High Sox2 expression predicts taste lineage competency of lingual progenitors in vitro Free

Gustation is mediated by taste buds in specialized structures on the tongue: in rodents, each fungiform papilla (FFP) on the anterior tongue houses a single bud, whereas posterior foliate and circumvallate papillae (CVP) each house hundreds of taste buds. Each taste bud contains ∼60 taste receptor cells (TRCs) categorized as follows: type I glial-like support cells; type II cells that detect sweet, bitter or umami stimuli; and type III cells that respond to sour and some salty stimuli (Roper and Chaudhari, 2017). TRCs and surrounding non-taste epithelium are continuously replenished by basal progenitor cells; these give rise directly to non-taste lingual epithelium and to taste-fated daughter cells that enter buds, transiently express sonic hedgehog (SHH) and differentiate into each of the TRC types (see Finger and Barlow, 2021). To date, progenitors expressing LGR5 (leucine-rich repeat-containing G-protein coupled receptor 5) (Takeda et al., 2013; Yee et al., 2013), GLI1 (Liu et al., 2013), cytokeratin (KRT) 14 (Okubo et al., 2009), KRT5 (Gaillard et al., 2015) and SOX2 (SRY-related HMG box family) (Ohmoto et al., 2017) are known to give rise to both TRCs and non-taste epithelium in mice, but their potentially distinct roles in taste epithelial renewal are unexplored. In murine CVP, LGR5⁺ progenitors also function as taste stem cells in vitro, as isolated LGR5⁺ cells generate lingual organoids housing cycling progenitors, non-taste epithelium and TRCs (Ren et al., 2014). However, the potential of SOX2⁺ progenitors to generate TRC-replete organoids in vitro has not been explored.

SOX2 is a key regulator of homeostasis in many adult epithelia (Arnold et al., 2011; Novak et al., 2020), the function of which often depends on expression level (Hagey and Muhr, 2014; Que et al., 2009; Sarkar and Hochedlinger, 2013). SOX2 is expressed by most basal keratinocytes within taste epithelium (Castillo-Azofeifa et al., 2018; Suzuki, 2008), and long-term lineage tracing of SOX2⁺ cells labels taste and non-taste epithelia in mice (Ohmoto et al., 2022, 2017). However, lingual SOX2 expression is highly variable with high expression in a subset of TRCs and in progenitors immediately adjacent to buds, and lower expression in non-taste epithelium within taste papillae (Castillo-Azofeifa et al., 2018; Ohmoto et al., 2017; Okubo et al., 2006; Suzuki, 2008). These findings suggest that high SOX2-expressing progenitors in lingual epithelium replenish taste buds, whereas low SOX2-expressing basal cells may not. Here, we have used organoid technology to test whether SOX2 expression levels predict the ability of isolated progenitors to give rise to TRCs in vitro and we show that only the progenitors expressing SOX2 at higher levels are competent to produce TRC-replete organoids. Furthermore, although both hedgehog and WNT/β-catenin are required for taste homeostasis in adult mice (Castillo et al., 2014; Castillo-Azofeifa et al., 2017; Gaillard et al., 2017, 2015; Kumari et al., 2015; Xu et al., 2017), we show here that only WNT/β-catenin promotes TRC differentiation in vitro and that it does so in organoids derived from higher SOX2-expressing taste lineage-competent progenitors.

We first assessed SOX2 immunofluorescence (IF) in Lgr5^EGFP mice (Barker et al., 2007). As reported previously (Ohmoto et al., 2017; Suzuki, 2008), SOX2 expression is highly variable among epithelial cells of the mouse CVP. For example, SOX2 IF is strong in some TRCs and basal cells adjacent to buds but is low in non-taste epithelial cells between buds, as well as in many cells in the deep CVP trench (Fig. 1A-A″). As previously reported (Yee et al., 2013), LGR5-GFP is bright in cells in the deep CVP and less robust in basal cells along the trench walls (Fig. 1A-A″). Sox2 mRNA, detected via hybridization chain reaction in situ hybridization, is more highly expressed by cells in and around taste buds, but expression is low in non-taste epithelium and in the deep trench, consistent with SOX2 IF. However, in contrast to the pattern of LGR5-driven GFP, which is strongest deep in the trench, Lgr5 expression levels appear similar throughout the CVP epithelium (compare Fig. 1A with B). This discrepancy may be due to perdurance of GFP (Arnone et al., 2004), a phenomenon likely enhanced in the slower cycling LGR5⁺ cells of the deep trench (Yee et al., 2013). Thus, although CVP basal keratinocytes co-express Sox2 and Lgr5, they do so with differing relative intensities, and their expression appears to vary independently of one another.

