Mendelian randomization and experimental IUGR reveal the adverse effect of low birth weight on lung structure and function

Intrauterine growth restriction (IUGR) and low birth weigth (LBW) are risk factors for neonatal chronic lung disease. However, maternal and fetal genetic factors and the molecular mechanisms remain unclear. We investigated the relationship between LBW and lung function with Mendelian randomisation analyses and studied angiogenesis in a low protein diet rat model of IUGR. Our data indicate a possible association between LBW and reduced FEV1 (p = 5.69E−18, MR-PRESSO) and FVC (6.02E-22, MR-PRESSO). Complimentary, we demonstrated two-phased perinatal programming after IUGR. The intrauterine phase (embryonic day 21) is earmarked by a reduction of endothelial cell markers (e.g. CD31) as well as mRNA expression of angiogenic factors (e.g., Vegfa, Flt1, Klf4). Protein analysis identified an activation of anti-angiogenic mTOR effectors. In the postnatal phase, lung capillaries (< 20 µm) were significantly reduced, expression of CD31 and VE-Cadherin were unaffected, whereas SMAD1/5/8 signaling and Klf4 protein were increased (p < 0.01). Moreover, elevated proteolytic activity of MMP2 and MMP9 was linked to a 50% reduction of lung elastic fibres. In conclusion, we show a possible link of LBW in humans and reduced lung function in adulthood. Experimental IUGR identifies an intrauterine phase with inhibition of angiogenic signaling, and a postnatal phase with proteolytic activity and reduced elastic fibres.

Intrauterine growth restriction (IUGR) is a multifactorial disease affecting approximately 10% of the newborn population and is classically defined as a birthweight two standard deviations below the 50th percentile for the gestational age 1 . Recently, IUGR has been characterised as a failure of a foetus to reach its full growth potential due to genetic or environmental factors, including maternal, placental, and fetal causes, which then leads to restricted nutrient and oxygen supply 1,2 . Infants with IUGR are at higher risk for adverse perinatal complications, such as prematurity, respiratory distress, and bronchopulmonary dysplasia (BPD) 2 . It has also been demonstrated that IUGR followed by postnatal catch-up growth increases the susceptibility to chronic adult's diseases beyond infancy, including hypertension and vasculopathies 3 .

Scientific Reports
| (2020) 10:22395 | https://doi.org/10.1038/s41598-020-79245-7 www.nature.com/scientificreports/ Intrauterine exposure to adverse nutritional, metabolic, and hormonal alterations can interfere with organ development and induce life-long changes in structure and physiology. This was initially described as fetal, perinatal or metabolic programming by Barker and colleagues in the 1990s 4,5 . Previous studies have shown that IUGR impairs alveolar formation and lung growth, leading to reduced lung function in adult rats. These structural and functional changes may be related to a disruption of key developmental signaling pathways, such as TGFβ signaling, and inflammatory response. This may induce matrix remodeling, including distorted elastic fibre assembly 6,7 . While clinical studies associated IUGR with lower diffusion capacity, increased susceptibility to infections, and obstructive pulmonary disease [8][9][10] , the genetic disposition remain elusive. Recent studies demonstrated that low birth weight (LBW), used as a proxy for IUGR, is causally related to an increased susceptibility to coronary artery disease, and hightens the risk for pulmonary arterial hypertension (PAH) in infants with BPD 11 . These clinical data suggest a causual genetic association of IUGR and susceptibility to lung diseases that needs to be elucidated.
Angiogenesis is essential for alveolar formation and lung growth. The activation of endogenous pro-angiogenic pathways increases endothelial cell survival, proliferation and migration, thereby driving alveolarisation 12 . The concerted interaction of growth factor signaling, such as vascular endothelial growth factor (VEGF) and bone morphogenetic protein (BMP) signaling, is central in endothelial cell function 13,14 . The disruption of these critical angiogenic pathways impairs microvessel formation and arrests lung growth, as is seen in BPD 13 www.nature.com/scientificreports/ Previous studies have shown that IUGR decreases pulmonary vessel growth in sheep and impairs proliferation, migration, and tube formation of pulmonary artery endothelial cells 16 . However, the molecular mechanisms disrupting developmental processes and angiogenesis in lungs following IUGR remain unclear. Based on the association of IUGR resulting in LBW and BPD we pursued two approaches: Firstly, we studied the long-lasting causal relationship of LBW and reduced lung function in humans with Mendelian randomisation; secondly, we investigated whether disruption of developmental pathways at intrauterine and postnatal time points contributes to impaired angiogenesis and elastic fibre metabolism after IUGR, thereby causing failed alveolarisation and functional impairment (Fig. 1A).

