Neonatal Hyperoxia Activates Activating Transcription Factor 4 to Stimulate Folate Metabolism and Alveolar Epithelial Type 2 Cell Proliferation

Abstract

Oxygen supplementation in preterm infants disrupts alveolar epithelial type 2 (AT2) cell proliferation through poorly understood mechanisms. Here, newborn mice are used to understand how hyperoxia stimulates an early aberrant wave of AT2 cell proliferation that occurs between Postnatal Days (PNDs) 0 and 4. RNA-sequencing analysis of AT2 cells isolated from PND4 mice revealed hyperoxia stimulates expression of mitochondrial-specific methylenetetrahydrofolate dehydrogenase 2 and other genes involved in mitochondrial one-carbon coupled folate metabolism and serine synthesis. The same genes are induced when AT2 cells normally proliferate on PND7 and when they proliferate in response to the mitogen fibroblast growth factor 7. However, hyperoxia selectively stimulated their expression via the stress-responsive activating transcription factor 4 (ATF4). Administration of the mitochondrial superoxide scavenger mitoTEMPO during hyperoxia suppressed ATF4 and thus early AT2 cell proliferation, but it had no effect on normative AT2 cell proliferation seen on PND7. Because ATF4 and methylenetetrahydrofolate dehydrogenase are detected in hyperplastic AT2 cells of preterm infant humans and baboons with bronchopulmonary dysplasia, dampening mitochondrial oxidative stress and ATF4 activation may provide new opportunities for controlling excess AT2 cell proliferation in neonatal lung disease.

Approximately 10% of births occur before 37 weeks of gestation and are thus considered preterm (1). Preterm infants are at risk of developing bronchopulmonary dysplasia (BPD), a chronic form of lung disease seen in preterm infants and baboons receiving supplemental oxygen therapies (2, 3). Survivors of preterm birth also face increased risks of developing chronic lung and cardiovascular disease that may shorten their lifespan (4, 5). In fact, a recent study looking at mortality of 6.2 million citizens in four Nordic nations found that preterm birth was associated with a two-fold increased risk of death from cardiovascular disease, diabetes, and chronic lung disease after age 15 (6). Among the many potential causes of chronic neonatal lung disease, early exposure of the lung to oxygen is considered particularly suspect because it inhibits angiogenesis and impairs alveolar development in a variety of animal models (7, 8).

Although hyperoxia clearly inhibits angiogenesis and thus capillary development, its effects on the developing alveolar epithelium appear to be more complex. Proliferation of alveolar epithelial type 2 (AT2) cells is low in rodents at birth and transiently rises during alveolar development (Postnatal Days [PNDs] 4 to 10) with peak proliferation around PND7 (9). Hyperoxia inhibits AT2 cell proliferation that normally rises on PND4 (10, 11). Although hyperoxia inhibits AT2 cell proliferation, genetic lineage mapping and bromodeoxyuridine (BrdU) labeling in neonatal mice revealed hyperoxia rapidly and paradoxically stimulates proliferation within the first 24 and 48 hours of exposure before suppressing it on PND4 (12, 13). Hyperoxia also reinitiates AT2 cell proliferation in newborn rabbits exposed for 10 days, and this is associated with increased expression of fibroblast growth factor 7 (FGF7), a potent mitogen and cytoprotective molecule for AT2 cells (14). The effects of hyperoxia on FGF7 expression may be specific to rabbits, because they are not seen in newborn mice (15). Despite this, the combination of an early and a late wave of AT2 cell proliferation may explain why AT2 cell hyperplasia is often seen in animal models exposed to hyperoxia for a week or more (13, 14, 16).

To understand how hyperoxia stimulates the early wave of AT2 cell proliferation, we analyzed the transcriptome of AT2 cells isolated from PND4 mice exposed to room air or hyperoxia since birth. It revealed that neonatal hyperoxia stimulates the expression of genes involved in serine synthesis and one-carbon coupled folate metabolism. This pathway begins when L-serine is produced in the cytoplasm from 3-phosphoglycerate by the sequential actions of phosphoglycerate dehydrogenase (PHGDH), phosphoserine aminotransferase 1 (PSAT1), and phosphoserine phosphatase (PSPH) (17). Serine shuttles into the mitochondria where serine hydroxymethyltransferase 2 (SHMT2), the bifunctional methylenetetrahydrofolate dehydrogenase/cyclohydrolase (MTHFD2), and the monofunctional C1-tetrahydrofolate synthase (MTHFD1L) convert it to formate through sequential oxidation/reduction reactions. Formate then shuttles from the mitochondria to the cytoplasm, where it is converted back to serine by the actions of SHMT1, MTHFD1, and MTHFD1L. The actions of mitochondrial and nonmitochondrial shuttling of carbon units create reducing equivalents and one-carbon moieties that cells need to proliferate. Interestingly, the same set of genes also increased when AT2 cells normally proliferate on PND7 and in response to the mitogen FGF7. However, hyperoxia selectively stimulates their expression via mitochondrial stress activation of the activating transcription factor 4 (ATF4), thus providing new insight into how neonatal hyperoxia disrupts alveolar lung development.

