Centrioles are amplified via rosette formation in cycling progenitors of olfactory sensory neurons

Olfaction in most animals is mediated by sensory neurons bearing receptors on cilia that are accessible to the environment. Within chordates, olfactory sensory neurons (OSNs) usually have multiple cilia, each with a centriole at its base. OSNs differentiate from stem cells in the olfactory epithelium, and how mature cells with multiple centrioles are generated during this process is not yet understood. OSNs in the mouse olfactory epithelium have about 15 cilia each, and we show that centrioles are amplified in precursor cells via formation of centriole rosette structures both during embryonic development and during turnover of the olfactory epithelium in adults. We also found free centrioles present in rosette-bearing cells, suggesting that more than one pathway contributes to total centriole number. Cells with amplified centrioles can go on to divide, with clustered centrioles at each pole. Additionally, we found that centrioles are amplified in early immediate neuronal precursors, coincident with elevation of mRNA for Plk4 and Stil, two key regulators of centriole duplication. Our findings highlight the importance of accounting for centriole amplification in the development of olfactory epithelium-derived neuron regeneration therapies.


Introduction
Olfaction, the primary way that animals sense their chemical environment, begins in olfactory sensory neurons (OSNs). In many chordates, each OSN has multiple cilia which protrude from the end of a dendrite at the apical surface of the olfactory epithelium. At the apical surface, odorants contact receptors on the surface of cilia, initiating a signaling event in the OSN. At the base of each cilium, a centriole organizes the structure ( Figure 1A). Cilia are necessary for olfaction, as are the centrioles which organize their microtubule structures. To have multiple cilia, each OSN must have multiple centrioles, raising the question: How are these many centrioles made?
The centriole number in OSNs lies between that of two well-studied states. A common state is for cells to have exactly two centrioles, with the older of the two serving as a basal body for a primary cilium. This older centriole is often referred to as the mother centriole and the newer centriole as the daughter centriole. The daughter centriole forms orthogonally to the mother centriole in G1/S of the cell cycle and is engaged to the mother until mitosis (Kuriyama and Borisy, 1981). Upon passage through mitosis, it becomes disengaged, and in the ensuing cell cycle, it acts as a mother centriole upon which another new daughter centriole can form. In contrast, multiciliated epithelial cells have as many as hundreds of centrioles, each serving as a basal body for a motile cilium. In this state, cells exit the cell cycle and initiate a transcriptional program that facilitates this centriole amplification (Hoh et al., 2012). Centriole amplification in multiciliated epithelial cells occurs by two means: 1) centriole growth from deuterosomes, structures that are specific to multiciliated epithelial cells, and 2) by growth of multiple daughter centrioles from each mother centriole, forming rosettes (Sorokin, 1968). Centriole rosettes are thought to contribute only a small percentage of the total number of centrioles in multiciliated epithelial cells (Al Jord et al., 2014). Cycling cells can also be induced to form centriole rosettes by overexpression of Polo-like kinase 4 (Plk4), a kinase required for centriole duplication, or by overexpression of certain other centriole duplication proteins (Habedanck et al., 2005).
Interestingly, centrioles in the olfactory epithelium were previously described to be arranged in a rosette-like array in some cells (Cuschieri and Bannister, 1975). However, the nature of these centriole rosettes and their possible relationship to OSN formation have not been investigated. Our work builds upon these findings by describing a role for rosettes in centriole amplification, identifying developmental timing of centriole amplification, and presenting a possible mechanism for driving centriole amplification in the mouse olfactory epithelium.

