Lasanius, an exceptionally preserved Silurian jawless fish from Scotland

The fossil record of non‐biomineralizing, soft‐bodied taxa is our only direct evidence of the early history of vertebrates. A robust reconstruction of the affinities of these taxa is critical to unlocking vertebrate origins and understanding the evolution of skeletal tissues, but these taxa invariably have unstable and poorly supported phylogenetic positions. At the cusp between mineralized bony vertebrates and entirely soft‐bodied vertebrates is the enigmatic Lasanius, a purported anaspid from the Silurian of Scotland. Interpretations of its affinity and significance are conflicted, principally because of its poorly understood anatomy due to taphonomic distortion and loss of soft‐tissues. Here we use an array of modern techniques to reassess the anatomy of Lasanius via a comprehensive study of 229 complete and partial specimens. A new reconstruction clarifies the identity and position of preserved features, including paired sensory organs, a notochord, and digestive tract, supporting the vertebrate affinities of this genus. SEM‐EDS trace element mapping suggests a bone‐like composition of mineralized parts, but finds no evidence for mineralized dermal armour. Phylogenetic analysis recovers Lasanius as an early stem‐cyclostome, and subsequent analysis supports the rejection of alternative placements (such as stem‐gnathostome). We highlight that while distinguishing between the early cyclostome and gnathostome condition is problematic, increasing confidence in the anatomy of key taxa, such as Lasanius, is vital for increased stability throughout the early vertebrate tree.

V E R T E B R A T E S comprise two clades: the jawed vertebrates and their relatives (Gnathostomata), and the jawless vertebrates and their relatives (Cyclostomi) (Janvier 1996). While modern cyclostomes are restricted to lamprey and hagfish, gnathostomes are extremely diverse and disparate: they include all other extant vertebrate taxa, and possess many other traits that cyclostomes lack, including paired fins and bony biomineralized tissues. These traits may have been key to the comparative success of the gnathostomes, and understanding their origins is essential for a better understanding of the evolution of vertebrates as a whole. Fortunately, this period is well documented by a diverse range of early vertebrates, including many softbodied, non-biomineralizing taxa that are vital for establishing events before, during, and after the advent of bone, and during the gnathostome-cyclostome divergence. However, establishing the timing of character acquisition requires a robust phylogenetic tree, and due to the notorious problems encountered analysing their fossils, the soft-tissue taxa are often some of the most intractable in phylogenetic analyses, thus obscuring their evolutionary history.
One such problematic soft-bodied taxa is Lasanius: a genus of extinct jawless fish (Fig. 1A, B), established by R.H. Traquair in 1898 on the basis of several specimens found in the Silurian (c. 443.8 Ma) fish beds of Lesmahagow, Scotland (Traquair 1898). This genus stands out as something unique within the early vertebrates due to an unusual mix of character traits. In comparison to the 'ostracoderms' (armoured stem-gnathostomes), Lasanius is 'naked', seemingly lacking armour. Lasanius also differs from the 'soft-bodied' non-biomineralizing vertebrates in that it has highly localized areas of ostensibly biomineralized tissues (e.g. dorsal row of sharp scutes, akin to those of Anaspida) (Fig. 1E). The presence of biomineralization in an otherwise soft-bodied taxon hints at the importance of Lasanius for understanding vertebrate evolution, potentially bridging the morphological gap between the nonbiomineralizing taxa and the heavily dermally-armoured ostracoderms.
The lack of consensus and the small number of diagnostic traits recognized in Lasanius has a large impact on the phylogenetic analysis of the genus (Sansom 2015). Lasanius appears to be wildly unstable within the vertebrate tree, due to different approaches and interpretations resulting in multiple conflicting hypotheses of affinity (Fig. 2). The uncertainty is further amplified by the poor resolution of the affinities of the Anaspida ( Fig. 2A) (Keating & Donoghue 2016), euphaneropid polytomies (Gess et al. 2006), and the historic debates surrounding cyclostome monophyly (see Donoghue (2017) for a thorough review), which often destabilize the relationships of the other ostracoderm groups.
The debate over the anatomy of Lasanius is driven by the nature of the fossils. Lasanius were primarily softtissue organisms, which are notorious for producing fossils that are hard to understand and easy to misinterpret (Briggs 2003;Sansom et al. 2014;Purnell et al. 2018). The intrinsic taphonomic processes that lead to softtissue fossilization also distort remains, create preservation biases, and cause loss of phylogenetically important morphological information (Cunningham et al. 2014;Sansom 2014;Briggs & McMahon 2016;Purnell et al. 2018;Saleh et al. 2020;Walker et al. 2020). One key problem for phylogenetic reconstruction is to correctly identify the absence of traits as true evolutionary absences, rather than absences created due to factors such as decay or environmental conditions of preservation. Misidentification of a 'lost' trait as a true absence can result in a phenomenon known as 'stem-ward slippage', where during phylogenetic analysis crown taxa are incorrectly positioned as stem-taxa (Sansom et al. 2010a(Sansom et al. , 2011Sansom 2015). Since the original description of Lasanius, many techniques have been developed that aid palaeontologists in the study of soft-tissue fossils. Application of comparative taphonomy and experimentally derived data has been used in early vertebrates to constrain the interpretation of fossil anatomy in the precise geological context (Briggs & Kear 1993;Sansom et al. 2013;Mounce et al. 2016;Parry et al. 2018;Purnell et al. 2018). Taphonomically informed interpretations still do not mitigate another important factor belying interpretation of exceptionally preserved soft tissue fossils: potential circularity in homology statements arising from initial choice of modern comparator organism. Instead, systematic and objective approaches to fossil descriptions (Donoghue & Purnell 2009;Sansom et al. 2010b) are necessary to separate description from homology. Similarly, innovative uses of geo-analytical techniques, such as SEM-linked energy-dispersive x-ray spectroscopy (SEM-EDS) and the more powerful synchrotron rapid scanning x-ray fluorescence (SRS-XRF) (Bergmann et al. 2012;Edwards et al. 2013Edwards et al. , 2018, combined with the breakthrough discovery that fossils can retain traces of original compounds (Manning et al. 2009;Briggs & Summons 2014; Ann e et al. 2019) have created a previously unexpected avenue for research and data collection and interpretation of anatomy (Manning et al. 2013;Rossi et al. 2019;Prado et al. 2021). Although the application of these techniques has been used with great success in clarifying the identity of unknown or ambiguous characters in early vertebrates, they have not been widely used when considering Lasanius specimens (Sansom et al. 2010b;Clements et al. 2016;Gabbott et al. 2016;Miyashita et al. 2019).
In this study we aim to reconstruct the anatomy and affinity of Lasanius by taking a comprehensive and objective approach. Anatomical elements are first considered in terms of their morphology, composition, and comparative taphonomy before reference to a modern comparator organism (Donoghue & Purnell 2009;Sansom et al. 2010b). This includes direct study of multiple specimens and characterization of the chemical composition using SEM-EDS. This morphology is then interpreted in terms of homology with comparator organisms, and placed in a phylogenetic context to address the questions: What is Lasanius, and what are its implications for vertebrate evolution?

Material
A total of 229 whole and partial Lasanius specimens were included in this study from a range of institutions (Data S1). Taxonomy. Traquair assigned two species to Lasanius: L. problematicus Traquair, 1898 and L. armatus Traquair, 1899. As dorsal scute morphology is the only diagnostic synapomorphy of Lasanius species, topology and anatomy have been considered here at genus level, with notes on species variation where appropriate. The status of a third species (L. 'altus') is unclear. While Ritchie included L. 'altus' throughout his PhD thesis (Ritchie 1963) it was with reference to more detailed work by Dr I. C. Smith from the University of Glasgow in 1958. This work consisted of a restudy of Lasanius, and also established L. 'altus' as a third species. However, Ritchie commented that it had not been published at the time of his thesis (Ritchie 1963) and, as far as we are aware, it does not appear to have been published subsequently. Ritchie described a comparable increase of scute overlap for L. 'altus' specimens, a 'deeper body', and notable variation in scute morphology (Ritchie 1963). However, scute overlap can be increased by different degrees of curvature in the dorsal aspect of the fossil, the deep body may be a taphonomic distortion effect, and the described scute morphology is seen in dorsoventrally preserved scutes in L. problematicus, suggesting that the L. 'altus' diagnostic characters are possibly preservation artefacts. Unfortunately, comparisons are limited as very few specimens labelled L. 'altus' exist, and the most complete L. 'altus' fossil within this study (NHMUK P11048) shows signs of distortion and folding in the scute line. As Smith's work is unavailable, there is no detailed description or established holotype for this species so positive identification of L. 'altus' specimens is impossible and further investigation of the validity of this species is hindered, leaving L. 'altus' as nomen nudum for the time being. Within this dataset, original collection names have been maintained, however L. 'altus' specimens are considered as Lasanius sp., except where a valid species name is also provided, in which case this name takes precedent (e.g. L. problematicus (altus)).

