The development of blood protein profiles in extremely preterm infants follows a stereotypic evolution pattern

doi:10.21203/rs.3.rs-2429783/v1

Download PDF

Article

The development of blood protein profiles in extremely preterm infants follows a stereotypic evolution pattern

https://doi.org/10.21203/rs.3.rs-2429783/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 02 Aug, 2023

Read the published version in Communications Medicine →

Version 1

posted

You are reading this latest preprint version

Preterm birth is the leading cause of neonatal mortality and morbidity. Early diagnosis and interventions are critical to improve the clinical outcomes of extremely premature infants. Here, we have investigated the blood protein profiles during the first months of life in preterm infants to shed light on the role of early extrauterine development. The blood protein levels were analyzed using “next generation blood profiling” on 1335 serum samples, collected longitudinally at nine time points from birth to full term (40 weeks postmenstrual age) from 182 extremely preterm infants. We found a unified pattern of protein development for all included infants, regardless of gestational age and clinical characteristics, suggesting an age-dependent stereotypic development of blood proteins after birth. This knowledge should be considered in neonatal settings and might alter the clinical approach within neonatology, where postmenstrual age is today the most dominant age variable.

Biological sciences/Developmental biology

Health sciences/Biomarkers/Prognostic markers

Health sciences/Biomarkers/Diagnostic markers

Preterm birth is the leading cause of neonatal morbidity and mortality and a significant risk factor for long-term neurodevelopmental and respiratory impairment^1–3. The number of births before 28 weeks of gestation, defined as extremely preterm, was estimated to be approximately 600,000 in 2014 worldwide⁴. Although significant advances in neonatal intensive care over the last few decades resulted in increased survival rates^4–7, extremely preterm birth is associated with a high rate of neonatal and childhood diseases⁸.

The etiology behind preterm morbidities is multifactorial and poorly understood. Infants born before 28 gestational weeks spend the third trimester outside the womb in an environment to which the newborn is not adapted. Blood protein profiling could provide insight into the postnatal development of these immature infants^9,10. Studies of the proteome and its evolution can be used to find biomarkers for disease and diagnosis, where the information provided may serve as a basis for personalized medicine¹¹. Knowledge about the postnatal proteome evolution in relation to postmenstrual age (PMA) and postnatal age (PNA) respectively can shed light on the role of early extrauterine development in health and disease. Thus, there is an urgent need to understand the patterns and dynamics of blood protein profiles during the early development of preterm infants.

The proximity extension assay (PEA) technology, also referred to as “next generation blood profiling”, allows analysis of blood levels of hundreds and even thousands of proteins simultaneously in a small blood volume without sacrificing accuracy or sensitivity^12,13. Immaturity and excessive blood sampling may increase the risk of anemia in preterm infants, and treatment with blood transfusions is associated with neonatal morbidities and impaired neurological development ^14–16. Thus, the assay platform is ideal for the comprehensive analysis of minute quantities of blood.

Longitudinal proteomic studies in preterm infants are scarce. A recent pilot study on 14 infants born at GAs 22 to 27 weeks used a large panel of proteins based on the PEA principle to investigate the serum protein profiles⁹. The study showed dramatic changes in protein profiles unrelated to GA from day one through the early weeks of life. The extremely preterm infants had a distinct unified protein profile in serum on the first day after birth. Another study by Olin et al. presented significant changes in protein expression by comparing cord blood from preterm and term infants with peripheral blood at later ages¹⁷. In a small cohort of term infants, age was the most prominent factor influencing serum protein expression in the first three years of life.¹⁸ Nonetheless, the influence of prematurity and the protein development during the first weeks of life remains unresolved.

In the present study, we use the PEA method to investigate longitudinal serum protein profiles from birth to term equivalent age (postmenstrual week (PMW) 40) and relate the results to GA, PMA, PNA, sex, and mode of delivery in a large cohort of extremely preterm infants. Thus, our results fill a critical knowledge gap on the evolution of the protein profiles in extremely preterm infants from premature birth to term-equivalent age. We identified a postnatal time-dependent stereotypic development of blood proteins. A developmental proteomics reference map for extremely preterm infants has been established to facilitate future pediatric research. It pinpoints a period where the degree of immaturity seems to significantly impact the proteome, thus directing a window for potential diagnostics and therapeutics.