Results
To better define the range of centriole number in OSNs in our preparations, we counted centrioles in OSNs from mice expressing eGFP-centrin2, a marker of the centriole (Bangs et al., 2015). In nasal septa from adult mice, OSNs had an average of 15.7 centrioles per cell, with wide variation around the mean (6 to 37 centrioles/cell, see Figure  3F) but no apparent trend across the anterior-posterior axis. This number of centrioles is similar to previous reports of cilium and centriole number in OSNs (Challis et al., 2015;Uytingco et al., 2019).
We next considered the potential means by which cells amplify centriole number during differentiation from stem cells to OSNs. Centrioles in the olfactory epithelium were previously described to be arranged in a rosette-like array in some cells (Cuschieri and Bannister, 1975). To assess the presence and role of centriole rosettes in the olfactory epithelium, we visualized centrioles by transmission electron microscopy (TEM) and fluorescence microscopy in both adult and embryonic tissue. First, ultrathin sections were made from dissected olfactory turbinates taken from adult mice and examined by TEM. We observed dendritic knob structures with multiple centrioles, typical of OSNs (Figure 1A), as well as horizontal basal cells with centriole pairs and primary cilia (not shown). Near the basal lamina, where OSN progenitor cells are typically found, we found cells with centriole rosettes ( Figure 1B). Next, we determined whether centriole rosettes were present in cryosections of embryonic (E12.5) olfactory epithelia from mice expressing eGFP-centrin2 (Figure 1C, S1A). Rosettes were apparent as clusters of eGFP-centrin2 foci with the expected dimensions. Note that the cell shown in the inset has two rosettes, consistent with rosette formation on both preexisting centrioles. In addition, we found that rosette-bearing cells were positive for the neuronal marker β-tubulin III, confirming that these cells were committed to a neuronal cell fate ( Figure 1C). Our results suggest that centrioles are amplified by rosettes in both adult and embryonic olfactory epithelium. By observing olfactory epithelia of adult mice by TEM and embryonic mice by fluorescence microscopy, we also found cells that had centrioles in addition to two rosettes. In olfactory epithelia from adult mice, we found cells with two rosettes, as well as free centrioles by TEM ( Figure 1D, S1B-E). Similarly, puncta of eGFP-centrin2 were observed near rosettes in embryonic olfactory epithelia by fluorescence microscopy (see Figure 1C, S1A). Whether free centrioles formed by detaching from a canonical rosette or free of a parental centriole (i.e. de novo) requires further investigation.
Next, we next asked whether centriole amplification can occur in cycling cells or only in non-dividing differentiated cells, using fluorescence microscopy in adult and embryonic olfactory epithelium. We used stage-specific markers to assess centriole amplification in cells in different stages of the cell cycle. In the olfactory epithelium of wildtype adult mice, some cells with nuclear PCNA, a marker for S phase, had centriole rosettes ( Figure S2A). To determine whether cells which amplify centrioles in S phase proceed into mitosis, we probed for phospho-H3, a marker for mitosis, and confirmed the mitotic state by presence of condensed DNA. In the olfactory epithelium of wild-type mice, many mitotic cells had clusters of centrioles ( Figure S2B), and in instances in which both spindle poles could be imaged, both poles had amplified centrioles (Figure 2A). Similarly, we identified newlyforming sister cells with condensed chromatin and amplified centrioles in embryonic olfactory epithelium at E12.5 ( Figure 2B). It was not possible to precisely count centrioles in most mitotic cells due to limitations of imaging, particularly in adult olfactory epithelium sections. Instead, we applied a quantitative fluorescence method, measuring the total area of centrin fluorescence signal ( Figure 2C). This method was calibrated on hTERT RPE-1 cells with and without Plk4 overexpression to generate centriole rosettes ( Figure S2C). In the olfactory epithelium many of the mitotic cells had area measurements consistent with centriole clusters (rosettes +/-extra centrioles) rather than centriole pairs ( Figure 2C). These results demonstrate that OSN precursors are able to divide after centriole amplification in both adult and embryonic olfactory epithelium, and that both sister cells of a We specifically examined genes known to be upregulated in association with DNA synthesis, including Rrm2, which encodes ribonucleotide reductase 2 (Thelander and Berg, 1986). We found that the mRNA for many of these genes was abundant only in cells in the GBC and INP1 states, consistent with these being mitotically active ( Figure 3A). Next, we analyzed the scRNAseq data for the expression pattern of genes encoding proteins associated with centriole duplication. Remarkably, the mRNA for Plk4 was strongly elevated in INP1s and INP2s, and less so in GBCs ( Figure 3B). Plk4 effects centriole formation in conjunction with a binding partner, Stil (Arquint et al., 2012;Vulprecht et al., 2012). We found that the mRNA for Stil was also elevated in INP1s and INP2s (Figure 3B, S3B). The mRNA for Cep152, another binding partner of Plk4, also followed this pattern (Figure S3A), whereas those for most other centriole-associated genes did not. Given that upregulation of Plk4 or Stil RNA drives centriole rosette formation in cell culture, we hypothesized that elevated Plk4 and Stil might drive centriole amplification in early INPs.
To determine if elevated Plk4 and Stil RNA levels correlate with the timing of centriole amplification, we used NeuroD1 as a marker of developmental timing within the differentiation pathway for OSNs. NeuroD1 is a transcription factor specifically upregulated in early INPs ( Figure 3C) (Packard et al., 2011). We identified OSN progenitors in sections of adult olfactory epithelium by their localization near the basal lamina and presence of nuclear NeuroD1 immunofluorescence signal.
We found examples of cells with two centrioles and cells with on-pair centrioles within the NeuroD1-positive progenitor population ( Figure  3D, 3E). We used cells dissociated from olfactory epithelia of adult mice expressing eGFP-centrin2 to quantify centriole number in NeuroD1-positive precursors (Figure 3F, S3D). We compared these counts to the number of centrioles per OSN imaged in septa from adult mice expressing eGFP-centrin2. As in the sections, we found two groups amongst the NeuroD1-positive cells: a minority (n=4) of cells that had only one or two visible centrioles, suggesting that they had not yet amplified centriole number, and a majority (n=36) that had many more centrioles per cell (6 to 39 centrioles per cell). This distribution of centriole numbers suggests that centrioles are amplified in NeuroD1positive cells, which also have high levels of Plk4 and Stil mRNA ( Figure S3C). Together, our data support a model in which elevated Plk4 and Stil drive amplification via centriole rosettes in the olfactory epithelium.