Methods
Our first stage was objectively describing the morphological features of the organisms without reference to homology in order to try and minimize circularity and bias in description (Donoghue & Purnell 2009;Sansom et al. 2010b). This involved identifying features and their morphology preserved In differed planes and orientations to reconstruct three-dimensional nature (Briggs & Williams 1981;Purnell & Donoghue 1999) and the composition of those features. The reconstruction was then used to provisionally select and justify the choice of comparison taxon group, and interpret homologies (Fig. 3).
Specimens were analysed using light microscopy and photography (see Appendix S1 for details). SEM-EDS was performed using a FEI Quanta 650 FEG fitted with a Bruker XTrace microXRF and a Bruker XFlash 30 EDS system located in the Department of Earth and Environmental Sciences, University of Manchester, UK. For chemical maps, the electron beam was moved in a raster pattern across the target areas using a low accelerating voltage of 15 kV to avoid sample charging issues. The resolution of the spectrometer is 129 eV (@ Mn Ka) and c. 100 eV for the C K line. Individual point analysis was also performed on selected specimens (Tables S2, S3). As this is an environmental chamber, specimens were not prepared or carboncoated for SEM-EDS, and the water vapour pressure was 0.5 mbar. Complementary scanning electron microscope backscatter electron (SEM-BSE) images were also created at the same time using the same equipment.
Size and dimensions of specimens were assessed using 15 continuous morphological measurements (Fig. S1 Phylogenetic analysis. The character dataset from Miyashita et al. (2019) was used for the phylogenetic analysis as it is the most recent, and combines several previous works. Due to the uncertainty of the validity of Achanarella, Ciderius and Cornovichthys these taxa were removed from the analysis (Van der Brugghen 2017; Miyashita et al. 2019). Pipiscius zangerli was removed due to uncertainty about its affinities with the agnathans (Shu et al. 1999;Gabbott et al. 2021), while Pikaia gracilens was also removed due to debate over its position within the chordates (Mallatt & Holland 2013). In all analyses, Hemichordata were considered the outgroup.
To assess the support of different phylogenetic topologies, we applied Bayes factor analysis (Jeffreys 1935;Kass & Raftery 1995). Using the methods described in Bergsten et al. (2013) andDembo et al. (2015), separate Bayesian analyses were performed, only varying in the topological constraints used to reflected competing hypotheses, placing Lasanius: (1) within the cyclostome total group within Anaspida; (2) within the gnathostome total group within Anaspida; (3) as a stem-cyclostome, with other anaspids as stem-gnathostomes; (4) as a stem-gnathostome outside of the anaspid group (Table S1; Figs S2-S5). MCMC sampling was used to acquire the marginal likelihood of each model via the harmonic means (HM) of the marginal log-likelihoods, a ratio of these was then used to produce a Bayes factor number. Assessment of the resulting Bayes factors for support for the 'best' model (i.e. the one with the highest marginal likelihood) used the interpretation guidelines established by Kass & Raftery 1995, p. 777). This procedure was repeated using a steppingstone method of sampling, which allows for a more direct calculation of the marginal likelihood and is considered a more accurate assessment (Xie et al. 2011;Bergsten et al. 2013). However, these analyses failed to resolve a tree after 100 million generations so the results have not been reported.

Topological description
Body shape and size. The mirrored location of several anterior paired structures indicate Lasanius was a bilaterally symmetrical organism. Comparison of specimens preserved in dorsoventral and lateral orientations suggest a laterally-compressed, elongated, fusiform body plan. The body curvature angle indicates the row of scutes were positioned on the dorsal aspect. Differing terminal morphologies of Lasanius allow for the provisional assignment of anterior and posterior. The anterior is indicated by a bluntly rounded end containing several paired circular structures and dark stains. The posterior end is split into two lobes, one thin and elongated with a ventral slope, and the other smaller, more faintly preserved, and dorsally sloped. Two distinct morphotypes are preserved: a 'thick' morphotype with a deep anterior portion of the torso, wider body shape, and generally more robust outline, and a 'thin' morphotype which is more gracile with a shallower torso. Increasing body length was not found to increase width of the torso in complete Lasanius specimens (linear regression (LR), p > 0.05).
The body ratio of Lasanius is approximately 1:3:1 (head:torso:tail) with some variation due to the angle of preservation. Complete specimens range between 13.3 and 74.5 mm long. A significant difference was found between the body length complete specimens of L. armatus and L. problematicus (T-test with Welch correction: p < 0.01, df = 9.59) although this is based on an uneven and small sample size (L. armatus n = 3, L. problematicus n = 12). The entire body of Lasanius is rarely preserved completely; the caudal lobes are often missing, and incomplete specimens suggest a larger maximum size. A positive linear relationship between the total body length and the length of the dorsal scute series was recovered (LR, b = 0.54235, p < 0.01, t = 10.178, df = 12), identifying the dorsal series as a potential proxy for body size. However, establishing if the dorsal series is complete can be problematic due to the variation of the number of scutes in individuals.
Anterior 'head' structures. The anteriormost area of Lasanius specimens contains a number of paired and nonpaired features.
The anteriormost structure is a circular-oval dark stained area that encompasses the entire anterior terminal edge of Lasanius. It is commonly preserved as an unpaired solid dark stain, but counterpart fossils often exhibit the stain as a dark ring surrounding lighter material ( Fig. 4A, B, DTS). Termed here 'dark terminal stain', it has a high preservation potential, and is found in a wide range of specimens, including those that preserve no other soft tissue. Chemical element mapping of this area indicates the dark material is associated with carbon (C), while the lighter areas are correlated with calcium (Ca) (Fig. 4C, D, DTS). Phosphorus (P) showed no correlation with this area.
Immediately behind the dark terminal stain are two large laterally paired circular structures (Fig. 4A, PLC). These structures, termed 'paired lateral circles', are seen in specimens that also preserve at least an impression of the body imprint. The circles occupy approximately 50% of the head region, with morphology varying between full dark circles and a dark ring around a light centre, dependent on whether counterpart or part. Comparison of laterally preserved specimens (where structures lie 'stacked' vertically with some overlap) and dorsoventrally preserved specimens (where circles are parallel but on opposite sides of the body) indicates a probably symmetrical placement on either lateral side of the head. Rare specimens display a 'bulging', with the arc of the circle lying outside of the natural line of the head, possibly signifying a more dorsal position in life. In light of this, the overprinting of the circles in laterally preserved specimens can be explained by some degree of distortion or twisting during the preservation process of the head. Chemical element analysis of these structures shows a concentration of C in the dark material, and a strong association of Ca with the white areas ( Fig. 4C, D, PLC).
Posterior to the large paired lateral circles are two smaller circular-oval shaped masses positioned close to the median line, level with the paired circular structures and approximately one-quarter the size (Fig. 4A, SPM). The masses, termed 'small paired masses', are preserved as a lighter colour than the surrounding body imprint, with a relatively smooth, flat texture. They display a similar vertical stacking as the circular structures, and comparison of different planes of orientation reveals that they too were symmetrically placed on either side of the head. The masses show a strong concentration of Ca (Fig. 4E, SPM), but little or no correlation with P ( Fig. 4F, SPM), silicon (Si) or other elements (Figs S8, S9). The masses lack texture or ornamentation, and are only seen within the body print, never near to the external margins in either lateral or dorsoventrally preserved specimens, suggesting that they are internal structures.
In addition to the anterior structures described, some isolated structures that are poorly characterized are observed in a limited number of specimens. A small circle or ring shape can be observed between the larger paired anterior circles, termed here the 'unpaired median ring'. This circle appears to be positioned in a central location on the dorsal aspect of the head, and is most clearly visible in GLAHM V8041 (Fig. 4G, H, UMR). Although present in the specimen BGS GSE 9713 (Fig.  4A, UMR), no distinct chemistry was detected during SEM-EDS analysis, unlike the nearby features. A small number of studied specimens of Lasanius exhibit a thin dark line on the dorsal medial aspect of the 'head' (Fig.  4B, G, H, ?DDM). This 'dark dorsal material' appears to be constrained to this region, but similar dark material directly below the scute line observed in other specimens may be a continuation of the same tissue. The dark dorsal material is only seen in well preserved specimens with a body imprint, although it is not consistently present within this group. Only one dorsoventral specimen (GLAHM V8041) preserves a similar structure (Fig. 4H), but it is not clear if this is the same feature, so details of the width or coverage of the dark dorsal material are not available.