The study cohort

A total of 182 extremely preterm infants born before 28 weeks of gestation from the Mega Donna Mega study were selected based on the availability of longitudinal serum samples (figure. 1a)¹⁹. Altogether 105 boys and 77 girls were included. Serum samples were collected repeatedly at nine planned time points (visits) from birth to term-equivalent age. Forty (22%) infants had nine complete samples, and 165 (91%) infants had at least six samples. Among the 182 infants included in the study, 177 (97.3%) survived to 40 weeks PMA. The enrolled infants were classified into three groups depending on GA at birth: group 1, born at less than 25 + 0 (weeks + days) (N = 61); group 2, born at 25 + 0 to 26 + 6 (weeks + days) (N = 81); and group 3, born at 27 + 0 to 27 + 6 (weeks + days) of gestation (N = 40) (Fig. 1b). The birth weights varied from 425 to 1345 grams (Supplementary Fig. 1a). Almost all (98.4%) of the infants had a birthweight appropriate for gestational age with standard deviations (SDS) > -2 (Supplementary Fig. 1b)²⁰. Vaginal delivery was more common among infants born at lower gestational age (Mann–Whitney U-test, P = 0.017). We observed no significant differences in sex distribution or postnatal growth between GA groups (Fig. 1b, Supplementary Fig. 1b-d). The clinical characteristics of the three GA groups were summarized in Supplementary Data 1.

Proteome profiling demonstrated dynamic changes in blood proteins after birth

We analyzed 538 unique protein targets measured by six Olink PEA panels, including cardiometabolic, cardiovascular II and III, development, inflammation, and metabolism, for all 1335 collected blood serum samples (Supplementary Data 2). Protein levels measured as Normalized Protein Expression (NPX) were determined for each target and sample. An example of the protein expression determined by the Olink PEA technique can be seen in Figs. 2a, where the levels of fibroblast growth factor 21 (FGF-21) is shown from birth to full-term age (PMW 40). FGF-21 is a protein involved in metabolism and growth by regulating insulin sensitivity and glucose uptake²¹. We observed higher levels with increasing postnatal age of the infant (Fig. 2a), which is consistent with our previous report that FGF-21 serum levels were elevated after birth⁹. No differences in FGF-21 expression between sexes could be seen.

To explore variance of protein levels, inter-individual and intra-individual variations were calculated for each protein across all 182 infants and nine visits (Fig. 2b, Supplementary Data 3). Most of the proteins were observed with considerable variability in both, and FGF-21 was the most variable protein in the analysis. To get a comprehensive overview of the postnatal changes in the blood proteins, we analyzed the time-course expression patterns of the variable blood proteins from birth to 40 weeks PMA. Over time, differentially expressed proteins (451 out of 538) were identified using linear mixed-effect modeling with Benjamini-Hochberg adjusted p-value < 0.01, including 196 up-regulated proteins and 255 down-regulated proteins across the study visits (Fig. 2c). The longitudinal changes in protein expressions for each differentially expressed protein can be seen in the circular heatmap in Fig. 2d. Unsupervised hierarchical clustering analysis was further performed on the longitudinal expression profiles of the proteins based on Pearson correlation. A total of eight separate clusters, ranging in size from 34 to 84 proteins, were identified with variable time-course patterns (Fig. 2d and 2e, Supplementary Data 4). As seen in Fig. 2e and supplementary Fig. 2a and 2b, five clusters display overall declining protein levels and three increasing trends. The proteins most strongly changed over time included leptin (LEP), LDL-receptor, and several placenta elevated proteins, including Fc fragment of IgG receptor IIa (FCGR2A) and CGA (supplementary Fig. 2c). The effect of GA at birth on the clustering trends was further explored. As seen in Supplementary figures S3a and S3b, protein levels in the three GA-groups were almost the same at the birth, indicating similar expression patterns of proteins in different GA groups at birth.

Functional analysis of the protein clusters

The functions of the proteins in each of the eight identified clusters were explored. Tissue specificity of the proteins in the clusters was analyzed based on the Human Protein Atlas (HPA) classification^22,23. This classification, elsewhere described, considers the level of gene expression in each tissue to determine the degree of specificity. Of the 451 proteins in the eight clusters, 101 proteins (22%), were annotated as tissue enriched proteins according to the HPA classification (Fig. 3a and Supplementary Data 5). The analysis showed that most of the liver, lymphoid tissue, and salivary gland enriched proteins were increased after birth, indicating the development of hepatic functions and immune and metabolic shifts during the neonatal period. Two examples include carboxylesterase 1 (CES1), a primary liver enzyme that functions in liver drug clearance²⁴, and Fc fragment of IgE receptor II (FCER2), which has essential roles in B cell growth and differentiation, as well as the regulation of IgE production (Fig. 3b)²⁵. Many proteins that decreased after preterm birth were associated with the placenta, pancreas, and bone marrow, consistent with our previous findings (Fig. 3a)⁹. For example, Carboxypeptidase A1 (CPA1), a pancreas-enriched protein, is produced in the pancreas and preferentially cleaves C-terminal branched-chain and aromatic amino acids from dietary protein²⁶. Hepatocyte growth factor (HGF) is an acidic protein with a strong mitogenic effect on hepatocytes. Still, it is also enriched in the placenta with strong expression in the villous syncytium, extravillous trophoblast, and amnionic epithelium (Fig. 3b)^27,28.