Discussion
Olfactory sensory neurons in mammals have a configuration of centrioles and cilia that distinguishes them from most other cells. We have found the centrioles in OSNs can be amplified from the progenitor cell's centrosome via centriole rosettes prior to cell division, and that this is correlated with increased expression of the centriole duplication proteins Plk4 and Stil (Figure 4).
The number of centrioles in OSNs, an order of magnitude separated from most well-studied cell types, prompted us to ask how the observed number is achieved by cells. Our results show that centriole rosettes form on both of the pre-existing centrioles in progenitor cells in the olfactory epithelium and that both daughter cells receive an amplified set of centrioles, presumably one rosette each. If daughter centrioles in rosettes are engaged orthogonally to the mother as they are in cycling cells, then we estimate that the maximum total number of centrioles made by a rosette to be approximately nine (eight daughters plus one mother centriole), limited by the surface area around the base of the mother centriole. This is consistent with the number of daughter centrioles in rosettes observed in a single plane by TEM (see Figure  1B). However, the distribution of centriole numbers in OSNs shows that most have more than nine centrioles, thus, inheriting a single rosette of centrioles is insufficient to explain the observed number.
Within the constraints imposed by rosette size, there are several possible ways that OSNs might achieve the desired number of centrioles. One possibility is that more centrioles can be formed in a single amplification event. If centriole formation is not limited to the area around the base of the mother centriole, either by disengagement of daughter centrioles or by de novo synthesis, then many more centrioles can form in a single amplification event. Our observation of free centrioles in cells with rosettes supports this possibility, although our data do not distinguish between disengagement and de novo synthesis. We note that the presence of free centrioles in addition to rosettes would allow the possibility of asymmetric segregation of centrioles during mitosis, which would yield cells with centriole numbers near the extremes of the observed distribution. The second possibility for how OSNs achieve the desired number of centrioles is that additional centriole amplification might occur in subsequent cell cycles or after the final cell division in OSN differentiation. There is precedence for the latter in multiciliated epithelial cells, the only other widely-studied example of centriole amplification in vertebrates, where cells only amplify centrioles post-division.
Our data show that centriole amplification can occur in mitotically active progenitors of the olfactory epithelium. This is surprising because division with amplified centrioles is usually considered to be detrimental due to chromosome missegregation (Ganem et al., 2009;Silkworth et al., 2009). However, several features of the process might mitigate the potential problem of mitosis with amplified centrioles. First, many of the amplified centrioles are contained within rosette structures that likely function as single organizing centers during spindle formation, based on the single mother centriole within each rosette (Figure S1B-E). Indeed, we found that each rosette in this case had only a single focus of γ-tubulin ( Figure S2A). This is similar to what appears to occur in Viviparus spermatogenesis, during which cells with rosettes go through meiosis (Gall, 1961;Pollister and Pollister, 1943). Cosenza et al. (2017) showed that the fidelity of mitosis in cells with overexpression-induced rosettes is sensitive to asymmetry in the number of daughter centrioles per rosette. It is unknown whether this phenomenon plays a role in the OSN lineage, and this would require live imaging of mitoses in the olfactory epithelium to resolve. Second, newly-formed free centrioles would not have undergone the centrioleto-centrosome conversion that would promote their ability to form a spindle pole in the mitosis immediately following their formation (Wang et al., 2011). Even if the free centrioles were capable of microtubule nucleation, known mechanisms could enforce bipolar spindle formation, for example by HSET-dependent centriole clustering (Kwon et al., 2008).
We showed that the transcripts for key proteins in centriole duplication, Plk4 and Stil, are transiently upregulated during OSN differentiation. This raises the question of whether this upregulation is sufficient to coordinate the formation of centriole rosettes. In multiciliated epithelial cells, Plk4 and other centriole-associated genes are upregulated to drive centriole amplification via deuterosomes and rosettes during differentiation (Hoh et al., 2012). OSN differentiation closely resembles that of multiciliated epithelial cells, except that no deuterosomes form  In summary, our work highlights a system in which centriole amplification and cell division occur as a normal part of development and organ maintenance. These findings outline an important window for therapeutic potential of olfactory stem cells and also reveal more information about the basic biology of centriole formation.