Paired 'chains and rods' apparatus. Two classes of structures show markedly three-dimensional preservation in comparison to the rest of the morphology and have historically been interpreted as derived from some sort of skeletal tissue: (1) laterally paired apparatus made up of 'chains and rods'; and (2) a series of dorsal 'scutes'. Different preserved orientations demonstrate that the cage-like 'chains and rods' apparatus were located on the anteroventral lateral sides of Lasanius (Fig. 1A, B lateral preservation, Fig. 5A dorsoventral preservation). Both paired elements of the apparatus are very rarely found displaced from each other, and individual units are only very rarely disarticulated from within their respective series. Throughout the apparatus there are no signs of ornamentation; a smooth texture is apparent when examined using both light and scanning electron microscopy. EDS chemical maps of these structures show a strong correlation with both Ca and P in contrast to the surrounding material ( Fig. 6B, C). EDS point analysis demonstrates a clear association between Ca and P, and the 'rods'. A significant difference was recovered between the percentage concentrations of Ca and P in locations in the tissue and the matrix, allowing for the unambiguous distinction between the two (T-test with Welch correction, Ca p < 0.01, P p < 0.01, both df = 4) (concentration was calculated as atomic weight percent assuming oxygen as the balancing anion). In addition, P concentrations are almost undetectable in areas without tissue interpreted as skeletal tissue (Fig. 6E). While less distinct, Ca concentrations show a similar pattern (Fig. S12). Ca and P exhibit a positive linear correlation; areas with high P concentrations also have high Ca concentrations (Fig.  S10). However, this relationship is not perfect, as there are other minerals containing Ca present and these influence the overall concentrations of Ca in the specimen. The anterior paired series of the 'chains and rods' apparatus is made of six separate units (12 in total), spaced at regular intervals forming the oblique anteroposterior dorsoventral 'chains' (Fig. 5A, B, Ch) The anteriormost unit is positioned, close to the midline of Lasanius, approximately level with the anterior origin of the dorsal scute line. The posteriormost unit sits next to the anterior unit of the associated 'rod' series, although they do not connect. The individual 'chain' units are acute isosceles triangles, with deeply concave bases and elongated dorsal points that taper to a sharp apex. Within the 'chain', each element is positioned below and behind the proceeding unit, with the base approximately level to the middle of the next unit's elongated dorsal point. The units are angled towards the anterior end, and a relaxation of this angle along the 'chain' creates a curved appearance.
The paired series of 'rods' are positioned horizontally along the ventral most aspect of the torso. They contain between 6-9 individual overlapping units ( Fig. 5A, B, Rd). Generally, each unit is a triangular shape with a greatly elongated thin dorsal point that tapers into a sharp tip angled towards the anterior of the body. The specific morphology of each unit varies slightly dependent on the position within the series, only the posteriormost unit has an exposed extended posterior point. The length of the elongated point reduces along the series, as does the anterior angle, creating the appearance of rib-like structures that lean towards the 'head'. Light microscope examination and BSE imaging of the 'rod' series, along with comparison of different preservation orientations, suggests the anterior point at the base of each unit created a 'cradle' under the ventral surface. This would have caused the units to project slightly from either side of Lasanius, with the posterior spine of each unit pointing diagonally backwards away from the organism. Overlaps of the units seen in lateral views appear to be a F I G . 5 . Paired 'chains and rods' apparatus; dorsoventral preservation. A-B, GMRC 03-15.8c visible light photograph and interpretative illustration. C-D, UMZC AG.95 visible light photograph and interpretative illustration outline. E, BGS GSE 5985, BSE composite image of laterally preserved 'rod' and 'chain' structure. Abbreviations: Ch, paired 'chain' apparatus; Ds, dorsal medial scutes; Rd, paired 'rod' apparatus. Scale bars represent: 10 mm (A-D); 2 mm (E). taphonomic artefact possibly due to decay-lead collapse (e.g. Figs 1A, B, 5B, 7B).
Dorsal scute series. Like the paired 'chains and rods' structure, the dorsal scute series is a commonly preserved, seemingly highly robust feature, that often exhibits threedimensionality. The series contains between 11-23 overlapping individual scutes, with the total number varying considerably between individuals and species. The anterior origin is located behind the 'head' region, in line with the anterior aspect of the 'rod' series ( Fig. 1A, B). The posteriormost point is found at the thinnest region of the body, immediately before the 'caudal lobes' (Fig.  7A). Each scute is roughly triangular in profile with a wide base and strongly recurved posterior edge that results in sharply-pointed uppermost apex which faces posteriorly, and that is in line with the lower posterior point. Anteriormost scutes are often flatter and less distinct than the posteriormost units, which tend to have a more elongated base. The overlapping of the scutes results from the recurved edge creating a concave recess at the posterior of each scute that houses the anterior apex of the next scute. Inter-scute spacing is determined by the dorsal curvature of the specimen; as no evidence of interscute connective tissue has been observed, it appears that the dorsal series was somewhat flexible as Lasanius moved. Disarticulation of the scutes is extremely rare and only observed as the displacement of large sections, not individual scutes. Some specimens appear to have a small number of anterior scutes (varying between one and three) in a reversed position, with the sharp posterior apex 'pointing' anteriorly instead of posteriorly, (Fig. 1B, arrow indicates position). This is most apparent in, but not exclusive to, specimens labelled 'L. altus'. As discussed, (see Material and Method, above), this morphology is not unsimilar to the outline of the scutes as seen from a dorsal perspective (Fig. 7D), so it is unclear if this is a true morphology, or a consequence of taphonomic influences. The inconsistency of this orientation and form throughout the specimens analysed suggests the latter is most likely. EDS chemical mapping shows a negative correlation with Si and the fossil material, but the concentrations of Ca, P and manganese (Mn) (Fig. 8B-D) are strongly associated with the scutes.
Variation in the scutes, both in terms of number and morphology, is seen between the two species. The variable F I G . 6 . Paired 'rod' apparatus. A-C, 'rod' structure in BGS GSE 9713: A, backscatter electron (BSE) image; B-C, false colour SEM-EDS chemical maps: B, phosphorus; C, calcium (arrowheads indicate location of one 'rod'). D, BSE image of base of one 'rod', each mark indicates location of one EDS point analysis. E, plot of P concentration at each EDS point analysis location, expressed as atomic weight percent assuming oxygen as the balancing anion for all components; 'skeletal tissue' points are located on the 'rod', 'matrix' points are located on the matrix, 'mixed' indicates points for which exact positioning was unclear; bars indicate maximum 2-sigma error. False colour maps have been enhanced to improve visibility. Scale bars represent: 0.6 mm (A-C); 0.3 mm (D).
number of scutes has been used to distinguish between L. armatus and L. problematicus (Ritchie 1963). Unlike previous interpretations, we find considerable overlap in the number of scutes (11-15 and 11-22 respectively), but their distributions remain significantly different (T-test with Welch's correction, p < 0.01, t = 4.3655, df = 5.2161). While an overall positive linear relationship between body size and scute number in the Lasanius genus is recovered (LR, b = 0.11125, p < 0.05, t = 3.418, df = 12), this is not found within the individual 'species' (LR, L. problematicus p > 0.05, L. armatus, p > 0.5), suggesting no evidence for larger individuals to have more scutes than smaller individuals within a species. Similarly, no linear relationship was recovered between the number of scutes and the length of the series (all species p > 0.05, L. problematicus p > 0.05, L. armatus, p > 0.05), even though the series length increases with body length. This suggests that it is unlikely that the number of scutes increased with growth, but rather that the scutes may themselves have increase size with growth. However, it is prudent to acknowledge that the small sample sizes may restrict the robustness of the interpretations.
Oblique 'surface' stripes. On the main body, thin parallel lines in an anterodorsally-posteroventral direction create a striped pattern (Fig. 9). This pattern is usually patchily preserved and most commonly found on the ventral half of the torso; it has not been observed in the 'head' or 'tail' of Lasanius. The continuous stripes are a gentle sigmoidal shape, and do not exhibit branching or fragmentation. No apparent evidence of texture or ornamentation independent from the underlying matrix topology was observed. Only one specimen studied preserves the pattern across the entirety of the torso: BGS GSE 5983/4 ( Fig. 9A, C). Unfortunately, neither the 'head' nor 'tail' is preserved in BGS GSE 5983/4 so the orientation and shape of the stripes outside the torso is unknown.
Elemental analysis mapping shows a stripe pattern distribution of carbon (Fig. 10B); however, this fails to consistently match the stripe pattern in the original specimen. Ca and P are also present (Fig. 10C, D) but do not show a distribution pattern that correlates with the preserved stripes.
Axial lines. Within the body imprint of Lasanius, two major antero-posterior longitudinal axial lines are present (Fig. 11A, B). While the clarity and fidelity of the preservation varies, these lines remain visible even when almost all of the body impression is not. The dorsalmost line, termed the 'dorsal axial line' sits between the dorsal scute line and the midline of the torso (Fig. 11A, B, DAL). The exact anterior origin is not clear but the line can be seen clearly within the anterior 'head' region. It extends across the full length of the body, following the natural curvature, and terminates in the ventral caudal lobe, preserved as either a darker material than the surrounding body or as a void space. Occasionally an additional thin strip of dark material can be seen near the dorsal edge of the major dorsal axial line.
The other major antero-posterior line, here termed the 'ventral axial line' is located below the torso midline. The anterior origin is also unclear, but it appears to emerge from within the paired 'rod' apparatus in the torso. This line curves down the body, terminating at the ventral edge, in line with the thinnest part of the torso, prior to the caudal lobes. This line also preserves as either a dark impression or void space (Fig. 11A, B, VAL).
