratinsseeFuchs&Weber,2 Parry&Steinert3andHatzfeld&Franke16)theA3 Materials and Methods probe detected only B2 and B1, and the B2 probe Animal specimens detected only A3 in a collection of ten different lan- celetIFproteins(Figure2;foraweakreactivityof The collection of B. ¯oridae adult and early larvae was A3 on B1, see below). This strong preference at the asdescribed.13Specimenswerefrozeninliquidnitrogen  dimer level is indicated also by the tissue-speci®c and kept at 80 C. expression patterns. B2 and A proteins are expressedintheintestinalepithelium(Figures7 Nucleic acid and protein techniques and 8), in which no normal keratin has been detected. Site-directed mutagenesis of the B. ¯oridae cDNAs for In addition to the heteropolymeric assembly proteins A3 and B2, cloned into the pET23a expression vector,12wasperformedusingtheQuiaChangemutagen- pathways of true keratin orthologues and the esis kit (Stratagene). newly de®ned keratin-like A/B2 molecules, there Expression of the rod domains of A1 and A3 as GST is a third IF assembly mechanism in Branchio- fusionproteinswereasdescribed.7Expression,puri®- stoma. IF protein B1, which shares 54 % identity cation and in vitro ®lament assembly of recombinant A3 intheroddomainwithB2,7istheonlylancelet andB2proteinsweredoneasdescribed.6The®lament proteinabletoformhomopolymericIFinvitro.12 buffer was 10 mM Tris-HCl (pH 7.5). In vivo, it is the only IF protein restricted to meso- Biotinylation of recombinant proteins was achieved dermallyderivedtissues13andthuscanbefunc- with the EZ-LinkTM Sulfo-NHS-LC-Biotinylation kit from tionally considered as a type III orthologue. Pierce (Rockford, IL) in 20 mM sodium borate (pH 8.5) Interchain ionic interactions, which are known to containing 4 M urea and 1 mM 2-mercaptoethanol. The be an important factor in specifying assembly part- overlay experiments were performed on nitrocellulose membranes containing the ten immobilized, non-modi- ners in coiled coil molecules, have been calculated ®ed lancelet IF proteins A3, B1, B2, C2, D1, E1, E2, k1, and shown to be consistent with the observed pre- Y1andX1.6Membraneswereblockedbyincubationfor ference of A3/B2 heterodimers and B1 homodi- one hour at room temperature in 5 % (w/v) skim milk in mers(Table1).However,thecalculationsshow TBS (20 mM Tris-HCl (pH 7.4), 150 mM NaCl), 0.05 % that a small number of other chain combinations (v/v) Tween 20 (TBST) and then incubated for two are also favoured. Since the role of the head and hours at room temperature with the biotinylated A3 or tail domains on IF assembly is not predictable at B2 probes in 20 mM Tris-HCl buffer (pH 7.5) containing the residue level, and because not all of the poss- 4 M urea. After washing in TBST, in TBST, 1 % (v/v) ible ionic interactions are likely to be made in vivo, Triton, in TBST, 0.5 % (w/v) NaCl and again in TBST, further discrimination between chain combinations each for ®ve minutes, the membranes were treated with horse-radish peroxidase-conjugated streptavidin (Pierce) is not possible at present. used at a concentration of 0.2 mg/ml. The blots were Currently we do not know why the A/B branch ®nally developed using the ECL chemiluminescence kit of the evolutionary trees calculated by two lab- from Amersham according to the manufacturer0s instruc- oratories and using both the distance and the tions. maximum-parsimonymethods(Figure10(a);6,12,14) Isolation of the polymerisation-competent heterodimer covers several functional and structural distinct of B. ¯oridae A3 and B2 containing a single disul®de proteins: the A group, which shows theoretical pre- bridgewasessentiallyasdescribed.9Brie¯y,equimolar ference for type II chains, the functional type III amounts of both recombinant proteins were dialysed for protein B1 and the related protein B2, which both three hours at room temperature against 2 mM Tris-HCl were theoretically designated to have a type I (pH 9), 1 mM 2-mercaptoethanol and then overnight structure(Table2).Onepossibleexplanationis against 10 mM Tris-HCl (pH 7.5) ®lament buffer in redu- cing agent. An aliquot was removed to monitor ®lament giveninFigure10(b).Thishypotheticalviewof formation by electron microscopy. Filaments were then Branchiostoma IF evolution proposes (a) that the dialysed for 20 hours against the ®lament buffer without more or less de®ned keratin-like character of the A reducing agent and harvested by centrifugation. The ®la- and B chains resulted from evolutionary