HOXA5-driven transcriptional program in lung development

Abstract

Despite the fundamental role of Hoxa5 in mouse development revealed by the well-characterized phenotypes of Hoxa5 mutant mice, HOXA5-dependent regulatory networks remain ill-defined. We generated a Hoxa5^FLAG epitope-tagged mouse line to perform ChIP-seq experiments and uncover genome-wide occupancy of the HOXA5 protein. This was done in the developing lung tissue, in which Hoxa5 plays a predominant role since Hoxa5^−/− mouse mutants die at birth from respiratory defects. ChIP-seq allowed us to define an in vivo HOXA5 binding motif and its widespread genome distribution in the embryonic lung. Combined with ATAC-seq assays and epigenetic analyses, HOXA5 targets were identified. They include Hox genes known to show expression changes in lungs from Hoxa5 null mutant embryos. Moreover, several key actors of lung morphogenesis were found to possess HOXA5-binding sites and appeared as potential targets of HOXA5. Impact of the loss of Hoxa5 function on their expression was confirmed by in situ hybridization. These targets include members of the FGF10, SHH, BMP4 and WNT2 signaling pathways. Altogether, these data unveil the crucial role of HOXA5 in the coordinated control of the signaling networks instructing lung development.

1. Introduction

Proper control of cell-type specific transcriptional programs is critical for normal development and mutation of a single transcriptional regulator often produces pleiotropic phenotypes with organ-specific consequences. HOX proteins are evolutionary conserved transcription factors that play critical regulatory roles in embryogenesis, organogenesis and tissue homeostasis in postnatal life. Organized in tight clusters localized on different chromosomes, single mutations of the 39 Hox genes lead to multiple and various developmental defects and diseases (Afzal and Krumlauf, 2022; Hubert and Wellik, 2023; Krumlauf, 1994; Quinonez and Innis, 2014; Shenoy et al., 2022). Despite their broad expression in numerous cell types, HOX proteins perform their regulatory functions in a context-dependent fashion as revealed by the phenotypic outcomes in mouse mutants. HOX proteins bind to similar DNA-binding motifs through their highly conserved homeodomain, but they confer different regulatory actions in vivo, raising the fundamental question of how specificity of HOX transcriptional outputs is attained (De Kumar and Darland, 2021; McGinnis et al., 1984). While the developmental role of Hox genes is clearly established, understanding how they act still awaits the identification of target genes. Several tools including chromatin accessibility and interaction assays now offer the opportunity to study Hox gene action at the genome scale (Bridoux et al., 2020; Bulajić et al., 2020; Loker et al., 2021; Slattery et al., 2011).

The loss of Hoxa5 gene function in mice results in a plethora of well-characterized phenotypes and, most importantly, in death at birth from breathing failure due to respiratory system malformations (Jeannotte et al., 2016). Hoxa5 expression in the mesenchyme of the respiratory tract is essential for tracheal patterning, lung branching and cell differentiation as well as post-natal alveolar development (Aubin et al., 1997; Boucherat et al., 2012; Boucherat et al., 2013; Landry-Truchon et al., 2017b; Mandeville et al., 2006). In parallel, Hoxa5 expression in phrenic motor neurons is key for diaphragm innervation and musculature, which impacts lung growth (Landry-Truchon et al., 2017a; Philippidou et al., 2012). No other single Hox mutants present such an overt lung phenotype revealing the functional predominance of Hoxa5 in respiratory development (Boucherat et al., 2013). Hoxa5 thus provides an ideal model for studying how an individual HOX protein achieves cell-specificity to orchestrate organ system development.

In mice, lung organogenesis initiates around embryonic day (E) 9 with the outpocketing of the ventral anterior foregut endoderm into the surrounding splanchnic mesenchyme to form the laryngotracheal groove and two primary lung buds. Up to the formation of saccules at E16.5, a stereotypic and repetitive branching program of the epithelial buds in the mesenchyme gives rise to the airway tree (Alescio and Cassini, 1962). The distal region of lung outgrowths becomes signaling center where a network of pathways coordinates reciprocal interactions between mesenchyme and epithelium to govern proper lung development. Among the signaling molecules, FGF10 occupies a central role in lung branching. It is expressed in the mesenchyme flanking the distal tip of lung buds, and it exerts a chemoattractant effect on endoderm promoting its proliferation (Bellusci et al., 1997b; Weaver et al., 2000). FGF10 acts on epithelial cells through its receptor FGFR2 to direct expression of key regulator genes, such as Shh, a negative regulator of lung branching, and Bmp4 (Abler et al., 2009; Bellusci et al., 1997a; Lebeche et al., 1999; Pepicelli et al., 1998). Bmp4 is also expressed in lung mesenchyme (Weaver et al., 2003). However, the distinct functions of Bmp4 in lung epithelium and mesenchyme are still elusive (Bragg et al., 2001; Guo et al., 2014; Luo et al., 2024). Moreover, Fgf10 expression is tightly regulated by members of the T-box transcription factor gene family, Tbx4 and Tbx5. Loss of Tbx4 and Tbx5 gene functions leads to a complete ramification arrest resulting from Fgf10 decreased expression (Arora et al., 2012; Goering et al., 2003). The WNT/β-catenin canonical pathway is also crucial for lung morphogenesis, since it regulates both cell proliferation and differentiation (Aros et al., 2021). For instance, the WNT2 canonical ligand is expressed in the sub-mesothelial layer of lung mesenchyme where it activates cell proliferation (De Langhe et al., 2005; Yin et al., 2008).

Changes in expression of members of these different pathways were reported in Hoxa5 single mutants as well as in Hox5 triple compound mutants suggesting a regulation by HOXA5 (Hrycaj et al., 2015; Landry-Truchon et al., 2017b). To address the role of Hoxa5 in the signaling networks that instruct lung development, we performed chromatin immunoprecipitation sequencing (ChIP-seq) assays in lungs from E15.5 embryos using a newly developed Hoxa5^FLAG mouse line. Combined with Transposase-Accessible Chromatin sequencing (ATAC-seq) experiments and epigenetic analyses, we identified HOXA5 direct target genes that were further validated by in situ hybridization assays for differential expression in Hoxa5 mutants versus controls. Altogether, these data establish HOXA5 as a key driver of gene networks in lung morphogenesis.