Different from the preservation style, morphology and location of the dorsal and ventral axial lines, another longitudinal line, termed the 'lateral line', is apparent in a single specimen (BGS GSE 5983/4; Fig. 9A, C, LL). It is a thin, non-continuous (i.e. broken), dark, anteroposterior line located on the lateral surface of the posterior half of the torso. Within this specimen the anterior origin of this lateral line appears to be at the midpoint of the torso, but a lack of further specimens displaying this trait means the true origin cannot be confirmed. Likewise, the partial preservation of BGS GSE 5983/4 prevents identification of the posteriormost point. The line appears to be mainly disrupted by the oblique surface stripes of the body, although it also can clearly be seen to be traversing the stripes with no interruption to the line. It is unclear if the non-continuous nature of this line is a true topology or an artefact of preservation. SEM-EDS analysis demonstrated this line is associated with P (Fig. 10D, LL), but no other associations were recovered.
'Caudal' lobes. The posterior terminal end of Lasanius is split into two 'caudal' lobes. The ventral lobe is significantly longer than the dorsal lobe, usually preserved as a solid dark imprint, and exhibits a strong to very strong downwards slope (Fig. 1A). The dorsal lobe is more faintly preserved, often seen with a diagonal striping pattern of thin parallel lines along the whole length, that connect the body of the organism to the posterior end of the lobe (Fig. 11C, D, CSt). Rarely, a faint suggestion of material between the lobes is present, seemingly representing a tissue connecting the dorsal and ventral lobes to give a fork-like appearance. SEM-EDS mapping of this F I G . 9 . Oblique stripes. A-B, BGS GSE 5984: A, visible light photograph of specimen immersed in water, with polarized light; red outline indicates scan area shown in Figure 10; B, interpretative illustration. C-D, BGS GSE 5983: C, visible light photograph of specimen, immersed in water, with polarized light; D, interpretative illustration. E, visible light photograph of BGS GSE 5985. F, visible light photograph of NHMUK P11056. Abbreviations: Ch, paired 'chain' apparatus; DAL, dorsal axial line; Ds, dorsal medial scutes; LL, lateral line; St, oblique stripes; Rd, paired 'rod' apparatus; VAL, ventral axial line. Scale bars represent 10 mm. area failed to showed a clear correlation between the stripes in the dorsal caudal lobe and the distribution of the elements detected.
External ventral material. Below the ventral edge of the main torso, a thin line is occasionally present, and is often associated with a patch of ill-defined material (Fig. 11,EVM). This line is preserved in a similar colour to the body imprint and runs roughly parallel, but does not appear to connect to the main body. In dorsoventral orientations the line appears to be within the body outline and is, on occasion, paired with a similar line of tissue. The start of the line varies between specimens, but is usually constrained to the posterior two-fifths of the specimen, ending at the origin of the caudal lobes. The line is seen in specimens with varying levels of preservation, but is not consistently seen in specimens with the same degree of preservation. EDS mapping of this region failed to establish correlation with the distribution of any elements.

Taphonomy
The lack of correlation of body length and width suggests the two morphotypes (thick and thin) may be products of taphonomic processes, as suggested by previous authors (Stetson 1927; Van der Brugghen 2010). Softtissue organisms generally do not undergo lateral expansion due to compression or decay (Briggs & Williams 1981), but decay driven collapse can create imperfect fidelity, and so variation, in the preservation of the body outline (Fig. 12). The lack of superimposition of the laterally placed features of Lasanius indicates that some laterally oriented specimens have undergone some level of distortion.
The morphological similarities of Lasanius to the anaspids, and possibility of non or only weakly mineralized armoured scales, suggest that Lasanius could represent a juvenile stage, especially as there is some evidence to support similar taxa, such as Euphanerops, developing armour as they grew (Janvier & Arsenault 2007). However, the upper limit for body size in Lasanius is approximately 21 cm, exceeding the size ranges F I G . 1 0 . Oblique stripes, BGS GSE 5984. A, BSE image of the scan area indicated in Figure 9A. B-D, false colour SEM-EDS chemical maps: B, carbon; C, calcium; D, phosphorus. Abbreviations: St, oblique stripes; LL, lateral line. False colour maps have been enhanced to improve visibility. Scale bars represent 1 mm. of all anaspids known from body fossils, and with no evidence for increased mineralization with size, it is unlikely that Lasanius is a juvenile of any known anaspid.

Model for homology
Lasanius was undoubtedly a bilaterian, but without interpretation it lacks unequivocal evidence for any of the small number of vertebrate, or even chordate, synapomorphies that are expected to be preserved (Janvier 2001). Nevertheless, Lasanius shares many morphological similarities with unambiguous vertebrates. The dorsal scute series is highly comparable to that seen in the Anaspida, while within the 'head' region the location of paired, and unpaired topological structures match those identified in the vertebrates Euphanerops and Jamoytius. The mineralized tissues of Lasanius suggest a calcium phosphate, or apatite, composition, and although this material is not exclusive to vertebrates (e.g. Bentov et al. 2016), it is the key component of skeletal bone. It is prudent therefore to assume that these shared characteristics are indicative of a shared evolutionary history, and the chordate-vertebrate model is the most suitable for initial identification of homologous structures. A summary of the interpretations of topology in light of this comparison model can be found in Table 1.  Anterior 'head' structures Dark stain as mouth. The size, position and location of the dark terminal stain (Fig. 4, DTS) is consistent with the positioning of the oral opening of Jamoytius (Sansom et al. 2010b) and shares similarities with the enigmatic 'head stains' reported in Euphanerops (Janvier & Arsenault 2007). The high preservation potential suggests that it is resistant to decay. The dark area is correlated with carbon, while it is the internal area that has a high correlation with Ca distribution (Fig. 4C, D, DTS). This pattern indicates a possibility of an internal calcified component with a carbonaceous surrounding. This is further supported by the consistency in preservation, shape and position, which identify it as a fairly rigid structure that resisted distortion. The rigidity of the oral opening provides an explanation for the suspected twisting and distortion of the head region, indicated by the preserved position of the paired circles and masses (Figs 4, 12). A large rigid structure at the anterior of the head would lead to distortion during collapse due to movement of the attached soft-tissue.
The circular-oval morphology, in particular the ring shape, is reminiscent of the morphology seen in decaying lamprey oral cavities, where annular cartilage creates an 'inner circle' surrounded by soft tissue (Sansom et al. 2011). Annular cartilage is also a decay resistant tissue (Sansom et al. 2011) and so the internal component of the oral cavity of Lasanius may be a similar calcified cartilaginous tissue. Some caution must be taken with this interpretation as there has yet to be unequivocal evidence for cartilage preservation with a carbonaceous film style of preservation (Sansom et al. 2010b). It is unknown if the inner area of this oral opening contained plates or the teeth-like structures seen in lampreys or conodonts (Purnell et al. 1995;Manzon et al. 2015). There is no evidence T A B L E 1 . Summary of the proposed identity of the topological features of Lasanius highlighted in Figure 3. for a jaw or jaw-like structure to allow movement so it is unlikely that Lasanius was able to 'bite', or manipulate the shape of this feature greatly. The dark carbon stain may also represent an area of pigmentation, as dark carbonaceous areas are welldocumented as containing the pigment carrying melanosomes in early vertebrate fossils (e.g. Clements et al. 2016;Gabbott et al. 2016). However, SEM-EDS mapping of this region failed to find a correlating distribution between copper (Cu) or zinc (Zn) the two key markers for melanosomes (Wogelius et al. 2011).
An alternative explanation for the distribution of calcium within the head region is the presence of a weakly mineralized head shield. When overlaid, the part and counterpart maps of BGS GSE 9713 appear to correlate, showing a constant, non-overlapping distribution of calcium across the head region (Fig. 13). This may have resulted from one continuous layer of a calcium rich tissue being broken as the slab the fossil sits in was split, and the parts of the tissue divided onto either side of the split. However, this interpretation must be taken with extreme caution for several reasons. Firstly, this pattern has only been recovered in one specimen so far, and not been tested in others. Secondly, head shields have not been suggested previously for Lasanius, and they are unknown in euphaneropids and lampreys. While anapsids do have dermal armour, they do not have a solid head shield, but one formed from many smaller plates (Blom & M€ arss 2010). This would make the presence of a solid head shield somewhat unique in Lasanius.
Paired circles as eyes. The shape and the location of the large paired anterior circular structures is consistent with the optic organs of vertebrates (Fig. 4, PLC). EDS analysis suggests a carbonaceous tissue surrounding a Ca-rich internal structure (Fig. 4C, D). The presence of a dark C ring around the paired optic region may support hypotheses inferring a cartilaginous sclerotic ring (see Ritchie 1963 andVan der Brugghen 2010). Yet, differentiation between pigmentation and cartilage can be problematic without further analyses (e.g. Gabbott et al. 2016), so confident support for a sclerotic ring is not currently possible. While the optic region does appear to be a rigid structure, it is rarely found misshapen, the rigidity is more likely to be provided by the calcium-rich internal structure or an unpreserved optic structure, such as a lens. Lenses have been described by Parrington (1958) (although potentially in error, see below), but were not observed during this study. However, complex eyes including a lens have been theorized as the ancestral state in vertebrates (Clements et al. 2016;Gabbott et al. 2016) so it is not unreasonable to consider their presence in Lasanius. While these structures are undoubtedly the eyes of Lasanius further analysis is needed to give a detailed insight into the nature of the organ.
