CHARACTERIZATION OF EARLY LIFE EXPOSURE
TO ENVIRONMENTAL CHEMICALS AND ITS
IMPACTS ON HEALTH
School of Science and Technology
Örebro University
Sweden
Lisanna Sinisalu
Supervisors from Örebro University: Tuulia Hyötyläinen, Leo Yeung Examiner: Ingrid Ericson Jogsten
1
List of abbreviations
12-epiCA 3α, 7α, 12β-trihydroxy-5β-cholan-24-oic acid 12-oxo-LCA 12-oxolithocholic acid
18O2-PFHxS Perfluoro-1-hexane [18O2] sulfonic acid
4:2FTSA 4:2 fluorotelomer sulfonic acid
6:2 Cl-PFESA 6:2 chlorinated polyfluoroalkyl ether sulfonic acid 6:2FTSA 6:2 fluorotelomer sulfonic acid
7-oxo-DCA 7-oxodeoxycholic acid 7-oxo-HDCA 7-oxohyodeoxycholic acid
8:2 Cl-PFESA 8:2 chlorinated polyfluoroalkyl ether sulfonic acid 8:2FTSA 8:2 fluorotelomer sulfonic acid
BA Bile acid
CA Cholic acid
CA-d4 Deuterated cholic acid
CD Celiac disease
CDCA Chenodeoxycholic acid
CDCA-d4 Deuterated chenodeoxycholic acid CE Cholesterol ester
Cer Ceramide
CYP7A1 Cholesterol-7α-hydroxylase DCA Deoxycholic acid
DCA-d4 Deuterated deoxycholic acid
DG Diglyceride
DHCA 3α,7α-dihydroxycholestanoic acid ECF Electrochemical fluorination
FA Fatty acid
FFA Free fatty acid
FOSA Perfluorooctanesulfonamide GCA Glycocholic acid
GCA-d4 Deuterated glycocholic acid GCDCA Glycochenodeoxycholic acid
GCDCA-d4 Deuterated glycochenodeoxycholic acid GDCA Glycodeoxycholic acid
GDHCA Glycodehydrocholic acid GHCA Glycohyocholic acid GHDCA Glycohyodeoxycholic acid
GL Glycerolipid
GLCA Glycolithocholic acid
GLCA-d4 Deuterated glycolithocholic acid GP Glycerophospholipid
GUDCA Glycoursodeoxycholic acid
GUDCA-d4 Deuterated glycoursodeoxycholic acid
HCA Hyocholic acid
HDCA Hyodeoxycholic acid
HPLC High-performance liquid chromatography ISTD Internal standard
2 LCA-d4 Deuterated lithocholic acid
LC-qTOF Liquid chromatography quadrupole time-of-flight LPC Lysophosphatidylcholine
LPE Lysophosphatidylethanolamine L-PFBS Linear-perfluorobutane sulfonic acid L-PFDoS Linear -perfluorododecane sulfonic acid L-PFDS Linear-perfluorodecane sulfonic acid L-PFHpS Linear-perfluoroheptane sulfonic acid L-PFHxS Linear-perfluorohexane sulfonic acid L-PFNS Linear-perfluorononanoic sulfonic acid L-PFOS Linear-perfluorooctane sulfonic acid L-PFPeS Linear-perfluoropentane sulfonic acid
M2-6:2FTSA 1H, 1H, 2H, 2H-perfluoro-1-[1,2-13C2] octane sulfonic acid
M2-8:2FTSA 1H, 1H, 2H, 2H-perfluoro-1-[1,2-13C2] decane sulfonic acid
M2PFDA Perfluoro-n-[1,2-13C
2] decanoic acid
M2PFDoA Perfluoro-n-[1,2-13C2] dodecanoic acid
M2PFHxA Perfluoro-n-[1,2-13C
2] hexanoic acid
M2PFHxDA Perfluoro-n-[1,2-13C2] hexadecanoic acid
M2PFTeDA Perfluoro-n-[1,2-13C2] tetradecanoic acid
M2PFUndA Perfluoro-n-[1,2-13C
2] undecanoic acid
M3PFBS Sodium perfluoro-1-[2,3,4-13C3] butane sulfonic acid
M3PFPeA Perfluoro-n-[1,2,3-13C
3] pentanoic acid
M4PFBA Perfluoro-n-[1,2,3,4-13C4] butanoic acid
M4PFHpA Perfluoro-n-[1,2,3,4-13C4] heptanoic acid
M4PFOA Perfluoro-n-[13C
8] octanoic acid
M4PFOS Sodium perfluoro-[13C8] octane sulfonic acid
M5PFNA Perfluoro-n-[13C
9] nonanoic acid
M8FOSA-M Perfluoro-1-[13C8] octanesulfonamide
NAFLD Non-alcoholic fatty-liver disease PC Phosphatidylcholine
PC ether Phosphatidylcholine ether PE Phosphatidylethanolamine PE ether Phosphatidylethanolamine ether PFAS Per- and polyfluoroalkyl substances PFBA Perfluorobutanoic acid
PFBS Perfluorobutane sulfonic acid PFDA Perfluorodecanoic acid PFDoDA Perfluorododecanoic acid PFDoDS Perfluorododecane sulfonic acid PFDS Perfluorodecane sulfonic acid PFHpA Perfluoroheptanoic acid PFHpS Perfluoroheptane sulfonic acid PFHxA Perfluorohexanoic acid PFHxDA Perfluorohexadecanoic acid PFHxS Perfluorohexane sulfonic acid PFNA Perfluorononanoic acid PFNS Perfluorononane sulfonic acid PFOA Perfluorooctanoic acid
3 PFOS Perfluorooctane sulfonic acid