2. Materials & methods

2.1. Generation of Hoxa5^FLAG epitope-tagged mouse line, genotyping and tissue collection

A targeting construct was designed in which a 3xFLAG sequence (5′-ttccgccccATCGACTACAAAGACCATGACGGTGATTATAAAGATCATGACATCGATTACAAGGATGACGATGACAAGTGACTGACTAGGAATTGgagcggcca-3′) was inserted in-frame into the C-terminus of HOXA5. The construct was generated following a strategy like the one used to produce the Hoxa5 conditional allele (Tabariès et al., 2007). The vector was targeted into KV1 (B6/129) hybrid mouse embryonic stem (ES) cells by standard electroporation and selection techniques at the Columbia University Transgenic Core Facility (NY, USA). Chimeras generated from a correctly targeted positive ES cell clone were bred to 129S6/SvEvTac mice (Taconic) to establish lines. To remove the neomycin selection cassette and generate mice carrying the Hoxa5^FLAG allele, Hoxa5^FLAGneo heterozygous mice were crossed with the Sox2cre (Edil3^{Tg(Sox2–cre)1Amc}) mouse line obtained from Dr. Andrew McMahon (Fig. 1A) (Hayashi et al., 2002).

Fig. 1.

Generation of a Hoxa5^FLAG mouse line

(A) Schematic of the targeting strategy to generate the Hoxa5^FLAG epitope-tagged mouse line. The targeting vector features a loxP-PGKneomycinpA-loxP cassette inserted into the Hoxa5 intron and the FLAG insertion at the end of the HOXA5 protein (orange box). The Hoxa5^FLAG allele resulting from the specific action of the Cre recombinase is indicated. Primers used for PCR genotyping are represented by open arrowheads underneath Hoxa5 exon 2. B, BsrGI; E, EcoRI; G, Eagl; H HindIII; R, EcoRV. (B-F) Detection of HOXA5 and HOXA5-FLAG proteins by immunofluorescence for E15.5 wt (B) and Hoxa5^FLAG/FLAG mouse embryos (C-F). Sagittal sections of respiratory tract are shown (B-C). White dashed line boxes represent enlargement of the trachea (D), lung (E) and diaphragm (F) regions from the Hoxa5^FLAG/FLAG mouse embryo. Panels C′-F′ correspond to HOXA5 signal, C″-F″ to FLAG staining, and C-F to the merge of both signals. c, cartilage; Di, diaphragm; e, epithelium, H, heart; L, lung; Li, liver; m, mesenchyme; Tr, trachea. Scale bars: 100 μm (B-F). (G) Western blot analysis of protein extracts from lung of E15.5 wt, Hoxa5^−/− and Hoxa5^+/FLAG mouse embryos. The HOXA5 protein was detected in wt and Hoxa5^+/FLAG specimens, while the HOXA5-FLAG protein is present only in the Hoxa5^+/FLAG specimen. Vinculin was used as a loading control.

The Hoxa5 null (Hoxa5^tm1Rob) mouse line was previously described (Jeannotte et al., 1991, 1993). Age of embryos was estimated by considering the morning of the day of the vaginal plug as embryonic (E) day 0.5. Experimental specimens were genotyped by PCR on DNA from yolk sac or tail biopsies (Holzman et al., 2018). To genotype Hoxa5^FLAG mice, the primers were: Forward: 5′-ATTCAGGACAAAGAGATGAACAGAAGCAG-3’; Reverse: 5′-CTCGCTACCAGACCCTGGAGCTGG-3′, producing bands of 418 base pairs (bp) for the Hoxa5 wild-type allele and 495 bp for the Hoxa5^FLAG allele. All procedures were performed in respect of the guidelines of Canadian Council on Animal Care and agreed by the Université Laval Animal Care Committee or done according to the NIH Guide for Care and Use of Laboratory Animals and approved by the Columbia University IACUC.

2.2. Western blot assays

Lungs were dissected from E15.5 mouse embryos, and they were lysed in 20 mM Tris pH7.4, 150 mM NaCl, 1 mM EDTA, 1 % NP-40, 0.5 % Sodium deoxycholate, 10 mM β-glycerophosphate, 50 mM NaF, 10 mM Sodium pyrophosphate, 10 % Glycerol, and cOmplete Protease inhibitor (Roche). 20 μg of total proteins per well were resolved on SDS-PAGE gels, electrotransferred onto nitrocellulose membrane (Amersham) and probed with the following antibodies: a rabbit polyclonal antibody against HOXA5 (1/100; gift from Dr. Jeremy Dasen), an anti-FLAG HRP-conjugated antibody (1/2,000; Sigma, Cat. No. A8592), and a mouse anti-vinculin (1/2,000; Sigma, Cat. No. V9131) as loading control. Antigen-antibody complexes were detected with an anti-IgG antibody coupled to horseradish peroxidase (HRP) and revealed by chemiluminescence with ECR Clarity MAX (Biorad). Visualization of immunoreactive bands was obtained with Fluo-S MAX MultiImager.

2.3. Immunostaining analysis

Paraffin-embedded E15.5 Hoxa5^FLAG/FLAG and wild-type (wt) embryos were serially sectioned at 5 μm. Immunostaining experiments were performed as described (Gendronneau et al., 2012). The primary antibodies were a rabbit polyclonal antibody against HOXA5 (1/10,000; Dr. Jeremy Dasen) and a mouse monoclonal antibody against FLAG (1/200; Sigma, Cat. No. F3165). Biotinylated secondary antibodies were a goat anti-rabbit or an anti-mouse antibodies (1/250; Jackson ImmunoResearch Labs). Signal revelation was performed with Vectastain ABC kit (Vector Laboratories). HRP activity was detected with the Tyramide Signal Amplification plus Cyanine 3 (1/150; Akoya Biosciences, Cat. No. NEL744001KT) and Cyanine 5 (1/150; Akoya Biosciences, Cat. No. NEL745001KT) for FLAG and HOXA5 detection, respectively. Nuclei were stained with DAPI. Sections were imaged using Zeiss Axio Observer 7 microscope inverted with Zeiss Axiocam 705 color camera and Zeiss 20x EC Plan-Neofluar 0.50 NA DIC.

2.4. In situ hybridization

Paraffin-embedded E12.5 wt and Hoxa5^−/− embryos and lungs from E15.5 embryos were serially sectioned at 5 μm. In situ hybridization was achieved by RNAscope assays using the Multiplex fluorescent V2 kit for mRNA detection (ACD) according to manufacturer’s instructions. The probes used were: Fgf10 (Cat. No. 446371), Bmp4 (Cat. No. 401301-C2), Tbx4 (Cat. No. 483291), Tbx5 (Cat. No. 519581-C2), Ptch1 (Cat. No. 402811) and Wnt2 (Cat. No. 313601). The TSA plus Cyanine 3 and Cyanine 5 were used for probe detection at a range of 1/1,500 to 1/2,000 dilutions depending on the probe. Sections were imaged with a confocal microscope Zeiss LSM 900 AiryScan II inverted with C-Apochromat 40x 1.1 NA water lens.

