1. Epidemiology and Global Health

Early life infection and proinflammatory, atherogenic metabolomic and lipidomic profiles in infancy: a population-based cohort study

  1. Toby Mansell
  2. Richard Saffery
  3. Satvika Burugupalli
  4. Anne-Louise Ponsonby
  5. Mimi LK Tang
  6. Martin O'Hely
  7. Siroon Bekkering
  8. Adam Alexander T Smith
  9. Rebecca Rowland
  10. Sarath Ranganathan
  11. Peter D Sly
  12. Peter Vuillermin
  13. Fiona Collier
  14. Peter Meikle
  15. David Burgner  Is a corresponding author
  16. Barwon Infant Study Investigator Group
  1. Murdoch Children's Research Institute, Australia
  2. Department of Paediatrics, University of Melbourne, Australia
  3. Metabolomics Laboratory, Baker Heart and Diabetes Institute, Australia
  4. The Florey Institute of Neuroscience and Mental Health, Australia
  5. Royal Children’s Hospital, Australia
  6. Deakin University, Australia
  7. Department of Internal Medicine, Radboud Institute for Molecular Life Sciences, Radboud University Medical Centre, Netherlands
  8. Child Health Research Centre, University of Queensland, Australia
  9. Child Health Research Unit, Barwon Health, Australia
  10. Department of Paediatrics, Monash University, Australia
https://doi.org/10.7554/eLife.75170
Share this article
Cite this article
  1. Toby Mansell
  2. Richard Saffery
  3. Satvika Burugupalli
  4. Anne-Louise Ponsonby
  5. Mimi LK Tang
  6. Martin O'Hely
  7. Siroon Bekkering
  8. Adam Alexander T Smith
  9. Rebecca Rowland
  10. Sarath Ranganathan
  11. Peter D Sly
  12. Peter Vuillermin
  13. Fiona Collier
  14. Peter Meikle
  15. David Burgner
  16. Barwon Infant Study Investigator Group
(2022)
Early life infection and proinflammatory, atherogenic metabolomic and lipidomic profiles in infancy: a population-based cohort study
eLife 11:e75170.
https://doi.org/10.7554/eLife.75170
  2. Download BibTeX
  3. Download .RIS

Abstract

Background:

The risk of adult onset cardiovascular and metabolic (cardiometabolic) disease accrues from early life. Infection is ubiquitous in infancy and induces inflammation, a key cardiometabolic risk factor, but the relationship between infection, inflammation, and metabolic profiles in early childhood remains unexplored. We investigated relationships between infection and plasma metabolomic and lipidomic profiles at age 6 and 12 months, and mediation of these associations by inflammation.

Methods:

Matched infection, metabolomics, and lipidomics data were generated from 555 infants in a pre-birth longitudinal cohort. Infection data from birth to 12 months were parent-reported (total infections at age 1, 3, 6, 9, and 12 months), inflammation markers (high-sensitivity C-reactive protein [hsCRP]; glycoprotein acetyls [GlycA]) were quantified at 12 months. Metabolic profiles were 12-month plasma nuclear magnetic resonance metabolomics (228 metabolites) and liquid chromatography/mass spectrometry lipidomics (776 lipids). Associations were evaluated with multivariable linear regression models. In secondary analyses, corresponding inflammation and metabolic data from birth (serum) and 6-month (plasma) time points were used.

Results:

At 12 months, more frequent infant infections were associated with adverse metabolomic (elevated inflammation markers, triglycerides and phenylalanine, and lower high-density lipoprotein [HDL] cholesterol and apolipoprotein A1) and lipidomic profiles (elevated phosphatidylethanolamines and lower trihexosylceramides, dehydrocholesteryl esters, and plasmalogens). Similar, more marked, profiles were observed with higher GlycA, but not hsCRP. GlycA mediated a substantial proportion of the relationship between infection and metabolome/lipidome, with hsCRP generally mediating a lower proportion. Analogous relationships were observed between infection and 6-month inflammation, HDL cholesterol, and apolipoprotein A1.

Conclusions:

Infants with a greater infection burden in the first year of life had proinflammatory and proatherogenic plasma metabolomic/lipidomic profiles at 12 months of age that in adults are indicative of heightened risk of cardiovascular disease, obesity, and type 2 diabetes. These findings suggest potentially modifiable pathways linking early life infection and inflammation with subsequent cardiometabolic risk.

Funding:

The establishment work and infrastructure for the BIS was provided by the Murdoch Children’s Research Institute (MCRI), Deakin University, and Barwon Health. Subsequent funding was secured from National Health and Medical Research Council of Australia (NHMRC), The Shepherd Foundation, The Jack Brockhoff Foundation, the Scobie & Claire McKinnon Trust, the Shane O’Brien Memorial Asthma Foundation, the Our Women’s Our Children’s Fund Raising Committee Barwon Health, the Rotary Club of Geelong, the Minderoo Foundation, the Ilhan Food Allergy Foundation, GMHBA, Vanguard Investments Australia Ltd, and the Percy Baxter Charitable Trust, Perpetual Trustees. In-kind support was provided by the Cotton On Foundation and CreativeForce. The study sponsors were not involved in the collection, analysis, and interpretation of data; writing of the report; or the decision to submit the report for publication. Research at MCRI is supported by the Victorian Government’s Operational Infrastructure Support Program. This work was also supported by NHMRC Senior Research Fellowships to ALP (1008396); DB (1064629); and RS (1045161) , NHMRC Investigator Grants to ALP (1110200) and DB (1175744), NHMRC-A*STAR project grant (1149047). TM is supported by an MCRI ECR Fellowship. SB is supported by the Dutch Research Council (452173113).

Editor's evaluation

This paper provides data from a population-based cohort study on early life infection and proinflammatory, atherogenic metabolomic and lipidomic profiles at 12 months of age. The authors generated matched infection, metabolomics and lipidomics data from 555 infants in a pre-birth longitudinal cohort and they showed that frequent infant infections are associated with adverse metabolomic and lipidomic profiles. They also report that similar profiles are noted with higher glycoprotein acetyls (GlycA), but not hsCRP. The paper is interesting and assesses the role of infection and markers of inflammation on lipid and metabolic profile of patients. It provides a comprehensive analysis of lipids and metabolites in infants in response to infection.

https://doi.org/10.7554/eLife.75170.sa0

Introduction

Infectious diseases are ubiquitous in infancy and childhood, with potential long-term impacts on health across the life course. Infection has been recognised as a potential contributor to atherosclerotic cardiovascular disease (CVD), one of the leading causes of adult morbidity and mortality, since the 19th century (Nieto, 1998). More recent adult studies link previous infection with long-term risks of disease (Bergh et al., 2017; Cowan et al., 2018; Wang et al., 2017). The mechanisms are largely unknown, but likely include immune activation and heightened inflammation (Shah, 2019), which are pathways central to CVD pathogenesis (Donath et al., 2019b; Ferrucci and Fabbri, 2018) and therefore offer potentially druggable targets in high-risk individuals (Donath et al., 2019a; Ridker et al., 2017). High-sensitivity C-reactive protein (hsCRP) has been extensively used as a marker of chronic inflammation in adult studies but is an acute phase reactant in children and may not reflect chronic inflammation in early life. Glycoprotein acetyls (GlycA) is a nuclear magnetic resonance (NMR) composite measure that is suggested to better reflect cumulative, chronic inflammation (Connelly et al., 2017). GlycA is an emerging biomarker for cardiometabolic risk (Connelly et al., 2017) that outperforms hsCRP as a predictor of CVD events and mortality (Akinkuolie et al., 2014; Duprez et al., 2016), and of infection-related morbidity and mortality (Ritchie et al., 2015). For example, in the Multi-Ethnic Study of Atherosclerosis (n = 6523), higher GlycA was associated with increased risk of incidence CVD and death, even after adjustment for hsCRP and other inflammatory markers. Conversely, prediction of these outcomes by hsCRP attenuated to null after mutual adjustment (Duprez et al., 2016).

Cardiovascular and metabolic (cardiometabolic) disease pathogenesis begins in early life and accrues across the life course (Nakashima et al., 2008). Infections occur disproportionally in early childhood (Cromer et al., 2014; Grüber et al., 2008; Troeger et al., 2018; Tsagarakis et al., 2018), and there is a dose-response relationship between childhood infections, adverse cardiometabolic phenotypes (Burgner et al., 2015c), and CVD events (Burgner et al., 2015a) in adulthood. Infection is linked to proatherogenic metabolic perturbations in later childhood and adulthood (Feingold and Grunfeld, 2019; Khovidhunkit et al., 2000), including higher triglycerides and oxidised low-density lipoprotein (LDL), and lower high-density lipoprotein (HDL) cholesterol and apolipoprotein A1 (ApoA1) (Liuba et al., 2003; Pesonen et al., 1993), and to acute and chronic inflammation (Burgner et al., 2015b; Ritchie et al., 2015), but little is known about these relationships in early life, when most infections occur.

We therefore aimed to characterise metabolomic and lipidomic profiles at 6 and 12 months of age and their relationship to infection burden during the first year of life. We also investigated the extent to which inflammation mediated the relationship between infection burden and metabolomic and lipidomic differences.

Materials and methods

Study cohort

Request a detailed protocol

This study used available data from 555 mother-infant dyads in the Barwon Infant Study (BIS), a population-based pre-birth longitudinal cohort (n = 1074 mother-infant dyads). The cohort details and inclusion/exclusion criteria have been detailed elsewhere (Vuillermin et al., 2015); in brief, mothers were eligible if they were residents of the Barwon region in south-east Australia and planned to give birth at the local public or private hospital. Mothers were recruited at approximately 15 weeks’ gestation and provided informed consent. They were excluded if they were not a permanent Australian resident, aged <18 years, required an interpreter to complete questionnaires, or had previously participated in BIS. Infants were excluded if they were very preterm (<32 completed weeks gestation) or had a serious illness or major congenital malformation identified during the first few days of life. Ethics approval was granted by the Barwon Health Human Research Ethics Committee (HREC 10/24).

Parent-reported infections

Request a detailed protocol

At the 4-week, 3-month, 6-month, 9-month, and 12-month time points following birth, mothers were asked to report each episode of infant illness or infection since the previous time point using standardised online questionnaires. The number of parent-reported infections from birth to 12 months was defined as the total number of respiratory tract infections, gastroenteritis, conjunctivitis, and acute otitis media episodes. In secondary analyses, numbers of parent-reported infections from birth to 6 months and from 6 to 12 months were considered. It was not possible to identify the proportion of parent-reported infections that lead to health service utilisation (Rowland et al., 2021).

Other maternal and infant measures

Request a detailed protocol

Questionnaires during pregnancy and at birth were used to collect self-reported data on maternal age, household income, maternal education, and prenatal smoking (considered here as a dichotomous any/none exposure). Residential postcode was used to determine neighbourhood disadvantage using the Index of Relative Socio-Economic Disadvantage (IRSD) from the 2011 Socio-Economic Indexes for Areas (SEIFA) (Pink, 2013), with a lower score corresponding to greater socioeconomic disadvantage. Pre-eclampsia (based on International Association of Diabetes and Pregnancy Study Groups criteria; Tranquilli et al., 2014) and gestational diabetes (based on International Society for the Study of Hypertension in Pregnancy criteria; Nankervis et al., 2013) diagnoses were extracted from hospital records. Infant gestational age, birth weight, and mode of delivery (categorised as vaginal, planned caesarean section, or unplanned caesarean section delivery) were collected from birth records, and the age- and sex-standardised birth weight z-score was calculated using the 2009 revised British United Kingdom World Health Organisation (UK-WHO) growth charts (Cole et al., 2011). Postnatal smoking data was collected from questionnaire data, with mothers asked the average number of hours each day someone smoked near or in the same room as the child (Gray et al., 2019). This was dichotomised as any postnatal smoke exposure if >0 hr reported at any time point up to 12 months of age, or no postnatal smoke exposure. Breastfeeding duration up to 12 months of age was collected from maternal questionnaire data. As most evidence for the protective effect of breastfeeding on early life infection is from comparisons between any breastfeeding and no breastfeeding (Victora et al., 2016), and in light of previous evidence in BIS for an association between even a short duration of breastfeeding and lower odds of infection in early infancy (Rowland et al., 2021), we first looked at breastfeeding as a binary (any/none) measure in models (presented in the main text). As most infants (98.2%) were breastfed to some extent, and it is unknown the degree to which breastfeeding, and the timing of breastfeeding, might affect infant metabolomics and lipidomics, we also considered duration of breastfeeding as a continuous variable for sensitivity analyses.

Metabolomic and lipidomic profiling

Request a detailed protocol

Venous peripheral blood was collected from infants at the 6- and 12-month time points in sodium heparin and generally processed within 4 hr, with a minority (197 of 555) of 12-month samples processed after 4 hr (median time for those 197 samples = 19.9 hr, inter-quartile range [IQR] [18.7, 21.4]). The time interval between collection and post-processing storage of samples was included as a covariate in analyses. Due to the bimodal distribution of 12-month sample collection times as samples were either processed same day of collection or the following day, sensitivity analysis excluding participants with a plasma storage time greater than 4 hr (197 out of 555 infants, predominantly processed the following day) was performed, as described in the Statistical analysis section below. Plasma was stored at –80°C, and aliquots were shipped on dry ice to Nightingale Health (Helsinki, Finland) for NMR metabolomic quantification and Baker IDI (Melbourne, Australia) for liquid chromatography/mass spectrometry (LC/MS) lipidomic quantification, as described below. For secondary analyses investigating possible ‘reverse causality’, that is, whether metabolomic or lipidomic profile at birth was associated with number of parent-reported infections from birth to 6 months of age, metabolomics and lipidomics data using the same platforms from venous cord blood collected at birth, as previously described (Burugupalli et al., 2022; Mansell et al., 2021), was used.

The NMR-based metabolomics platform has been described in detail (Kettunen et al., 2016; Soininen et al., 2015), and quantified a broad range of metabolic measures including lipoprotein size subclasses, triglycerides, cholesterols, fatty acids, amino acids, ketone bodies, glycolysis metabolites, and GlycA. In brief, plasma samples were mixed with a sodium phosphate buffer prior to NMR measurements with Bruker AVANCE III 500 MHz and Bruker AVANCE III HD 600 MHz spectrometers (Bruker, Billerica, MA). Samples were kept at 6°C using the SampleJet sample changer (Bruker) to prevent degradation. After initial measurements, samples went through a multiple-step lipid extraction procedure using saturated sodium chloride solution, methanol, dichloromethane, and deuterochloroform. The lipid extracts were then analysed using the 600 MHz instrument (Soininen et al., 2015). The utility of this platform in epidemiological research has been detailed elsewhere (Würtz et al., 2017). Using the Nightingale Health 2016 bioinformatics protocol, 228 metabolomic measures were generated for the 12-month samples. From a subset of participants, replicate samples were quantified, and these showed a low percentage coefficient of variation (<10%). Subsequently, the Nightingale Health 2020 bioinformatics protocol was used to generate 250 metabolomic measures for the 6-month samples, with 224 of these measures also present in the 12-month data. As a large proportion of the NMR metabolomic measures are ratios and are strongly correlated with each other in children and adults (Ellul et al., 2019), an informative subset of 51 measures that captured the majority of variation in the metabolomic dataset, primarily absolute metabolite concentrations, were included in analysis presented in the main text. Analyses for excluded metabolomic measures are presented as supplementary data. To complement GlycA as a measure of inflammation, hsCRP was also quantified in 6- and 12-month plasma using enzyme-linked immunosorbent assay (ELISA) (R&D Systems, Minneapolis, MN, cat. no. DY1707), as per the manufacturer’s instruction.