PFPeA Perfluoropentanoic acid PFPeS Perfluoropentan sulfonic acid PFTDA Perfluorotetradecanoic acid PFTeDA Perfluorotetradecanoic acid PFTriDA Perfluorotridecanoic acid PFUnDA Perfluoroundecanoic acid PI Phosphatidylinositols POPs Persistent organic pollutants PS Phosphatidyleserine
RSD Relative standard deviation
SM Sphingomyelins
SP Sphingolipid
ST Sterol lipid
T1D Type 1 diabetes T2D Type 2 diabetes TCA Taurocholic acid
TCA-d4 Deuterated taurocholic acid TCDCA Taurochenodeoxycholic acid
TCDCA-d9 Deuterated taurochenodeoxycholic acid TDCA Taurodeoxycholic acid
TDHCA Taurohyodeoxycholic acid
TG Triglyceride
TG mufa Triglyceride Monounsaturated fatty acids TG pufa Triglyceride Polyunsaturated fatty acids TG sfa Triglyceride Saturated fatty acids THCA Taurodeoxycholic acid
THDCA Taurohyodeoxycholic acid TLCA Taurolithocholic acid
TLCA-d4 Deuterated taurolithocholic acid TUDCA Tauroursodeoxycholic acid
TUDCA-d4 Deuterated tauroursodeoxycholic acid TαMCA α-tauromuricholic acid
TβMCA β-tauromuricholic acid TωMCA ω-tauromuricholic acid UDCA Ursodeoxycholic acid
UDCA-d4 Deuterated ursodeoxycholic acid
UHPLC-qTOF/MS Ultra-performance liquid chromatography quadrupole time-of-flight mass spectrometry UNEP United Nations Environment Programme
αMCA α-muricholic acid βMCA β-murocholic acid ωMCA ω-murocholic acid
4
Abstract
The exposome is a complex study that includes both endogenous and exogenous markers related to different diseases as part of the research. Connecting both endogenous and exogenous markers can help explain many associations between environmental chemicals and metabolic profiles, which can be stressors for causing different diseases. This study investigates the impact of per- and polyfluorinated alkyl substances (PFAS) exposure during pregnancy on lipid profiles and measures the concentrations in blood and also in infants cord blood at birth. In addition, we are interested in lipids and bile acids (BA) levels in blood and if these three groups of analytes of interest are associated with each other. The infant blood-based samples come from a Chinese cohort and the maternal samples are from a Finnish cohort.
The samples were extracted for PFAS and BA and for lipid analysis. Statistical analysis was performed for both cohorts. The main interest was to investigate the interactions between the identified compounds. The results also include concentrations to show the variety of values measured.
Type 2 diabetes (T2D) and non-alcoholic fatty-liver disease (NAFLD) are usually not acquired early in life. Our results show a strong positive correlation between PFASs and triglycerides (TGs), as well as strong correlation between PFASs and conjugated primary and secondary BA particularly in cord blood samples, which can indicate higher risk of developing T2D or NAFLD. In maternal samples, the interactions between PFASs, BAs and lipids were much lower and showed less significant correlation. This indicates that the impact of PFAS exposure is much stronger during fetal development than in adult life.
Acknowledgement
The author of this study is thankful for support and help from Professor Tuulia Hyötyläinen and Dr. Leo Yeung. In addition, the author would like to thank Dr. Samira Salihovic and laboratory technician Daniel Duberg from Örebro University for guidance and help throughout the project. Professor Jia-Yin Dai and his students and staff from Chinese Academy of Sciences, the Institute of Zoology have given practical help and are being thanked for the support and kindness. The author is especially thankful for Jinghua Wang, from Chinese Academy of Sciences, for lots of practical help and for sharing a part of her research with this study. In addition, to FinnBrain cohort project for the cooperation. The author is grateful for Mobily grant provided by The Swedish Foundation for International Cooperation in Research and Higher Education (STINT) for the Joint China Sweden Mobility Programme 2018 (Dnr: CH2018-7805).