2.5. Chromatin immunoprecipitation sequencing (ChIP-seq)

ChIP experiments were performed on lungs from E15.5 Hoxa5^FLAG/FLAG mouse embryos as described (Frenette et al., 2024). Fragmented chromatin (20 μg) was incubated overnight at 4°C with 5 μl of FLAG antibody (Sigma, Cat. No. F1804) or with 4 μl of IgG (Millipore, Cat. No. PP64) as control. Libraries and sequencing for two biological replicates were done at the Next-Generation Sequencing Platform, Genomics Center, Centre de recherche du CHU de Québec-Université Laval (Québec, Canada). ChIP-seq librairies were pair-end sequenced at 100 bp reads on Illumina NovaSeq 6000. The mean coverage was 170M of reads for IP samples and 140M for input samples.

Quality of sequencing reads was evaluated using FastQC software (v0.12.1) (http://www.bioinformatics.babraham.ac.uk/projects/fastqc). Reads were trimmed and filtered using Trim Galore (v0.6.10) (https://github.com/FelixKrueger/TrimGalore) with the default parameters. Filtered reads were mapped with Bowtie2 aligner (v2.5.2) (Langmead and Salzberg, 2012) to ENSEMBL mouse reference genome GRCm38.p6 (mm10, release 100) (Mudge and Harrow, 2015). Duplicates were removed using the markdup command from Samtools package (v1.18) (Danecek et al., 2021). Exclusion regions were removed with BEDTools (v2.31.0) (Quinlan and Hall, 2010) using the ENCODE exclusion list (file accession: ENCFF547MET) (Amemiya et al., 2019). Peaks corresponding to regions enriched in HOXA5 binding sites were identified with the MACS3 callpeak module (v3.0.1) (Zhang et al., 2008) with a cut-off of False Discovery Rate (FDR) value of 0.01. Irreproducibility Discovery Rate (IDR) analysis was performed on MACS3 callpeak files to evaluate concordance between replicates (IDR <0.05) (Landt et al., 2012; Li et al., 2011). Replicates were then merged, and a new peak call was performed on the merged data. Only peaks with IDR <0.05 from the previous IDR analysis were kept. For visualization, coverage tracks were generated with the BamCompare command from deepTools (Ramírez et al., 2016) and the number of reads was normalized with the signal extraction scaling algorithm (SES) (Diaz et al., 2012). Peaks in assembly gaps or alignment artifacts previously identified in the mouse genome were removed with BEDTools (Vu and Ernst, 2023). Track visualizations were obtained with pyGenomeTracks (v3.8) (Lopez-Delisle et al., 2020). Pseudogenes were removed for better presentation.

ChIPseeker package from R was used to annotate the localization of HOXA5 binding sites along the genome (Wang et al., 2022). Research of motif sequences was achieved using the findMotifsGenome.pl command from HOMER software in DNA regions containing ± 25 bp from the summit of HOXA5 binding sites (v4.11) (Heinz et al., 2010). AnnotatePeaks.pl command was also applied to characterize the distribution of de novo motifs in HOXA5 binding site regions covering 1 kb in length (Fig. 2E).

Fig. 2.

Genome-wide distribution of HOXA5 binding sites

(A) Pie chart illustrating the distribution of HOXA5 binding genes over genome features. Promoter is defined as a region ± 1 kb from TSS. (B) Chromatin accessibility as determined by ATAC-seq for the 1,308 HOXA5 summit peaks identified. (C) Histone mark frequencies are shown for the five chromatin states defined by ChromHMM analysis for lungs of E15.5 mouse embryos. A representation of genome coverage and enrichment for specific genomic characteristics, genes, TSS and experimentally defined enhancers, for the five chromatin states is shown. The pie chart illustrates the distribution of the 1,308 HOXA5 summit peaks over the five chromatin states. (D) HOMER output of the three top de novo HOXA5 motifs (left side) with the three best matches to known motifs for the top motif (right side). (E) Distribution of the top de novo motif in HOXA5 binding site regions ± 500 bp from peak summit as defined with AnnotatePeaks.pl command. (F) Protein family characterization of the top 50 known motifs enriched at HOXA5 binding sites as found by HOMER. The three top motifs are represented.

rGenomic Regions Enrichment Annotations Tools package from R (rGREAT; v2.8.0) (Gu and Hübschmann, 2022) was used with the default parameters to associate HOXA5 ChIP-seq peaks with genes and perform a functional enrichment analysis for specific gene sets: Gene Ontology (GO) Biological Process terms and Molecular Signatures Database for canonical pathways (Ashburner et al., 2000; Castanza et al., 2023).

2.6. Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq)

Lungs from E15.5 mouse embryos were isolated, snap-frozen in liquid nitrogen and kept at −80°C until use. Nucleus extraction and tagmentation were performed as described (Grandi et al., 2022). Tagmentation with Tn5 transposase (Diagenode, Cat. No. C01070012) was done on 60,000 nuclei. Library preparation and sequencing of four biological replicates were performed at the Molecular Biology and Functional Genomics Core Facility from Institut de Recherches Cliniques de Montréal (Montréal, Canada). Librairies were pair-end sequenced at 100 bp reads on an Illumina NovaSeq 6000 with an average depth of 72M of reads per sample.

As described for ChIP-seq assays, quality of the reads was evaluated with FastQC. Reads were trimmed with Trim Galore and mapped with Bowtie2 to ENSEMBL mouse reference genome GRCm38.p6 (mm10, release 100). Duplicates, exclusion regions and reads mapping to the mitochondrial genome were removed. Reads were then adjusted for Tn5 offset using ATACseqQC package from R (v1.26.0) (Grandi et al., 2022; Ou et al., 2018). MACS3 callpeak module was used to determine the open chromatin regions according to the following parameters: -shift 37, –extsize 73, –nomodel and –nolambda (v3.0.1) (Zhang et al., 2008). Peaks were also called on merged replicates. Only peaks present in at least three replicates with an overlap of at least 50 % were kept using the ChIPpeakAnno package from R (v3.40.0) (Zhu et al., 2010). ChIPpeakAnno was also used to integrate ATAC-seq and HOXA5-FLAG ChIP-seq data and define chromatin openness at the summit of HOXA5 binding sites. Coverage track was generated with the BamCoverage command from deepTools, and number of reads were normalized in bin per millions (BPM). Track visualizations were obtained with pyGenomeTracks.