SOX2⁺ and LGR5⁺ cells generate TRCs and non-taste epithelium in vivo, but only LGR5⁺ cells have been shown to produce TRC-replete organoids (Ren et al., 2017, 2014, 2020; Shechtman et al., 2021). Thus, we employed Sox2^GFP mice where GFP reliably reports SOX2 expression in taste epithelium (Castillo-Azofeifa et al., 2018; Okubo et al., 2006) to determine whether isolated SOX2⁺ progenitors generate TRC-containing organoids. SOX2-GFP⁺ and LGR5-GFP⁺ cells were isolated and cultured as described previously (Shechtman et al., 2021) (Fig. 2A). Quantitative PCR (qPCR) revealed that mRNA for markers of type I (Entpd2 and Kcnj1) (Bartel et al., 2006; Dvoryanchikov et al., 2009), type II (Gnat3 and Pou2f3) (Boughter et al., 1997; Matsumoto et al., 2011) and type III (Car4 and Pkd2l1) (Kataoka et al., 2008; Wilson et al., 2017) TRCs was expressed by SOX2-derived organoids (‘SOX2 organoids’) but at significantly lower levels than in LGR5-derived organoids (‘LGR5 organoids’) (Fig. 2B). Using IF, we found most LGR5 organoids contained type I (NTPDase2⁺), type II (gustducin⁺) and type III (CAR4⁺) TRCs, as expected, whereas most SOX2 organoids lacked TRCs (Fig. 2C). Furthermore, LGR5 organoids had significantly more TRCs of each type (>10 taste cells/organoid), whereas the few SOX2 organoids with TRCs had fewer cells per organoid (Fig. 2D). Sparse NTPDase2⁺ and gustducin⁺ cells in many SOX2 organoids appeared poorly differentiated, lacking the typical fusiform morphology of type I and II cells in LGR5 organoids (Fig. 2C). Nonetheless, a small number of SOX2 organoids were TRC replete, suggesting a subset of SOX2⁺ progenitors are taste competent.

As SOX2⁺ progenitors generate few organoids with TRCs, we posited the remainder comprised primarily non-taste epithelium. Consistent with this, Kcnq1 (a general TRC marker) (Wang et al., 2009a) and Krt14 (a marker of cycling progenitors) were significantly lower in SOX2 than in LGR5 organoids, whereas Krt13 (a non-taste marker) expression was comparable between organoid types (Fig. 2E). In LGR5 organoids, KRT14⁺ progenitors made up the organoid exterior, whereas differentiated TRCs (KRT8⁺) and non-taste cells (KRT13⁺) were situated internally, reflecting the basal/external-apical/internal organization of lingual organoids (Fig. 2F) (Ren et al., 2014). KRT14⁺ cells were also external in SOX2 organoids, but internally these organoids comprised mainly KRT13⁺ non-taste cells (Fig. 2G). Additionally, KRT8 and KRT14 were frequently co-expressed by cells at the periphery of SOX2 organoids (Fig. 2G). In adult CVP, as progenitors produce new TRCs, KRT14 is downregulated and KRT8 upregulated, so that immature cells within taste buds are transiently both KRT8⁺ and KRT14⁺ (Asano-Miyoshi et al., 2008). In SOX2 organoids, these KRT8⁺ and KRT14⁺ cells may therefore be immature TRCs (Fig. 2G). In summary, although some SOX2 organoids are taste competent, most are composed of non-taste epithelium with little TRC differentiation.

Because TRC production is limited in SOX2 organoids, and higher SOX2-expressing cells are associated with taste buds in vivo, we hypothesized that higher SOX2⁺ populations would have gene profiles consistent with taste lineage production. Because relative SOX2 expression levels in different CVP epithelial cell populations have not been quantified, the selection of the number and size of brightness bins was based on qualitative estimates of SOX2 IF (see Fig. 1). We estimated for following: SOX2-bright cells in buds are the least frequent (high 5%), cells around taste buds that are moderately SOX2 bright are slightly more numerous (HiMed 10%), dimmer SOX2⁺ non-taste epithelial cells are common in the CVP (MedLow 25%) and very low SOX2-expressing cells are likely abundant non-taste epithelium included in the dissection (Low 50%). Cells from each brightness bin were separately collected via FACS from dissociated Sox2^GFP CVP epithelium (see Materials and Methods for full details) and processed for bulk RNA sequencing and analysis. Relative expression of Sox2 and Gfp confirmed the accuracy of sorted SOX2-GFP⁺ brightness bins (Fig. 3A, Table S1). Gene expression profiles across biological replicates within brightness bins were highly consistent; transcriptomes of SOX2^HiMed and SOX2^MedLow cells were similar, and SOX2^High and SOX2^Low cells were most distinct (Fig. 3B).

