Loss of growth differentiation factor 15 exacerbates lung injury in neonatal mice

Abstract

Growth differentiation factor 15 (GDF15) is a divergent member of the transforming growth factor-β (TGF-β) superfamily, and its expression increases under various stress conditions, including inflammation, hyperoxia, and senescence. GDF15 expression is increased in neonatal murine bronchopulmonary dysplasia (BPD) models, and GDF15 loss exacerbates oxidative stress and decreases cellular viability in vitro. Our overall hypothesis is that the loss of GDF15 will exacerbate hyperoxic lung injury in the neonatal lung in vivo. We exposed neonatal Gdf15^−/− mice and wild-type (WT) controls on a similar background to room air or hyperoxia (95% $\text{F}{\mathrm{I}}_{{\text{O}}_2}$) for 5 days after birth. The mice were euthanized on postnatal day 21 (PND 21). Gdf15^−/− mice had higher mortality and lower body weight than WT mice after exposure to hyperoxia. Hyperoxia exposure adversely impacted alveolarization and lung vascular development, with a greater impact in Gdf15^−/− mice. Interestingly, Gdf15^−/− mice showed lower macrophage count in the lungs compared with WT mice both under room air and after exposure to hyperoxia. Analysis of the lung transcriptome revealed marked divergence in gene expression and enriched biological pathways in WT and Gdf15^−/− mice and differed markedly by biological sex. Notably, pathways related to macrophage activation and myeloid cell homeostasis were negatively enriched in Gdf15^−/− mice. Loss of Gdf15 exacerbates mortality, lung injury, and the phenotype of the arrest of alveolarization in the developing lung with loss of female-sex advantage in Gdf15^−/− mice.

NEW & NOTEWORTHY We show for the first time that loss of Gdf15 exacerbates mortality, lung injury, and the phenotype of the arrest of alveolarization in the developing lung with loss of female-sex advantage in Gdf15^−/− mice. We also highlight the distinct pulmonary transcriptomic response in the Gdf15^−/− lung including pathways related to macrophage recruitment and activation.

INTRODUCTION

Bronchopulmonary dysplasia (BPD) is one of the most common causes of morbidity among surviving premature infants. Despite the decreased mortality rate among preterm neonates due to advances in neonatal care, the incidence of BPD among this vulnerable population has stayed the same, leading to an increased disease burden (1). The key pathological feature in “New BPD” is the arrest in lung development, characterized by aberrant vascular development and abnormal alveolar septation (2). BPD also leads to long-term morbidities and may predispose to the early development of chronic lung diseases in adulthood (3).

Growth differentiation factor 15 (GDF15) is a divergent member of transforming growth factor-β (TGF-β) superfamily cytokines. It is also known as macrophage inhibitory cytokine-1 (MIC-1) (4), nonsteroidal anti-inflammatory drug-activated gene-1 (5), and placental transforming growth factor-β (PTGFβ) (6). GDF15 expression increases during pregnancy, and under pathological and stress-related conditions such as hypoxia, inflammation, hyperoxia, infection, senescence, and intense exercise (7, 8). GDF15 plays a central role in the coordination of tolerance to inflammation by initiating metabolic adaptation (9), including triglyceride metabolism and energy homeostasis (10).

In adult literature, GDF15 has been extensively studied and has been associated with multiple cardiopulmonary disorders such as chronic obstructive pulmonary disease (COPD), idiopathic pulmonary fibrosis (IPF), pulmonary viral infections, and pulmonary hypertension (11–14). GDF15 has been identified as a potential biomarker, prognostic factor, and a possible therapeutic target (8, 15). GDF15 was upregulated in human IPF lungs (16) and was identified as mediating the association between age, interstitial lung abnormality, and mortality in two large prospective cohorts: Framingham Heart Study and the COPD gene cohort (17).

Our previous studies in vivo have shown increased pulmonary GDF15 expression in mice when exposed to hyperoxia (18). Hyperoxia also induces GDF15 expression in pulmonary endothelial and epithelial cells, and GDF15 loss increases oxidative stress and decreases cellular viability in vitro (19). Whether increased serum concentrations indicate ongoing cellular injury or represent a protective response to biological stress is still an open question, and the answer might depend on the organ and cellular environment. Gdf15 is a part of the in vivo gene expression signature of oxidative stress (20). GDF15 has been shown to have anti-inflammatory (21–23), proangiogenic (24), and antiapoptotic (25) effects. Whether the loss of GDF15 modulates the response of the neonatal murine lung to postnatal hyperoxia is not known. Using global knockouts for the Gdf15^−/− gene, we hypothesized that loss of GDF15 will exacerbate neonatal hyperoxic lung injury with reduced survival and will adversely impact alveolar and pulmonary vascular development.

METHODS

Animals

The study protocol was approved by the Institutional Animal Care and Use Committee (IACUC) at Baylor College of Medicine. Wild-type mice (C57BL/6) were obtained from Charles River Laboratories (Wilmington, DE). Gdf15^−/− mice were obtained from Dr. Se-Jin Lee from Johns Hopkins University (Baltimore). Active colonies of mice were maintained by breeding in our animal facility. Newborn mice were randomly allocated, wild-type mice and Gdf15^−/−, to the normoxia or hyperoxia groups within 12 h of birth.

