Cell-type heterogeneity in the zebrafish olfactory placode is generated from progenitors within preplacodal ectoderm

Vertebrate olfactory placodes consists of a variety of neuronal populations, which are thought to have distinct embryonic origins. In the zebrafish, while ciliated sensory neurons arise from preplacodal ectoderm (PPE), previous lineage tracing studies suggest that both Gonadotropin releasing hormone 3 (Gnrh3) and microvillous sensory neurons derive from cranial neural crest (CNC). We find that the expression of Islet1/2 is restricted to Gnrh3 neurons associated with the olfactory placode. Unexpectedly, however, we find no change in Islet1/2+ cell numbers in sox10 mutant embryos, calling into question their CNC origin. Lineage reconstruction based on backtracking in time-lapse confocal datasets, and confirmed by photoconversion experiments, reveals that Gnrh3 neurons derive from the anterior/medial PPE. Similarly, all of the microvillous sensory neurons we have traced arise from preplacodal progenitors. Our results suggest that rather than originating from separate ectodermal populations, cell-type heterogeneity is generated from overlapping pools of progenitors within the preplacodal ectoderm.


Introduction 1
A fundamental question in developmental neurobiology is how different neuronal 2 subtypes arise from fields of pluripotent progenitors. At the end of gastrulation, the 3 anterior neural plate border of vertebrate embryos gives rise to two specialized 4 regions of ectoderm: the preplacodal ectoderm (PPE) that will ultimately produce the 5 cranial placodes, and the cranial neural crest (CNC). Specification of the PPE is 6 achieved through the action of so-called preplacodal competence factors such as 7 tfap2a, tfap2c, foxi1 and gata3 (Kwon et al., 2010). During a similar time-window, key 8 neural crest specifier genes, such as foxd3 (Lister et al., 2006;Montero-Balaguer et 9 al., 2006;Stewart et al., 2006), tfap2a (Barrallo-Gimeno et al., 2004)  where they have been reported to contribute to a large number of cell types, 17 including neurons associated with derivatives of the cranial placodes. These include 18 sensory neurons of the various peripheral sensory ganglia (D'Amico-Martel and 19 Noden, 1983;Harlow and Barlow, 2007), as well as sensory and neurosecretory cells 20 of the olfactory system (Whitlock et al., 2003;Saxena et al., 2013). established, a subset of EONs dies by apoptosis (Whitlock and Westerfield, 1998) 29 suggesting the existence of distinct subtypes of neurons within the population, but 30 specific markers for these different neurons have yet to be described. Neural subtype 31 heterogeneity is also detected early within the OSN population; in zebrafish the 32 predominant subtypes are ciliated sensory neurons that have long dendrites and 33 express olfactory marker protein (OMP) and microvillous sensory neurons, which third neural subtype associated with the early olfactory placode in zebrafish 3 expresses gonadotropin releasing hormone 3 (gnrh3). Rather than projecting 4 exclusively to the olfactory bulb, Gnrh3 neurons send their axons caudally to various 5 brain regions, including the hypothalamus (Abraham et al., 2008). While laser-6 ablating Gnrh3+ cells leads to sterility, animals homozygous for TALEN-induced 7 mutations of gnrh3 are fertile, pointing to the need of identifying other genes 8 expressed in these cells that might underlie the differences between these 9 phenotypes (Abraham et al., 2010;Spicer et al., 2016). 10 Although the major neural cell types associated with the olfactory epithelium 11 appear to be conserved across vertebrates, there is no coherent vision as to their 12 lineage origin between species. For instance, while Gnrh cells associated with the 13 developing olfactory placode are reported to be of preplacodal origin in chick, in the 14 zebrafish they have been shown to derive from the neural crest (Whitlock et al., 15 2003;Sabado et al., 2012); in mouse, Cre/lox experiments suggest that Gnrh cells 16 are of mixed lineage origin, coming from both the ectoderm and CNC (Forni et al., 17 2011). To identify additional markers of cell-type heterogeneity in the developing 18 zebrafish olfactory placode we screened expression of molecules known to label 19 discreet sets of neurons in other regions of the nervous system. We found that an 20 antibody that recognizes the Islet family (Islet1/2) of LIM-homeoproteins labels Gnrh3 21 neurons in the olfactory placode from an early developmental stage (Ericson et al., 22 1992). With this new marker we find no change in the numbers of Islet1/2+ cells in 23 the olfactory system in sox10 -/mutants, which are deficient in many CNC lineages. 24 This is in contrast with previous studies and calls into question the proposed CNC 25 origin of Gnrh+ cells. Consistent with these findings, lineage reconstructions of time-26 lapse confocal movies show that most if not all Gnrh3+ neurons, as well as 27 microvillous sensory neurons, derive from the PPE. Thus, cell-type heterogeneity 28 within the olfactory placode is likely established entirely from progenitors within the 29

