Discovery of a Vertebral Skeletal Stem Cell Driving Spinal Metastases

Abstract

Vertebral bone is subject to a distinct set of disease processes from those of long bones, notably including a much higher rate of solid tumor metastases that cannot be explained by passive blood flow distribution alone. The basis for this distinct biology of vertebral bone has remained elusive. Here we identify a vertebral skeletal stem cell (vSSC), co-expressing the transcription factors ZIC1 and PAX1 together with additional cell surface markers, whose expression profile and function are markedly distinct from those of long bone skeletal stem cells (lbSSCs). vSSCs display formal evidence of stemness, including self-renewal, label retention and sitting at the apex of their differentiation hierarchy. Lineage tracing of vSSCs confirms that they make a persistent contribution to multiple mature cell lineages in the native vertebrae. vSSCs are physiologic mediators of spine mineralization, as genetic blockade of the ability of vSSCs to generate osteoblasts results in defects in the vertebral neural arch and body. Human counterparts of vSSCs can be identified in vertebral endplate specimens and display a conserved differentiation hierarchy and stemness. Multiple lines of evidence indicate that vSSCs contribute to the high rates of vertebral metastatic tropism observed clinically in breast cancer. Specifically, when an organoid system is used to place both vSSCs and lbSSCs in an identical anatomic context, vSSC-lineage cells are more efficient than lbSSC-lineage cells at recruiting metastases, a phenotype that is due in part to increased secretion of the novel metastatic trophic factor MFGE8. Similarly, genetically targeting loss-of-function to the vSSC lineage results in reduced metastasis rates in the native vertebral environment. Taken together, vSSCs are distinct from other skeletal stem cells and mediate the unique physiology and pathology of vertebrae, including contributing to the high rate of metastatic seeding of the vertebrae.

Extended Data Fig. 1 |. Exclusion of potential lineage marker for vertebrae.

a, Representative fluorescence images of vertebrae and long bone from Prrx1-cre mTmG, Pax7-cre mTmG, Mpz-cre mTmG and Shh-cre ai9 mice at P1. NP, nucleus pulposus. n=2 for each mouse line. Scale bar, 100 μm. b, Gating strategy for FACS of SSCs in long bones (lbSSCs) and candidate SSCs in vertebrae (candidate vSSCs). c, Flow cytometry for the contribution of different cre lineage cells to lbSSCs or candidate vSSCs. n=2 or 3 for each mouse line.

Extended Data Fig. 2 |. EMB labels a non-stem population

a, Clustering heatmap showing the differentially expressed genes in candidate vSSCs (Lin-THY1-6C3-CD200+CD105-) versus lbSSCs (Lin-THY1-6C3-CD200+CD105-). Each row represents one gene, and each column represents one sample. b&c, Immunofluorescence staining for COL10A1, EMB and OPN on P10 mouse vertebrae (b) and tibia (c) sections, showing the expression of EMB in osteoblasts and hypertrophic chondrocytes. Images are representative of two independent experiments. Scale bar, 50 μm.

Extended Data Fig. 3 |. Identification of vSSC markers.

a, Violin plot showing the expression of Pax1, Zic1, Pax9, Prdm6 and Meox2 genes in candidate vSSCs and lbSSCs as determined by RNA-seq. n=4, data are presented as mean±s.d., unpaired, two-tailed Student’s t test. b, Violin plot showing the expression of Dpt, Ramp2, Avpr1a and Ropo3 genes in candidate vSSCs and lbSSCs as determined by RNA-seq. n=4, data are presented as mean±s.d., unpaired, two-tailed Student’s t test. c, Heatmap showing the top differentially expressed transcription factors in candidate vSSCs and lbSSCs. d, Diagram of the screen of vertebral-specific transcription factors in lbSSCs (top) and RT-PCR showing the activation of vertebral reporter genes in lbSSCs overexpressing single vertebral-specific transcription factors. n=4, data are presented as mean±s.d., unpaired, two-tailed Student’s t test.

Extended Data Fig. 4 |. Generation of Zic1-cre and Pax1-creER^T2 knock-in mouse lines.

a, Generation of Zic1-cre line using CRISPR-Cas9. Structure of Zic1 locus (top), targeting vector (middle) and knock-in allele (bottom). b, DNA sequencing result of the knock-in region. The sequence of knock-in element is labeled as red. c, Generation of Pax1-creER^T2 line using CRISPR-Cas9. Structure of Pax1 locus (top), targeting vector (middle) and knock-in allele (bottom). d, DNA sequencing result of the knock-in region. The sequence of knock-in element is labeled as red.