Transmission electron microscopy
Mice were euthanized by CO2 in accordance with Stanford's APLAC guidelines. Facial bones were removed in a dish of cold Tyrode's solution (140mM NaCl, 5mM KCl, 10mM HEPES, 1mM CaCl2, 1mM MgCl2, 1mM sodium pyruvate, 10mM glucose in ddH2O), as in other reports (Dunston et al., 2013), and turbinate scrolls were mechanically separated from septa. Epithelia from turbinate scrolls were removed mechanically and fixed immediately in a solution of 2% glutaraldehyde and 4% PFA in 0.1M Na cacodylate buffer for 3 to 4 hours at 4°C. Samples were then rotated in a 1% solution of OsO4 for 1 hour at room temperature, washed four times gently in water, then rotated in a 1% solution of uranyl acetate overnight at 4°C. Samples were then dehydrated in a graded ethanol series (30%, 50%, 70%, 95%, 100%, 100%) for 15 to 20 minutes per step, rotating at room temperature. Samples were washed twice for 10 minutes each in propylene oxide (PO), then embedded through a graded PO:EMBED resin series (2:1 for 1 hour, 1:1 for 1 hour, 1:2 overnight). Samples were then rotated in pure EPON with lids open for 5 hours to evaporate remaining PO before embedding in molds at 50°C for 4 days. Semi-thin sections were taken and imaged on a dissecting scope to find samples in the correct orientation. 80 nm sections were treated with uranyl acetate and mounted on grids before imaging on a JEOL JEM-1400 transmission electron microscope.

