In vitro modeling of cranial placode differentiation: recent advances, challenges, and perspectives

Abstract

Cranial placodes are transient ectodermal thickenings that contribute to a diverse array of organs in the vertebrate head. They develop from a common territory, the pre-placodal region that over time segregates along the antero-posterior axis into individual placodal domains: the adenohypophyseal, olfactory, lens, trigeminal, otic, and epibranchial placodes. These placodes terminally differentiate into the anterior pituitary, the lens, and contribute to sensory organs including the olfactory epithelium, and inner ear, as well as several cranial ganglia. To study cranial placodes and their derivatives and generate cells for therapeutic purposes, several groups have turned to in vitro derivation of placodal cells from human embryonic stem cells (hESCs) or induced pluripotent stem cells (hiPSCs). In this review, we summarize the signaling cues and mechanisms involved in cranial placode induction, specification, and differentiation in vivo, and discuss how this knowledge has informed protocols to derive cranial placodes in vitro. We also discuss the benefits and limitations of these protocols, and the potential of in vitro cranial placode modeling in regenerative medicine to treat cranial placode-related pathologies.

Graphical Abstract

Introduction

The post-gastrulation embryo is comprised of three germ layers: ectoderm, mesoderm, and endoderm. Each germ layer gives rise to specific lineages that over time develop into the various tissues and organs of the body. In vertebrates, the major lineages of the ectoderm are the neuroectoderm, the non-neural ectoderm, the neural crest, and the cranial placodes. These four lineages are closely related in their positioning and development in early embryogenesis. During gastrulation, the ectoderm is induced to form neuroectoderm and non-neural ectoderm. Neuroectoderm develops into the central nervous system, while non-neural ectoderm gives rise to the epidermis of the skin. As neurulation begins, the neuroectoderm transitions to become the neural plate, while the region between the neural plate and non-neural ectoderm is established as the neural plate border. The neural plate border subsequently gives rise to the neural crest and the pre-placodal region (PPR) (Schlosser, 2010; Streit, 2007; Streit, 2004), with some contribution to the neural and non-neural ectoderm (Roellig et al., 2017; Williams et al., 2022). Neural crest cells eventually delaminate and migrate to form various derivatives of the head and trunk, while the PPR gives rise to all cranial placodes.

Cranial placodes are transient ectodermal thickenings that give rise to the hormone producing cells of the pituitary, portions of the sensory organs in the developing head (lens, olfactory epithelium, inner ear), and contribute to several cranial ganglia (von Kupffer, 1891; LeDouarin et al., 1986; Schlosser, 2010; Baker and Bronner-Fraser, 2001). Cranial placodes, along with neural crest, are believed to have arisen in vertebrates evolutionarily from the epidermal nerve plexus found in basal protochordates (Northcutt, 2005; Patthey et al., 2014). While cranial placodes develop from a common territory – the PPR – they segregate over time along the antero-posterior axis into anterior (adenohypophyseal, olfactory, and lens), intermediate (trigeminal), and posterior (otic and epibranchial) placodes (Singh and Groves, 2016). They also differ in their mode of development, forming either by invagination of the thickened epithelium into a vesicle (adenohypophyseal, olfactory, lens and otic) that will in most cases separate from the surface ectoderm (adenohypophyseal, lens and otic), or by delamination of cells from the thickened ectoderm and their subsequent coalescence into the underlying tissues (trigeminal and epibranchial). Delamination of cells is also occurring in placodes undergoing invagination with the exception of the lens. In all cases, the resulting structures undergo extensive proliferation, remodeling, and differentiation to produce the diverse array of cell types characteristic for each placode-derived organ (Schlosser, 2010; Grocott et al., 2012; Saint-Jeannet and Moody, 2014). In amniotes, organs derived from cranial placodes include the anterior pituitary, olfactory epithelium, lens, inner ear, and the trigeminal, geniculate, petrosal, and nodose ganglia (Fig. 1).

Figure 1: Cranial placodes and their derivatives originate from a common territory, the pre-placodal ectoderm.

The pre-placodal ectoderm (PPR) gives rise to the adenohypophyseal placode (AdP), olfactory placode (OlP), epibranchial placodes (EpP), otic placode (OtP), lens placode (LeP), and trigeminal placode (TrP), which will differentiate into the anterior pituitary, olfactory epithelium, geniculate, petrosal and nodose ganglia, inner ear, lens, and trigeminal ganglion, respectively.

Several classes of signaling molecules and transcription factors have been implicated in the multiple steps leading to the induction and specification of individual cranial placodes. Defining how cranial placodes develop is critical to understanding the pathologies that may arise from defects in these processes. In this review we briefly describe the factors and mechanisms underlying cranial placode formation and diversification in vivo and discuss how this fundamental knowledge has informed protocols to derive cranial placodes in vitro. We will also discuss recent advances and challenges in cranial placode generation and in vitro modeling, and their potential use in cell replacement therapies.

Development and differentiation of cranial placodes

The development of cranial placodes in vivo can be separated into three stages: (i) induction of the PPR, (ii) regionalization of the anterior, intermediate, and posterior placodal regions, and (iii) specification of individual placodes. Here we summarize some of the key signaling cues at play at these different stages to generate placodal cells and their derivatives in vivo (Table 1), and the subset of transcription factors that define their identity (Table 2).

Table 1:

Signaling cues regulating cranial placode formation in vivo, and their use for cranial placode derivation in vitro from hESCs or hiPSCs. +, indicates the type and number of cell lines tested. hi, high concentration; lo, low concentration; i, inhibition; D0-D60, days in culture.

