TGF-βR2 signaling coordinates pulmonary vascular repair after viral injury in mice and human tissue

Abstract

Disruption of pulmonary vascular homeostasis is a central feature of viral pneumonia, wherein endothelial cell (EC) death and subsequent angiogenic responses are critical determinants of the outcome of severe lung injury. A more granular understanding of the fundamental mechanisms driving reconstitution of lung endothelium is necessary to facilitate therapeutic vascular repair. Here, we demonstrated that TGF-β signaling through TGF-βR2 (transforming growth factor-β receptor 2) is activated in pulmonary ECs upon influenza infection, and mice deficient in endothelial Tgfbr2 exhibited prolonged injury and diminished vascular repair. Loss of endothelial Tgfbr2 prevented autocrine Vegfa (vascular endothelial growth factor α) expression, reduced endothelial proliferation, and impaired renewal of aerocytes thought to be critical for alveolar gas exchange. Angiogenic responses through TGF-βR2 were attributable to leucine-rich α-2-glycoprotein 1, a proangiogenic factor that counterbalances canonical angiostatic TGF-β signaling. Further, we developed a lipid nanoparticle that targets the pulmonary endothelium, Lung-LNP (LuLNP). Delivery of Vegfa mRNA, a critical TGF-βR2 downstream effector, by LuLNPs improved the impaired regeneration phenotype of EC Tgfbr2 deficiency during influenza injury. These studies defined a role for TGF-βR2 in lung endothelial repair and demonstrated efficacy of an efficient and safe endothelial-targeted LNP capable of delivering therapeutic mRNA cargo for vascular repair in influenza infection.

Fig. 1.. Up-regulation of TGFBR2 and TGF-βR2–mediated signaling after viral injury.

(A) Reanalysis of published snRNA-seq datasets of healthy (n = 7 donors) and COVID (n = 19 donors) lungs from Melms et al. (30), with EC clusters labeled. (B) Left: Endothelial clusters were subsetted from (A) and confirmed by signature EC gene PECAM1 (CD31). Right: Uniform Manifold Approximation and Projection (UMAP) plot of reclustered EC populations. (C) Volcano plot displaying differentially expressed genes in healthy and COVID lung ECs. FC, fold change. (D) Violin plot of TGFBR2 mRNA expression (log-normalized) in ECs. P = 1.24 × 10⁻¹³ was determined by Wilcoxon rank-sum test. (E) Left: Representative immunofluorescence image of endothelial (ERG⁺) TGF-βR2–expressing cells in both healthy and post-COVID lung tissue. Scale bar, 50 μm. Right: Quantification of TGF-βR2 immunostaining (healthy donors, n = 3; COVID donors, n = 3). InDen, integrated density. Data are means ± SEM, unpaired two-tailed t test, *P = 0.038. DAPI, 4′,6-diamidino-2-phenylindole; AU, arbitrary units. (F) qPCR analysis of TGFBR2 in sorted human lung ECs from both healthy (n = 5) and post-COVID (n = 4) donors. Data are means ± SD, unpaired two-tailed t test, **P = 0.002. (G) scRNA-seq analysis for mouse lung ECs sorted from uninjured (D0) and on 20 and 30 days after influenza infection (marked as D20 and D30, respectively). (H) UMAP plot showing EC marker gene (Pecam1). (I) UMAP plot of EC populations annotated in scRNA-seq data for adult mouse lung (34, 57). (J) Violin plots showing Tgfbr2 expression in mouse lung ECs sorted from D0, D20, and D30, respectively. P value was determined by Wilcoxon rank-sum test, D0 versus D20, P < 10⁻³⁰⁷; D0 versus D30, P = 1.45 × 10⁻²⁴⁸. (K) qPCR analysis of Tgfbr2 in isolated lung ECs (CD45⁻CD31⁺) sorted on day 0 (uninjured) and on days 10 and 20 after influenza infection. n = 3 or 4 per group; data are presented as means ± SEM, unpaired two-tailed t test. *D0 versus D10, P = 0.0326; D0 versus D20, P = 0.0198. (L) Heatmap comparing TGF-β pathway gene expression in mouse lung ECs sorted from D0, D20, and D30, respectively.

Fig. 2.. Endothelial Tgfbr2 deletion invivo prevents EC repair after influenza injury.