Methods

Detailed methods for this study may be found in the data supplement.

Studies with Mice

Mice of the C57BL/6J strain were exposed to room air or hyperoxia (100% oxygen) and injected with 0.7 μg/g mitoTEMPO as described (12, 18). AT2 cells were isolated using dispase with agarose instillation (19). AT2 cells collected from Sftpc^EGFP mice were further purified using a cell sorter (Figures E1A and E1B in the data supplement). Cells were spun onto slides, and purity was determined by staining and counting SP-C⁺ or enhanced green fluorescent protein (EGFP)⁺ cells (Figure E1C).

RNA-Sequencing and Analysis

RNA isolated from AT2 cells was used to generate 25 million reads using the Illumina sequencing library construction. The R 3.4.3/Bioconductor 3.5/Limma 3.32.1 was used to estimate 2,830 differentially expressed genes at a 5% false discovery rate (FDR) among 19,718 genes detected (20). Gene set enrichment was tested among genes differentially expressed at 1% FDR and >2 fold-change using enrichGO in clusterProfiler 3.4.4 (21). Data were deposited (#GSE140915) in the Gene Expression Omnibus of the National Center for Biotechnology Information.

qRT-PCR

PCR products were amplified on a CFX384 Touch Real-Time PCR Detection System using sequence-specific primers (Table E1). Relative gene expression was determined by averaging duplicate runs using the △△CT method and normalizing to 18S RNA.

Cells and siRNA Transfection

Mouse AT2 or MLE15 cells were transfected with Lipofectamine RNAiMAX Transfection Reagent containing siRNA molecules to Mthfd2 (M-042690–01; Dharmacon), Atf4 (L042737–01; Dharmacon), or Phgdh (M-045115–01; Dharmacon) versus a nontargeting control (D-001210–02–05; Dharmacon).

Immunohistochemistry

Lung sections were stained with antibodies to ATF4 (D4B8; Cell Signaling), Mthfd2 (PA5–28169; ThermoFisher), Ki67 (ab15580; Abcam), GFP (ab6663; Abcam), and ABCA3 (13-H2–57; Seven Hills Bioreagents) that were detected using fluorescently labeled secondary antibodies. Tissues were counterstained with DAPI and positive cells quantified as described (22).

Western Blot Analysis

Western blot membranes were immunoblotted with rabbit anti-ATF4 (D4B8; Cell Signaling), anti-GRP78/BiP (ADI-SPA-826; Enzo), phosphor-eiF2a and eiF2a (#9721 and #9722; Cell Signaling), anti-MTHFD2 (PA5–76502; Thermofisher), or anti–β-ACTIN antibody (A2066; Sigma) as described (22).

Reactive Oxygen Species Measurements

Cells were incubated with 5 μm MitoSox (M36008; ThermoFisher) washed, spun onto glass slides, stained with Hoechst dye 33342 (62249; ThermoFisher), and visualized under fluorescent microscopy (23).

Human and Baboon Lung Tissues

Donor human lungs were obtained from six infants: three infants who died with BPD (19-month-old female, 21-month-old female, and 15-month-old male) and three infants who died of nonpulmonary causes (19-month-old female, 15-month-old male, and 20-month-old female). Preterm baboon lung tissues described by Maniscalco and colleagues (16) were provided by the Southwest Foundation for Biomedical Research. Sections were stained for Ki67 (ab15580; Abcam), Mthfd2 (PA5–28169; ThermoFisher), or ABCA3 (13-H2–57; Seven Hills Bioreagents).

Statistical Analysis

Data were evaluated using JMP15 software (SAS Institute) and graphed with Prism 8 (GraphPad Software) as means ± SEM of three to four experiments. The analysis was used to determine overall significance, followed by comparisons for all pairs using Tukey-Kramer HSD tests, where P < 0.05 was considered statistically significant.

Study Approval

All experiments with mice (#2007–121R) and humans (#RSRB00047606) were evaluated and approved by the appropriate review board at the University of Rochester.

Results

Neonatal Hyperoxia Stimulates Expression of One-Carbon Coupled Folate Metabolism Genes in Proliferating AT2 Cells