The details of the high-performance LC/MS lipidomics platform have been described elsewhere (Beyene et al., 2020). In addition, we used medronic acid to passivate the LC/MS system to avoid peak tailing for acidic phospholipids (Hsiao et al., 2018). In brief, this platform quantified 776 lipid features in 36 lipid classes, including sphingolipids, glycerophospholipids, sterols, glycerolipids, and fatty acyls. Analysis was performed on an Agilent 6490 QQQ mass spectrometer with an Agilent 1290 series high-performance liquid chromatography system and two ZORBAX eclipse plus C18 column (2.1 × 100 mm 1.8 mm) (Agilent, Santa Clara, CA) with the thermostat set at 45°C. Mass spectrometry analysis was performed in both positive and negative ion mode with dynamic scheduled multiple reaction monitoring.

Quantification of lipid species was determined by comparison to the relevant internal standard. Lipid class total concentrations were calculated as the sum of individual lipid species concentrations, except in the case of triacylglycerols (TGs) and alkyl-diacylglycerols, where we measured both neutral loss and single ion monitoring (SIM) peaks, and subsequently used the SIM species concentrations for summation purposes.

Statistical analysis

Request a detailed protocol

Analyses were performed in R (version 3.6.3) (R Development Core Team, 2018). All metabolomic and lipidomic measures had their lowest observed non-zero value (considered the lower limit of detection) added to their value before they were natural log-transformed and scaled to a standard distribution (standard deviation units). Pearson’s correlations were calculated for number of infections from birth to 12 months with 12-month GlycA and hsCRP.

The estimated effect of number of parent-reported infections from birth to 12 months as an exposure on 12-month metabolomic and lipidomic profile was investigated using linear regression models for each metabolomic/lipidomic measure. Standard errors were used to calculate 95% confidence intervals for estimated effects. All models were adjusted for infant sex, exact age at 12-month time point, birth weight z-score, gestational age, maternal household income, exposure to maternal smoking during pregnancy, breastfeeding, and time from collection to storage for the plasma sample. Linear regression models adjusted for the same covariates were used to investigate 12-month GlycA and hsCRP as exposures and each metabolomic/lipidomic measure as an outcome. Two-tailed p-values were adjusted for multiple comparisons within each dataset (NMR metabolomics, LC/MS lipidomic species, and LC/MS lipidomic classes) using the Benjamini-Hochberg method (Benjamini and Hochberg, 1995). To investigate the robustness of the estimates, mean model coefficients and bias-corrected accelerated percentile bootstrap confidence intervals were calculated from nonparametric bootstrap resampling (1000 iterations) using the ‘boot’ package (Davison and Hinkley, 1997) (version 1.3–25) in R, included in Source Data. The assumption of linearity was investigated post hoc using plots of residual values for the top 10 metabolomic and lipidomic differences (ranked by p-value) for each of number of infections, GlycA, and hsCRP.

To further investigate the other sources of potential confounding and variation that could affect findings from the primary models, several sensitivity analyses were performed. These were: (i) additional adjustment of the primary model for postnatal smoking exposure, gestational diabetes, and pre-eclampsia; (ii) analyses excluding twins (five infants); (iii) analyses excluding infants with hsCRP >5 mg/L (24 infants) as a marker of acute infection (Lemiengre et al., 2018; Verbakel et al., 2016); (iv) analyses excluding plasma samples with >4 hr from collection to storage (197 samples); and (v) analyses adjusting for breastfeeding duration instead of any breastfeeding. Analyses of the primary models adjusted for different measures of socioeconomic position (SEIFA or maternal education) instead of household income were also considered. These models are presented as supplementary forest plots (Supplementary file 1A-1F).

Secondary analyses to investigate 6-month periods of infections were performed. The birth to 6-month models investigated the relationship between infections up to 6 months of age, 6-month inflammation, and 6-month metabolomic/lipidomic measures. The 6- to 12-month models investigated the relationship of infections between 6 and 12 months of age, 12-month inflammation, and 12-month metabolomic/lipidomic measures, with adjustment for the corresponding 6-month measures. For secondary analyses to investigate reverse causality, quasi-Poisson regression models (Zeileis et al., 2008) were used, either for (i) each metabolomic/lipidomic measure from cord blood at birth as exposure and total number of infections from birth to 6 months of age as outcome or (ii) each 6-month metabolomic/lipidomic measures as exposure and total number of infections from 6 to 12 months of age as the outcome, with adjustment for number of infections from birth to 6 months of age. All secondary models were adjusted for the same covariates as the primary models described above: that is, infant sex, gestational age, exact age (at 6- or 12-month time point), birth weight z-score, maternal household income, exposure to maternal smoking during pregnancy, and sample time between collection and post-processing storage. Models with birth metabolomic/lipidomic measures as the exposure were additionally adjusted for mode of delivery, which is associated with differences across many NMR metabolomic measures in cord serum in this cohort (Mansell et al., 2021).

To investigate the potential role of inflammation in mediating associations between infection and metabolic differences, the ‘medflex’ package (Steen et al., 2017) (version 0.6–7) in R was then used in a counterfactual-based framework to estimate the natural direct effect (not mediated by inflammation) and natural indirect effect (mediated by inflammation) of infection. Specifically, 12-month GlycA or hsCRP were considered separately as mediators for the effect of number of infections from birth to 12 months of age on metabolomic and lipidomic differences at 12 months, for all metabolomic and lipidomic measures associated with number of infections with an adjusted p-value < 0.1 from the linear regression models described above. Percentage mediation was calculated as the estimated natural indirect effect divided by the total effect (natural direct effect plus natural indirect effect). A representative directed acyclic graph (DAG) of the mediation model is shown in Figure 1.

Figure 1
Download asset Open asset
Representative directed acyclic graph (DAG) for causal model investigated in this study.

The natural indirect effect (mediated by glycoprotein acetyls [GlycA] or high-sensitivity C-reactive protein [hsCRP]) and natural direct effect (not mediated by GlycA/hsCRP) of parent-reported infections on metabolomic and lipidomic measures were calculated, with adjustment for confounders. Confounders were considered to be confounders for all associations (exposure to outcome, exposure to mediator, and mediator to outcome).

As models investigating reverse causality, described above, suggested that 6-month GlycA or lipidomic measures may influence number of infections from 6 to 12 months of age, we additionally performed mediation analyses to estimate the percentage mediation of 12-month GlycA and hsCRP for the effect of number of infections from 6 to 12 months of age on 12-month metabolomic and lipidomic differences, with additional adjustment for the corresponding 6-month measures (number of infections from birth to 6 months of age, 6-month GlycA or hsCRP, and the 6-month metabolomic/lipidomic measure).

Results

The flowchart for the 555 infants included in this study is shown in Figure 2, and the cohort characteristics for these infants are shown in Table 1. The median number of total parent-reported infections from birth to 12 months of age was 5 (IQR = [3–8]). Median infections from birth to 6 months of age was 2 [1–3], and median infections from 6 to 12 months of age was 3 [2–5]. Median 12-month hsCRP and GlycA were 0.25 mg/L [0.08–0.96] and 1.30 mmol/L [1.16–1.48], respectively. Total number of parent-reported infections between birth and 12 months of age was more strongly correlated with 12-month GlycA (r = 0.20) than hsCRP (r = 0.11). The distributions of metabolomic and lipidomic measures at each time point for the cohort are shown in Supplementary file 2A and B.

Table 1
Cohort characteristics (n = 555).
Measure
Sex (female)269 (48.4)
Mean (SD)
Maternal age at delivery (years)31.7 (4.5)
n (%)
Maternal smoking during pregnancy (any)69 (12.5)
Gestational diabetes (cases) (n = 84 missing data)26 (5.5)
Pre-eclampsia (cases) (n = 1 missing data)21 (3.8)
Maternal annual household income (AUD) 
 <$25,00011 (2.0)
 $25,000 to $49,99941 (7.5)
 $50,000 to $74,99994 (17.2)
 $75,000 to $99,999145 (26.5)
 $100,000 to $149,999191 (34.9)
 ≥$150,00065 (11.9)
Maternal education (highest level completed) (n = 8 missing data) 
 Less than year 10 of high school2 (0.4)
 Year 10 of high school or equivalent20 (3.7)
 Year 12 of high school or equivalent86 (15.7)
 Trade/certificate/diploma135 (24.9)
 Bachelor’s degree199 (36.4)
 Postgraduate degree105 (19.2)
Mode of birth 
 Vaginal374 (67.4)
 Planned caesarean section103 (18.6)
 Unplanned caesarean section78 (14.1)
Breastfed (any breastfeeding)545 (98.2)
Infant postnatal smoke exposure to 12 months (any) (n = 47 missing data)14 (2.8)
Median [IQR]
SEIFA index of disadvantage*1031 [996–1066]
Breastfeeding duration to 52 weeks (weeks)40 [16–52]
 
Mean (SD)
Gestational age (weeks)39.5 (1.5)
Birth weight (g)3,538 (521)
Birth weight z-score0.32 (0.95)
Age at 6-month time point (months)6.5 (0.4)
Weight at 6 months (kg)7.9 (1.0)
Weight z-score at 6 months0.1 (1.0)
Age at 12-month time point (months)13.0 (0.8)
Weight at 12 months (kg)10.1 (1.3)
Weight z-score at 12 months0.4 (1.0)
 
Median [IQR]
Number of parent-reported infections from birth to 12 months5 [3–8]
 Infections from birth to 6 months2 [1–3]
 Infection from 6 to 12 months3 [2–5]
 
GlycA at 6 months (mmol/L)0.76 [0.68–0.85]
hsCRP at 6 months (mg/L)0.14 [0.05–0.94]
GlycA at 12 months (mmol/L)1.30 [1.16–1.48]
hsCRP at 12 months (mg/L)0.25 [0.08–0.96]

  1. All n = 555 infants had complete covariate data for primary models, missing data of secondary exposures is indicated next to the relevant measure.

  2. *

    A lower SEIFA value indicates greater socioeconomic disadvantage based on postcode.

Figure 2
Download asset Open asset
Flowchart of Barwon Infant Study participants included this study (bolded boxes).

Included participants had complete infection data from all five time points between birth and 12 months of age, and 12-month plasma nuclear magnetic resonance (NMR) metabolomics data. Almost all included participants (n = 550 out of 555) had 12-month plasma liquid chromatography/mass spectrometry (LC/MS) lipidomics data.

Infection and inflammation burden and plasma NMR metabolomic profile at 12 months

There was evidence for higher number of infections associating with higher inflammatory markers (GlycA (0.06 SD per 1 infection, 95% CI [0.04–0.08]) and hsCRP (0.06 [0.03–0.08])), lower HDL (−0.04 [−0.07 to −0.02]), HDL2 (−0.04 [−0.07 to −0.02]), and HDL3 (−0.04 [−0.06 to −0.01]) cholesterols, lower ApoA1 (−0.04 [−0.06 to −0.01]), lower citrate (−0.04 [−0.07 to −0.01]), higher phenylalanine (0.04 [0.02–0.07]), and to a lesser extent with higher triglycerides (0.03 [0.00–0.05]) and lower sphingomyelins (−0.03 [−0.05 to −0.01]) (Figure 3a). In models with GlycA as the marker of inflammation burden, metabolomic differences observed for higher GlycA were largely similar to, but more marked than, those for parent-reported infections; including cholesterols (lower HDL (−0.38 SD per 1 SD increase in log GlycA [−0.46 to −0.30]), higher LDL (0.18 [0.09–0.26]), and very-large-density lipoprotein (0.50 [0.42–0.58]) cholesterols), apolipoproteins (lower ApoA1 (−0.23 [−0.31 to −0.14]), higher apolipoprotein B (ApoB) (0.39 [0.31–0.48])), higher total fatty acids (0.36 [0.28–0.45]), higher total triglycerides (0.48 [0.40–0.56]) and cholines (0.18 [0.10–0.27]), amino acids (higher phenylalanine (0.17 [0.09–0.25]), isoleucine (0.14 [0.05–0.22]), and glycine (0.16 [0.08–0.23]), lower histidine (−0.16 [−0.24 to −0.09])), glycolysis-related metabolites (higher pyruvate (0.06 [0.01–0.11]), lower citrate (−0.17 [−0.26 to −0.08])), and lower acetoacetate (−0.18 [−0.26 to −0.09]) (Figure 3b). Bootstrap estimates were generally similar to standard regression estimates for all models. Estimated effect sizes were generally similar across most sensitivity analyses (Supplementary file 1A-C), though excluding samples with processing time greater than 4 hr slightly reduced the magnitude of estimated effects of parent-reported infections on HDL cholesterols and ApoA1 (Supplementary file 1A). Higher hsCRP was associated with lower HDL cholesterol (−0.24 SD per 1 SD increase in log hsCRP, [−0.32 to −0.16]), ApoA1 (−0.26 [−0.34 to −0.18]) and histidine (−0.20 [−0.27 to −0.12]), and higher phenylalanine (0.14 [0.06 to 0.22]), as observed for GlycA, and with lower levels of most other amino acids and albumin (−0.11 [−0.19 to −0.03]), and higher LDL triglycerides (0.16 [0.08–0.25]) (Figure 4a).

Figure 3 with 2 supplements see all
Download asset Open asset
Difference in 12-month plasma nuclear magnetic resonance (NMR) metabolomic measures for each increase in parent-reported infection (birth to 12 months) and for each SD increase in 12-month glycoprotein acetyls (GlycA) (n = 555).

Forest plots of the estimated 12-month metabolomic differences for each additional parent-reported infection from birth to 12 months (a, circle points) or SD log 12-month GlycA (b, square points) from adjusted linear regression models, and the correlation of estimated metabolomic differences for these two exposures (c). Error bars are 95% confidence intervals. Closed points represent adjusted p-value < 0.05. All models were adjusted for infant age, sex, gestational age, birth weight, maternal household income, smoking during pregnancy, breastfeeding status, and sample processing time. Infection and GlycA exposure model estimates and details for all NMR metabolomic measures are shown in Figure 3—source data 1.

Figure 3—source data 1

Summary of regression models for difference in 12-month nuclear magnetic resonance (NMR) metabolomic measures per 1 increase in parent-reported infection from birth to 12 months of age or per SD increase in 12-month log glycoprotein acetyls (GlycA).

https://cdn.elifesciences.org/articles/75170/elife-75170-fig3-data1-v1.xlsx
Figure 3—source data 2

Summary of regression models for difference in 6-month nuclear magnetic resonance (NMR) metabolomic measures per 1 increase in parent-reported infection from birth to 6 months of age or per SD increase in 6-month log GlycA.

https://cdn.elifesciences.org/articles/75170/elife-75170-fig3-data2-v1.xlsx
Figure 3—source data 3

Summary of regression models for difference in 12-month nuclear magnetic resonance (NMR) metabolomic measures (adjusted for corresponding 6-month measure) per 1 increase in parent-reported infection from 6 to 12 months of age (adjusted for infections from birth to 6 months) or per SD increase in 12-month log glycoprotein acetyls (GlycA) (adjusted for 6-month GlycA).

https://cdn.elifesciences.org/articles/75170/elife-75170-fig3-data3-v1.xlsx
Figure 4 with 2 supplements see all
Download asset Open asset
Difference in 12-month plasma nuclear magnetic resonance (NMR) metabolomic measures for each SD increase in 12-month high-sensitivity C-reactive protein (hsCRP) (n = 555).