Extended Data Fig. 5 |. Zic1-cre and Pax1-creER^T2 label vertebrae.

a, Flow cytometry analysis of Zic1-cre cells with mTmG reporter in vertebrae at P10. n=5. b, Representative fluorescent image (left) and flow cytometry analysis (right) of femur from Pax1-creER^T2 mTmG mice 2 weeks after Tamoxifen injections at P8 and P9. Data are representative of three independent experiments. Scale bars, 100 μm. c, Immunofluorescence images of 6-month-old Pax1-creER^T2 mTmG mouse lumbar vertebral section (Tamoxifen induction at P8&P9), showing the contribution of Pax1-creER^T2 cells to osteoblasts with anti-OPN antibody staining (top), and osteocyte in trabecular bone. Data are representative of three independent experiments. Scale bars, 20 μm. d, Immunofluorescence staining for EMB and PAX1 on 4-week-old mouse vertebrae sections. NP, nucleus pulposus, EP, endplate, BM, bone marrow. Data are representative of two independent experiments. Scale bars, 100 μm. e, RT-PCR analysis of Pax1, Zic1, and Emb genes in FACS sorted Zic1-lineage vSSCs, lbSSCs and vertebral CD45+ cells. n=4, data are presented as mean±s.d., one-way ANOVA followed by Tukey’s multiple comparison test.

Extended Data Fig. 6 |. vSSCs are label retaining cells.

a&b, Flow cytometry of H2B-GFP^hi label retaining cells in the vertebrae of H2B-GFP/rtTA mice before and after 6 months of doxycycline treatment. c, Diagram of label retention experiment: doxycycline food was administrated to the H2B-GFP/tTA mice to turn on GFP expression from 4-week-old to 9-week-old and replaced with chow diet after 9-week-old to turn off GFP expression, label retaining cells were analyzed 2 months or 14 months after doxycycline treatment. d, Representative flow cytometry plots showing the total GFP+ cells in vertebrae of H2B-GFP/tTA mice at indicated time points. e-g, Flow cytometry analysis of H2B-GFP^hi label retaining cells in vertebrae of H2B-GFP/tTA mice before and after 2-month or 14-month doxycycline treatment.

Extended Data Fig. 7 |. Zic1-lineage EMB+ cells lack stem cell properties.

a, Flow cytometry of the cell populations derived from Zic1-lineage/Lin-THY1-6C3-CD200+CD105-EMB+ cells after the first round (top panels) and second round (bottom panels) of intramuscular transplantation. Plots are representative of 3 independent experiments.

Extended Data Fig. 8 |. Deletion of key osteoblast factors in Zic1-cre cells lead to vertebral mineralization defects and hindlimb paralysis.

a, Representative images showing the hindlimb paralysis phenotype observed in 4-week-old Osx^zic1 mice. b, Splay reflex test showing that 4-week-old Osx^f/f mice have a normal limb splaying reflex while Osx^zic1 mice display spasticity due to paraplegia. c, Whole-mount skeletal staining of 4-week-old Osx^zic1 and Osx^f/f mice at the thoracolumbar region showing the mineralization defect of the dorsal vertebrae in Osx^zic1 mice. n=3.

Extended Data Fig. 9 |. The tropism of breast cancer cells vertebrae.

a, Representative bioluminescence images showing cancer metastasis in long bones and vertebrae 4 weeks after injection of 4T1.2 cells through the caudal artery. Scale bar, 5 mm. b, Representative bioluminescence images showing cancer metastasis in long bones and vertebrae 3 weeks after injection of E0771 cells through the caudal artery. c, Diagram of the blood flow distribution experiment. d, Representative flow cytometry plots showing the microspheres in long bone and vertebrae 1min after caudal artery injection. e, Quantification of the microspheres in long bone and vertebrae 1 min after caudal artery injection. n=6, data are presented as mean±s.d., unpaired, two-tailed Student’s t test. f, Representative histology images of H&E staining showing metastasized PY8119 cells in lbSSC-derived and vSSC-derived bone organoids 3 weeks after cancer cells injection, scale bar 200 μm. g, Diagram of the study of 4T1.2 metastasis to bone organoids. h, Quantification of the metastasis rate of 4T1.2 cells to lbSSC-derived and vSSC-derived bone organoids 3 weeks after cancer cells injection. n=42. Chi-square test. i, Representative bioluminescence images showing 4T1.2 metastasis in lbSSC-derived and vSSC-derived bone organoids 3 weeks after cancer cells injection. Scale bar, 5 mm. j, Diagram of the bone organoid early seeding experiment. k, Quantification of the early seeding cancer cells in lbSSC-derived and vSSC-derived bone organoids 20h after cancer cells injection. n=9, data are presented as mean±s.d., unpaired, two-tailed Student’s t test. l. Representative flow cytometry plots showing the early seeding cancer cells in lbSSC-derived and vSSC-derived bone organoids 20h after cancer cells injection. m, Representative histology images of three independent experiments showing the early seeding of cancer cells in lbSSC-derived and vSSC-derived bone organoids 20h after 4T1.2 cancer cells injection, scale bar, 20 μm.