Cranial placodes	In vivo induction and regionalizationcues	In vivo specificationcues	In vitro strategies	hESCs	hiPSCs	References
Adenohypophyseal	FGFhi BMPhi Wntlo	FGFhi BMPhi Shhhi	D0-D2: TGFβi+BMP4 D3: TGFβi D4-D15: TGFβi+SHH+FGF8/10 D15-D30: SHH+FGF8/10 D30-D60: FGF8+BMP2	+++	+	Zimmer et al., 2016
			D6-D24: SHH+BMP4	++		Ozone et al., 2016
Olfactory	FGFhi BMPhi Wntlo	FGFhi BMPhi	D0-D10: dSMADi D10-D20: FGF8 D20-D29: FGF8+DAPT	+	++	Lund et al., 2016
			D3-D9: BMPi+WNTi D9-D18: BMPi+WNTi+Rai+TGFα+FGF8 D18-D24: BMPi+TGFα+FGF8		+	Bricket et al., 2022
Lens	FGFhi BMPhi Wntlo	FGFhi BMPhi	D0-D6: Noggin D6-D20: BMP4+BMP7+FGF2	+	+	Dewi et al., 2020
			ITS medium and cell sorting	+		Mengarelli and Barberi, 2013
			D0-D5: BMP4i D6-D13: FGF2+BMP4/7 D18-D35: FGF2+WNT3A	+		Yang et al., 2010
			DO: BMP4i D6-D14: FGF2+BMP4/7	+	++	Fu et al., 2017
			D0-D2: BMP4+TGFβi D3-D12: TGFβi+FGF2	+++++	+++	Tchieu et al., 2017
			D0-D2: TGFβi+BMPi D3 onward: TGFβi	+	++	Dincer et al., 2013
Trigeminal	FGFhi Wnthi	FGFhi Wnthi PDGFhi	D0-D1: BMP4+TGFβi D2: BMP4+TGFβi+WNT D3: TGFβi+WNT D4-D12: TGFβi	+++++	+++	Tchieu et al., 2017
Otic	FGFhi BMPlo Wnthi RAhi	FGFhi Wnthi Notchhi	D0-D12: FGF3/10 D13-D19: RA+EGF		+	Lahlou et al., 2018
			D0-D12: FGF3/10	++++		Chen et al., 2012
			D0-D14: Noggin+FGF2 D14-D19: EGF+FGF2	+	++	Gunewardene et al., 2014
			D3: BMP4+TGFβi D4-D5: FGF2+BMPi	+		Koehler et al., 2013
			D0-D3: TGFβi+BMP4+FGF2lo D4-D7: BMP4i+FGF2hi D8-D12: WNT	+	+	Koehler et al., 2017
			D0-D6: TGFβl+WNTi D6-D8: FGF2+RA D8-D12: FGF2+BMPi D12-D16: FGF2+BMP4+WNT	+	+	Ealy et al., 2016
			D8-D22: FGF2		+	Ohnishi et al., 2015
			D3-D5: TGFβi+BMP4+FGF2 D6-D10: TGFβi+BMPi+FGF2+WNTi D11-D17: FGF2+WNT3A+IGF1 D18-D24: SHH+RA+EGF+FGF2+IGF1 D25-D27: EGF+FGF2+IGF1	+++		Matsuoka et al., 2017
Epibranchial	FGFhi BMPlo Wnthi RAhi	FGFhi BMPhi Wntlo	No current protocols

Table 2:

Genes activated during cranial placode formation in vivo, and their use in vitro for the identification of cranial placodes derived from hESCs or hiPSCs.

Cranial placodes	In vivo induction and regionalization genes	In vivo specification genes	In vitro markers	References
Adenohypophyseal	Six1/3/6 Otx2 Pitx3 Pax6	Pitx1/2 Lhx3/4 Six6 Otx2	Pitx1/2, Lhx3/4, Hesx1, Six6, Prop1	Zimmer et al., 2016
			Pitx1, Lhx3	Ozone et al., 2016
Olfactory	Six1/3/6 Otx2 Pitx3 Pax6	Hes5 Ngn1 Foxg1 Msx1/2	Sox1, Pax6, Emx2, FoxG1, Six1, Eya1	Lund et al., 2016
			Pou2F1, Dlx5, Emx2, Ascl1, Nestin	Bricker et al., 2022
Lens	Six1/3/6 Otx2 Pitx3 Pax6	FoxE3 Pax6 Sox2 L-Maf Prox1 Pitx3	Pax6	Dewi et al., 2020
			Six1, Eya1, Dlx3, Dlx5, Six4,FoxE3, Prox1	Mengarelli and Barberi, 2013
			Pax6, Sox2, Six3	Yang et al., 2010
			Six1, Eya1, Dlx3, Pax6, Sox1, Prox1, FoxE3	Fu et al., 2017
			Six1, Pax6, Six3	Tchieu et al., 2017
			Pax6, Dlx3, Six1, Eya1, Dach1, FoxG1, Pitx3	Dincer et al., 2013
Trigeminal	Pax3	Pax3 Ngn1/2 Isl1	Six1, Pax3	Tchieu et al., 2017
Otic	Gbx2 Dlx Foxi3 Pax2/8	Pax2/8 Foxi3 Sox2	Sox2, Pax2	Lahlou et al., 2018
			Pax8, Pax2, Sox2, FoxG1	Chen et al., 2012
			Pax2	Gunewardene et al., 2014
			Dlx3, Pax8	Koehler et al., 2013
			Pax8, Sox2, TFAP2A, Ecad, Ncad, Pax2	Koehler et al., 2017
			Six1, Eya1, Pax2, Pax8, Sox2,	Ealy et al., 2016
			Six1, Pax2	Ohnishi et al., 2015
			Dlx5, Dlx3, Six4, Eya1, FoxG1, Sox2, Pax8,	Matsuoka et al., 2017
Epibranchial	Gbx2 Dlx Foxi3 Pax2/8	Pax2 Foxi2 Sox3	No current protocols

Induction and regionalization of the pre-placodal region

The PPR is defined by the expression of the Six family of transcription factors and their co-factors belonging to the Eya family (Moody and LaMantia, 2015; Singh and Groves, 2016). At the end of gastrulation, the expression of these genes abuts the anterior neural plate border in a U shape domain and is eventually maintained in most individual placodes (Kobayashi et al., 2000; Zou et al., 2004; Christophorou et al., 2009; Sato et al., 2010). This expression pattern is conserved across species from Ciona to mouse (Manni et al., 2004; Mazet et al., 2005). SIX proteins are master regulators of PPR formation by promoting expression of genes essential to maintain the PPR, while repressing genes associated with epidermal and neural crest fates (Brugmann et al., 2004). These proteins all contain DNA binding domains as well as co-factor binding domains (Kobayashi et al., 2001). EYA proteins are tyrosine phosphatases that act as transcription co-activators of SIX proteins (Li et al., 2003).

