A Pax3 lineage gives rise to transient haematopoietic progenitors

Deducing cell lineages in developmental biology research is commonly achieved with genetic lineage tracing in the mouse via Cre/LoxP-mediated DNA recombination (Li et al., 2018), whereby Cre recombinase expressed from a cell type-specific mouse promoter irreversibly excises a stop codon upstream of a fluorescent reporter, such as Rosa^Egfp, Rosa^Yfp or Rosa^tdTom (Madisen et al., 2010). Using this method with the Pax3^Cre knock-in allele (Lang et al., 2005), it has been shown that muscle cells and a subset of endothelial cells arise from Pax3-expressing paraxial mesoderm, whereas melanocytes and a subset of neurons arise from Pax3-expressing neural crest (Conway et al., 1997; Wu et al., 2008; Mayeuf-Louchart et al., 2014; Stone and Stainier, 2019; Tani et al., 2020). Recently, bulk RNA-sequencing (RNA-seq) of Pax3^Cre lineage-traced embryonic mouse limbs identified unexplained transcripts for ‘immune system-related genes’ alongside the expected musculoskeletal and neuronal markers typical of paraxial mesoderm and neural crest derivatives (Singh et al., 2018).

During embryonic development, haematopoiesis occurs in temporally and spatially overlapping waves that originate from well-defined tissues (Canu and Ruhrberg, 2021). Extra-embryonic mesoderm gives rise to yolk sac haemogenic endothelium that transitions into pro-definitive haematopoietic progenitors, including erythro-myeloid progenitors (EMPs) and lympho-myeloid progenitors (LMPs) (Zambidis et al., 2006; Canu and Ruhrberg, 2021; Biben et al., 2023). By contrast, intra-embryonic lateral plate mesoderm produces haemogenic endothelium in the dorsal aorta that transitions into definitive haematopoietic stem cells (HSCs) (Wasteson et al., 2008; Seco et al., 2020; Canu and Ruhrberg, 2021). Neither type of mesoderm with haematopoietic potential is known to express Pax3, which instead is considered to be selectively expressed in paraxial mesoderm and neural crest (Conway et al., 1997; Wu et al., 2008; Mayeuf-Louchart et al., 2014; Stone and Stainier, 2019; Tani et al., 2020).

Here, we show that the mouse foetal liver and foetal circulation contain a transient population of Pax3 lineage cells with hallmarks of haematopoietic progenitors and the potential to generate both myeloid and erythroid cells. Our findings raise the possibility that haemogenic potential, observed for extra-embryonic mesoderm and intra-embryonic lateral plate mesoderm, may also extend to paraxial mesoderm.

Re-analysis of bulk RNA-seq data (Singh et al., 2018) from FACS-isolated embryonic day (E) 12.5 mouse limbs genetically lineage traced with Pax3^Cre;Rosa^Egfp detected the expected transcripts for the paraxial mesoderm-derived musculoskeletal (e.g. Myf5, Myog) and endothelial (e.g. Cdh5, Kdr) cell lineages, but also transcripts typical of haematopoietic progenitors (e.g. Kit, Vav1, Runx1), myeloid cells (e.g. Csf1r, Cd68) and erythroid cells (e.g. Hbb-bh1, Hbb-bs; Fig. 1A). Transcription factors required for lymphoid differentiation were also detected (e.g. Ebf1, Tcf7), but not markers of mature lymphocytes (e.g. Cd3g, Cd4, Cd19; Fig. 1A). The finding suggests that a Pax3 lineage may give rise to a subset of haematopoietic cells.

