Evolutionary Analysis
of "Hagfish Amelogenin"

ABSTRACT Hagfishes lack mineralized tissues and teeth. Part of a cDNA strand,alledgedly from amelogenin, the major gene involved in enamel formation inmammals, has recently been cloned in a hagfish (Slavkin and Diekwish,1996; Anatomical Record vol. 245). This cloning is of great interest because itcould change the current view about the evolution of mineralized tissues but nophylogenetic analysit nophylogenetic analysis of this piece of DNA has been made by the authors. Phylogenetic analysis of this part of cDNA has been conducted using bothphenetic and cladistic methods. The cDNA amplified in hagfish does not fit with a non-mammalian origin butfits well with a degraded rodent sequence. The gene cloned in hagfish is probably of mammalian origin due tocontamination during PCR. Key-words: amelogenin; teeth; evolution; hagfish; PCR contamination.

Hagfishes are eel-shaped, jawless craniates belonging to the Hyperotreti, thesister-group of vertebrates (Janvier, 1996). All extant hagfishes and all knownfossil Hyperotreti lack mineralized tissues; their skeleton is cartilaginous,and their mouth is armed with "horny teeth" (odontoids ortoothlets). The absence of mineralized tissues, which are otherwise found inall gnathostomes and in many groups of fossil jawless vertebrates, haess vertebrates, has beenvariously interpreted. While many early phylogenies implied that hagfishes hadlost the ability to produce mineralized tissues, the current consensus is thatthe ancestors of hagfishes never possessed this ability (Janvier, 1993). Therecent cloning of a piece of hagfish (Eptatretus stoutii) cDNA byRT-PCR, using amelogenin primers (Slavkin and Diekwish, 1996; Anatomical Recordvol. 245), is of great interest in the perspective of the evolution ofmineralized tissues, but is quite surprising. Indeed, amelogenin has been knownfor a few years to be involved in the formation of mammalian tooth enamel(review by Deutsch, 1989), and recently, we have shown that the amelogenin geneis probably absent in toothless sauropsids, such as turtles and birds (Girondotand Sire, 1998). This reinforces the idea that the only role of amelogenin inamniotes is to contribute to enamel formation, and that it is lost in taxa thatlack selective pressure to maintain its integrity (in less than 200 My in thecase ofin thecase of turtles, and in less than 100 My in the case of birds). The presenceof amelogenin in hagfishes implies either that this gene has another function(at least in hagfishes), or, if we assume that the ancestors of hagfishes oncehad a mineralized skeleton (as suggested by most early phylogenies), that thisapparently inactive gene has been retained for over 300 My (the age of theoldest known hagfish).

The existence of an amelogenin gene in hagfishes has been used by Slavkin andDiekwish (1996) to validate the observed cross-reactivity in hagfish toothletsof polyclonal antibodies against mammalian amelogenin (Slavkin et al., 1982,1983, 1991). However, the published 50 amino-acid sequence of the hagfishamelogenin is very similar to the known eutherian sequences. The authors havenot performed any quantitative measure of divergence between these sequences,although such an analysis either would have permitted confirmation of thenon-mammalian origin of the sequence, or woulquence, or would have detected any sourceof contamination during the PCR phase of the cloning protocol. Therefore, todetermine the origin of the "hagfish amelogenin", we have performeda phylogenetic test. *Correspondance to: Marc Girondot, URA 1137, Evolution et Adaptations des SystèmesOstéo-musculaires, CNRS and Université Paris 7, Case 7077, 2place Jussieu, 75251 Paris cedex 05, France. E-mail: mgi@ccr.jussieu.fr Received 12 March 1998; Accepted 25 June 1998

Recently, we have demonstrated that the amelogenin gene can be useful forstudies of mammalian phylogeny (Girondot and Sire, 1998). It is possible todemonstrate a non-mammalian origin of any sequence by establishing its basalposition in the phylogeny, although no outgroup is available because onlymammalian sequences are known. The phylogenetic relationships betweenhagfishes, metatherians (marsupials), and eutherians (placental mammals) withincraniates are known without ambiguity, and this phylogeny can be used as a"true phylogeny" (Janvier, 1996). The mammalian amelogeninphylogeny should include an early divergence between the opossum and theeutherian sequences, anrian sequences, and a hagfish amelogenin sequence should fit outsideMammalia. However, since no outgroup is available for this analysis, severalrooting options are possible, but only two are biologically meaningful. First,the trees can be rooted between hagfishes and mammals (Fig. 1A, 2A). If thereported sequences truly belong to hagfishes, the "correct" dichotomy betweenmetatherians and eutherians should be found. Second, the trees can be rootedbetween metatherians (the opossum) and the other taxa (Fig. 1B, 2B). If thereported sequences were the result of contamination from a eutherian, the"hagfish" sequences should form a clade with the species from which the DNAactually originated.