Materials and methods
Mendelian randomisation. We performed two-sample Mendelian randomisation analyses to infer causality between birth weight and forced expiratory volume in 1 s (FEV1) and forced vital capacity (FVC), used as proxies for lung function. We selected and used as exposure the genome-wide significant independent hits (p value ≤ 5E10−8) associated with own and offspring birth weight (Supplemental Table 1) from the genome-wide association meta-analyses of own and offspring birth weight performed by the Early Growth Genetics (EGG) Consortium 17 . We used the genome-wide association studies (GWAS) summary statistics of FEV1 (field ID 3063) and FVC (field ID 3062) performed in the UK Biobank by Neale et al. (http://www.neale lab.is/uk-bioba nk/) as outcomes.
We used four separate methods to estimate causal effects: the standard inverse-variance weighted (IVW) regression with and without MR-PRESSO 18 (Mendelian Randomization Pleiotropy RESidual Sum and Outlier; to minimise the risk of horizontal pleiotropy); as well as two robust regression methods, the weighted medianbased method, and Egger regression 19 . We applied robust methods with special assumptions about the behavior of pleiotropic variants, such as MR-Egger 20 , which assumes pleiotropic effects are uncorrelated with the genetic associations with the risk factor, the InSIDE assumption; and the MR-PRESSO 18 , that excludes outlying variants as being potentially pleiotropic. In addition, we performed leave-one-out sensitivity analyses to identify if a single SNP was driving an association. We estimated statistical power for the MR analyses assuming a clinically relevant fixed effect size of 0.1 SD with an alpha threshold of 0.05. Power for MR analyses was estimated using the method reported by Burgess et al. 21,22 . We performed the two-sample MR analyses 19,23 with the R package TwoSampleMR.
Tissue assays. RNA extraction and real-time qPCR. Total RNA extraction and quantitative RT-PCR were performed as previously described 7 . Quantitative changes in mRNA expression were assessed with Taqman or SYBR Green PCR Master Mix (Invitrogen, 11743500 & 11760500, Germany) using a 7500 Real-Time PCR System (Applied Bioscience). Primers were designed using Primer Express Software (Applied Biosystems) (Supplemental Table 2). The ΔΔC t method, as previously described, was used for quantification 25 .
Protein extraction, quantification, and immunoblot. Protein was isolated from homogenised whole-lung tissue and quantified, followed by immunoblotting as previously described 7 . The primary antibodies and the secondary peroxidase-conjugated anti-mouse, anti-rabbit or anti-rat antibodies (commercially available and tested) are listed in Supplemental Table 3. Quantitative analysis was performed with densitometry (Bio-Rad ImageLab software, Bio-Rad, Munich, Germany) using hypoxanthine-guanine phosphoribosyltransferase (HPRT) and β-Actin as loading controls. All data represent contiguous lanes, complete blots are shown in the Supplemental Figures.
Zymography. Protease activity of metalloproteinase 2 (MMP2) and MMP9 was analysed by gelatin-based zymography as previously described 7 . Activity was quantified by densitometric analysis on negative images (Image Lab Software, Bio-Rad Laboratories, Germany). All data represent contiguous lanes, blots are not cropped.
Histology and immunhistochemistry. Microvessel count. Randomly selected lung sections (3 µm thickness) were stained with the primary antibody anti-von Willebrand Factor, followed by a secondary antibody (for details see Supplemental Table 3). The sections were then scanned using the slide scanner (Leica SCN400). Microvessels (2-20 µm and 20-100 µm, displaying a lumen) were quantified in a total of ten fields of 20× view per tissue section and 6 random tissue sections per animal (6 animals per group).
Elastic fibre quantification. Lung sections were stained for elastic fibres using Resorcin Fuchsin from Weigert (Weigert's Iron Resorcin and Fuchsin Solution; Carl Roth, X877.3) 26 , and counterstained yellow with Tatrazine [0.5% in 0.25% acetic acid (Dianova, cat. no. TZQ999, USA)]. The elastic fibre density as an index of parenchymal elastin content was analysed in up to ten fields of 20× view per tissue section and 6 random tissue sections

Statistical analysis.
All results are displayed as mean ± SEM. The unpaired non-parametric T-test was used to compare the control (normal protein) to the IUGR (low protein) group. For a quantitative comparison between time points, a two-way ANOVA followed by Bonferroni post-test was performed. The p-value was adjusted using Bonferroni post-test and less than 0.05 was considered significant. The statistical analyses were carried out using the Graph Pad Prism software (GraphPad software Version 6.0 and 7.0, San Diego, CA, USA).