To identify potential mechanisms by which neonatal hyperoxia stimulates AT2 cell proliferation, transgenic Sftpc^EGFP mice were exposed to room air or hyperoxia between PND0 and PND4 (Figure 1A). AT2 cells were isolated and enriched to high purity using flow cytometry and gating for intrinsic green fluorescence afforded by their transgenic expression of EGFP (Figure E1). RNA was isolated from these cells and used for RNA-sequencing (RNA-seq) studies. It revealed that the RNA was highly enriched for genes expressed by epithelial cells, particularly AT2 cells, because of the high expression of Sftpc and Abca3 (Table E2). Out of 19,718 genes detected, 2,830 were differentially expressed between the room air- and the hyperoxia-exposed mice using an FDR of 5%. Neonatal hyperoxia stimulated the expression of 1,861 genes and suppressed the expression of 969 genes. The 10 most abundantly suppressed and 50 most abundantly stimulated genes are depicted in a heat map (Figure 1B). Gene ontology analysis detected 78 gene sets with significant (FDR <5%) enrichment of differentially expressed genes. The cellular amino acid metabolism, carboxylic and biosynthetic process, and small molecule biosynthetic process pathways spanned most of the genes included in the top 10 sets (Figures 1C, 1D, and E2). Among these three groups, 27 genes were involved in cellular amino acid metabolism, 22 were involved in carboxylic and biosynthetic process, and 35 were involved in small molecule biosynthetic process. Interestingly, 15 genes were involved in serine synthesis and one-carbon coupled folate metabolism (Table E3).

Figure 1.

RNA-sequencing analysis of AT2 cells isolated from PND4 mice exposed to room air or hyperoxia. (A) Cartoon model showing newborn Sftpc^EGFP mice were exposed to room air or hyperoxia between PND0 and PND4. EGFP⁺ AT2 cells were sorted by flow cytometry on PND4 for RNA-sequencing studies. (B) Heatmap shows relative log-fold changes in genes expressed by AT2 cells with FDR of less than 5% isolated from mice exposed to room air or hyperoxia. Each column represents data from a single mouse. Genes in bold and marked with an arrow were confirmed by PCR and studied further. (C) Gene ontology analysis of differentially expressed genes. The size of the dot reflects the number of genes in the group, whereas color reflects significance relative to room air. (D) Network visualization of differentially expressed (FDR < 1%) individual genes within the cellular amino acid metabolism, carboxylic and biosynthetic, and small molecule biosynthetic process pathways. AT2 = alveolar epithelial type 2; EGFP = enhanced GFP; exprs. = expressions; FDR = false discovery rate; PND = postnatal days.

Genes involved in serine synthesis and one-carbon coupled folate metabolism are important for cell proliferation because they generate reducing equivalents required for epigenetic maintenance and redox defense and one-carbon moieties required for amino acid homeostasis and biosynthesis of purines and thymidine (Figure 2A). RNA was isolated from AT2 cells isolated from a new group of mice exposed to hyperoxia. qRT-PCR confirmed that neonatal hyperoxia stimulated expression of Mthfd2, Shmt2, and Mthfd1l that collectively produce formate in the mitochondria (Figure 2B). Neonatal hyperoxia also stimulated Phgdh, Psat1, and Psph used to generate L-serine. The increased expression of Mthfd2 mRNA in hyperoxic AT2 cells was particularly intriguing to us because MTHFD2 also traffics to the nucleus, where it stimulates proliferation through a presently undefined mechanism that is independent of its dehydrogenase functions (24). To further investigate Mthfd2 expression, AT2 cells isolated from PND4 mice exposed to room air or hyperoxia were stained for the proliferation marker Ki67 and MTHFD2. Hyperoxia increased the number of AT2 cells with nuclear-speckled staining of both Ki67 and MTHFD2 (Figure 2C). The speckled staining of MTHFD2 in the nucleus has been shown to reflect sites of DNA replication where Ki67 also resides (24). We were not able to colocalize Ki67 with MTHFD2, because the Ki67 and MTHFD2 antibodies were both made in rabbits. Hyperoxia did not affect the expression of Shmt1 and Mthfd1, two related enzymes in the parallel cytoplasmic one-carbon coupled folate pathway (Figure 2B).

Figure 2.

Newborn hyperoxia stimulates the expression of genes involved in one-carbon folate coupled metabolism in AT2 cells. (A) Cartoon model showing how serine is produced in the cytosol and donates methyl groups through mitochondrial and cytoplasmic one-carbon coupled folate metabolism. (B) AT2 cells were isolated on PND4 from three mice exposed to room air or hyperoxia and pooled to create a single sample. RNA was then isolated from three to four separate exposures and used to assess the expression of bifunctional methylenetetrahydrofolate dehydrogenase/cyclohydrolase (MTHFD2) Mthfd2, serine hydroxymethyltransferase (Shmt) 2, and the monofunctional C1-tetrahydrofolate synthase (Mthfd1l), phosphoglycerate dehydrogenase (Phgdh), phosphoserine aminotransferase 1 (Psat1), and phosphoserine phosphatase (Psph), Mthfd1, and Shmt1 by qRT-PCR. Data are graphed as individual values where the line reflects the mean value. *P < 0.05 and **P < 0.01 compared with room air. (C) AT2 cells were isolated from PND4 mice and stained for Ki67 or MTHFD2 (red) and DAPI (blue). Values below each set of images represent mean ± SEM with individual values reflected as dots. *P < 0.05, **P < 0.01, and ***P < 0.001 compared with room air control. Scale bar, 20 μm.