To explore the postnatal development of the immune system in preterm infants, we investigated the immune cell specificity of the proteins with longitudinal dynamic changes. The cellular specificity is determined based on the gene expression levels in the 18 different immune cell types from the HPA (Fig. 3c and 3d)²³. Here, 43 (9.5%) proteins were annotated as immune cell type enriched according to the HPA classification, which were at least four-fold higher expressed in one cell type than all other cell types (Supplementary Data 6). A large fraction of the proteins in clusters 6, 7, and 8 with increasing levels after birth are enriched in plasmacytoid dendritic cells. In contrast, most proteins with decreasing trends (clusters 1, 2, and 5) are enriched in basophils or neutrophils. The proteins in cluster 3 with an increasing trend followed by decreased protein levels have a mixed immune cell origin, including proteins enriched in T-cells, eosinophils, neutrophils, and basophils. This suggests that the expression activity of the dendritic cells increases after birth in these preterm infants, while the expression of proteins from basophils and neutrophils decreases.

In addition, functional enrichment analyses were performed to explore the modulated pathways for each identified cluster (Supplementary Fig. 4 and Supplementary Data 7a and 7b). As expected, multiple immune-related pathways were activated after preterm birth, including the T cell activation, inflammatory response pathway, interleukin-17 signaling pathway, and hematopoiesis pathway. Interestingly, the receptor for the advanced glycation end products (RAGE) pathway, which plays a vital role in leukocyte recruitment and have relatively high blood levels in extremely premature infants²⁹, was deactivated during postnatal development.

Variance analysis of blood protein profiling after birth

To assess how clinical aspects affect the blood protein expression levels, we established a linear regression model for each protein target and included four factors: PNA, GA at birth, sex, and mode of delivery (see Fig. 4a and 4b). The model revealed that the protein expression variations were primarily associated with PNA. This indicates that regarding protein variation, postnatal time is dominant compared to gestational age at birth. Nonetheless, GA at birth is the second most explanatory factor. Sex and delivery mode impacted a few specific proteins; however, overall substantially less influential compared to postnatal age. The contribution of the predictor variables was summarized in Supplementary Data 8. Moreover, to further expand the evaluation of impact of PNA on blood proteome expression, we investigate the potential of blood proteome as a predictor for PNA. We employed generalized linear models with an elastic-net penalty and identified a ‘blood proteomic clock’ comprising 151 proteins (Supplementary Data 9). The predictive PNA had a high level of consistency with chronological age, as demonstrated by a Pearson correlation coefficient of 0.98 (Fig. 4c). This suggests that the blood proteins are reliable measurements for estimating PNA of preterm infants. Interestingly, it was observed that among the samples that were predicted to be younger than the actual PNA, the weight gain tended to be lower than normal (Supplementary Fig. 5).

The proteins with the highest effect within each analyzed factor have been highlighted in Fig. 4d-g. The influence of PNA was most prominent in the delta-like non-canonical Notch ligand 1 (DLK-1), with 75.1% of the serum protein level variance explained by PNA (Fig. 4d). DLK-1, also known as preadipocyte factor 1 (PREF1), is a marker of preadipocytes and inhibits adipogenesis³⁰. It has been proposed that its function is to shift metabolism from lipid storage to peripheral lipid oxidation and act as a mediator of metabolic adaptation in early life³¹. DLK-1 levels were constant up to 30 weeks PMA, whereas after they decreased considerably, as shown in Fig. 4d.

As mentioned above, PNA seems more important than GA at birth in determining protein variance; however, for some proteins, variance is more associated with the GA at birth (seen to the right in Fig. 4a). One example was the tissue factor pathway inhibitor (TFPI), the primary inhibitor of the extrinsic coagulation pathway³². As illustrated in Fig. 4e, apparent differences in protein expression for TFP1 were observed between the three GA groups. The infants born at younger gestational ages had a persisting higher expression until full-term, when the expression levels converged. Furthermore, we show that the glycoprotein hormones alpha chain (CGA) levels decreased rapidly during the first days after birth and with a more significant decline in males, resulting in longitudinally lower levels in the male infants (Fig. 4f). CGA is one of the subunits that form the hormones human chorionic gonadotropin (hCG), luteinizing hormone (LH), follicle-stimulating hormone (FSH), and thyroid-stimulating hormone (TSH)³³. Several proteins related to the mode of delivery were identified (Fig. 4a). The strongest association was seen for surfactant protein D (PSP-D, also called SP-D). An elevated level of PSP-D was observed in infants delivered by cesarean section (Fig. 4g).