Alignment has been done using Clustal X 1.64b (Thompson et al., 1994), but thesequences can be aligned without any ambiguity by hand. Only one gap isrequired in the human X gene [the amelogenin gene is located on the heterosomalpart of the sex chromosomes in eutherians; (Girondot and Sire, 1998)]. Sire, 1998)]. Analyseshave been performed using both a distance method (phenetic) and a parsimonymethod (cladistic) (see legend of figures for details of computing procedures).For the distance method, a recent algorithm (BIONJ; Gascuel, 1997) hasbeen used to minimize the effect of different substitution rates in X and Ymammalian chromosomal lineage of amelogenin (Huang et al., 1997).

Fig. 2: Strict consensus tree of the 22 shortest trees (41 steps) obtained bythe parsimony method using an exhaustive search (2 027 025 trees analyzed) inPAUP3.1.1 (Swofford, 1993), with collapsing zero-length branches option.The putative hagfish sequence is indicated by "S&D". The treesare rooted at the divergence between "S&D" and other sequencesin panel A or rooted at the divergence of metatherian and othersequences in panel B. Indicated below each tree is the expected topologyif "S&D" sequence is a hagfish seques a hagfish sequence (A) or if"S&D" sequence is a eutherian sequence (B). Bootstrapvalues in % were obtained using 1000 replicates with the Branch-and-Boundalgorithm. The 50% majority-rule consensus of the resulting trees has the sametopology as the strict consensus of the shortest trees obtained by theexhaustive search.

Using both a BIONJ distance tree (fig. 1) and a consensus parsimony tree(fig. 2), the topology of the inferred phylogeny is never consistent with anon-eutherian origin of the putative hagfish sequences reported by Slavkin andDiekwish (1996). In both cases (fig. 1A and 2A), if the trees are rootedbetween the presumed "hagfish" sequences and the other taxa, theopossum is deeply nested in Eutheria. This result suggests that the"hagfish" sequence actually represents a eutherian contaminant. To assess the plausibility of a contaminant origin of the "hagfish"sequence, and to identify the actual source ofthe actual source of the DNA sequence, the trees werere-rooted between the opossum and the other taxa (fig. 1B and 2B). In thiscase, the "hagfish" sequences clustered with the rodent sequences(rat, mouse and hamster; fig. 1B and 2B). Twenty-two shortest trees (41 steps)are obtained by the parsimony method, and in all these trees the"hagfish" sequence clustered with the rodent ones (fig. 2B). The"correct" basal topology is obtained in only three of the 118 trees thatrequire up to one extra step, and a strict consensus of these trees give atopology inconsistent with the actual knowledge on the evolution of amelogeninbased on the entire gene sequence (Girondot and Sire, 1998). Furthermore, the"hagfish" sequence still clustered with at least one rodentsequence in the remaining 115 of these 118 trees. This suggests that theposition of hagfishes as sister-group of all the placental sequences in threeof these 118 trees is not significant. Finally, 63% and 60% of the 1000bootstrap replicates using respectively parspectively parsimony or BIONJ methods link"hagfish" with rodent sequences whereas the analyzed sequences arerelatively short (fig.1B and 2B). The classical neighbor-joining method(Saitou and Nei, 1987) gives exactly the same tree topology as the BIONJ method(not shown).

The divergence time between hagfishes and the other species analyzed here isapproximately 470 million years, whereas the divergence time betweenmetatherians (marsupials) and eutherians (placental mammals) is only 120million years. Thus, the results obtained here cannot be explained byrapid speciation events. More probably, the published "hagfish"sequence is a mammalian sequence obtained by contamination during PCRamplification. This artifact may result from the use of either somepotentially degenerate primers, a low annealing temperature, or a high numberof PCR cycles (the primer sequences, the annealing temperature, and the numberof PCR cycles are not described in the original article). Such a contaminSuch a contaminationis frequent when cloning genes of distant species using PCR [see for examplethe fungal and angiosperm origin of putative dinosaur ribosomal genes (Wang etal., 1997) or the human origin of putative dinosaur mitochondrial gene (Colluraand Stewart, 1995)], and might be detected before publication by a phylogeneticanalysis of the produced sequence. The differences between the putative"hagfish" sequence and the most similar mammalian one could be dueto the sequencing of a formerly unsequenced mammalian amelogenin gene, or moreprobably, to the sequencing of a contaminant, degraded mouse gene. Our conclusion that the putative "hagfish" amelogenin gene sequencepublished by Slavkin and Diekwish (1996) probably originates from a mammaliancontaminant does not invalidate the results obtained using mammalian antibodies(Slavkin et al., 1982, 1983, 1991), but this sequence cannot be used tovalidate them. The primary sequence of amelogenin cDNA in hagfish remains to befound. ACKNOWLEDGMENT>ACKNOWLEDGMENTS We thank Patricia Lai for correction of this manuscript and Jean-Yves Sire andArmand de Ricqlès (URA 1137) for critical reading and many valuablesuggestions. Michel Laurin was supported by the Alexander von HumboldtFoundation.