We next examined the Pax3 lineage trace in the mouse foetal liver, which harbours haematopoietic progenitors from midgestation until birth to support foetal haematopoiesis (Canu and Ruhrberg, 2021). Immunofluorescence staining identified Pax3 lineage KDR⁺ endothelial and CD45⁺ haematopoietic cells in E12.5 Pax3^Cre;Rosa^tdTom liver (Fig. 1B). Flow cytometry corroborated the presence of Pax3 lineage KDR⁺ and CD45⁺ cells in E12.5 liver (Fig. 1C; 46.98±11.36% of KDR⁺ endothelial cells and 25.70±7.15% of CD45⁺ haematopoietic cells in the E12.5 liver were lineage-traced with Pax3^Cre). A small fraction of cells (0.40±0.10%) in the foetal liver expressed higher levels of tdTOM, but these cells were negative for both KDR or CD45, excluding endothelial and haematopoietic identity (Fig. 1C). Pax3 lineage cells were also identified in the circulation but were less abundant than in the liver (E12.5 liver: 15.83±5.09%; E12.5 blood: 2.99±3.20%; Fig. 2A,B). In both the liver and blood, the proportion of Pax3 lineage cells peaked at E12.5 and then rapidly declined (Fig. 2A,B). We investigated the identity of Pax3 lineage cells at E12.5 and determined that 16.89±3.46% of Pax3 lineage cells in the liver and 9.64±4.88% in the blood were CD45⁺ haematopoietic cells (Fig. 2C,D). The Pax3 lineage-traced populations at both sites also included TER119⁺ erythroid cells (E12.5 liver: 39.31±5.73%; E12.5 blood: 31.84±24.18%; Fig. 2C,D). Moreover, a subset of CD45⁺Pax3 lineage cells co-expressed the haematopoietic progenitor marker KIT (Fig. 2C,D). Consistent with haematopoietic progenitors being enriched in the foetal liver at midgestation (Canu and Ruhrberg, 2021), Pax3 lineage CD45⁺KIT⁺ cells were more abundant in the liver than blood (E12.5 liver: 12.27±2.00%; E12.5 blood: 3.06±2.73%; Fig. 2C,D).

Next, we quantified the contribution of Pax3 lineage cells to the developing haematopoietic system at E12.5. In the liver, more than 25% of CD45⁺ cells overall and CD45⁺KIT⁺ progenitors were Pax3 lineage-traced (CD45⁺: 25.70±7.15%; CD45⁺KIT⁺: 26.49±7.47%; Fig. 3A). The proportion of Pax3 lineage-traced CD45⁺KIT⁺ cells in the blood was approximately half of that in the liver at E12.5 (12.63±6.89%, Fig. 3B), again consistent with haematopoietic progenitor enrichment in the midgestation foetal liver. Pax3 lineage TER119⁺ erythroid cells were present in the E12.5 liver but rare in the blood (liver: 10.71±2.38%; blood: 0.52±0.27%; Fig. 3A,B). Considering that most TER119⁺ cells in the E12.5 circulation are primitive erythrocytes (Canu and Ruhrberg, 2021), foetal liver presence of Pax3 lineage TER119⁺ cells, but their absence from the circulation, suggests that Pax3 lineage haematopoietic cells are not derived from the so-called primitive haematopoietic wave. Instead, these findings suggest that Pax3 lineage haematopoietic cells home to the foetal liver to differentiate into TER119⁺ erythrocytes, as described for erythroid cell production from pro-definitive EMPs at midgestation (Canu and Ruhrberg, 2021). Although Pax3 lineage haematopoietic cells expressed markers typical of EMPs (Lin⁻KIT⁺CD16/32⁺CD41^low) in the foetal liver (10.95±5.73%; Fig. 3C), these cells were not detected in the E8.5 yolk sac, where EMPs arise (Fig. 3D). These findings suggest that Pax3 lineage cells are EMP-like progenitors generated from a haemogenic source inside the embryo, or that they express Pax3 only after leaving the yolk sac.

During foetal development, HSCs home to the foetal liver from around E12.5 onwards, before they seed the bone marrow to sustain adult haematopoiesis (Canu and Ruhrberg, 2021). Pax3 lineage cells contributed to the LSK (Lin⁻SCA1⁺KIT⁺) haematopoietic stem/progenitor cell pool in the E12.5 foetal liver (9.81±4.36%; Fig. 3E). Therefore, we next investigated whether Pax3 lineage haematopoietic progenitors also migrated to the bone marrow. Flow cytometry of postnatal day (P) 6 Pax3^Cre;Rosa^tdTom bone marrow detected rare Pax3 lineage cells (1.15±0.64%) that were mostly CD45⁺CD11b⁺F4/80⁺ macrophages and CD45⁺CD11b⁺F4/80⁻ myeloid cells; by contrast, Pax3 lineage KIT⁺CD45⁺ progenitors or TER119⁺CD45⁻ erythrocytes were not observed (Fig. 3F). Fewer Pax3 lineage cells were detected in P10 than P6 bone marrow, and they were undetectable in adult bone marrow (Fig. 3F). Absent adult bone marrow colonisation is inconsistent with Pax3 lineage haematopoietic progenitors being a subset of HSCs but consistent with a pro-definitive, EMP-like identity.