Distinct and coherent evolution of blood protein profiles over time

To investigate the global molecular dynamics of preterm infants, we performed several dimensional reduction analyses, including Uniform Manifold Approximation and Projection (UMAP),, based on the longitudinal protein expression profiles of all 182 infants (Fig. 5a, Supplementary Fig. 6a and 6b). Visualizing all 1335 samples the UMAP results revealed a distinct and stereotypic evolution of blood protein profiles from birth to term-equivalent age. The majority of the infants' proteome followed a predestined pathway, regardless of sex and mode of delivery (Supplementary Fig. 7a and 7b). Correspondingly, the results of principal component analysis (PCA) (Supplementary Fig. 8a) and the related diffusion map (Supplementary Fig. 8b) both demonstrated similar results with the samples following a clear pattern based on time since birth.

Interestingly, the protein profile trajectory was most coherent right at birth and at full term (Fig. 5b). Moreover, the most pronounced diversity in protein expression was observed at one week postnatal age. This suggests that the infants start life with similar protein profiles, followed by an interval where internal or external factors might be more influential before most infants converge their protein expressions again.

To examine the observed protein evolution in relation to neonatal immaturity, the UMAP result was investigated based on GA at birth as a proxy for fetal maturation. As seen in Fig. 5c, GA at delivery seemed to be of minor importance for protein expression at birth, as no distinction between GA-groups can be seen on the first day of life. However, the degree of immaturity plays an increasing part in the differentiation, with a peak around one to two weeks after birth where the samples are clearly separated depending on GA at birth. Further on, the profiles converge once again as the infants grow older. From PMA 30 weeks, and especially 32 and persistently to 40 weeks PMA, no separation can be seen between GA-groups. In addition, comparing the growth trajectories of infants in three GA groups, it was observed that those with smaller GA tended to have a slower rate of growth (Fig. 5d).

Gestational effects on blood protein profiling

The importance of GA on protein expression was additionally illustrated in Fig. 6a, where the numbers of proteins affected by the factors GA, sex, and delivery mode are presented per sampling time point. Consistent with Fig. 5c, GA at birth is most influential at one-week of PNA, with a drastic decline at later postnatal ages. To further explore how the GA group differs, the differentially expressed proteins (DEPs) between GA groups on PNA day 7 were analyzed by ANOVA (Supplementary Data 10) and visualized in a volcano plot (Fig. 6b). In total, 86 DEPs were identified, with some examples of proteins with decreased or increased levels in the infants with more advanced GA (Fig. 6c). The top 30 most significant DEPs were further analyzed in the radar plots (Fig. 6d and Supplementary Fig. 9). This analysis revealed the same pattern as seen in Fig. 5b and 6a. The three GA groups have similar protein profiles at birth, diverge into clearly varying trends at one-week PNA but converge at full-term.

We report on a longitudinal study of postnatal blood protein profile development in extremely preterm infants in which “next generation blood proteome profiling” has been used. The fact that only a minimal blood volume is needed, without sacrificing accuracy or sensitivity, makes the assay platform ideal for analysis of such individuals. The combination of a large study cohort with an in-depth multiplex, high-throughput protein analysis platform, provides high-resolution insight into the complex physiology of preterm postnatal development. The study demonstrates a dynamic, stereotypic evolution of serum protein profiles from birth to full-term age. The pattern is consistent for the entire cohort, regardless of GA at birth, clinical differences, sex or growth. Protein variations were highly associated with postnatal age rather than any other factor, demonstrating the immense physiological impact of intrauterine-extrauterine transition. In addition, we show that the first weeks after birth is a period where protein expression differs most between individuals and the degree of immaturity seems to significantly impact the proteome, thus directing a window for potential diagnostics and therapeutics.

The fact that PNA is a more critical determinant than PMA for the development of the protein profiles might seem surprising, since in most neonatal clinical settings, the medical decisions, care and treatment are based on PMA rather than the PNA. Our findings thus suggest that it might be relevant to reevaluate this mindset and to include the aspect of PNA in the assessment of each patient. Our study suggests that the impact of GA at birth, a proxy used today for the degree of immaturity, is most prominent around one week after birth, but the effect of GA is heavily reduced after 30 weeks postmenstrual age. These observations are interesting since the infants born at lower GAs are more likely to be longer on parenteral nutrition, have a higher need for ventilator support and oxygen treatment, receive transfusions, and be subject to severe infections like sepsis³⁴. One might speculate that the altered protein expression at one week of PNA may play a causal role, where the more immature infants may struggle to activate specific pathways for homeostasis, which leads to dysregulation and pathology. The reverse may also be true; several interventions such as transfusion with adult blood products, broad-spectrum antibiotics, and parenteral nutrition might influence the protein expression.

An analysis of the proteins with significantly higher expression in the most immature infants show that many of these proteins are involved in skin development, pulmonary maturation, myelinization, metabolism, and infection/inflammation. In contrast, the more mature group of infants have higher expression of proteins involved in osteogenesis, immune system maturation, and angiogenesis. This suggests that the more immature infants express proteins associated with reactive processes, such as inflammation, while the more mature infants express higher levels of proteins related to the development of the immune system.