Mouse Model of BPD

Mouse pups were exposed to hyperoxia (95% $\text{F}{\mathrm{I}}_{{\text{O}}_2}$) within 12 h of birth up to postnatal day 5 (PND 5) to induce arrest in alveolarization with the goal to replicate the lung phenotype seen in human infants with BPD. Mice are born at the saccular stage of lung development, which corresponds to 26–36 wk of fetal lung development in humans (26). Exposing neonatal mice to hyperoxia in the first 5 days and then transitioning them to room air mimics the majority of preterm human neonates’ clinical course in the neonatal intensive care unit (27, 28), albeit most of the preterm neonates do not need 95% oxygen until 36 wk postmenstrual age, which is a drawback of this model. The mouse pups were randomly assigned to a normoxia group (21% oxygen) and a hyperoxia group (95% oxygen). The litter sizes were limited to six pups in all groups to avoid overcrowding and limit its effects on nutrition and growth. The dams were rotated daily between room air- and hyperoxia-exposed litters to prevent oxygen toxicity and eliminate the maternal effects among different groups. Hyperoxia was achieved using Plexiglass chambers, and the oxygen analyzer was used to measure the $\text{F}{\mathrm{I}}_{{\text{O}}_2}$ in the chambers. The hyperoxia-exposed mice were transitioned to normoxia to recover in room air till PND 21. All mice were euthanized at PND 5 (immediately after hyperoxia exposure) and PND 21 (after recovery from normoxia). A schematic illustration of the experiment design is shown in Fig. 1. Lung weight/body weight ratios were measured at euthanasia on PND 5. Body weights were measured for all the surviving mice at PND 21.

Figure 1.

Does loss of Gdf15 worsen arrest in alveolarization and pulmonary vascular development neonatal hyperoxic lung injury? Schematic showing exposure of WT and Gdf15^−/− neonatal mice to hyperoxia (95% $\text{F}{\mathrm{I}}_{{\text{O}}_2}$, PND 1–5) during the saccular stage of lung development. Euthanasia was performed at PND 21 with evaluation of alveolarization, pulmonary vascular development, and assessment of the pulmonary transcriptome. $\text{F}{\mathrm{I}}_{{\text{O}}_2}$, fraction of inspired oxygen; Gdf15, growth differentiation factor 15; PND, postnatal day; WT, wild type. [Image created with BioRender.com and published with permission.]

Lung Histology and Morphometry

During euthanasia on PND 21, the trachea was cannulated, and the lungs were inflated with 4% paraformaldehyde administered endotracheally at 25 cmH₂O pressure for 15 min. The lungs were collected, and 5 µm sections were obtained and stained with hematoxylin-eosin stain. The slides were examined under a light microscope. Mean linear intercept (MLI) and radial alveolar count (RAC) were used to measure the alveolar development at PND 21 using methods previously described (28). We used 10–15 randomly nonoverlapping fields per biological replicate for analysis. Fields with large airways and vessels were excluded.

Pulmonary vascular development.

The pulmonary vessels were identified using immunofluorescence for von Willebrand factor (vWF), an endothelial cell-specific marker (1:4,000 dilution, Abcam; Cat. No. ab6994). Pulmonary vessel density was measured by counting the stained vessels with external diameters of <100 µm (×40 magnification) per lung field. Ten random nonoverlapping fields per specimen were used for analysis. Fields with large airways and vessels were excluded.

Analysis of inflammation.

To measure the degree of inflammation, we quantified the macrophages per high power field in the distal lung by immunofluorescent staining using F4/80 antibody, a specific antibody for macrophages (1:500 dilution, Bio-Rad Laboratories: Cat. No. MCA497GA). Fifteen random nonoverlapping fields per biological replicate were used for analysis. Fields with large airways and vessels were excluded.