Work across several species has shown that during gastrulation, the different regions of the ectoderm are specified in response to varying levels of fibroblast growth factor (FGF), bone morphogenetic protein (BMP) and Wnt signaling (Table 1), and the relative importance of these factors differs in a species-specific manner. The intricate interplay between the signaling molecules driving PPR specification and its subsequent regionalization has been extensively reviewed (Schlosser, 2010; Grocott et al., 2012; Saint-Jeannet and Moody, 2014; Moody and LaMantia, 2015), and is only briefly summarized here.

The neural ectoderm is induced by FGF signals and by antagonizing Wnt and BMP signaling pathways (Streit et al., 2000; Wilson et al., 2000; De Robertis and Kuroda, 2004). BMP inhibition is mediated by the activity of BMP antagonists (Noggin, Chordin, Cerberus, and Follistatin), which sequester BMP proteins in the extracellular space preventing binding and activation of the cognate receptors (Zimmerman et al., 1996; Larrain, 2000; Liu et al, 2000; Piccolo et al., 1999; Hemmati-Brivanlou et al., 1994; Papanayotou et al., 2008; Rogers et al., 2011; Lunn et al., 2007). By contrast non-neural ectoderm fate is driven by active BMP and Wnt signaling (Bhat et al., 2013; Kwon et al., 2010; Hong and Saint-Jeannet, 2007). The region in between these territories, the neural plate border, forms in response to partial attenuation of BMP signaling, mediated by the same BMP antagonists, and active FGF and Wnt signaling (Groves and LaBonne, 2014; Brugmann and Moody, 2005; Streit and Stern, 1999; McLarren et al, 2003, Garcia-Castro et al., 2002). The neural plate border has been primarily described as a region that expresses neither neural nor non-neural genes, and defined by expression of genes of the Tfap2, Dlx, Msx, Zic, and Pax families (Lou et al., 2003; Yang et al., 1998; Phillips et al., 2006; Hong and Saint-Jeannet, 2007; Schlosser and Ahrens, 2004; Khudyakov and Bronner-Fraser, 2009; Grocott et al., 2012; Pieper et al., 2012). However, recent evidence in chicken embryos indicates that until early neurulation, the neural plate border is a territory of cells with heterogeneous gene expression profiles rather than having a unique molecular signature (Roellig et al. 2017; Williams et al., 2022; Thiery et al., 2023). As neurulation begins, Wnt inhibition at the neural plate border cooperates with BMP antagonists and FGF ligands to promote PPR fate (Brugmann et al., 2004; Litsiou et al., 2005; Ahrens and Schlosser, 2005; Hong and Saint-Jeannet, 2007; Pieper et al., 2012), while active Wnt signaling and BMP attenuation are both required for neural crest fate (Saint-Jeannet et al., 1997; LaBonne and Bronner-Fraser, 1998; Monsoro-Burq et al. 2003; Litsiou et al., 2005; Kwon et al., 2010). In addition, retinoic acid (RA) signaling is important for restricting the posterior boundary of the PPR (Dubey et al., 2018), thereby confining placode development to the head region as compared to neural crest formation, which occurs along the entire body axis (Chen et al., 2001; Chen et al., 1994).

Lineage and fate map analyses of the olfactory, lens, trigeminal, epibranchial, and otic placodes performed in fish, frog, or chicken embryos (Bhattacharyya et al., 2004; Streit, 2002; Xu et al., 2008; Pieper et al., 2011; Dutta et al., 2005; Kozlowski et al., 1997; Whitlock and Westerfield, 2000) indicate that in the PPR precursors for different placodes are initially intermingled and eventually separate into distinct domains along the antero-posterior axis. Gradients of key signaling molecules help establish the necessary microenvironment for individual placode specification and differentiation to proceed along the antero-posterior axis. Without these signaling gradients, the placodal regions cannot be properly initiated, leading to a loss of these domains.

The anterior placodal region relies on gradients of FGF, BMP, and Wnt signaling for proper formation - high levels of FGF and BMP activity are required, while Wnt signaling must be kept at low levels (Hong and Saint-Jeannet, 2007; Bailey et al., 2006; Glavic et al., 2004; Park and Saint-Jeannet, 2008). These inductive cues originate from the underlying mesoderm as well as the PPR itself (Chapman et al., 2004; Ogita et al., 2001; Esterberg and Fritz, 2009). This includes Wnt antagonists of the Dickkopf family (Dkk1 for example) that maintain low levels of Wnt in this region (Brugmann et al., 2004; Groves and LaBonne, 2014). Gain- and loss-of-function studies in frog, fish, and chicken embryos have shown that factors important for anterior placode induction include Six1/3/6, Otx2, Pitx3, and Pax6 (Christophorou et al., 2009; Sato et al., 2010; Kobayashi et al., 2001; Bhattacharyya et al., 2004; Steventon et al., 2012; Dutta et al., 2005; Schlosser, 2010; Singh and Groves, 2016; Koontz et al., 2023).

For the intermediate placodal region, high levels of FGF and Wnt activity are required. Pax3 is an important mediator for intermediate placode identity (Singh and Groves, 2016; Koontz et al., 2023; Stark et al., 1997; Canning et al., 2008), and repression of Pax6 by Pax3 helps maintain proper regionalization of this domain (Wakamatsu, 2011).