Forest plot for the estimated 12-month metabolomic differences for each additional SD log 12-month hsCRP (a, diamond points) from adjusted linear regression models, and the correlation of estimated metabolomic differences for infection and hsCRP (b) and for glycoprotein acetyls (GlycA) and hsCRP (c). Error bars are 95% confidence intervals. Closed points represent adjusted p-value < 0.05. All models were adjusted for infant age, sex, gestational age, birth weight, maternal household income, smoking during pregnancy, breastfeeding status, and sample processing time. hsCRP exposure model estimates and details for all NMR metabolomic measures are shown in Figure 4—source data 1.

Figure 4—source data 1

Summary of regression models for difference in 12-month nuclear magnetic resonance (NMR) metabolomic measures per SD increase in 12-month log high-sensitivity C-reactive protein (hsCRP).

https://cdn.elifesciences.org/articles/75170/elife-75170-fig4-data1-v1.xlsx
Figure 4—source data 2

Summary of regression models for difference in 6-month nuclear magnetic resonance (NMR) metabolomic measures per SD increase in 6-month log high-sensitivity C-reactive protein (hsCRP).

https://cdn.elifesciences.org/articles/75170/elife-75170-fig4-data2-v1.xlsx
Figure 4—source data 3

Summary of regression models for difference in 12-month nuclear magnetic resonance (NMR) metabolomic measures (adjusted for corresponding 6-month measure) per SD increase in 12-month log high-sensitivity C-reactive protein (hsCRP) (adjusted for 6-month hsCRP).

https://cdn.elifesciences.org/articles/75170/elife-75170-fig4-data3-v1.xlsx

There was a stronger correlation for the metabolomic differences related to infection and those related to GlycA (r = 0.74, Figure 3c) than for infection and hsCRP (r = 0.62, Figure 4b). The correlation of metabolomic differences related to GlycA and those for hsCRP was r = 0.61 (Figure 4c).

In models investigating the relationship between number of parent-reported infections from birth to 6 months of age and 6-month metabolomic profile, there was evidence of a similar, but less marked, pattern of associations to that seen at 12 months of age. Higher number of infections was modestly associated with higher inflammatory markers (GlycA (0.07 SD per 1 infection, [0.02–0.12]) and hsCRP (0.06 [0.01–0.11])), lower HDL cholesterol (−0.06 [−0.11 to −0.01]), and lower ApoA1 (−0.07 [−0.12 to −0.02]) (Figure 3—figure supplement 1a). At 6 months of age, higher GlycA was associated with a similar pattern of 6-month metabolomic differences as seen at 12 months (Figure 3—figure supplement 1b), as was hsCRP (Figure 4—figure supplement 1a). In models investigating the relationship between number of parent-reported infections from 6 to 12 months of age and 12-month metabolomic profile (i.e. infections within the preceding 6 months) with adjustment for birth to 6-month infections, associations between number of infections and metabolomic profile were similar to models considering number of infections from birth to 12 months of age (Figure 3—figure supplement 2a). Models considering 12-month GlycA or hsCRP as the exposure with adjustment for 6-month GlycA or hsCRP, respectively, also resembled those without adjustment for 6-month inflammation (Figure 3—figure supplement 2b, Figure 4—figure supplement 2a).

In secondary analyses, there was little evidence for associations between serum NMR metabolomic measures at birth at number of parent-reported infections from birth to 6 months of age (Supplementary file 3A). Similarly, there was little evidence for metabolomic measures at 6 months of age associating with number of infections from 6 to 12 months of age. GlycA at 6 months showed a modest association with a higher number of infections from 6 to 12 months (average 0.10 higher infections per 1 SD higher log 6-month GlycA, [0.04–0.16]), adjusted for the number of infections from birth to 6 months; for hsCRP there was little evidence (0.02 [−0.04–0.08]) (Supplementary file 3B).

Infection and inflammation burden and plasma LC/MS lipidomic profile at 12 months

In regression models with number of parent-reported infections as exposure and LC/MS lipids as the outcomes, infants with more infections had, on average, lower levels of the dehydrocholesteryl ester (−0.05 SD per 1 infection, [−0.08 to −0.03]) and trihexosylceramide (−0.04 [−0.06 to −0.01]) class lipids. There was also evidence to a lesser extent for associations between higher number of infections and lower cholesteryl esters (−0.03 [−0.05 to −0.01]) and plasmalogen classes (lysoalkenyl phosphatidylcholines, −0.02 [−0.04–0.00]; alkenylphosphatidylethanolamines, −0.03 [−0.06 to −0.01]) (Figure 5a). The 10 lipid species with the strongest statistical evidence for association with number of infections were hexosylceramides (HexCer(d18:2/18:0), –0.05 [−0.07 to −0.02]; HexCer(d18:2/22:0), –0.04 [−0.07 to −0.02]; HexCer(d16:1/24:0), –0.04 [−0.07 to −0.02]), trihexosylceramides (Hex3Cer(d18:1/18:0), –0.04 [−0.07 to −0.02]; Hex3Cer(d18:1/20:0), –0.05 [−0.07 to −0.02]; Hex3Cer(d18:1/22:0), –0.04 [−0.07 to −0.02]), phosphatidylethanolamines (PE(18:0/20:3), 0.04 [0.02–0.07]), cholesteryl esters (CE(22:5), –0.04 [−0.07 to −0.02]; CE(22:6), –0.04 [−0.07 to −0.02]) and dehydrocholesteryl esters (DE(18:2), –0.05 [−0.08 to −0.03]) (Figure 5—figure supplement 1a).

Figure 5 with 3 supplements see all
Download asset Open asset
Difference in 12-month plasma liquid chromatography/mass spectrometry (LC/MS) lipidomic class totals for each increase in parent-reported infection (birth to 12 months) and for each SD increase in 12-month glycoprotein acetyls (GlycA) (n = 550).

Forest plots of the estimated 12-month lipidomic differences in class totals for each additional parent-reported infection from birth to 12 months (a, circle points) or SD log 12-month GlycA (b, square points) from adjusted linear regression models, and the correlation of estimated differences for these two exposures across all lipidomic measures (c). In (a) and (b), error bars are 95% confidence intervals. Closed points represent adjusted p-value < 0.05. All models were adjusted for infant age, sex, gestational age, birth weight, maternal household income, smoking during pregnancy, breastfeeding status, and sample processing time. Forest plots depicting individual lipid species within each group are shown in Figure 5—figure supplement 1. Infection and GlycA exposure model estimates and details for all LC/MS lipidomic measures are shown in Figure 5—source data 1.

Figure 5—source data 1

Summary of regression models for difference in 12-month liquid chromatography/mass spectrometry (LC/MS) lipidomic measures per 1 increase in parent-reported infection from birth to 12 months of age or per SD increase in 12-month log glycoprotein acetyls (GlycA).

https://cdn.elifesciences.org/articles/75170/elife-75170-fig5-data1-v1.xlsx
Figure 5—source data 2

Summary of regression models for difference in 6-month liquid chromatography/mass spectrometry (LC/MS) lipidomic measures per 1 increase in parent-reported infection from birth to 6 months of age or per SD increase in 6-month log glycoprotein acetyls (GlycA).

https://cdn.elifesciences.org/articles/75170/elife-75170-fig5-data2-v1.xlsx
Figure 5—source data 3

Summary of regression models for difference in 12-month liquid chromatography/mass spectrometry (LC/MS) lipidomic measures (adjusted for corresponding 6-month measure) per 1 increase in parent-reported infection from 6 to 12 months of age (adjusted for infections from birth to 6 months) or per SD increase in 12-month log glycoprotein acetyls (GlycA) (adjusted for 6-month GlycA).

https://cdn.elifesciences.org/articles/75170/elife-75170-fig5-data3-v1.xlsx

Compared to models with number of infections, lipidomic differences were more pronounced when GlycA was considered as the exposure, including higher ceramides (0.15 SD per 1 SD increase in log GlycA, [0.06–0.24]), dihexosylceramides (0.29 [0.21–0.38]), di- (0.32 [0.24–0.40]) and TGs (0.27 [0.19–0.36]), and phospholipid classes (e.g. phosphatidylethanolamines, 0.28 [0.19–0.36]; phosphatidylglycerols, 0.23 [0.15–0.32]), and lower plasmalogen classes (alkenyl phosphatidylcholines, −0.28 [−0.37 to −0.21]; alkenylphosphatidylethanolamines, −0.22 [−0.31 to −0.14]; lysoalkenyl phosphatidylcholines, −0.14 [−0.20 to −0.08]), cholesteryl esters (−0.25 [−0.34 to −0.17]), and dehydrocholesteryl esters (−0.41 [−0.49 to −0.33]) (Figure 5b).

Higher hsCRP was also associated with differences across several lipid classes, particularly lower plasmalogen classes (e.g. lysoalkenyl phosphatidylcholines, −0.19 SD per 1 SD increase in log hsCRP, [−0.25 to −0.14]; alkenyl phosphatidylcholines, −0.26 [−0.35 to −0.18]), sulfatides (−0.26 [−0.34 to −0.17]), and alkyl phosphatidylcholines (−0.29 [−0.38 to −0.21]) (Figure 6a). For all lipidomic models, bootstrap estimates were similar to the standard regression estimates. All sensitivity analyses (Supplementary file 1D-1F) showed similar estimated effect sizes, except for hsCRP models excluding samples with processing times greater than 4 hr, where estimated differences in several classes (including lysophosphatidylcholines, lysoalkenyl phosphatidylcholines, and lysoalkenylphosphatidylethanolamines) were substantially larger compared to the primary models (Supplementary file 1F).

Figure 6 with 3 supplements see all
Download asset Open asset
Difference in 12-month plasma liquid chromatography/mass spectrometry (LC/MS) lipidomic class totals for each SD increase in 12-month high-sensitivity C-reactive protein (hsCRP) (n = 550).

Forest plot for the estimated 12-month lipidomic differences for each additional SD log 12-month hsCRP (a, diamond points) from adjusted linear regression models, and the correlation of estimated differences across all lipidomic measures for infection and hsCRP (b) and glycoprotein acetyls (GlycA) and hsCRP (c). In (a), error bars are 95% confidence intervals. Closed points represent adjusted p-value < 0.05. All models were adjusted for infant age, sex, gestational age, birth weight, maternal household income, smoking during pregnancy, breastfeeding status, and sample processing time. Forest plots depicting individual lipid species within each group are shown in Figure 6—figure supplement 1. hsCRP exposure model estimates and details for all LC/MS lipidomic measures are shown in Figure 6—source data 1.

Figure 6—source data 1

Summary of regression models for difference in 12-month liquid chromatography/mass spectrometry (LC/MS) lipidomic measures per SD increase in 12-month log high-sensitivity C-reactive protein (hsCRP).

https://cdn.elifesciences.org/articles/75170/elife-75170-fig6-data1-v1.xlsx
Figure 6—source data 2

Summary of regression models for difference in 6-month liquid chromatography/mass spectrometry (LC/MS) lipidomic measures per SD increase in 6-month log high-sensitivity C-reactive protein (hsCRP).

https://cdn.elifesciences.org/articles/75170/elife-75170-fig6-data2-v1.xlsx
Figure 6—source data 3

Summary of regression models for difference in 12-month liquid chromatography/mass spectrometry (LC/MS) lipidomic measures (adjusted for corresponding 6-month measure) per SD increase in 12-month log high-sensitivity C-reactive protein (hsCRP) (adjusted for 6-month hsCRP).

https://cdn.elifesciences.org/articles/75170/elife-75170-fig6-data3-v1.xlsx

There was stronger correlation for lipidomic differences related to infection and GlycA (r = 0.77, Figure 5c) than for infection and hsCRP (r = 0.36, Figure 6b) or GlycA and hsCRP (r = 0.55, Figure 6c).

In models investigating parent-reported infections from birth to 6 months of age and 6-month lipidomic measures, a similar, but less marked, pattern of associations was seen as at 12 months (Figure 5—figure supplement 2a). Models of 6-month GlycA (Figure 5—figure supplement 2b) or 6-month hsCRP (Figure 6—figure supplement 2a) and 6-month lipidomic measures were largely the same as the 12-month models with some exceptions, such as higher GlycA associating with lower GM3-gangliosides at 6 months (−0.10 SD per 1 SD higher log GlycA, 95% CI [−0.18 to −0.02]) in contrast to associating with higher GM3-gangliosides at 12 months (0.14 [0.05–0.23]). In models investigating infections from 6 to 12 months of age and 12-month lipidomic measures, with adjustment for infection from birth to 6 months, a similar pattern of associations was seen as in models for number of infections from birth to 12 months (Figure 5—figure supplement 3a). Likewise, models for 12-month GlycA or hsCRP and 12-month lipidomic measures showed very similar results with adjustment for 6-month measures compared to models without 6-month adjustment (Figure 5—figure supplement 3b, Figure 6—figure supplement 3a).

As with the NMR metabolomic measures, there was little evidence for associations between serum LC/MS lipidomic measures at birth and number of infections from birth to 6 months of age in secondary analyses (Supplementary file 3C). For 6-month measures, higher cholesteryl esters and dehydrocholesteryl esters were associated with lower number of subsequent infections from 6 to 12 months of age (average −0.10 lower infections per 1 SD higher 6-month log total cholesteryl esters, 95% CI [−0.15 to −0.05]; −0.10 infections per 1 SD higher 6-month log total dehydrocholesteryl esters, 95% CI [−0.16 to −0.05]) (Supplementary file 3D).

Mediation analysis

We next assessed whether inflammation (i.e. GlycA or hsCRP at 12 months) mediated the effects of infection on specific metabolite and lipid measures (adjusted p-value < 0.1) (Figure 7). For the NMR metabolomic measures considered in mediation models, there was evidence of an indirect effect of infection mediated by GlycA on all measures, and an indirect effect mediated by hsCRP for phenylalanine, ApoA1, and the HDL, HDL2, and HDL3 cholesterols. For all measures except ApoA1 and HDL3 cholesterol, GlycA was estimated to mediate a larger proportion of the total effect of infections on metabolomic measures than hsCRP. For the LC/MS lipidomic mediation models, there was evidence of indirect effects of infection on all the considered lipid classes mediated by GlycA, and indirect effects mediated by hsCRP for all classes except dehydrocholesteryl esters. GlycA was estimated to mediate a larger proportion of the total effect on infections on all lipid classes except lysoalkenyl phosphatidylcholines. GlycA was similarly estimated to mediate a larger proportion of the total effect of infection on individual lipid species than hsCRP for all lipid species included in mediation analyses.

Figure 7 with 1 supplement see all
Download asset Open asset
Total effect of infection on 12-month metabolomic and lipidomic measures (purple, circle points) and the estimated natural indirect effect component of these mediated by glycoprotein acetyls (GlycA) (orange, square points) or high-sensitivity C-reactive protein (hsCRP) (green, diamond points).

Units of change are 1 infection for parent-reported infections, and 1 SD change for GlycA, hsCRP, and metabolomic/lipidomic measures on log scale. Error bars are 95% confidence intervals. Closed points represent p-value < 0.05. All models were adjusted for infant age, sex, gestational age, birth weight, maternal household income, smoking during pregnancy, breastfeeding status, and sample processing time. Model details are in Figure 7—source data 1.