MR analyses indicate a possible association of birth weight with adult lung function.
With more than 80% statistical power our data suggest an association of birth weight with lung function (FVC and FEV1) in all Mendelian randomisation analyses performed. The leave-one out sensitivity analysis did not highlight any SNPs with a large effect on the results. After excluding outlying variants as being potentially pleiotropic using MR-PRESSO, the analysis showed neither significant heterogeneity or directional horizontal pleiotropy. The analyses using MR-Egger and weighted median methods consistently yielded similar effect estimates. The direction of the effect was positive (i.e., LBW was associated with reduced lung function). Full results are shown in Table 1 and Supplemental Figure 1.

IUGR was established in a low-protein diet rat model. Low protein diet of dams during gestation
induced IUGR in the offspring when compared to controls. The experimental rat model of IUGR is illustrated in Fig. 1B. As shown in Fig. 1C, offspring had LBW at (E21) and P3, whereas their weight was significantly higher at P23, indicating a postnatal catch-up growth after IUGR.

IUGR alters pulmonary vessel growth and regulates endothelial cell markers. Quantification
of lung microvessels with CD31 immunostaining at P3 and P23 ( Fig. 2A) revealed a significant formation of larger microvessels (20-100 µm) between P3 und P23 in control pups, not visible in IUGR. Small microvessels (< 20 µm) were significantly decreased in IUGR when compared to control at P23, suggesting dysfunctional postnatal angiogenesis after IUGR (Fig. 2B). The mRNA expression of Pecam1 (CD31) was significantly reduced at E21 and immunoblot showed a non-significant reduction of VE-Cadherin protein expression at P3, but not at P23 after IUGR, compared to control (Fig. 2C,D).
IUGR dysregulates the pro-angiogenic BMP signaling pathway in lungs.  (Bmpr1a) was significantly reduced at E21, whereas Bmpr1b mRNA expression was significantly reduced after IUGR at P23. Bmpr2 mRNA was also expressed differentially, with a slight reduction at E21 and a significant upregulation at P3 after IUGR (Fig. 4A). Immunoblot of the downstream signaling cascade of BMPRs showed activation of SMAD1/5/8 (pSMAD1/5/8) after IUGR at P3, but no difference at P23 (Fig. 4B). The mRNA expres- www.nature.com/scientificreports/ sion of apelin (Apln), inhibitor of differentiation (Id1), and Krüppel-like factor 4 (Klf4), regulators of endothelial cell homeostasis or BMP interactors, were also reduced after IUGR at E21. Apln remained significantly lower in IUGR than Control at P3, whereas Klf4 gene expression was significantly downregulated at P23 (Fig. 4C). The protein abundance of Klf4, a regulator of stem cell capacity, was slightly reduced at E21 and more than twofold increased at P23 (Fig. 4D).
IUGR dysregulates cell metabolism through AMPKα and mTOR signaling. Immunoblot analysis revealed that IUGR significantly inhibited AMPKα signaling during the intrauterine phase (E21) (Fig. 5A). In contrast, assessment of mTOR signaling using 4E-BP1 as a downstream effector showed a marked activation at E21 (Fig. 5B). However, during postnatal lung development neither AMPKα nor mTOR signaling were significantly regulated by IUGR.
IUGR increases lung protease activity and reduces lung elastic fibre content. A Hart stain for elastic fibres (Fig. 6A) revealed a reduction of relative elastic fibre content by 50% at both time points P3 and P23 (Fig. 6B). Assessment of Eln (elastin) mRNA expression did not significantly differ between IUGR and Control group (Fig. 6C), suggesting increased degradation as a result of proteolytic activity. To test this notion we used zymography, and determined a significant increase of metalloproteinase 2 (MMP2) and MMP9 activity after IUGR at P3 (Fig. 6D). ). The respective quantification is shown below the immunhistochemical stainings (n = 6/group). (C) Assessment of mRNA expression of pulmonary endothelial cell markers, VE-Cadherin (Cdh5) and CD31 (Pecam1) on embryonic day 21 (E21), P3 and P23 using qRT-PCR; Glyceraldehyde 3-phosphate dehydrogenase (Gapdh) served as housekeeping gene; the control group was set at 1. (D) Measurement of protein abundance using immunoblot; the respective densitometric quantification of protein expression of VE-Cadherin on P3 and P23 is shown below the immunblots; VE-Cadherin was related to the loading control HPRT (n = 5-6/group). Mean ± SEM. A non-parametric T-test was used to compare IUGR to the control group,*p < 0.05. Comparison between P3 and P23 with a two-way ANOVA test, # p < 0.05, ## p < 0.01.