Lung mRNA Extraction and RNA-Seq Analysis

Total RNA from flash-frozen lung samples was isolated using the Zymo micro kit (Zymo Research, Irvine, CA). RNA concentration and quality check were assayed using Nanodrop-8000 (Thermo Scientific, Wilmington, DE). RNA quality parameters were as follows: the 260/280 and 260/230 ratios needed to be >1.9. Furthermore, the RNA integrity number (RIN) was analyzed using the Agilent Bioanalyzer. The samples needed to have RIN values of 7–10 and with a range of 1–1.5. Sequencing was done on the Illumina NovaSeq S4 PE150. Salmon (version 1.9.0) (29) was used to build an index based on the GRCm39 reference transcriptome. It was also used to quantify using—seqBias and—gcBias flags to correct for sequence and fragment-level biases. Tximeta (version 1.16.1) (30) was used to load our quantified data along with our metadata. We perform a prefiltering step to keep only genes that have at least 10 reads present. We then use DESeq2 (version 1.38.3) (31) to perform differential expression analysis. We filtered these results using the results() function from DESeq2 with an α value of 0.05. We only kept genes with less than 0.05 adjusted P value and greater than 0.3219 absolute log2 fold change. We then mapped the resulting ensemble IDs to gene symbols using biomaRt (version 2.54.0) (32) and performed Shrinkage of effect size using DESeq2 lfcShrink() function for better gene ranking and visualization. Fgsea (version 1.24.0) (33) was used for gene set enrichment analysis using the stat value for ranking. We filtered these results to get pathways with a P value less or equal to 0.05 and a gene number of more than 5. Other programs used for sorting, analyzing, and exporting data were Org.Mm.eg.db (version 3.16.0), ggvenn (version 0.1.9), VennDiagram (version 1.7.3), tibble (version 3.1.8), ggplot2 (version 3.4.1), tidyverse (version 2.0.0), and openxlsx (version4.2.5). All the programs, versions, and references are also provided in Supplemental Table S1.

Data Availability

All code used in this manuscript can be found in Supplemental Material; the raw data have been uploaded to National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO); GSE211744. All supplemental data are accessible at https://doi.org/10.6084/m9.figshare.22771034.

Western Blotting and qRT-PCR

Lung tissue from WT and Gdf15^−/− mice exposed to room air and hyperoxia as described before was obtained after euthanasia at PND 7. Lung whole protein (20 μg of protein) was prepared as described before and subjected to SDS polyacrylamide gel electrophoresis in 10% acrylamide gels and then transferred to polyvinylidene difluoride membranes, followed by Western blotting. After the membranes were blocked in 5% nonfat dry milk, they were incubated overnight with primary antibodies after which, the membranes were washed and incubated with the appropriate secondary antibodies. Vinculin was used as the loading control. The primary antibodies were GDF15 (Santa Cruz; sc-515675,1:1,200 dilution) and vinculin (Cell Signaling; Cat No. 4650, 1:1,200 dilution). This was followed by electrochemical detection of bands. Band intensities were quantified using Bio-Rad ImageLab Software. Total RNA from lung samples in mice exposed to room air or hyperoxia was isolated on PND 21 using the mRNeasy kit (Qiagen, Valencia, CA). RNA concentration was assayed using a Nanodrop-8000 (Thermo Scientific, Wilmington, DE). NanoDrop spectrophotometer was used to measure RNA concentration and quality and subjected to one-step real-time quantitative TaqMan RT-PCR using 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA). Gene-specific primers were purchased from life science technologies in the presence of TaqMan reverse transcription reagents and RT reaction mix (Applied Biosystems, Foster City, CA) were used to reverse transcribe RNA, and TaqMan Gene Expression probes and TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA), were used for PCR amplification. Furthermore, 18S was used as the reference gene. The primers used were Ucp1 (Mm01244861_m1), Gdf15 (Mm00442228_m1), Siglec f (Mm00523987_m1), Spp1 (Mm00436767_m1), and 18 s (Hs99999901_s1).

Statistical Analyses

Data analysis was performed using GraphPad Prism 9 software. Survival analysis was performed using the Log-rank (Mantel–Cox) test. Data are displayed as means ± SD. Two-way ANOVA following Tukey’s posttest was performed for statistical evaluation. The main effects of hyperoxia, genotype, and interaction between the independent variables were also assessed. In other analyses, three-way ANOVA was used to assess the main effects of sex, genotype, and treatment, and the interaction among these variables. The significance level was set at P < 0.05. We calculated the sample size based on our preliminary results for hyperoxia-induced alveolar simplification in untreated WT mice. A sample of six mice per group will be needed to detect a 25% increase (two SD) in the MLI in Gdf15^−/− mice to provide 80% power with an α of 0.05.

RESULTS

Gdf15^−/− Mice Had Higher Mortality and Lower Body Weight than WT When Exposed to Hyperoxia

Gdf15 expression was increased in the hyperoxia-exposed WT lung both at the mRNA and the protein level after hyperoxia exposure (95% $\text{F}{\mathrm{I}}_{{\text{O}}_2}$ till PND 5) (Supplemental Fig. S1). At the time of euthanasia (PND 21), the body weights and lung weights of the surviving mice were measured. Within the Gdf15^−/− mice group, hyperoxia-exposed mice had significantly lower body weight than room air-exposed mice. Between the two genotypes, upon exposure to hyperoxia, Gdf15^−/− mice had lower body weight than the WT mice. There were no significant differences between the two genotypes at baseline under normoxia. As an independent variable, treatment (hyperoxia or normoxia) was statistically significant, whereas genotype and the interaction term were not (Fig. 2A). Lung weight/body weight ratio (LW/BW) was measured as an indicator of lung injury. LW/BW ratio was increased in both WT and Gdf15^−/− mice groups upon exposure to hyperoxia with no differences between the genotypes (Fig. 2B). When analyzed by biological sex, hyperoxia-exposed WT male mice and Gdf15^−/− male and female mice showed a significant increase in LW/BW ratio, whereas the WT female mice did not (Fig. 2C). Figure 2D shows the survival curves in WT and Gdf15^−/− mice. Under normoxia, WT (n = 32 mice) and Gdf15^−/− mice (n = 42 mice) had no significant differences in the survival rates (100% vs. 98%, respectively). Upon exposure to hyperoxia, the survival in Gdf15^−/− mice was significantly decreased compared with WT mice, 26% versus 57%, respectively. Survival data by sex was not collected.