5
Table of content
List of abbreviations ... 1 Abstract ... 4 Acknowledgement ... 4 Appendix list ... 7 1. Introduction ... 8 1.1 The exposome ... 81.2 Per- and polyfluoroalkyl substances (PFASs) ... 8
1.3 Bile acids ... 9
1.4 Lipidomics ... 10
1.5 Impacts on health ... 10
The aim ... 11
2. Materials and Methods ... 12
2.1 Materials ... 12
Chemicals ... 12
China ... 12
Sweden ... 12
2.2 Methods... 13
2.2.1 One-phase extraction method MMC (MeOH/MTBE/CHCl3) ... 13
2.2.2 Matyash extraction method (MTBE/MeOH) ... 14
2.2.3 Bile acids and PFASs analysis ... 14
2.2.4 Lipid analysis ... 14
2.2.5 Quality control ... 14
3. Experimental part ... 15
3.1 Samples from China ... 15
3.1.1 Samples ... 15
3.1.2 Sample preparation for the Chinese samples ... 15
3.2 Samples from Finland ... 15
3.2.1 Samples ... 15
3.2.2 Sample preparation for Turku samples ... 16
3.3 Sample analysis ... 16
3.3.1 Lipid analysis ... 16
3.3.2 Bile acid and PFAS analyses ... 17
6 3.5 Statistical analyses ... 18 4. Results ... 18 4.1 Method development ... 18 4.2 Chinese samples ... 19 4.2.1 Lipid profiles ... 19 4.2.2 Bile acids ... 20 4.2.3 The exposome ... 21 4.2.4 Statistical analyses ... 23
4.3 Finnish samples – FinnBrain project ... 29
4.3.1 Lipid profiles ... 29 4.3.2 Bile acids ... 29 4.3.3 Exposome ... 30 4.3.4 Statistical analyses ... 31 5. Discussion ... 37 Conclusions ... 39 References ... 40 Appendix ... 45
7
Appendix list
1. Figure 20 – Normalisation results after log transformation and auto-scaling on Chinese blood samples in measured variables.
2. Figure 21 – Normalisation results after log transformation and auto-scaling on Chinese blood samples. Presented 50 samples.
3. Figure 22 – Normalisation results after log transformation and auto-scaling done on FinnBrain blood samples in measured variables.
4. Figure 23 - Normalisation results after log transformation and auto-scaling on FinnBrain blood samples. Presented 50 samples.
5. Table 8 – Correlation heat map between all measured and identified compounds in Chinese blood samples. Identified compounds are lipids, PFASs and BAs.
6. Table 9 – Correlation heat map between all measured and identified compounds in FinnBrain blood samples. Identified compounds are lipids, PFASs and BAs.
7. Table 10 - Lipid median concentrations with minimum and maximum values in blood samples from Wuhan and Beijing, China.
8. Table 11 – BA median concentrations with minimum and maximum values in blood samples from Wuhan and Beijing, China.
9. Table 12 – PFAS median concentrations with minimum and maximum values in blood samples from Wuhan and Beijing, China.
10. Table 13 – Non-target analysis results as median values in blood samples from Wuhan and Beijing, China, with minimum and maximum values.
11. Table 14 – Lipid median concentrations with minimum and maximum values for FinnBrain blood samples.
12. Table 15 – BA median concentrations with minimum and maximum values for all FinnBrain blood samples.
13. Table 16 – PFAS median concentrations for FinnBrain blood samples with minimum and maximum values.
14. Table 17 – Non-target analysis results as median values for all Finnbrain blood samples with minimum and maximum values.
8
1. Introduction
1.1 The exposome
Human health and the quality of living is an important research topic. Significant development in the healthcare has been contributed by new discoveries in research. A large part of the medical science focuses on preventing diseases with a broad interest of looking into different possible factors, which may cause or induce several diseases. Nowadays scientists do not only focus on endogenous or exogenous stressors separately but connect these two to be able to see correlations on a larger combined profile. One of the largely growing terms for this complex study is called the exposome. The exposome consists of both environmental exposure and metabolites as biomarkers for different diseases. The exogenous compounds that can trigger negative health effects can be all different environmental pollutants. These compounds come from industries or are released and leached from objects that are used in everyday life. For example, some substances are released to the environment as water soluble compounds from factories or are emitted to the air as volatile compounds. Other compounds are leached from various materials or released to the environment in other ways. All biochemical substances in the human organism are considered as endogenous compounds. They are usually produced by the organism, which include different metabolites or also food that is ingested and is broken down to simpler compounds (Miller, 2013; Sarigiannis, 2019).
1.2 Per- and polyfluoroalkyl substances (PFASs)
One specific group of environmentally relevant chemicals are the per- and polyfluoroalkyl substances (PFASs) that have been widely used for more than 60 years. Perfluoroalkyl substances are fully fluorinated surfactants with all hydrogens replaced with fluorine atoms. Polyfluoroalkyl substances are instead partially fluorinated, containing both hydrogen and fluorine atoms in the molecule. The hydrocarbon segment in polyfluoroalkyl substances is known to have a better solubility in commonly used solvents and due to that the partially fluorinated surfactants have an advantage over the fully fluorinated ones (Kissa, 2001). These substances do not occur naturally but are a diverse group of man-made chemicals. The productions of fluorinated surfactants first started in 1938 with the invention of Teflon. After that, in the 1950s, the production of PFAS chemical began to a larger scale. There are two methods mainly used to synthesize PFAS chemicals – telomerization and electrochemical fluorination (ECF). The telomerization yields only linear isomers, but ECF can produce up to 30% branched isomers in addition to linear isomers (Kissa, 2001).
PFAS chemicals have been widely used for many of their unique purposes in industries and also in households. They are extremely persistent and due to having hydro- and lipophobic properties, they have been greatly used as surfactants on many materials. These substances degrade very slowly and because of that have a greater change to bioaccumulate in the nature and in living organisms and have been found everywhere around the world. They can travel long distances due to their properties and as of early 2000s, PFASs have even been found in the eastern Arctic marine species, far from the main production areas (Kissa, 2001; Tomy et
9 The first detection of PFAS chemicals in human blood was in the 1970s from workers in a PFAS chemical production factory. Elevated levels of perfluorooctanic acid (PFOA) were found in their urine (Ubel, Sorenson and Roach, 1980). Another fluorinated substance, perfluorooctane sulfonic acid (PFOS), was found in the general populations of the U.S. in 1998. Pooled samples from blood banks were analysed and showed significant detections of PFOS (3M, 1999).