Gnrh3 neurons 13
Gnrh3+ neurons associate closely with the olfactory placode from early stages 14 (Gopinath et al., 2004). However, a lack of Cre lines specific for placodal progenitors 15 precluded using a Cre/lox approach to address if Gnrh3 neurons derive from these 16 cells. As an alternative, we developed an unbiased backtracking approach using 17 time-lapse confocal movies. Briefly, synthetic mRNAs encoding Histone2B-RFP 18 (H2B-RFP) were injected into Tg(gnrh3:eGFP) transgenic embryos, which were 19 subsequently imaged from 12-36 hpf ( Figure 3A); delamination and migration of CNC 20 begins approximately 2 hours after the initiation of the time-lapse acquisition, and 21 eGFP from the transgene is robustly expressed at 36 hpf (Schilling and Kimmel, 22 1994;Abraham et al., 2008). The lineage of various populations of cells was then 23 manually retraced by backtracking H2B-RFP+ nuclei to their position at the beginning 24 of the time-lapse series using Imaris software ( Figure 3A and Video 1). To test our 25 approach relative to well-established fate maps already generated for zebrafish 26 cranial placodes, we first backtracked H2B-RFP+ nuclei of gnrh3:eGFP-negative 27 cells in the olfactory placode as well as lens cells, which can be identified at the end 28 than from a region posterior to the lens progenitors as would be expected for CNC 5

cells. 6
To confirm the anterior/medial PPE origin of Gnrh3 neurons, we used 7 photoconversion to label the cells. We loaded Tg(gnrh3:eGFP) embryos with NLS-8 mEos2 by mRNA injection and at 12 hpf photoconverted either the left or right half of Similar to the Gnrh3+ cells, we used photoconversion of NLS-mEos2 to 8 confirm our backtracking results for the microvillous population. This time, however, 9 photoconversion was focused on the left or right two-thirds of the anterior/lateral PPE 10 of Tg (-4.9sox10:eGFP) embryos ( Figure 6A,B). As before, the number of eGFP+ 11 cells was unaffected by the photoconversion (8+/-0.6 versus 9.8+/-0.7 cells per 12 placode, n=9 and 6 respectively; Figure 6C-E), and less than one NLS-13 mEos2 PC ;sox10:eGFP+ cell per placode was detected on the control side of the 14 embryo (0.3+/-0.3 cells per placode, n=6; Figure 6E). On the photoconverted side, 15 however, an average of just under 60% of the total sox10:eGFP+ population was 16 also NLS-mEos2 PC -positive (4.6+/-0.6 cells per placode, n=9; Figure 6D,E). This is 17 consistent with the fact that we photoconverted approximately two-thirds of the cells 18 in the PPE that gives rise to the olfactory placode. 19 Taken together, our results obtained using a combination of backtracking and 20 photoconversion, indicate that Gnrh3 and microvillous sensory neurons associated 21 with the zebrafish olfactory placode derive from the anterior/medial and 22 anterior/lateral PPE, respectively. neurons we have revisited this lineage and its origins within the placode. By 5 combining Cre/lox, backtracking in 4D confocal datasets and photoconversion, we 6 show that Gnrh3+ neurons derive from progenitors in the PPE; similar live imaging 7 techniques also indicate a PPE origin for microvillous sensory neurons. These results 8 support a common PPE origin for all of the neuronal populations within the olfactory 9 system and argue against any CNC contribution. More generally, they suggest a 10 mechanism by which cellular heterogeneity arises progressively within a field of 11 neuronal progenitors. 12 In zebrafish at least two Islet genes, islet1 and islet2b, are expressed in the our study and others show that in zebrafish these two populations overlap significantly in the PPE (reviewed in (Toro and Varga, 2007)). In mice Cre/lox lineage 1 analysis suggests that GnRH cells associated with the olfactory system are of mixed 2 origin, being derived 70% from the ectoderm and 30% from CNC (Forni et al., 2011). 3 This does not appear to be the case in chick, however, as grafted neural folds 4 expressing GFP do not contribute to the olfactory placode (Sabado et al., 2012). 5 Furthermore, similar to our results in zebrafish no change in GnRH cells numbers is 6 detected in Sox10-null mutant mice (Pingault et al., 2013). As the original lineage 7 assignment in mouse was established using only a single ectodermal (Crect; (Reid et 8 al., 2011)) and neural crest (Wnt1Cre; (Danielian et al., 1998)) Cre line, respectively, 9 our results suggest that revisiting the lineage assignment in the mouse with other 10 genetic tools or other approaches is needed. 11 The origin of zebrafish microvillous sensory neurons is controversial (Saxena 12 et al., 2013;Torres-Paz and Whitlock, 2014). Our results indicate that this lineage is 13 derived from progenitors in the PPE. Furthermore, they imply that the expression of 14 eGFP in the olfactory placode of Tg(sox10:eGFP) embryos does not reflect a CNC 15 origin for these cells but rather an ectopic site of transgene expression. Why ectopic 16 expression of the transgene in the olfactory placode appears restricted to 17 microvillous sensory neurons is unclear but highlights the need to be cautious when 18 using transgenic tools in lineage analyses. In this regard, the backtracking approach 19 we have developed provides a powerful alternative for ascertaining lineage 20 assignments during zebrafish embryogenesis. While in this study we used transgenic 21 lines to identify cell types for backtracking, with the caveats that this obliges, we have 22 also backtracked cells identified using antibody markers. For this, cells for 23 backtracking can be identified by comparing the 3D architecture of a tissue described 24 by nuclear position at the end of a time-lapse series to that of the same embryo after 25 fixation and antibody labeling. Our analysis also highlights difficulties with using 26 morpholinos when studying tissues for which a detailed comparison between the 27 morphant and mutant phenoptypes has not been undertaken. Thus, while sox10 28 morphants recapitulate a variety of phenotypes described in strong alleles of sox10 29 mutants (Dutton et al., 2001), our data suggests that this cannot be extrapolated to 30 the olfactory placode. 31 In conclusion, our results argue that cell-type heterogeneity in the zebrafish 32 olfactory placode is generated from progenitors within the PPE, and begin to provide 33 coherence for the lineage assignment of olfactory neural subtypes between vertebrate species. Identifying the mechanisms underlying the segregation of the 1 various olfactory lineages from overlapping progenitor pools is an important avenue 2 for future research. Health (NIH R01 DE13828 and AR67797 to TS); Fondation pour la Recherche 9 Médicale (DEQ20131029166 to PB); Fédération pour la Recherche sur le Cerveau; 10 and the French Ministère de la Recherche. We would like to thank Brice Ronsin and 11 the Toulouse RIO Imaging platform (LITC), and Aurore Laire, Richard Brimicombe-12 Lefevre and Ines Gehring for fish husbandry. We thank the Alsina, Fisher, Granato, 13 Kelsh, Neuhauss, and Zohar labs for providing fish lines and other reagents. We also 14 thank Christian Mosimann, Tatjana Sauka-Spengler and Trevor Williams for 15 comments and suggestions. 16 17