In addition to the top 50 differentially expressed genes (FDR adjusted P<0.05) (Fig. 3C), type I TRC markers were enriched in SOX2^High cells (Fig. 3D). Published reports suggest that SOX2 is expressed by type I cells in CVP taste buds (Suzuki, 2008; Takeda et al., 2013); however, identifying individual type I TRCs in tissue sections is problematic (see Miura et al., 2014). NTPDase2 IF of dispersed CVP cells from Sox2^GFP mice revealed many SOX2-GFP^Bright cells were NTPDase2 and KRT8⁺, supporting a type I TRC identity (Fig. 3F). KRT8⁺ and NTPDase2^neg cells, which are likely type II and III TRCs, were SOX2-GFP^neg (Fig. 3F), consistent with limited SOX2 expression in type II and III TRCs (Suzuki, 2008). Shh was also highly enriched in SOX2^High cells (Fig. 3E). Hybridization chain reaction in situ hybridization for Shh confirmed colocalization of Shh and SOX2 IF in one or two cells in the basal compartment of taste buds, consistent with the known location of Shh⁺ post-mitotic precursors (classically defined as type IV TRCs) (Kinnamon, 1987) (Fig. 3G). Thus, SOX2^High cells include both type I TRCs and SHH⁺ postmitotic taste precursor cells (Miura et al., 2006, 2014).

We next interrogated the dataset for GO term enrichment (Table S2). With significance at P<0.01, SOX2^High cells were enriched for 145 GO terms, SOX2^HiMed for 24 and SOX2^Low for 73, whereas SOX2^MedLow cells lacked significantly enriched terms. For SOX2^High, 30/145 top terms were relevant to development, differentiation or morphogenesis of non-neural tissues. Additional enriched terms were relevant to: (1) nervous system development and function (27/145); and (2) cell migration/locomotion (14/145). SOX2^HighMed cells were also enriched, but less so, for terms we categorized as development, differentiation or morphogenesis of non-neural tissues (4/24) with no neural-associated terms. Finally, SOX2^Low GO terms were primarily relevant to development, differentiation or morphogenesis of non-neural tissues (34/73), whereas nervous system-related terms were less frequent (6/73).

Because SOX2^High and SOX2^Low cells had the most significantly enriched GO terms, we analyzed these further (Table S2). Terms were grouped into three categories relevant to development, differentiation or function of: (1) nervous system; (2) soft tissue – kidney, muscle, lung and gut; or (3) hard tissue – bone, teeth, hair and skin. SOX2^High were enriched for both neural (27/143) and soft (20/143) tissue, with only two hard tissue terms, whereas SOX2^Low were enriched for hard tissue (13/73, seven of the top 20) and soft tissue (10/73) GO terms, with only 7/73 related to the nervous system.

In summary, neural and endodermal terms were enriched in SOX2^High cells, terms relevant to ‘hard’ tissue formation were over-represented in the SOX2^Low population, whereas intermediate populations (SOX2^HiMed and SOX2^MedLow) were less defined. As TRCs are modified epithelial cells with neural characteristics (Roper, 2007), which in the CVP are derived from endoderm (Rothova et al., 2012), and non-taste epithelium is keratinized (Cane and Spearman, 1969), GO analysis suggested that SOX2^High cells have greater taste lineage potential and that SOX2^Low cells are progenitors of non-taste keratinized epithelium; however, the potency of intermediate SOX2-expressing cells was unresolved. Thus, we next tested whether differential SOX2 expression correlates with TRC production in organoids.

Lingual organoids were generated from SOX2-GFP⁺ cells from each fluorescence bin as in Fig. 3A. qPCR for TRC markers revealed all SOX2 organoids expressed Kcnj1 (type I marker); Gnat3, Tas1r2, and Pou2f3 (type II); and Pkd2l1, Snap25, and Ascl1 (type III) but at significantly lower levels than LGR5 organoids (Fig. 4A, Fig. S1A). Among SOX2 organoids, those from SOX2^High and SOX2^HiMed progenitors tended to have higher TRC marker expression than SOX2^MedLow and SOX2^Low organoids (Fig. 4A, Fig. S1A). However, whole organoid IF revealed significant differences in TRC differentiation across organoid types. Specifically, SOX2^High and SOX2^HiMed organoids generated all TRC types – albeit with limited production of type III TRCs by SOX2^HiMed organoids (Fig. 4B,C), suggesting taste lineage competency of SOX2^High and SOX2^HiMed progenitors differ. Further, SOX2^MedLow and SOX2^Low organoids exhibited little TRC differentiation; organoids generally lacked type III TRCs and had only occasional type I and II TRCs. In sum, although qPCR data suggested some TRC marker expression across all organoid types, IF revealed TRC differentiation was limited primarily to higher expressing SOX2 organoids.