Discussion
The present study shows a possible association between LBW in humans, used as a proxy of IUGR, and reduced lung function in adulthood. In addition, it provides a pathomechanism of acute and long-term impact of IUGR on pulmonary microvascular formation and elastic fibre formation in an experimental rat model of IUGR. The combined results of these two approaches paint the picture of dynamic intrauterine programming of pulmonary development. Specifically, during the intrauterine phase, restricted nutrient supply and cellular stress are related to a significant disruption of pro-angiogenic VEGFA-and BMP-signaling, activation of anti-angiogenic mTOR signaling, downregulation of angiogenic growth and transcription factors, and loss of endothelial cell markers. Whereas during the postnatal phase, angiogenic signaling caught up, but small microvessels were significantly reduced and associated with an increased lung proteolytic activity, resulting in a persistent reduction of elastic fibre content. These processes accumulate in a loss of microvasculature and alveolarisation, evidenced by a loss of lung function in adulthood.
Linking LBW to reduced lung function using Mendelian randomization. With Mendelian randomisation, we have shown a positive association between LBW and reduced lung function (FEV1, FVC), indicating a possible reduced functional lung volume in individuals with LBW. It has been previously shown that a reduction of FVC is associated with lung emphysema 27 , defined as a decrease in elastic recoil 28 . In addition, previous studies have shown an association between LBW and type 2 diabetes and cardiovascular diseases 11 . Our data provide a direct link between maternal effects and LBW-associated lung function that is in line with the Barker Hypothesis.

Strengths and limitations of Mendelian randomization.
Our study included the large sample size and the most recent and powerful GWAS summary statistics used as exposures and outcomes. In addition, we used a wide range of sensitivity analyses increasing robustness of our findings. Our study also has several limitations. First, the participants included in the genetic analyses were of European ancestry. Hence, our results may not be generalizable to other ethnic groups with significantly different prevalence/predispositions with regard to the outcome. Second, we analyzed birth weight as a continuous variable without specifically looking at LBW. Moreover, we cannot exclude the importance of additional environmental factors implicated in the relationship between birth weight and lung function. Finally, samples of the GWAS for the exposure (the genome-wide association meta-analyses of own and offspring birth weight) and the GWAS for the outcome [FEV1 (field ID Immunoblots illustrating protein abundance of total ERK1/2 and phosphorylated ERK1/2 (pERK1/2) on E21, P3 and P21; pERK1/2 was related to βActin or to total ERK1/2; the densitometric analysis is shown next to the respective immunoblot (n = 5-6/group). Mean ± SEM. A non-parametric T-test was used to compare IUGR to the control group,*p < 0.05, **p < 0.01. Translational approach: dysregulation of angiogenic signaling is associated with impaired lung growth in rats with IUGR . VEGF promotes angiogenesis and alveolarisation in animal models of neonatal lung injury 29 . In the present study, we demonstrate a significant downregulation of Vegfa and Flk1 (VEGF-R2) in the intrauterine phase, which is followed by a catch-up to reach Control levels. Interestingly, the downstream effector ERK1/2 protein 30 is reduced on E21, and increased after birth. However, the activation of   Immunoblots showing the analysis of AMP-Activated protein kinase (AMPK)α-pathway in total lung homogenate on embryonic day (E21), postnatal day 3 (P3), and P23; total AMPKα and phosphorylated AMPKα (pAMPKα) were assessed; βActin served as loading control; pAMPKα was related to β-Actin or to total AMPKα; densitometric summary data are shown next to the respective immunoblot (n = 6/group). (B) Total 4E-BP1 and phosphorylated 4E-BP1 (p4E-BP1) as a downstream effector of mTOR-pathway were assessed with immunoblot on E21, P3, and P23; βActin served as loading control; p4E-BP1 was related to βActin or to total 4E-BP1; densitometric summary data are shown next to the respective immunoblot (n = 6/group). Mean ± SEM. A non-parametric T-test was used to compare IUGR to the Control group, *p < 0.05, **p < 0.01, ***p < 0.001. www.nature.com/scientificreports/ liferation and migration 31 . Reduced activation of the AMPKα signaling pathway during the intrauterine period may further contribute to the inhibition of the VEGF signaling machinery. BMP signaling drives angiogenesis and is essential for the maintenance of endothelial cell homeostasis, and mutations in this family are associated with hereditary PAH. The transduction of BMP signaling is facilitated by complex formation between BMP receptors 1 and 2, activating its downstream canonical SMAD1/5/8 and non-canonical ERK1/2 signaling 14 . Our data show that IUGR significantly downregulates Bmpr1a and Bmpr2 during the intrauterine period, whereas Bmpr1b was unaffected. In the postnatal phase, we found opposing effects, with similar gene expression of Bmpr1a and Bmpr2, but significant lower Bmpr1b in IUGR than control lungs. IUGR may block the proliferative Bmpr1a/Bmpr2 pathways during the intrauterine phase, while it is shifted to pro-proliferative signaling by postnatal reduction of Bmpr1b as an attempt to promote vascularisation and compensate for the intrauterine growth restriction 32 . This notion is further sustained by a postnatal activation of the downstream effector pSMAD1/5/8. This early developmental blockade of VEGF and BMP-receptor signaling in the lung may impair angiogenesis, induce arrest of alveolarisation and contribute to early origin of chronic lung diseases, such as BPD and PAH.

