Evidence for multiple roles for grainyhead-like 2 in the establishment and maintenance of human mucociliary airway epithelium

GRHL Protein Expression in Human Airways and HBE Cells in Culture.

In the normal human lung, immunohistochemistry (IHC) of sections of small and large airways shows GRHL2 in the nuclei of both luminal and p63⁺ KRT5⁺ basal cells. By contrast, GRHL1 is expressed predominantly in MUC5AC⁺ goblet cells and GRHL3 in acetylated tubulin⁺ ciliated cells (Fig. 1 A–E).

The in vitro culture of primary HBE cells presents a convenient method for testing the function of GRHL proteins in airway epithelium. When cultured in plastic dishes submerged in growth medium, all cells remain undifferentiated and express the basal markers p63 and KRT5. Even when confluent, they have very few intercellular junctions containing TJP1 (ZO-1). The cells robustly express nuclear GRHL2, but have low levels of GRHL1 and no detectable GRHL3. By contrast, when transferred to air–liquid interface (ALI) culture in differentiation medium, the cells differentiate into a mucociliary epithelium in about 4 wk. In the initial 3–7 d after seeding, most cells are still undifferentiated and express p63, KRT5, and GRHL2, with lower levels of GRHL1 (Fig. 1F, Top). When confluent, the cells now express obvious intercellular junctions with TJP1 and Claudin 4 (CLDN4) (see Fig. 3B). By ALI day (D) 10, p63⁻ nuclei are seen above the p63⁺ basal layer and GRHL1 is present at higher levels in these suprabasal nuclei (Fig. 1F, Middle). Only by ALI D35, when differentiated ciliated and secretory cells are present, is GRHL3 detectable in ciliated cells (Fig. 1F, Bottom).

GRHL2 Regulates the Expression of Multiple Genes in Undifferentiated HBE Cells.

We used RNA-seq to identify genes regulated by GRHL2 in undifferentiated HBE cells from three human donors (Fig. 2A). Specifically, we identified transcripts differentially expressed 24 h after inhibiting GRHL function with an inducible lentivirus driving a dominant-negative (DN) mutant form of the protein (Fig. S1). This DN-GRHL2 protein, in which the N-terminal transactivating domain is replaced with EGFP, is identical to a mouse DN-GRHL2 mutant protein previously shown to displace GRHL2 from a known target site in the Cdh1 gene and to block its transcription (9). Control virus expressed nuclear H2B-GFP.

There were 980 unique genes differentially expressed in the DN-GRHL2 vs. control cells with an adjusted P value <0.05 and an average change in expression greater than 50% (Dataset S1). A similar proportion of genes was up- (n = 523) and down-regulated (n = 457) (Fig. 2B). Annotating the differentially expressed genes with the Gene Ontology suggests that GRHL2 regulates diverse functions in undifferentiated HBE cells, including those associated with morphogenesis, adhesion, and migration. In addition, GRHL2-regulated proteins were significantly more likely to be localized to the cell periphery, especially the apical and apicolateral membranes (P = 7.5 × 10⁻³ and P = 6.9 × 10⁻³, respectively). Full Gene Ontology enrichments are shown in Dataset S2.

Some genes differentially expressed in response to DN-GRHL2 may lie downstream of direct targets, including microRNAs (19). Therefore, to identify direct GRHL2 targets, we performed ChIP-seq on HBE cells grown under the same conditions as for the RNA-seq studies. There were 7,687 GRHL2 binding sites reproducibly identified among all the three donors (Dataset S3). The sites were enriched for the known GRHL2-binding motif (Fig. 2C) and included the few genes that had been shown from previous ChIP studies to have GRHL2 binding sites [e.g., CLDN4 and CDH1 (9, 16)]. Representative sites for MPZL2, KRT5, SEMA4B, and HMBB1 are shown in Fig. 2E. Functional annotation analysis of all genes near GRHL2 binding sites using GREAT (20) reveals a specific enrichment for binding near genes involved both in epithelial cell–cell adhesion [false discovery rate (FDR), 5.2 × 10⁻⁸] and in cell–matrix adhesion [false discovery rate, 9.7 × 10⁻⁴].