Figure 2.

Loss of Gdf15 decreases survival and increases lung injury. A: body weights from room air- and hyperoxia (95% oxygen, PND 1–5)-exposed WT and Gdf15^−/− mice at PND 21; n = 11–15/group. B: lung weight/body weight ratios from room air- and hyperoxia-exposed WT and Gdf15^−/− mice at PND 21; n = 10–23/group. C: lung weight/body weight ratios from room air- and hyperoxia-exposed WT and Gdf15^−/− male and female mice at PND 21; n = 4–12/group. Statistically significant differences between the indicated groups shown by *P < 0.05; #P < 0.05, **P < 0.01 and ***P < 0.001. Statistical analysis performed using two-way or three-way ANOVA. D: survival curves of WT and Gdf15^−/− neonatal mice exposed to room air or hyperoxia. WT-room air group n = 32, Gdf15^−/− room air group: n = 42, WT hyperoxia group n = 42, Gdf15^−/− hyperoxia group n = 54. Values are presented as means ± SD. Survival differed significantly between the four groups, P < 0.0001, using the log-rank test. Gdf15, growth differentiation factor 15; PND, postnatal day; WT, wild type.

Upon Exposure to Hyperoxia, Gdf15^−/− Female Mice Had Greater Alveolar Simplification than WT

Lung morphometry was assessed at PND 21, after a period of hyperoxia exposure (95% oxygen) during PND1 5 (saccular stage of lung development). We measured the mean linear intercept (MLI) and radial alveolar count (RAC) to quantify the alveolar simplification as per the techniques described in the methodology. Representative lung sections are shown in Fig. 3, using ×10 and ×20 magnifications. The MLI and RAC were significantly altered upon exposure to hyperoxia in WT and Gdf15^−/− mice, however, the effect was greater in Gdf15^−/− mice (Fig. 3C). When stratified by biological sex, alveolar simplification was attenuated in WT females compared with WT males and Gdf15^−/− female mice. There was no difference between Gdf15^−/− male and female mice (Fig. 3, E and F).

Figure 3.

Loss of Gdf15 worsens arrest in alveolarization after hyperoxia exposure during the saccular stage of lung development. Representative hematoxylin and eosin-stained lung sections at ×10 (A) and ×20 (B) magnification from room air and hyperoxia (95% oxygen, PND 1–5)-exposed WT and Gdf15^−/− mice euthanized at PND 21. C and D: lung morphometry assessed by mean linear intercept (MLI) and radial alveolar count from room air and hyperoxia (95% oxygen, PND 1–5)-exposed WT and Gdf15^−/− mice at PND 21; n = 8–17/group. E and F: lung morphometry in male and female neonatal mice assessed by mean linear intercept (MLI) and radial alveolar count from room air and hyperoxia (95% oxygen, PND 1–5)-exposed WT and Gdf15^−/− mice at PND 21; n = 3–11/group. Values are presented as means ± SD. Statistically significant differences between the indicated groups shown by #P < 0.05, ##P < 0.01, and ***P < 0.001. Statistical analysis performed using two-way or three-way ANOVA. Gdf15, growth differentiation factor 15; PND, postnatal day; WT, wild type.

Gdf15^−/− and WT Mice Showed Greater Arrest in Angiogenesis upon Exposure to Hyperoxia than Mice Kept on Room Air, but with No Significant Differences between the Two Genotypes

We sought to determine whether the loss of Gdf15 affected vascular development in response to hyperoxia. Pulmonary vascular density was decreased within both genotypes upon exposure to hyperoxia, Gdf15^−/− and WT mice. These results are shown in Fig. 4. The degree of hyperoxia-induced arrest in angiogenesis was not significantly different between the two genotypes (P > 0.05). Interestingly, at baseline under normoxia, Gdf15^−/− mice had lower vascular density compared with WT mice.

Figure 4.

Loss of Gdf15 worsens arrest in pulmonary vascular development after hyperoxia exposure during the saccular stage of lung development. A: representative immunohistochemistry image for pulmonary vascular density using immunohistochemistry for von Willebrand factor (vWF). Arrows point to the pulmonary blood vessels. Scale bar = 100 µm. B: quantitative analysis of the pulmonary vascular density. Green dots: Room Air; Red dots: Hyperoxia. n = 8–17/group (number of vWF-stained vessels per high power field). C: pulmonary vascular density in male and female WT and Gdf15^−/− mice exposed to room air or hyperoxia (n = 3/11 per group). Values are presented as means ± SD. Statistically significant differences between the indicated groups shown by *, #P < 0.05, and ***P < 0.001. Statistical analysis performed using two-way or three-way ANOVA. Gdf15, growth differentiation factor 15.