Figure 7—source data 1

Summary of mediation models for total, natural direct, and natural indirect effects (mediated by glycoprotein acetyls [GlycA] or high-sensitivity C-reactive protein [hsCRP]) of infection from birth to 12 months on 12-month metabolomic and lipidomic measures.

https://cdn.elifesciences.org/articles/75170/elife-75170-fig7-data1-v1.xlsx
Figure 7—source data 2

Summary of mediation models for total, natural direct, and natural indirect effects (mediated by glycoprotein acetyls [GlycA] or high-sensitivity C-reactive protein [hsCRP]) of infection from 6 to 12 months on 12-month metabolomic and lipidomic measures, with adjustment for infections from birth to 6 months of age, 6-month inflammation, and the corresponding 6-month metabolomic/lipidomic measure.

https://cdn.elifesciences.org/articles/75170/elife-75170-fig7-data2-v1.xlsx

To consider potential confounding from earlier inflammation or metabolomic/lipidomic measures, we additionally performed mediation analyses for inflammation measured at 12 months mediating the relationship between number of infections from 6 to 12 months of age and the 12-month metabolomic/lipidomic measures with adjustment for number of infections from birth to 6 months of age, 6-month inflammation, and the corresponding 6-month metabolomic/lipidomic measure. For most metabolomic and lipidomics measures, evidence for indirect effects mediated by GlycA or hsCRP varied minimally in this 6- to 12-month model compared to the birth to 12-month mediation models (Figure 7—figure supplement 1).

Discussion

In this study, cumulative parent-reported infection burden from birth to 12 months was associated with adverse NMR metabolomic and LC/MS lipidomic profiles at 12 months of age. Similar but more marked effects on these profiles were evident when considering GlycA, a cumulative inflammation marker, as an exposure. In contrast, differences in metabolomic and lipidomic profiles associated with higher hsCRP were largely distinct from, and less marked than, those for GlycA, suggesting that GlycA may be superior to hsCRP as an early life marker of infection burden. There was evidence that inflammation (with GlycA generally showing stronger evidence as a mediator than hsCRP) may partly mediate many of the largest metabolomic and lipidomic differences.

These findings are novel and of potentially considerable significance; the burden of infection falls largely in infancy and early childhood, and these are among the first data to explore the cardiometabolic associations with common infections in this age group. Cardiometabolic risk accrues across the entire life course; the American Heart Association states that ‘with primordial and primary prevention, CVD is largely preventable’ and that risk stratification and intervention are more likely to be successful if done earlier in life (Weintraub et al., 2011). Therefore, understanding associations in early life in general populations is an important step towards risk stratification and targeted interventions. There are few data investigating LC/MS lipidomics in response to common infections in high-income countries; indeed some of the only analogous studies investigated lipidomics following Ebola (Kyle et al., 2019) or acute lower respiratory tract infection (Gao et al., 2019) in adults. These are the only analyses to explore the potential mediation of these associations between infections and omics profiles by inflammation.

The metabolomic profiles observed in this study in infants with greater infection burden (higher triglycerides, lower HDL cholesterol, and lower ApoA1) reflect those previously linked to infection, including SARS-CoV-2 (Alvarez and Ramos, 1986; Bruzzone et al., 2020; Gallin et al., 1969; Liuba et al., 2003; Madsen et al., 2018), and CVD risk (particularly higher phenylalanine) in adults (Würtz et al., 2015; Würtz et al., 2012). Acute and chronic infection-related metabolic differences have been implicated in atherosclerosis (Khovidhunkit et al., 2000), including increased carotid intima-media thickness (Burgner et al., 2015c; Liuba et al., 2003) and arterial stiffness (Charakida et al., 2009) in older children, and in adult CVD risk (Bergh et al., 2017; Burgner et al., 2015a). The mechanisms underpinning the link between inflammation and metabolism are not well characterised, but there is evidence they may share regulatory pathways. For example, lipid-activated nuclear receptors such as peroxisome-proliferator-activated receptors and liver X receptors regulate both lipid metabolism and inflammation (Bensinger and Tontonoz, 2008). Synthesis of leptin, a key regulatory hormone of metabolism, including energy homeostasis (Park and Ahima, 2015), is increased by proinflammatory cytokines during acute infection (Behnes et al., 2012), and is in turn implicated in many inflammatory processes (Abella et al., 2017).

In lipidomic analyses, infection burden was associated with lower levels of several ether phosphatidylethanolamine species (PC(O) and PC(P)), consistent with the reported decrease in these polyunsaturated-fatty acid (PUFA)-containing species in SARS-CoV-2 infected adults (Schwarz et al., 2020). Importantly, these decreases contrast with observed increases in diacyl-phosphatidylethanolamine species, and this discordance in phosphatidylethanolamine species is characteristic of conditions characterised by inflammation, including Alzheimer’s disease (Huynh et al., 2020). The role of phosphatidylethanolamines in infection is incompletely understood, but omega-3 fatty acids (a major class of PUFA) have been implicated in anti-inflammatory pathways (Calder, 2013) and ether phosphatidylethanolamine species are a potential source of PUFAs. Lower trihexosylceramide lipid class, associated with more infections, has been linked to increased BMI and pre-diabetic phenotypes in adults (Meikle et al., 2013; Weir et al., 2013).

The different relationships of the two inflammatory markers GlycA and hsCRP with metabolomic and lipidomic profiles are consistent with a recent study in pregnant women reporting GlycA was more strongly correlated with NMR metabolomic differences than hsCRP (Mokkala et al., 2020). While hsCRP has been widely used as a measure of chronic inflammation, primarily in adults, it is an acute phase reactant that increases rapidly following acute stimulus and returns to baseline levels within a matter of days and is therefore mostly used as a diagnostic adjunct in children with acute infection or inflammation (Gabay and Kushner, 1999). In contrast, there is evidence that GlycA can remain elevated for up to a decade in young adults (Ritchie et al., 2015), and it is considered a superior marker of cumulative inflammation burden, though data of GlycA in early life are sparser. Several studies in adults have reported that these two markers are only moderately correlated (Akinkuolie et al., 2014; Gruppen et al., 2015b), and it is suggested that these markers reflect different (albeit overlapping) inflammatory processes. This is consistent with our findings of distinct metabolomic and lipidomic profiles for these two markers and reflects other findings that show different relationships of GlycA and hsCRP with cardiovascular and metabolic phenotypes. For example, GlycA is independently associated with risk of CVD and with enzymatic esterification of free cholesterol, even after adjustment for hsCRP (Duprez et al., 2016; Gruppen et al., 2015a; Muhlestein et al., 2018). Associations between GlycA and lipolysis rates (Levine et al., 2020) and gut microbiome diversity (Mokkala et al., 2020) are also stronger than those reported for hsCRP.

Strengths and limitations

This large prospective cohort study is the first to examine the relationship between infection, inflammation, and plasma NMR metabolomic and LC/MS lipidomic profiles in early life, with implications for later CVD risk. The associations were consistent using either parental-reported infections (potentially prone to reporting bias) or 12-month GlycA as a measure of cumulative inflammatory burden. We found no evidence that birth plasma metabolomic or lipidomic profiles were associated with infection burden in the first 6 months of life (i.e. reverse causation; Supplementary file 3A and C). There was some limited evidence that GlycA at 6 months may be associated with subsequent number of infections (Supplementary file 3B), but we were able to adjust for this in mediation models to avoid exposure-induced mediator-outcome confounding (Vanderweele and Vansteelandt, 2009). The use of metabolomics and lipidomic data from three early life time points (birth, 6 months, and 12 months) is unique and allowed longitudinal analyses to be performed.

Limitations include the use of cross-sectional data for mediation analyses (i.e. 12-month inflammatory markers and 12-month metabolomics/lipidomics), which we addressed in part with mediation models adjusted for infections from birth to 6 months of age and 6-month inflammation. Moreover, GlycA is a marker of cumulative inflammation (Collier et al., 2019; Ritchie et al., 2015) and is believed to reflect inflammatory events occurring prior to the 12-month time point. Evidence from randomised controlled trials in adults support a causal role of inflammation in CVD risk (Chunfeng et al., 2021; Ridker et al., 2017), however this study in infants is observational. While we have used a casual framework for mediation analyses, our findings do not demonstrate causality. Despite rigorous adjustment and sensitivity analyses, there may be unmeasured confounding, such as from environmental factors that may modify changes in metabolomic/lipidomic measures following infection. We considered total parent-reported infections, as defining clinical categories of infection given the non-specific symptoms and signs in infancy is challenging. Validation of all parent-reported infections was not feasible and only a minority (70 infections in 555 infants) resulted in medical attention (Rowland et al., 2021). However, the strong correlation between metabolomic and lipidomic differences for number of reported infections and for GlycA suggests that parent report is a reasonable measure of infection-induced inflammatory burden. Most childhood infections are viral, and microbial testing is impractical in non-hospitalised children and cannot necessarily differentiate between colonisation and infection. Notwithstanding, distinct lipid differences have been reported for children with bacterial versus viral infections (Wang et al., 2019), and different infectious aetiologies in adults have been linked to differential CVD risk (Cowan et al., 2020). Finally, the relative lack of racial/ethnic diversity in our cohort may limit generalisability of our findings.

Conclusions

In summary, we present evidence for higher infection burden in early life leading to proatherogenic and prodiabetic plasma metabolome and lipidome at 12 months of age, and for inflammation partly mediating these relationships. GlycA may be a better marker of early life infection and inflammation burden than hsCRP. These findings suggest that the impact of the cumulative infection and inflammation burdens previously implicated in adult cardiometabolic disease may begin in infancy, thereby offering opportunities for early prevention. Further work is required to determine the potential consequences these adverse metabolomic profiles in early life have on later risk of disease, and how the relationships between infections, inflammation, and metabolomic and lipidomic profiles might differ across age groups, pathogen type, and clinical severity of infection.

Data availability

Given the ethics for this study, the individual participant data cannot be made freely available online. Interested parties can access the data used in this study upon reasonable request, with approval by the BIS data custodians. As part of this process, researchers will be required to submit a project concept for approval, to ensure the data is being used responsibly, ethically, and for scientifically sound projects.

Source data files have been uploaded for each of the results figures (Figures 3—7) showing the model summary data for all metabolomic and lipidomic measures, including those not presented in figures.

Source code for all analyses have been uploaded as Source code 1 and Source code 2.

Data availability

With the approved ethics for this study, the individual participant data cannot be made freely available online. Interested parties can access the data used in this study upon reasonable request, with approval by the Barwon Infant Study data custodians. As part of this process, researchers will be required to submit a project concept for approval, to ensure the data is being used responsibly, ethically, and for scientifically sound projects. Source data files have been uploaded for each of the results figures (Figures 3 to 7) showing the model summary data for all metabolomic and lipidomic measures, including those not presented in figures. Source code for all analyses have been uploaded as Source Code 1 and 2.

References

  28. Book
    1. Feingold KR
    2. Grunfeld C
    (2019)
    The Effect of Inflammation and Infection on Lipids and Lipoproteins
    Endotext.
    1. Pesonen E
    2. Rapola J
    3. Viikari J
    4. Turtinen J
    5. Akerblom HK
    (1993)
    Altered serum lipid profile after systemic infection in children: risk factor for CHD?
    European Heart Journal 14 Suppl K:7–11.
  55. Book
    1. Pink B
    (2013)
    Socio-Economic Indexes for Areas (SEIFA) 2011
    Canberra: Australian Bureau of Statistics.
  56. Software
    1. R Development Core Team
    (2018) R: A language and environment for statistical computing
    R Foundation for Statistical Computing, Vienna, Austria.
    http://www.r-project.org
    1. Troeger C
    2. Blacker B
    3. Khalil IA
    4. Rao PC
    5. Cao J
    6. Zimsen SRM
    7. Albertson SB
    8. Deshpande A
    9. Farag T
    10. Abebe Z
    11. Adetifa IMO
    12. Adhikari TB
    13. Akibu M
    14. Al Lami FH
    15. Al-Eyadhy A
    16. Alvis-Guzman N
    17. Amare AT
    18. Amoako YA
    19. Antonio CAT
    20. Aremu O
    21. Asfaw ET
    22. Asgedom SW
    23. Atey TM
    24. Attia EF
    25. Avokpaho EFGA
    26. Ayele HT
    27. Ayuk TB
    28. Balakrishnan K
    29. Barac A
    30. Bassat Q
    31. Behzadifar M
    32. Behzadifar M
    33. Bhaumik S
    34. Bhutta ZA
    35. Bijani A
    36. Brauer M
    37. Brown A
    38. Camargos PAM
    39. Castañeda-Orjuela CA
    40. Colombara D
    41. Conti S
    42. Dadi AF
    43. Dandona L
    44. Dandona R
    45. Do HP
    46. Dubljanin E
    47. Edessa D
    48. Elkout H
    49. Endries AY
    50. Fijabi DO
    51. Foreman KJ
    52. Forouzanfar MH
    53. Fullman N
    54. Garcia-Basteiro AL
    55. Gessner BD
    56. Gething PW
    57. Gupta R
    58. Gupta T
    59. Hailu GB
    60. Hassen HY
    61. Hedayati MT
    62. Heidari M
    63. Hibstu DT
    64. Horita N
    65. Ilesanmi OS
    66. Jakovljevic MB
    67. Jamal AA
    68. Kahsay A
    69. Kasaeian A
    70. Kassa DH
    71. Khader YS
    72. Khan EA
    73. Khan MN
    74. Khang Y-H
    75. Kim YJ
    76. Kissoon N
    77. Knibbs LD
    78. Kochhar S
    79. Koul PA
    80. Kumar GA
    81. Lodha R
    82. Magdy Abd El Razek H
    83. Malta DC
    84. Mathew JL
    85. Mengistu DT
    86. Mezgebe HB
    87. Mohammad KA
    88. Mohammed MA
    89. Momeniha F
    90. Murthy S
    91. Nguyen CT
    92. Nielsen KR
    93. Ningrum DNA
    94. Nirayo YL
    95. Oren E
    96. Ortiz JR
    97. Pa M
    98. Postma MJ
    99. Qorbani M
    100. Quansah R
    101. Rai RK
    102. Rana SM
    103. Ranabhat CL
    104. Ray SE
    105. Rezai MS
    106. Ruhago GM
    107. Safiri S
    108. Salomon JA
    109. Sartorius B
    110. Savic M
    111. Sawhney M
    112. She J
    113. Sheikh A
    114. Shiferaw MS
    115. Shigematsu M
    116. Singh JA
    117. Somayaji R
    118. Stanaway JD
    119. Sufiyan MB
    120. Taffere GR
    121. Temsah M-H
    122. Thompson MJ
    123. Tobe-Gai R
    124. Topor-Madry R
    125. Tran BX
    126. Tran TT
    127. Tuem KB
    128. Ukwaja KN
    129. Vollset SE
    130. Walson JL
    131. Weldegebreal F
    132. Werdecker A
    133. West TE
    134. Yonemoto N
    135. Zaki MES
    136. Zhou L
    137. Zodpey S
    138. Vos T
    139. Naghavi M
    140. Lim SS
    141. Mokdad AH
    142. Murray CJL
    143. Hay SI
    144. Reiner RC Jr
    (2018) Estimates of the global, regional, and national morbidity, mortality, and aetiologies of lower respiratory infections in 195 countries, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016
    The Lancet Infectious Diseases 18:1191–1210.
    https://doi.org/10.1016/S1473-3099(18)30310-4
    1. Zeileis A
    2. Kleiber C
    3. Jackman S
    (2008)
    Regression models for count data in R
    Journal of Statistical Software 27:1–25.