(A and B) VECAD^CreERT2; Tgfbr2^flox/flox or WT (VECAD^CreERT2 or Tgfbr2^flox/flox) mice were administered five doses of tamoxifen, followed by 3 weeks of chase and influenza infection. Time course of changes in capillary oxygen saturation (A) and body weight (B) in WT and Tgfbr2^ECKO mice, n = 5 to 8 per group. Data are means ± SEM, unpaired two-tailed t test; *P < 0.05 and **P < 0.01. WT versus Tgfbr2^ECKO in (A): D17, P = 0.028; D19, P = 0.018; D26, P = 0.034. WT versus Tgfbr2^ECKO in (B): D16, P = 0.01; D20, P = 0.0013; D26, P = 0.019. (C) Kaplan-Meier survival curves after influenza infection, log-rank test. Data are means ± SEM, *P = 0.0477. (D to F) Total protein (D), MPO activity (E), and perfused dextran (F) were quantified in BALF on day 0 (uninjured, n = 3 mice), day 10 (n = 4 mice), day 25 (n = 5 mice), and day 35 (n = 5 mice) after influenza infection. Data are means ± SEM, unpaired two-tailed t test, *P < 0.05 and **P < 0.01. WT versus Tgfbr2^ECKO in (D): D20, P = 0.009; D35, P = 0.047. WT versus Tgfbr2^ECKO in (E): D20, P = 0.028. WT versus Tgfbr2^ECKO in (F): D20, P = 0.032. (G and H) Intracellular flow cytometry quantification of proliferative ECs (CD45⁻/EpCAM⁻/CD31⁺/Ki67⁺) at day 15 after influenza. WT, n = 6; Tgfbr2^ECKO, n = 8. Data are means ± SEM, unpaired two-tailed t test, *P = 0.0412. (I) qPCR analysis of Tgfbr2 and Mki67 in isolated lung ECs (CD45⁻CD31⁺) from WT and Tgfbr2^ECKO mice sorted 20 days after influenza infection. Data are means ± SEM (n = 5), unpaired two-tailed t test, *P < 0.05 and ****P < 0.0001. Tgfbr2, P < 0.0001; Mki67, P = 0.036. (J) Left: Tile scan images of H&E stain at 27 days after infection; demarcated boxes indicate different injury zones. Right: Clustered injury zone maps produced from left H&E images. Scale bars, 1 mm. (K) Zoomed-in images from the demarcated boxed area in (J). Scale bars, 50 μm. (L) Quantification of injured area in different injury zones in (J). Data are means ± SEM (n = 5), unpaired two-tailed t test, *P < 0.05. Total injured zone, P = 0.024; damaged zone, P = 0.039. (M) Tile scan images of immunostaining of vascular endothelial cadherin (VECAD) at day 27 after infection. Scale bars, 1 mm. (N) Images of VECAD staining in different injury zones in (M). Scale bars, 25 μm. (O) Quantification of vessel percentage judged by VECAD staining in different injury zones in (M). Data are means ± SEM (n = 5), unpaired two-tailed t test, *P < 0.05. Total injured zone, P = 0.08; severe zone, P = 0.011.

Fig. 3.. Endothelial Tgfbr2 deletion in vivo impairs aerocyte regeneration after influenza injury.