Fish Husbandry and lines 19
Fish were maintained at the CBD (Toulouse) and UCI (Irvine) zebrafish facilities in 20 accordance with the rules and protocols in place in the respective locations. The 21 sox10 t3 , Tg (-2.4gnrh3:egfp) zf103 and Tg   obtained through natural crosses and staged according to (Kimmel et al., 1995). Embryos from the Tg(gnrh3:egfp) zf103 and Tg (-4.9sox10:eGFP) ba2Tg transgenic lines 30 were injected with synthetic mRNA encoding an NLS-mEos2 (mEOS2 fused to a 31 nuclear localization sequence) fusion protein (Sapede et al., 2012). Embryos were 32 then grown to 12 hpf, dechorionated and embedded for photoconversion/imaging in 33 0.7% low-melting point agarose in embryos medium in 35mm circular petri dish (Nunc TM ; 153066) bearing a silicone sealed 22mm circular cover slip (Thermo 1 Scientific TM ; 174977). Mounted embryos were first imaged for NLS-mEos2 2 expression prior photoconversion at very low laser levels (confocal stack 2 mm 3 slice/80 mm deep). Subsequently, a region of interest (ROI) was photoconverted 4 using a 405nm diode (100% laser, 41sec), after which embryos were imaged again 5 to assess the extent of NLS-mEos2 conversion. Photoconversion and imaging was 6 done on an inverted SP8 Leica confocal with an HC PL APO CS2 40x/1.3 oil 7 objective. Full z-stacks were acquired for each photoconverted embryo 24h after the 8 photoconversion (confocal stack 1 mm slice/80 mm deep) to determine the 9 contribution of progenitors located in the ROI at the time of photoconversion to the 10 Gnrh3 or microvillous lineages. 11 12

Additional information 19
Funding 20

30
The funders had no role in study design, data collection and interpretation, or the 31 decision to submit the work for publication.                            s.e.m. P values are calculated using a two-tailed Student's t-test, n.s. not significant, 20 ***p=0.0001. 21

22
The following source data is available for figure 6: 23 Source data 1. sox10:eGFP+ and mEos2 PC + cell number quantification. 24 25 Video 1. Lineage reconstruction reveals an anterior preplacodal ectoderm origin for 26

Gnrh3 neurons. 27
Movie showing an acquisition series and backtracking of a Gnrh3 neuron in an 28 Histone2B-RFP loaded Tg(gnrh3:eGFP) transgenic embryo. The movie is divided 29 into 5 parts: an acquisition phase, a phase showing a z-stack of the olfactory placode 30 at 36 hpf, an initial backtracking phase, the unique mitosis detected during the 31 backtracking, and a final backtracking phase. During backtracking, the nucleus being 32 followed is labelled with a blue dot. During the mitosis, the sister cell is labelled with a pink dot for three frames until the two sister nuclei fuse. The movie finishes with a 1 schematic representation of the anterior neural plate and adjacent preplacodal 2 ectoderm at 12 hpf showing the results obtained from backtracking 30 gnrh3:eGFP-3 positive cells. The cell backtracked in the movie is indicated (arrowhead).