We also assessed non-taste lineage production in organoids from different SOX2⁺ progenitors. All SOX2 organoids had comparable Krt14 expression and KRT14⁺ progenitors were consistently located externally (Fig. 4E-H). As expected, SOX2^High and SOX2^HiMed organoids expressed Kcnq1 more highly (Fig. 4D) and contained more KRT8⁺ TRCs (Fig. 4E″,F″) than SOX2^MedLow and SOX2^Low organoids (Fig. 4G″,H″). While SOX2^High and SOX2^HiMed organoids contained non-taste cells (Fig. 4E‴,F‴), SOX2^MedLow and SOX2^Low organoids expressed strikingly high levels of Krt13 (Fig. 4D) and comprised mostly KRT13⁺ cells (Fig. 4G‴,H‴). KRT8⁺ cells were essentially absent in SOX2^Low organoids (Fig. 4H″). Although present sporadically in SOX2^MedLow organoids, KRT8⁺ cells were consistently KRT14⁺ (Fig. 4G′-G‴), which is reminiscent of immature taste-fated daughters (see Fig. 2 above). Cells that were both KRT8⁺ and KRT14⁺ were not observed in SOX2^Low organoids, suggesting further differences in competency between the two dim SOX2⁺ populations.

In rodents, non-taste epithelium renews in 5-7 days (Potten et al., 2002), whereas taste cells renew every 10-30 days (Beidler and Smallman, 1965; Perea-Martinez et al., 2013). SOX2^MedLow and SOX2^Low organoids, composed of rapidly renewing non-taste epithelium, grew noticeably more than SOX2^High and SOX2^HiMed organoids, which house taste and non-taste lineages and had growth comparable with LGR5 organoids (Fig. S2A,B). Compared with other SOX2 populations or LGR5⁺ cells, SOX2^High cells generated ∼50% fewer organoids (Fig. S2C), consistent with our demonstration that the SOX2^High cell population also contains many post-mitotic SHH⁺ precursor cells and type I TRCs (see Fig. 3). Thus, we surmise roughly half of SOX2^High cells are taste lineage-competent progenitors.

In vivo, progenitors adjacent to taste buds are sensitive to Hh, i.e. they express the Hh target gene Gli1 (Miura et al., 2001). Our transcriptome analysis revealed Hh pathway genes are differentially expressed across SOX2⁺ populations, including target genes Gli1 and Ptch1 (Fig. 5A). In CVP tissue sections, we found that many SOX2⁺ basal cells outside buds are Gli1⁺ (Fig. 5B-B‴), suggesting that some SOX2⁺ progenitors are H responsive in vivo. In fact, Hh signaling is required for TRC differentiation, and forced SHH expression induces formation of taste buds comprising type I, II and III TRCs, as well as SOX2 expression in adult mouse tongue (Castillo et al., 2014; Castillo-Azofeifa et al., 2018; Kumari et al., 2015; Liu et al., 2013). These findings suggested that increased Hh signaling could boost taste lineage production in organoids.

Organoids from each SOX2⁺ bin were treated with the Smoothened agonist SAG during the differentiation phase of culture (Fig. 5C). SAG treatment increased Gli1 expression in all organoid types and Sox2 expression was also upregulated (Fig. 5D), the latter consistent with findings in mice where Hh signaling positively regulates SOX2 expression (Castillo et al., 2014). Despite activating the pathway, however, SAG minimally impacted TRC marker gene expression. Although Kcnj1 (type I cell marker) was upregulated in all but SOX2^MedLow organoids, changes in Gnat3 and Pou2f3 (type II), and Pkd2l1 and Ascl1 (type III) expression were inconsistent, highly variable and limited across all organoid conditions (Fig. 5E, Fig. S3). Our results are consistent with previous reports, where excess SHH did not drive TRC differentiation in LGR5 organoids (Ren et al., 2014). Hh pathway inhibition has been reported to reduce expression of a proliferation marker, Mki67, in LGR5 organoids (Ren et al., 2017); here, however, SAG did not increase Mki67 expression in SOX2 organoids (Fig. 5F). In summary, Hh appears to be dispensable for TRC homeostasis in SOX2-derived organoids.