(A) Representative gating scheme for identification of CAR4-expressing aerocyte ECs at day 35 after influenza infection. SSC-A, side scatter area. (B) Intracellular flow cytometry quantification of Car4-expressing aerocyte ECs at day 0 (uninjured; WT, n = 4; Tgfbr2^ECKO, n = 4), day 20 (WT, n = 5; Tgfbr2^ECKO, n = 5), and day 35 (WT, n = 6; Tgfbr2^ECKO, n = 5) after influenza infection. Data are means ± SEM, unpaired two-tailed t test, *P < 0.05. D20, P = 0.033; D35, P = 0.026. (C) The percentage of lung ECs was compared between WT and Tgfbr2^ECKO mice at day 0 (uninjured; WT, n = 3; Tgfbr2^ECKO, n = 3), day 20 (WT, n = 5; Tgfbr2^ECKO, n = 5), and day 35 (WT, n = 6; Tgfbr2^ECKO, n = 5) after influenza infection. Data are means ± SEM, unpaired two-tailed t test, *P = 0.037. (D) Representative immunostaining of CAR4-expressing aerocytes in WT and Tgfbr2^ECKO mice on day 20 after infection. Scale bars, 100 and 25 μm (inset). (E) qPCR analysis of aerocyte genes Car4, Apln and Tbx2 in isolated total lung ECs (CD45⁻CD31⁺) from WT and Tgfbr2^ECKO mice sorted on day 20 after influenza infection. WT, n = 5; Tgfbr2^ECKO, n = 5. Data are means ± SEM, unpaired two-tailed t test, *P < 0.05, **P < 0.01, and ***P < 0.001. Vegfa, P = 0.00014; Car4, P = 0.009; Apln, P = 0.0009; Tbx2, P = 0.024. (F) Feature plots showing Car4 and Ednrb expression in aerocyte ECs. (G) Representative gating scheme for identification of EDNRB-expressing ECs at day 35 after influenza infection. (H) Intracellular flow cytometry quantification of EDNRB-expressing ECs at day 35 (WT, n = 6; Tgfbr2^ECKO, n = 5) after influenza infection. Data are means ± SEM, unpaired two-tailed t test, *P = 0.026.

Fig. 4.. Dose-dependent effects of LRG1 and TGF-β1 regulate angiogenic proliferation.

(A) Violin plots of Lrg1 mRNA expression (log-normalized) in mouse lung ECs on day 0 (D0), day 20, and day 30 after influenza infection. (B) qPCR analysis of Lrg1 in isolated lung ECs (CD45⁻CD31⁺) sorted on days 0 (uninjured), 10, 19, and 27 after influenza infection. n = 3 or 4 per group; data are means ± SD, ANOVA, followed by Dunnett’s multiple comparison test. *P < 0.05 and ****P < 0.0001 (D0 versus D10, P < 0.0001; D0 versus D19, P = 0.032). (C and D) The concentrations of active TGF-β1 (solid black line) and LRG1 (dashed red line) in BALF (C) and peripheral blood serum (D) were measured by enzyme-linked immunosorbent assay at days 0 (uninjured), 15, 20, and 30 after influenza infection. n = 3 or 4 per group; data are presented as means ± SEM, ANOVA, followed by Dunnett’s multiple comparison test. *P < 0.05, **P < 0.01, and ***P < 0.001. TGF-β1 in BALF: D30 versus D0, P = 0.009; TGF-β1 in serum: D30 versus D0, P < 0.0001; LRG1 in BALF: D15 versus D0, P < 0.0001; D20 versus D0, P = 0.004; LRG1 in serum: D15 versus D0, P = 0.0149. (E) C57BL6/J mice were treated with LuLNPs encapsulating control noncoding siRNA [si–negative control (NC), 1 mg/kg] or Lrg1 siRNA (si-Lrg1; 1 mg/kg) by tail-vein injection on days 15 and 18 after infection, and EdU (50 mg/kg) was administrated 24 hours before analysis. i.v., intravenously. (F) qPCR analysis of Lrg1 in sorted ECs from mice on day 19 after infection. Data are means ± SEM (n = 4), unpaired two-tailed t test, *P = 0.0218. (G) Quantification of SMAD1/5 and SMAD3 phosphorylation in ECs (ERG⁺pSMAD1/5⁺ or ERG⁺pSMAD3⁺) in mouse lungs from (E), representative images in fig. S9 (C and D). Data are means ± SEM (n = 4), unpaired two-tailed t test, pSMAD1/5: ***P < 0.0001. (H) Intracellular flow analysis of proliferative ECs (EdU⁺) in mouse lungs from (E). Data are means ± SEM (n = 4), unpaired two-tailed t test, *P = 0.0429. (I) Immunoblotting analysis of indicated proteins in primary human lung ECs treated ± LRG1 (5 μg/ml) with or without low (TGF-β1^lo, 3 ng/ml) or high (TGF-β1^hi, 10 ng/ml) concentrations of TGF-β1 for 1 hour; values were quantified by densitometry, normalized to β-actin. (J) Quantification of indicated proteins in (I) of cells treated ± LRG1 with or without low concentration TGF-β1 (TGF-β1^lo). (K) Quantification of indicated proteins in (I) when cells were treated ± LRG1 combined with or without high concentration of TGF-β1 (TGF-β1^hi). Data are means ± SEM (n = 3), unpaired two-tailed t test, *P < 0.05 and **P < 0.01; pSMAD1/5 in (J) LRG1 versus TGF-β1 + LRG1, P = 0.034; pSMAD2/3 in (J) PBS versus TGF-β1, P = 0.0024; pSMAD1/5 in (K) PBS versus TGF-β1, P = 0.038; pSMAD2/3 in (K) PBS versus TGF-β1, P = 0.0008, LRG1 versus TGF-β1 + LRG1, P = 0.0046. (L and M) Cell proliferation of primary human lung ECs treated ± LRG1 (5 μg/ml) with or without low (3 ng/ml) or high (10 ng/ml) concentrations of TGF-β1 for 6 hours, as assessed by EdU incorporation assay. (L) Representative immunofluorescence for nuclei (blue) and EdU incorporation (red). (M) Quantification of percentage of proliferating cells (EdU⁺/DAPI) in (H). Data are means ± SEM (n = 3), ANOVA, followed by Dunnett’s multiple comparison test. *P < 0.05 and **P < 0.01; control versus TGF-β1 (10 ng/ml), P = 0.013; TGF-β1 (3 ng/ml) versus TGF-β1 (3 ng/ml) + LRG1, P = 0.0085; control versus TGF-β1 (3 ng/ml) + LRG1, P = 0.029; TGF-β1 (3 ng/ml) + LRG1 versus TGF-β1 (10 ng/ml) + LRG1, P = 0.011.