Immunofluorescent staining of cryosections
Olfactory epithelia were dissected as described above. Whole olfactory epithelia, turbinate epithelia, or E12.5 embryo heads were fixed immediately in 4% PFA in PBS at 4°C for 3 to 24 hours. Samples were then washed in phosphate buffered saline (PBS) and stored at 4°C. Before mounting, samples were equilibrated in 1 to 5mL of 30% sucrose solution in water for a minimum of 12 hours at 4°C. Samples were embedded in OCT compound (Sakura Tissue-Tek) on dry ice and stored at -80°C. Embedded samples were sectioned at 8 to 14 µm on a Leica cryostat and adhered to charged slides by drying at room temperature for approximately 1 hour. Slides were stored with drying pearls (Thermo Fisher) at -80°C and thawed under desiccation no more than twice. Samples were pretreated as needed (see antibody summary chart) and rehydrated and blocked for 0.5 to 4 hours in 5% milk in 0.1% Triton-x 100 that had been spun in a tabletop centrifuge to pellet un-dissolved milk particles. Slides were incubated in primary antibody for approximately 3 hours, washed in PBS, incubated in secondary antibody for approximately 1 hour, washed in PBS, incubated in DAPI for 1 to 5 minutes, washed in PBS, and mounted in MOWIOL. Embryonic samples were imaged on a spinning disk confocal microscope, and adult samples were imaged on a Zeiss inverted widefield microscope using MicroManager (Edelstein et al., 2010). Images were processed in FIJI (Schindelin et al., 2012). Images from the widefield microscope were deconvolved using the Iterative Deconvolve plug-in (Dougherty, 2012) and theoretically-generated point spread functions (Diffraction PSF 3D). For images with high background, contrast in the representative images was adjusted uniformly across the image such that the area outside of cells was black and areas of high signal were just below saturation.

Analysis of centriole structure area
To quantify the area of centriole structures, samples were imaged on a Zeiss inverted widefield microscope. For proof of concept, centrioles from RPE-1 cells with immunofluorescent staining (see below) were imaged in z-stacks with 0.5 µm steps to include all centrioles in the field of view. Images were processed in FIJI. Z-stacks were converted into a maximum projection image, and the green channel was deconvolved using the Iterative Deconvolve plug-in (Dougherty, 2012). Engaged structures were selected based on the presence of anti-Sass6 immunofluorescence signal between adjacent anti-GFP puncta. We measured the area of anti-GFP fluorescence in these structures and normalized the area such that the average area of centriole pairs was exactly 2. The normalized fluorescence area had approximately a 1:1 ratio with actual centriole number (0.9208), demonstrating that it is an appropriate approximation for centriole number (Figure S2C). Actual area measurements were approximately the dimensions expected for two centrioles, modeled as the projection of two overlapping spheres (analysis not shown). Next, we used this method to assess centriole structures from the olfactory epithelium ( Figure 3C). We estimated the probability density distribution of centriole pair areas to be a Gaussian function. We used this distribution to estimate a cutoff area (0.7085 µm 2 ) above which a structure has less than 1% probability of belonging to the centriole pairs' dataset. As a proof of concept, 73.0% of rosettes measured in cell culture were above this cutoff. For olfactory epithelia, single-plane images were used for analysis shown here, though similar results were obtained with z-stacks (data not shown). We applied the cutoff to mitotic cells of the olfactory epithelium because, in contrast to S-phase cells, mitotic cells' centrioles separate in preparation for spindle formation, reducing overlap and making structures more amenable to measurement. Images of anticentrin immunofluorescence signal in mitotic cells in cryosections from adult mice were deconvolved using the Iterative Deconvolve plug-in, and area was measured by outlining puncta in the anti-centrin channel. Images in which centriole pairs were clearly visible were categorized as such. All other images were categorized as "non-pair" structures. 87.2% of area measurements in this group fell above the cutoff. Probability calculations and modeling of area projections were performed in R. Dot plots were generated with Statistika (Weissgerber et al., 2017).