Immunofluorescence staining of E12.5 liver showed that Pax3 lineage CD45⁺ haematopoietic cells expressed KI67, a marker of proliferating cells (Fig. 4A), indicating that they could be isolated and placed into colony-forming unit (CFU) assays to determine their haematopoietic potential. Thus, we isolated E12.5 liver CD45⁺KIT⁺ cells lacking differentiated lineage markers for CFU assays, in which they gave rise to multilineage CFU-GEMM along with erythroid and myeloid colonies (Fig. 4B-D). The number and type of colonies produced was similar for tdTOM⁻ and tdTOM⁺ progenitors (Fig. 4D). Detecting uniform tdTOM expression in differentiated CFU colonies excluded the possibility that the haematopoietic potential of explanted progenitors was derived from contaminating tdTOM⁻ cells (Fig. 4E). Further, the Pax3 lineage CFU colonies expressed blood lineage markers consistent with differentiated erythroid and myeloid cells (Fig. 4F). Finding that Pax3 lineage haematopoietic progenitors have erythroid and myeloid differentiation potential, in the context of their transient presence in the foetal liver but absence from bone marrow, again agrees with the idea that these cells are more similar to pro-definitive EMPs than definitive HSCs. The residual presence of Pax3 lineage macrophages in the neonatal bone marrow, in the absence of Pax3 lineage haematopoietic progenitors, is reminiscent of EMP-derived macrophages, which persist as tissue-resident in adults in an organ-specific manner even though their progenitors are only transient (Canu and Ruhrberg, 2021). Nevertheless, future studies should investigate whether Pax3 lineage macrophages also persist in adult organs.

As Pax3 is an established marker of paraxial mesoderm and neural crest, both populations might be considered as potential sources of Pax3 lineage haematopoietic progenitors. In particular, it is conceivable that these progenitors might arise from a subset of paraxial mesoderm-derived endothelial cells with haemogenic potential, in a process analogous to that described for extra-embryonic mesoderm-derived haemogenic endothelium in the yolk sac or lateral plate mesoderm-derived haemogenic endothelium in the dorsal aorta (Canu et al., 2020; Seco et al., 2020; Biben et al., 2023). Alternatively, Pax3 may be expressed in a distinct and hitherto unrecognised source of progenitors with haematopoietic potential that may or may not pass through an endothelial intermediate, in which case Pax3-mediated lineage tracing would be insufficient to attribute cellular origins to paraxial mesoderm.

We have not investigated whether Pax3 lineage haematopoietic progenitors may be involved in haematopoietic malignancies. Nevertheless, it is conceivable that genetic alterations that cause leukemic transformation in haematopoietic stem and progenitor cells might also affect Pax3 lineage haematopoietic progenitors. Furthermore, the Pax3 lineage origin of these transient progenitors raises the possibility that they may also be affected by genetic alterations that are not commonly taken into consideration for haematological studies. In this context, we note that a t(2;13)(q35;q14) genetic translocation produces a tumorigenic PAX3-FOXO1 fusion protein in patients with paediatric alveolar rhabdomyosarcoma (aRMS) (Douglass et al., 1987; Skapek et al., 2019). This mutation increases the proliferation of Pax3-expressing cells while inhibiting their terminal differentiation (Milewski et al., 2021) and also promotes the trans-differentiation of endothelial cells towards a myogenic fate (Searcy et al., 2023). Thus, it should be examined whether PAX3-FOXO1 could also induce myogenic trans-differentiation in haematopoietic cells, or whether it might alter haematopoietic development via an endothelial-derived Pax3 lineage haematopoietic progenitor. These are interesting considerations, because some aRMS patients with PAX3-FOXO1 fusion protein have a bone marrow phenotype resembling acute leukaemia (Sandberg et al., 2001), which is currently thought to result from metastatic aRMS infiltrating the bone marrow, even in cases with no identifiable primary tumour (Karagiannis et al., 2015). Existing mouse models (Keller et al., 2004; Nishijo et al., 2009) could be used to determine whether such translocation can cause primary haematopoietic malignancies.