The toxicological testing has been mainly done on rats and mice to understand the absorption, distribution, metabolism, and excretion of fluorinated substances. When introduced to the organism, PFASs are well absorbed in the gastrointestinal tract. These substances are also distributed to the kidney, liver, and serum. Significant levels have also been found in lungs and heart, as well as in spleen, skin, and testis. Much lower levels from other mentioned organs have been found in the brain and in adipose tissue (DeWitt, 2015). PFASs were shown to also transfer from mother to the fetus through placenta and in addition, after birth, with the breastmilk (Croes et al., 2012).
1.3 Bile acids
Bile acids (BAs) are endogenous chemicals that are actively synthesized in the liver in parenchymal cells from cholesterol molecules and are stored in the gall bladder as in the bile. They are then released to the intestinal lumen, where they help with digestion of food (Lefebvre
et al., 2009). BAs are also metabolites that are part of many biochemical reactions. Not only
do they help with the digestion and absorption of lipids, cholesterol, and fat-soluble vitamins but they also regulate the homeostasis of lipid and glucose metabolism (Prawitt, Caron and Staels, 2011; Haeusler et al., 2013).
An average person has a pool of 2 grams of BAs in the organism that cycles around 12 times per day. The flux is regulated by the liver and the intestine by transporting ca 24 grams of BAs daily. Most of the BAs are moving in circulation and are being reabsorbed. About 5% of the circulated BAs move to duodenum and are then excreted with feces. The loss is rebalanced in the liver with new synthesized BAs (Lefebvre et al., 2009).
BAs are divided into two groups - primary and secondary bile acids. Primary BAs such as cholic acid (CA) and chenodeoxycholic acid (CDCA) are directly synthesized from cholesterol in the liver. Bile acids get conjugated in the liver with taurine and glycine. Secondary BAs are bio-transformed from unconjugated primary BAs by the gut microbiota, mainly to deoxycholic acid (DCA) and lithocholic acid (LCA). Conjugated and unconjugated BAs are both in circulation for liver uptake, where the secondary BAs are re-conjugated with taurine and glycine. These movements keep the BAs pool balanced and regulated (Di Ciaula et al., 2017). The key enzyme that catalyses the first steps of cholesterol synthesized to BAs is cholesterol-7α-hydroxylase (CYP7A1) (Prawitt, Caron and Staels, 2011). It has been speculated that CYP7A1 may be down-regulated by PFASs, which suppresses the synthesis of BAs if high levels of PFASs are introduced to or accumulated in the human organism (Beggs et al., 2016). This can then cause increased levels of reabsorbed BAs. Especially PFOS and PFHxS have shown to decrease the levels of BA excretion. Current knowledge also suggest that BAs and
10 PFAS are being reabsorbed in a similar manner and there is a possible observed association between the levels of BAs, PFAS and lipids (Salihović et al., 2019).
1.4 Lipidomics
Lipidomics is a discipline that is increasingly being used in health-related research for finding biomarkers to predict and detect different diseases. It studies lipids with analytical chemistry principles and presents the cellular lipid profiles on a wide scale. Lipids have an essential part in the human organism and have many important roles. For example in cellular functions, lipids are part of cellular membranes and they also act as vital energy depots as described in more detail in literature (Vance and Vance, 2008; Hyötyläinen and Orešič, 2015; Yang and Han, 2016).
Fatty acids are usually built up with a straight alkyl chain and a carboxyl end-group. The number of carbons in the fatty acid chain varies and double bonds in the chain determine if the lipid is saturated (containing no double bonds) or unsaturated (containing one or many double bonds) (National-Research-Council, 1989). Lipids are divided into smaller classes mainly based on their structures but also because of their functions. The main lipid classes found in human plasma are sterol lipids, sphingolipids, glycerolipids and glycerophospholipids (Risé et
al., 2007).
1.5 Impacts on health
Previous studies suggest that many serious health problems and diseases are related to high levels of lipids in the blood. Changes in lipid profiles can already be seen in infants and small children years before the onset of any disease. For example, altered lipid profiles have been detected in children later developing type 2 diabetes (T2D) (Alderete et al., 2019) or non-alcoholic fatty-liver disease (NAFLD), both of which are indicated by having abnormal concentrations of specific lipids in the blood (Orešič, Hyötyläinen, et al., 2013). Cord blood is often used as a sample matrix when analysing blood from new-borns. This shows new-borns blood profile and also reflects directly, which environmental pollutants have been transferred from mother to child through the umbilical cord.