Immunofluorescent staining of RPE-1 cells
Cells cultured on poly-L-lysine-coated coverslips were fixed in methanol at -20°C for 20 minutes and washed in PBS. Samples were blocked for a minimum of 30 minutes in 5% dry milk in 0.1% Triton-x 100 that had been spun to pellet undissolved milk particles. Samples were then washed 3 times in PBS, incubated with primary antibodies for 1 to 2 hours at room temperature, washed 3 times in PBS, incubated with secondary antibodies for 0.5 to 2 hours, washed 3 times in PBS, incubated with DAPI for 5 minutes, washed 3 times in PBS, and mounted in MOWIOL. Cells were imaged on a Zeiss inverted widefield microscope with MicroManager (Edelstein et al., 2010). Images were processed in FIJI.

Quantification of centrioles in OSNs
Samples were dissected as described above, except that dissections were performed in PBS. Septa were immediately transferred to 4% paraformaldehyde (PFA) in PBS and fixed for 3 to 24 hours at 4°C. Septa were washed and stored in PBS at 4°C. For imaging, septa were mounted in a chamber of double-sided tape on glass slides with SlowFade Gold mountant (Invitrogen) and high precision 1.5 weight coverslips (Deckglässer) sealed with nail polish. Samples were imaged on a Leica SP8 scanning confocal microscope. For each sample, five fields of view were spaced approximately evenly along the anteriorposterior axis of the olfactory epithelium. Within each field of view, the cell at the center of each quadrant of the field was imaged such that the z-stack included all centrioles within the dendritic knob. Individual dendrites can be identified by their tightly-clustered centrioles (data not shown). Image stacks were processed by semi-automated detection in the program Imaris x64 9.2.1 using the Surfaces function and separating touching objects by seed points of 0.3 µm diameter. Dot plots were generated using Statistika (Weissgerber et al., 2017).

Olfactory epithelium dissociation
The turbinate region of olfactory epithelia was dissected in cold Tyrode's solution, as described above. Samples were incubated in 1 to 2mL of 0.25% trypsin (Thermo-Fisher, #MT-25-053-CI) and minced with a feather scalpel periodically, between incubations at 37°C, for 10 to 15 minutes in total. Trypsin was inactivated by adding 10 mL of DMEM with 10% serum. Samples were poured over a 40 µm cell strainer to remove bone fragments and other large debris. Samples were spun at 800g for 5 minutes to pellet, washed in PBS, and spun again. Samples were resuspended and fixed in 4% PFA in PBS overnight at 4°C, then washed and stored in PBS at 4°C.

Quantification of centriole number in progenitor cells
Dissociated samples were stained to identify NeuroD1-positive progenitors by first spinning at 8000rpm for 2 minutes in a tabletop centrifuge to remove PBS, then pretreating by resuspending in 0.5% (w/v) SDS in water for 1 minute. Cells were spun to remove SDS, then resuspended and blocked for 0.5 to 1 hour at 4°C in a solution of 5% dry milk in 0.1% Triton-x 100 that had been spun to remove un-dissolved milk particles. Samples were washed in PBS, incubated in primary antibody overnight at 4°C, washed in PBS, incubated in secondary antibody for 30 minutes at room temperature, washed in PBS, incubated in DAPI solution for 2 minutes, washed in PBS, then resuspended in MOWIOL. Samples were mounted with 1.5 weight coverslips, and centrioles were imaged and counted by the same method as OSN centrioles, described above.