Decision letter

  1. Hossein Ardehali
    Reviewing Editor; Northwestern University, United States
  2. Matthias Barton
    Senior Editor; University of Zurich, Switzerland
  3. Noah Snyder-Mackler
    Reviewer; Arizona State University, United States

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Early life infection and proinflammatory, atherogenic metabolomic and lipidomic profiles at 12 months of age: a population-based cohort study" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Noah Snyder-Mackler (Reviewer #3).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1. The paper contains a number of grammatical, spelling and structural errors. The authors are strongly encouraged to review the paper carefully prior to resubmitting the paper.

2. Figures 4 and 5 are difficult to understand. What do each point in the blot represent? What does the line around the data points represent? It is important the authors provide more explanation for readers who do not have familiarity with these plots.

3. The authors need to better define the significance of their findings. For example, although the link between inflammation and changes in metabolomics and lipidomics is known, the authors need to provide a better explanation on why their specific findings are important in the setting of infection in infants.

4. The background and rationale of the introduction should clarify the importance of transient inflammation with infections compared with cumulative burden of inflammation. The authors are suggesting that a higher burden of infections, which are common in the first year of life, are associated with long-lasting inflammation. This does not pan out in terms of hsCRP but does in terms of GlycA but this is not a known robust marker of long-term inflammation. This should be tempered.

5. Are social factors available? For example, a child with two working parents who is placed in daycare is likely to have a higher burden of infections in contrast with one with a stay at home parent? Household income and daycare status would be useful covariates to consider.

6. Are pregnancy-related covariates available? Particularly adjustment for gestational age at delivery, chronic or gestational hypertension, preeclampsia, and/or chronic or gestational diabetes.

7. All results in the text should include measures of imprecision such as 95% CI.

8. Were the children screened and excluded if they had an active infection at the time of measures of inflammatory markers, metabolomics, and lipidomics? How was this assessed?

9. The authors may consider the following reference to align with the AGReMA guidelines for reporting mediation analyses. Lee et al. A Guideline for Reporting Mediation Analyses of Randomized Trials and Observational Studies: The AGReMA Statement. JAMA. 2021;326(11):1045-1056. doi:10.1001/jama.2021.14075; Particularly, a DAG would be useful.

10. Table 2 should include direct effects and clarity on what unit of change for each exposure, mediator, and outcome are the statistics based on? What is a 1-unit change in hsCRP reflect? since these units may vary significantly, use of 1 SD change would be optimal.

11. One technical concern is the variation in the amount of time the blood samples spent between post-processing and storage at -80C (and how they were held during that interval). The authors should be commended for their sensitivity analysis looking at just those samples that were stored for < 4 hours. However, the justification for using only those < 4 hours is not clear. It is also unclear and a bit difficult for a reader to look at Supplementary files (1A and 1B) and draw the authors' conclusion that "this had little difference on the estimated effect sizes observed in analyses with the full cohort". Could the authors make more direct comparisons between the effect sizes to hammer this point home? Given that the sample size is smaller for this analysis, the p-values should be higher (less power), but the effect sizes should be unbiased and relatively similar. Perhaps calculating the correlation between effect sizes would help improve this sensitivity analysis and be good to report in the main text.

12. The authors show that while the correlation between GlycA (or CRP) and number of infections is relatively low (albeit a bit higher for GlycA), the effect sizes of GlycA and infections on the metabolome and lipidome are strongly correlated (and to a lesser extent between CRP and infections). A third comparison here that would be very useful, would be between GlycA and CRP effects on the metabolome and lipidome.

13. The mediation analysis is the least convincing part of the manuscript. It is appreciated that the authors are trying to identify how infections might be translating to adverse metabolomic and lipidomic profiles. However, the justification for GlycA or hsCRP being the mediating mechanisms is not convincing. The authors are implying (statistically through their mediation models) that the effect of the number of infections on many metabolites and lipids is due to the changes in GlycA (or hsCRP). Since both GlycA and number of infections are cumulative measures at 12mo, it is unlikely that one (e.g., GlycA) precedes the other. Rather, as the authors state, GlycA is a marker of cumulative infections. This would preclude it from being a mediator unless the authors have data from earlier timepoints (like previous months) that would provide more support for a mediation effect.

Reviewer #1 (Recommendations for the authors):

This is an interesting paper and a large amount of data are provided. I have the following comments:

1. The paper contains a number of grammatical, spelling and structural errors. The authors are strongly encouraged to review the paper carefully prior to resubmitting the paper.

2. Figures 4 and 5 are difficult to understand. What do each point in the blot represent? What does the line around the data points represent? It is important the authors provide more explanation for readers who do not have familiarity with these plots.

3. The authors need to better define the significance of their findings. For example, although the link between inflammation and changes in metabolomics and lipidomics is known, the authors need to provide a better explanation on why their specific findings are important in the setting of infection in infants.

4. Additional environmental variables and their link to changes in childhood serum lipids and metabolome after infection need to be studied and included in the paper.

Reviewer #2 (Recommendations for the authors):

Mansell et al. report on associations between infections between birth and 12 months of age and metabolomic/lipidomic profiles. There are several suggestions that may strengthen the study and manuscript:

1) The background and rationale of the introduction should clarify the importance of transient inflammation with infections compared with cumulative burden of inflammation. The authors are suggesting that a higher burden of infections, which are common in the first year of life, are associated with long-lasting inflammation. This does not pan out in terms of hsCRP but does in terms of GlycA but this is not a known robust marker of long-term inflammation. This should be tempered.

2) The introduction suggests that inflammation may be a causal target but this analysis can only identify potential markers. This should be tempered to distinctly argue against any causal findings in this observational analysis?

3) Are social factors available? For example, a child with two working parents who is placed in daycare is likely to have a higher burden of infections in contrast with one with a stay at home parent? Household income and daycare status would be useful covariates to consider.

4) Are pregnancy-related covariates available? Particularly adjustment for gestational age at delivery, chronic or gestational hypertension, preeclampsia, and/or chronic or gestational diabetes.

5) Are higher order gestations (e.g., twins, triplets) excluded from the analysis? If not, a sensitivity analysis excluding them should be considered due to confounding from relatedness and shared maternal characteristics in models

6) All results in the text should include measures of imprecision such as 95% CI

7) Were the children screened and excluded if they had an active infection at the time of measures of inflammatory markers, metabolomics, and lipidomics? How was this assessed?

8) The authors may consider the following reference to align with the AGReMA guidelines for reporting mediation analyses. Lee et al. A Guideline for Reporting Mediation Analyses of Randomized Trials and Observational Studies: The AGReMA Statement. JAMA. 2021;326(11):1045-1056. doi:10.1001/jama.2021.14075; Particularly, a DAG would be useful.

9) The language that GlycA outperforms hsCRP is unclear how this was demonstrated. The authors should provide direct comparisons and supportive data or remove these statements.

10) Table 2 should include direct effects and clarity on what unit of change for each exposure, mediator, and outcome are the statistics based on? What is a 1-unit change in hsCRP reflect? since these units may vary significantly, use of 1 SD change would be optimal

11) since metabolomics and lipidomics were measured at birth, adjustment for these variables in the mediation analysis may be helpful to mitigate concerns of residual confounding

Reviewer #3 (Recommendations for the authors):

1. I think the authors should take care in avoiding technical jargon and/or explaining it before using it. For instance, there are multiple points where acronyms are introduced without being first defined, like NMR, TGs, and TG-Os (which I think are nuclear magnetic resonance, and two types of triglycerides). Additionally, technical modifications like lines 211-213 are difficult to parse. What does "used medronic acid to passivate… to avoid peak tailing.." mean to a non-expert?

2. L229: I understand that 1 is added so that you can log-transform any zero values, but is 1 a good value for all metabolites and lipids? Could this offset be scaled to each individual metabolite and lipid based on their lowest non-zero value, such that it doesn't affect the results as much? Additionally, what proportion of the metabolites/lipids measures had 0 values in the samples (and how many 0 values)? In other words: how big of an issue could this +1 offset be?

3. The reverse-causality analysis is really interesting. It would be great if the authors had intermediate time-points (6mo?) rather than just birth to see if there could be some link between inflammatory levels at 6mo (controlling for infection burden) and later life levels.

4. The use of the forest plots is much appreciated, but the effect sizes and CIs are overlapping, so it's really hard to interpret. Especially the lipidomic figures (Figures4 and 5). Here, it might be best to just show the "class totals" and not every species underlying those classes.

The mediation analysis results could also be presented as a forest plot instead of a table. That would make interpretation much easier for the reader.

https://doi.org/10.7554/eLife.75170.sa1

Author response

Essential revisions:

S1. The paper contains a number of grammatical, spelling and structural errors. The authors are strongly encouraged to review the paper carefully prior to resubmitting the paper.

Thank you for flagging this as an issue. There were some errors introduced during late-stage edits. The revised manuscript has been carefully reviewed by several authors for errors.

S2. Figures 4 and 5 are difficult to understand. What do each point in the blot represent? What does the line around the data points represent? It is important the authors provide more explanation for readers who do not have familiarity with these plots.

These points and lines were described in the legend of these figures. E.g., from the old Figure 4 (now Figure 5): ‘Error bars are 95% confidence intervals. In (a) and (b), black points represent class totals, closed dark grey points represent the top 10 species differences by p-value, and hollow points with a black outline represent other species with adjusted p<0.05.’ In response to this comment and the suggestion by Reviewer #3 (see comment R3.4 below), these forest plots have now been simplified; we now only show the lipid class totals in the main figures, instead presenting the forest plot with individual lipid species as a figure supplement. We have also updated these figures to indicate whether a difference in lipid class total has adjusted p-value<0.05 by using closed points, consistent with the NMR metabolomics figures (Figures 3 and 4). For the figure supplement forest plots showing lipid species, we now do not highlight the top 10 species associations by p-value and now do not show the confidence intervals for individual species. We have also revised the description in the figure legends for clarity.

Eg, now for Figure 5: ‘In (a) and (b), error bars are 95% confidence intervals. Closed points represent adjusted p-value<0.05. All models were adjusted for infant age, sex, gestational age, birth weight, maternal household income, smoking during pregnancy, breastfeeding status, and sample processing time. Forest plots depicting individual lipid species within each group are shown in Figure 5—figure supplement 1.’

S3. The authors need to better define the significance of their findings. For example, although the link between inflammation and changes in metabolomics and lipidomics is known, the authors need to provide a better explanation on why their specific findings are important in the setting of infection in infants.

The evolving COVID-19 pandemic has brought the relationship between acute infection and cardiometabolic risk into sharp focus. Notwithstanding, data linking metabolomic and lipidomic perturbations following infection are almost exclusively from adults and older children (Alvarez and Ramos, 1986; Bruzzone et al., 2020; Gallin, Kaye, and O'Leary, 1969; Liuba, Persson, Luoma, Ylä-Herttuala, and Pesonen, 2003; Madsen, Varbo, Tybjærg-Hansen, Frikke-Schmidt, and Nordestgaard, 2018). We therefore consider the findings to be novel and of potentially considerable significance for the following reasons: First, the burden of infection falls largely on pre-school children and ours are among the first data to explore the cardiometabolic associations with common infections in this age group. Second, cardiometabolic risk accrues across the entire life course, as recognised by the American Heart Association who have stated that “with primordial and primary prevention, CVD is largely preventable” and that risk stratification and intervention are more likely to be successful if done earlier in life (Weintraub et al., 2011); therefore understanding associations in early life in general populations is an important step towards risk-stratification and targeted interventions. Third, there are few data investigating mass-spec lipidomics in response to common infections in high income countries; indeed, some of the only analogous studies investigated lipidomics following Ebola (Kyle et al., 2019) or acute lower respiratory tract infection (Gao et al., 2019) in adults. Finally, these are the only analyses to explore the potential mediation of these associations by inflammation.

We have therefore added a paragraph to the Discussion that summarises these points (page 7-28, line 510-523):

“The findings are novel and of potentially considerable significance; the burden of infection falls largely on pre-school children and these are among the first data to explore the cardiometabolic associations with common infections in this age group. Second, cardiometabolic risk accrues across the entire life course; the American Heart Association state that “with primordial and primary prevention, CVD is largely preventable” and that risk stratification and intervention are more likely to be successful if done earlier in life (Weintraub et al., 2011) Therefore, understanding associations in early life in general populations is an important step towards risk-stratification and targeted interventions. Third, there are few data investigating LM/CS lipidomics in response to common infections in high income countries; indeed some of the only analogous studies investigated lipidomics following Ebola (Kyle et al., 2019) or acute lower respiratory tract infection (Gao et al., 2019) in adults. These are the only analyses to explore the potential mediation of these associations between infections and omics profiles by inflammation.”

S4. The background and rationale of the introduction should clarify the importance of transient inflammation with infections compared with cumulative burden of inflammation. The authors are suggesting that a higher burden of infections, which are common in the first year of life, are associated with long-lasting inflammation. This does not pan out in terms of hsCRP but does in terms of GlycA but this is not a known robust marker of long-term inflammation. This should be tempered.

We used two markers of inflammation (GlycA and hsCRP). We are well aware that (especially in early childhood) hsCRP is an acute phase reactant and in unlikely to reflect chronic inflammation; in part we included hsCRP as it is so widely used in the adult cardiovascular literature and therefore gives a familiar frame of reference to the use of the more novel marker, GlycA. We acknowledge that GlycA is a suggested rather than proven marker of chronic inflammation, especially in childhood (although we have previously published some evidence from this cohort (Collier et al., 2019)). We have edited the introduction to reflect these points as follows (page 6, line 69-78):

“High-sensitivity C-reactive protein (hsCRP) has been extensively used as a marker of chronic inflammation in adult studies, but is an acute phase reactant in children and may not reflect chronic inflammation in early life. Glycoprotein acetyls (GlycA), is a nuclear magnetic resonance (NMR) composite measure that is suggested to better reflect cumulative, chronic inflammation (Connelly, Otvos, Shalaurova, Playford, and Mehta, 2017). GlycA is an emerging biomarker for cardiometabolic risk (Connelly et al., 2017) that out-performs hsCRP as a predictor of CVD events and mortality (Akinkuolie, Buring, Ridker, and Mora, 2014; Duprez et al., 2016), and of infection-related morbidity and mortality (Ritchie et al., 2015).”

S5. Are social factors available? For example, a child with two working parents who is placed in daycare is likely to have a higher burden of infections in contrast with one with a stay at home parent? Household income and daycare status would be useful covariates to consider.

Annual household income was self-reported by mothers at the 28-weeks gestation time point and included as a covariate in all models. As suggested, child care utilisation may be associated with greater number of infections, but in the opinion of the authors does not seem likely to be itself a confounder of metabolomic/lipidomic outcomes. However, we agree that socio-demographic factors warrant further consideration, we have now performed sensitivity analyses adjusting for two other measures of socioeconomic position and disadvantage available in this cohort instead of household income: (i) maternal education, collected by questionnaire at the time of recruitment, and (ii) the Index of Relative Socio-Economic Disadvantage (IRSD) from the 2011 Socio-Economic Indexes for Areas (SEIFA) (Pink, 2013), which is a composite of socioeconomic indexes for postcodes within Australia and is based on census data collected every 5 years. With these additional sensitivity analyses, we have now thoroughly considered adjustment for sociodemographic measures capturing different information at a household and neighbourhood level.