Another group of diseases that has been linked with lipid abnormalities are autoimmune diseases such as type 1 diabetes (T1D) and celiac disease (CD). T1D is a serious autoimmune disease that has been increasing in the younger populations during the last decade. It is mostly known to be caused by destruction of insulin-secreting pancreatic beta-cells (Atkinson, Eisenbarth and Michels, 2014; Lamichhane et al., 2019). In addition, there is a strong suggestion that PFASs, especially PFOS and PFOA are tightly connected with the increased risk of developing T1D, particularly in individuals that have a genetic risk (Mcglinchey et al., 2019). Both PFOS and PFOA have been listed as persistent organic pollutants (POPs). These substances are regulated, PFOS from year 2009 and PFOA, its salts and related compounds from year 2019, by the Stockholm Convention (Templeton et al., 2019). PFAS exposure during pregnancy can increase the risk of developing T1D in the offspring (Mcglinchey et al., 2019). CD is, similarly to T1D, an autoimmune disorder and causes strong gluten intolerance (Caio et
al., 2019). CD diagnoses have recently increased in the developed countries and it is suggested
11 2017). The connection between PFASs and CD has been less studied than the interaction between T1D and PFASs, but it has been hypothesized that early life PFAS exposure might support the development of CD in long term, and that the PFAS exposure is linked with changes in lipid and bile acid metabolism (Sinisalu et al., 2020).
The aim
The aim of this study is to investigate the impact of prenatal PFAS exposure on concentrations of BAs and lipids using both infant blood-based samples (cord blood) as well as maternal samples during pregnancy. A comparison was done between two cities in China (Beijing and Wuhan, cord blood) and Finland (Turku, maternal samples).
Specific objectives:
1. Further investigate the interplay of PFASs and BAs and their association with lipid profiles;
2. Develop methods using smaller sample volumes and more efficient analysis, that can reduce the total analysis time and gives the ability to analyse many analytes of interest with one extraction step;
3. Study the concentrations of BAs and lipids in relation to prenatal exposure to PFASs in cord blood of infants collected from two cities in China;
4. Study the concentrations of BAs and lipids in relation to exposure to PFASs in blood samples of mothers during pregnancy.
12
2. Materials and Methods
Since parts of the experiments were carried out in China and Sweden, the chemicals used are given from respective sources.
2.1 Materials
Chemicals
China
Methanol HPLC Grade (MeOH) from Fisher chemical (Hampton, New Hampshire, the United
States of America). Chloroform A.R. (CHCl3) with 0.3-1.0% ethanol (C2H5OH) from Beijing
Chemical Works (Beijing, China). Reagent alcohol – methyl alcohol 90%, ethyl alcohol 4.5% and isopropyl alcohol 5% from Fisher chemical (Hampton, New Hampshire, the United States
of America). Hexane HPLC Grade (C6H14) from Fisher chemical (Hampton, New Hampshire,
the United States of America). Methylene chloride (CH2Cl2) from Fisher chemical (Hampton,
New Hampshire, the United States of America). Methyl tert-butyl ether HPLC Grade (C5H12O)
from Fisher chemical (Hampton, New Hampshire, the United States of America).
Sweden
Methanol LC-MS Grade 99.9% (MeOH) from Fisher chemical (Hampton, New Hampshire,
the United States of America), chloroform (CHCl3) for HPLC ≥99.9% from Sigma-Aldrich
(Saint Louis, Missouri, the United States of America). Contains amylenes as stabilizer. Ethanol 99.7% (C2H5OH) from Solveco AB (Rosenberg, Sweden). Sodium chloride reagent grade from
Scharlau (Barcelona, Spain). Acetonitrile LC/MS Grade 99.99% from Fisher chemical
(Hampton, New Hampshire, the United States of America). LC-MS Grade water (H2O) from
Fisher chemical (Hampton, New Hampshire, the United States of America). Tert-butyl methyl
ether (C5H12O) for HPLC Grade ≥99.8% from Sigma-Aldrich (Saint Louis, Missouri, the
United States of America). N-Hexane (CH3(CH2)4CH3) from Merch (Darmstadt, Germany).
Dichloromethane (CH2Cl2) from Honeywell (Charlotte, North-Carolina, the United States of
America). 2-Propanol (C3H8O) LC-MS ≥99.9% from Honeywell (Charlotte, North-Carolina,
the United States of America). Ammonium acetate (C2H7NO2) ≥99.0% solid form from
Sigma-Aldrich (Saint Louis, Missouri, the United States of America). Formic Acid (CH2O2) from
Fisher chemical (Hampton, New Hampshire, the United States of America). Quality control
(QC) of pooled human blood samples from donors from Örebro University Hospital, Sweden. PFAS internal and calibration standards M8FOSA-M, M2PFHxDA, M2PFTeDA, M2-6:2FTSA, M2-8:2FTSA, M3PFBS, M3PFPeA, M4PFHpA, M4PFBA, M2PFHxA, M4PFOA, M5PFNA, M2PFDA, M2PFUndA, M2PFDoA, 18O2-PFHxS, M4PFOS, PFBA, PFPeA,
PFHxA, PFHpA, PFOA, PFNA, PFDA, PFUnDA, PFDoDA, PFTrDA, PFTeDA, PFHxDA, PFODA, PFBS, PFHxS, PFOS, PFDS, FOSA, 4:2FTSA, 6:2FTSA, 8:2FTSA, L-PFHpS, L-PFDoS, L-PFNS and L-PFPeS were purchased from Wellington Laboratories (Guelph, Ontario, Canada). Internal standards were 13C-/18O2 labelled PFAS compounds.