WNT/β-catenin regulates taste bud homeostasis in vivo. Specifically, β-catenin is required for stem cell proliferation and maintenance (Gaillard et al., 2017, 2015; Xu et al., 2017), and drives differentiation of TRCs both in vivo and in LGR5 organoids (Ren et al., 2014). We found that WNT pathway transcripts are differentially expressed among SOX2⁺ cell populations: target genes Tcf7, Tcf7l1, Lef1, Lgr5 and Lgr6, and frizzled receptors (FZDs) Fzd1, Fzd2, Fzd3, Fzd4, Fzd8 and Fzd10 are enriched in SOX2^High and SOX2^HiMed cells; and ligands Wnt3, Wnt3a, Wnt10a and Wnt10b are enriched in SOX2^MedLow and SOX2^Low progenitors (Fig. 6A). Our culture medium includes ample WNT3A, but as FZDs are low in SOX2^MedLow and SOX2^Low cells, exogenous WNT ligand may be insufficient to drive TRC formation in organoids from these progenitors. To test whether increased β-catenin increased or promoted TRC production in SOX2^High and SOX2^HighMed, or SOX2^MedLow and SOX2^Low organoids, respectively, cultures were treated with CHIR99021 (CHIR), which upregulates β-catenin signaling downstream of FZDs (An et al., 2010) (Fig. 6B). Lef1, a WNT/β-catenin target gene, was significantly increased in all CHIR-treated SOX2 organoids (Fig. 6C). Although WNT/β-catenin upregulates SOX2 in embryonic tongues (Okubo et al., 2006), the effect of CHIR on Sox2 expression was inconsistent in organoids derived from adult progenitors (Fig. 6C). Consistent with our hypothesis, however, CHIR substantially increased type I, II and III TRC marker expression in SOX2^High and SOX2^HiMed taste-competent organoids (Fig. 6D), as well as in SOX2^MedLow and SOX2^Low organoids, which, under control conditions, have limited taste differentiation potential.

In the CVP of adult mice, TRC renewal is supported by LGR5⁺ and SOX2⁺ progenitors (see Finger and Barlow, 2021 for a review); however, questions remain as to whether LGR5⁺ and SOX2⁺ progenitors have comparable taste lineage potential. Furthermore, SOX2 is variably expressed by basal keratinocytes within and around taste epithelium, and by a subset of postmitotic cells within taste buds, suggesting SOX2 cells likely represent a mixed cell population, not all of which are taste-competent progenitors.

Individual LGR5⁺ progenitors generate CVP taste and non-taste lineages in organoids. To test the taste potency of SOX2⁺ CVP cells, we isolated differentially expressing SOX2-GFP⁺ cells from adult mouse CVP and compared their ability to generate lingual organoids containing both taste and non-taste lineages with that of LGR5⁺ progenitors. We found that higher expressing SOX2⁺ progenitors are taste competent, and produce organoids composed of all TRC types and non-taste cells. Additionally, our analysis revealed that roughly half of SOX2^High cells are either type I glial-like TRCs, confirming the findings of Suzuki (Suzuki, 2008), or SHH⁺ taste precursor cells (Miura et al., 2014), which are both postmitotic and therefore unlikely to generate organoids. Progenitors expressing low levels of SOX2, by contrast, are minimally taste-competent and generate organoids composed primarily of non-taste epithelium. Our results are consistent with previous suggestions that basal keratinocytes outside of taste buds expressing high levels of SOX2 are taste progenitors (Castillo-Azofeifa et al., 2018; Ohmoto et al., 2017; Okubo et al., 2006). Importantly, organoid technology allows the dissection of the potency of discrete progenitor populations, at least in vitro, which is difficult to accomplish with genetic mouse models in vivo.

SOX2 dose affects the behavior of embryonic stem cells (ESCs), as small experimentally induced changes in expression can drive ESC differentiation (e.g. Kopp et al., 2008). In embryonic neural ectoderm, the levels of SOX2 and other SOXB family members tightly regulate the balance between progenitor proliferation and differentiation, with overexpression driving proliferation and depletion, leading to cell cycle exit and differentiation (see Sarkar and Hochedlinger, 2013). In murine hippocampus, SOX2 is required for stem cell maintenance (Ellis et al., 2004) and in trachea it is required for progenitor proliferation and genesis of the proper proportions of functional cell types (Que et al., 2009). The dose-dependent effects of SOX2 are mediated by interactions with transcriptional co-factors, which themselves are dependent on SOX2 level (Sarkar and Hochedlinger, 2013). In taste epithelium, SOX2 is required for continual renewal of TRCs in FFP and CVP (Castillo-Azofeifa et al., 2018; Ohmoto et al., 2020; Okubo et al., 2006), and higher SOX2 expression is associated with progenitors adjacent to taste buds in each taste field (Castillo-Azofeifa et al., 2018; Okubo et al., 2006). Here, our organoid data suggest that differential SOX2 expression predicts the taste lineage potential of CVP progenitors; however, the transcriptional co-factors that affect SOX2 function remain to be explored.