Small masses as otic. Using the paired lateral circles as landmarks within the head, the small white paired posterior masses probably represent part of the otic structures of Lasanius (Fig. 4, SPM). Otic regions of vertebrates contain calcified components and the composition of these organs is diagnostic: in gnathostomes the calcified components (otoliths) are most commonly made of calcium carbonate, while the homologous otolith-like structures of cyclostomes are composed of calcium phosphate (apatite) (Maisey 1987). Our analyses find a strong concentration of calcium in these masses, but no evidence of phosphorus (Fig. 4E, F, SPM). We do however find strong correlations between phosphorus and calcium in other parts of the anatomy of Lasanius, as such, it is unlikely that these masses were apatitic (i.e. the cyclostome condition).
The interpretation of the masses as the otic region has been previously proposed by Smith (1957), Ritchie (1963) and Van der Brugghen (2010) as an alternative to Parrington (1958), who suggested that the masses were the remains of a hard lens from the eye. Smith (1957) made reference to two specimens with associated labyrinth structures. Unfortunately, no full description is available, and we were not able to identify the specimens nor similar structures in the specimens examined. We find no support for the possibility that the masses represent external openings. While some ostracoderms, and modern lampreys, do have paired nasal openings these are usually located on the dorsal medial line (Janvier 2008), and the composition, and preservation of these features as solid masses, is inconsistent with an opening.

'Chains and rods' apparatus
The chemical maps of these unique, three-dimensionally preserved structures show a clear association of calcium and phosphorus (Fig. 6B, C). This association indicates a potential original composition of calcium phosphate (apatite), the dominant compound found in vertebrate bone (Edwards et al. 2013). EDS point analysis data further support this interpretation, recovering a strong association of P and Ca with the mineralized 'rod' areas ( Fig.  6D, E) and with each other (Fig. S10). It also indicates a clear distinction between 'rods' and the matrix, with less Ca and almost no P detected in the matrix. While even at the highest concentrations there is notably less P and Ca than would be expected in pristine apatite (1.2% and 23.5% compared to 18.6% and 39.9% respectively), this is not unexpected nor does it undermine the compositional interpretation. The age of the fossil should be considered, some loss is expected due to the natural diagenetic processes which would reduce the signal. As no additional specimen preparation was permitted for SEM-EDS analysis, any remaining overlying matrix would partially obscure the fossil and filter out the low energy x-rays. For example, 5 lm of iron oxide would absorb approximately 65% of the Ca fluorescence at c. 3.7 keV, but would absorb 90% of the P emission at c. 2 keV (Henke et al. 1993). The combination of the high preservation potential of P relative to Ca by the embedding matrix (because of the lower energy of P fluorescence), and due to the fact that we also typically see preferential P loss from hard tissue as a function of age in fossil specimens, the low P to Ca ratios in these regions are unsurprising. However, enough P is present to allow us to unambiguously distinguish the chemistry of the 'rods' from the matrix (Fig.  6E).
The high preservation potential, texture and appearance, and the geochemical data strongly supports the hypothesis that these structures were biomineralized tissues in the living animal, composed of material from the apatite family. This may either be as the apatite mineral phase sensu stricto or as secondary mineral phases most probably derived from an apatitic precursor based on the relatively high concentrations of P and Ca with typical associated trace elements such as Zn and Sr. In the case of postulated fossil bone the assignment of apatite affinity may be strengthened by consideration of structural or anatomical details of the specimen. As the other mineralized structures in Lasanius share similar structure and geochemical features, it is not unreasonable to infer they also shared the same chemical composition. , suggesting that this conformation is an ancestral condition of all vertebrates. The deeply concave and almost semi-circular base of each unit in Lasanius (Fig. 5, Ch), is consistent with the circular morphology of these branchial openings. A similar morphology is also present in the plates of Birkenia and other anaspids, which create a round or horseshoe shaped branchial openings (Stetson 1928;Blom 2012). With a shared location and morphology, it seems likely that the 'chain' shares functional homology too, with the units demarcating the branchial openings in Lasanius, as previously proposed (e.g. Kiaer 1924;Simpson 1926;Stetson 1927;Ritchie 1963). , may indicate a shared function, although homology between the structures has been questioned (Blom 2012). In anapsids, the tri-radiate spines are proposed to have acted as fin supports, either for a ribbonlike fin connected posteriorly to a paired anal spine (Janvier 1987) or for a separate pectoral fin, as seen in Rhyncholepis parvula (Ritchie 1980). In Lasanius, there is no evidence of an anal spine, and only the posteriormost unit unequivocally has a long projection. This is opposite to the condition in most anaspids, where the tri-radiate spine is located in a more anterior position. While it has been suggested that all the units in Lasanius had a similar long posterior projection (e.g. Stetson 1927) or even an extra 'missing' fourth projection, evident by damage or holes in the units (Parrington 1958), this study failed to find clear evidence of either of these morphologies. Some specimens do demonstrate damage or a bluntness to this area, (Fig. 5B, Rd) but the majority of specimens have a solid smooth surface, with holes only rarely observed, and in inconsistent locations, suggesting a cause other than the loss of additional projections. As the posterior elongated edge is routinely preserved, it would not be unreasonable to expect the preservation of other projections, but no additional elongated projection is preserved in any specimen used in this study, so the extrapolation of a missing or additional long spine cannot be supported. There is no sign of ornamentation or texture on the surface of the 'rods'. This is potentially evidence for a subdermal endoskeleton position (e.g. Stetson 1927), which would be consistent with a function of internal fin supports. If the 'rod' series are the fin supports of Lasanius then the lack of anal spine would suggest a fin-shape similar to the short pectoral fin of Rhyncholepis (Ritchie 1980) rather than a long ribbon-like fin. However, no finlike material has been found preserved in the location of the 'rod' series, even though fin-like material is preserved in the caudal region. Likewise, no fin material is preserved in the multi-spine region in Pterygolepis which raises doubts about its function as a fin support in that genus (Ritchie 1980). The lack of fin-like material and additional projections, alongside the posterior position of the elongated projection, reduces the similarity of the 'rods' of Lasanius and the fin-bearing anaspid spines. This, in turns, reduces support for the 'rod' series as fin bearing structures.
Alternatively, it may be that the 'rod' series is related to respiration and branchial support (e.g. Van der Brugghen 2010). Other authors (Stetson 1927) have suggested that the 'rod' series is a precursor to gill arches, which is difficult to determine without a clear phylogenetic position. The series does share a superficially morphological similarity to branchial baskets seen in Euphanerops (assumed to be mineralized), (Janvier et al. 2006;Janvier & Arsenault 2007), as well as in lampreys, although the lamprey branchial baskets lack bone (Janvier 2008;Potter et al. 2015). However, the 'rod' series is positioned behind the branchial openings in Lasanius, unlike the condition seen in the branchial basket arrangement in which each 'arch' is associated with an opening (Potter et al. 1982). The presumed internal nature of this series may point to the 'rods' as acting as a supporting structure for additional branchial openings along the body, but no other evidence is currently available that would support this hypothesis.
While ultimately the function of the 'rods' is unknown, it may be that they provided some combination of the theories proposed. Although it does not seem that the 'rods' were true fin-supports, their position on the body, with the triangular base angled slightly away from the torso, may have provided a similar function to fins, giving Lasanius stability in the water. Similarly, although they are not identical to branchial baskets seen in other taxa, the long dorsal projections may have had some interaction with the respiratory system of Lasanius, potentially in a passive way as it swam.
It is also not totally apparent whether the 'rod' series was internal or external. They lack the ornamentation associated with external elements, but the angle they form on the body would lead to some projection away from the body. No evidence for tissue covering is found (e.g. skin) but this does not preclude that they may have been covered in life. It is most likely that both the 'rod' and 'chain' series were external, due to their similarity to the biomineralized structures seen in anaspids, and the lack of internal skeletal elements in other early ostracoderm vertebrates.

'Trunk' structures
Dorsal scute series. The robust nature of the scutes along with the strong association with Ca and P suggests a composition of calcium phosphate. Mn is also highly constrained to the scutes (Fig. 8C). Mn substitutes for Ca in the phosphate phase (Pan & Fleet 2019) so its presence is consistent with an apatite composition. Mn is a vital element, important for osteoclast and osteoblast activity, controlling growth and reabsorption of bone in vertebrates (Strause & Saltman 1987), so would be expected in an active bone-like tissue. However, Mn is common in geological environments, and geological processes can lead to the precipitation of Mn as an inorganic oxide on fossilized tissues (e.g. Egerton et al. 2015). The lack of Mn mineralization in cracks in the matrix or along the bedding plane suggests that it is most likely to be original to the tissue, and not a postmortem substitution.