BA calibration standards were purchased from different companies. 7H-dodecafluoroheptanoic acid was purchased from ABCR (Karlsruhe, Germany). CA, CDCA, DCA, DHCA, GCA,
13 GCDCA, LCA, TCA, TCDCA, TDCA, TDHCA, THCA, THDCA, TLCA, and TUDCA were purchased from Sigma-Aldrich (Saint Louis, Missouri, the United States of America), HDCA, HCA, αMCA, βMCA, ωMCA, 7-oxo-HDCA, 7-oxo-DCA, 12-oxo-LCA, TαMCA, TβMCA, TωMCA, GDHCA, GHCA, and GHDCA were purchased from Steraloids (Newport, RI,U.S.A), GLCA and GUDCA from Calbiochem (Gibbstown, NJ, the United States of America), and GDCA and UDGA from Fluka (Buchs, Switzerland). Internal standards CA-d4,
LCA-d4, UDCA-d4, CDCA-d4, DCA-d4, GCA-d4, GLCA-d4, GUDCA-d4 and GCDCA-d4
were purchased from Qmx laboratories Ltd. (Essex, UK). TCDCA-d9, TCA-d4, TUDCA-d4
and TLCA-d4 were purchased from Cambridge Isotope Laboratories, Inc. (Massachusetts, the
United States of America). All PFAS and bile acids standards were prepared in methanol and were stored at 4 °C.
Lipid internal standards CE (17:0) and TG (17:0/17:0/17:0) (Triheptadecanoin) were purchased from Larodan Research Grade Lipids, Inc. (Monroe, the United States of America). Ceramide (d18:1/17:0), LysoPC (17:0), PC (16:0d31-18:1), PC (17:0), PE (17:0) and SM (17:0) were purchased from Avanti Polar Lipids, Inc. (Alabama, the United States of America). Calibration standards 1,3-Dilinolein (DG(18:1/18:1)), Cholesteryl Linoleate (CE(18:1)), Cholesteryl Palmitate (CE(16:0)) and TG (18:0/18:0/18:0) were purchased from Larodan Research Grade Lipids, Inc. (Monroe, the United States of America). Cer (d18:0/18:1(9z)), Cer (d18:1/18:1(9z)), LysoPC (18:0), LysoPC (18:1), LysoPE (18:1), PC(16:0/16:0), PC (16:0-18:1), PC (18:0/18:0), PE (16:0-(16:0-18:1), SM (d18:1/18:1(9z)) and TG (16:0/16:0/16:0) were purchased from Avanti Polar Lipids, Inc. (Alabama, the United States of America).
2.2 Methods
We tested two methods for extraction and compared the results with two previously validated methods for lipids, BAs and PFASs. The two methods tested were the Matyash method (MTBE) and the one-phase extraction (MMC method) (Gil et al., 2018).
The chosen method used in China was one-phase MMC (MeOH/MTBE/CHCl3) extraction
method, which main focus is on using methanol/methyl tert-butyl ether/chloroform (MeOH/MTBE/CHCl3) 1.33:1:1 solution mixture for extraction. This method gave us the best recovery out of two methods tested and good quality results.
2.2.1 One-phase extraction method MMC (MeOH/MTBE/CHCl
3)
In total 40 µl of serum sample was added to Eppendorf tube and mixed with 270 µl of MeOH/MTBE/CHCl3 (1.33:1:1, v/v/v) solution mixture. A volume of 10 µl of BA and PFAS
internal standard (200 ng/mL PFASs and 1000 ng/mL bile acids in acetonitrile) and 12 µl of lipid internal standard (25 ppm lipids in chloroform and methanol 2:1 solution) were added to the tube and vortexed for 20 seconds. The Eppendorf tubes were then incubated on a shaker at 900 rpm for one hour at room temperature. The tubes were then centrifuged at 9600 relative centrifugal force (RCF) for 10 minutes at room temperature. In total, 250 µl of the supernatant was transferred to a glass vial and the extracted sample was evaporated to dryness. After that, 200 µl of CHCl3/MeOH (2:1) solution mixture was added and the sample was divided into two
parts. A volume of 50 µl was taken for lipid analysis and the remaining 150 µl was taken for BA and PFAS analysis. The sample was evaporated to dryness and 40 µl of methanol/water (40%/60%) was then added.
14
2.2.2 Matyash extraction method (MTBE/MeOH)
In total, 40 µl of serum sample was added to Eppendorf tube and mixed with 100 µl of methanol, 10 µl of BA and PFAS internal standard (200 ng/mL PFASs and 1000 ng/mL bile acids in acetonitrile) and 12 µl of lipid internal standard (25 ppm lipids in chloroform and methanol 2:1 solution). An additional 335 µl of MTBE was added to the tube and vortexed for 20 seconds. The tubes were then incubated at 900 rpm for one hour at room temperature. A total volume of 85 µl of LC-grade water was added, vortexed for 10 seconds and after that the tubes were incubated for additional 10 minutes. After incubation, the sample tubes were centrifuged at 9600 RCF for 10 minutes at room temperature. In total, 300 µl of the upper phase was collected to glass vial and evaporated with nitrogen to dryness. A volume of 200 µl of CHCl3/MeOH (2:1) was then added and the sample was divided to 150 µl for BA and PFAS
analysis and 50 µl for lipid analysis. The 150 µl of sample was evaporated to dryness again and after that 40 µl of methanol/water (40%/60%) solution mixture was added.