Fig. 5.. TGF-βR2 signaling induces autocrine Vegfa expression through SMAD activation.

(A) qPCR analysis of Vegfa in isolated lung ECs (CD45⁻ CD31⁺) from uninjured WT and Tgfbr2^ECKO mice. Data are means ± SD (n = 3), unpaired two-tailed t test, *P = 0.044. (B) qPCR analysis showing the expression of Vegfa in WT and TGFBR2-KO iMVECs treated ± TGF-β1 (10 ng/ml) for 24 hours. Data are means ± SD (n = 4), unpaired two-tailed t test, *P < 0.05. WT: PBS versus TGF-β1, P = 0.014; TGF-β1: WT versus KO, P = 0.005. (C) Timeline for generating blood vessel organoids from hiPSCs and ITD-1 treatment. For vascular network analysis, ITD-1 (10 μM) was added after vascular colony embedding into collagen I–Matrigel solution (day 7) for 3 days, with medium changed every day. For vascular organoid formation analysis, ITD-1 (10 μM) was added on day 10 when single vascular organoids were isolated from the three-dimensional (3D) matrix and then harvested on day 15, with one medium change at day 13. (D) Images of hiPSC differentiation into vascular networks and blood vessel organoids. (i) Cell aggregate formation at day 2 (D2). Scale bar, 100 μm. (ii) Cell aggregates differentiated into mesoderm (D5). Scale bar, 100 μm. (iii) Induction of mesoderm differentiation into vascular progenitor cells (D7), before embedding into the 3D collagen I–Matrigel matrix. Scale bar, 100 μm. (iv) Cell aggregates of early blood vessels grow outward in 3D collagen I–Matrigel matrix to form vascular networks (D10). Scale bar, 100 μm. (v) Higher power image of the vascular networks at day 10. Scale bar, 100 μm. (vi) Immunostaining of CD31 for ECs at day 10, CD31 (green) and nuclei (blue). Scale bar, 100 μm. (vii) Bright field of well-formed vascular organoids at day 15. Scale bar, 200 μm. (viii) Endothelial tubes (CD31, green) in vascular organoids covered by pericytes [platelet-derived growth factor receptor beta (PDGFRβ), red], nuclei (DAPI, blue). (ix) Immunostaining of CD31 showing the vessel tubes in vascular organoids. Scale bar, 50 μm. (x to xii) 3D reconstruction of capillary organization (CD31, green) in a vascular organoid covered by pericytes (PDGFRβ, red) at day 15. (x) Merged, (xi) CD31, and (xii) PDGFRβ. Scale bars, 200 μm. (E) Cell aggregates were embedded into 3D collagen I–Matrigel matrix and treated with ITD-1 (10 μg/ml), combined with or without VEGFA (100 ng/ml), or vehicle (dimethyl sulfoxide). Vascular network density was analyzed on day 3 by immunostaining of CD31 and then quantification of vessels percentage area. Scale bars, 100 μm. Data are presented as means ± SEM (n = 3), ANOVA, followed by Dunnett’s multiple comparison test, *P < 0.05. Vehicle versus ITD-1, P = 0.029. (F) ITD-1 (10 μM) combined with or without VEGFA (100 ng/ml) was added on day 10 when single vascular organoid was isolated from 3D matrix and then cultured in ultralow attachment cell culture plates for 5 days. Networks successfully assembled into vascular organoids with the round, smooth, and well-demarcated border. Left: The representative images of fully formed vascular organoids in vehicle group and failed vascular organoids after ITD-1 treatment. Right: Illustration depicting the processing of vascular organoids in the left images. (G) The vascular organoid formation efficiency was assessed by the proportions of mature vascular organoids [VOs; as shown in (F)]. Scale bars, 200 μm. Data are presented as means ± SD (n = 3), ANOVA, followed by Dunnett’s multiple comparison test, ***P < 0.001. Vehicle versus ITD-1, P < 0.0001; ITD-1 versus ITD-1 + VEGFA, P < 0.0001. (H) qPCR analysis of TGFBR2 in primary human lung ECs after transfection with si-TGFBR2 (5 nM) or si-NC (5 nM) for 48 hours. Data are means ± SD (n = 3), unpaired two-tailed t test, ***P = 0.0002. (I) Tube formation assays of primary human lung ECs were performed 48 hours after transfection with si-TGFBR2 (5 nM) or si-NC (5 nM). Left: Representative images of tube networks after ECs were treated with or without VEGFA for 6 hours (20 ng/ml). Right: Tube networks were quantified by counting the average rings/tubes per field under a light microscope at ×100 magnification. Dashed circles represent vascular rings. Scale bars, 100 μm. Data are means ± SEM (n = 3), unpaired two-tailed t test, *P < 0.05, **P < 0.01, and ***P < 0.001. si-NC: PBS versusVEGFA, P = 0.023; si-TGFBR2: PBS versus VEGFA, P = 0.015; PBS: si-NC versus si-TGFBR2, P = 0.00015.