2.7. Chromatin states

Chromatin state discovery based on the multivariate Hidden Markov Model analysis (ChromHMM) was performed to characterize HOXA5 binding sites identified by ChIP-seq (Ernst and Kellis, 2017). Raw ChIP-seq data for lungs of E15.5 mouse embryos obtained from ENCODE were used for H3K27ac, H3K4me3, H3K27me3, H3K4me1, H3K9ac, H3K36me3 histone marks and input (Table S1). Quality of the data, trimming, mapping and de-duplication were performed as described above for ChIP-seq analysis. Technical replicates were merged after mapping and isogenic (biological) replicates were merged after filtering the alignment. Our ATAC-seq data were also included in the analysis. De-duplicated BAM files for each sample, along with their input control for ChIP-seq data, were binarized with the BinarizeBam function of ChromHMM with a 100 bp resolution. Models considering 2 to 20 states were learned using LearnModels with the same resolution. The CompareModels function was then used to compare decreasing numbers of states from the 20-state model down to the 2-state model. A 5-state model was chosen as it corresponds to the point after which any further increase in the number of states provides no additional information when compared to the 20-state model. To characterize the 5-state model, a functional enrichment of the different states over the specific DNA regions as annotated in NCBI RefSeq Functional Elements (https://www-ncbi-nlm-nih.gov.acces.bibl.ulaval.ca/refseq/functionalelements/) was made: gene sequences, TSS (transcription start site), and experimentally defined enhancers. Annotation of the five states was established as described (van der Velde et al., 2021). The genome coverage of the five chromatin states was used to characterize the summit of HOXA5 binding sites with ChIPpeakAnno.

ChromHMM was also applied to characterize the HOXA5 binding sites detected in lungs from E15.5 mouse embryos compared to other tissues and developmental ages: hindbrain, neural tube, lung, stomach, intestine, and heart at E12.5, E13.5, E14.5, E15.5, E16.5 and postnatal (P) day 0 (Fig. 5). BinarizeBam function was applied to de-duplicated BAM files for H3K27ac, H3K4me1 and H3K27me3 marks from ENCODE with a 100 bp resolution (Table S1). Data were concatenated to generate one shared model with tissue-specific annotations. As described above, multiple models were learned with the LearnModels function and then compared with CompareModels. A 5-state model reflecting the five most different states was chosen. Track visualization was obtained with pyGenomeTracks.

Fig. 5.

Epigenetic analysis of HOXA5 binding sites in target genes

(A) The left panel shows a heatmap of expression for Hoxa5 and the identified targets Fgf10, Bmp4, Tbx4, Tbx5, Ptch1 and Wnt2. It is represented in z-score of Log₂ for the mean of TPM+1 for different tissues and developmental ages. ChromHMM analysis was performed for the three histone marks H3K27ac, H3K4me1 and H3K27me3 and the resulting five state-model is represented at the right end side with a color code. The middle panel represents the chromatin state that intersects with each HOXA5 binding summit associated with the target genes. (B) Representation of the chromatin state distribution over HOXA5 binding site regions ± 1 kb from peak summit using the same color code and the same tissues and developmental ages as in (A). Statistically significant HOXA5 peaks are represented by orange boxes with the black line indicating peak summit. Characterized enhancers for Tbx4, Tbx5 and Wnt2 are shown by pink boxes.

2.8. Link between chromatin states and gene expression

RNA-seq data from ENCODE for the corresponding tissues analyzed for the ChromHMM analysis described in the preceding section were used to link expression of HOXA5 targets to chromatin states and Hoxa5 expression (Fig. 5A; Table S2). Mean of TPM (Transcripts per Million) from Gene quantifications file (ENCODE) for biological replicates was used.

3. Results

3.1. Characterization of the Hoxa5^FLAG allele

To facilitate in vivo assays for HOXA5 genome occupancy, an epitope-tagged Hoxa5 knock-in allele was generated (Fig. 1A). This allele encodes a 3xFLAG tag fused in-frame into the C-terminus of the HOXA5 protein. Removal of the neomycin cassette following Cre recombinase action generated a functional Hoxa5^FLAG allele. Hoxa5^FLAG/FLAG mice are viable, healthy and fertile with a normal lifespan and an apparent wild-type phenotype. By immunofluorescence staining, we showed that the HOXA5-FLAG fusion protein presents the same distribution as the endogenous HOXA5 protein with expression of both proteins in the mesenchymal component of trachea and lung, and in diaphragm (Fig. 1B-F) (Landry-Truchon et al., 2017b). Moreover, HOXA5-FLAG protein was detected in other structures and tissues expressing the Hoxa5 gene, such as the developing spinal cord with an anterior boundary corresponding to the caudal region of the myelencephalon, the myenteric plexus of the gastrointestinal tract and the ribs (Fig. S1) (Aubin et al., 1998, 1999; Boucherat et al., 2013; Mitchel et al., 2022). HOXA5-FLAG protein expression was also assessed by Western blot assay. A band at ~38 kDa corresponding to the endogenous HOXA5 protein was detected in specimens with at least one Hoxa5 wt allele (Coulombe et al., 2010). In specimens carrying a Hoxa5^FLAG allele, a larger band expected for the epitope-tagged HOXA5-FLAG protein was revealed with either HOXA5 or FLAG antibodies (Fig. 1G). Both proteins appeared expressed in similar proportions indicating that the insertion of the epitope did not affect expression levels. Therefore, the HOXA5-FLAG protein behaves as the endogenous HOXA5 and the Hoxa5^FLAG knock-in allele directs normal development mimicking the Hoxa5 wt allele. This mouse model thus provides a valuable tool for in vivo assays by the presence of a tag bypassing the necessity for a ChIP-validated HOXA5 antibody.

3.2. Genome-wide mapping of HOXA5 binding in the developing lung

To address genome occupancy by HOXA5 during lung formation, we first performed chromatin immunoprecipitation followed by sequencing (ChIP-seq) for the HOXA5-FLAG protein in lungs from E15.5 Hoxa5^FLAG/FLAG mouse embryos using a validated FLAG antibody (Ibrahim et al., 2013). At E15.5, lung branching morphogenesis is still in progress concomitant with the differentiation of the respiratory epithelium into specialized cells (Chang et al., 2013; Swarr and Morrisey, 2015). As explained in the Materials & Methods section, we applied a stringent approach to identify HOXA5-FLAG (thereafter named HOXA5) binding regions displaying high concordance between replicates. A total number of 1,308 HOXA5 binding sites was obtained. Genome-wide distribution of these HOXA5 binding sites revealed they were mainly found in intron (42.2 %) and intergenic (42.3 %) regions while only 9.0 % of the peaks were localized in annotated promoters (± 1 kb of known transcription start site (TSS); Fig. 2A). We also performed ATAC-seq assays on lungs from E15.5 wt embryos to distinguish open chromatin regions. Comparison between HOXA5 binding sites and ATAC-seq data revealed an overall enrichment of HOXA5 binding in open chromatin locations (84 %; Fig. 2B).

