Lateral line placodes of aquatic vertebrates are evolutionarily conserved in mammals

ABSTRACT Placodes are focal thickenings of the surface ectoderm which, together with neural crest, generate the peripheral nervous system of the vertebrate head. Here we examine how, in embryonic mice, apoptosis contributes to the remodelling of the primordial posterior placodal area (PPA) into physically separated otic and epibranchial placodes. Using pharmacological inhibition of apoptosis-associated caspases, we find evidence that apoptosis eliminates hitherto undiscovered rudiments of the lateral line sensory system which, in fish and aquatic amphibia, serves to detect movements, pressure changes or electric fields in the surrounding water. Our results refute the evolutionary theory, valid for more than a century that the whole lateral line was completely lost in amniotes. Instead, those parts of the PPA which, under experimental conditions, escape apoptosis have retained the developmental potential to produce lateral line placodes and the primordia of neuromasts that represent the major functional units of the mechanosensory lateral line system.


INTRODUCTION
In vertebrate embryos, many developmental processes take place in the presence of cell death (Glücksmann, 1951;Sanders and Wride, 1995). The 'programmed' character of such cell death events was confirmed by the formulation of the apoptosis concept (Kerr et al., 1972). In continuation of this ground-breaking concept, several types of programmed cellular demise are currently distinguished with caspase-driven apoptosis being the major type activated during ontogenetic development (Duprez et al., 2009). Yet we still know remarkably little about the exact ontogenetic functions of apoptosis, especially during early embryogenesis. The present study addresses this issue with the example of placode morphogenesis in the posterior placodal area (PPA) of mice.
The multiplacodal PPA is located at the posterior end of the panplacodal primordium and, in amniotes, gives rise to at least three pairs of epibranchial placodes (Fig. 1A,B). These are situated at the dorsal aspects of the branchial arches, and produce visceral sensory neurons of the distal ganglia of cranial nerves VII, IX and X, respectively, that serve roles in the detection of baroreceptive, osmoreceptive or chemoreceptive information from many visceral organs (Baker and Bronner-Fraser, 2001;Baker et al., 2008;Schlosser, 2006Schlosser, , 2010. Other derivatives of the amniote PPA are the otic placode and, in birds, the paratympanic placode which probably provides a barometric or altimetric measurement system (O'Neill et al., 2012;von Bartheld and Giannessi, 2011). The currently held consensus is that only the anamniote PPA is capable of developing lateral line placodes which produce mechanosensory neuromasts or electrosensory ampullary organs unique to aquatic vertebrates (Ghysen and Dambly-Chaudiere, 2004;Northcutt, 1992;Schlosser, 2002aSchlosser, , 2006. Three preotic and three post-otic pairs of lateral line placodes represent the primitive condition. However, subsets of lateral line placodes have gone astray, to different degrees, in different gnathostome lineages (Northcutt, 1992(Northcutt, , 1997Schlosser, 2002aSchlosser, , 2006, and their order of ontogenetic appearance reveals interspecies variability (Northcutt and Brändle, 1995;Schlosser and Northcutt, 2000;Fig. 2C).
During the morphogenesis of epibranchial placodes, embryonic mice demonstrate functionally unexplained apoptosis in three main loci Knabe, 2013, 2017): (1) at the ventral margin of the otic pit [embryonic day (E) 9], (2) between the detachment site of the otic vesicle and the epibranchial placodes 1, 2, and/or 3 (E9.25-E9.5, peak period of apoptosis, Fig. 1A), and (3) in peripheral parts of the three mature epibranchial placodes (E9.5-E11.5). Surprisingly, these apoptotic events predominantly eliminate ectodermal cells that express the general placode marker Six1, a member of the Sine oculis homeobox (Six) family of transcription factors (Washausen and Knabe, 2017). We therefore aimed to determine the developmental potential of these physiologically eliminated placode precursor cells by pharmacologically inhibiting apoptosis in cultured mouse embryos. It turned out that, contrary to previous assumptions, amniotes have retained the capability to produce morphologically and molecularly typical lateral line placodes as well as the primordia of neuromasts. Available evidence suggests that apoptosis may eliminate vestigial lateral line placodes also in other amniotes [Washausen et al. (2005); our unpublished observations in chick embryos]. Our findings further support the hypothesis that lateral line placodes may constitute the default fate of the PPA.