processes ment pellet was washed with 10 mM Tris-HCl (pH 8), independent from those that drove the diversi®ca- 5 mM EDTA and again centrifuged. The ®nal pellet was 136 Novel Heterotypic IF System in Amphioxus resuspended in the same buffer and left for 20 hours. An Bigger values indicate a stronger preference for the par- aliquot was removed to monitor ®lament formation by ticular chain type. electron microscopy and to characterize the composition Only 259 amino acid long sequences of the rod of oxidized ®laments on SDS/10 % (w/v) polyacryl- domains (the linker and other regions of questionable amide gels run under non-reducing as well as reducing homology were excluded) of all 13 Branchiostoma IF conditions. Filament were dissolved in 10 mM Tris-HCl proteins and of 22 other vertebrate and invertebrate IF (pH 8), containing 8.5 M urea without reducing agent, sequences were aligned and submitted to weighted followed by Mono Q chromatography using the same (rescaled consistency index over an interval of 1-1000) buffer and a gradient formed with 0.4 M NaCl in the maximumparsimonyphylogeneticanalysis.22Random same buffer. An aliquot of the major peak eluted from additions (ten cycles) of the data were analysed with a the Mono Q column was subjected to one-dimensional heuristic method with a branch-swapping algorithm (SDS/10 % polyacrylamide) and two-dimenstional gel (tree bisection-reconnection, TBR). A bootstrap analysis electrophoresis using the IPGphor system from Pharma- (1000 replications) was used to infer a majority rule con- cia (pH 3-10). Gels were stained with coomassie brilliant sensus phylogram that was rooted with the Homo sapiens blue. Another aliquot was dialysed for four hours lamin C as outgroup. Accession numbers of the Bran- against 10 mM Tris-HCl (pH 7.5) without reducing agent chiostoma sequences used in this study are presented in to monitor ®lament assembly of the crosslinked heterodi- Table2;othersequenceswerereportedelsewhere.14 mer by electron microscopy.

Antibody preparations and immunofluorescence analyses Acknowledgements The A3 rabbit and B2 guinea pig antisera were obtained using the synthetic A3 peptide We thank Wolfgang Berning-Koch, Uwe Plessmann CVAEFQKKVDTLRAEA (amino acid residues 236-250 of and Susanne Brandfass for expert technical assistence. theA3proteininFigure1)andB2peptide CRLREIAENKVAIAEL (residues 288-302 of the B2 pro- teininFigure1),respectively.Peptideswerecoupledvia References their extra N-terminal cysteine residue to hemocyanin. Both antisera were af®nity-puri®ed and, if necessary, 1. McLean, W. H. & Lane, E. B. (1995). Intermediate preabsorbed on other recombinant proteins coupled to ®laments in disease. Curr. Opin. Cell Biol. 7, 118-125. cyanogen bromide-activated Sepharose (Pharmacia). The 2. Fuchs, E. & Weber, K. (1994). Intermediate ®la- speci®city of each antibody was veri®ed by immunoblot- ments: structure, dynamics, function and disease. tingonpuri®edrecombinantIFproteinsasdescribed.6 Annu. Rev. Biochem. 63, 345-382. Frozen sections of B. ¯oridae early larvae and adults 3. Parry, D. A. D. & Steinert, P. M. (1995). Intermediate were processed for indirect immuno¯uorescence Filament Structure, Springer, New York. microscopyasdescribed.6 4. Erber, A., Riemer, D., Hofemeister, H., Bovenschulte, M., Stick, R., Panopoulou, G. et al. (1999). Characterisation of the Hydra lamin and its Sequence analyses gene; a molecular phylogeny of metazoan lamins. J. Mol. Evol. 49, 260-271. Potential interchain ionic interactions between charged 5. Stuurman, N., Heins, S. & Aebi, U. (1998). Nuclear residue pairs in positions 2e0-1 g, 1 g0-2e, 2a0-1 g, 1 g0-2a, lamins: their structure, assembly, and interactions. 1e0-1d,and1d0-1ewerecalculatedasdescribed.21 J. Struct. Biol. 122, 42-66. The following methodology was used to designate the 6. Karabinos, A., Riemer, D., Panopoulou, G., Lehrach, chain type. Consensus sequences were derived for rod H. & Weber, K. (2000). Characterisation and tissue- domain segments 1a, 1b, 2a, L2 and 2b in type I and speci®c expression of the two keratin subfamilies of type II mammalian keratins. Variability in lengths and intermediate ®lament proteins in the cephalochor- character precluded a similar consensus being obtained date Branchiostoma. Eur. J. Cell. Biol. 79, 1-10. for segments L1 and L12. For some positions in the 7. Riemer, D., Karabinos, A. & Weber, K. (1998). Anal- sequences, considerable variability was