Gdf15^−/− Showed Lower Macrophage Recruitment in Response to Hyperoxia Compared with WT Mice

We next explored if the loss of Gdf15 altered the inflammatory response induced by hyperoxia. We assessed lung macrophages by immunohistochemistry (Fig. 5). Interestingly, under normoxic conditions, Gdf15^−/− mice had a lower macrophage count in the lung tissue compared with WT mice. Upon exposure to hyperoxia, Gdf15^−/− mice showed no increase in macrophage recruitment compared with the room air group. WT mice showed a significant increase in macrophage recruitment in response to hyperoxia compared with the room air group. Treatment, genotype, and interaction term were significant in the statistical analysis.

Figure 5.

Decreased lung macrophages following loss of Gdf15. A: representative lung sections from wild-type (WT) and Gdf15^−/− mice exposed to normoxia or hyperoxia (95% oxygen, postnatal days 1–5). Immunohistochemistry performed for macrophages using F4/80 antibody. Scale bar = 100 µm. B: quantitative analysis of macrophage recruitment (macrophage count per high power field). Values are presented as means ± SD. Statistically significant differences between the indicated groups shown by ###P < 0.001 and ***P < 0.001. Statistical analysis performed using two-way ANOVA. Gdf15, growth differentiation factor 15.

The Pulmonary Transcriptome Shows Marked Differences in the Gdf15^−/− Developing Lung upon Exposure to Hyperoxia Compared with WT Mice

We subjected lungs from room air- and hyperoxia-exposed WT and Gdf15^−/− mice at PND 21 to RNA-Seq analysis. The number of differentially expressed genes (DEGs) in each genotype (hyperoxia vs. room air) was analyzed. We then analyzed the overlap of the hyperoxia response between the WT and Gdf15^−/− mice. Among the upregulated genes, there were 42 genes (out of 475) and 23 downregulated genes (among 424) that were common between the two genotypes highlighting the marked differences in the transcriptomic response to hyperoxia (Fig. 6, A and B). The enriched biological pathways also had minor overlap with seven pathways common between WT and Gdf15^−/− mice (Fig. 6C). Volcano plots showing the DEGs in WT and Gdf15^−/− mice are shown in Fig. 6, D and E. The WT hyperoxia response has been described in previous publications (34). One of the upregulated genes in the Gdf15^−/− mice is Suv39h2, a histone lysine methyltransferase that plays a role in chromatin remodeling. Data from LungMAP (35) localize expression in the myeloid/macrophage population at PND 7, 14, and 28 (Supplemental Fig. S2). Increased Suv39h2 expression decreased trained immunity in monocytes (36). Endoglin (Eng) was one of the top downregulated genes in the Gdf15^−/− mice, which is expressed on macrophages (37), and mice lacking Eng in macrophages show an impaired immune response, including macrophage recruitment (38, 39). Representative biological pathways are unique to WT and Gdf15^−/− mice and pathways modulated in opposite directions between WT and Gdf15^−/− mice are highlighted in Fig. 6F. Circulatory system development and organ morphogenesis-related pathways were negatively enriched in the WT hyperoxia-exposed mice. Interestingly, macrophage activation and regulation of cellular response to stress were negatively enriched in Gdf15^−/− mice. Differentially expressed genes and biological pathways in all genotypes, biological sex, and treatment groups are provided in Supplemental Table S2.

Figure 6.

Distinct pulmonary transcriptomic differences in the Gdf15^−/− mice exposed to hyperoxia (95% oxygen, PND 1–5): Differentially expressed genes (DEGs) upregulated (A) and downregulated (B) in WT and Gdf15^−/− mice. Venn diagrams showing genes that are common to and distinct in WT and Gdf15^−/− mice. C: Venn diagrams showing biological pathways that are common to and distinct in WT and Gdf15^−/− mice. Volcano plots showing upregulated and downregulated genes in WT (D) and Gdf15^−/− (E) mice exposed to hyperoxia compared with room air controls. F: biological pathways enriched in WT and Gdf15^−/− mice exposed to hyperoxia compared with room air controls. Normalized enrichment score (NES) plotted on the x axis. Gdf15, growth differentiation factor 15; PND, postnatal day; WT, wild type.