A clear pattern of blood protein changes during the weeks after birth is evident in these preterm infants. Here, we have classified the most variable proteins into eight clusters, each showing a particular blood protein level trend during the weeks and months after birth, with five groups of proteins (clusters) displaying overall declining protein levels and three increasing trends after birth. Interestingly, many proteins with increased levels after birth have their origin in the liver, thus indicating the development of hepatic functions during the neonatal period. Similarly, the analysis of the proteins with origin from immune cells suggests that proteins from the dendritic cells increase after birth, while the expression of proteins from basophils and neutrophils decrease.

Overall, the current study has not found any strong correlation between protein levels in blood with sex or delivery mode. In most analyses, the sex had no or negligible effect on protein profiles. Nonetheless, a handful of proteins were subject to sex-related differences, exemplified with the lower level of protein CGA in male subjects. Similarly, delivery mode was found to influence a limited number of proteins, of which the protein PSP-D showed the highest correlation between blood protein levels and delivery mode. Thus, cesarean-delivered infants had elevated levels of PSP-D as compared to vaginally delivered infants which is interesting since PSP-D, a lung-specific protein that participates in the innate immune defense of the lungs, and this protein is considered a marker of lung injury. In this context, it is noteworthy that several earlier studies have demonstrated an increased risk for pulmonary disease after a cesarean section³⁷. The data presented here thus supports earlier reports of associations between this protein and severe pediatric acute respiratory distress syndrome (PARDS) and bronchopulmonary dysplasia (BPD)^35,36. Elevated levels of PSP-D have been described as indicating more severe PARDS, whereas low levels seem to predict worse pulmonary outcomes in BPD. Our findings support the suggestion that the administration of recombinant human PSP-D could be a promising therapy to prevent lung injury and BPD as suggested earlier³⁶.

Here, we report on a longitudinal study based on extremely preterm infants showing a consistent and stereotypic evolution of the blood protein profiles during the weeks and months after birth, mostly independent of gestational age, sex and delivery. This comprehensive analysis of blood proteins contributes to filling the knowledge gap regarding the expression of proteins in extremely preterm infants in their transition from immaturity to term equivalent age with clinical implications for guiding the diagnostic platform to improve the treatment regime of these infants.

Patients and nutritional management

The current study is based on the multicenter, open-label, randomized controlled trial MegaDonnaMega (Clinical Trial.gov identifier NCT03201588). Details of the MegaDonnaMega-study are described elsewhere¹⁹. In summary, infants born before 28 weeks of gestation and treated at the neonatal intensive care unit (NICU) in Gothenburg, Lund, or Stockholm, Sweden, between December 2016 and December 2019, were randomized to receive the triglyceride oil supplement Formulaid→ (DSM Nutritional Products Inc) containing arachidonic acid and docosahexaenoic acid or no extra supplement/standard care. Randomization was stratified according to the center and three GA groups: less than 25 weeks, 25 to 26 weeks, or 27 weeks. Twins or triplets were randomized to the same group. The supplement emulsion was administrated orally from within 72 hours of age to term-equivalent age. Written informed consent was obtained from the parents or guardians before inclusion.

The MegaDonnaMega-study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline³⁴. The regional ethical board of Gothenburg approved the MegaDonnaMega-study, and the Swedish Ethical Review Authority approved this extended study. The current study cohort included all infants in MegaDonnaMega with available longitudinal blood samples up to full term age²⁰.

Sample and medical data collection

Blood samples were taken in line with clinical medical needs at birth, postnatal day three, seven, fourteen, twenty-eight, followed by samples taken at PMW 30, 32, 36, and term-equivalent age corresponding to 40 weeks PMA. Whole blood was collected in serum separator tubes with clot activator and centrifuged once before storing the remaining serum at -70 degrees Celsius without delay. Clinical data regarding birth, growth, nutrition and morbidities were collected prospectively according to the study protocol.

Serum protein profiling

Serum proteins were analyzed using a multiplex proximity extension assay (PEA) technology (Olink Bioscience, Uppsala, Sweden) as previously described^9,13. Briefly, each kit consists of a microtiter plate for measuring 92 protein biomarkers in all 88 samples, and each well contained 96 pairs of DNA-labeled antibody probes. To minimize inter- and intra-run variation, the data were normalized using both an internal control (extension control) and an interplate control and then transformed using a pre-determined correction factor. I. This study uses six Olink panels including Cardiometabolic, Cardiovascular II, Cardiovascular III, Development, Inflammation, and Metabolism, resulting in 552 protein assays and 541 unique proteins. The pre-processed data were provided in the arbitrary unit NPX on a log2 scale, where a high NPX represents high protein concentration. Limit of detetion (LOD) for each protein was defined as three standard deviations above the background. Protein panels from samples with more than 10% under LOD values were removed from the analysis. Three proteins with drastic fluctuations between visits were considered to have a problematic batch effect and were removed. After quality control, 538 proteins from 1335 samples were kept.