In addition to FGF, BMP, and Wnt signaling, the posterior placodal region requires high RA levels to establish the posterior placodal boundary (Dubey et al., 2018). Similar to the intermediate region, the posterior placodal region also requires high levels of FGF and Wnt signaling (Freter et al., 2008; Nechiporuk et al., 2005; McCarroll et al., 2013; Culbertson et al., 2011). FGF is the main inducer of the otic-epibranchial placode domain (OEPD) (Groves and Fekete, 2012), the precursor territory for both otic and epibranchial placodes (Ladher et al., 2000, 2005; Wright and Mansour, 2003; Zelarayanet al., 2007). This region also requires BMP inhibition, as BMP is an anterior inducing signal (Singh and Groves, 2016). Key regulators of posterior placode identity are Gbx2, Dlx, Foxi3, and Pax2/8 (Khatri et al., 2014; Edlund et al., 2014; Birol et al., 2015; Hans et al., 2007; Padanad and Riley, 2011).

A comprehensive description of the events pertaining to PPR regionalization, and the role of the various transcription factors defining these domains have been reviewed in recent years (Schlosser, 2010; Grocott et al., 2012; Saint-Jeannet and Moody, 2014; Moody and LaMantia, 2015).

Specification of placodes

Once the PPR is regionalized, it is further subdivided into six domains with unique placodal identities. At the most anterior part of the embryo, the adenohypophyseal placode requires activation of FGF and BMP for the formation of Rathke’s pouch, and sonic hedgehog (SHH) signaling from the adjacent ectoderm and ventral midbrain for specification. In this context, SHH signaling represses lens and olfactory fates (Singh and Groves, 2016; Koontz et al., 2023; Sheng et al., 1996; Treier et al., 1998; Ericson et al., 1998, Sheng et al., 1997; Karlstrom et al., 1999) (Table 1). At this stage, the factors that define the molecular identity of the adenohypophyseal placode are Pitx1/2, Lhx3/4, Six6, and Otx2 (Singh and Groves, 2016; Kelberman et al., 2009; Sanchez-Arrones et al., 2015; Sheng et al., 1996; Steventon et al., 2012) (Table 2).

The lens and olfactory placodes form in adjacent regions and rely on the timing of exposure to gradients of FGF and BMP to elicit their separation (Sjödal et al., 2007). Short exposure to BMP and high FGF will specify the olfactory placode, while long exposure to BMP and low FGF will specify the lens placode (Bhattacharyya et al., 2004) (Table 1). Wnt signaling plays a role in patterning the lens and olfactory placodes by repressing Pax6, thereby defining the prospective lens placode (Canning et al., 2008; Lassiter et al., 2007; Grocott et al., 2012). RA is also an important lateral signal from the underlying frontonasal mesenchyme (Kawauchi et al, 2004; LaMantia et al., 2000; Bhasin et al., 2003). Key factors involved in olfactory specification include Hes5, Ngn1, Foxg1, and Msx1/2, while these factors for the lens include FoxE3, Pax6, Sox2, L-Maf, Prox1, and Pitx3 (Bailey et al., 2006; Litsiou et al., 2005; Patthey et al., 2008; Choi and Goldstein, 2018; DeHamer et al., 1994; Goudreau et al., 2002; Liu et al., 2006; Purcell et al., 2005; Smith et al., 2009) (Table 2).

The trigeminal placode is specified by the simultaneous activation of FGF, platelet derived growth factor (PDGF), and Wnt signaling pathways. FGF and Wnt signaling induce and maintain Pax3 expression in this intermediate domain (Canning et al., 2008; Lassiter et al., 2007; Shigetani et al., 2008). PDGF along with Notch signaling, induce the ophthalmic subdivision of the trigeminal placode, and helps regulate neurogenesis (McCabe and Bronner-Fraser, 2008; Lassiter et al., 2010; Voelkel et al., 2014; Lassiter et al., 2014) (Table 1). A majority of these specifying signals arise from the brain, which is immediately adjacent to the trigeminal placode domain (Stark et al., 1997; Canning et al., 2008). The major transcription factors involved in trigeminal specification comprise Pax3, Ngn1/2, and Isl1 (Singh and Groves, 2016; Stark et al., 1997) (Table 2).

The otic and epibranchial placodes develop initially from a common domain, the OEPD, which is induced by signals from the paraxial mesoderm (Groves and Bronner-Fraser, 2000; Kil et al., 2005; Ladher et al., 2000). A long exposure to FGF and low BMP signaling lead to epibranchial specification, while a short pulse of FGF and high BMP signaling result in otic specification (Ladher et al., 2000). The otic placode also relies on active Wnt signaling, whereas for epibranchial placode formation Wnt signaling must be inhibited (Singh and Groves, 2016; Freter et al., 2008; Ohyama et al., 2006; Park & Saint-Jeannet, 2008) (Table 1). Key transcriptional regulators of otic placode specification include Pax2/8, Foxi3, and Sox2, while factors regulating epibranchial specification comprise Pax2, Foxi2, and Sox3 (Freter et al., 2008; Nechiporuk et al., 2005; McCarroll and Nechiporuk, 2013; Ohyama et al., 2006; Solomon et al., 2004; Hans et al., 2004) (Table 2).

Terminal differentiation of cranial placodes

Following this multistep process, cranial placodes ultimately differentiate into several key organs in the head. The adenohypophyseal placode gives rise to the anterior pituitary, an endocrine gland located under the hypothalamus, which plays a major function in growth, metabolism regulation, and reproduction (Rizzoti, 2015; Sheng et al., 1997). The olfactory placode invaginates to become the olfactory pit that develops into the olfactory epithelium of the nose, which is composed of olfactory and GnRH neurons, along with basal (stem) cells and sustentacular (supporting) cells (Koontz et al., 2023; Carter et al., 2004). The lens placode provides the epithelium and the crystallin-rich fibers of the lens of the eye (Chow and Lang, 2001). The trigeminal placode develops into the trigeminal ganglia, composed of the ophthalmic and maxillomandibular lobes, which gather sensory input from the head and face, and control the motor function of muscles responsible for chewing (Schlosser and Northcutt, 2000). The major components of the inner ear, the cochlea and the semicircular canals – including mechanoreceptors, neurons, structural, and support cells – are involved in hearing, balance, and detection of acceleration, and are derived from the otic placode (Chen and Streit, 2013). The epibranchial placodes (geniculate, petrosal, and nodose) contribute to the distal portion of the ganglia of cranial nerves VII (facial), IX (glossopharyngeal), and X (vagus), generating the large neurons of these ganglia (Saint-Jeannet and Moody, 2014; Schlosser, 2003).