Genes that were differentially expressed with DN-GRHL2 were enriched for nearby GRHL2 binding, particularly within 40 kb of the transcription start site (Fig. 2D and Fig. S2A), suggesting that those genes are direct targets of GRHL2. Based on this analysis, we considered genes that are both regulated by DN-GRHL2 and bound by GRHL2 within 40 kb of the transcription start site to be direct targets (n = 296). These genes were significantly more likely to have reduced expression after DN-GRHL2 (206 vs. 90; P = 7 × 10⁻²⁹ using a two-sided Fisher’s exact test). More generally, the proportion of genes with reduced expression after DN-GRHL2 directly and significantly correlated with the proximity of GRHL2 binding to the transcription start site (Spearman’s ρ = 0.09, P < 1 × 10⁻⁵⁷; Fig. S2 B and C).

Clearly, there are many more GRHL2 binding sites in the genome than genes modulated over 24 h in undifferentiated HBE cells in response to DN-GRHL2. Therefore, to better understand the specific function of candidate direct targets, we repeated the functional enrichment analysis using only differentially expressed genes with GRHL2 binding within 40 kb of the transcription start site (Dataset S4). These genes were specifically enriched in multiple functional categories. Examples include those related to multicellular organismal development (morphogenesis) (SEMA4A/B, CTNNBIP1, GJB5, TGM1, MPZL2, COL7A1, OVOL2, and GPR56 were among those down-regulated by DN-GRHL2), hemidesmosome assembly (LAMA3 and KRT5), and cell “motility” (VEGFA, EPHA1, HMGB1, SUN2, NRP2, and PDGFA were down-regulated and KDR and PRKD2 were up-regulated). The last category includes genes that are likely involved in epithelial-to-mesenchymal signaling (KDR, VEGFA, HMGB1, and PDGFA).

GRHL2 Is Required to Establish Barrier Function and Differentiation of HBE Cells in ALI Culture.

After transfer to ALI culture, undifferentiated HBE cells give rise to a well-differentiated, polarized mucociliary epithelium with transepithelial resistance (TER). To evaluate the role of GRHL2 in the formation of this epithelium, we used the doxycycline (Dox)-inducible DN-GRHL2 lentivirus system to inhibit protein function from the time the cells first reach confluence (around ALI D7) (Fig. 3A). At this time, the cultures have distinct apical junctional complexes and localized CDH1, CLDN4, and TJP1 (Fig. 3B). In the experiment shown in Fig. 3C, TER began to increase in control cultures at D11 and reached a plateau around D18. However, following induction of DN-GRHL2, the cultures completely failed to establish TER (Fig. 3C). By ALI D14, although the number and density of cells on the membranes were essentially the same, cells expressing DN-GRHL2 (visualized by GFP staining) expressed much lower levels of CDH1, CLDN4, and TJP1 than controls (Fig. 3D). Significantly, CDH1 and TJP1 proteins had a more punctate distribution. In the case of TJP1, the protein was often localized in the nucleus, suggesting that inhibition of GRHL function interferes with the shuttling of TJP1 between the nucleus and cell membrane (21).

To confirm the specificity of the effect of DN-GRHL2 on HBE cells in ALI culture, we used the same lentiviral protocol to induce expression of an shRNA against GRHL2, with red fluorescent protein (RFP) as a reporter. Control Western blot and quantitative (q)PCR experiments at ALI D28 established that this shRNA reduced protein and RNA levels by 64% and 75%, respectively, compared with scrambled shRNA (Fig. 4A). Even though inhibition of GRHL2 protein levels was not complete, Fig. 4B shows that GRHL2 shRNA inhibited the development of TER compared with control RNA, with a statistically significant difference in levels at ALI D21/22. Moreover, RFP⁺ cells within the cultures that express GRHL2 shRNA do not differentiate into ciliated or secretory cells at ALI D28 (Fig. 4B). Taken together, these results are further evidence that GRHL2 is required for the de novo establishment of barrier function and the differentiation of primary HBE cells.

Returning to the experiments using DN-GRHL2, we used Affymetrix microarray analysis to identify transcriptional changes induced by inhibiting GRHL2 function between ALI D7 and D14 (Dataset S5). We found 607 unique genes that had a significant change in expression between cells expressing DN-GRHL2 versus EGFP (P < 0.05; absolute fold change in expression >1.5). Of those, 258 (43%) had lower expression, whereas 349 (57%) had increased expression. The changes were highly concordant with the results of DN-GRHL2 expression in undifferentiated HBE cells, with 72% of genes regulated in the same direction. Of the 61 genes with significant differential expression in both experiments, 92% were regulated in the same direction. Genes with lower expression after DN-GRHL2 during ALI D7–14 were strongly enriched in the GRHL2 binding sites that we documented in undifferentiated HBE cells, suggesting that a subset of the same GRHL2 binding sites is functional in the ALI D7–14 cells. Of the 258 down-regulated genes, 125 (48%) had GRHL2 binding within 40 kb of the transcription start site. In comparison, only 77 (22%) of the up-regulated genes had GRHL2 binding sites. As in the undifferentiated HBE cells, the relative enrichment of genes with reduced expression after DN-GRHL2 was stronger for GRHL2 binding sites located closer to the transcription start site. This further suggests that GRHL2 binding primarily increases expression of nearby genes.