Clustering analysis

For hierarchical clustering, the NPX values of each protein were first standardized by scaling with a standard deviation of 1 and centered at 0. The scaled values from all 1335 samples were used to create the Euclidean distance matrix for dendrogram generation. Dendrograms showing gene expression in heatmaps have been clustered using the Ward2 algorithm, an implementation of' ‘Ward's minimum variance method implemented as "Ward.D2" in R package pheatmap³⁵. UMAP analysis has been performed on NPX values of samples by using the R packages umap with default parameters. The diffusion map analysis was performed using the R package destiny³⁶ and the principal component analysis was performed using the prcomp function in the basic R package stats.

Time-series expression analysis

Mixed-effect modeling was performed using the lme4 package³⁷, and Kenward-Roger approximation³⁸ was used to calculate P-values which were subsequently adjusted for multiple testing based on false discovery rate. P-values were considered significant if less than 0.01. In total, 451 differentially expressed proteins across nine visits were identified. The average NPX of each protein per visit was used for the hierarchical clustering analysis to create the dendrogram based on Euclidean distance. Ward's minimum variance method implemented as "Ward.D2" in the hclust function in the R package stats, where clusters are chosen at each stage such that the increase in cluster variance is minimized after merging. As a result, eight clusters were identified. For regression analysis, the NPX values from each protein in the particular cluster were standardized by scaling with a standard deviation of 1 centered at 0. Non-parametric local weighted regression (LOESS) was applied to generate the regression curve. Gene Ontology term (Biological Process) and BioPlanet pathway enrichment analyses for proteins in different time-series clusters were conducted using EnrichR³⁹ based on Fisher exact test.

Prediction of postnatal age (PNA)

Generalized linear models with an elastic-net penalty was performed using glmnet⁴⁰ package in R. Specifically, the alpha value was set to 0.5, and 100 lambda values were tested. The "lambda.min" value, which is a measure of shrinkage, was determined after conducting a tenfold cross-validation analysis. Two-thirds (n = 934) of the samples were used for training the model, and the remaining samples (n = 401) were used as a validation.

Statistical analysis and visualization

Data analysis and visualization were performed using the R project⁴¹ with the tidyverse suite of R packages⁴² and the ggplot2 R package⁴³. Variance analysis of the protein levels was conducted using multiple linear regression model with all protein significantly associated PNA, GA, sex, and mode of delivery as variables in the model. The fraction of explained variability was measured using Eta-squared (η²), representing the proportion of the total sum of square explained by the factor, and was determined using analysis of variance (ANOVA) method with the build-in R function anova(). Differential expression analysis was conducted using ANOVA. The false discovery rate (FDR) was calculated using the p.adjust() function in R, which uses the Benjamini-Hochberg method. Proteins with FDRs less than 0.01 were considered differentially expressed proteins. Radar chart was generated using a R package fsmb⁴⁴.

Data availability

All summary statistics and association data are available in the supplementary material. The proteomic data of the preterm infant cohort is available in the BioStudies database (http://www.ebi.ac.uk/biostudies) under accession number S-BSST843.

Author Contributions

W. Z., H.D., N.B., L.F., M.U., and A.H., designed and analyzed the data and wrote the manuscript. A.K.N, and U.S. were responsible for sample collection, handled the samples, and approved the manuscript. D.W., I.P., K.S., D.L., provided data, and advice, performed the clinical data collection and approved the manuscript. The authors declare no competing financial interests.

Acknowledgments

We thank the Plasma Profiling Facility at SciLifeLab in Stockholm for conducting the Olink analyses and the nurses at Sahlgrenska University Hospital, Karolinska University Hospital and Lund University Hospital for sample and data collection and Chatarina Löfqvist who contributed to the sample handling before the measurements. We acknowledge the entire Human Protein Atlas program staff and the Science for Life Laboratory for their valuable contributions. We also thank Chatarina Löfqvist for sample handling and Mattias Karlén for the illustrations. Funding was provided from the Erling Persson Foundation, the Swedish Medical Research Council #2016-01131, the Swedish Research Council #2022-01562, Government grants under the ALF agreements ALFGBG-717971 and ALFGBG-812951 and The Wallenberg Clinical Scholars, LillaBarnetsFond and Spädbarnsfonden (Swedish infant death foundation). This work was supported by the SciLifeLab & Wallenberg Data Driven Life Science Program (grant: KAW 2020.0239).