Sex as a Biological Variable Modulates the Changes in the Pulmonary Transcriptome in Response to Hyperoxia and Is Distinct in WT and Gdf15^−/− Mice

We next wanted to analyze how biological sex modulates the transcriptomic response to hyperoxia in WT and Gdf15^−/− mice. The number of up- and downregulated DEGs in WT and Gdf15^−/− male and female mice are shown in Fig. 7A. In the WT mice, there was a 5% (20/404 DEGs) and 11.5% (50/433 DEGs) overlap between male and female mice for up- and downregulated genes, respectively. Gdf15^−/− mice showed 7% (22/321 DEGs) and 8% (18/234 DEGs) for up- and downregulated genes between male and female mice. Similarly, when examined for overlap in biological pathways this was very different among male and female mice with WT mice showing only 18% overlap and Gdf15^−/− mice with 3% overlap (Fig. 7B). This is further highlighted in Fig. 7C, where the gene expression and the biological pathways common to and distinct in each genotype and biological sex are represented. These results underscore the crucial role of sex as a biological variable in regulating the gene expression in the lung upon exposure to hyperoxia. Differences between the transcriptomic responses in WT male and female neonatal lungs have been highlighted in our prior publications (34). Volcano plots showing the DEGs in male and female WT and Gdf15^−/− mice are shown in Fig. 7D. One of the top downregulated genes in Gdf15^−/− male mice is Adarb1 (Adar2), which is involved in myeloid cell differentiation (40). SiglecF was decreased in Gdf15^−/− female mice, a known marker for lung alveolar macrophages (41). Representative enriched biological pathways in WT and Gdf15^−/− male and female mice are shown in Fig. 7E. Noticeably, immune system development and myeloid cell differentiation are negatively enriched in Gdf15^−/− male mice, as in response to oxidative stress. Lipid metabolic pathways were negatively enriched in WT males, whereas carbohydrate-derivative metabolic process was positively enriched in WT females. We validated our results from the RNA-Seq experiment with qRT-PCR of some relevant genes based on the curation of our enriched biological pathways. With the marked decrease in the number of macrophages in the Gdf15^−/− lung both under room air and upon exposure to hyperoxia, we measured the gene expression for Spp1 and SiglecF. Expression of both genes was decreased in Gdf15^−/− hyperoxia-exposed mice compared with WT. On the other hand, the expression of Ucp1 was significantly increased in the hyperoxia-exposed Gdf15^−/− mice compared with WT (Fig. 8). Uncoupling protein 1 (Ucp1), a well-known thermogenic and mitochondrial gene was increased in the Gdf15^−/− mice. This gene is also highly expressed in the myeloid/macrophage population in the developing lung (Supplemental Fig. S2). Uncoupling proteins are members of the family of mitochondrial anion carrier proteins, that separate oxidative phosphorylation from ATP synthesis with energy dissipated as heat. Deficiency of Ucp1 promotes macrophage recruitment (42). The role of GDF15 as a mitochondrial stress-induced cytokine is well established (43), but the role in the setting of neonatal hyperoxic lung injury in a cell-specific manner needs to be elucidated (44, 45).

Figure 7.

Remarkable sex-specific differences in the pulmonary transcriptome in WT and Gdf15^−/− mice exposed to hyperoxia (95% oxygen, PND 1–5). A: differentially expressed genes (DEGs) upregulated and downregulated in male and female WT and Gdf15^−/− mice exposed to hyperoxia compared with room air controls. Venn diagrams showing genes that are common to and distinct in WT and Gdf15^−/− mice. B: Venn diagrams showing biological pathways that are common to and distinct in hyperoxia-exposed WT and Gdf15^−/− male and female mice compared with room air controls. C: four-way Venn plots showing overlap and distinct DEGs and pathways between male and female WT and Gdf15^−/− mice exposed to hyperoxia compared with room air controls. D: volcano plots showing upregulated and downregulated genes in WT and Gdf15^−/− male and female mice exposed to hyperoxia compared with room air controls. E: biological pathways enriched in WT and Gdf15^−/− male and female mice exposed to hyperoxia compared with room air controls. Normalized enrichment score (NES) plotted on the x axis. Gdf15, growth differentiation factor 15; PND, postnatal day; WT, wild type.

Figure 8.

Validation of RNA-Seq data by qRT-PCR. Differentially expressed genes in RNA-Seq were validated by qRT-PCR. A: Spp1 mRNA expression B: Ucp1 mRNA expression, C: Siglec F mRNA expression. Values are represented as means ± SD. Independent biological replicates in each group are shown (n = 6–8/group). Statistical analysis was performed using two-way ANOVA. Significant differences between groups are shown by #P < 0.05, ###P < 0.001, **P < 0.01, and ***P < 0.001.

DISCUSSION

GDF15 is a stress-responsive protein that has recently received significant scientific attention as a key modulator of metabolic adaptation to stress and regulation of appetite (10). The role of GDF15 in many pulmonary diseases has been studied, and they range from disease modulation to a marker of severity and mortality (12). The significance of GDF15 in bronchopulmonary dysplasia (BPD) is yet to be elucidated. To our knowledge, this study is the first that focuses on the novel hypothesis that the loss of Gdf15 exacerbates neonatal hyperoxic lung injury in the mouse model.

Our study showed that Gdf15 deficiency decreased the tolerance to hyperoxia and decreased survival upon exposure to hyperoxia in Gdf15^−/− compared with WT mice. This significant result points to the protective effect of GDF15 levels in hyperoxic conditions. The sources of increased GDF15 production could be both pulmonary and extrapulmonary, leading to increased circulating GDF15 levels (46). The survival results of this study comply with our published in vitro study, which showed that GDF15 loss in the pulmonary epithelial and endothelial cells increases cellular oxidative stress and decreases cell viability (19).