Molecular signature of lateral line placodes and neuromasts in Q-VD-OPh-treated embryos
Lateral line placodes of Q-VD-OPh-treated mice also exhibit the molecular signature of PPA derivatives (Schlosser, 2006(Schlosser, , 2010. Firstly, they express the general placode marker Six1 ( Fig. 3B; Fig. S8I-K). This retrospectively clarifies why apoptosis, seemingly located 'interplacodally' between the otic and epibranchial placodes of in utero-developed mice Knabe, 2013, 2017) (Fig. 1A), predominantly removed Six1 immunopositive (Six1 + ) cells (Washausen and Knabe, 2017). Secondly, they spring from paired homeobox (Pax) 2 and/or Pax8 expressing parts of the PPA but, under experimental conditions, lose protein expression of these specific posterior placode markers shortly thereafter (Fig. 4A,B). Thirdly, they develop in tandem with the otic placode, as is apparent from their common expression of Sox10 (Schlosser, 2010), a member of the group E Sox (sex determining region Y-box) family of transcription factors. Prior to the peak period of PPA apoptosis, this common expression domain was observed in both control and Q-VD-OPhtreated embryos. Furthermore, Sox10 + epithelial rosettes indicate the onset of neuromast formation (Fig. 4C,D). During the peak period of PPA apoptosis, control embryos reveal disorganized Sox10 + epithelia, decreased Sox10 immunoreactivity, and massive apoptosis in the positions of 'dormant' lateral line placodes (Fig. 4E,F). In contrast, lateral line placodes of Q-VD-OPh-treated mouse embryos escape apoptosis and demonstrate mosaics of viable Sox10 + /Sox10 − cells as well as intact Sox10 + neuromast primordia (Fig. S10C,D). Fourthly, some but not all lateral line placodes and/or neuromasts of Q-VD-OPh-treated mouse embryos express mosaics of Tbx3 + /Tbx3 − cells (Fig. 4G-I). This T-box transcription factor is specifically upregulated in the lateral line placodes of anamniotes (Schlosser, 2006(Schlosser, , 2010. Furthermore, Q-VD-OPh-treated (Fig. 4G) as well as in utero-developed mice (Bollag et al., 1994) express Tbx3 in the otic vesicle.
As a next step, we examined whether the lateral line placodes of Q-VD-OPh-treated mice express Sox2 which maintains the proliferative status and determines the neural fate of placode precursor cells (Schlosser, 2010). In zebrafish, Sox2 immunopositivity was found in neuromast mantle and support cells, with the latter serving as the source for hair cell replacement (Hernández et al., 2007). Correspondingly, neuromast primordia of Q-VD-OPh-treated mice show Sox2 + mantle and support cells (Fig. 5A,B). Altogether, Sox2 + cells were observed in 36 out of 66 embryos (Figs S1-S6F), and up to 24 Sox2 + cells were present in the lateral line placodes of a single individual (Fig. S2A).