Goldenberg, R. L., Culhane, J. F., Iams, J. D. & Romero, R. Epidemiology and causes of preterm birth. Lancet 371, 75–84 (2008). https://doi.org:10.1016/S0140-6736(08)60074-4
Saigal, S. & Doyle, L. W. An overview of mortality and sequelae of preterm birth from infancy to adulthood. Lancet 371, 261–269 (2008). https://doi.org:10.1016/S0140-6736(08)60136-1
Walani, S. R. Global burden of preterm birth. Int J Gynaecol Obstet 150, 31–33 (2020). https://doi.org:10.1002/ijgo.13195
Chawanpaiboon, S. et al. Global, regional, and national estimates of levels of preterm birth in 2014: a systematic review and modelling analysis. Lancet Glob Health 7, e37-e46 (2019). https://doi.org:10.1016/S2214-109X(18)30451-0
Cheong, J. L., Spittle, A. J., Burnett, A. C., Anderson, P. J. & Doyle, L. W. Have outcomes following extremely preterm birth improved over time? Semin Fetal Neonatal Med 25, 101114 (2020). https://doi.org:10.1016/j.siny.2020.101114
Blencowe, H. et al. National, regional, and worldwide estimates of preterm birth rates in the year 2010 with time trends since 1990 for selected countries: a systematic analysis and implications. Lancet 379, 2162–2172 (2012). https://doi.org:10.1016/S0140-6736(12)60820-4
Morgan, A. S. et al. Birth outcomes between 22 and 26 weeks' gestation in national population-based cohorts from Sweden, England and France. Acta Paediatr 111, 59–75 (2022). https://doi.org:10.1111/apa.16084
Glass, H. C. et al. Outcomes for extremely premature infants. Anesth Analg 120, 1337–1351 (2015). https://doi.org:10.1213/ANE.0000000000000705
Zhong, W. et al. Dramatic changes in blood protein levels during the first week of life in extremely preterm infants. Pediatr Res (2020). https://doi.org:10.1038/s41390-020-0912-8
Danielsson, H. et al. Blood protein profiles related to preterm birth and retinopathy of prematurity. Pediatr Res (2021). https://doi.org:10.1038/s41390-021-01528-0
Ignjatovic, V. et al. Mass Spectrometry-Based Plasma Proteomics: Considerations from Sample Collection to Achieving Translational Data. J Proteome Res 18, 4085–4097 (2019). https://doi.org:10.1021/acs.jproteome.9b00503
Zhong, W. et al. Next generation plasma proteome profiling to monitor health and disease. Nature Communications (2021).
Assarsson, E. et al. Homogenous 96-plex PEA immunoassay exhibiting high sensitivity, specificity, and excellent scalability. PLoS One 9, e95192 (2014). https://doi.org:10.1371/journal.pone.0095192
Hellstrom, W., Martinsson, T., Hellstrom, A., Morsing, E. & Ley, D. Fetal haemoglobin and bronchopulmonary dysplasia in neonates: an observational study. Arch Dis Child Fetal Neonatal Ed 106, 88–92 (2021). https://doi.org:10.1136/archdischild-2020-319181
Hellstrom, W., Forssell, L., Morsing, E., Savman, K. & Ley, D. Neonatal clinical blood sampling led to major blood loss and was associated with bronchopulmonary dysplasia. Acta Paediatr 109, 679–687 (2020). https://doi.org:10.1111/apa.15003
Crawford, T. M., Andersen, C. C., Hodyl, N. A., Robertson, S. A. & Stark, M. J. The contribution of red blood cell transfusion to neonatal morbidity and mortality. J Paediatr Child Health 55, 387–392 (2019). https://doi.org:10.1111/jpc.14402
Olin, A. et al. Stereotypic Immune System Development in Newborn Children. Cell 174, 1277–1292 e1214 (2018). https://doi.org:10.1016/j.cell.2018.06.045
Lietzen, N. et al. Characterization and non-parametric modeling of the developing serum proteome during infancy and early childhood. Sci Rep 8, 5883 (2018). https://doi.org:10.1038/s41598-018-24019-5
Hellstrom, A. et al. Effect of Enteral Lipid Supplement on Severe Retinopathy of Prematurity: A Randomized Clinical Trial. JAMA Pediatr 175, 359–367 (2021). https://doi.org:10.1001/jamapediatrics.2020.5653
Fenton, T. R. & Kim, J. H. A systematic review and meta-analysis to revise the Fenton growth chart for preterm infants. BMC Pediatr 13, 59 (2013). https://doi.org:10.1186/1471-2431-13-59
Wu, C. T., Chaffin, A. T. & Ryan, K. K. Fibroblast Growth Factor 21 Facilitates the Homeostatic Control of Feeding Behavior. J Clin Med 11 (2022). https://doi.org:10.3390/jcm11030580
Uhlen, M. et al. Proteomics. Tissue-based map of the human proteome. Science 347, 1260419 (2015). https://doi.org:10.1126/science.1260419
Uhlen, M. et al. A genome-wide transcriptomic analysis of protein-coding genes in human blood cells. 366, eaax9198 (2019). https://doi.org:10.1126/science.aax9198%J Science