observed, and ysis of eight cDNAs and six genes for intermediate thesesiteshavebeenrepresentedbyadot(Figure9).In ®lament proteins in the cephalochordate Branchios- addition, where the same consensus residue was present toma reveals differences in the multigene families of in both the type I and the type II chains, both residues lower chordates and the vertebrates. Gene, 211, 361- were designated X, and were thus not used further in 373. attempts to assess the typeI/II character of Branchiosto- 8. Riemer, D. & Weber, K. (1998). Common and var- ma chains. The numbers of identities between the Bran- iant properties of intermediate ®lament proteins chiostoma chains and the type I and type II consensus from lower chordates and vertebrates; two proteins sequences were calculated both for the entire sequence from the tunicate Styela and the identi®cation of a and for those residues falling in the non-a and non-d pos- type III homologue. J. Cell Sci. 111, 2967-2975. itions of the heptad repeat. Finally, the results were 9. Wang, J., Karabinos, A., SchuÈ nemann, J., Riemer, D. scaled to allow for the fact that the numbers of con- & Weber, K. (2000). The epidermal intermediate served residues unique to the type I and type II consen- ®lament proteins of tunicates are distant keratins; a sus sequences differed (144 and 157 respectively in all polymerisation-competent hetero coiled coil of the positions, and 105 and 120 respectively in non-a and Styela D protein and Xenopus keratin 8. Eur. J. Cell. non-d positions). The differences between the numbers of Biol. 79, 478-487. identities for the type I and type II consensus sequences 10. Erber, A., Riemer, D., Bovenschulte, M. & Weber, K. islistedinTable2.Positiveandnegativevaluesindicate (1998). Molecular phylogeny of metazoan intermedi- a preference respectively for type I and type II chains. ate ®lament proteins. J. Mol. Evol. 47, 751-762. Novel Heterotypic IF System in Amphioxus 137

11. Weber, K., Plessmann, U. & Ulrich, W. (1989). Cyto- of heterotypic complexes and intermediate-sized plasmic intermediate ®lament proteins of invert- ®laments by homologous and heterologous recom- ebrates are closer to nuclear lamins than are binations of puri®ed polypeptides. J. Cell. Biol. 101, vertebrate intermediate ®lament proteins; sequence 1826-1841. characterization of two muscle proteins of a nema- 17. Ruppert, E. E. (1997). Cephalochordata (Acrania). In tode. EMBO J. 11, 3221-3227. Microscopic Anatomy of Invertebrates, vol. 15, pp. 349- 12. Karabinos, A., Riemer, D., Erber, A. & Weber, K. 504, Wiley-Liss, New York. (1998). Homologues of vertebrate type I, II and III 18. McLachlan, A. D. & Stewart, M. (1975). Tropo- intermediate ®lament (IF) proteins in an invert- myosin coiled-coil interactions: evidence for an ebrate; the IF multigene family of the cephalochor- unstaggered structure. J. Mol. Biol. 98, 293-304. date Branchiostoma. FEBS Letters, 437, 15-18. 19. Parry, D. A. D., Crewther, W. G., Fraser, R. D. B. & 13. Karabinos, A., Wang, J., Wenzel, D., Panopoulou, MacRae, T. P. (1977). Structure of a-keratin: struc- G., Lehrach, H. & Weber, K. (2001). Developmen- tural implication of the amino acid sequences of the tally controlled expression pattern of intermediate type I and type II chain segments. J. Mol. Biol. 113, ®lament proteins in the cephalochordate Branchios- 449-454. toma. Mech. Dev. 101, 283-288. 20. Conway, J. F. & Parry, D. A. D. (1988). Intermediate 14. Luke, G. N. & Holland, P. W. (1999). Amphioxus ®lament structure. 3. Analysis of sequence hom- type I keratin cDNA and the evolution of intermedi- ologies. Int. J. Biol. Macromol. 10, 79-98. ate ®lament genes. J. Exp. Zool. 285, 50-56. 21. Citi, S., D0Atri, F. & Parry, D. A. D. (2000). Human 15. Hesse, M., Franz, T., Tamai, Y., Taketo, M. M. & and Xenopus cingulin a modular organization of the Magin, T. M. (2000). Targeted deletion of keratin 18 coiled-coil rod domain: predictions for intra- and and 19 leads to trophoblast fragility and early intermolecular assembly. J. Struct. Biol. 131, 135-145. embryonic lethality. EMBO J. 19, 5060-5070. 22. Swofford, D. L. (1998). PAUP*4.0-Phylogenetic Ana- 16. Hatzfeld, M. & Franke, W. W. (1985). Pair formation lyses using Parsimony (* and Other Methods), Sinauer and promiscuity of cytokeratins: formation in vitro Assoc., Sunderland, MA.

Edited by W. Baumeister

(Received 5 September 2001; received in revised form 13 November 2001; accepted 13 November 2001)