Extended Data Fig. 10 |. Identification of vSSC secreted proteins mediating breast cancer metastatic tropism.

a, Heatmap showing the top differentially expressed secreted proteins in Zic1-lineage vSSCs versus lbSSCs. b, qPCR analysis of candidate vertebral-derived metastasis tropism factors in FACS isolated Zic1-lineage vSSCs (Lin-THY1-6C3-CD200+CD105-EMB-GFP+) and lbSSCs (Lin-THY1-6C3-CD200+CD105-EMB-). n=4, data are presented as mean±s.d., unpaired, two-tailed Student’s t test. c, Quantification of bone volume/total volume (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th) and trabecular separation (Tb.Sp) of L5 vertebrae in 8-week-old Mfge8^−/− and WT mice. n=4, data are presented as mean±s.d., unpaired, two-tailed Student’s t test.

Fig. 1 |. Identification of vSSC markers.

a, Principal component analysis (PCA) of RNA-Seq following FACS of Lin-THY1-6C3-CD200+CD105- populations from the vertebrae (candidate vSSCs) and long bone(lbSSCs) of P10 mice. b, Representative flow cytometry analysis of the EMB+ and EMB− fraction of lbSSCs and candidate vSSCs. n=3. c, Principal component analysis (PCA) of RNA-seq following FACS of Lin-THY1-6C3-CD200+CD105-EMB+ and Lin-THY1-6C3-CD200+CD105-EMB- populations from the vertebrae and long bones of P10 mice. d, RNA-seq analysis using a Violin plot showing the expression of Alpl, Spp1, Col10a1, Ihh, Sp7, Runx2 and Col2a1 in EMB− candidate vSSCs and EMB+ candidate vSSCs. n=4, data are presented as mean±s.d., unpaired, two-tailed Student’s t test. e, Immunostaining for EMB and COL10A1 on thoracic vertebrae (T10) from Ocn-cre mTmG mice at 1-month of age. Data are representative of two independent experiments. Scale bar, 100 μm. f, Flow cytometry analysis of Ocn-cre mTmG vertebral cells. n=5 mice. Data are presented as mean±s.d. In the histogram pictures, grey peaks represent FMO control cells for EMB staining and red peaks represent GFP+lbSSCs or GFP+ Candidate vSSCs. g, Flow cytometry analysis of cells derived from EMB+ vs EMB− candidate vSSCs 7 days after mammary fat pad transplantation and organoid formation. Plots are representative of 3 independent experiments.

Fig. 2|. Zic1-cre and Pax1-creER^T2 label candidate vSSCs.

a, Volcano plot highlighting the differentially expressed transcriptional factors in candidate vSSCs and lbSSCs (FDR<0.05). b, Representative fluorescence images of the lumbar vertebra (L3, left) and femur (right) from Zic1-cre mTmG mice at P10. Scale bar, 100 μm. Data are representative of three independent experiments. c, Flow cytometry analysis of the mGFP+ Zic1-lineage cells within immunophenotypic SSCs (Lin-THY1-6C3-CD200+CD105-EMB- populations) in lumbar vertebrae (L1-L6) and femurs. n=4. d, Immunofluorescence staining of Zic1-cre mTmG mouse caudal vertebrae for Perilipin, showing the contribution of Zic1-lineage cells to marrow adipocytes. Data are representative of two independent experiments. Scale bar, 50 μm. e, Immunofluorescence staining of Zic1-cre mTmG mouse lumbar vertebrae with anti-OPN antibody, showing the contribution of Zic1-cre cells to osteoblasts. Data are representative of two independent experiments. Scale bar, 50 μm. f, Representative fluorescence images of the cervical, thoracic, lumbar and sacral vertebrae from Zic1-cre mTmG mice at 6-weeks of age. n=2 biological replicates. Scale bar, 100 μm. g, Tracing of Pax1-lineage cells in the endplate region of Pax1-creER^T2 mTmG mice after a tamoxifen pulse on P8 and P9. n=4 biological replicates. Scale bar, 50 μm. h, Flow cytometry of Pax1-creER^T2 mTmG mice 24h (n=6), 2w (n=8), 4w (n=10), or 1 year (n=6) after a tamoxifen pulse. Data are mean±s.d. i, Immunofluorescence staining of P4 Zic1-cre mTmG mouse lumbar vertebrae for PAX1, showing PAX1 expression in Zic1-lineage cells at the resting zone of vertebral endplate. Data are representative of two independent experiments. Scale bar, 20 μm.