The composition, position and morphology of the scutes show considerable similarity to the dorsal ridge scales seen in true anaspids, although they appear to lack the ornamentation reported in some genera, such as Birkenia (Blom et al. 2002). The difference in shape along the dorsal series is also consistent with anaspid dorsal ridge scales, which has been linked to the increase of lateral movement in the posterior of the body (Blom et al. 2002;Blom 2012).This combined with the seeming lack of connection between the scutes, which would increase flexibility through the dorsal line, which may have implications for the movement and manoeuvrability of Lasanius.
Dorsal scute morphology has been suggested as a phylogenetically important trait, with a reduction in height and hook shape potentially indicating a more derived state in the anaspids. However, the evaluation of this character state can be subjective, and lack of a clear outgroup or a robust anaspid phylogeny means interpretations of this manner should be treated with caution (Blom & M€ arss 2010). The internal histological structure of anaspid scales has been used to challenge hypotheses of affinity and the acquisition of biomineralization in vertebrates (Keating & Donoghue 2016). The SEM and associated images of the scutes (Fig. 8) do not allow for the identification of growth-related structures, and so the exact pattern of growth remains unknown. However, they do show an apparent void-like area towards the base. Damage in this region is not uncommon in Lasanius, and can be seen to differing extents in Figures 1A-B, 7C and 9F. This may simply be a taphonomic effect, driven by the shape of the scutes where the thinner base proves less robust than the more solid tops. However, the void space (Fig. 8) is reminiscent of the cavities seen in thin sections of the body scales of the heterostracan Lepidaspis serrata (Keating & Donoghue 2016). In L. serrata, scales are odontodes formed by a crown of dentine and enameloid that overlays a pulp cavity. It is this cavity that produces a similar morphology to the void seen in Lasanius. If the void space of the scutes of Lasanius is a true feature, rather than taphonomic, then it might be indicative of a similar internal structure, potentially indicative that the scutes of Lasanius are also formed of individual odontodes.
Knowledge of the internal structure of the dorsal scutes of Lasanius is therefore crucial for a better understanding of their phylogenetic position, and for testing hypotheses as to whether their 'nakedness' is a derived state or not. However, attempts to undertake such work for this study were unsuccessful (as they have been historically: Traquair 1898; Van der Brugghen 2010). Specimens with the necessary three-dimensional preservation are rare in collections, which prevented further attempts.
The exact function of the dorsal scutes is unknown, but an obvious modern analogue is the stickleback fish (Gasterosteidae). These small teleosts use a row of modified dorsal fin supports as defence against predation (Hoogland et al. 1956). The scutes of Lasanius may also have provided a defensive weapon against predators, such as large eurypterids that shared their habitats (Ritchie 1963;Lomax et al. 2011), but this does not preclude other potential uses, such as hydrodynamic stabilizing similar to a true dorsal fin (Webb 1984;Fish et al. 2003).
Stripes: myomeres or scales?. The stripes of Lasanius are one of the more contentious features of this genus; hypothesized to be either myomeres (e.g. Traquair 1905; Simpson 1926; Van der Brugghen 2010) or weakly mineralized scales (e.g. Traquair 1899Traquair , 1905Kiaer 1924;Stetson 1927). Although several different approaches were tried during this study, no unambiguous evidence was found to support or reject either hypothesis (Table 2).
For mineralized scales, some relationship with Ca and P and the stripe pattern would be expected, but although these elements are present, there is no clear correlation (Fig. 10C, D). The resolution of the distribution maps is approximately 5 lm per pixel, so if the stripe pattern did exist it should be visible in the maps. However, if mineralization was only weak, resulting in relatively low content of Ca and P, it may not have been possible to resolve the stripe pattern. Therefore, the lack of correlation with Ca or P cannot be taken as evidence for a lack of mineralized scales. Examination under a light microscope of BGS GSE 5983/4 failed to find similar ornamentation and topology of the stripes previously described in other specimens as evidence for mineralized scales (Stetson 1927;Ritchie 1963). The overall texture of the fossil does differ from the surrounding matrix, and the stripes appear to be raised in comparison to the spaces separating them, but there is no discernible difference in surface texture between the stripes and the gaps at higher magnification. The troughs between stripes could be created by a lack of underlying material either by the shrinkage of myomere T A B L E 2 . Evidence for the two main alternative hypotheses for the identity of the stripe pattern of Lasanius.

Myomeres Mineralized scales
No Ca or P associated with stripes Lack of Ca or P may be due to resolution of analysis Possible C association Association of C not well restrained to stripes No ornamentation or topology unique to stripes Topology is unique to whole fossil and not seen in matrix No breaks or sharp lines in stripes No preserved muscle fibre seen Not composed of single units Preservation pattern inconsistent with muscle decay blocks that occurs during decay, creating separation and gaps (Briggs & Kear 1993;Sansom et al. 2013), or by the decay of tissue more labile than mineralized scale armour. Black material observable in the stripes may be the same small, black units identified by Parrington (1958), used to indicate denticles and skin armour, but they do not match in description or distribution. Although patchy, this black material seems to be part of a continuous line, with topology determined by the underlying texture of the fossil, and no evidence was found to support the presence of identifiable individual units. This suggests it represents a continuous band of material, which may support a muscle identity rather than armour, but closer inspection failed to find muscle fibres or similar within the material. Comparison with other taxa that have similar stripe patterns interpreted as mineralized scales (e.g. Jamoytius; Sansom et al. 2010b) showed that the stripes of Lasanius differ in several key areas. The stripes of Jamoytius show signs of fracture and displacement, indicating their rigidity (Sansom et al. 2010b); this is not seen in Lasanius where the stripes are continuous and changes in direction are smooth. Likewise, Lasanius lacks both the same texture and relief reported in Jamoytius, and, critically, lacks the correlation with Ca and P distribution used as evidence for mineralization in Jamoytius (although see above). Clear differences are also apparent in the dermal armour patterns of anaspids. Anaspid dermal armour is formed from many long, overlapping, biomineralized scales, arranged in vertical stripes that form a horizontal 'V' shape across the body (although this pattern becomes less apparent towards the posterior section of Birkenia specimens). Unlike the sigmoidal, almost straight, stripes of Lasanius, the anaspid 'V' pattern is not continuous but broken into five defined sections along each stripe, and the change in orientation of the scales is distinct (Blom et al. 2002). Additional dissimilarity is evident when considering the morphology of the individual anaspid scales; anaspid scales are complex, often with tubercles, spines and ridges which Lasanius seems to lack (Blom et al. 2002;Keating & Donoghue 2016). Although the dorsal scutes exhibit great similarity with the scutes of Lasanius, there is little to support Lasanius stripe pattern being produced by a similar complex of biomineralized scales as seen in the anaspids.
While no unambiguous evidence supports the stripes as preserved mineralized scales, the support for the muscle hypothesis is also weak. Muscle blocks do separate during decay, creating a stripe pattern, and the muscle fibre pattern is lost quickly to decay so it is not unexpected that it is not preserved or observed (Briggs & Kear 1993). But muscles have also been shown to decay along the ventral surface first (Briggs & Kear 1993;Sansom et al. 2013), which is in contrast to the Lasanius fossils where the majority of specimens examined preserve the stripe pattern ventrally. Preservation of ventral muscle would indicate a low-level of pre-fossilization decay and produce the expectation of dorsal muscle preservation, which is exceptionally rarely seen (and never without ventral preservation too). Again, the shape of the stripes raises doubts about their identity as muscles. When preserved in their entirety, they are a gently curved sigmoidal shape. Decay can quickly cause loss of the classic 'W' shape of myomeres to 'Z' then 'V' (Sansom et al. 2013) but it is unclear how the almost-straight sigmoid shape could be produced by a 'softening' of these myomere patterns or how this could be uniformity achieved.
One recurring hypothesis that combines the ambiguous lines of evidence is that Lasanius scales mineralized as the specimen matured, so stripes are either muscle or scales dependent on the size of the specimen (e.g. Kiaer 1924;Simpson 1926). But, comparison of body length measurements fails to find a significant relationship between the presence of stripes and the total length of Lasanius (ANOVA on complete species p > 0.5, and p > 0.05 on incompletely preserved specimens). While the largest specimen analysed (BGS GSE 5983/4) does preserve the best representation of the stripes, this is an outlier within the data and the pattern is not seen in other specimens.
It is unclear exactly what the stripes of Lasanius were; there is no positive evidence for either of the competing hypotheses. Indeed, the myomere or scales question may be a false dichotomy, and there is a need to examine alternative hypotheses. Potentially the stripes could be preservation of pigmentation, either created during decay as a replication of the position of the muscles under the skin (Sansom et al. 2010a(Sansom et al. , 2013, or as a record of a true striped patternation in life, as seen in other Palaeozoic vertebrates (Gabbott et al. 2016). However, we found no correlation of the key chemical markers Cu or Zn within the stripes that would be expected to be present if they were melanosome pigmentation (Wogelius et al. 2011), leaving this hypothesis unsupported. A simpler explanation is that the stripes and zonation of C (Fig. 10B) may have been created via a 'peak and trough' corrugated pattern, formed by undulating areas of thick and thin tissue. This may have been driven by underlying soft-tissue elements, which in turn could provide support to muscles as the cause of the stripes. However, while this may go some way to explaining the loss of fibres and texture, it does not account for the dissimilarity between the stripes and usual vertebrate muscle patterns. Muscles need not be the only cause of thick and thin areas. It may also be due to thickened dermal elements, for example, the keratinized tubercles and skin of many species of catfish, where the additional lipids provide protection against pathogens and mechanical damage (Kumar Mittal & Whitear 1979). However, keratin preservation is controversial (e.g. Saitta Schweitzer et al. 2018) and we find no evidence for a correlating zonation of the potential markers of Ca or P and the stripes.