Here, we have identified Pax3 lineage cells with hallmarks of transient embryonic haematopoietic progenitors. Our results should open new lines of investigations to determine the origins and role of Pax3 lineage progenitors during normal haematopoietic development, and their relevance for haematological malignancies.

Raw bulk RNA-seq reads from Pax3^Cre;Rosa^Egfp mouse embryo limbs (PRJNA422253) (Singh et al., 2018) were downloaded from the European Nucleotide Archive, aligned to GRCm39 and annotated using Mus_musculus.GRCm39.110.gtf at http://www.ensembl.org/info/data/ftp/index.html. Transcripts per million (TPM) values were plotted in Prism 9 (GraphPad).

Animal procedures were performed with Animal Welfare Ethical Review Body (AWERB) and UK Home Office approval. C57BL/6J mice carrying the Pax3^Cre knock-in allele (Lang et al., 2005) were timed-mated to mixed C57BL/6J;129/sv mice carrying a Rosa^tdTom recombination reporter (Madisen et al., 2010). Cryosections of formaldehyde-fixed E12.5 liver were blocked in PBS containing 10% serum-free protein block (DAKO), 2.5% bovine serum albumin (BSA), and 0.1% Triton X-100 before staining with the following antibodies as indicated in the figures: goat anti-mouse KDR (R&D Systems, #AF644, Lot COA0420091, 1:50), rat anti-mouse CD45 (BD Biosciences, #550539, Lot 9301732, 1:50), rabbit anti-mouse RFP (MBL, #PM005, Lot 048, 1:250), rabbit anti-mouse KI67 (Abcam, #ab16667, Lot GR8416681-9, 1:100), goat anti-mouse RFP (OriGene, #AB1140-100, Lot 081290121, 1:250). This was followed by Alexa Fluor 647-conjugated donkey anti-goat Fab fragments (Stratech, #705-606-147, 1:200), Alexa Fluor 488-conjugated donkey anti-rat Fab fragments (Stratech, #712-547-003, 1:200), Cy3-conjugated donkey anti-rabbit Fab fragments (Stratech, #711-166-152, 1:200), Alexa Fluor 647-conjugated donkey anti-rabbit Fab fragments (Stratech, #711-607-003, 1:200), or Cy3-conjugated donkey anti-goat Fab fragments (Stratech, #705-167-003, 1:200). DAPI-counterstained sections were imaged on a Ti2 microscope with NIS-Elements automatic deconvolution (Nikon) and on a LSM710 confocal microscope (Zeiss).

Embryos were collected in ice-cold FACS buffer: RPMI with 2.5% foetal bovine serum (Thermo Fisher Scientific), 100 µg/ml heparin and 50 µg/ml DNAse I (Sigma-Aldrich). Extra-embryonic and head tissues were removed for blood collection before livers were dissected and dissociated in FACS buffer with 100 µg/ml collagenase/dispase (Sigma-Aldrich) for 20 min and TrypLE (Gibco) for 3 min. Single cell suspensions from foetal liver, foetal blood or postnatal femur and tibia bone marrow, were incubated with Fc block (BioLegend) for 30 min before labelling with the following antibodies (BioLegend): KIT (clone 2B8, 1:50), CD45 (clone 30-F11, 1:50), TER119 (clone TER-119, 1:100), CD11b (clone M1/70, 1:50), CD41 (clone MWReg30, 1:50), F4/80 (clone BM8, 1:50), Lineage (Lin) cocktail (CD3ε, clone 145-2C11; Ly-6G/Ly-6C, clone RB6-8C5; CD11b, clone M1/70; CD45R/B220, clone RA3-6B2; TER-119, clone Ter-119; 1:100), CD16/32 (clone 93, 1:50), SCA1 (clone D7, 1:50). Live cells were analysed using a Fortessa X-20 (BD Biosciences) or sorted on a FACSAriaIII (BD Biosciences). Data were analysed using FlowJo VX (FlowJo) and Prism 9 (GraphPad).

Haematopoietic differentiation potential was tested in CFU assays using Methocult GF-M3434 (STEMCELL Technologies), following the manufacturer's instructions.

We thank the staff of the Biological Resources, FACS and Imaging Facilities at the UCL Institute of Ophthalmology.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202924.reviewer-comments.pdf