Fig. 3 |. vSSCs fulfill stemness criteria.

a, In vivo clonal analysis of Pax1-creER^T2 cells at the vertebral endplate region in 3-month-old mice after tamoxifen induction at 1 month. Representative of 4 biologic replicates. Scale bar, 20 μm. b, Summary diagram for label retention studies: doxycycline chow was administrated to H2B-GFP/rtTA mice to suppress de novo H2B-GFP expression starting at 8-weeks of age. Label retaining cells were analyzed 6 months after initiating doxycycline treatment. c, Flow cytometry analysis of H2B-GFP^hi cells (label retaining cells) before and after doxycycline treatment. Representative of 4 biologic replicates. d, Immunostaining for PAX1 showing expression of PAX1(magenta) in H2B-GFP (green) label retaining cells at the vertebral endplate in 8-month-old H2B-GFP/rtTA mice, after 6 months of chase. Representative of 2 independent experiments. Scale bar, 50 μm. e, uCT analysis of lbSSC (Lin-THY1-6C3-CD200+CD105-EMB-) and Zic1-lineage vSSC-derived bone organoids 6 weeks after intramuscular transplantation showing 3D reconstruction (left) and quantification of bone volume (right). n=7 organoids/host mice per group. Data are mean±s.d., unpaired, two-tailed Student’s t test. Scale bar: 0.5 mm. f, Representative Von Kossa staining (black) for mineralized bone in lbSSC and Zic1-lineage vSSC-derived bone organoids 8 weeks after intramuscular transplantation. n=4. Scale bars, 500 μm. g, Representative images of Safranin O staining (red) for cartilage in lbSSC and Zic1-lineage vSSC-derived bone organoids 2 weeks after intramuscular transplantation. n=4. Scale bars, 200 μm. h, Representative images of H&E staining showing marrow recruitment in the bone organoids 4 month after intramuscular transplantation. n=3. Scale bars, 500 μm. i, Immunostaining of Perilipin+ adipocytes in lbSSCs and Zic1-lineage vSSC-derived bone organoids 4 months after intramuscular transplantation. n=3. Scale bars, 500 μm. j, Schematic diagram for the in vivo serial transplantation experiment. k, Flow cytometry analysis of vSSC-derived cell populations after the first round (top panels) and second round (bottom panels) of intramuscular transplantation. Plots are representative of 3 independent experiments.

Fig. 4|. vSSCs contribute to physiologic mineralization.

a, Representative X-ray images showing vertebral mineralization defects in 4-week-old Osx^zic1 mice. n=2. b, 3D reconstruction of spine μCT images from 4-week-old Osx^zic1 and Osx^f/f mice. Scale bar, 5 mm. c, Dorsal view of a 3D reconstruction of thoracic and lumbar spine μCT images showing the absence of the neural arch in 4-week-old Osx^zic1 mice. Scale bar, 1 mm. d, 3D reconstruction of L5 vertebra μCT from 4-week-old Osx^zic1 and Osx^f/f mice. Scale bar, 100 μm. e, Quantification of bone volume/total volume (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th) and trabecular separation (Tb.Sp) of L5 vertebra in 4-week-old Osx^zic1 and Osx^f/f mice. n=6, data are presented as mean±s.d., unpaired, two-tailed Student’s t test. f, 3D reconstruction of L5 vertebrae μCT from 4-week-old Stat3^zic1 and Stat3^f/f mice. Scale bar, 100 μm. g, Quantification of bone volume/total volume (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th) and trabecular separation (Tb.Sp) of L5 vertebrae in 8-week-old Stat3^zic1 and Stat3^f/f mice. n=5 or 7, data are mean±s.d., unpaired, two-tailed Student’s t test. h, 3D reconstruction of L5 vertebra (top) and femur (bottom) μCT from 8-week-old Shn3^zic1 and Shn3^f/f mice. Scale bar, 100 μm. i, Quantification of bone volume/total volume (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th) and trabecular separation (Tb.Sp) of L5 vertebrae (top) and femur (bottom) in 8-week-old Shn3^zic1 and Shn3^f/f mice. n=7–8, data are mean±s.d., unpaired, two-tailed Student’s t test. j, Immunostaining for PAX1 and ZIC1 in a human endplate specimen showing co-expression of PAX1 and ZIC1. Images are representative of 3 independent experiments. Scale bar, 20 μm. k, Flow cytometry analysis of human vSSCs in a human endplate specimen. Plots are representative of 5 independent experiments. l, uCT analysis showing 3D reconstruction of human vSSCs and sponge control derived bone organoids 6 weeks after intramuscular transplantation. n=3, scale bar 0.5 mm. m, Representative images of Von Kossa staining (left) for mineralized bone and Safranin O staining (right) for cartilage in the bone organoids derived from human vSSCs and sponge control 6 weeks after intramuscular transplantation. n=3. Scale bars, 500 μm. n, Flow cytometry analysis of human vSSC-derived cell populations 7 days after intramuscular transplantation. Plots are representative of 2 independent experiments.