Her, L. & Zhu, H. J. Carboxylesterase 1 and Precision Pharmacotherapy: Pharmacogenetics and Nongenetic Regulators. Drug Metab Dispos 48, 230–244 (2020). https://doi.org:10.1124/dmd.119.089680
Acharya, M. et al. CD23/FcepsilonRII: molecular multi-tasking. Clin Exp Immunol 162, 12–23 (2010). https://doi.org:10.1111/j.1365-2249.2010.04210.x
Uhlig, R. et al. Carboxypeptidase A1 (CPA1) Immunohistochemistry Is Highly Sensitive and Specific for Acinar Cell Carcinoma (ACC) of the Pancreas. Am J Surg Pathol 46, 97–104 (2022). https://doi.org:10.1097/PAS.0000000000001817
Wolf, H. K., Zarnegar, R., Oliver, L. & Michalopoulos, G. K. Hepatocyte growth factor in human placenta and trophoblastic disease. Am J Pathol 138, 1035–1043 (1991).
Kreicberga, I., Junga, A. & Pilmane, M. Assessment of apoptosis and appearance of hepatocyte growth factor in placenta at different gestational ages: A cross-sectional study. Int J Reprod Biomed 19, 505–514 (2021). https://doi.org:10.18502/ijrm.v19i6.9372
Buschmann, K. et al. RAGE controls leukocyte adhesion in preterm and term infants. BMC Immunol 15, 53 (2014). https://doi.org:10.1186/s12865-014-0053-0
Wang, Y., Kim, K. A., Kim, J. H. & Sul, H. S. Pref-1, a preadipocyte secreted factor that inhibits adipogenesis. J Nutr 136, 2953–2956 (2006). https://doi.org:10.1093/jn/136.12.2953
Charalambous, M. et al. DLK1/PREF1 regulates nutrient metabolism and protects from steatosis. Proc Natl Acad Sci U S A 111, 16088–16093 (2014). https://doi.org:10.1073/pnas.1406119111
Wood, J. P., Ellery, P. E., Maroney, S. A. & Mast, A. E. Biology of tissue factor pathway inhibitor. Blood 123, 2934–2943 (2014). https://doi.org:10.1182/blood-2013-11-512764
Nwabuobi, C. et al. hCG: Biological Functions and Clinical Applications. Int J Mol Sci 18 (2017). https://doi.org:10.3390/ijms18102037
Schulz, K. F., Altman, D. G., Moher, D. & Group, C. CONSORT 2010 statement: updated guidelines for reporting parallel group randomised trials. PLoS Med 7, e1000251 (2010). https://doi.org:10.1371/journal.pmed.1000251
Kolde, R. & Kolde, M. R. Package ‘pheatmap’. R package 1, 790 (2015).
Angerer, P. et al. destiny: diffusion maps for large-scale single-cell data in R. Bioinformatics 32, 1241–1243 (2016). https://doi.org:10.1093/bioinformatics/btv715
Bates, D., Mächler, M., Bolker, B. & Walker, S. Fitting linear mixed-effects models using lme4. arXiv preprint arXiv:1406.5823 (2014).
Halekoh, U. & Højsgaard, S. A kenward-roger approximation and parametric bootstrap methods for tests in linear mixed models–the R package pbkrtest. Journal of Statistical Software 59, 1–30 (2014).
Chen, E. Y. et al. Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool. BMC Bioinformatics 14, 128 (2013). https://doi.org:10.1186/1471-2105-14-128
Friedman, J., Hastie, T. & Tibshirani, R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw 33, 1–22 (2010).
Team, R. C. R: A language and environment for statistical computing. (2013).
Wickham, H. The tidyverse. R package ver 1, 1 (2017).
Wickham, H. ggplot2. Wiley Interdisciplinary Reviews: Computational Statistics 3, 180–185 (2011).
Nakazawa, M. fmsb: Functions for medical statistics book with some demographic data. R package version 0.5 2 (2018).

There is NO Competing Interest.

SupplementaryData110.xlsx
Supplementary Datasets
PretermSupplementarymaterialsfinal.docx

Download PDF

Journal Publication

published 02 Aug, 2023

Read the published version in Communications Medicine →

Version 1

posted

You are reading this latest preprint version

The development of blood protein profiles in extremely preterm infants follows a stereotypic evolution pattern

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Results

The study cohort

Proteome profiling demonstrated dynamic changes in blood proteins after birth

Functional analysis of the protein clusters

Variance analysis of blood protein profiling after birth

Distinct and coherent evolution of blood protein profiles over time

Gestational effects on blood protein profiling

Discussion

Materials And Methods

Patients and nutritional management

Sample and medical data collection

Serum protein profiling

Clustering analysis

Time-series expression analysis

Prediction of postnatal age (PNA)

Statistical analysis and visualization

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1