In adult mice, Hh and WNT/β-catenin signaling pathways regulate taste epithelial homeostasis. Hh is required for taste bud maintenance, as inhibition abolishes pro-TRC differentiation signals (Castillo-Azofeifa et al., 2017, 2018; Lu et al., 2018). Furthermore, Hh functions upstream of SOX2, as overexpression of SHH in lingual progenitors induces excess and ectopic taste buds that express high SOX2 (Castillo et al., 2014; Golden et al., 2021); SOX2 is also required for the formation of endogenous and ectopic taste buds downstream of Hh (Castillo-Azofeifa et al., 2018; Okubo et al., 2006). In lingual organoids derived from differentially expressing SOX2⁺ progenitors, SAG treatment increased Sox2 expression; however, TRC differentiation markers were little changed. Similarly, although SOX2 is required for embryonic taste bud development, its forced expression is not sufficient to drive taste bud formation (Okubo et al., 2006). We have argued that Hh may function in vivo to promote changes in cell adhesion and locomotion that are permissive to TRC differentiation from progenitors located outside taste buds, processes that are, in part, regulated by SOX2 (Golden et al., 2021). However, as taste bud structures do not form in organoids, we hypothesize that Hh and its regulation of SOX2 may be dispensable for differentiation of scattered TRCs in vitro.

In mice, WNT/β-catenin is required for progenitor proliferation and survival, and, at higher levels, impacts TRC fate (Gaillard and Barlow, 2021; Gaillard et al., 2017, 2015; Xu et al., 2017). WNT is also required for the production of many epithelial organoid systems, including lingual organoids to support growth (Ren et al., 2014; Shechtman et al., 2021). Transcriptome analysis revealed variable WNT pathway gene expression in different SOX2⁺ populations, suggesting TRC production from cells expressing lower levels of SOX2 might require pathway augmentation. Consistent with this model, pharmacologically increased WNT/β-catenin enhanced TRC production in taste-competent SOX2^High, SOX2^MedHi and SOX2^LowMed organoids, but to a much lesser extent in SOX2^Low organoids, suggesting that not all SOX2⁺ progenitors are competent to respond to increased WNT signaling with TRC production. These observations are consistent with findings in vivo in anterior tongue, where genetic induction of stabilized β-catenin in lingual progenitors throughout the tongue increases differentiation of TRCs but only in taste bud-bearing fungiform papillae (FFP) and not elsewhere in tongue epithelium (Gaillard et al., 2015). Notably, SOX2 is expressed at higher levels, albeit variably, in FFP epithelium (Castillo-Azofeifa et al., 2018), where production of additional TRCs is induced by increased WNT/β-catenin, and expressed at lower levels in non-taste epithelium, where elevated β-catenin does not drive excess TRC production. Overall, these data support a model where SOX2 levels predict taste competency in vivo as well as in organoids.

In summary, lingual organoid technology provides a rapid means of testing the competency of adult taste stem cells. To date, only LGR5⁺ progenitor cells have been shown to have taste stem properties in vitro (Shechtman et al., 2021; Ren et al., 2017, 2014, 2020). Here, we have used organoids to functionally define for the first time another taste stem population within adult taste epithelium – high SOX2-expressing cells. Overall, our findings support a model where numerous adult progenitor populations coordinate taste epithelial homeostasis, as has been reported in many other adult epithelia (e.g. intestine) (de Sousa and de Sauvage, 2019). Going forward, it will be important to determine the extent to which LGR5⁺ and/or SOX2⁺ progenitors are required for taste bud cell renewal to better understand how taste homeostasis is perturbed by disease and drug treatments (Gaillard and Barlow, 2021; Wang et al., 2009b). Elucidating the lineage relationship between these stem populations will provide additional insight for devising therapies to mitigate taste dysfunction. Finally, this in vitro approach can complement and extend findings in vivo, and will continue be an important tool for exploring taste epithelial homeostasis.

Mice were obtained from the Jackson Laboratory (Lgr5^{EGFP-IRES-CreERT2}, 008875; Sox2^GFP, 017592) and maintained in an AAALAC-accredited facility in compliance with the Guide for Care and Use of Laboratory Animals, Animal Welfare Act and Public Health Service Policy. Both male and female adult mice between 8-20 weeks were used. Procedures were approved by the Institutional Animal Care and Use Committee at the University of Colorado Anschutz Medical Campus.