Clinically, infants with BPD show poor growth and nutritional status, which are detrimental to postnatal lung growth (47, 48). Experimentally, mouse models of BPD show lower body weight in hyperoxia-exposed neonatal mouse pups (28). GDF15 regulates appetite, food intake, and body weight via its interaction with its receptor, glial cell line-derived neurotrophic factor (GDNF) family receptor α-like (GFRAL), a distant relative of receptors for a distinct class of the TGF-β superfamily ligands, in the hindbrain (49, 50). Tsai et al. (51) and other studies have shown that Gdf15^−/− mice weigh more than WT mice, and that this difference manifests at 4 wk of age. In contrast to these findings, our study showed that Gdf15^−/− mice had lower body weight than WT mice upon exposure to hyperoxia. That may be because GDF15 deficiency augments the intolerance to hyperoxic lung injury, hence exaggerating the weight loss in that group. Evaluation at PND 21 following neonatal hyperoxia exposure might provide insufficient recovery time, which may halt the expected regain in body weight in the Gdf15^−/− mice.

BPD is characterized by the arrest in lung development with aberrant vascular development and abnormal alveolar septation (2). Our results suggest that in the postnatal hyperoxia model, loss of Gdf15 exacerbates neonatal hyperoxic lung injury with respect to alveolarization with a worse phenotype in Gdf15^−/− mice compared with WT mice at PND 21. Significantly, the sex-specific differences noted in WT mice with the female mice being more resilient to hyperoxic injury compared with male mice, were not seen in Gdf15^−/− mice. Lung weight/body weight ratios immediately after hyperoxia injury at PND 5 also showed similar results with a significant increase in WT male mice, but not in female mice. However, this sex-specific difference was lost in female mice. Prior studies have shown sex-specific differences in hyperoxic lung injury regarding alveolarization, angiogenesis, and inflammation via different molecular mechanisms (28, 34, 52–57).

Hyperoxia causes a decrease in pulmonary vascular development. We did not observe robust differences between Gdf15^−/− mice and WT regarding vascular development upon exposure to hyperoxia, but there was a difference at baseline under normoxia. Prior studies showed that GDF15 plays an important role in angiogenesis via the induction of other angiogenic factors, including Hif-1 α and Vegf (24, 58).

Inflammation is one of the major contributors to lung injury and the development of BPD (59). Inhibiting inflammatory cell recruitment in the lung leads to improved alveolarization and angiogenesis (60). Many studies on Gdf15^−/− mice showed higher macrophage and inflammatory cell infiltration in different pathological states (21, 61). Interestingly, we observed that Gdf15^−/− mice had significantly lower macrophage recruitment compared with WT mice under hyperoxia and normoxia. In addition, Gdf15 deficiency blunted the expected increase in macrophage recruitment in the Gdf15^−/− mice in response to hyperoxia. The disparity between study results can be secondary to the significant complexity of the regulatory mechanisms and differences in the time courses of inflammatory cell recruitment and activation. In a model of ricin-induced acute lung injury, Gdf15^−/− mice showed decreased alveolar macrophages (24 h postinjury) and decreased number of Ly6c^lo (circulating/patrolling) monocytes (72 h postinjury). However, the same study showed increased interstitial macrophages in the Gdf15^−/− lung following ricin-induced injury compared with WT mice (62). Regarding macrophage function, Gdf15 decreases macrophage production of tumor necrosis factor and nitric oxide by inhibiting TGF-β-activated kinase TAK1 (63). Gdf15 is required for oxidative metabolism in macrophages and leads to M2-like polarization (64). A recent study highlighted the critical role of myeloid-derived GDF15 in the regulation of both the infiltration and transition to a reparative phenotype following injury (65). Gdf15 was identified in the Ly6c^lo monocyte-derived macrophage population in this study and single-cell sequencing identified a unique Gdf15 expressing repair-type macrophage subpopulation (65). More studies need to explore the interaction and mechanistic role of GDF15 in macrophage modulation and activation in hyperoxic lung injury and BPD with close attention to the time course of macrophage composition and cell state after hyperoxic injury.

Genes related to macrophage activation, recruitment, and homeostasis were differentially expressed in the Gdf15^−/− lung. Spp1 or Osteopontin (Opn) is an arginine-glycine-aspartate-containing adhesive glycoprotein that is secreted as soluble cytokine (66) and is known to be produced by alveolar macrophages as well as other cells in the lung (67, 68). Spp1 also plays an important role in macrophage recruitment, function, and accumulation at sites of injury (69–71). Notably, Spp1 was downregulated in Gdf15^−/− mice in our study. Other genes such as Csf1r (macrophage colony-stimulating factor I receptor) were downregulated in the Gdf15^−/− male lung upon exposure to hyperoxia. Csf1r plays a crucial role in the development of tissue macrophages and modulates macrophage survival and chemotaxis (72). Other macrophage function-related genes include Eng (38, 39) and Siglec F, both of which are downregulated in the Gdf15^−/− mice.