Trends in in vitro placode derivation

Pathologies associated with defects in cranial placode development may result in hormone imbalance and a broad array of sensory deficit disorders affecting the orofacial complex. Hypopituitarism refers to a condition in which the pituitary gland is absent or deficient in its functions (Regal et al., 2001). Defects in the production of olfactory sensory neurons can lead to congenital anosmia (Boesveldt et al. 2017), while defects in the production of GnRH neurons can cause hypogonadotropic hypogonadism (Kim, 2015). Improper development of the lens placode can cause congenital cataracts and associated loss of vision (Yang et al., 2010; Fu et al., 2017), whereas disruption of otic placode development can result in hearing loss and balance disorders (Bermingham-McDonogh and Reh, 2011; Brigande and Heller, 2009).

In order to study cranial placodes and their derivatives and generate cells for therapeutic purposes that can be used for compound screening or cell replacement therapies, several groups have turned to in vitro derivation of placodal cells from human embryonic stem cells (ESCs) or induced pluripotent stem cells (iPSCs). These protocols are primarily driven by the information accumulated over the years on the in vivo signaling cues necessary to generate cranial placodes in various species, and transposing and/or adjusting these cues to the in vitro system. The majority of these protocols are two-dimensional (2D) cell differentiation methods, as our understanding of most sensory systems development is rather limited to build three-dimensional (3D) organoids the way they are available for other organ systems. Here we summarize some of the key protocols developed in recent years to generate (Table 1) and identify (Table 2) placodal cells and their derivatives in vitro (Fig. 2).

Figure 2: In vitro modeling of human cranial placode differentiation.

Summary of the signaling pathways commonly involved in the production of adenohypophyseal placode (AdP), olfactory placode (OlP), lens placode (LeP), trigeminal placode (TrP), and otic placode (OtP) cell types in vitro from human embryonic stem cells (ESCs) or induced pluripotent stem cells (iPSCs). The repertoire of genes commonly used to characterize each cranial placode derivative in vitro is indicated.

Adenohypophyseal placode

The positioning and patterning of the adenohypophyseal placode during development relies on opposing activities of BMP and SHH signaling. In vitro development of these adenohypophyseal cells has attempted to use similar developmental cues. For example, one protocol employs a stepwise developmental trajectory that includes inhibition of transforming growth factor β (TGFβ) signaling, via the small molecule inhibitor SB431542, and addition of SHH, BMP, and FGF ligands (Zimmer et al., 2016). The cell environment is changed throughout differentiation to mimic the shifts in signaling pathways seen in vivo. This protocol results in induction of pituitary placodal cells by day 15, characterized by the expression of Pitx1/2, Lhx3/4, Hesx1, Six6, and Prop1, while lacking expression of hypothalamus-specific genes such as Nkx2.1. By day 30 of differentiation, pituitary-like cells are a mix of ACTH⁺ corticotrophs, GH⁺ somatotrophs, and FSH⁺LH⁺ gonadotrophs. The expression of these genes is maintained through day 60. While this protocol closely resembles the in vivo developmental cues with good efficiency in generating hormone-producing cells across three hESC lines and one hiPSC line, the method does not model 3D aspects of pituitary development and shows bias toward a subset of hormone-producing cells.

As an alternative approach, another group developed a pituitary differentiation pipeline that simply relies on the addition of SHH and BMP ligands, two important regulators of pituitary development (Ozone et al., 2016). This method is more straightforward and recapitulates the in vivo physical environment by using a co-induction paradigm of ventral hypothalamus and non-neural ectoderm. Ventral hypothalamus (Rx⁺, Nkx2.1⁺) is induced in these culture conditions via SHH activation (smoothened agonist, SAG), while non-neural ectoderm (pan-cytokeratin⁺) is induced via BMP4. This co-culture system allows for the formation of superficial layers of non-neural ectoderm surrounding the ventral hypothalamus, resulting in the expression of the early pituitary markers Pitx1 and Lhx3. Terminal differentiation of these cells gives rise to ACTH⁺ corticotrophs, Pou1f1⁺GH⁺ somatotrophs, and LH⁺FSH⁺ gonadotrophs, which are representative of the pituitary hormone producing cells. Although this protocol allows for 3D aspects of pituitary development, with cues from neighboring hypothalamic cells, the overall efficiency is less than optimal, with a strong bias toward generation of corticotrophs over other hormone-producing cell types.

Olfactory placode

The olfactory placode invaginates to form the olfactory sensory epithelium, which is composed of olfactory neurons and GnRH expressing neurons. In vivo, the olfactory placode requires Wnt inhibition and expression of FGF and BMP ligands for induction and specification. Currently, the in vitro protocols deviate slightly from the in vivo situation; for example, inhibition of BMP (dorsomorphin) and TGFβ (SB431542) signaling, followed by exposure to an FGF ligand elicit GnRH neuron formation (Lund et al., 2016). The intermediate placodal cells express Sox1, Pax6, Emx2, and FoxG1, with low expression levels of Six1 and Eya1. These progenitor cells continue to differentiate into GnRH neurons, expressing Gnrh1 and Tuj1, and maintain FoxG1 expression. This method of GnRH neuron generation has relatively high efficiency for neuronal cells (15% GnRH⁺ cells), but functional assays are lacking. While this protocol generates GnRH neurons from neural progenitor cells, as shown by the expression of FoxG1, Dlx2, and Dlx5 (Duggan et al., 2008; Givens et al., 2005; Iyer et al., 2010), there is very low activation of Six1 or Eya1 during the differentiation process.