Serine and One-Carbon Coupled Folate Genes Are Highly Expressed When AT2 Cells Proliferate in Room Air

Two approaches were used to investigate whether Mthfd2 and other genes involved in one-carbon coupled folate metabolism are induced whenever AT2 cells proliferate or only in response to hyperoxia. We began by investigating whether Mthfd2 expression changed during normal postnatal alveolar lung development. Mthfd2 mRNA was detected in the whole lung of PND1 mice exposed to room air. It remained constant at PND4, increased fourfold on PND7, and then declined through PND42 before diminishing to nearly undetectable concentrations by PND56 (Figure E3). This pattern of Mthfd2 expression was intriguing because it correlated with the increased expression of Ki67 mRNA. Also, peak Mthfd2 and Ki67 were detected on PND7 when AT2 cells transiently proliferated during rodent alveolar development (25). To confirm AT2 cells were proliferating at this time, the lungs of PND4 and PND7 Sftpc^EGFP mice exposed to room air were stained for Ki67 and EGFP. Ki67 was detected on PND4 in approximately 5% of AT2 cells, whereas on PND7, it was detected in nearly 40% of AT2 cells, demonstrating a marked increase in proliferation of these cells as the lung enters alveolar development (Figures 3A and 3B). RNA was then isolated from AT2 cells harvested from PND4 and PND7 mice exposed to room air. qRT-PCR revealed expression of Mthfd2, Shmt2, Mthfd1l, Phgdh, Psat1, and Psph was significantly higher in AT2 cells isolated from PND7 mice than those from PND4 mice (Figure 3C). There was no change in expression of the cytosolic Mthfd1 and Shmt1. These findings reveal a postnatal proliferation of AT2 cells is associated with increased expression of Mthfd2 and other genes involved in serine synthesis and mitochondrial one-carbon coupled folate metabolism.

Figure 3.

One-carbon coupled folate metabolism gene expression increases when AT2 cells proliferate in PND7 mice exposed to room air. (A) Lungs of PND4 and PND7 Sftpc^EGFP mice exposed to room air were stained for EGFP (green), Ki67 (red), and DAPI (blue). White arrows point to dual positive cells. Scale bar, 50 μm. (B) The proportion of EGFP⁺ AT2 cells expressing Ki67 was quantified and graphed as mean ± SEM. Dots represent the values of individual mice. (C) AT2 cells were isolated from four PND4 and three PND7 mice exposed to room air. RNA was then isolated and used to assess the expression of Mthfd2, Shmt2, Mthfd1l, Phgdh, Psat1, Psph, Mthfd1, and Shmt1 by qRT-PCR. Data are graphed as individual values where the line reflects the mean value. *P < 0.05 and ***P < 0.001 compared with room air PND4 mice.

We next investigated whether these genes were increased by FGF7, a potent mitogen for AT2 cells (9, 26). AT2 cells were isolated from adult mice and cultured with FGF7 for 48 hours in the presence or absence of siRNA oligonucleotides targeting Mthfd2 (Figure 4A). FGF7 stimulated nuclear staining of Ki67 and MTHFD2 (Figure 4B) as it increased the number of AT2 cells by approximately fourfold (Figure 4C). FGF7 also stimulated the expression of Mthdf2, Shmt2, Mthfd1l, Phgdh, Psat1, and Psph mRNA (Figures 4D–4K). It did not affect the expression of the cytoplasmic Shmt1 and Mthfd1. siRNA silencing of Mthfd2 expression inhibited basal and FGF7-dependent proliferation and overall survival (Figure 4C). It also rescued the expression of enzymes involved in serine synthesis and mitochondrial tetrahydrofolate production (Figures 4D). Silencing Mthfd2 did not affect the expression of the cytoplasmic Shmt1 or Mthfd1, which did not change during hyperoxia.

Figure 4.

FGF7 stimulates the expression of one-carbon coupled folate metabolism genes and AT2 cell proliferation. (A) Cartoon model showing isolation of adult AT2 cells that were cultured in the presence or absence of fibroblast growth factor 7 (FGF7) and siRNA to Mthfd2 for 48 hours. (B) AT2 cells were stained for Ki67 or MTHFD2 (red) and DAPI (blue). Positive cells in boxes are enlarged below each image. Scale bars, 50 μm. (C) AT2 cells were transfected with siRNA to Mthfd2 or scrambled control and then cultured with or without FGF7. The number of AT2 cells was quantified before (0 h) and after 48 hours of culture. Each line represents the values from an individual culture. (D) Expression of Mthfd2, Shmt2, Mthfd1l, Shmt1, Phgdh, Psat1, Psph, Mthfd1, and ATF4 was quantified by qRT-PCR after 48 hours of culture. Values reflect mean ± SEM fold change compared with control cells cultured in the absence of FGF7 or siRNA to Mthfd2. Dots and whiskers represent values from cells isolated from a single mouse. Columns with the same letter are not different from each other, whereas columns with different letters are significantly different from each other. FGF7 significantly increased expression (b) P < 0.01 relative to (a). n.d. = not determined.