Organoid production was as described previously (Shechtman et al., 2021). For each experiment, organoids were derived from tongue epithelium from four to eight Lgr5^EGFP or Sox2^GFP mice aged 8-20 weeks. Briefly, collagenase (2 mg/ml) and dispase (5 mg/ml) in phosphate-buffered saline (PBS) were injected beneath and beside the CVP, the epithelium peeled and then dissociated for 45 min in collagenase (2 mg/ml), dispase (5 mg/ml) and elastase (2 mg/ml) at 37°C. Cells were centrifuged (320 g at 4°C), the pellet resuspended in FACS buffer [1 mM EDTA, 25 mM HEPES (pH 7.0), 1% FBS, 1×Ca²⁺/Mg²⁺-free dPBS], passed through a 30 µm cell strainer, stained with DAPI (Thermo Fisher Scientific) and subjected to FACS on a MoFlo XDP100 (Cytomation). Debris and doublets were gated out via side-scatter (FSC and SSC-width, respectively), live (DAPI⁺) cells were enriched, then the gating of GFP⁺ signal was determined relative to background fluorescence of epithelial cells from a wild-type control mouse processed in parallel (see Shechtman et al., 2021 for gating strategy). LGR5-GFP⁺ and SOX2-GFP⁺ cells, as well as cells within the four SOX2-GFP⁺ brightness bins, were gated and collected as outlined in Figs. 2A and 3A, respectively; GFP^neg cells in all FACS experiments were discarded. Sorted GFP⁺ cells were plated in 48-well plates at 200 cells/well in 15 µl Matrigel in WENR+AS to support growth (days 0-6) and moved to WENR to promote TRC differentiation (days 6-12) (Shechtman et al., 2021). WENR is 50% WRN (WNT/RSPO/Noggin)-conditioned media (Miyoshi and Stappenbeck, 2013); 1× Glutamax, 1× HEPES, 1× penicillin-streptomycin, 1× B27 supplement, 1× gentamicin (Gibco); 1× primocin (InvivoGen); 25 ng/ml murine Noggin, 50 ng/ml murine EGF (Peprotech); and 1 mM nicotinamide, 1 mM N-acetyl-L-cysteine (Sigma). WENR+AS is WENR with 500 nM A8301 (Sigma) and 0.4 µM SB202190 (R&D Systems). Y27632 (10 µm; Stemgent) was added on days 0-2 to promote survival. Medium was changed every 2 days.

Lgr5^EGFP and Sox2^GFP mice were perfused transcardially with periodate-lysine-paraformaldehyde, tongues dissected from the jaw, incubated for 3 h at 4°C in 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB) and placed in sucrose (20% in 0.1 M PB) overnight at 4°C. Samples were embedded in OCT Compound (Tissue-Tek), 12 µm cryosections collected on SuperFrost Plus slides (Thermo Fisher Scientific) and stored at −80°C.

Sections were washed in 0.1 M PBS, incubated in blocking solution (BS: 5% normal goat or donkey serum, 1% bovine serum albumin, 0.3% Triton X-100 in 0.1 M PBS at pH 7.3) for 1.5 h at room temperature followed by primary antibodies diluted in BS for 2 nights at 4°C. Sections were rinsed in PBS+0.1% Triton, incubated for 1 h at room temperature in secondary antibodies in BS, followed by DAPI nuclear counterstaining. Slides were coverslipped with ProLong Gold (Thermo Fisher Scientific). Organoids were harvested as described previously (Shechtman et al., 2021). Briefly, organoids were incubated in Cell Recovery Solution (Corning) at 4°C, washed in 0.1 M PBS, fixed in 4% PFA and stored at 4°C in PBS with 1% BSA. For IF, organoids were incubated in BS (2 h), then with primary antibodies in BS for 3 nights at 4°C, washed with PBS+0.2% Triton and incubated with secondary antibodies in BS overnight at 4°C. Organoid nuclei were counterstained with DAPI, washed with 0.1 M PB and mounted on SuperFrost Plus slides in Fluoromount (Southern Biotech). For dispersed cell immunostaining, CVP epithelia from two Sox2^GFP mice were dissociated, and cells mounted on poly-D-lysine (1:10 in H₂O) and fibronectin (1:100 in 1× PBS)-coated coverslips, fixed in 4% PFA for 2 min at room temperature, washed with 0.1 M PBS, and stored at −80°C until processed for IF as for tissue sections. Antibodies are listed in Table S3.

Molecular Instruments designed and produced probes against Sox2 (NM_011443.4), Lgr5 (NM_010195.2), Gli1 (AB025922.1) and Shh (NM_009170.3). Methods were adapted from the manufacturer's protocol. Frozen sections were incubated in 4% PFA for 10 min at room temperature, washed with 0.1 M PBS and incubated in 2 µg/ml Proteinase K for 2 min at room temperature. Sections were incubated in triethanolamine solution, 12 M HCl and acetic anhydride in DEPC-treated water for 10 min at room temperature, washed with PBS, incubated in hybridization buffer for 55 min at 37°C, and then with 1.2 pmol probe in hybridization buffer overnight at 37°C. Sections were washed with 75% wash buffer/25% 5× SSCT (20× SSC and 10% Tween20, ultrapure water), 50% wash buffer/50% 5× SSCT, 25% wash buffer/75% 5× SSCT, and 100% 5× SSCT, each for 20 min at 37°C, then with 100% 5× SSCT for 20 min at room temperature. Sections were incubated in amplification buffer for 1 h, then in denatured hairpin solution (6 pmol hairpin 1 and 2 in amplification buffer) overnight. Slides were washed with 5× SSCT and mounted in DAPI-containing ProLong Gold.