With the lack of SEM-EDS evidence to support for mineralization, it seems most likely that the stripes of Lasanius correlate to a dermal or subdermal soft-tissue structure. However, without clear evidence to support the possible hypotheses the identity of this structure remains unknown.
Axial lines, guts and notochord?. The two major longitudinal lines appear to preserve well, and be fairly decay resistant. They are obscured by other features, such as the 'rod' structure, suggesting they lie beneath those structures and are internal. Three main possibilities exist for internal axial lines in early vertebrates: notochord, nerve chord and intestinal tract (Sansom et al. 2013).
The dorsal axial line extends from the head of the animal to the end of the ventral caudal lobe, consistent with the path of a notochord or dorsal nerve cord (Figs 1,11,VAL). Notochords are formed of highly decay-resistant tissue (Briggs & Kear 1993;Sansom et al. 2010a), which matches the high preservation potential of the dorsal axial line. Notochords decay from the centre, often leaving just the sheath behind (Briggs & Kear 1993;Sansom et al. 2010a). In the chordate Branchiostoma, this sheath can be observed as two defined parallel lines, which may account for the void-like preservation seen in some Lasanius specimens (Briggs & Kear 1993). A void space is seen above the dorsal axial line, which is a similar pattern to the separation of the notochord from the supporting tissue seen during decay in Branchiostoma (Briggs & Kear 1993).
Identifying the dorsal axial line as the notochord indicates that the thin strip of darker material near the dorsal aspect of the line is most likely to be the dorsal nerve cord (Fig. 11C, Dn). It is consistent with the colouration and location of a pigmented nerve cord seen in taphonomic experiments by Briggs & Kear (1993). Additionally, the nerve cord has a similar decay-resistance as the notochord (Sansom et al. 2010a), so would be expected to be present when the notochord is preserved.
The ventral axial line is consistent with the position of the gastro-intestinal tract in vertebrates. This is supported by its posterior limit being on the ventral surface of the body, anterior to the caudal lobes. This location, and the dark colouration of this feature, is comparable to other fossilized digestive tracts identified in taxa such as Euphanerops (Janvier & Arsenault 2007).
Caudal lobes of the tail. The identification of the notochord terminating in the well-preserved ventral caudal lobe supports the previously assumed hypocercal condition of the tail (e.g. Stensi€ o 1927). The fainter preservation of the dorsal lobe may indicate a more delicate tissue, but may also be due to the lack of notochord. The faint evidence of material connecting the two lobes indicates the presence of a fin web between the two lobes, as described by Bulman (1930). The diagonal striping pattern seen in the dorsal lobe (Fig. 11C, CSt) is consistent with the presence of fin supports, a vertebrate synapomorphy. Chemical mapping of this area failed to find any correlations leaving the composition of the 'supports' unknown. The preservation of dorsal lobe tissue is common (when the tail is present), which correlates with experimental data suggesting that fins with supports are less labile than other tissues (Sansom et al. 2013).
Miscellaneous other structures. Some structures are poorly characterized, observed in a very limited number of specimens and difficult to homologize with any confidence. The position and shape of the 'unpaired median ring' (Fig. 4G, UMR) suggests two possibilities: a single nasal opening or a pineal organ. A single nasal opening is well documented both extant and extinct organisms (e.g. lampreys, osteostracans and galaespids;Gai et al. 2011;Janvier 2015), while lampreys and anaspids both have either a pineal gland or a pineal opening in a similar dorsal location (Cole & Youson 1982;Blom & M€ arss 2010).
The identity of the 'black dorsal material', seen most clearly in the head, is unclear. It is an uncommon trait, not seen consistently in specimens with similar styles of preservation. This raises doubt over its status as a true morphological feature, and it may be a taphonomic artefact. Comparison of slightly different lateral orientations reveals that the dark line is dorsal and medial in position, but a similar line, possibly another artefact, is observed in a ventral position in some specimens. The 'dorsal line' black material may be the same feature described as a 'dorsal rod' by Van der Brugghen (2010).
Of the axial lines observed, the dorsal and ventral lines are well characterized and homologized above. A third longitudinal line (lateral line), distinct from the dorsal and ventral, is observed in only one specimen BGS GSE 5983/4 ( Fig. 9A-D, LL). It is unclear whether it is an internal or external structure. There are two potential options for such a feature: blood vessel or lateral sensory line. The preserved material includes phosphate (Fig.  10D, LL) and is similar to a black line in Euphanerops, described tentatively as a blood vessel (Janvier & Arsenault 2007). The possibility of phosphatization in Euphanerops was raised, a known phenomenon in blood vessel preservation, although in a 'white', not black line (Janvier & Arsenault 2007). If internal in Lasanius then this structure may be homologous with lines in Euphanerops proposed to be blood vessels. Superficially, this line in Lasanius is reminiscent of the lateral sensory line seen in extant aquatic vertebrates, and present in heterostracans and other ostracoderms, that may be a plesiomorphic condition of vertebrates (Janvier 2001). The noncontinuous nature could be indicative of the pore canal system associated with the lateral line, but this is speculative without knowledge of the microstructure (Denison 1966). This hypothesis would not explain the association with P, and would need clarification of the full position and length of this line, and if the non-continuous nature is reflective of the in-life condition or an artefact produced during the fossilization process. The limited number of specimens preserving this feature restrict exploration, so, for now, the identity of this structure cannot be clarified.
Absence of anal fin. 'External ventral material' (Fig. 11, EVM) is identified in the expected location for an anal fin and seems to share a morphology similar to the anal fins of anaspids (Blom & M€ arss 2010) and the lateral fin folds of Jamoytius (Sansom et al. 2010b). The anal fin of Lasanius has historically been reconstructed on the basis of phylogenetic bracketing with taxa that have clear fossil evidence for such a fin (e.g. Bulman 1930;Parrington 1958;Van der Brugghen 2010), yet, so far, no unequivocal physical evidence has been found. The phylogenetic instability of early vertebrates means that bracketing is an unreliable method for character reconstruction as it inserts an element of circularity to trait identification and phylogenetic placement.
There are several discrepancies that prevent confident identification as a fin. First, the preservation of this feature is inconsistent: specimens that preserve a high level of detail often do not preserve any posterior ventral material, while specimens with a fairly poor level of detail do. The lack of correlation between preservation fidelity and presence of the material suggests the absence is not due to preservation bias, but is a consequence of another factor. Second, the material is rarely connected to the main body imprint, as would be expected with a fin, and seen in other fins in Lasanius. In NHMUK P11051, the material is reminiscent of the small anal fin seen in Euphanerops, which may the 'low rounded structure' briefly mentioned in Lasanius by Ritchie (1963, p. 81). However, it has an indistinct outline and the proximity to the anus means the possibility that this is fluid expelled during decay cannot be dismissed (Briggs & Williams 1981). Finally, in BGS GSE 3893 and counterpart BGS GSE 3894, both preserved with a twisted orientation, a seemingly thickened paired area of tissue in the area can be seen. This is in line with the base of the 'rod' structure, suggesting that it was on the ventral surface in life. Although possibly indicative of paired anal fins, this morphology may also be an artefact of the decay process. The carcass may have split, and during collapse the resulting edges displaced and folded giving a false impression of additional material.
While similarly preserved taxa show clear evidence for an anal fin, the evidence for this feature in Lasanius is weak. Taphonomic processes are able to produce similar structures, and no unambiguous morphology consistent only with a fin has been identified. It seems more likely that the ventral material is a taphonomic artefact and that, for now, it must be assumed the absence of an anal fin is a true absence.

Phylogenetic analysis
The new data and interpretations of homology presented above were used to recode the character states of Lasanius within the matrix created by Miyashita et al. (2019) (Reeves et al. 2023). This revised matrix was then used to assess the phylogenetic affinities of Lasanius within the vertebrates.
Overall, the trees produced by the two analyses are largely similar, but the Bayesian analysis recovers more ambiguity in the stem-gnathostome topology. Maximum parsimony searches find Lasanius as sister-taxon to Euphaneropids (i.e. Euphanerops and Jamoytius) (Fig.  14A). These taxa were recovered as a sister-group to the Birkeniida anaspids Birkenia and Rhyncholepis, which are together positioned as an early diverging stem-cyclostome clade. Bayesian analysis also recovers the stem-cyclostome position, but with Lasanius outside of the Anaspida within a polytomy with the cyclostome clade and the Euphaneropids (Fig. 14B).