Hyperoxia Selectively Stimulates ATF4 Expression in Proliferating AT2 Cells

In the small molecule biosynthetic process, the third predominantly enriched pathway, we observed that hyperoxia increased expression of activating transcription factors Atf3 (8.6-fold), Atf4 (3-fold), and Atf5 (24-fold). These genes, along with the C/EBP homologous protein (Chop), are activated by the integrated stress response (ISR), a set of signaling pathways that help cells adapt to changes in their environment (27). There are four arms to the ISR that reflect activation by endoplasmic reticulum stress, amino acid deprivation, double-stranded RNA, and heme deprivation. Although ATF4 and CHOP are responsive to endoplasmic reticulum stress, they did not appear to reflect this type of stress, because hyperoxia did not affect the expression of the canonical endoplasmic reticulum (ER) stress-responsive genes GRP78/BiP, Perk, or Atf6 (Table E4). This is consistent with results from our prior study showing hyperoxia by itself does not activate ER stress signaling in A549 cells (28). Instead, it may reflect mitochondrial stress because hyperoxia also stimulated LonP1 and the mitochondrial heat shock proteins Hsp60 and Hspa9/Mortalin, genes activated by mitochondrial oxidative stress and unfolded proteins (29). Interestingly, hyperoxia did not change the expression of Pink1 and Park2, two genes required for mitophagy.

The expression of ATF4 was studied further because it is a member of the cAMP response element-binding protein family of transcription factors that plays a critical role in autophagy, the unfolded protein response, hypoxia, and nutrient deprivation (30). It is also believed to be more proximal or upstream of other ATFs. Faint ATF4 staining was detected by immunohistochemistry in some AT2 cells of PND4 mice exposed to room air, while intense ATF4 staining was detected in many AT2 cells of mice exposed to hyperoxia (Figure 5A). Some of these cells were also proliferatively defined by BrdU labeling. In contrast, minimal ATF4 staining was detected in PND7 mice exposed to room air even though BrdU label or Ki67 was now detected in EGFP-positive AT2 cells (Figure 5B). Hyperoxia increased Atf4 mRNA when AT2 cells proliferate on PND4 (Figure 5C). Atf4 expression was not induced in PND7 mice exposed to room air even though the proliferation of AT2 cells was detected.

Figure 5.

Neonatal hyperoxia stimulates ATF4 in AT2 cells. (A) Lungs of Sftpc^EGFP mice exposed to room air or hyperoxia between PND0 and PND4 were stained for ATF4 (red), EGFP (green), BrdU (bromodeoxyuridine) (purple), and DAPI (blue). Scale bars, 50 μm. Boxed regions are enlarged below each figure. (B) AT2 cells were isolated from PND7 Sftpc^EGFP mice exposed to room air and stained for ATF4 (red), EGFP (green), and DAPI (blue) or Ki67 (red), EGFP (green), and DAPI (blue). Scale bars, 50 μm. Boxed regions are enlarged below each figure. (C) AT2 cells were isolated from PND4 mice exposed to room air or hyperoxia and PND7 mice exposed only to room air. RNA was isolated and used to quantify the expression of Atf4 by qRT-PCR. Values reflect mean ± SEM of three mice per group. Dots represent the values of individual mice. ***P < 0.001 compared with PND4 room air.

Antioxidants that scavenge superoxide produced by mitochondria or nicotinamide adenine dinucleotide phosphate oxidase can alleviate and even preserve normal lung development in rodents exposed to hyperoxia (10, 31, 32). To test whether ATF4 was activated by mitochondrial stress, we investigated whether scavenging mitochondrial superoxide using mitoTEMPO would block ATF4 activation in mice exposed to hyperoxia. Newborn Sftpc^EGFP mice were administered mitoTEMPO daily while exposed to room air or hyperoxia (Figure 6A). AT2 cells were harvested and stained with the redox-sensitive dye mitoSOX (Invitrogen) followed by counterstaining with Hoechst dye. Increased mitoSOX red staining was detected in EGFP⁺ AT2 cells isolated from mice exposed to hyperoxia and was markedly diminished in mice administered mitoTEMPO (Figure 6B). MitoTEMPO also inhibited the oxygen-dependent increases in Atf4 and Mthdf2 mRNA seen in AT2 cells isolated from the mice (Figure 6C). It also decreased ATF4 staining in EGFP⁺ AT2 cells (Figures 6D–6F) and thus reduced the number of EGFP⁺ AT2 cells detected in mice exposed to hyperoxia (Figures 6D and 6G).

Figure 6.