CVP sections and organoids were imaged with a Leica TCS SP8 laser-scanning confocal microscope with LAS X software. Sequential z-stacks of CVP sections were acquired as 0.75 µm optical sections, and organoids via 2 µm optical sections. For analysis, investigators were not made aware of the condition. Immunolabeled cells per organoid were tallied when: (1) a cell is immunomarker-positive; and (2) an immunostained cell has a DAPI⁺ nucleus. Each organoid (typically 5-25 per well) was tallied for the presence of TRC types, and each immunostaining experiment was repeated two or three times.

Organoids were harvested as described previously (Shechtman et al., 2021). Briefly, plates were placed on ice for 30 min and organoids freed from matrix by scratching with a pipet tip. Organoid samples were centrifuged and resuspended in RLT buffer (Qiagen) with β-mercaptoethanol. RNA was extracted using a RNeasy Micro Kit (Qiagen), quantified via Nanodrop (Thermo Fisher Scientific) and reverse transcribed with an iScript cDNA synthesis kit (Bio-Rad). Power SYBR Green PCR Master Mix (Applied Biosystems) was used for qPCR reactions on a StepOne Plus Real-Time PCR System (Applied Biosystems, Life Technologies). Relative gene expression was assessed using the ΔΔCT method, with Rpl19 as the housekeeping gene. Primers are listed in Table S3. Organoids from three wells of a 48-well plate were pooled per RNA sample and three samples were collected per experiment for RT-PCR.

Growth curves were generated from Incucyte Live Cell Analysis System (Sartorius) 2D images of organoids taken daily from days 6-12. ImageJ ‘analyze particles tool’ was used to (1) set the range for expected organoid size (0.003-0.4 mm²) and (2) obtain areas of organoids within this range. Organoids were then manually reviewed and non-organoids outside the range criteria were excluded from analysis. To address the problem of spatially overlapping organoids in the z dimension, these were individually outlined manually, allowing manual area measurement to made of each organoid. If manual delineation was not possible these organoids were excluded from area measurements but included in calculating plating efficiency. After manual review, total organoid number identified in each well at day 12 was divided by the starting number of cells (200/well) to obtain plating efficiency.

CVP epithelia from 11- to 12-week-old Sox2^GFP mice (7-13 mice/replicate) were dissociated and GFP⁺ cells isolated via FACS as above. RNA was extracted with an Arcturus PicoPure RNA Isolation Kit (Thermo Fisher Scientific) and stored at −80°C. Poly-A-selected sequencing libraries were prepared using Nugen Universal Plus mRNA kit and sequenced (2×150 bp) using an Illumina NOVASeq6000 by the Genomics and Microarray Core Facility at the University of Colorado AMC. Reads were trimmed and filtered using BBDuk (v38.50; sourceforge.net/projects/bbmap/), resulting in an average of 49.6±8.8 million (±s.d.) read pairs per sample. Transcript abundance was quantified using Salmon (v1.416.1) (Patro et al., 2017) and a decoy-aware transcriptome index (Gencode M26) (Frankish et al., 2021) at a mapping rate of 85±2.5%. Transcript abundance was summarized at the gene-level using ‘tximport’ (Soneson et al., 2015) and differential expression was tested with DESeq2 (v1.28.1) (Love et al., 2014). A likelihood ratio test was performed to identify differentially expressed genes (DEGs) across all four samples. Significant DEGs were identified by padj (FDR-adjusted P<0.05). To identify DEGs specific to each SOX2-GFP bin, differential expression was tested on each group versus the remaining three. DEG lists from these analyses were used as input to TopGO (v2.50.0) for gene ontology analysis, with inclusion criteria of average normalized expression across all bins>100, log2FoldChange>1, padj<0.05. A final inclusion criterion included padj<0.05 by the above-mentioned likelihood ratio test analysis. Heatmaps were generated using ‘pheatmap’ (CRAN.R-project.org/package=pheatmap) in R using default column and row clustering methods. Display values are relative expression levels.

Normally distributed data were analyzed by ordinary one- or two-way ANOVA with Tukey's multiple comparisons post-hoc test using GraphPad Prism software. Dunn's multiple comparison test was used when data were not normally distributed. Data are mean±s.e.m. and significance was taken as P<0.05 with a confidence interval of 95%.

The authors thank Monica Brown (University of Colorado Anschutz Medical Campus Organoid and Tissue Modeling Shared Resource) and our Rocky Mountain Taste and Smell Center colleagues for valuable technical support and discussions. We also thank the Gates Center for Regenerative Medicine Genomics Core, and the University of Colorado Cancer Center Cell Technologies and Flow Cytometry Shared Resources, especially Dmitry Baturin, for cell-sorting expertise. Finally, we thank Benjamin Tiano and David Castillo-Azofeifa for early pilot data that contributed to study design.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201375.reviewer-comments.pdf