Another study follows a stepwise inhibition of BMP (LDN-193189) and Wnt (IWP2) signaling, followed by RA inhibition (pan-RAR antagonist, AGN193109) and expression of TGFα and FGF ligands to generate olfactory neurons (Bricker et al., 2022). This protocol leads to robust expression of olfactory placode genes Pou2F1, Dlx5, and Emx2, as well as neural progenitor genes Ascl1 and Nestin. Terminal differentiation of these cells gave rise to both Omp⁺ olfactory neurons and Gnrh1⁺ neurons. Although this protocol generates PPR-like cells, the PPR-induction is much lower than neural ectoderm induction in these conditions (6-10-fold vs. 3000-fold). Early time points also show expression of optic cup genes along with PPR and neural ectoderm genes, indicative of the shared origin of the lens and olfactory placodes, not something that is seen at later time points as the protocol is specific for olfactory placode cells.

Lens placode

Multiple studies have been focused on the generation of lens cells in vitro. A key factor in lens placode development and its segregation from the olfactory placode is the timing of exposure to BMP and FGF signals (Sjödal et al., 2007). The lens placode requires a prolonged exposure to BMP in combination with FGF in order to differentiate in vivo. However, not all in vitro protocols have followed this pattern. Some studies have developed lens placodal cells without BMP and/or FGF ligands (Yang et al., 2010; Dewi et al., 2020; Megarelli and Barberi, 2013), while others have generated lens progenitor cells using base medium with the BMP antagonist Noggin and TGFβ inhibition (SB431542) (Dewi et al., 2020). These progenitor cells then become lens cells with the addition of BMP4/7 and FGF2 ligands. While this protocol has been tested across hESCs and hiPSCs, each cell line requires troubleshooting to adjust enzymatic digestion, and account for non-lens cell contamination. Lens placodal cells have also been obtained via spontaneous differentiation of hESCs (Mengarelli and Barberi, 2013). The spontaneously differentiating cells gave rise to cells expressing early placodal genes, such as Six1/4, Eya1 and Dlx3/5. After cell sorting, a subset of these cells eventually activates the lens epithelium-specific genes FoxE3 and Prox1, as well as crystallin genes CRYAA and CRYAB. This is a streamlined protocol, varying greatly in efficiency prior to cell sorting, but allowing for careful characterization of the lens cell population, and a negative selection panel permits the generation of pure populations.

Following more closely the in vivo situation, one group has produced developing lens cells by BMP inhibition (Noggin) followed by exposure to FGF and BMP ligands (Fu et al., 2017). This ‘fried egg’ protocol involves neuroectoderm induction via Noggin, followed by placode induction via FGF and BMP, and finally lens epithelial cell induction via exposure to Wnt3a. Placode cells emerge in a window of 8 days, characterized by expression of Six1, Eya1, Dlx3, and Pax6. These cells eventually express early lens markers (Sox1, Prox1, FoxE3) and lens fiber markers (CryaA, CryaB, Mip). The downside to this protocol is the selection step, which is labor intensive as it requires manual selection of colonies. Another group also showed similar results using Noggin, FGF, and BMP to induce lens placode via a neuroectodermal state (Yang et al., 2010). This protocol used Noggin to induce neuroectoderm (Pax6⁺ cells) that became prospective lens placode (Sox2⁺ and Six3⁺), and eventually lens cells characterized by Pax6 maintenance and CryaA and CryaB expression. This protocol produces fairly immature lens cells compared to the other protocols, as the nuclei are retained in the majority of the cells at the endpoint. This protocol also lacked robust testing across cell lines, as only one hESC line has been used.

Other studies have shown that BMP and FGF ligands coupled with TGFβ inhibition (SB431542) can give rise to lens placodal cells (Tchieu et al., 2017; Dincer et al., 2013). BMP has an early role in establishing placode competence. Development of a de-repression differentiation protocol – in which Noggin is withdrawn from the medium to remove the BMP inhibition, while TGFβ inhibition is maintained – allowed for robust expression of Dlx3, Six1, and Eya1, with a switch from 82% Pax6⁺ cell clusters to 71% Six1⁺Eya1⁺Dach1⁺FoxG1⁺ cell clusters. After this induction, inhibition of lens placode requires FGF inhibition (SU5402), leading to Pitx3 induction, followed by CryaB expression at later time points. While the generation of lens cells is robust in this protocol, the initial generation of Six1⁺ cells is relatively low, with only a 40-50% yield (Tchieu et al., 2017).

Trigeminal placode

In the embryo, Wnt, FGF, PDGF, and Notch signaling are all key components of trigeminal placode induction (Canning et al., 2008; McCabe and Bronner-Fraser, 2008; Lassiter et al., 2010). Currently there is only one published protocol for trigeminal placode cell generation in vitro, which involves partial BMP attenuation (LDN193189) and TGFβ inhibition (SB431542), followed by Wnt activation (CHIR), then complete removal of BMP activity, and finally removal of Wnt activation (Tchieu et al., 2017). This stepwise modulation of the cell environment allows for the formation of Six1⁺ cells that are able to further differentiate into sensory neurons (Tuj1⁺, Brn3A⁺), which eventually express peripherin, glutamate, TRK receptors, Trpv1, and Trpm8 - genes typically associated with differentiation of trigeminal neurons. This protocol is robust across multiple hESC and hiPSC lines, yet lacks the contribution of neural crest, which give rise to the supporting glia, therefore preventing any study on neuron-glia interactions during trigeminal placode development.