We used qPCR to confirm the down-regulation of some of the most highly differentially expressed genes, as well as other genes selected for their known involvement in epithelial functions and for having occupied GRHL2 sites in HBE cells (Fig. 3D). These included known targets of GRHL2 in other tissues (CLDN4 and ARHGEF19), even though the relative decreases in these genes in the microarray (∼1.4-fold) were just below the cutoff for inclusion in Dataset S5. The confirmed genes encode proteins that fall into a number of different classes, including membrane-associated proteins such as PROM2, RHBDL2 [a membrane-associated protease up-regulated in epidermal cells undergoing wound repair (22)], SMAGP [small trans-membrane and glycosylated protein (23)], STYK1 [a novel membrane-associated tyrosine kinase (24)], and SCRIB [implicated in planar cell polarity and epidermal wound repair (25)]. Another class encompasses small GTPases and other adaptor proteins, including RAB25 [associated with the recycling of membrane proteins and in GRHL2 function (13)], RAP1A (a Ras-related protein active in cell locomotion), and VAV1 [a Rho-GEF implicated in barrier function in colon epithelium (26)]. A third class includes proteins connected with cell adhesion, including CGN, CGNL1, CTNND1, MPZL2 [also known as epithelial V-like antigen (EVA1) and implicated in barrier formation between Sertoli cells (27) and as a target of Grhl2 in mouse embryonic neurectoderm (17)], and PVRL4 [poliovirus receptor-related 4, also known as nectin4, required for adherens junction formation in skin (28)]. Finally, we confirmed down-regulation of genes encoding ESRP1, an epithelial spicing regulatory protein (29), and various transcription factors and cofactors, including VGLL1 (30), WDR5B [a protein closely related to WDR5, a core component of the Trithorax histone methyltransferase complex that interacts with GRHL3 in keratinocyte differentiation (10)], ZNF750 [associated with the differentiation and development of barrier function of keratinocytes (31)], and ZNHIT6 [zinc finger protein implicated in RNA processing and emphysema (32)].

As shown in Fig. 3D, qPCR also confirmed the differential expression of transcripts for genes most highly up-regulated in the array. All except IFIT1, IFI6, and IFI44L (which encode IFN-inducible proteins) have occupied GRHL2 binding sites in basal cells. They were CA2, KRT6B, NPR3 (natriuretic peptide receptor 3), SERPINE2 (a peptidase inhibitor), SMOC1 (SPARC-related modular calcium binding 1), and STMN3 (also known as SCLIP, encoding a member of a family of proteins that destabilizes microtubules and promotes the motility of non-small-cell lung cancer cells (33).

Finally, the qPCR analysis confirmed that there was no change in the level of transcripts for the junction proteins CDH1 and TJP1 (Fig. 3D), even though these genes have distinct GRHL2 binding by ChIP-seq analysis of basal cells and CDH1 has been identified as a direct target in other cell types.

GRHL2 Is Required to Maintain Barrier Function in ALI Cultures.

We next asked whether GRHL2 is required to maintain the integrity of the epithelium once barrier function has been established. To achieve this, Dox was added to cells transfected with either DN-GRHL2 or H2B-GFP virus at a time when TER values had reached a maximum (Fig. 5A). In the experiment shown in Fig. 5B, TER starts to decline 48 h after adding Dox and 7–10 d later reaches a new lower baseline. At this time, control cultures have a cobblestone epithelial appearance (Fig. 5C). By contrast, cells expressing DN-GRHL2, although still confluent, appear more extended and less closely adherent to each other. This difference in morphology is reflected in histological sections of the coherent epithelial layer (Fig. 5D). There was no evidence for significant cell death (by staining for cleaved caspase 3) or for a difference in proliferation (Ki67 staining, Fig. 5E) between control and experimental cultures.