To characterize the chromatin state of the HOXA5 binding sites, we used ChromHMM analysis to compare HOXA5 binding sites identified by ChIP-seq to the distribution of open chromatin regions from our ATAC-seq data and histone post-translational modifications (Fig. 2C). For the latter, we exploited ENCODE data from lungs of E15.5 mouse embryos for the following six epigenetic marks: H3K4me3, H3K9ac, H3K4me1, H3K27ac, H3K27me3 and H3K36me3 (Table S1). As described in the Materials & Methods section, the five-state model defined by the ChromHMM analysis broadly recapitulated chromatin states previously identified in the mouse genome (Gorkin et al., 2020). A functional enrichment for each of the five states across gene coding sequences, TSS and enhancer regions was also performed to further characterize genomic features in each state (Fig. 2C). H3K4me3 and H3K9ac are mainly associated with promoter regions, H3K27ac is enriched in promoters and enhancers of transcribed genes, while H3K4me1 is primarily present in enhancer regions (Creyghton et al., 2010; Heintzman et al., 2007; Santos-Rosa et al., 2002; Wang et al., 2008). These four histone active marks combined with an open chromatin structure are highly enriched in state 1, which was therefore associated with a permissive state. State 2 harbors an active enhancer signature with open chromatin and H3K4me1 and H3K27ac active marks while state 3 corresponds to an active transcription state with a high frequency of the H3K36me3 positive mark in the coding sequences of actively transcribed genes (Bannister et al., 2005). The repressive H3K27me3 mark is associated mostly with inactive genes and is predominant in state 4 (Cao et al., 2002). Finally, absence of histone marks represents the quiescent state 5. HOXA5 binding sites were found mostly in ChromHMM state 2 (86 %), which displays an active enhancer signature enriched in experimentally defined enhancer annotations (NCBI RefSeq Functional Elements). This result correlates with the high proportion of HOXA5 binding sites present in intergenic regions (Fig. 2A).

To define HOXA5 binding motifs, we applied HOMER analysis to first search genome-wide for enriched de novo motifs in a ± 25 bp sequence from each peak summit. The top enriched motif TGATTCAT showed a high percentage of similarity with HOXA2 (90 %) and HOXA1 (89 %) binding sites previously identified by ChIP-seq assays with the main difference occurring at position 6, which is a C for HOXA5 and a G for HOXA2 and HOXA1 (Fig. 2D) (De Kumar et al., 2017; Donaldson et al., 2012). Moreover, when we searched HOXA5-binding regions covering 1 kb in length, the de novo HOXA5 motif TGATTCAT was found enriched at the summit sequence of 43 % of HOXA5 peaks with less than 3 % in background regions (Fig. 2E). We also compared HOXA5 peak sequences to known DNA binding motifs, and we observed that more than half of the top 50 known enriched motifs match homeobox-containing protein family motifs. The top three known motifs correspond to homeobox-containing proteins PDX1, PBX2 and HOXC9 (Fig. 2F). In much lesser proportions, motifs associated with forkhead box, basic leucine zipper, T-box and POU domains were also detected, some corresponding to previously identified interactors of HOXA5 in other tissues (Catela et al., 2016; Foucher et al., 2002; Kc et al., 2024; Lu & Kamps, 1996). In summary, using the Hoxa5^FLAG allele as a tool for ChIP-seq in vivo assays allowed us to demonstrate that in the developing lung, HOXA5 protein specifically binds to active enhancer regions via specific recognition motifs related to known HOX motifs.

3.3. Direct action of HOXA5 protein on Hox gene expression

We have previously shown the impact of the loss of Hoxa5 function on Hox gene expression. In the developing lung of Hoxa5 null mutants, a negative effect was observed for Hoxa1, Hoxa3, Hoxa4, Hoxa5, Hoxa6 and Hoxb1 expression while Hoxc5 and Hoxd8 expression was augmented (Frenette et al., 2024). To resolve the role of HOXA5 in Hox gene regulation, we searched for HOXA5 binding sites along Hox clusters. In the HoxA cluster, statistically significant HOXA5 binding sites were detected near the Hoxa1, Hoxa2 and Hoxa3 genes, which are all expressed in the developing lung (Fig. 3A) (Frenette et al., 2024). For Hoxa1, the HOXA5 binding site coincides with the TSS of the gene. In contrast, HOXA5 binding sites in the Hoxa2 and Hoxa3 loci overlap with enhancer regions previously described to target expression in rhombomeres (Manzanares et al., 1999; Nonchev et al., 1996; Tümpel et al., 2007; Lampe et al., 2008). A region encompassing the Hoxa5 TSS also showed an enrichment of signal for HOXA5 binding but it was not significant (Fig. 3B). Interestingly, this region is located near a previously reported Hoxa5-responsive element (Zhao et al., 1996). Moreover, all these binding sites showed chromatin openness when aligned with ATAC-seq data, strengthening the notion that HOXA5 regulatory action on flanking HoxA genes occurs in trans via binding to these DNA regions. Finally, additional HOXA5 binding sites were observed 3′ of the HoxA cluster with sites located between ~68 and 178 kb from the Hoxa1 gene (Fig. 3B).

Fig. 3.

Direct regulation of Hox gene expression by HOXA5

(A-D) ChIP-seq profile for HOXA5 protein (grey) and ATAC-seq profile for chromatin accessibility (blue) are shown for HoxA and HoxB genes. Peak call for HOXA5-FLAG is shown by orange boxes with the black line indicating peak summit. (A) Hoxa1, Hoxa2, Hoxa3 genes are represented with enhancers previously described (pink boxes for E1 (Tümpel et al., 2007), E2 (Lampe et al., 2008), E3 (Nonchev et al., 1996), and E4 (Manzanares et al., 1999)). (B) HoxA cluster and the 3′ flanking region with significant HOXA5 binding sites labeled in grey and non-significant HOXA5 binding site found in the Hoxa5 gene labeled in green. (C) Hoxb2 gene. (D) HoxB cluster and its 3′flanking region with significant HOXA5 binding sites labeled in grey.