Otic placode

Multiple in vitro protocols are available for the development of inner ear cells, including the precursor otic placodal cells. The main signaling pathway required for otic placode induction in vivo is FGF signaling (Groves and Fekete, 2012), and all published protocols for otic placode derivation in vitro require FGF as well. For some protocols, this is the only factor required (Lahlou et al., 2018; Chen et al., 2012; Ohnishi et al., 2015). For example, one protocol uses FGF3 and FGF10 to induce Pax8⁺Pax2⁺ cells. They then isolate triple positive Pax8^hiSox2^hiFoxG1^hi cells that generate otic epithelial and otic neural progenitors, which eventually become Atoh1⁺Brn3C⁺Myo7A⁺ and Brn3A⁺β-Tubulin III⁺NF200⁺ cells, respectively (Chen et al., 2012). This protocol is efficient across four hESC lines, but is labor intensive, with a manual selection step during sensory neuron differentiation. Another protocol uses FGF3 and FGF10 to induce Sox2⁺Pax2⁺ otic placode cells, and then utilizes RA and epidermal growth factor (EGF) or Notch activation to terminally differentiation the cells into cochlear hair cells (Atoh1⁺Myo7A⁺) and support cells (S100A1⁺Jag1⁺Sox2⁺Gjb1⁺Lgr5⁺) (Lahlou et al., 2018). These hair cells, however, lack stereocilia and appear to be immature, with no proper functional characterization. This protocol also was not tested across multiple cell lines (only one hiPSC line was used), so robustness is unknown. An alternative protocol has relied on the spontaneous differentiation of pre-placodal ectoderm before addition of FGF2 to induce otic placode (Ohnishi et al., 2015). Here, the pre-placodal ectoderm cells are Six1⁺, otic placode cells are Pax2⁺, and the terminally differentiated cochlear hair cells are MyoVIIa⁺β-Tubulin III⁺. However, the efficiency of this protocol is extremely low (0.01% cells becoming hair cells) and robustness across cell lines was not assessed, with only one hiPSC line used.

Other available protocols require additional signals, such as BMP inhibition (Gunewardene et al., 2014) or BMP activation followed by BMP inhibition, coupled with TGFβ inhibition (Koehler et al., 2013; Koehler et al., 2017). BMP inhibition (Noggin) and FGF ligand in combination, followed by simultaneous exposure to EGF and FGF ligands generate otic placode cells, with a robust induction of Pax2 (Gunewardene et al., 2014). These cells further differentiate into sensory neurons (NeuroD1⁺Brn3A⁺Islet1⁺) and auditory neurons (Gata3⁺β-Tubulin III⁺NFM⁺VGlut1⁺). However, this differentiation protocol is not homogeneous, with other neural lineages forming throughout the process, and the protocol is highly variable across cell lines, specifically hESC vs. hiPSC lines.

Two publications from the same group focused on generating inner ear cells via non-neural ectoderm and the OEPD (Koehler et al., 2013; Koehler et al., 2017). The first approach used TGFβ inhibition (SB431542) and BMP4 ligand to induce non-neural ectoderm, which is characterized by Dlx3 expression and subsequent differentiation into Krt5⁺p63⁺ epithelium (Koehler et al., 2013). At day 4.5, the treatment is changed to BMP inhibition (LDN193189) and exposure to FGF2, inducing pre-placodal ectoderm, which then developed into Pax8⁺ OEPD cells. The second approach generates organoids by combining TGFβ inhibition and BMP4 ligand with low FGF2 to induce non-neural ectoderm (Tfap2a⁺Dlx3⁺Ecad⁺), followed by exposure to high FGF2 levels and BMP inhibition to induce the OEPD (Pax8⁺Sox2⁺Tfap2a⁺Ecad⁺Ncad⁺), and finally Wnt activation (CHIR) to induce terminal inner ear cell differentiation as evidenced by Pax2, Pax8, Sox2, Sox10, and Jag1 expression (Koehler et al., 2017). Despite the push toward terminal differentiation, the sensory cells are still electrophysiologically immature, and hair cells only arise in a subset (20%) of organoids produced (Koehler et al., 2017).

While some groups are focused on generating placode progenitors, others have put their efforts into developing terminally differentiated cell types. To generate otic progenitors in a stepwise manner, one protocol begins with TGFβ (SB431542) and Wnt (FH535) inhibition to give rise to a subset of Six1⁺Eya1⁺ cells amongst Tfap2A⁺ non-neural ectoderm cells (Ealy et al., 2016). These cells are then expanded using exposure to FGF2 and RA for two days followed by FGF2 and BMP inhibition to give rise to Pax2 and Pax8 expressing cells. Finally, the cells are cultured in the presence of FGF2, BMP4, and Wnt ligands to generate otic progenitors expressing Sox2, Jag1, and Atoh1 (Ealy et al., 2016). This protocol has limitations across cell lines, with hiPSCs showing co-expression of PAX genes, not something that is seen during in vivo development. Matsuoka and colleagues have developed a protocol to generate spiral ganglion neurons of the inner ear via placodal progenitors. The placodal progenitors are obtained via TGFβ inhibition and exposure to BMP4 and FGF2 ligands, followed by Wnt inhibition (Matsuoka et al., 2017). Early pre-placodal cells first express Dlx3/5, and then activate Six4 and Eya1. Later stage cells are specified as otic neural progenitors expressing FoxG1, Gata3, Sox2, Pax8, and Nestin, and terminally differentiated sensory neurons expressing β-Tubulin III, BRN3A, TrkB, Vglut1/2/3, and Map2 (Matsuoka et al., 2017). These neurons still remain electrophysiologically immature.

Limitations and the future of in vitro placode modeling

Here we have summarized some of the advances in recent years to develop protocols to generate in vitro a broad array of cells with placode characteristics including placodal precursor cells, placodal cells, and the terminally differentiated cells of the anterior pituitary, sensory tissues and ganglia (Fig. 2). While these efforts have tremendous potential and promises in regenerative medicine, still a number of challenges persist in the emerging field of in vitro placode modeling.

A number of available protocols for placode cell generation deviate from the conditions typically required for their in vivo formation and development. While this is presumably necessary in some instances to account for the in vitro environment, adhering more closely to the in vivo developmental cues and their timing could improve on these protocols. It is also important to further expand our basic understanding of cranial placode formation in vivo using a broad range of animal models to refine and improve these protocols. For example, the neuropeptides, somatostatin and nociceptin, have been shown to control anterior placode progenitors’ fate in zebrafish (Lleras-Forero et al., 2013), and a recent study has implicated natriuretic peptide signaling in the early stage of cranial placode formation in Xenopus (Devotta et al., 2023), therefore modulation of these signaling pathways should also be considered when generating placodal cells in vitro.