Glücksmann
shape changes, 'histogenetic' ones that control cell numbers, and 'phylogenetic' ones which eliminate vestigial or larval organs. We here provide evidence that, in normally developing mouse embryos, lateral line placodes represent a new case of 'vestigial organs' which, unlike larval organs, undergo apoptosis largely prior to the onset of differentiation and, therefore, are particularly difficult to assess. Pharmacological inhibition of apoptosis in Q-VD-OPhtreated mice not only enables short time protection of these vestigial placodes but, far beyond that, allows development of distinct, morphologically and molecularly typical lateral line placodes which even produce neuromast primordia with ciliated hair cells. In this sense, our discovery refutes the long-held evolutionary theory that the whole lateral line sensory system was completely lost in amniotes (Northcutt, 1992;Schlosser, 2002aSchlosser, , 2006. Our results substantiate, but do not conclusively prove, the hypothesis that lateral line placodes may be considered the default fate of the PPA (Fritzsch et al., 1998;Schlosser, 2010). Originally, this hypothesis had been formulated to explain two observations made in anamniotes: (1) the apparent lack of specific lateral line placode inducers, and (2) the large distance that separates lateral line placodes from known signalling centres in the hindbrain and pharyngeal pouches (Schlosser, 2010). Accordingly, lateral line placodes of Q-VD-OPh-treated mice develop in the absence of externally supplied inducing signals. However, only very recently it was demonstrated that, in zebrafish, anterior lateral line placodes require Fgf signalling, whereas posterior lateral line placodes depend on retinoic acid that inhibits Fgf signalling (Nikaido et al., 2017). Nevertheless, the additional finding that ectopic activation of Fgf or Wnt signalling suppresses posterior lateral line placodes but increases the size of the otic placode is consistent with the hypothesis that ( posterior) lateral line placodes may represent the default fate of the PPA (Nikaido et al., 2017).
In view of the fact that culture of E9 mice should not markedly exceed 18-24 h to keep the embryos healthy (Martin and Cockroft, 1999), it is presently impossible to fully explore the developmental potential of vestigial lateral line precursor cells. Thus, for instance, we were unable to determine whether, in embryonic mice, rudiments of the lateral line system can generate migratory primordia, or whether neuromasts principally arise in unmigrated lateral line placodes. Nor can we comment on whether the observed rare occurrence of hair cell kinocilia may be due to culture conditions, or whether these supernumerary placodes normally do not complete the full lateral line differentiation pathway apart from rare exceptions. Both scenarios could explain why we found no evidence for the expression of Atoh1, which is thought to be required for hair cell differentiation (Sarrazin et al., 2006). However, Atoh1 may well have been expressed in the precursors of the few hair cells that differentiated.
Considering that, compared with mice, largely identical patterns of apoptosis were observed in the PPA of the primate-related Tupaia belangeri (Tupaiidae, Scandentia, Mammalia) (Washausen et al., 2005) and chick embryos (our unpublished data), apoptotic elimination of vestigial lateral line placodes may prove a widespread phenomenon among amniotes. Our results also shed new light on the possible developmental origin of the mechanosensory paratympanic and spiracular organs (Baker et al., 2008;O'Neill et al., 2012). Recent evidence suggests that the amniote paratympanic and the anamniote spiracular organs are homologous. Furthermore, a Sox2 + placode that resides dorsally adjacent to the first epibranchial placode could be identified as the source of the paratympanic organ in chicken embryos (O'Neill et al., 2012). Whether this previously undiscovered placode and, thus, paratympanic and spiracular organs develop independent of both lateral line and epibranchial placodes is not yet fully resolved. Our finding that, in embryonic mice, a latent conservation of mechanisms exists to develop lateral line placodes which are able to recapitulate at least part of the lateral line developmental program increases the probability of a lateral line origin of both organs. Modified versions of our experimental setting may provide an innovative possibility to further explore hitherto unknown developmental links between lateral line, otic and epibranchial placodes (Baker et al., 2008), and the molecular mechanisms that underlie placode morphogenesis in the PPA.

Whole embryo culture of mouse embryos
The culture was performed according to established protocols (Gray and Ross, 2011;Martin and Cockroft, 1999;New, 1978). The gravid uterus was removed and rinsed in a petri dish filled with Hank's balanced salt solution (HBSS; L2035, Biochrom, Berlin, Germany) at room temperature. Further dissection was performed under a clean bench (HERAGuard HPH 12/95, Thermo Fisher Scientific) using a M165 FC stereomicroscope (Leica, Wetzlar, Germany) and Dumont forceps (11251, 11255, Fine Science Tools, Heidelberg, Germany). The myometrium was torn off in order to expose the decidual swellings that ensheath the embryos. Without damaging the ectoplacental cone, the decidua was peeled from the conceptus and Reichert's membrane was removed from the yolk sac. For staging, each embryo was photographed (microscope camera: DFC450 C, Leica) using the Leica Application Suite (LAS) software, and head length was measured using ImageJ (Rasband, 1997(Rasband, -2016. According to these measurements, somite stages were determined with the help of mouse developmental tables (van Maele-Fabry et al., 1993). Embryos with 9-15 somite pairs [embryonic day (E) E8.5 to less advanced E9] were selected for wec only on the condition that neither their yolk sacs nor their ectoplacental cones had been damaged. Dissected embryos were transferred to the roller culture system using sterile transfer pipettes, and were randomly assigned either to Q-VD-OPh treatment or to one of the above specified control groups 1, 2 or 3 (2-4 embryos per bottle, 1 embryo/ml culture medium). At constant gas supply (40% O 2 , 5% CO 2 , 55% N 2 , gas flow rate: 25 ml/min), embryos were then incubated at 37.5°C and 30 rpm for 12, 18 or 24 h in the dark. For extended incubation periods (30 or 36 h in the dark), gas supply was modified to 70% O 2 , 5% CO 2 , and 25% N 2 (gas flow rate: 25 ml/min) following 22 h in wec. At the end of culture, embryos were transferred to HBSS (37.5°C) and carefully examined for viability and developmental stage using established morphological criteria and measurements (van Maele-Fabry et al., 1990, 1993. Embryos from all experimental and control groups were selected for further analysis only on the condition that an appropriate developmental status had been reached compared with the corresponding in uterodeveloped embryos.