For the HoxB cluster, only one statistically significant HOXA5 binding site was seen in the first exon of Hoxb2 gene and it is within an open chromatin region (Fig. 3C). As for the HoxA cluster, a HOXA5 binding site was also found in the 3′ flanking region of the HoxB cluster at ~290 kb from the Hoxb1 gene (Fig. 3D). No HOXA5 binding was detected in HoxC and HoxD clusters, a result in agreement with the absence and very low expression of HoxC and HoxD genes in the developing lung (Fig. S2A-B) (Frenette et al., 2024). Therefore, specific Hox genes appear as bona fide HOXA5 transcriptional targets in the developing lung.

3.4. HOXA5 protein binds to lung regulatory genes

Loss of Hoxa5 function in lung mesenchyme was shown to perturb expression of members of several signaling pathways involved in lung morphogenesis (Landry-Truchon et al., 2017b). To specify the role of HOXA5 in lung formation, we first performed a functional enrichment analysis using the Genomic Regions Enrichment Annotations Tool (rGREAT), which revealed several terms related to lung development, lung branching morphogenesis, lung mesenchymal cell proliferation and lung epithelial cell differentiation as well as enrichment for WNT, FGF10, BMP4 and SMO-activated signaling pathways (Fig. S3). These results agreed with the lung phenotypes previously observed in Hoxa5 mouse mutants (Boucherat et al., 2012; Boucherat et al., 2013; Landry-Truchon et al., 2017b). We then searched for the presence of HOXA5 binding sites in lung regulatory genes shown to have their expression affected in Hoxa5 null mutants, such as Fgf10, Bmp4, Wnt2, Tbx4 and Shh (Landry-Truchon et al., 2017b). Statistically significant HOXA5 binding sites were detected in sequences upstream Fgf10 at position −35 kb (from TSS), upstream Bmp4 at position −101 kb, downstream Wnt2 at position +43 kb, and in the third intron of Tbx4 at position +7 kb (Fig. 4). We also looked at Tbx5, since it acts with Tbx4 to modulate Fgf10 expression during lung growth and branching (Arora et al., 2012). Four HOXA5 binding sites were found upstream and downstream of Tbx5 at positions −7 kb, −23 kb, −35 kb and +240 kb (Fig. 4). No HOXA5 binding site was detected near Shh, a predictable result since Shh is specifically expressed in lung epithelium, while HOXA5 expression is restricted to lung mesenchyme (Fig. 1E’) (Bellusci et al., 1997a). However, a HOXA5 binding site was found upstream Ptch1, encoding the receptor of SHH, at position −113 kb (Fig. 4E) (Kugler et al., 2015). ATAC-seq data showed that most binding sites are found within open chromatin domains. Moreover, in several cases HOXA5 binding sites are contained inside established enhancer sequences. For instance, the HOXA5 binding site found in Tbx4 is localized in a lung/trachea enhancer (Menke et al., 2008). The HOXA5 binding site associated with Wnt2 is inside a TBX5-responsive cis-regulatory sequence that directs expression into cardiopulmonary progenitors (Steimle et al., 2018). As well, the HOXA5 binding site at position +240 kb of Tbx5 overlaps with the topologically associating domain (TAD) boundary located between Tbx5 and Lhx5 genes that possesses a lung-specific enhancer signature (Rajderkar et al., 2023). Finally, the position of the HOXA5 binding site near Bmp4 corresponds to an evolutionarily conserved mesoderm-specific cis-regulatory sequence (Chandler et al., 2009).

Fig. 4.

Direct action of HOXA5 on regulators of lung development

(A-F) ChIP-seq profile for HOXA5 protein (grey) and ATAC-seq profile for chromatin accessibility (blue) are shown for Fgf10 (A), Bmp4 (B), Tbx4 (C), Tbx5 (D), Ptch1 (E) and Wnt2 (F) genes. Peak call for HOXA5-FLAG is shown by orange boxes with the black line indicating peak summit. Characterized enhancers are represented by pink boxes for Tbx4 (Menke et al., 2008), Tbx5 (Rajderkar et al., 2023) and Wnt2 (Steimle et al., 2018).

We wanted to further determine if the HOXA5 binding sites found in these putative targets are associated with a lung-specific chromatin state corresponding to active enhancer regions. To do so, we investigated the dynamics of histone mark distribution over HOXA5 binding regions in several mouse tissues, including hindbrain, neural tube, lung, heart, stomach and intestine, at different ages between E12.5 to birth. Hoxa5 is expressed in all tissues except heart (Fig. 5A). H3K27ac, H3K4me1 and H3K27me3 histone marks were selected to distinguish permissive, quiescent and silent enhancer signatures (Fig. 5A). The histone ChIP-seq data were obtained from ENCODE and used in ChromHMM analysis (Table S1). They were also compared to RNA-seq expression results for each target gene in the different tissues studied (Table S2). As expected, all six genes were strongly expressed in the developing lung between E14.5 to birth and the chromatin was in a permissive state in most conditions (Fig. 5A). A closer analysis of the HOXA5 binding regions supported the association of HOXA5 binding with a permissive enhancer signature for the six targets in lung. For instance, Tbx4 and Tbx5 lung-specific enhancers showed a clear lung tropism with an active enhancer chromatin state, while in the other tissues, the chromatin was either quiescent or silent (Fig. 5B). The three upstream HOXA5 binding sites of Tbx5 showed also a permissive chromatin state in heart, in agreement with the strong expression of Tbx5 in the developing heart and its role in cardiac morphogenesis (Bruneau et al., 2001). The lack of Hoxa5 expression in heart implies that this DNA region can be recognized by other transcription factors. A lung tropism was also observed for Fgf10, Ptch1, Bmp4 and Wnt2 genes although in the last two cases, a permissive chromatin state was detected in additional tissues, some co-expressing Hoxa5. This raises the possibility that HOXA5 binding may be involved in Bmp4 and Wnt2 regulation in other tissular contexts. Altogether, these results support a specific link between HOXA5 binding sites and the transcriptional activity of the targets in lung development.