These sensitivity analyses are now mentioned in the methods (page 14) and the results (pages 20, 24), and presented as figures in Supplementary Files 1A to 1F. The information on maternal education and SEIFA have been added to the methods (page 9) and Table 1.

Please note that in light of this and other requested analyses (see comments S.6, S.8, R1.4, R2.5), we have restructured the Supplementary Files to present these various sensitivity analyses (including the breastfeeding duration and plasma processing time exclusion sensitivity analyses that were in the previous version of this manuscript) as combined forest plots (Supplementary Files 1A to 1F) allowing for easier visual comparison.

S6. Are pregnancy-related covariates available? Particularly adjustment for gestational age at delivery, chronic or gestational hypertension, preeclampsia, and/or chronic or gestational diabetes.

These covariates (gestational age at delivery, gestational hypertension, preeclampsia, and gestational diabetes) are available. All models have now been adjusted for gestational age. This has attenuated the relationship previously observed between number of infections and the DHA and omega 3 fatty acid measures, and modified some of the associations in the lipidomics data, but the overall findings of the study are otherwise unchanged.

Only small proportions of infants in this cohort with available data were exposed to pre-eclampsia (3.8%) or gestational diabetes (5.5%), so additional adjustment of the primary models for pre-eclampsia and gestational diabetes (and postnatal smoke exposure, see response to R1.4) as now been performed as sensitivity analyses. As gestational hypertension was used to diagnose pre-eclampsia, we have only adjusted for pre-eclampsia and not gestational hypertension.

These sensitivity analyses are now mentioned in the methods (page 14) and the results (pages 20, 24), and presented as figures in Supplementary File 1A to 1F. The information on pre-eclampsia and gestational diabetes data have been added to the methods (page 9) and Table 1.

S7. All results in the text should include measures of imprecision such as 95% CI.

This was originally excluded from the main text (and instead all included in the relevant Source Data) out of concern that it would make the results harder to parse due to the number of results reported. However, following this request, we have now included model estimates and the corresponding 95% CI for all results in the text.

S8. Were the children screened and excluded if they had an active infection at the time of measures of inflammatory markers, metabolomics, and lipidomics? How was this assessed?

Active infection was not an exclusion criterion for participation in this analysis. However, following this comment, we have considered sensitivity analyses excluding infants that may have had an active acute infection based on elevated hsCRP (>5 mg/L as a conservative cut-off indicative of acute infection (Lemiengre et al., 2018; Verbakel et al., 2016)) at 12 months of age.

These sensitivity analyses are now mentioned in the methods (page 14) and the results (pages 20, 24), and presented as figures in Supplementary Files 1A to 1F.

S9. The authors may consider the following reference to align with the AGReMA guidelines for reporting mediation analyses. Lee et al. A Guideline for Reporting Mediation Analyses of Randomized Trials and Observational Studies: The AGReMA Statement. JAMA. 2021;326(11):1045-1056. doi:10.1001/jama.2021.14075; Particularly, a DAG would be useful.

We have now added some revised the mediation analyses and the accompanying text to better align the manuscript with the AGReMA guidelines. This includes:

a) Revising the statistical package used and the estimates of interest. Previously, we used the ‘mediation’ package to estimate average causal direct and indirect effects and calculate percentage mediation from those. However, we felt it would be more appropriate in the context of the AGReMA guidelines and review comment S.10 to instead estimate the natural direct and indirect effects and include those in the results alongside the percentage mediation. For this, we have now used the ‘medflex’ package (Steen, Loeys, Moerkerke, and Vansteelandt, 2017) in R, and have revised the methods text accordingly (page 15-16). Note that we have now calculated percentage mediation based on the natural direct and indirect effects. These percentages are similar (within ~3%) to those using the previous method with average casual effects, and this change in mediation method has not impacted on our overall findings. However, while this revised method has not changed our findings, we feel it is more appropriate with regards to aligning with reporting principles laid out in the AGReMA guidelines.

b) The addition of a representative DAG for the mediation model as a new Figure (Figure 1).

c) Clarification in the methods that a counterfactual-based framework has been used for mediation analyses, and that we are estimating natural direct and indirect effects (page 15, line 297-298).

We have also revised the display item: we have now included a new Figure (Figure 7) to visually display these total and indirect effects as requested in R3.4. What was previously Table 2 has been revised in line with the comment S.10 and is now the Source Data 1 for Figure 7.

S10. Table 2 should include direct effects and clarity on what unit of change for each exposure, mediator, and outcome are the statistics based on? What is a 1-unit change in hsCRP reflect? since these units may vary significantly, use of 1 SD change would be optimal.

As described in the response to comment S.9, we have now added direct and indirect effects to what was previously Table 2 (now Source Data 1 for Figure 7). As in the other analyses in this study, the units are 1 infection change for number of parent-reported infections, and 1 SD change for log-transformed inflammation, metabolomic, and lipidomic measures. This has been included in the figure legend and has now also been added to the footnotes of the Source Data for clarity.

S11. One technical concern is the variation in the amount of time the blood samples spent between post-processing and storage at -80C (and how they were held during that interval). The authors should be commended for their sensitivity analysis looking at just those samples that were stored for < 4 hours. However, the justification for using only those < 4 hours is not clear. It is also unclear and a bit difficult for a reader to look at Supplementary files (1A and 1B) and draw the authors' conclusion that "this had little difference on the estimated effect sizes observed in analyses with the full cohort". Could the authors make more direct comparisons between the effect sizes to hammer this point home? Given that the sample size is smaller for this analysis, the p-values should be higher (less power), but the effect sizes should be unbiased and relatively similar. Perhaps calculating the correlation between effect sizes would help improve this sensitivity analysis and be good to report in the main text.

The <4-hour threshold was selected as a relatively conservative cut-off given the long tail of the sample processing time (see Author response image 1 for a histogram of sample processing time). The majority of samples that were processed within the same day were processed within 4 hours, while samples with >4h processing time were predominantly samples processed the following day after collection. We have now clarified this in the methods section (page 10, lines 170-174). We agree that visually comparing the effect sizes would make the comparison between models easier and have now included a visual comparison for these as a forest plot including this and other supplementary analyses that have been added (e.g., responses to S.6 and S.7) as a supplementary file (Supplementary Files 1A to 1F). We feel that direct comparison of the forest plots may be more readily informative than providing correlation coefficients.

Author response image 1
Download asset Open asset
Distribution of 12-month plasma processing times for the participants in this study.

S12. The authors show that while the correlation between GlycA (or CRP) and number of infections is relatively low (albeit a bit higher for GlycA), the effect sizes of GlycA and infections on the metabolome and lipidome are strongly correlated (and to a lesser extent between CRP and infections). A third comparison here that would be very useful, would be between GlycA and CRP effects on the metabolome and lipidome.

We have now added this comparison between GlycA and hsCRP for both the metabolomic and lipidomic data to the results text (page 21, line 369, and page 24, line 445) and as additional panels in the relevant figures for hsCRP effect sizes (Figures 4C, 6C, and corresponding figure supplements).

S13. The mediation analysis is the least convincing part of the manuscript. It is appreciated that the authors are trying to identify how infections might be translating to adverse metabolomic and lipidomic profiles. However, the justification for GlycA or hsCRP being the mediating mechanisms is not convincing. The authors are implying (statistically through their mediation models) that the effect of the number of infections on many metabolites and lipids is due to the changes in GlycA (or hsCRP). Since both GlycA and number of infections are cumulative measures at 12mo, it is unlikely that one (e.g., GlycA) precedes the other. Rather, as the authors state, GlycA is a marker of cumulative infections. This would preclude it from being a mediator unless the authors have data from earlier timepoints (like previous months) that would provide more support for a mediation effect.

First, we agree that the distinction between infections and GlycA was not made as clearly in the previous version of this manuscript as it should have been. For clarity, we do not consider early life GlycA to solely reflect cumulative infections, but rather consider that early life infections may be a key determinant of early life GlycA, with other non-infectious (and unquantified infectious and non-infectious) exposures also contributing to GlycA levels. Therefore, the intention was not to convey GlycA as a marker solely of cumulative infection burden. We appreciate that describing GlycA in parts of the manuscript as a marker of infection burden did not make this distinction clear, so we have revised these parts to describe GlycA as a marker of inflammation burden instead (e.g., lines 334, 343, 400, 628).

There are few studies that have considered longitudinal changes in GlycA, however studies with longitudinal GlycA measurements in adolescents and adults have reported that recent infection (within 2-3 weeks prior of blood collection) was associated with higher GlycA (Chiesa et al., 2022; Ritchie et al., 2015). There are also several studies investigating COVID-19 that have reported higher GlycA following acute infection compared to non-infected controls or reference populations (Ballout et al., 2021; Bizkarguenaga et al., 2022; Lodge et al., 2021). Together, these studies support the direction of the mediation model we have investigated in this manuscript. However, there is also some evidence in adults that GlycA may predict subsequent risk of infection (Julkunen, Cichońska, Slagboom, Würtz, and Initiative, 2021; Ritchie et al., 2015), and it has been speculated that higher GlycA may reflect a heightened responsiveness of the immune system or immune ‘training’, as we have recently reported from this cohort (Collier et al., 2022).

While originally we did not anticipate having data from an earlier time point than 12 months of age for this study, we very recently received NMR metabolomics data from 6-month plasma, sooner than anticipated, so we have now been able to perform additional analyses utilising these 6-month data. We believe this additional time point greatly strengthens the main findings of this manuscript and in particular addresses the valid queries raised by the reviewers in relation to the causal model.

While some of these additional analyses were not directed requested by the reviewers, we felt it important to include these to strengthen the paper alongside the analyses that were suggested.

Specifically, we have added the following additional analyses to address specific points as follows:

a) We present the additional cross-sectional analysis to provide evidence for a relationship between infection burden from birth to 6 months, 6-month inflammation, and 6-month metabolomic/lipidomic profiles to highlight that some of these associations between infection and omic profiles are evident at 6 months, albeit less so than at 12 months of age.

We therefore provide analysis investigating the relationship between number of parent reported infections from birth to 6 months of age, 6-month inflammation, and 6-month metabolomic/lipidomic measures (methods on page 14, lines 274-276; results on pages 21 and 24, Figure 3—figure supplement 1, Figure 4—figure supplement 1, Figure 5—figure supplement 2, Figure 6—figure supplement 2).

b) To provide evidence that the relationship between infection and adverse omics profiles is less likely to be confounded by unmeasured/unknown early life factors, we further analysed the associations between infections from 6-12 months with 12 month omics, adjusting for the corresponding exposure at 6-months (number of infections from birth to 6 months of age, 6-month GlycA, or 6-month hsCRP) and the corresponding 6-month metabolomic/lipidomic measure (methods on page 14-15, lines 277-280; results on pages 21-22 and 24-25, Figure 3—figure supplement 2, Figure 4—figure supplement 2, Figure 5—figure supplement 3, Figure 6—figure supplement 3).

c) To investigate reverse causality in greater detail, we provide an extension of the previous models of birth metabolomic/lipidomic measures potentially associating with subsequent number of infections. We now have additional data investigating the association of the 6-month metabolomic/lipidomic measurements with number of infections from 6 to 12 months of age, with adjustment for number of infections from birth to 6 months of age (as suggested in comment R3.3) (methods on page 15, lines 283-286; results on pages 22 and 25, Supplementary Files 3B and 3D). Please note that as part of these more in-depth analyses, we have revised the previous reverse causality models with birth metabolomic/lipidomic measures to have infections from birth to 6 months as the outcome instead of infections from birth to 12 months (Supplementary Files 3A and 3C).

d) To strengthen the mediation analyses by adjusting for potential confounding by earlier inflammation, we provide additional mediation analyses investigating 12-month inflammation as a mediator of the effect of number of infections from 6 to 12 months of age on 12-month metabolomic/lipidomic measures, with adjustment for number of infections from birth to 6 months of age, 6-month inflammation, and corresponding 6-month metabolomic/lipidomic measure (to address the concerns raised in this comment and in R2.11) (methods on page 16, results on page 26, Figure 7—figure supplement 1).

Of particular relevance to this specific comment, is evidence from these additional analyses that indicates some limited evidence that 6-month GlycA is modestly associated number of infections from 6 to 12 months of age, after adjustment for number of infections from birth to 6 months of age. This is in keeping with data from adults that GlycA may be a marker of inflammatory potential and predicts long-term risk of severe infection and infection-related mortality (Julkunen et al., 2021; Ritchie et al., 2015). Therefore, we have now included the additional mediation analyses (as mentioned above) investigating 12-month inflammation as a mediator in the relationship between infections from 6-12m and 12-month metabolomic/lipidomic measures, with adjustment for birth-6m infection, 6-month inflammation, and 6-month metabolomic/lipidomic measures. We believe that we have now controlled for the potential of reverse causality of 6-month GlycA on 6m-12m infections and reduced the risk of exposure-induced mediator-outcome confounding and similar confounding. The estimates of proportion mediated in these additional models were similar to those considering infections from birth to 12 months of age.

With these clarifications and additional analyses, we feel these additional analyses strengthen both the rationale and the findings from the mediation analyses.

Reviewer #1 (Recommendations for the authors):

This is an interesting paper and a large amount of data are provided. I have the following comments:

R1.1. The paper contains a number of grammatical, spelling and structural errors. The authors are strongly encouraged to review the paper carefully prior to resubmitting the paper.

This is S.1 above, please refer to our response there.

R1.2. Figures 4 and 5 are difficult to understand. What do each point in the blot represent? What does the line around the data points represent? It is important the authors provide more explanation for readers who do not have familiarity with these plots.

This is S.2 above, please refer to our response there.

R1.3. The authors need to better define the significance of their findings. For example, although the link between inflammation and changes in metabolomics and lipidomics is known, the authors need to provide a better explanation on why their specific findings are important in the setting of infection in infants.

This is S.3 above, please refer to our response there.

R1.4. Additional environmental variables and their link to changes in childhood serum lipids and metabolome after infection need to be studied and included in the paper.

We have now considered models with additional adjustment for postnatal smoke exposure (methods: page 9, line 147-151, results: Supplementary Files 1A to 1F). We have not adjusted for body weight at 12 months, as effects on 12-month body weight may be in the casual pathway from infections to metabolomic/lipidomic differences. However, all models are adjusted for birth weight z-score. We have also now additionally adjusted all models for gestational age, as well as considered additional adjustment for pre-eclampsia and gestational diabetes (as suggested in S.6) alongside postnatal smoke exposure (2.8% of infants exposed to any postnatal smoke up to 12 months) as sensitivity analysis.

We have also rephrased the comment in the limitations section of the discussion about potential unmeasured confounding to specifically mention environmental factors linked to changes in metabolomic/lipidomic measures after infection (page 31, line 607-609):

“Despite rigorous adjustment and sensitivity analyses, there may be unmeasured confounding, such as from environmental factors that may modify changes in metabolomic/lipidomic measures following infection.”