Fig. 5 |. vSSCs drive the preferential metastasis of breast cancer to the vertebrae.

a, Representative bioluminescence images showing metastasis of PY8119 cells to long bones and vertebrae 4 weeks after injection via the caudal artery. Scale bar, 5 mm. b-d, Quantification of metastasis rates to long bones and vertebrae 3–4 weeks after caudal artery injection with PY8119 (b, 4 weeks, n=32), 4T1.2 (c, 4 weeks, n=30), and E0771 cells (d, 3 weeks, n=25). Chi-square test. e, Schematic diagram of the early seeding experiment. f. Representative flow cytometry plots showing the initially seeding cancer cells in long bones and vertebrae after caudal artery injection. g, Quantification of the early seeding cancer cells in vertebrae and long bone 20h after cancer cells injection. n=13, data are mean±s.d., unpaired, two-tailed Student’s t test. h, Representative histology images of three independent experiments showing the early seeding cancer cells in vertebrae and long bones. i, Schematic diagram of the bone organoid metastasis experiment. j, Representative fluorescent histology images showing PY8119 (Green) metastasis to lbSSC-derived and vSSC-derived bone organoids (Red) 3 weeks after cancer cells injection, scale bar 200 μm. k, Quantification of the metastasis rate of PY8119 cells to lbSSC-derived and vSSC-derived bone organoids 3 weeks after cancer cells injection. n=40. Chi-square test. l, Representative bioluminescence images showing PY8119 metastasis in lbSSC-derived and vSSC-derived bone organoids 3 weeks after cancer cells injection. Scale bar, 5 mm. m, Quantification of the metastasis rate of PY8119 cells to WT or Stat3^zic1 vertebrae 4 weeks after cancer cells injection. n=31 for WT mice and n=21 for Stat3^zic1 mice. Chi-square test. n&o, Quantification (n) and representative images (o) of PY8119 transwell migration in the presence of the indicated concentrations of recombinant MFGE8 in the lower chamber. n=4, data are mean±s.d., one-way ANOVA followed by Tukey’s multiple comparison test. p&q, Quantification (p) and representative images (q) of PY8119 transwell migration in the presence of bone marrow stromal cells in the lower chamber isolated from long bones or vertebrae of 3-week-old WT or Mfge8^−/− mice. n=4 or 5, data are mean±s.d., two-way ANOVA followed by Tukey’s multiple comparison test. r, Representative bioluminescence images showing cancer metastasis in long bones and vertebrae of WT and Mfge8^−/− mice 4 weeks after PY8119 cells injection through caudal artery. Scale bar, 5 mm. s, Quantification of the metastasis rate to long bones and vertebrae 4 weeks after caudal artery injection with PY8119 cells, n=51 for WT mice and n=61for Mfge8^−/− mice. Chi-square test. t, Quantification of the early seeding of cancer cells in the vertebrae and long bones of WT and Mfge8^−/− mice 20h after cancer cells injection. n=11 for WT and n=10 for Mfge8^−/− mice, data are mean±s.d. Two-way ANOVA followed by Tukey’s multiple comparison test. u, Representative flow cytometry plots showing the early seeding of cancer cells in long bone and vertebrae of WT and Mfge8^−/− mice 20h after cancer cells injection. v, Quantification of the metastasis rate of PY8119 cells to WT or Mfge8^−/− vSSC-derived bone organoids 3 weeks after cancer cells injection. n=54. Chi-square test. w, Representative bioluminescence images showing metastasis of PY8119 cells to WT or Mfge8^−/− vSSC-derived bone organoids 3 weeks after caudal artery injection. Scale bar, 5 mm.