Coupling of apical-basal polarity and planar cell polarity to interpret the Wnt signaling gradient in feather development

Directional sensing is essential for the construction of appropriate spatial forms. This is often achieved through the formation and interpretation of a signaling gradient in the morphogenetic field (Yang and Mlodzik, 2015; Aw and Devenport, 2017; Sagner and Briscoe, 2017; Lander, 2007). A classical example is the Drosophila wing, in which the directions of epithelial hairs and bristles are coordinated globally via a core mechanism termed planar cell polarity (PCP), in response to a supposed global signaling gradient cue (Yang and Mlodzik, 2015; Bayly and Axelrod, 2011; Zallen, 2007). A similar mechanism has been shown to control limb bud elongation in higher vertebrates (Gao et al., 2011; Gao and Yang, 2013).

It remains highly controversial as to how the morphogen gradient is established and interpreted (Akiyama and Gibson, 2015; Nahmad and Lander, 2011). Although passive diffusion could be a driving force, recent work suggested that the Wnt ligands do not diffuse; rather, they remain cell bound and distribute only with cell division (Farin et al., 2016; Boutros and Niehrs, 2016). Alternatively, the signaling molecules can be transported via long cytoprojections, the so-called cytoneme structures (Kornberg and Roy, 2014; Hamada et al., 2014; Sanders et al., 2013). Therefore, it is of fundamental importance to clarify how cells sense and interpret the signaling gradient to coordinate cell shape change and pattern formation.

Avian feathers are composed of terminally differentiated keratinocytes with complex structures (Lucas and Stettenheim, 1972; Chen et al., 2015; Li et al., 2017; Lin et al., 2013; Yu et al., 2004; Feo et al., 2016; Prum, 2005; Prum and Williamson, 2001). A feather can either be bilaterally symmetric, with a central axis (rachis) into which branches (barbs) insert, or radially symmetric, with only a short calamus to which barbs attach (Chen et al., 2015; Yu et al., 2004; Prum, 2005; Yue et al., 2006). Our previous work has demonstrated that the emergence of an anterior-posterior Wnt signaling gradient in the developing feather follicle breaks the symmetry and induces the tilting of barbs, and thus the formation of bilaterally symmetric feathers (Yue et al., 2006). It remains unclear how this gradient is interpreted by feather keratinocytes and whether the PCP pathway is involved.

In recent years, we have developed methods to overexpress or knock down gene expression in the feather follicle through lentiviral-mediated gene transfer in vivo with high efficiency (Chu et al., 2014; Chen et al., 2014; Xie et al., 2015; Cheng et al., 2018). In our effort to screen gene functions in the feather follicle, we found that the core PCP gene Prickle1 (herein referred to as Pk1) controls directional sensing in feather development. Further work revealed that Wnt/Fzd signaling is involved, which in turn is controlled by the apical-basal polarity of feather keratinocytes, namely Par3/aPKC. This work sheds new light on how the global signaling gradient is interpreted locally to build complex 3D structures.

The two basic feather forms differ in their symmetric levels (Fig. 1A,B). The radially symmetric feathers have no central rachis, and the branches (barbs) are perpendicularly attached to the calamus. In bilaterally symmetric feathers, there is a central rachis, toward which the barbs are helically tilted (Yue et al., 2006). Histologically, each barb is composed of two columns of marginal plate cells in the periphery, and two columns of barbule plate cells with distinct elongated shape, which are separated by axial plate cells. The anterior barbule plate cells are tilted before the posterior barbule plate cells, with no obvious tilting in radially symmetric feathers (Fig. 1C,D). This change in cell shape is a gradual process, with characteristic elongated barbule plate cells even in radial feathers (Fig. 1E, Fig. S1; see Fig. S2 for the quantification of barbule plate cell parameters).