To assess whether the loss of Hoxa5 function alters putative target expression and spatial distribution, we performed in situ hybridization in lungs from wt and Hoxa5^−/− mutant embryos at E12.5, the midst of lung branching morphogenesis, and E15.5 (Fig. 6). At E12.5, Fgf10, Ptch1 and Wnt2 are expressed in the distal mesenchyme surrounding endodermal buds, while Tbx4 and Tbx5 expression is detected throughout lung mesenchyme. Bmp4 is strongly expressed in epithelial tips of distal buds with a fainter expression in the flanking mesenchyme (Fig. 6) (Bellusci et al., 1997b; Chapman et al., 1996; De Langhe et al., 2005; Weaver et al., 2003). Decreased Fgf10, Bmp4, Tbx4, Tbx5, Ptch1 and Wnt2 expression was detected in lung mesenchyme from E12.5 mutant embryos (Fig. 6). At E15.5 in wt condition, Fgf10, Ptch1 and Wnt2 expression appears more broadly distributed in lung mesenchyme, Tbx4 and Tbx5 stay strongly expressed in mesenchyme, whereas Bmp4 expression profile in lung epithelium and mesenchyme remains similar to what was observed at E12.5. Again, all six target genes showed reduced expression in E15.5 Hoxa5^−/− mutant lung specimens (Fig. 6). The strong correlation between the presence of HOXA5 binding sites in the cis-regulatory sequences of these genes and the changes in their expression levels in lung mesenchyme of Hoxa5 null mutants supports the notion that their transcription is directly activated by HOXA5. We can thus assume that in absence of Hoxa5 function, expression of these critical lung regulatory genes is altered causing lung malformations that participate in the subsequent death of Hoxa5 mutant newborns.

Fig. 6.

Comparative in situ hybridization analysis of HOXA5 targets in lung from wt and Hoxa5^−/− embryos

RNAscope assays were performed for (A) Fgf10, Bmp4, (B) Tbx4, Tbx5, (C) Ptch1 and (D) Wnt2 genes on lung tissues from E12.5 and E15.5 wt and Hoxa5^−/− mouse embryos. A white dashed line delimitates lung epithelium. DAPI staining is detected in blue in the merge picture. Scale bars: 50 μm. e, epithelium; me, mesenchyme.

4. Discussion

To decipher how HOXA5 acts in lung organogenesis, we assessed HOXA5 genome-wide occupancy in lungs from E15.5 mouse embryos by a series of ChIP-seq, ATAC-seq and epigenetic analyses. This was made possible by the generation of a FLAG-tagged Hoxa5 knock-in mouse model to detour the need for a ChIP-validated antibody. Epitope-tagged HOX proteins have been successfully used in cell culture-based systems, and Hox epitope-tagged knock-in mouse models were recently reported (De Kumar et al., 2017; Echevarría-Andino et al., 2025; Ibrahim et al., 2013; Jerkovíc et al., 2017). Our Hoxa5^FLAG mouse line presents a FLAG expression profile that recapitulates the endogenous pattern of HOXA5, providing a powerful tool for genome-wide binding studies in different in vivo contexts.

HOXA5 binding sites in embryonic lung are enriched in intergenic and intronic regions, a data in agreement with ChIP-seq results obtained for other HOX proteins (Donaldson et al., 2012; Jerković et al., 2017). Moreover, the de novo HOXA5 motif TGATTCAT was specifically found at the summit of HOXA5 peaks reinforcing our confidence in the authenticity of this motif. Even though this de novo motif does not correspond to HOXA5 motifs previously reported, it does present similitudes with the top motif identified in HOXA5-PBX1 common peaks reported for the developing spinal cord (Kc et al., 2024). In fact, our de novo motif is comparable to motifs identified as functional binding site for HOX and PBX proteins (TGATNNAT) (Bridoux et al., 2020; Donaldson et al., 2012; Mann et al., 2009). This is compatible with the high Pbx1 expression in the developing lung, and it suggests that HOXA5 and its known cofactor PBX1 could act together to regulate expression of target genes in embryonic lung (McCulley et al., 2018). Moreover, the identification of binding motifs for other transcription factors may provide cues to define additional candidate interactors of HOXA5 in the lung.

HOXA5 binding occurs predominantly in open chromatin regions (Fig. 2B). This data concurs with a study showing that binding of HOX transcription factors from central paralog groups (HOX6-8) harbors higher chromatin openness than binding sites from posterior HOX groups (HOX9-13) (Bulajíc et al., 2020). Furthermore, ChromHMM analysis of HOXA5 specific lung targets indicates that HOXA5 is not a determining factor for chromatin accessibility. For instance, the high expression of Hoxa5 in hindbrain does not correlate with an open chromatin structure in the six identified targets (Fig. 5A). Moreover, in the heart, HOXA5 binding sites located upstream Tbx5 and downstream Wnt2 are associated with a permissive chromatin signature even though Hoxa5 is not expressed in this organ (Fig. 5B).

The presence of HOXA5 binding sites in HoxA and HoxB clusters largely matches with the regulatory effect of the Hoxa5 mutation on Hox gene expression and confirms a direct action of HOXA5 (Frenette et al., 2024). In the HoxA cluster, some HOXA5 binding sequences coincide with rhombomere-specific enhancer regions, supporting a link between HOXA5 binding and Hox gene expression. However, in the case of Hoxa2, the presence of HOXA5 binding sites does not result in variation in Hoxa2 expression in Hoxa5 mutant lungs (Frenette et al., 2024). Regulatory sequences can be shared between Hox genes depending on the tissue context (Sharpe et al., 1998). The HOXA5 binding sites found near Hoxa2 could therefore regulate other Hox genes, such as Hoxa4 and Hoxa6, which do not possess HOXA5 binding sites in their close vicinity and show reduced expression in Hoxa5 mutant lungs (Frenette et al., 2024). On the other hand, the impact of the Hoxa5 mutation on Hoxa4 and Hoxa6 expression may be indirect.

HOXA5 binding sites were found in the 3′ flanking region of HoxA cluster (Fig. 3B). Some of these sites are located near early Hox enhancers identified in epiblast stem cells that were shown to make close contact with the HoxA cluster (Neijts et al., 2016; Neijts and Deschamps, 2017). This suggests that HOXA5 could have a remote action on HoxA transcription. As HOXA5 binding sites are also detected in flanking regions of HoxB and HoxD clusters, this points to a possible role for HOXA5 in Hox expression control via a conserved mechanism implying long-range interactions (Fig. 3D-S2C) (Lonfat and Duboule, 2015; Nolte et al., 2019). However, this has to be experimentally validated. In the context of lung development, even though the loss of Hoxa5 function causes only subtle effects on other Hox gene expression, it is possible that the sum of these alterations may contribute to Hoxa5 mutant lung phenotype (Frenette et al., 2024). However, which Hox genes are involved and how they act remain to be elucidated.