Histological procedures
Embryos were removed from the yolk sacs and fixed in 4% paraformaldehyde in phosphate buffered saline (PBS: 137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 2 mM KH 2 PO 4 , pH 7.4) for 24 h at room temperature. Following two washes in PBS for 30 min, specimens were dehydrated via an ascending ethanol series. Hereby, embryos were preembedded in 1% low gelling-point agarose (Seakem LE,50001,Lonza,Köln,Germany) at the 50% ethanol level. Following complete dehydration, specimens were cleared in chloroform and subsequently infiltrated as well as embedded in Surgipath Formula 'R' paraffin (3801450, Leica). Finally, embryos were serially sectioned at 5 µm using rotary microtomes (RM2245, RM2265, Leica). Serial sections were consecutively placed on two sets of slides (Knabe et al., 2002) that were used either for Mayer's haematoxylin stainings (Romeis, 1948) or for immunostainings with different primary antibodies.

Immunohistochemistry
Deparaffinised and rehydrated sections were washed in Tris-buffered saline (TBS: 0.05 M Tris, 0.15 M NaCl, pH 7.4). Next, antigen retrieval was performed by high-pressure cooking in citrate buffer (0.01 M, pH 6). Following cooling to room temperature, endogenous peroxidase activity was blocked and sections were permeabilized by incubation with 1% H 2 O 2 and 0.3% Triton X-100 in TBS for 30 min. Thereafter (and in between all following incubation steps), sections were washed three times in TBS (5 min per wash). For anti-Ngn2 staining, reagents and procedures from the mouseon-mouse kit (BMK-2002, Vector Laboratories) were used to block the sections and to perform the primary and secondary antibody incubation steps. For all other immunostainings, primary antibodies were diluted in Dako REAL antibody diluent which contains background reducing agents (S202230-2, Agilent Technologies, Waldbronn, Germany). Immunoreacted sections were incubated with the appropriate biotinylated secondary antibody diluted 1:100 in TBS with 2% normal serum of the same species as the secondary antibody (S-1000, S-2000, Vector Laboratories) for 1 h at room temperature. For all immunostaining, sections were finally incubated with the avidin-biotin peroxidase complex (Elite ABC reagent, PK-7100, Vector Laboratories) for 1 h at room temperature. Peroxidase reactions were developed with 0.06% 3,3′-diaminobenzidine (DAB; D5637, Sigma-Aldrich) and 0.007% H 2 O 2 in Tris-HCl buffer (0.1 M, pH 7.6). Afterwards, sections were thoroughly rinsed in distilled water, counterstained with Mayer's haematoxylin (Romeis, 1948), and, following dehydration and clearance, embedded with DePeX mounting medium (18243, Serva, Heidelberg, Germany). Negative controls were performed by omission of the primary antibody and resulted in the absence of immunolabelling.

Histological analysis
Histological examinations were carried out for Q-VD-OPh-treated embryos (18 h wec: n=29, 24 h wec: n=57, 36 h wec: n=56) and for the corresponding control specimens (18 h: n=5, 24 h: n=47, 36 h: n=20). For orientation purposes, lower numbers of embryos were additionally examined at 12 h wec (Q-VD-OPh: n=9, DMSO: n=2) or at 30 h wec (Q-VD-OPh: n=1, DMSO: n=1). Individual Q-VD-OPh-treated embryos were numbered chronologically (#001 to #152). In utero-developed control embryos were taken from the developmental period between E9 and E10.5 (n=50). Embryo numbers in the Q-VD-OPh and control groups are consistent with, or considerably exceed, those employed in previous studies which were based on comparable methodologies (Huang et al., 2012;Lassiter et al., 2009;Massa et al., 2009). During all histological examinations, the group allocations were not blinded to the investigators. However, analysis was performed according to predefined, objective criteria, and findings were evaluated independently by the two authors.