We wondered how this change in cell shape is translated into tilting of the feather branches. The tilting of the anterior barbule plate cells leads to an anterior expansion of the barb, which creates a tangential disparity, d_A−d_P (Fig. 1D). The calculated tilting angle is tgσ=(d_A−d_P)/h, where h is the height of the barb. We then measured the actual tilting (helical) angle of barbs in feather development (Fig. 1F). It turns out that the calculated angle σ fits well with the helical angle θ (Fig. 1F). Alternatively, it is also possible that θ is achieved through the cumulative effect of barbule cell tilting. That is, σ does not necessarily always correspond to θ, but cumulatively contributes to θ. Nonetheless, the coordinated cell shape change underlies the tilting of feather branches.

In an effort to systematically dissect gene functions in feather development, we performed RNA interference (RNAi) screening using our established method of lentiviral-mediated gene knockdown in vivo (Chu et al., 2014; Chen et al., 2014; Xie et al., 2015; Cheng et al., 2018). We found that the core PCP gene Pk1 controls feather cell shape change and bilateral feather formation (Fig. 2). Pk1 is expressed in a graded pattern, with higher expression in the anterior follicle, similar to the pattern of expression of the Wnts (Yue et al., 2006; Fig. 3). In each barb ridge, Pk1 mRNA does not show a clear enrichment pattern (Fig. 2A). However, when examined by a specific antibody (Fig. S3), Pk1 is initially homogenous and then enriched in the barbule plate cells facing the axial plate (Fig. 2B; quantified in Fig. 2D). This polarized distribution suggests that Pk1 might regulate cell shape change in feather development.

We then examined the functional role of Pk1 in feather development. The RNAi knockdown efficiency for Pk1 is ∼90% when examined in DF-1 chicken fibroblast cells and ∼80% in vivo (Fig. S4). When the RNAi virus was locally injected into the developing feather follicle, the barbule plate cells were unable to undergo the programmed shape change (Fig. 2C; quantified in Fig. 2D). Knockdown of Pk1 expression moderately reduced cell proliferation, as shown by proliferating cell nuclear antigen (PCNA) staining (Fig. 2E). The impact of Pk1 knockdown was documented at both the whole-mount level and final feather morphology level. Local injection of the RNAi virus led to perpendicular barbs which lose their coordinated tilting toward the rachis (Fig. 2F,G). At the final feather morphology level, the perturbed feathers show transition toward radial symmetry (Fig. 2H). Together, these data suggest that the core PCP gene Pk1 regulates cell shape change and barb tilting in feather development.

We wondered whether the polarized distribution and function of Pk1 is downstream of, and thus regulated by, Wnt/Fzd signaling. In our previous effort to map gene expression in the feather follicle (Chu et al., 2014; Cheng et al., 2018; GSE42017 and GSE110591), we found that the Wnt ligands expressed in the feather epithelium are mainly Wnt6 and Wnt5a. In situ hybridization revealed that both show an anterior-posterior graded expression (Fig. 3A; the higher magnification view shows that Wnt6 is enriched in barbule plate cells). This graded Wnt expression is similar to that of keratin A, which indicates terminal differentiation of the feather epithelium (Fig. 3A). Within each barb, the anterior barbule plate cells differentiate earlier than the posterior cells, as indicated by keratin A expression (Fig. S5).

The expression patterns of Wnt ligands and the receptor Fzd10 were examined by specific antibodies (Fig. 3B). Wnt6 was diffusive in early barbs, but was later enriched in barbule plate cells facing the axial plate (Fig. 3B). A similar expression pattern was found for Wnt5a (Fig. S6) and Fzd10 (Fig. 3B; quantified in Fig. 3D). The polarized distribution of Wnt ligands and the Fzd10 receptor suggests that they are upstream of and control Pk1 localization and function. Indeed, when we perturbed the Wnt gradient by local injection of Wnt6 protein, barbule plate cells failed to change their shape and Pk1 asymmetric distribution was lost (Fig. 3C; quantified in Fig. 3D). Consistent with previous work (Yue et al., 2006), Wnt6 overexpression led to reduced barb growth, whereas a constitutively active β-catenin promoted barb growth (Fig. 3E). Therefore, it is crucial to coordinate the canonical and noncanonical pathways downstream of Wnt signaling. As expected, perturbation of the Wnt6 gradient leads to re-orientation of the feather branch (Fig. 3F). Similar results were also obtained for Wnt5a (Fig. S6; the specificity of the antibodies is verified in Fig. S7). Thus, the barb tilting is controlled by an anterior-posterior gradient of Wnt signaling as reported previously (Yue et al., 2006), and this regulation is via the function of Pk1 to coordinate cell shape change.