The use of the Hoxa5^FLAG mouse line has allowed us to identify HOXA5 targets enriched in biological processes and signaling pathways governing lung development that can enlighten the Hoxa5 mutant lung phenotype (Fig. S3) (Jeannotte et al., 2016; Landry-Truchon et al., 2017b). The impact of the Hoxa5 null mutation on Fgf10 expression in the developing lung, as shown by RT-qPCR and in situ hybridization, can be explained by the presence of HOXA5 binding site upstream Fgf10 (Fig. 4A and 6A) (Landry-Truchon et al., 2017b). The action of HOXA5 on Fgf10 could also be mediated by the direct regulation by HOXA5 of Tbx4 and Tbx5 gene expression, which both control Fgf10 expression (Arora et al., 2012). The effect of the Hoxa5 mutation on Tbx4 and Tbx5 expression was previously demonstrated by RT-qPCR and confirmed by in situ hybridization in this study (Fig. 6B) (Landry-Truchon et al., 2017b). Thus, by acting directly on Fgf10, Tbx4 and Tbx5, HOXA5 promotes lung bud outgrowth (Fig. 7).

Fig. 7.

Driver role for HOXA5 in the mesenchyme of the developing lung.

The presence of HOXA5 binding sites combined to the mesenchymal coexpression of Hoxa5 with its targets and the changes in expression of Fgf10, Bmp4, Tbx4, Tbx5, Ptch1 and Wnt2 in Hoxa5 null mutants support a direct role for HOXA5 in the transcriptional regulation of these genes in the developing lung emphasizing the importance of Hoxa5 in the fine-tuning of lung branching morphogenesis.

It is well established that the stereotypic and repetitive branching program requires finely balanced control between ramification and proliferation that implies a complex network of several signaling cascades (Morrisey and Hogan, 2010). HOXA5 appears to be involved in the regulation of several of these pathways. For instance, HOXA5 directly positively controls the mesenchymal expression of Ptch1, which encodes the receptor of SHH, a negative regulator of lung branching (Bellusci et al., 1997a; Pepicelli et al., 1998). SHH binding to PTCH1 releases the receptor SMO and ultimately allows the GLI transcription factors to translocate to the nucleus and control transcription of effectors (Kugler et al., 2015; Taipale et al., 2002). The SHH pathway is also closely modulated by the cell surface co-receptors BOC, CDON and GAS1 that act positively on the pathway (Allen et al., 2011; Martinelli and Fan, 2007; Tenzen et al., 2006). HOXA5 binding sites were found in the vicinity of genes encoding members of the SHH pathway, such as the transcription factor GLI2 and the co-receptors BOC, CDON and GAS1 (not shown). Altogether, these data indicate a role for HOXA5 in the feedback-loop that regulates Fgf10 spatiotemporal expression, supporting the notion that HOXA5 participates in the fine tuning of lung branching (Fig. 7).

The impact of HOXA5 on lung morphogenesis is also shown by the direct regulation of Bmp4 expression in lung mesenchyme, a result supported by the decreased Bmp4 expression shown here by in situ hybridization assay (Fig. 6A) and previously detected by RT-qPCR (Hrycaj et al., 2015; Landry-Truchon et al., 2017b). Although the roles of Bmp4 in lung formation remain unclear, the positive action of HOXA5 on mesenchymal Bmp4 might generate a diffusible feedback signal that can control lung cell epithelial proliferation (Weaver et al., 2000).

Action of HOXA5 on Bmp4 could also be mediated via the direct binding of HOXA5 on Wnt2, a canonical WNT ligand specifically expressed in lung mesenchyme that shows decreased expression in Hoxa5 null mutants (Fig. 6D) (De Langhe et al., 2005; Landry-Truchon et al., 2017b). Indeed, Bmp4 was shown to be a downstream target of WNT/β-catenin pathway (Shu et al., 2005). Reduction in Wnt2 and Bmp4 expression was previously reported in Hoxa5;Hoxb5;Hoxc5 compound mutants and is explained by the role of Hox5 genes in the control of the Wnt2/Bmp4 axis through Wnt2 (Hrycaj et al., 2015). Here, we demonstrated that the loss of Hoxa5 alone produces a decrease of Wnt2 expression, strengthening evidence for the predominant function of HOXA5 among Hox5 paralogs in lung morphogenesis. Moreover, the direct control of Wnt2 by HOXA5 might explain the lung hypoplasia phenotype observed in Hoxa5^−/− mutant lungs (Boucherat et al., 2013). Thus, by acting on Wnt2 expression, HOXA5 regulates mesenchyme proliferation (Yin et al., 2008). Finally, HOXA5 binding sites were also detected nearby several key activators of the WNT/β-catenin canonical pathway, such as Rspo2/3, Lrp5, Dvl1 and Fzd1 (not shown) (Aros et al., 2021). Together, these data support the notion that HOXA5 promotes mesenchymal proliferation by positively acting on the WNT canonical pathway via Wnt2 and likely other members of the cascade (Fig. 7).

Interestingly, Tbx4 and Tbx5 genes are part of the Tbx2/3/4/5 gene family of T-box transcriptional factors that arose by tandem duplication (Ruvinsky and Silver, 1997). Tbx2 is located upstream Tbx4 while Tbx3 is upstream Tbx5. Like Tbx4 and Tbx5, Tbx2 and Tbx3 are expressed in the mesenchyme of the developing lung. Moreover, Tbx2 and Tbx3 genes play redundant role in maintaining mesenchymal proliferation during lung branching by repressing expression of the cyclin-dependent kinase inhibitors Cdkn1a and Cdkn1b and the Wnt antagonists Frzb and Shisa3 (Lüdtke et al., 2016). Several HOXA5 binding sites were detected near Tbx2 and Tbx3 genes, suggesting a putative role of HOXA5 via Tbx2 and Tbx3 in mesenchyme proliferation (not shown). The presence of HOXA5 binding sites in all four Tbx genes also suggests a conserved action of HOXA5 through evolution of this cluster gene family.

The combination of ChIP-seq, ATAC-seq, epigenetic and expression analyses has led to the identification of HOXA5 direct target genes. However, additional experiments testing the transcriptional activity of these HOXA5 responsive cis-acting sequences and defining the role of the HOXA5 homeodomain in DNA binding will be necessary to establish that they are truly HOXA5 functional downstream effectors. In summary, discovery of HOXA5 direct targets via a new Hoxa5^FLAG mouse model allows to propose that HOXA5 is a strategic driver of transcriptional programs instructing lung development and controlling the fine-tuning of lung branching (Fig. 7).

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.ydbio.2025.08.010.

Data availability

The accession numbers of ChIP-seq and RNA-seq data obtained from ENCODE are provided in Tables S1 and S2. ChIP-seq and ATAC-seq data are available through NCBI GEO (accession numbers GSE299119 and GSE299120).