Fig. 6.. LuLNP-mediated mRNA delivery to lung ECs.

(A) C57BL6/J mice were treated with LuLNPs encapsulating luciferase mRNA (Luc-LuLNP; 0.2 mg/kg) or empty control LuLNPs (Ctrl-LuLNP) by tail-vein injection 12 hours before analysis. Transfection efficiency was detected by IVIS. IVIS imaging of luciferase mRNA delivery to the lung (left) and heart, liver, spleen, and kidneys were dissected for luminescence imaging (right), n = 3 to 5 mice per group. (B) Mice were treated with Luc-LuLNP (0.2 mg/kg) by tail-vein injection and imaged by IVIS. Quantification of luciferase signal at 6, 24, 48, and 72 hours after injection. Data are means ± SEM (n = 4). (C) C57BL6/J mice were administrated LuLNPs encapsulating GFP mRNA (GFP LuLNP; 0.5 mg/kg) or equal volume of PBS, and liver enzymes ALT and AST were quantified 12 hours after injection. Data are means ± SEM (n = 5). (D) Analysis of GFP⁺ cells 18 hours after GFP-LuLNP or Ctrl-LuLNP administration in C57BL6/J mice. (E) Representative gating scheme for identification of GFP⁺ ECs (CD45⁻/CD31⁺/GFP⁺). (F) The proportion of GFP⁺ cells in the lung by cell type, including immune cells, ECs, epithelial cells and others (mesenchymal). Data are means ± SD (n = 3). (G) Distribution of total GFP⁺ cells in each cell type. n = 3 mice. (H) Immunostaining showing that GFP⁺ cells colocalize with EC marker PECAM1. Scale bar, 25 μm. (I) lsl-Ai14-tdTomato (R26-lsl; tdTomato) mice were administrated LuLNPs encapsulating Cre mRNA (Cre-LuLNP) or equal volume of vehicle empty LuLNPs (control) 1 week before analysis. (J) The proportion of LuLNP-Cre–traced cells (tdTomato⁺) in the lung by cell type, including immune cells, ECs, epithelial cells, and mesenchymal (“other”). Data are means ± SD (n = 3). (K) Immunostaining showing that LuLNP-Cre–traced cells (tdTomato⁺) cells colocalize with EC marker PECAM1. Scale bar, 25 μm.