Reviewer #2 (Recommendations for the authors):

Mansell et al. report on associations between infections between birth and 12 months of age and metabolomic/lipidomic profiles. There are several suggestions that may strengthen the study and manuscript:

R2.1) The background and rationale of the introduction should clarify the importance of transient inflammation with infections compared with cumulative burden of inflammation. The authors are suggesting that a higher burden of infections, which are common in the first year of life, are associated with long-lasting inflammation. This does not pan out in terms of hsCRP but does in terms of GlycA but this is not a known robust marker of long-term inflammation. This should be tempered.

This is S.4 above, please refer to our response there.

R2.2) The introduction suggests that inflammation may be a causal target but this analysis can only identify potential markers. This should be tempered to distinctly argue against any causal findings in this observational analysis?

There is evidence in adults from a randomised control trial of anti-inflammatory medication (e.g. CANTOS, the Canakinumab Anti-Inflammatory Thrombosis Outcomes Study) (Ridker et al., 2017) and trials of colchicine (Chunfeng, Ping, Yun, Di, and Qi, 2021) that supports inflammation as a causal target to reduce risk of cardiovascular disease. While we had been careful to avoid suggesting that the findings from our study in infants were causal findings, we have now added some text in the Discussion to make this distinction more explicit (page 31, line 602-606):

“Evidence from randomised control trials in adults support a causal role of inflammation in cardiovascular disease risk (Chunfeng et al., 2021; Ridker et al., 2017), however this study in infants is observational. While we have used a casual framework for mediation analyses, our findings do not demonstrate causality.”

R2.3) Are social factors available? For example, a child with two working parents who is placed in daycare is likely to have a higher burden of infections in contrast with one with a stay at home parent? Household income and daycare status would be useful covariates to consider.

This is S.5 above, please refer to our response there.

R2.4) Are pregnancy-related covariates available? Particularly adjustment for gestational age at delivery, chronic or gestational hypertension, preeclampsia, and/or chronic or gestational diabetes.

This is S.6 above, please refer to our response there.

R2.5) Are higher order gestations (e.g., twins, triplets) excluded from the analysis? If not, a sensitivity analysis excluding them should be considered due to confounding from relatedness and shared maternal characteristics in models

Twins were not excluded from the model, but we have now added sensitivity analyses excluding the 5 twins (2 twin pairs, and 1 infant who was a twin to sibling not included in this study due to a lack of 12-month data) from the primary model (methods on page 14, results figures in Supplementary File 1A to 1F).

R2.6) All results in the text should include measures of imprecision such as 95% CI

This is S.7 above, please refer to our response there.

R2.7) Were the children screened and excluded if they had an active infection at the time of measures of inflammatory markers, metabolomics, and lipidomics? How was this assessed?

This is S.8 above, please refer to our response there.

R2.8) The authors may consider the following reference to align with the AGReMA guidelines for reporting mediation analyses. Lee et al. A Guideline for Reporting Mediation Analyses of Randomized Trials and Observational Studies: The AGReMA Statement. JAMA. 2021;326(11):1045-1056. doi:10.1001/jama.2021.14075; Particularly, a DAG would be useful.

This is S.9 above, please refer to our response there.

R2.9) The language that GlycA outperforms hsCRP is unclear how this was demonstrated. The authors should provide direct comparisons and supportive data or remove these statements.

There are data from adults, as cited in the introduction, that GlycA improves risk stratification for CVD, even after adjusting for hsCRP. We have added a sentence to the introduction that briefly elaborates on this statement in relation to CVD risk (page 6, lines 78-82):

“For example, in the Multi-Ethnic Study of Atherosclerosis (n=6523), higher GlycA was associated with increased risk of incidence CVD and death, after adjustment for hsCRP and other inflammatory markers. Conversely, prediction of these outcomes by hsCRP attenuated to null after mutual adjustment (Duprez et al., 2016).”

R2.10) Table 2 should include direct effects and clarity on what unit of change for each exposure, mediator, and outcome are the statistics based on? What is a 1-unit change in hsCRP reflect? since these units may vary significantly, use of 1 SD change would be optimal

This is S.10 above, please refer to our response there.

R2.11) since metabolomics and lipidomics were measured at birth, adjustment for these variables in the mediation analysis may be helpful to mitigate concerns of residual confounding

We have now performed adjusted mediation analysis (discussed in more detail in the response to S.13 above) by assessing mediation by 12-month inflammation in the effect of number of infections from 6 months to 12 months on 12-month metabolomic/lipidomic measures, with adjustment for number of infection from birth to 6 months of age, 6-month inflammation, and 6-month metabolomic/lipidomic measure. We believe this adjusted mediation analysis addresses this concern of residual confounding.

Reviewer #3 (Recommendations for the authors):

R3.1. I think the authors should take care in avoiding technical jargon and/or explaining it before using it. For instance, there are multiple points where acronyms are introduced without being first defined, like NMR, TGs, and TG-Os (which I think are nuclear magnetic resonance, and two types of triglycerides). Additionally, technical modifications like lines 211-213 are difficult to parse. What does "used medronic acid to passivate… to avoid peak tailing.." mean to a non-expert?

We have now spelled out all acronyms in full at first use. We have also revised the lipidomics Results sections and figures to include the lipid class names in full instead of predominantly using acronyms for lipid classes.

We appreciate that the omics methodological section is relatively technical; we have attempted to strike a balance being concise and providing sufficient methodological details to reassure those with relevant laboratory expertise that the methods were robust. We would therefore prefer to leave this text unchanged if the editors agree.

R3.2. L229: I understand that 1 is added so that you can log-transform any zero values, but is 1 a good value for all metabolites and lipids? Could this offset be scaled to each individual metabolite and lipid based on their lowest non-zero value, such that it doesn't affect the results as much? Additionally, what proportion of the metabolites/lipids measures had 0 values in the samples (and how many 0 values)? In other words: how big of an issue could this +1 offset be?

Only a small number of metabolomic/lipidomic measures had any 0 values, and most of these measures had only a small number of participants with 0 values. We have added a column for each measure in Supplementary Files 2A and 2B stating the number of 0 values for that measure.

Given that the magnitude of the metabolomic/lipidomic measures differs substantially, we have now redone the analysis as suggested, by offsetting each measure by its minimum non-zero value prior to log-transformation, instead of offsetting all measures by 1. We have also added a column to the tables in Supplementary Files 2A and 2B stating this minimum non-zero measure for each measure.

R3.3. The reverse-causality analysis is really interesting. It would be great if the authors had intermediate time-points (6mo?) rather than just birth to see if there could be some link between inflammatory levels at 6mo (controlling for infection burden) and later life levels.

As described in the response to S.13, we have recently received corresponding metabolomic data for the 6-month time point. As such, we have been able to update the reverse causality modelling: now, we consider the association between birth metabolomic/lipidomic measures and number of infections from birth to 6m, and the association between 6-month metabolomic/lipidomic measures and number of infections from 6m to 12m with adjustment for number of infections from birth to 6m.

R3.4. The use of the forest plots is much appreciated, but the effect sizes and CIs are overlapping, so it's really hard to interpret. Especially the lipidomic figures (Figures4 and 5). Here, it might be best to just show the "class totals" and not every species underlying those classes.

The mediation analysis results could also be presented as a forest plot instead of a table. That would make interpretation much easier for the reader.

We have made some revisions to the figures to make them easier to read. Specifically:

a) We have reduced the scale of the x-axis and also increased the width of the NMR metabolomics forest plots for GlycA and hsCRP to avoid the issue of the effect size and CIs overlapping.

b) We have simplified the main LC/MS lipdiomics forest plots to now only show the lipid class totals and not the individual lipid species. We have also now indicated adjusted p>0.05 with an open point and adjusted p<0.05 with a closed point for the lipid class totals to be consistent with the NMR forest plots. We have included simplified versions of the forest plot showing the individuals species as figure supplements: these simplified versions now do not highlight the top 10 species by p-value and also do not show the CIs for the individual lipid species. We have also modified the shades for the individual species to make them less obtrusive on the class total effect sizes and CIs.

As requested, we have also now included a new figure (Figure 7) showing the indirect and total effects of infection for each of the metabolomic/lipidomic measures investigated in the mediation analyses. The Source Data for this figure includes the direct effect as well.

References

Akinkuolie, A. O., Buring, J. E., Ridker, P. M., and Mora, S. (2014). A novel protein glycan biomarker and future cardiovascular disease events. Journal of the American Heart Association, 3(5), e001221.

Alvarez, C., and Ramos, A. (1986). Lipids, lipoproteins, and apoproteins in serum during infection. Clinical Chemistry, 32(1), 142-145.

Ballout, R. A., Kong, H., Sampson, M., Otvos, J. D., Cox, A. L., Agbor-Enoh, S., and Remaley, A. T. (2021). The NIH Lipo-COVID Study: A pilot NMR investigation of lipoprotein subfractions and other metabolites in patients with severe COVID-19. Biomedicines, 9(9), 1090.

Bizkarguenaga, M., Bruzzone, C., Gil‐Redondo, R., SanJuan, I., Martin‐Ruiz, I., Barriales, D.,... Laín, A. (2022). Uneven metabolic and lipidomic profiles in recovered COVID‐19 patients as investigated by plasma NMR metabolomics. NMR in Biomedicine, 35(2), e4637.

Bruzzone, C., Bizkarguenaga, M., Gil-Redondo, R., Diercks, T., Arana, E., de Vicuña, A. G.,... San Juan, I. (2020). SARS-CoV-2 infection dysregulates the metabolomic and lipidomic profiles of serum. Iscience, 23(10), 101645.

Chiesa, S. T., Charakida, M., Georgiopoulos, G., Roberts, J. D., Stafford, S. J., Park, C.,... Ala‐Korpela, M. (2022). Glycoprotein Acetyls: A Novel Inflammatory Biomarker of Early Cardiovascular Risk in the Young. Journal of the American Heart Association, e024380.

Chunfeng, L., Ping, L., Yun, Z., Di, L., and Qi, W. (2021). Colchicine for Coronary Heart Disease: A Meta-Analysis of Randomized Controlled Trials. Heart Surg Forum, 24(5), E863-e867. doi:10.1532/hsf.3609

Collier, F., Chau, C., Mansell, T., Faye-Chauhan, K., Vuillermin, P., Ponsonby, A.-L.,... Burgner, D. (2022). Innate Immune Activation and Circulating Inflammatory Markers in Preschool Children. Frontiers in Immunology, 12. doi:10.3389/fimmu.2021.830049

Collier, F., Ellul, S., Juonala, M., Ponsonby, A.-L., Vuillermin, P., Saffery, R.,... on behalf of the Barwon Infant Study Investigator, G. (2019). Glycoprotein acetyls (GlycA) at 12 months are associated with high-sensitivity C-reactive protein and early life inflammatory immune measures. Pediatr Res, 85(5), 584-585. doi:10.1038/s41390-019-0307-x

Connelly, M. A., Otvos, J. D., Shalaurova, I., Playford, M. P., and Mehta, N. N. (2017). GlycA, a novel biomarker of systemic inflammation and cardiovascular disease risk. Journal of translational medicine, 15(1), 219.

Duprez, D. A., Otvos, J., Sanchez, O. A., Mackey, R. H., Tracy, R., and Jacobs Jr, D. R. (2016). Comparison of the predictive value of GlycA and other biomarkers of inflammation for total death, incident cardiovascular events, noncardiovascular and noncancer inflammatory-related events, and total cancer events. Clinical Chemistry, 62(7), 1020-1031.

Gallin, J. I., Kaye, D., and O'Leary, W. M. (1969). Serum lipids in infection. New England Journal of Medicine, 281(20), 1081-1086.

Gao, D., Zhang, L., Song, D., Lv, J., Wang, L., Zhou, S.,... Zhang, J. (2019). Values of integration between lipidomics and clinical phenomes in patients with acute lung infection, pulmonary embolism, or acute exacerbation of chronic pulmonary diseases: a preliminary study. Journal of translational medicine, 17(1), 1-20.

Julkunen, H., Cichońska, A., Slagboom, P. E., Würtz, P., and Initiative, N. H. U. B. (2021). Metabolic biomarker profiling for identification of susceptibility to severe pneumonia and COVID-19 in the general population. eLife, 10, e63033.

Kyle, J., Burnum-Johnson, K., Wendler, J., Eisfeld, A., Halfmann, P. J., Watanabe, T.,... Waters, K. (2019). Plasma lipidome reveals critical illness and recovery from human Ebola virus disease. Proceedings of the National Academy of Sciences, 116(9), 3919-3928.

Lemiengre, M. B., Verbakel, J. Y., Colman, R., Van Roy, K., De Burghgraeve, T., Buntinx, F.,... De Sutter, A. (2018). Point-of-care CRP matters: normal CRP levels reduce immediate antibiotic prescribing for acutely ill children in primary care: a cluster randomized controlled trial. Scandinavian journal of primary health care, 36(4), 423-436.

Liuba, P., Persson, J., Luoma, J., Ylä-Herttuala, S., and Pesonen, E. (2003). Acute infections in children are accompanied by oxidative modification of LDL and decrease of HDL cholesterol, and are followed by thickening of carotid intima–media. European heart journal, 24(6), 515-521.

Lodge, S., Nitschke, P., Kimhofer, T., Wist, J., Bong, S.-H., Loo, R. L.,... Lindon, J. C. (2021). Diffusion and relaxation edited proton NMR spectroscopy of plasma reveals a high-fidelity supramolecular biomarker signature of SARS-CoV-2 infection. Analytical chemistry, 93(8), 3976-3986.

Madsen, C. M., Varbo, A., Tybjærg-Hansen, A., Frikke-Schmidt, R., and Nordestgaard, B. G. (2018). U-shaped relationship of HDL and risk of infectious disease: two prospective population-based cohort studies. European heart journal, 39(14), 1181-1190.

Pink, B. (2013). Socio-economic indexes for areas (SEIFA) 2011. Canberra: Australian Bureau of Statistics.

Ridker, P. M., Everett, B. M., Thuren, T., MacFadyen, J. G., Chang, W. H., Ballantyne, C.,... Anker, S. D. (2017). Antiinflammatory therapy with canakinumab for atherosclerotic disease. New England Journal of Medicine, 377(12), 1119-1131.

Ritchie, S. C., Würtz, P., Nath, A. P., Abraham, G., Havulinna, A. S., Fearnley, L. G.,... Aalto, K. (2015). The biomarker GlycA is associated with chronic inflammation and predicts long-term risk of severe infection. Cell Systems, 1(4), 293-301.

Steen, J., Loeys, T., Moerkerke, B., and Vansteelandt, S. (2017). Medflex: an R package for flexible mediation analysis using natural effect models. Journal of statistical software, 76, 1-46.

Verbakel, J. Y., Lemiengre, M. B., De Burghgraeve, T., De Sutter, A., Aertgeerts, B., Shinkins, B.,... Buntinx, F. (2016). Should all acutely ill children in primary care be tested with point-of-care CRP: a cluster randomised trial. BMC Med, 14(1), 1-7.