To identify GRHL-regulated genes, we treated ALI cultures transfected with either H2B-GFP or DN-GRHL2 virus with Dox when the TER reached a threshold level (between ALI D14 and D18, depending on the donor). RNA was isolated from cultures harvested just 48 h later to focus on early events in the uncoupling of junctional interactions, including those regulating TER. We found 94 unique genes expressed at lower levels and 209 expressed at higher levels in cells expressing DN-GRHL2 compared with control (P < 0.05; absolute fold change >1.5; Dataset S6). Again, genes expressed at lower levels were much more likely to have previously identified GRHL2 binding sites close to the transcription start site than genes expressed at higher levels (65% vs. 19% for GRHL2 binding within 40 kb). Genes with significant differential expression after DN-GRHL2 expression during ALI D7–14 and ALI D14–18 were highly concordant between the two experiments (Spearman’s ρ = 0.88). For no genes reaching our significance criteria did the expression level after DN-GRHL2 expression change in opposite directions. qPCR was used to confirm the down-regulation of the same set of genes down-regulated in cultures treated with Dox at ALI D7 (Fig. 5F).

Finally, we identified genes specifically regulated by DN-GRHL2 in undifferentiated HBE cells compared with cells that are both establishing and maintaining transepithelial resistance in ALI culture. The analysis confirmed that only a small fraction of genes was specifically regulated by DN-GRHL2 between progenitor cells and cells in ALI culture (Fig. S2D and Dataset S7).

Cell Culture.

Primary HBE cells were obtained from donors without pre-existing lung disease under University of North Carolina at Chapel Hill Biomedical IRB protocol #03-1396. ChIP-seq and RNA-seq experiments used the same three donors. Passage (P) 1 P1 cells were seeded into plastic dishes coated with bovine collagen (Advanced BioMatrix; PureCol, 5005-B) in BEGM Basal Epithelial Growth Medium (34) containing 0.11 mM Ca²⁺ and 25 ng/mL EGF (34). Confluent cultures were passaged to P2 in dishes without collagen. For ChIP-seq, P2 cells were collected at confluence. For RNA-seq, P2 cells at 40% confluence were infected with H2B-GFP or DN-GRHL2 lentivirus and 1 μg/mL puromycin was added 48 h later. At confluence, 0.5 μg/mL Dox was added for 24 h.

For ALI cultures, P0 HBE cells were seeded in collagen-coated 10-cm dishes. At 40% confluence, cultures were infected with lentivirus and grown for 48 h before puromycin selection. At 70–90% confluence, 1.25 × 10⁵ cells were seeded onto collagen IV (Sigma-Aldrich; C7521)-coated 12-mm Transwell permeable supports (Corning; 3460) in ALI culture medium (formula is given in ref. 34) with 1 mM Ca²⁺ and 0.5 ng/mL EGF. Medium was changed three times a week. Triplicate inserts were washed and 500 μL ALI medium was added before measuring TER every other day using a Millicell ERS2 (Millipore). The value was calculated in ohms per cm² and the resistance of the membrane inserts only was subtracted. The unit area resistance is obtained by multiplying the electric resistance by the effective surface area of the filter.

Lentivirus.

An shRNA targeting hGRHL2 (V2THS_176822) and a nonsilencing control shRNA (RHS4743) were from Thermo Scientific. To construct a Tet-on GFP-labeled dominant-negative mutant of GRHL2, DNA encoding a truncated version of mouse GRHL2 in which amino acids 1–232 were replaced with EGFP or H2B-GFP was cloned into pTRIPZ inducible lentiviral vector (Thermo Scientific) between AgeI and MluI. The vector was modified to replace RFP and insert bovine growth hormone polyA sequence after the truncated GRHL2 for its expression. Plasmids were transfected into 293T cells as described (35) using FuGENE 6 (Promega, E2691). Viral expression was confirmed by visualizing GFP after adding Dox.

Grhl3 Antibody.

Rabbit polyclonal Grhl3 antibody was produced by Harlan Bioproducts for Science against amino acids 75–230 of mouse Grhl3 (83% identical to human protein). GST-Grhl3 was used for primary immunization and His-Grhl3 for boosting. The antibody was affinity-purified using His-Grhl3 and specificity was confirmed by IHC in the presence of blocking peptide.

Immunohistochemistry.