Fig. 7.. LuLNP delivery of Vegfa mRNA alleviates the exacerbation of influenza injury caused by endothelial Tgfbr2 deficiency.

(A) Timeline for LuLNP administration and sampling. WT and Tgfbr2^ECKO mice were treated with empty LuLNPs (Ctrl LuLNP) or LuLNPs encapsulating Vegfa mRNA (Vegfa LuLNP) (0.5 mg/kg) on day 15 after infection, and samples were collected after 72 hours. Dexamethasone-21-phosphate (DEX) was injected i.p. (2 mg/kg) into the mice 30 min before LuLNP injection in all mice. (B) Intracellular flow cytometry quantification of proliferative ECs (CD31⁺/Ki67⁺) 72 hours after administration of control or Vegfa LuLNPs in lung ECs from WT and Tgfbr2^ECKO mice. Data are means ± SEM (n = 5), unpaired two-tailed t test, *P < 0.05. Tgfbr2^ECKO: Ctrl LuLNP versus Vegfa LuLNP, P = 0.048. (C) Representative immunostaining of proliferative ECs 72 hours after administration of control or Vegfa LuLNP in Tgfbr2^ECKO mice lungs. Scale bars, 100 μm. (D and E) Quantification of apoptotic (TUNEL⁺) ECs (D) and total apoptotic cells (E). n = 3 to 5 per group. Data are means ± SEM, unpaired two-tailed t test, *P < 0.05 and ***P < 0.001. (D) WT: Ctrl LuLNP versus Vegfa LuLNP, P = 0.018; Tgfbr2^ECKO: Ctrl LuLNP versus Vegfa LuLNP, P = 0.0002. (E) WT: Ctrl LuLNP versus Vegfa LuLNP, P = 0.06; Tgfbr2^ECKO: Ctrl LuLNP versus Vegfa LuLNP, P = 0.02. (F) Timeline for indicated LuLNP administration and sampling. WT and Tgfbr2^ECKO mice were treated with Ctrl LuLNP or Vegfa LuLNP (0.5 mg/kg) on days 15 and 21 after infection, and lungs were harvested on day 27. DEX was injected i.p. (2 mg/kg) into the mice 30 min before LuLNPs injection in all mice. (G and H) Body weight (G) and capillary oxygen saturation (H) for LNP-treated Tgfbr2^ECKO mice. Data are means ± SD (n = 8), unpaired two-tailed t test, *P < 0.05 and **P < 0.01. (G) D21: Ctrl LuLNP versus Vegfa LuLNP, P = 0.0013; D25: Ctrl LuLNP versus Vegfa LuLNP, P = 0.04. (H) D25: Ctrl LuLNP versus Vegfa LuLNP, P = 0.011. (I and J) Total protein (I) and MPO activity (J) were quantified in BALF collected from Tgfbr2^ECKO mice that received Ctrl LuLNP or Vegfa LuLNP treatment and harvested at day 27 after infection. Data are means ± SEM (n = 5), unpaired two-tailed t test, *P < 0.05 and **P < 0.01. (I) P = 0.008; (J) P = 0.023. (K to P) Tgfbr2^ECKO mice received Ctrl LuLNP or Vegfa LuLNP treatment, and lung samples were harvested on day 27 after infection. (K) Left: Tile scan images of H&E stain; demarcated boxes indicate different injury zones. Right: Clustered injury zone maps produced from left H&E images. Scale bars, 1 mm. (L) Zoomed-in images from the demarcated boxes area in (K). Scale bars, 50 μm. (M) Quantification of injury area in different injury zones in (K). (N) Tile scan images of immunostaining of vascular endothelial cadherin (VECAD). Scale bars, 1 mm. (O) Images of VECAD staining in different injury zones in (N). Scale bars, 25 μm. (P) Quantification of vessel percentage judged by VECAD staining in different injury zones in (N). Data are means ± SEM (n = 5), unpaired two-tailed t test, *P < 0.05. (M) Ctrl LuLNP versus Vegfa LuLNP: total injured zone, P = 0.023; damaged zone, P = 0.022. (P) Ctrl LuLNP versus Vegfa LuLNP: damaged zone, P = 0.03.