The stem-cyclostome affinity of Lasanius is not novel Due to this, and the many conflicting hypotheses on the affinity of Lasanius and allied taxa previously reported, it is important to assess the validity of the result found by this analysis: does the stem-cyclostome position reflect the true relationships of Lasanius? To answer this, the morphological dataset was subjected to Bayes factor analysis. This approach allows us to ask if there is enough evidence to reject or accept competing phylogenetic hypotheses (Jeffreys 1935(Jeffreys , 1961Kass & Raftery 1995;Dembo et al. 2015). The position of Lasanius within the cyclostome stem (without constraints applied) (Fig. 14B) had the highest marginal of likelihood (mll) of all the positions tested (À1327.263), initially establishing it as the best model for the data out of the tested topologies (Table S1). This best model was compared to the mll of the other constrained positions (including as a stemgnathostome, and position in and out of the Anaspida; see Methods, above) to create Bayes factors, which were assessed using guidelines detailed by Kass & Raftery (1995). In all cases, the Bayes factors provided enough evidence to reject all tested alternative topologies (Table  S1), demonstrating that the recovery of the phylogenetic affinity of Lasanius as a stem-cyclostome can be accepted with confidence.
Implications for trait evolution. The recovered topologies have important implications for the origin of biomineralization. Biomineralizing anaspids and a clade of less heavily biomineralized anaspid related taxa (Lasanius, Jamoytius, Euphanerops) as sister-group to all other cyclostomes suggests that the ancestral condition for vertebrates was some form of biomineralized dermoskeleton. This implies that the lack of biomineralization seen in modern cyclostomes is actually a secondary loss, as seen in other traits in this group (Gabbott et al. 2016). This is supported by the results of the ancestral state estimate (Figs S6, S7) which recover both biomineralization at the base of the vertebrate group, and a pattern of loss within the cyclostomi. Uncovering the precise timing and tempo of the evolution of biomineralization in vertebrates may require a greater knowledge of internal histological structures of Lasanius, such as the number and organization of different tissue layers. This could allow both a comparison to skeletal structures in other taxa (e.g. conodont elements; Donoghue et al. 2000) and an understanding of how the biomineralized tissues of Lasanius relate to the vertebrate plesiomorphic conditions established by Keating & Donoghue (2016). However, histological investigation of Lasanius has been unsuccessful, and no other of the Lasanius specimens surveyed for this study had the required preservation to undertake this work. Acquisition of more Lasanius specimens may help fill in some of the morphological gaps, both histologically and with more ambiguous features, but this is becoming an increasingly unlikely proposition. Lasanius is currently only known from localities that have been historically over-collected and mistreated, and as a result are now fairly unproductive. Understandably, in order to preserve what is left from further damage these sites are now designated Sites of Special Scientific Interest (SSSI), with highly restricted and controlled access (Scottish Natural Heritage 2019). Until the identification of new localities, or access to material held in private collections is available, many of the questions surrounding the anatomy of this curious little fossil will remain unanswered.

CONCLUSION
Our analysis has confirmed Lasanius is a vertebrate, possessing a head with a terminal ridged mouth and paired sensory organs (optic and otic), with evidence for a notochord, dorsal nerve cord, a digestive system, and non-terminal anus (Table 1; Fig. 15). The path of the notochord can be traced to the ventral caudal lobe, supporting the hypocercal state of the tail of Lasanius. For the first time, it has been demonstrated that the composition of the presumed mineralized tissues is most likely be calcium phosphate, the main component of vertebrate bone. Analysis of the otic region indicated that it contained a strong calcium content, but failed to find evidence for phosphorus. While a lack of phosphorus in the otic region is a diagnostic condition for gnathostomes, our phylogenetic analyses resolved Lasanius as a stemcyclostome, although with some slight variation seen between the methods used. Bayes factor analysis provided strong support for this stem-cyclostome position, and allowed the rejection of all other tested alternative hypotheses of affinity. The position of Lasanius (along with the anaspids and associated taxa) as a stemcyclostome has clear implications for the evolution of biomineralization in vertebrates. It indicates that the vertebrates had a biomineralized ancestral form, and the lack of biomineralization seen in modern cyclostomes is a result of secondary loss later in their history. This in turn implies that, rather than a gnathostome trait, biomineralized tissues (like bone) are a synapomorphy of vertebrata (Fig. 16).
The lack of resolution of the relationships seen in the trees (clearest in Fig. 14B) illustrates that there is still some instability in the tree. Many of the potentially informative taxa surrounding the vertebrate origin and the cyclostome-gnathostome divergence share similar bauplans, with little in the way of distinguishing features. Embryonic and transient developmental features (such as the embryonic origin of gills, which differs in extant cyclostomes and gnathostomes; Gillis & Tinswell 2017) may help to resolve some of the affinities of this group, but it is highly unlikely that these will be preserved (Donoghue 2017). The unambiguous identification of the presence or absence of key traits, such as true gills, would also be advantageous, but this has proven consistently problematic due to the confusing interplay of decay and preservation biases (Donoghue & Purnell 2009;Sansom et al. 2010aSansom et al. , 2011. Differentiating between cyclostome and gnathostome morphology to produce robust relationships in early vertebrates may thus require a resolution that is not currently available in the fossil record. Several features of Lasanius still remain ambiguous to varying degrees. Analysis of the striped torso pattern failed to provide unequivocal evidence for either myomere preservation or weakly mineralized scales. While other hypotheses were explored, it is still unclear what the stripes represent, however it seems most likely that they are the result of either a dermal or sub-dermal softtissue structure. Fin supports have been identified in the dorsal caudal lobe, and while suspected to be cartilaginous and possibly mineralized, analysis of their composition failed to produce any strong signal. An anal tail has often been postulated for Lasanius, but we found no evidence to support this claim; a lack of consistent preservation and varying morphology is more indicative of a decay or preservation artefact than a true anatomical feature. Finally, while the 'chain' series of the paired torso apparatus demarcate the location of branchial openings, the function of the unique 'rod' series is still unclear. The individual units do not appear adequate to provide fin support, and no fin-like tissue has been identified in this area. However, the position and biomineralized nature of the 'rods' suggests they may have provided some hydrodynamic advantage to Lasanius. Additionally, F I G . 1 6 . Implications for trait acquisition. Recovering Lasanius as a stem-cyclostome indicates the inferred origin of biomineralization to be on the vertebrate stem (green circle) instead of the previous inferred position (grey circle) on the gnathostome linage. A pattern of increasing loss of biomineralization can be seen on the cyclostome line, identifying an eventual total loss (red circle) within the Cyclostomi, demonstrating two alternate approaches to biomineralization in vertebrates. Illustrations not to scale. (All silhouettes except Lasanius and Anaspida courtesy of https://www.phylopic.org: ostracoderm, Philippe Janvier (vectorized by T. Michael Keesey) CC BY 3.0; others CC0 1.0). the proximity of the 'rods' to the branchial openings suggests that they may also have played a role in respiration, but the lack of internal preservation means no definite connection can be made.
This study has clarified many aspects of the anatomy of Lasanius, and most critically provided evidence for the presence of mineralized tissues composed of apatite. While factors inherent to the preservation style restrict our knowledge of the fine detail of the anatomy of these creatures, modern analytical techniques can provide a guide for interpretation of soft-tissue. Ultimately Lasanius remains a fascinating genus; its tantalizing mix of skeletal tissues on an otherwise non-mineralized body hints at the importance of this genus. Only by improving our knowledge of Lasanius and other associated soft-tissue vertebrates can we create a robust reconstruction of the evolutionary history of biomineralization in vertebrates we can be confident in.
Acknowledgements. The work contained in this publication contains work conducted during a PhD study by Jane Reeves supported by the Natural Environment Research Council (NERC) EAO Doctoral Training Partnership and is fully-funded by NERC whose support is gratefully acknowledged. Grant ref no is NE/L002469/1. The Bioimaging Facility microscopes used in this study were purchased with grants from BBSRC, Wellcome and the University of Manchester Strategic Fund. Special thanks to Roger Meadows for their help with the microscopy. Many other people deserve thanks for their help with access to collections and arranging loans of study material including: Eileen Callaghan (British Geological Survey), Neil Clark (Hunterian Museum), Stig Walsh (National Museums Scotland), Emma Bernard (Natural History Museum, London), Ann Ainsworth (Glasgow Museums Resource Centre), Matt Lowe (University Museum of Zoology, Cambridge) and Jason Sutcliffe (The Dick Institute, Kilmarnock). Additional thanks to Lewis Smith and Jon Fellows (University of Manchester) for all their help with SEM-EDS analysis, and to Gary Hoare for interesting discussions about Lasanius and the Scottish Silurian fossils beds. We would also like to thank the reviewers Dr Zerina Johanson and Dr Duncan Murdock for their thoughtful comments which helped improve this manuscript.
Author contributions. JCR accessed the specimens, planned and completed the analyses, and wrote the manuscript. RAW assisted in chemical data interpretation and manuscript preparation. JNK consulted on the phylogenetics and assisted in manuscript preparation. RS designed and supervised the project, and assisted in manuscript preparation.