MitoTEMPO inhibits oxygen-dependent changes in ATF4 and Mthfd2. (A) Cartoon model showing Sftpc^EGFP newborn mice were administered mitoTEMPO (arrows) or vehicle on PND0, PND1, and PND2 while being exposed to room air or hyperoxia between PND0 and PND4. (B) AT2 cells were isolated from Sftpc^EGFP mice exposed to room air or hyperoxia and stained for mitoSOX red and Hoechst dye. Red (mitoSOX; Invitrogen), green (EGFP), and blue (Hoechst) were then visualized by epifluorescence microscopy. Scale bar, 20 μm. (C) RNA was isolated from AT2 cells and used to quantify the expression of Atf4 and Mthfd2 by qRT-PCR. Values represent mean ± SEM of three to four mice per group compared with the room air vehicle. ***P < 0.001 compared with room air or MitoTEMPO treatment. (D) Lungs of PND4 Sftpc^EGFP mice exposed to room air or hyperoxia in the presence or absence of mitoTEMPO were stained for EGFP (green), ATF4 (red), and DAPI (blue). Scale bar, 50 μm. (E and G) The proportion of ATF4 cells in the lung (E), ATF4 in EGFP⁺ AT2 cells (F), and proportion of EGFP⁺ AT2 cells in the whole lung (G) were quantified and graphed. Values reflect mean ± SEM of three mice per group. **P < 0.01 and ***P < 0.001.

Interestingly, mitoTEMPO did not reduce the total number of AT2 cells below the number seen in room air, suggesting that it has no effect in room air. To test this further, mitoTEMPO was administered to Sftpc^EGFP mice exposed to room air on PND4, PND5, and PND6. Lungs were then harvested on PND7 and stained for EGFP and Ki67 or ATF4. MitoTEMPO did not affect proliferation in PND7 mice exposed to room air (Figure E4). Thus, mitoTEMPO specifically suppresses mitochondrial oxidative stress and ATF4 signaling when AT2 cells proliferate in response to hyperoxia. It does not affect the normal wave of AT2 cell proliferation seen when alveolar development occurs on PND7.

Mouse lung epithelial cells (MLE-15) were exposed to room air or hyperoxia to establish the epistatic relationship between ATF4 and MTHFD2. Asynchronously dividing cultures of MLE15 cells were transfected with siRNA against Atf4, Mthfd2, or scrambled oligonucleotide controls for 24 hours (Figure E5A). The cells were then exposed to room air or hyperoxia for an additional 24 hours. Although hyperoxia inhibits MLE15 proliferation through DNA damage signaling (33), it still increases the expression of Mthfd2, Shmt2, and Atf4. Silencing Atf4 suppressed oxygen-dependent changes in Mthfd2 and Shmt2. In contrast, silencing Mthdf2 had no detectable effect on the expression of Atf4, but it did suppress Shmt2, indicating that silencing Mthfd2 was impacting other genes in one-carbon coupled folate metabolism.

We then investigated whether mitoTEMPO suppresses oxygen-dependent changes in ATF4 and Mthfd2 in MLE15 cells. As expected, hyperoxia increased mitochondrial superoxide production as defined by increased mitoSOX red staining (Figure E5B). Increased mitoSOX staining was inhibited in cells cultured with mitoTEMPO. Hyperoxia also stimulated Mthfd2 and Atf4 mRNA, and this too was inhibited in cells cultured with mitoTEMPO (Figure E5C). ATF4 is encoded by a downstream open-reading frame that is translated when stress signaling stimulates eiF2a phosphorylation on serine 51. Western blot studies confirmed hyperoxia increased ATF4 protein along with a slight increase in p-eiF2a (Figure E5D). MitoTEMPO suppressed these changes and even suppressed ATF4 abundance in cells exposed to room air. Hyperoxia and mitoTEMPO did not affect the expression of the ER stress protein BiP.

Hyperplastic AT2 Cells in Human and Baboon BPD Express MTHFD2

We investigated the proliferation of AT2 cells and expression of MTHFD2 and ATF4 in the lungs of three human infants with BPD and three age-matched infants with no lung disease (NLD) (Table E5). The BPD donors are classified as type 2 by Abman and grade 4 by Jensen because they remained on mechanical ventilation at greater than or equal to 36 weeks post-menstrual age (34, 35). The lungs were stained for ABCA3 expressed by AT2 cells and Ki67 expressed by proliferating cells. ABCA3 was used to identify AT2 cells because the antibody was made in goats and therefore could be used with rabbit antibodies against MTHFD2 or ATF4. ABCA3 was detected in numerous cells of infants with BPD compared with NLD (Figures 7A and 7D). This was associated with a three- to four-fold increase in the number of ABCA3 cells expressing Ki67 (Figures 7A and 7E). Most of these proliferating AT2 cells were seen in clusters lining remodeled alveoli with thickened septae. In contrast, rare proliferating AT2 cells seen in NLD appeared uniformly dispersed as single alveolar cells. Similarly, increased staining for MTHFD2 was seen in BPD lungs compared with NLD (Figures 7B and 7F). Quantitation of these cells revealed 80% of AT2 cells in BPD lungs expressed MTHFD2 compared with 20% of AT2 cells in NLD. Likewise, increased staining of ATF4 was detected in BPD lungs compared with NLD (Figures 7C and 7G).