Several technical shortcomings plague the current protocols. The major one is the extended culture time of the placodal cells once they are formed. For all the published protocols, the placodal cells represent a temporary state during differentiation, as is the case with cranial placodes during in vivo development. The placodal cells appear within a window of about 7-10 days, before continuing differentiating toward their final fate as neurons or sensory cells. While this is helpful to recapitulate sensory organ development in vitro, this is not optimal to study placodal cells themselves. In order to analyze placodal cells on a large scale using transcriptomic or epigenomic assays it would be preferable to develop protocols in which placodal cells can be cultured for multiple passages without differentiating. Development of a protocol for prolonged placodal cell culture would be beneficial to the field in order to study this cell population in more detail.

Another technical shortcoming is the vast number of small molecules, signaling pathway activators and inhibitors, used in some of the protocols. While necessary to recapitulate the in vivo cues, it can be cumbersome to include in sequence upwards of four different factors into the differentiation medium. Eventually, it would be advantageous to generate some type of placode medium or medium kit that is less burdensome on the user, and that would improve the efficiency and reproducibility of the differentiation protocols, in a similar way to what has been developed for neural crest cell differentiation in vitro (STEMdiff^™ Neural Crest Differentiation Kit, StemCell Technologies). Other technical shortcomings include reliance in some protocol on media with not always clearly identified components, low efficiency of differentiation, and variability across cell lines.

Finally, it is important to point out that there are no current protocols to generate epibranchial placode cells. These cells begin their development as part of the posterior placodal region, in a shared domain with the otic placode (OEPD). However, they only segregate, upon Wnt signaling inhibition and active FGF and BMP signaling (Padanad and Riley, 2011; Sun et al., 2007). An in vitro approach to generating epibranchial placode cells would most likely begin with combined BMP inhibition and exposure to FGF2 ligand to form the pre-placodal cells, followed by BMP inhibition, FGF2, Wnt and RA signaling activation to induce the posterior placodal OEPD, and finally exposure to FGF2 with Wnt inhibition to generate epibranchial placodal cells. There are also very few protocols available for the generation of trigeminal and olfactory placode cells, with one or two protocols published, respectively. Additional work will be needed to further optimize these protocols, as we learn more about the diversity of these neuron populations. Interestingly, placode differentiation in knockout serum replacement (KSR) base medium conditions defaults to a trigeminal placode fate (Dincer et al., 2013) - a better understanding as to why this is the default placode state in these conditions would be of interest to the field.

The future of in vitro placode generation and modeling very much follows that of in vitro cell modeling in general. For example, the next steps in human stem cell studies include three-dimensional organoids and organs-on-a-chip. Placodal cells can be useful in both of these technologies to understand human organ development. Any 3D organoid of the sensory tissues of the human head will require placodal cells. Therefore, it is important to develop organoid protocols with placodal-like cells to fully recapitulate the early development of the sensory tissues.

While some organoid protocols for placode-derived structures exist (Ueda et al., 2023; van der Valk et al., 2023; Gabriel et al., 2023; Doda et al., 2023), including some discussed above (Koehler et al., 2017, Ozone et al, 2016), the field is still lacking protocols for most placode-derived structures in 3D. The most accessible sensory organoids are those of the inner ear, with several established protocols (Koehler et al., 2017; Nie and Hashino, 2020; Ueda et al., 2022; NIe et al., 2022; van der Valk et al., 2022; Ueda et al., 2023, Doda et al., 2023). Some of these protocols fail to assess the otic placode stage of development, with characterization beginning after otic progenitor specification. As these protocols begin with ESCs or iPSCs, it would be beneficial to consider the developmental trajectory of the cells at all stages, including the pre-placodal and placodal cells, to better recapitulate the in vivo development of these cells. While many of the protocols do include placodal stages, the still incomplete understanding of placode development in vivo results in significant gaps between the in vitro organoids and the in vivo organs they model.

Along with organoids, organs-on-a-chip represent the future of in vitro cell modeling. Placodal cells are a necessary component for any sensory tissue development on a chip. For example, it would be interesting to model the development of the four lineages of the embryonic ectoderm on a chip, neuroectoderm, neural crest, cranial placodes, and non-neural ectoderm, to gain further insights on how these lineages interact to generate the tissues of the vertebrate head.

Cell replacement therapies also mark a future direction for in vitro human cell development. Briefly, this involves taking cells from a patient, producing induced pluripotent stem cells via reprogramming with Yamanaka factors (Takahashi et al., 2007), differentiating these cells into the desired cell type, and replacing the patient’s own cells with the in vitro-derived cells. Some of the protocols discussed above have taken steps in this direction by testing the function of terminally differentiated cells in animal models (Zimmer et al., 2016; Ozone et al., 2016; Chen et al., 2012; Matsuoka et al., 2017). A number of the protocol-derived cells can be transplanted in vivo for functional assessment as well, paving the way for developing cell replacement therapies for patients with disorders of placode-derived structures. For structures where cell replacement therapies are not feasible, due to limitations such as invasiveness of cell replacement, or complexity of the structures, the patient-derived placodal cells could be used for disease modeling and/or compound screening, allowing for patient-specific treatments and therapies to be designed and implemented.

Conclusions

In this review we have summarized the current protocols for in vitro placodal cell generation and discussed the challenges and potential of these approaches in regenerative medicine. While many robust protocols exist, there are still areas of limitations that would benefit from improvements to advance the field. These protocols also pave the way for the future of in vitro cell generation, including 3D organoids, organs-on-a-chip, and their use in cell replacement therapies. Enhancing and expanding these protocols can help advance our understanding of cranial placode development and improve treatment of cranial placode-related pathologies.

Highlights.

• Cranial placodes originate from a common territory, the pre-placodal region (PPR)

• From the PPR emerge most sensory organs, the anterior pituitary and several cranial ganglia

• In vivo signaling cues has informed protocols to derive cranial placodes in vitro

• In vitro models of cranial placode development are powerful tools in regenerative medicine