2.2.3 Bile acids and PFASs analysis
An aliquot of 40 µl of serum sample was transferred to an Eppendorf tube. A volume of 10 µl of BA+PFAS internal standard (200 ng/mL PFASs and 1000 ng/mL Bile acids in acetonitrile) was added to the tube with 50 µl of acetonitrile and vortexed for 5 seconds. The sample extraction was ultrasonicated for 3 minutes and then centrifuged at 10000 RCF for 5 minutes at room temperature. In total, 80 µl of the sample was then transferred to a glass vial and evaporated to dryness. The sample was resuspended with 40 µl of methanol/water (40%/60%) solution mixture.
2.2.4 Lipid analysis
A volume of 10 µl of serum sample, 10 µl of 0.9% NaCl solution and 120 µl of lipid internal standard (2.5 ppm lipids in chloroform and methanol 2:1 solution) were added to an Eppendorf tube and vortexed until milk-like colour appeared in the tube. The samples were then put into ice for 30 minutes. After that, the tubes with samples were centrifuged at 9400 RCF for 3 minutes. In total, 60 µl of CHCl3/MeOH (2:1) solution mixture was added to a new glass vial
and 60 µl of the lower layer of the sample extraction in the Eppendorf tube as added to the glass vial as well. Samples were then ready for analysis.
2.2.5 Quality control
All used lab equipment was cleaned with ethanol, hexane, and dichloromethane in series. Glassware was burned for 12 hours at 450°C to remove contaminants before use. QC-pooled blood samples from Örebro University medical faculty were used to assess the quality of the analysis. A volume of 10 µl of extracted samples were pooled together to use as quality control. Extraction blanks were included after every 20 samples for additional quality control. Internal standards were used for quality assurance and quantification. Calibration curves were used for quantitative measurements and quality control. One ppm of lipid standard was used as quality control throughout the instrumental analysis. Methanol blanks were used for the instrument between injections for detecting any internal contamination. Batch standard mixture solution was used to ensure the quality of sample preparation and instrumental status.
15
3. Experimental part
First step of the experimental part was to validate the extraction method used for Chinese samples in Beijing, China. The aim was to use an effective method that includes all analytes of interest in one extraction step with no additional steps needed.
3.1 Samples from China
3.1.1 Samples
The cord plasma samples were collected from two cities (Beijing and Wuhan) in China in 2018. We obtained n=100 samples from Wuhan and n=104 samples from Beijing, i.e. in total 204 samples from China. The Beijing samples were kept in -40°C in the blood bank facility until they were transferred to the Key Laboratory of Animal Ecology and Conservation Biology, Institute of Zoology, Chinese Academy of Science, where they were kept at -80°C. Samples from Wuhan were kept at -80°C in their blood bank reserve. After transportation to the institute, the samples were kept at -80°C.
3.1.2 Sample preparation for the Chinese samples
Samples were prepared according to the MeOH/MTBE/CHCl3 method, justification is given in
the results section on method development. First, 40 µl of the serum sample was transferred to an Eppendorf tube. Then, 270 µl of MeOH/MTBE/CHCl3 (1.33:1:1, v/v/v), 10 µl of BA and
PFAS internal standard mixture (200 ng/mL PFAS and 1000 ng/mL bile acids in acetonitrile) and 12 µl of lipid internal standard mixture (25 ppm lipids in chloroform and methanol 2:1 solution) were added to the serum sample. The sample tube was vortexed for 20 seconds and then incubated in the room temperature at 900 rpm on a shaker for one hour. After incubation, the samples were centrifuged at 9600 RCF at room temperature for ten minutes. In total, 250 µl of the, supernatant, was collected to a new Eppendorf tube and evaporated with nitrogen at room temperature to dryness.
The extracted Chinese serum samples were transported to Örebro University, Örebro, Sweden. All samples were prepared for lipidomics analysis by resuspending CHCl3/MeOH 2:1 solution
into the dried vials and transferring the liquid to LC vials. After running the samples on LC-qTOF/MS (Agilent Technologies, Santa Clara, CA, he United States of America), additional evaporation to dryness was done to prepare samples for bile acid analysis. Methanol/water 70%/30% solution was added. The resuspended samples were run on LC-qTOF/MS (Agilent Technologies, Santa Clara, CA, he United States of America).
3.2 Samples from Finland
3.2.1 Samples
The blood plasma samples were collected from Turku, Finland. The samples were taken from pregnant women from FinnBrain Birth cohort project. All pregnancies were as far as week 24. In total, we obtained 200 samples. The samples were kept at -80°C in Turku and transported in dry ice and kept at -80°C at Örebro University until further experiments.
16
3.2.2 Sample preparation for Turku samples
All samples were randomized before extraction.
Bile acids and PFAS extraction. A volume of 40 µl of serum sample was taken for BA and
PFAS analyses and added to Eppendorf tube. A volume of 10 µl of BA and PFAS internal standard (200 ng/mL PFASs and 1000 ng/mL bile acids in acetonitrile) was added to the tube with 80 µl of ACN and vortex for 5 seconds. The tube was ultrasonicated for 3 minutes and then centrifuged at 10000 RCF for 5 minutes. After centrifugation, 90 µl of the supernatant was added to a glass vial with an insert and 10 µl of the sample was collected to a separate glass vial for pooled samples. The 90 µl of sample was evaporated to dryness and resuspended with 60 µl of MeOH:H2O mixture solution (70%:30%).