We asked how the polarized distribution of Wnt6 and Fzd10 is achieved. Morphologically, it appears that the nuclei of barbule plate cells were localized at one side of the cells (Fig. S8), similar to the situation in polarized simple epithelium. We therefore examined the apical-basal polarity of these cells. It turns out that the molecular determinants of the apical compartment, Par3 and aPKC, are all localized facing the axial plate, suggesting that the barbule plate cells show apical-basal polarity in this tangential axis (Fig. 4A; quantified in Fig. 4C). To examine the functional significance of this polarity, we designed short hairpin RNA (shRNA) that effectively knocked down the expression of Par3 and aPKC. This was confirmed both in vitro in DF-1 cells and in vivo in the developing feather follicle (Fig. S4). Knockdown of either Par3 or aPKC led to failed cell shape change (Fig. 4B; quantified in Fig. 4C), disoriented barbs and perturbed feather formation (Fig. S9). Furthermore, the polarized distributions of Fzd10 and Pk1 were also disrupted (Fig. 4D), and cell proliferation was inhibited (Fig. 4E). Finally, in radially symmetric feathers, the polarized localization of Par3/aPKC is readily established, whereas Fzd10 (Fig. 4F) and Pk1 (not shown) do not show polarization. These results suggest that the apical-basal polarity in barbule plate cells is upstream and controls the subsequent polarization of PCP pathway components.

The complex feather structure exemplifies how spatial forms are constructed via coordinated cell shape change. In feather development, there is a programmed change of cell shape: in early feather branches, the cells are almost round (length/width ratio ∼1), then gradually the barbule plate cells become elongated, and tilted in bilateral feathers. Our work illustrates how the Wnt signaling gradient is interpreted to coordinate this cell shape change and orient the feather branch. The PCP pathway is involved, which in turn is controlled by Wnt/Fzd signaling. However, the polarized distribution of these core members is not according to the global Wnt gradient. Rather, their localization is determined locally within each barb, which is under the control of apical-basal polarity, namely the Par3/aPKC machinery (summarized in a hypothetical model in Fig. 4G). In radially symmetric feathers, the apical-basal polarity is readily established; however, there is no Wnt signaling gradient and thus no PCP activity to polarize the barbule plate cells. This is apparently caused by the low levels of Wnt6/Wnt5a and Pk1, which showed graded expression in bilateral feathers and are only weakly expressed in radial feathers (Fig. S5).

The division of a continuous Wnt gradient into discrete signaling centers is due to the process of feather branching, which is mainly related to Notch and FGF signaling (Cheng et al., 2018). After branching, the marginal plate cells now directly face the pulp mesenchyme and become the basal epithelial cells. Thus, the two columns of barbule plate cells acquire distinct apical-basal polarity, in opposite directions. The shape change of barbule plate cells is subsequent to this branching process. The fact that in radially symmetric feathers, the barbule plate cells do not tilt, yet still have an elongated shape, suggests additional mechanisms independent of PCP to modulate the process. The PCP pathway kicks in relatively late to tilt the barbule cells and orient the feather branch. Nonetheless, the PCP pathway and Pk1, in particular, can also modulate the shape of barbule plate cells, possibly because they can regulate the shared key cytoskeleton molecules. Indeed, Pk1 knockout has been shown to disrupt the apical-basal polarity of mouse epiblast cells (Tao et al., 2009). To dissect the molecular mechanism of cell shape change in feather development, which is specifically in barbule plate cells but not marginal plate nor axial plate cells, precise gene manipulation in these different cell populations is required. Future work to characterize specific promoters for these different cell populations will be essential to design such experiments.