Figure 7.

Hyperplastic AT2 cells seen in humans with bronchopulmonary dysplasia express ATF4 and MTHFD2. Human lung tissues were obtained from three infants that died with BPD and three infants that died at similar postnatal ages (15–21 mo) from NLD. Lungs were stained for (A) Ki67 (red), ABCA3 (green), and DAPI (blue); (B) MTHFD2 (red), ABCA3 (green), and DAPI (blue); and (C) ATF4 (red), ABCA3 (green), and DAPI (blue). Scale bars: A, 40 μm; C, 50 μm. The proportion of (D) DAPI-positive cells expressing ABCA3, (E) ABCA3-positive cells expressing Ki67, (F) ABCA3 expressing MTHFD2, and (G) ABCA3-positive cells expressing ATF4 were quantified and graphed. Values reflect mean ± SEM with dots reflecting the values of individual donors. *P < 0.05, **P < 0.01, and ***P < 0.001 compared with NLD. Statistical values could be obtained with Panel G because ATF4 was not detected in ABCA3⁺ cells of patients with NLD. BPD = bronchopulmonary dysplasia; NLD = no lung disease.

To confirm these findings, we evaluated ATF4 and MTHFD2 expression in preterm baboons delivered prematurely at 125 days’ gestation and treated with oxygen and ventilation (pro re nata [PRN]) for 14 days as needed to survive. We previously showed how their distal alveolar epithelium is lined with hyperplastic AT2 cells (16). Analogous to preterm humans, increased ATF4 and MTHFD2 staining was seen in cuboidal epithelial cells lining the alveolar airspace of 14-day PRN animals when compared with gestationally matched controls (Figure E6). Taken together, these data suggest that early life oxygen exposure in neonatal mice or preterm humans and baboons with BPD promotes hyperplasia of AT2 cells expressing MTHFD2 and other enzymes involved in one-carbon folate coupled metabolism.

Discussion

It is widely accepted that hyperoxia impairs normal growth and development of the preterm infant lung, in part, by suppressing endothelial cell proliferation and angiogenesis. Its effects on the developing alveolar epithelium appear to be more complex, because it initially stimulates and then inhibits the proliferation of AT2 cells. Despite suppressing growth, AT2 cell hyperplasia is seen with prolonged exposure to hyperoxia and in autopsy samples of BPD human and preterm baboon lung. To provide clarity on this early wave of proliferation, we provide new evidence showing how hyperoxia stimulates the expression of genes required for serine synthesis and mitochondrial one-carbon coupled folate metabolism via mitochondrial stress activation of ATF4 (model in Figure E7). Interestingly, the same metabolic genes but not ATF4 were highly expressed when AT2 cells normally proliferated during alveolar development. Scavenging mitochondrial reactive oxygen species with mitoTEMPO blocked ATF4 activation and thus AT2 cell proliferation during hyperoxia, but it had no effect when AT2 cells normally proliferated later during alveolar development. Antioxidants and inhibitors of ATF4 activity may therefore prevent or diminish AT2 cell hyperplasia in BPD with little to no effect on normal lung development.

Hyperoxia stimulated the expression of many genes required for one-carbon coupled folate metabolism. The increased expression of Mthfd2 mRNA stood out because it is a nuclear-encoded mitochondrial bifunctional enzyme with MTHFD2 and activities. During fetal development, MTHFD2 provides an important source of carbon units needed for synthesizing purines and thymidine in DNA synthesis. MTHFD2 expression decreases as embryos mature and is nearly absent after birth (36). Its actions in the mitochondria are replaced by MTHFD2L, which has nicotinamide adenine dinucleotide-positive and nicotinamide adenine dinucleotide phosphate-positive dehydrogenase activities and is highly expressed as a housekeeping enzyme in both the brain and lung (37). Although MTHFD2 is not expressed after birth (38), it is highly induced in rapidly dividing tumors, where it is thought to produce additional one-carbon units needed to enable rapid growth and generate NADH (nicotinamide adenine dinucleotide + hydrogen) required for protection against reactive oxygen species produced by overly active mitochondrial respiration (38, 39). MTHFD2 has recently been detected at sites of DNA replication, where it stimulates cell proliferation independent of its dehydrogenase activity (24). Consistent with a direct role for MTHFD2 in DNA replication, we observed discrete punctate nuclear staining of MTHFD2 in Ki67-positive AT2 stimulated to proliferate with FGF7. siRNA targeting Mthfd2 suppressed proliferation and survival of AT2 cells isolated from mice and stimulated to proliferate with FGF7. Based upon these observations, we conclude proliferating AT2 cells require or at least use MTHFD2 and other genes involved in one-carbon coupled folate metabolism.