Weintraub, W. S., Daniels, S. R., Burke, L. E., Franklin, B. A., Goff, D. C., Jr., Hayman, L. L.,... Whitsel, L. P. (2011). Value of primordial and primary prevention for cardiovascular disease: a policy statement from the American Heart Association. Circulation, 124(8), 967-990. doi:10.1161/CIR.0b013e3182285a81

https://doi.org/10.7554/eLife.75170.sa2

Article and author information

Author details

  1. Toby Mansell

    1. Murdoch Children's Research Institute, Parkville, Australia
    2. Department of Paediatrics, University of Melbourne, Parkville, Australia
    Contribution
    Conceptualization, Data curation, Formal analysis, Methodology, Software, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    TM has received a postdoctoral fellowship from MCRI, is supported by NHMRC funding, has received travel support from MCRI and the University of Melbourne, and received a PhD scholarship from the University of Melbourne
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1282-6331

  2. Richard Saffery

    1. Murdoch Children's Research Institute, Parkville, Australia
    2. Department of Paediatrics, University of Melbourne, Parkville, Australia
    Contribution
    Conceptualization, Funding acquisition, Methodology, Resources, Supervision, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  3. Satvika Burugupalli

    Metabolomics Laboratory, Baker Heart and Diabetes Institute, Melbourne, Australia
    Contribution
    Data curation, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared

  4. Anne-Louise Ponsonby

    1. Murdoch Children's Research Institute, Parkville, Australia
    2. Department of Paediatrics, University of Melbourne, Parkville, Australia
    3. The Florey Institute of Neuroscience and Mental Health, Parkville, Australia
    Contribution
    Funding acquisition, Methodology, Writing – review and editing
    Competing interests
    ALP is an unpaid scientific advisor for, and has shares in, Dysrupt Labs. ALP has shares in Prevatex Pty Ltd
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6581-3657

  5. Mimi LK Tang

    1. Murdoch Children's Research Institute, Parkville, Australia
    2. Department of Paediatrics, University of Melbourne, Parkville, Australia
    3. Royal Children’s Hospital, Parkville, Australia
    Contribution
    Writing – review and editing
    Competing interests
    MLKT has received funding paid to Murdoch Childen's Research Institute (MCRI) from NHMRC, Prota Theraputics, Abbott Nutrition, the Allergy and Immunology Foundation of Australasia, and the National Children's Research Centre of Ireland, and has received internal research funding from MCRI. MLKT is inventor of 2 patents owned by MCRI relating to allergy treatment and a method to induce tolerance to an allergen. MLKT is a member of the Advisory Boards for Pfizer (has received personal fee) and Anaphylaxis & Anaphylaxis Australia, and of allergy/anaphylaxis-related Committees for the World Allergy Organisation, the International Union of Immunological Societies, the Asia Pacific Association of Allergy Asthma and Clinical Immunology, the American Academy of Allergy Asthma and Immunology, and the Australasian Society of Clinical Immunology and Allergy. MLKT is employee of, and has share options in, Prota Therapeutics. MLKT is an Associate Editor for the Journal of Allergy and Clinical Immunology: Global

  6. Martin O'Hely

    1. Murdoch Children's Research Institute, Parkville, Australia
    2. Deakin University, Geelong, Australia
    Contribution
    Methodology, Writing – review and editing
    Competing interests
    MOH has stocks in Prevatex Pty Ltd

  7. Siroon Bekkering

    1. Murdoch Children's Research Institute, Parkville, Australia
    2. Department of Internal Medicine, Radboud Institute for Molecular Life Sciences, Radboud University Medical Centre, Nijmegen, Netherlands
    Contribution
    Resources, Writing – review and editing
    Competing interests
    S Bekkering has received postdoctoral grants from the Dutch Heart Foundation and the Dutch Research Council, and travel support from the European Society for Atherosclerosis
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1149-466X

  8. Adam Alexander T Smith

    Metabolomics Laboratory, Baker Heart and Diabetes Institute, Melbourne, Australia
    Contribution
    Software, Writing – review and editing
    Competing interests
    No competing interests declared

  9. Rebecca Rowland

    Murdoch Children's Research Institute, Parkville, Australia
    Contribution
    Data curation, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared

  10. Sarath Ranganathan

    1. Murdoch Children's Research Institute, Parkville, Australia
    2. Department of Paediatrics, University of Melbourne, Parkville, Australia
    3. Royal Children’s Hospital, Parkville, Australia
    Contribution
    Writing – review and editing
    Competing interests
    SR is Director of the Lung Foundation Australia. SR has stocks/options in Prevatex Pty Ltd

  11. Peter D Sly

    1. Murdoch Children's Research Institute, Parkville, Australia
    2. Child Health Research Centre, University of Queensland, Brisbane, Australia
    Contribution
    Writing – review and editing
    Competing interests
    No competing interests declared

  12. Peter Vuillermin

    1. Murdoch Children's Research Institute, Parkville, Australia
    2. Deakin University, Geelong, Australia
    3. Child Health Research Unit, Barwon Health, Geelong, Australia
    Contribution
    Writing – review and editing
    Competing interests
    PV is an inventor on a patent relating to the relationship between maternal carriage of Prevotella. copri and offspring allergic disease, and has stocks/options in Prevatex Pty Ltd

  13. Fiona Collier

    1. Deakin University, Geelong, Australia
    2. Child Health Research Unit, Barwon Health, Geelong, Australia
    Contribution
    Data curation, Resources, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5438-480X

  14. Peter Meikle

    Metabolomics Laboratory, Baker Heart and Diabetes Institute, Melbourne, Australia
    Contribution
    Resources, Supervision, Writing – original draft, Writing – review and editing, Joint senior authors
    Competing interests
    No competing interests declared
    Additional information
    Joint senior authors
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2593-4665

  15. David Burgner

    1. Department of Paediatrics, University of Melbourne, Parkville, Australia
    2. Department of Paediatrics, Monash University, Clayton, Australia
    Contribution
    Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing – original draft, Writing – review and editing, Joint senior authors
    For correspondence
    david.burgner@mcri.edu.au
    Competing interests
    DB has received an Investigator Grant and Project Grant from the Australian National Health and Medical Research Council (NHMRC)
    Additional information
    Joint senior authors
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8304-4302

  16. Barwon Infant Study Investigator Group

    Competing interests
    TM has received a postdoctoral fellowship from MCRI, is supported by NHMRC funding, has received travel support from MCRI and the University of Melbourne, and received a PhD scholarship from the University of Melbourne
    1. John Carlin, Murdoch Children's Research Insitute, Royal Children's Hospital, Department of Paediatrics, University of Melbourne, Parkville, Australia
    2. Amy Loughman, Institute for Mental and Physical Health and Clinical Translation, School of Medicine, Barwon Health, Deakin University, Geelong, Australia
    3. Lawrence Gray, Child Health Research Unit, Barwon Health, Faculty of Health, School of Medicine, Deakin University, Geelong, Australia

Funding

National Health and Medical Research Council (Senior Research Fellowship 1008396)

  • Anne-Louise Ponsonby

National Health and Medical Research Council (Senior Research Fellowship 1064629)

  • David Burgner

National Health and Medical Research Council (Senior Research Fellowship 1045161)

  • Richard Saffery

National Health and Medical Research Council (Investigator Grant 1110200)

  • Anne-Louise Ponsonby

National Health and Medical Research Council (Investigator Grant 1175744)

  • David Burgner

National Health and Medical Research Council (NHMRC-A*STAR project grant)

  • Peter Meikle

Murdoch Children's Research Institute (ECR Fellowship)

  • Toby Mansell

Nederlandse Organisatie voor Wetenschappelijk Onderzoek (452173113)

  • Siroon Bekkering

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The authors thank the BIS participants for the generous contribution they have made to this project. The authors also thank current and past staff for their efforts in recruiting and maintaining the cohort and in obtaining and processing the data and biospecimens.

The members of the BIS Steering Committee are the following: Peter Vuillermin, Anne-Louise Ponsonby, John Carlin, Mimi LK Tang, Fiona Collier, Amy Loughman, Toby Mansell, Lawrence Gray, Martin O’Hely, Richard Saffery, Sarath Ranganathan, David Burgner, Peter Sly, and Leonard Harrison. We thank Terry Dwyer and Katie Allen for their past work as foundation investigators.

Ethics

Human subjects: Participating mothers provided informed consent at recruitment. Ethics approval for this study was granted by the Barwon Health Human Research Ethics Committee (HREC 10/24).

Senior Editor

  1. Matthias Barton, University of Zurich, Switzerland

Reviewing Editor

  1. Hossein Ardehali, Northwestern University, United States

Reviewer

  1. Noah Snyder-Mackler, Arizona State University, United States

Publication history

  1. Received: November 1, 2021
  2. Preprint posted: December 2, 2021 (view preprint)
  3. Accepted: April 24, 2022
  4. Version of Record published: May 10, 2022 (version 1)

Copyright

© 2022, Mansell et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 958
    Page views
  • 186
    Downloads
  • 4
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Toby Mansell
  2. Richard Saffery
  3. Satvika Burugupalli
  4. Anne-Louise Ponsonby
  5. Mimi LK Tang
  6. Martin O'Hely
  7. Siroon Bekkering
  8. Adam Alexander T Smith
  9. Rebecca Rowland
  10. Sarath Ranganathan
  11. Peter D Sly
  12. Peter Vuillermin
  13. Fiona Collier
  14. Peter Meikle
  15. David Burgner
  16. Barwon Infant Study Investigator Group
(2022)
Early life infection and proinflammatory, atherogenic metabolomic and lipidomic profiles in infancy: a population-based cohort study
eLife 11:e75170.
https://doi.org/10.7554/eLife.75170

Categories and tags

Research organism

Further reading

    1. Epidemiology and Global Health

    Perceived barriers to cervical cancer screening and motivators for at-home human papillomavirus self-sampling during the COVID-19 pandemic: Results from a telephone survey

    Susan L Parker, Ashish A Deshmukh ... Jane R Montealegre
    Research Article

    Background: Home-based self-sampling for human papillomavirus (HPV) testing may be an alternative for women not attending clinic-based cervical cancer screening.

    Methods: We assessed barriers to care and motivators to use at-home HPV self-sampling kits during the COVID-19 pandemic as part of a randomized controlled trial evaluating kit effectiveness. Participants were women aged 30-65 and under-screened for cervical cancer in a safety-net healthcare system. We conducted telephone surveys in English/Spanish among a subgroup of trial participants, assessed differences between groups, and determined statistical significance at p<0.05.

    Results: Over half of 233 survey participants reported that clinic-based screening (Pap) is uncomfortable (67.8%), embarrassing (52.4%), and discomfort seeing male providers (63.1%). The last two factors were significantly more prevalent among Spanish versus English speakers (66.4% vs. 30% (p=0.000) and 69.9 vs. 52.2% (p=0.006), respectively). Most women who completed the kit found Pap more embarrassing (69.3%), stressful (55.6%), and less convenient (55.6%) than the kit. The first factor was more prevalent among Spanish versus English speakers (79.6% vs. 53.38%, p=0.001) and among patients with elementary education or below.

    Conclusions: The COVID-19 pandemic influenced most (59.5%) to participate in the trial due to fear of COVID, difficulty making appointments, and ease of using kits. HPV self-sampling kits may reduce barriers among under-screened women in a safety-net system.

    Funding: This study is supported by a grant from the National Institute for Minority Health and Health Disparities (NIMHD, R01MD013715, PI: JR Montealegre).

    Clinical trial number: NCT03898167.

    1. Epidemiology and Global Health

    Approximating missing epidemiological data for cervical cancer through Footprinting: A case study in India

    Irene Man, Damien Georges ... Iacopo Baussano
    Research Article

    Local cervical cancer epidemiological data essential to project the context-specific impact of cervical cancer preventive measures are often missing. We developed a framework, hereafter named Footprinting, to approximate missing data on sexual behaviour, human papillomavirus (HPV) prevalence, or cervical cancer incidence, and applied it to an Indian case study. With our framework, we (1) identified clusters of Indian states with similar cervical cancer incidence patterns, (2) classified states without incidence data to the identified clusters based on similarity in sexual behaviour, (3) approximated missing cervical cancer incidence and HPV prevalence data based on available data within each cluster. Two main patterns of cervical cancer incidence, characterized by high and low incidence, were identified. Based on the patterns in the sexual behaviour data, all Indian states with missing data on cervical cancer incidence were classified to the low-incidence cluster. Finally, missing data on cervical cancer incidence and HPV prevalence were approximated based on the mean of the available data within each cluster. With the Footprinting framework, we approximated missing cervical cancer epidemiological data and made context-specific impact projections for cervical cancer preventive measures, to assist public health decisions on cervical cancer prevention in India and other countries.

    1. Epidemiology and Global Health
    2. Microbiology and Infectious Disease

    Integrating contact tracing and whole-genome sequencing to track the elimination of dog-mediated rabies: an observational and genomic study

    Kennedy Lushasi, Kirstyn Brunker ... Katie Hampson
    Research Article

    Background:

    Dog-mediated rabies is endemic across Africa causing thousands of human deaths annually. A One Health approach to rabies is advocated, comprising emergency post-exposure vaccination of bite victims and mass dog vaccination to break the transmission cycle. However, the impacts and cost-effectiveness of these components are difficult to disentangle.

    Methods:

    We combined contact tracing with whole-genome sequencing to track rabies transmission in the animal reservoir and spillover risk to humans from 2010-2020, investigating how the components of a One Health approach reduced the disease burden and eliminated rabies from Pemba Island, Tanzania. With the resulting high-resolution spatiotemporal and genomic data we inferred transmission chains and estimated case detection. Using a decision tree model we quantified the public health burden and evaluated the impact and cost-effectiveness of interventions over a ten-year time horizon.

    Results:

    We resolved five transmission chains co-circulating on Pemba from 2010 that were all eliminated by May 2014. During this period, rabid dogs, human rabies exposures and deaths all progressively declined following initiation and improved implementation of annual islandwide dog vaccination. We identified two introductions to Pemba in late 2016 that seeded re-emergence after dog vaccination had lapsed. The ensuing outbreak was eliminated in October 2018 through reinstated islandwide dog vaccination. While post-exposure vaccines were projected to be highly cost-effective ($256 per death averted), only dog vaccination interrupts transmission. A combined One Health approach of routine annual dog vaccination together with free post-exposure vaccines for bite victims, rapidly eliminates rabies, is highly cost-effective ($1657 per death averted) and by maintaining rabies freedom prevents over 30 families from suffering traumatic rabid dog bites annually on Pemba island.

    Conclusions:

    A One Health approach underpinned by dog vaccination is an efficient, cost-effective, equitable and feasible approach to rabies elimination, but needs scaling up across connected populations to sustain the benefits of elimination, as seen on Pemba, and for similar progress to be achieved elsewhere.

    Funding:

    Wellcome [207569/Z/17/Z, 095787/Z/11/Z, 103270/Z/13/Z], the UBS Optimus Foundation, the Department of Health and Human Services of the National Institutes of Health [R01AI141712] and the DELTAS Africa Initiative [Afrique One-ASPIRE/DEL-15-008] comprising a donor consortium of the African Academy of Sciences (AAS), Alliance for Accelerating Excellence in Science in Africa (AESA), the New Partnership for Africa's Development Planning and Coordinating (NEPAD) Agency, Wellcome [107753/A/15/Z], Royal Society of Tropical Medicine and Hygiene Small Grant 2017 [GR000892] and the UK government. The rabies elimination demonstration project from 2010-2015 was supported by the Bill & Melinda Gates Foundation [OPP49679]. Whole-genome sequencing was partially supported from APHA by funding from the UK Department for Environment, Food and Rural Affairs (Defra), Scottish government and Welsh government under projects SEV3500 & SE0421.