The exact source of GDF15 production in the injured neonatal lung needs to be elucidated. Apart from myeloid cells being a potential source, as detailed earlier, evidence from in vivo and in vitro experiments points toward lung epithelial cells (73–75). Data from single-cell RNA-sequencing data from human IPF lungs identified alveolar epithelial cells as the source for GDF15, and levels were increased both in the murine model and the human disease (76). High Gdf15 expression was reported in aberrant basaloid cells in IPF lungs at the edge of myofibroblastic foci (77). Some evidence also points to the pulmonary endothelial cells as sources of the protein (19, 78). GDF15 levels were recently associated with postcapillary pulmonary hypertension (79). However, the levels and the possible modulatory role of GDF15 in BPD-PH still need to be determined. In an ovalbumin-sensitization-based asthma murine model, GDF15 expression was induced by Notch in Treg cells and mediated a Type 2 innate lymphoid cell-dependent inflammation (80). Gdf15 deficiency attenuated the cigarette smoke-induced pulmonary inflammation (81), but cigarette smoke-induced GDF15 expression increased airway epithelial cell senescence (82) and mucin overexpression (83). In summary, there could be multiple cellular sources of GDF15 in different injury models and they could differ based on the kind of injury/disease process, developmental stage, acute/chronic phase of the disease/injury, and the cellular microenvironment.

The only known receptor for GDF15 is in the hindbrain (GFRAL) (49, 50). Upon binding to this receptor, it mediates anorexic effects and is responsible for weight loss and cachexia associated with GDF15 expression. Whether GDF15 exerts any receptor-mediated effects in the lung are not known. Zhang et al. (76) reported that there was no expression of GFRAL in the lungs. In vitro experiments in A549 cells suggest that the biological effects of GDF15 may be dependent on TGFBR2 expression (84). We recently reported that GDF15 levels are elevated in preterm babies, are inversely related to the gestational age at birth, and that longitudinal GDF15 levels were associated with increased respiratory support and need for oxygen at 36 wk postmenstrual age (85). This suggests that as a biomarker, sustained elevation in GDF15 levels is associated with worse clinical outcomes. The critical windows of intervention postinjury where modulation of GDF15 with either increasing physiological levels acutely or decreasing sustained elevated GDF15 levels need to be elucidated.

Among the limitations of this study is the use of the global knockout model for the loss of Gdf15. Timed loss of Gdf15 after the onset of injury or during the recovery would pinpoint the role of the gene during the acute lung injury and the chronic repair phase. In this study, we are not able to report the cellular source of Gdf15 production within the lung. Studies from adult lungs have mainly pointed to the epithelial origin of this protein. In the developing lung, the expression of Gdf15 is mainly in the myeloid/macrophage cell population (LungMAP database) and perhaps switches to epithelial cells later during development (Supplemental Fig. S3). Among the myeloid cells, Gdf15 expression is higher in the alveolar and interstitial macrophages (Supplemental Fig. S3) (86). Timed scRNA-Seq experiments and lineage tracing experiments using Gdf15 reporter mice will enable the cellular sources and fate of these cells in the lung. We did not assess the acute effects of loss of GDF15 at the acute stage of neonatal hyperoxic lung injury at PND 5. Also, given the role of GDF15 as a biomarker in adult pulmonary arterial hypertension, whether the loss of GDF15 is associated with a worse pulmonary hypertension phenotype following neonatal hyperoxia exposure was not studied. Finally, to answer the question whether the phenotype seen in Gdf15^−/− mice is due to the local pulmonary effects or centrally GFRAL-mediated effects of GDF15 will need to be elucidated using GFRAL loss of function murine models.

In conclusion, we show for the first time that loss of Gdf15 exacerbates mortality, lung injury, and the phenotype of the arrest of alveolarization in the developing lung with loss of female-sex advantage in Gdf15^−/− mice. We also highlight the distinct pulmonary transcriptomic response in the Gdf15^−/− lung including pathways related to macrophage recruitment and activation.

DATA AVAILABILITY

The raw data have been uploaded to the National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO) under Accession No. GSE211744.

SUPPLEMENTAL DATA

Supplemental Tables S1 and S2 and Supplemental Figs. S1–S3: https://doi.org/10.6084/m9.figshare.22771034.

GRANTS

This work was supported in part by National Institutes of Health Grants R01-HL144775, R01-HL146395, and R21-HD100862 (to K.L.) and P42 ES0327725 and R01HL129794 (to B.M.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

K.L. conceived and designed research; F.A.-M., W.J., L.W., X.D., and K.L. performed experiments; F.A.-M., M.C.G., A.C., W.J., L.W., X.D., B.M., and K.L. analyzed data; F.A.-M., M.C.G., and K.L. interpreted results of experiments; F.A.-M., M.C.G., A.C., L.W., and K.L. prepared figures; F.A.-M., M.C.G., A.C., and K.L. drafted manuscript; M.C.G., B.M., E.S., and K.L. edited and revised manuscript; M.C.G., W.J., E.S., and K.L. approved final version of manuscript.