There are subtle differences among manipulating the different aspects of Wnt signaling in feather development, such as Pk1 knockdown/Wnt6 overexpression/β-catenin overexpression. It appears that β-catenin promotes barbule cell growth (Fig. 3E). On the other hand, Wnt6 overexpression inhibited barb growth, which is consistent with our previous finding that Wnt3a overexpression does not elongate the barbule but disrupts barb patterning (Yue et al., 2006). The downstream events of Wnt signaling are complex and multifaceted. It is possible that Wnt protein overexpression triggers cell differentiation through the Wnt/calcium pathway, although the details will need further investigation. On the other hand, the patterning of barb ridges was disrupted in all samples (Par3 RNAi, PRKCI RNAi, Wnt6 overexpression and Pk1 RNAi; Fig. S10). One reason could be that cell shape change requires the function of these molecules (thus is not distinct after their perturbation). The reduced cell number (as is the case of Par3 RNAi, PRKCI RNAi, Wnt6 overexpression and, to a lesser extent, Pk1 RNAi) might inhibit the branching process, in addition to patterning the barb ridge. However, it remains unknown how the barb ridge is patterned, i.e. specification of the marginal plate cells, barbule plate cells and axial plate cells. This also requires further investigation.

It is likely that the preferential localization of Wnt/Fzd proteins in the apical side of the barbule plate cells is controlled by the directional transportation/retention mechanisms associated with the apical-basal machinery (Román-Fernández and Bryant, 2016; Galic and Matis, 2015; Lee and Streuli, 2014). Although the details remain unclear at this moment, our preliminary results showed that by destabilizing the microtubule network via colchicine treatment, the polarized distribution of Par3/Wnt6/Fzd10 were all disrupted (data not shown). In summary, our results reveal a new strategy through which the global morphogen gradient is interpreted locally, thus coordinating the construction of complex spatial structures.

Adult male chicken (Gallus gallus domesticus) aged 3-6 months were purchased from local farms and housed in the animal facility of Fuzhou University with free access to food and water. All experimental protocols were approved by Fuzhou University Experimental Animal Ethics Board, and comply with all relevant institutional and national animal welfare laws, guidelines and policies. For feather plucking and virus injection into the follicle cavity, anesthesia was not necessary; for virus injection into the growing follicle and collection of regenerating follicles, chickens were anesthetized by intraperitoneal injection of pentobarbital at 50 mg/kg body weight.

Feather samples were fixed in PBS containing 4% paraformaldehyde. Samples were then processed for paraffin embedding and sectioned at 6 µm. A standard protocol for Hematoxylin-Eosin staining was used. For immunofluorescence staining, antigen retrieval was performed using 10 mM citrate buffer, pH 6.0. After staining, the slides were counterstained with 1 µg/ml 4′,6-diamidino-2-phenylindole (DAPI) and mounted, and then photographed using a Leica fluorescence microscope. Antibodies used were as follows: chicken anti-Prickle1 (homemade, see supplementary Materials and Methods and Fig. S2; 1:100 dilution), anti-β-catenin (C2206, Sigma-Aldrich; 1:200 dilution), anti-PCNA (sc-7907, Santa Cruz Biotechnology; 1:200 dilution); anti-Wnt6 (24201-1-AP, Proteintech; 1:200 dilution), anti-Wnt5a (55184-1-AP, Proteintech; 1:200 dilution), anti-Fzd10 (18175-1-AP; Proteintech; 1:200 dilution), anti-Par3 (11085-1-AP, Proteintech; 1:200 dilution) and anti-aPKC (13883-1-AP, Proteintech; 1:200 dilution). The method for in situ hybridization has been described previously (Chu et al., 2014). Probes used were as follows: Prickle1 [nucleotides (nt) 1532-2187; XM_416036.5], Wnt5a (nt 382-1208; NM_204887.1), Wnt6 (nt 191-507; NM_001007594.2) and keratin A (nt 158-708; NM_001101732.2).