Analysis of apoptosis
To verify the diagnosis of apoptotic cells in tissue sections, a multiparametric approach has to be applied (Stadelmann and Lassmann, 2000;Taatjes et al., 2008). Accordingly, previously published patterns of apoptosis in the PPA of C57BL/6N mice (E8.5 to E11.5, n=65 embryos) had been determined by using combinations of (1) immunohistochemistry against cleaved caspase-3, (2) the TUNEL (terminal deoxynucleotidyl transferase-mediated dUTP nick end-labelling) method, and structural diagnosis in haematoxylin-stained (3) paraffin or (4) 'semithin' (1 µm thick) resin sections Knabe, 2013, 2017). It turned out that the peak period of PPA apoptosis takes place in embryos of approximately 19-27 somite pairs that had been classified either to E9.25 (detaching otic vesicle exhibits a small pore) or E9.5 (a solid stalk connects the closed otic vesicle to the overlying ectoderm) Knabe, 2013, 2017). To provide a baseline for the present experimental work, a summary scheme depicting PPA apoptosis during the peak period was compiled from previously studied embryos Knabe, 2013, 2017) (n=44 body sides; Fig. 1A). In the current work, apoptotic cells were identified by immunohistochemistry with antibodies against cleaved caspase-3 as well as by established structural criteria (Häcker, 2000;Sanders and Wride, 1995). Furthermore, representative sections were subjected to TUNEL staining according to the protocol published previously (Washausen and Knabe, 2013).

Reconstructions of the PPA
Fine-grained schematic reconstructions were performed by transferring histological and/or immunohistochemical data from the PPAs of completely serially sectioned Q-VD-OPh-treated or control embryos to basic schemes of the embryonic head that had been generated in CorelDRAW X4 (Corel, Unterschleißheim, Germany) using scanning electron micrographs and three-dimensional reconstructions of the corresponding embryonic stages as a reference (Tamarin and Boyde, 1977;Verwoerd and van Oostrom, 1979;Washausen et al., 2005). Depending on the embryonic stage and/or the plane of sectioning, 60 up to 240 serial sections (section thickness=5 µm) were evaluated per PPA. Case-dependent, either complete series (section interval=5 µm) or every second section (interval=10 µm) were used to reconstruct the PPA. The plane of sectioning was determined according to the positions of various topographical landmarks (e.g. optic and otic vesicles, branchial membranes). Epibranchial placodes were identified structurally as patches of high-grade thickened, pseudostratified epithelium located adjacent to the branchial membranes Knabe, 2013, 2017). Additionally, the otic anlage and, if present, the otic detachment site were mapped. Diagnosis of lateral line placodes and neuromast primordia was based on the criteria which have been previously established in anamniotes (Northcutt, 1992;Northcutt et al., 1994;Sato, 1976;Schlosser, 2002b;Schlosser and Northcutt, 2000;Stone, 1933;Winklbauer, 1989).

Photomicrographs
Histological sections were examined under an Axioskop 2 MOT microscope (Carl Zeiss, Göttingen, Germany). Micrographs were captured with an Axiocam HR digital camera (Carl Zeiss) and the KS400 image analysis software (v3.0, Carl Zeiss). Following shading correction in KS400, images were cropped, resized, and adjusted for brightness (including slight gamma changes), colour balance, and sharpness in Corel Photo-Paint X4. All adjustments were applied to the whole image and no specific features within the photographs were modified, removed, or inserted.

Statistics
Statistical analysis was performed using STATISTICA software (v12.0, StatSoft, Hamburg, Germany). Since the numbers of apoptotic cells in the PPA were not normally distributed (Kolmogorov-Smirnov test) and variances between the three groups were not homogeneous (Levene's test), the nonparametric Mann-Whitney test was used to compare the levels of PPA apoptosis between in utero-developed embryos, cultured embryos incubated with DMSO, and cultured embryos treated with Q-VD-OPh or Z-VAD-fmk, respectively (Fig. 1B). All tests were two-sided, and P values <0.05 were considered statistically significant. Boxplots of PPA apoptosis were created using STATISTICA software. Diagrams demonstrating the frequency of unilaterally or bilaterally developed lateral line placodes were produced in Microsoft Excel (Fig. 2D).