In addition to Mthfd2, proliferating AT2 cells also express Phgdh, Psat1, and Psph used to generate serine. As illustrated in Figure 2A, serine is converted to glycine in the cytoplasm by the trifunctional enzymes MTHFD1 and MTHFD2 in the mitochondria before birth or MTHFD2L after birth. Interestingly, PHGDH is also required for one-carbon coupled metabolism because it regulates the mass balance of carbons required for nucleotide synthesis and redox maintenance (40). This would explain why siRNA targeting Phgdh or Mthfd2 in isolated AT2 cells, MLE15 cells, and even human lung adenocarcinoma A549 cells (data not shown) suppressed expression of their target genes, as well as many other enzymes in one-carbon folate coupled metabolism. Modulating Phgdh or other enzymes required for central carbon metabolism may therefore provide an additional way to control the proliferation and survival of AT2 cells.

A recent paper showed how glycolytic and fatty acid metabolites are replaced by serine, glycine, methionine, and one-carbon metabolites as the mouse lung undergoes alveolar development (41). Moreover, this transition in amino acids and one-carbon metabolites can be stimulated by exposing mice to 85% oxygen. The current study suggests hyperoxia stimulates these metabolic changes in AT2 cells via ATF4, a transcription factor central to the ISR, to nutrient deprivation, viral infection, oxidative stress, and proteostasis. Four specialized kinases (PERK, GCN2, PKR, and HrI) sense these stresses and phosphorylate a single serine on the eukaryotic initiation factor eiF2, thus decreasing cap-dependent and increasing cap-independent translation of specific mRNAs like Atf4. We speculate this might involve GCN2 signaling because studies in newborn rats found that hyperoxia stimulates eiF2a phosphorylation needed to translate Atf4 and expression of GCN2, ATF3, and ATF4 in alveolar endothelial and epithelial cells (42, 43). ATF4 has emerged as a central mediator of mitochondrial stress responses where it activates cytoprotective genes that interestingly reprogram cellular metabolism toward the synthesis of key metabolites, one of them being serine (30). In fact, the unfolded protein response stimulates the expression of enzymes that divert metabolites from glycolysis into mitochondrial one-carbon metabolism (44). Taken together, these observations suggest oxygen-induced mitochondrial oxidative stress may activate ATF4 that redirects carbon sources, perhaps in an attempt to restore redox balance under severe oxidizing conditions. But it may also inadvertently stimulate the proliferation of AT2 cells, which use the same network of genes to proliferate during alveolar development and in response to mitogens such as FGF7.

It is important to point out some limitations in our study. The transcriptome of mouse AT2 cells was analyzed on PND4 because we felt this would give us the greatest opportunity to detect changes in gene expression. It revealed that hyperoxia markedly influenced the expression of genes controlling serine synthesis and one-carbon coupled folate metabolism, but interestingly, few changes in genes involved in cell cycle progression. The most logical explanation for this is that expression of these genes was already declining because hyperoxia was starting to inhibit AT2 cell proliferation. Consistent with this hypothesis, the RNA-seq studies revealed the cyclin-dependent kinase inhibitor cdkn1a (p21^{Cip1/Waft1/Sdi}) was significantly higher (4.01 log2fold, P = 2.1 × 10²⁵) in AT2 cells isolated from mice exposed to hyperoxia compared with room air. Transcriptional studies using AT2 cells isolated from mice exposed to hyperoxia for a shorter period of time may identify early response genes used by hyperoxia to start the process of proliferation. It was also challenging to show hyperoxia stimulates proliferation strictly through one-carbon coupled folate metabolism. We found that hyperoxia stimulates the expression of Mthfd2 and other folate metabolism genes in MLE15 cells. But hyperoxia did not stimulate the proliferation of these cells, perhaps because they are a tumor line that is already highly proliferative. Despite this, siRNA suppression of Mthfd2 clearly inhibited their proliferation and survival, thus supporting an important role for mitochondrial one-carbon coupled folate metabolism in AT2 cell proliferation. Finally, our study does not discern how chronic hyperoxia restimulates AT2 cell proliferation or causes AT2 cell hyperplasia. Because chronic hyperoxia stimulates FGF7 expression in neonatal rabbits (14) and suppressing FGF signaling (45) increases the severity of oxygen-induced lung disease, we speculate that hyperoxia might activate a second wave of AT2 cells proliferation via FGF7 (or another mitogen)-dependent stimulation of ATF4 and folate metabolism.

In summary, we uncovered a previously unappreciated mechanism by which hyperoxia stimulates AT2 cell proliferation through ATF4 activation of genes that AT2 cells normally use to proliferate. Activation of this network may reflect a mitochondrial stress response designed to generate one-carbon units needed to maintain a redox state under severe oxidizing conditions. But this adaptive response to hyperoxia may become maladaptive when the same network of genes inappropriately stimulate the proliferation of AT2 cells. Mitochondrial antioxidants and inhibitors of ATF4 signaling may therefore prove useful for reducing chronic lung disease in preterm infants.