Normal human lungs and ALI cultures were fixed in 4% (wt/vol) paraformaldehyde (PFA) at 4 °C for 3 h and overnight, respectively. Samples were embedded in paraffin, sectioned at 7 μm, and stained as described (36). For whole-mount staining of ALI cultures, membranes were washed with PBS and fixed with 4% PFA for 10 min at room temperature. Antibodies were rabbit GRHL1 (1:100; Sigma; HPA006420), rabbit GRHL2 (1:450; Sigma; HPA004820), rabbit Grhl3 (1:2,000), mouse monoclonal P63 (1:50; Santa Cruz; sc-8431), rabbit TJP1 (1:100; Invitrogen; 61-7300), mouse monoclonal MUC5A (1:200; Lab Vision; MS145-PO), rat CDH1 (1:500; Zymed; 13-1900), mouse acetylated tubulin (1:1,000; Sigma; T-7451), chicken GFP (1:500; Aves Labs; 1020), rabbit anti-Claudin4 (1:150; Invitrogen; 36-4800), mouse Ki67 (1:100; Vector Labs; VP-K452), and rabbit cleaved caspase 3 (1:600; Cell Signaling; 9664). Antigen retrieval on paraffin sections used 100 mM sodium citrate butter (pH 6) in a pressure cooker-like device (PickCell Laboratories; 2100 Retriever).

RNA-Seq.

We performed RNA-seq on untreated, control-infected, and DN-GRHL2–infected HBE cells as described (37). For details and accession numbers, see SI Materials and Methods. We estimated differential expression using DESeq (38) to compare a model that included donor and condition (untreated, control-infected, or DN-infected) with a model that only accounted for donor. To be called differentially expressed, genes were required to have an FDR less than 0.05 and a more than 50% change in expression between control and DN-GRHL–infected cells. Genes were analyzed for enriched categories from Gene Ontology as described (39).

ChIP-Seq.

ChIP-seq for GRHL2 was performed on HBE cells from three donors as described (40, 41) (see also SI Materials and Methods; Gene Expression Omnibus accession no. GSE46194). To control for sonication biases, we sequenced a matched input control library made from sonicated chromatin from each donor. We identified binding sites relative to the donor-specific input control using MACS (42). To identify reproducible GRHL2 binding sites for our final analyses, we calculated pairwise irreproducibility discovery rates (IDRs) (43, 44) and retained binding sites with an IDR of >0.05 in all three comparisons.

Integrative ChIP-Seq and RNA-Seq Analysis.

To identify genes most likely to be directly regulated by GRHL2, we identified the GRHL2 binding site nearest each RefSeq gene (42). To determine a characteristic binding distance for functional GRHL2 binding, we plotted the cumulative distribution of the distance to the nearest GRHL2 binding site for genes with increasingly significant differential expression as determined by RNA-seq. We observed a strong enrichment for GRHL2 binding within 40 kb of genes with a highly significant response to the GRHL knockdown (Fig. S2A). Based on this analysis, we considered differentially expressed genes within 40 kb of a GRHL2 binding site to be likely direct GRHL2 targets. To determine functional enrichment in likely direct GRHL2 target genes, we analyzed the genes for enriched categories from Gene Ontology (45) using the GOStats package (46).

Statistical Analysis of RNA-Seq and ChIP-Seq.

Statistical analysis for RNA-seq and the integrative analysis with the GRHL2 ChIP-seq were performed using the R statistical package (http://www.r-project.org); a script to reproduce that analysis is available upon request. All alignment, ChIP-seq peak calling, and ChIP-seq replicate analysis were performed using publicly available software (SI Materials and Methods).

Microarray Analysis.

RNA was extracted using an RNeasy Micro Kit (QIAGEN) and quality-checked with a 2100 Bioanalyzer (Agilent Technologies). RNA was processed using Ambion MessageAmp Premier by the Duke Microarray Facility. Standard Affymetrix protocols and human genome U133 Plus 2.0 chips were used to generate .cel files. Genomics Suite 6.5 (Partek) was used to perform data analysis. Robust multichip analysis normalization was performed on each dataset. Two-way ANOVA and fold-change analyses were performed to select target genes that were differentially expressed between control and DN-GRHL2 datasets. The top differentially expressed genes were selected with a P-value cutoff of <0.05 based on ANOVA test and a fold-change cutoff of >±1.5.

qRT-PCR.

RNA was extracted with the RNeasy Micro Kit (QIAGEN) and cDNA was synthesized with SuperScript III (Invitrogen). PCR was performed with SYBR Green chemistry in a StepOnePlus Real-Time PCR System (Applied Biosystems). Samples were analyzed in triplicate from two or three biological replicates and data were analyzed using the ΔΔCt method (47). Primer sequences are listed in Table S1.