A standard protocol for lentiviral production using 293T cells was followed and has been described previously (Chu et al., 2014). We used the pLL3.7 vector to construct short hairpin RNA (shRNA) for gene knockdown. Target sequences for genes were as follows: Prickle1, 5′-ATCCAAGAGCTGGACATG-3′; Par3, 5′-ACAGGAGACGTACTTACA-3′; aPKC, 5′-AAGTGTCACAAACTGGTC-3′. A scramble control was constructed with the target sequence 5′-AGATACGACAGAGGACACT-3′. To monitor the RNAi knockdown efficiency, constructs were transfected into the chicken fibroblast cell line DF-1 (GNO30, Cell Library of the Chinese Academy of Sciences) using Lipofectamine 2000, and cells were harvested 48 h post-transfection. Total RNAs were extracted and processed for RT-PCR and quantitative PCR analysis (Roche LightCycler 480). To monitor the knockdown efficiency in vivo, the follicles were plucked, infected with the virus and collected 4 days later. Total RNAs were extracted for gene expression analysis. In addition, the developing feather follicles were locally injected with the virus, and samples were collected 2 days later and stained for protein expression. Virus infection was confirmed by the expression of GFP on the viral vector. Primers used were as follows: β-actin (forward 5′-CTGACGGACTACCTCATGAA-3′, reverse 5′-CCTCTCATTGCCAATGGTGA-3′); Prickle1 (forward 5′-AGAAGATCTAAATCCCAGTCT-3′, reverse 5′-GTGATCCTGAGGTGAGTAAT-3′); Par3 (forward 5′-CTCCTAACAATCATGACCGG-3′, reverse 5′-CTTGTCGTTGCCGTTGCATT-3′); aPKC (forward 5′-ATCCAAAGGAGCGGTTAGGC-3′, reverse 5′-GCTGCACAGGTTCATTGGTG-3′).

For virus infection of the regenerating feather follicle, feathers were plucked and 100-200 µl virus solution was injected into the follicle cavity. Feathers were collected 1-2 months later to examine the morphological changes. For virus injection into the growing feather follicle, a small hole was punched in the follicle wall near the base of the follicle, and 2-3 µl virus solution was injected using a fine glass needle. Samples were collected 48 h postinjection to visualize the disrupted branches. Similar procedures were adopted for protein injection into the feather follicle. Plasmids containing Wnt5a or Wnt6 under the control of a CMP promoter in pEGFP vector were overexpressed in 293T cells. The cells were harvested by trypsin/EDTA digestion, and sonicated before injection. Nontransfected cells were used as a control. For β-catenin overexpression, a constitutively active form of β-catenin was cloned into the RCAS vector and infected the feather follicle (Widelitz et al., 2000). Samples were collected 2 weeks later for histological analysis.

Feather morphology was photographed using a Leica stereo 3D microdissection microscope. To visualize the developing feather branches, feathers were plucked and cut open under a dissection microscope (Yue and Xu, 2017). After removing the pulp mesenchyme, the epithelium was fixed by 4% paraformaldehyde at 4°C overnight, counterstained by 1 μg/ml DAPI in PBS for 1 h at room temperature, briefly washed in 3× PBS, and mounted for photography under an inverted Nikon fluorescence microscope.

Wing contour feathers and leg feathers in adult chicken were used as bilateral and (nearly) radial feathers. Feathers were plucked to induce regeneration for ∼2 weeks and collected when in active growing phase. The early, middle and later stages of barb maturation were defined as follows: starting from the initial branching, every 30 sections (6 µm thick) were collected, which correspond to ∼200 µm in distance. The tilting angles of barbule plate cells were measured (using Adobe Illustrator CS6) for both the anterior (αA) and posterior (αP) column of cells, with reference to the horizontal line as the tangential line of the follicle circle. The length and width of barbule plate cells were also measured (in Adobe Illustrator CS6) from at least three barbs and at least 20 cells. The asymmetry index (ASI) is defined as (A-P)/(A+P), where A and P define the anterior and posterior signal, respectively. The anterior and posterior signals were measured separately by using the ImageJ program after bisecting the cells into the anterior and posterior halves (Rodriguez et al., 2017; Fig. S2).

Each experiment was repeated at least three times. For gene transfer in the feather follicle, at least 15 follicles were each individually manipulated. Data are shown as mean±s.e. Statistical differences between two groups were determined by two-tailed Student's t-test.

We thank Drs Cheng-Ming Chuong and Randall Widelitz (University of Southern California, Los Angeles, USA) for their helpful input, and Zhang Juan for technical assistance.