Scientific Reports
To determine whether the disruption of developmental signaling after IUGR may be related to changes in growth and transcription factors, we assessed Id1, Apln, and Klf4 expression. In accordance with our previous data, Id1, Apln, and Klf4 were significantly reduced by 50% in fetal lungs with IUGR, and postnatally upregulated. Downregulation of these factors has been associated with the pathogenesis of PAH 15,33,34 . The growth factor apelin is pro-angiogenic and regulates endothelial cell migration, proliferation and survival 15 , and preserves lung growth in a model of BPD 35 . Similarly, the expression of Id1 is induced by BMPR2 signaling, and maintains endothelial cell function 33,36 . On the other hand Klf4, promotes angiogenesis through activation of VEGF signaling 34 , and regulates cell survival and differentiation. Loss of these angiogenic factors may impair vascular development and induce chronic vascular lung disease.
Linking nutrient sensing pathways to angiogenic signalling in lungs of rats with IUGR . The pathways that signal anti-angiogenic effects during lung development remain uncertain. Organ growth requires an increased metabolic rate, which requires oxygen and nutrient supply. Nutritive restriction or placental insufficiency forces adaptation of the metabolism by changing cellular energy consumption. The serine/therosine kinase, mammalian target of rapamycin (mTOR) senses nutrient status in growing vessels, and modulates cel-   www.nature.com/scientificreports/ lular responses during angiogenesis 37 . For example, rapamycin inhibits VEGF synthesis, has antiproliferative activity, and blocks angiogenesis in vivo 37 . Likewise, it was shown that mTOR signaling regulates angiogenic sprouting 38 . Our data supports the notion that activated mTOR signaling in the intrauterine phase may interact with BMP, and disrupt angiogenic signaling after IUGR.
Proteolytic activity and disruption of elastic fiber formation may contribute to impaired lung structure and function after IUGR . Despite the postnatal catch-up of angiogenic signaling after IUGR, the number of small microvessels remains reduced. This initial paradox let us study processes independent of angionic signaling. Lung matrix is crucial in alveolarisation and serves as a scaffold that directs secondary septation and microvascular formation 39 . Experimental studies demonstrate that microvascular formation is reduced in genetically modified mice with elastin haplo-insufficiency, possibly increasing the risk for PAH 40 . Similarly, disturbed elastic fibre assembly and distribution are related to inhibition of angiogenesis in animal models of BPD 41 . BMPR2 signaling might contribute to impaired elastic fibre assembly via the BMP4-TGFβ1 pathway 40,42 . Additionally, we show a postnatal activation of metalloproteinases MMP2 an MMP2, linked to 50% reduction of elastic fibre content, persisting until P23. Increased degradation of elastic fibres might contribute to the post-IUGR alteration of lung function throughout life, including reduced FVC and increased FEV1. While our model of nutrient deprivation is highly relevant in countries, in which pregnant women are exposed to malnutrition, it does not reflect the pathomechanisms of placental insufficiency. In future studies, we recommend to confirm our findings in other models of IUGR. With our translational data from genome wide association studies, however, we aimed to correct for all causes of IUGR by defining a genetic association between LBW and lung function, independent of the cause, in a cross-sectional cohort.

Conclusion
In conclusion, our translational data show a possible association between low birth weight and lung function.
Our study provides novel insight as to how IUGR disrupts lung angiogenic signaling, elastic fibre formation, and microvascular formation in a time dependent manner in an experimental rat model. Using this approach,  www.nature.com/scientificreports/ we identified two phases of perinatal vascular programming (Fig. 7): first, an intrauterine phase with inhibition of angiogenic signaling; and second, a postnatal phase, in which vascular signaling catches up, but lack of formation of small microvessels is related to increased proteolytic activity, and ultimately marked reduction of elastic fibre content. These data highlight how adverse intrauterine malnutrition programs the microvascular system, increasing the susceptibility to chronic lung diseases such as BPD and PAH.