Lipid extraction. The lipid extraction work was done on ice blocks to prevent lipid degradation.
A volume of 10 µl of serum sample was added to Eppendorf tube with 10 µl of 0.9% NaCl solution and 120 µl of lipid internal standard (2.5 ppm lipids in chloroform and methanol 2:1 solution). The tube was vortexed until milk-like colour and put on ice for 30 minutes. After that, the sample was centrifuged at 9400 RCF for three minutes. In total, 60 µl of CHCl3:MeOH
(2:1) was added to a glass vial with an insert. A volume of 60 µl of the lower layer of the sample in the Eppendorf tube was added to the same glass vial with an insert. In total, 50 µl of the leftover lower layer was added to a glass flask for pooled samples.
External calibration curves were made for both lipid and BA and PFAS analysis. The calibration curves were run twice, once in the beginning of the run and once in the end of the run. For lipids, the linearity was received with 7 points from 0.1-5 ppm from a mixture of 17 lipids in a calibration standard mixture. The calibration curve points were 0.1 ppm, 0.5 ppm, 1 ppm, 2 ppm, 3 ppm, 4 ppm and 5 ppm. For bile acids and PFASs, the linearity was received with 7 points from 0-60/320 (PFAS/BA) ng/ml. The starting point of 0 included only internal standard and no additional calibration standards. The curve points were made from bile acid and PFAS calibration mixtures. The points were, 0 ng/ml, 0.04 ng/ml, 0.1 ng/ml, 2 ng/ml, 4 ng/ml, 20 ng/ml and 60 ng/ml for PFASs and 0 ng/ml, 10 ng/ml, 20 ng/ml, 40 ng/ml, 80 ng/ml, 160 ng/ml and 320 ng/ml for BAs. The results from calibration curve run were showing similar peak intentions and an average point value was calculated with the two values from the results for every curve point, which were then used for further calculations.
3.3 Sample analysis
3.3.1 Lipid analysis
Samples were analysed on a UHPLC-qTOF/MS (Agilent Technologies, Santa Clara, CA, the United States of America) with separation using a Acquity UPLC®, BEH C18 (2.1x100 mm,
particle size 1.7 µm) (Waters corporation, Milford, MA, the United States of America) column kept at 50°C. The column was protected with C18 pre-column (Waters Corporation, Wexford, Ireland). Mobile phases used for the sample run were A: 10 mM NH4Ac and 0.1% formic acid
in water and B: 10 mM NH4Ac and 0.1% formic acid in acetonitrile/isopropanol (1:1). NH4Ac
is used as ionization agent. The samples were kept at 10°C during the whole run and 1 µl of the sample volume was injected. The flow rate was set on 0.4 ml/min and the gradient for the run started with 65%A and 35%B with a change after two minutes to 20%A and 80%B, which followed the change to 100%B from minute 7-14. The whole acquisition was 21 minutes with 14 min gradient program and additional 7 minutes of equilibration time including clean-up and
17 stabilization before start of next sample. Maximum pressure limit in the binary pump was set on 850 bar. Dual jet stream electrospray (dual ESI) ion source was used and the reference mass solution included masses between 121.050873 m/z and 922.009798 m/z. The ion polarity was on positive mode. The acquisition mass range was set between 100-1700 m/z and the instrument was in extended dynamic range. The capillary voltage and the nozzle voltage were kept at 3643 V and 1500 V. The N2 pressure was set on 21 psi, with the sheath gas flow as 11
L/min with the nebulizer temperature at 379°C. The data was acquired with MassHunter B.06.01 software (Agilent Technologies, Santa Clara, CA, the United States of America).
3.3.2 Bile acid and PFAS analyses
Bile acids and PFASs were analysed on a UHPLC-qTOF/MS (Agilent Technologies, Santa Clara, CA, the United States of America) with Acquity UPLC®, BEH C18 (2.1 x 100 mm,
particle size 1.7 µm) (Waters Corporation, Milford, MA, the United States of America) column at 50°C with a C18 pre-column for column protection (Waters Corporation, Wexford, Ireland). Mobile phases used for the sample analysis were A: 2mM NH4Ac in H2O:MeOH (70:30) and
B: 2mM NH4Ac in MeOH. NH4Ac was used as ionization agent. The samples were kept at
10°C during the whole sample acquisition and 1 µl of the sample volume was injected. The flow rate was set to 0.4 ml/min and the gradient started with 95%A and 5%B with a change after 1.5 minute to 70%A and 30%B, which followed a change after 4.5 minutes to 30%A and 70%B, the last change was after 7.5 minutes with 100%B until the end of run. Gradient program was 18 minutes long. 13 minutes for sample analysis and 5 minutes for clean-up, including equilibration in the end of the run. Maximum pressure limit in the binary pump was set on 850 bar. Dual jet stream electrospray (dual ESI) ion source was used and the ion polarity was on negative mode. The capillary voltage and the nozzle voltage were kept at 4500 V and 1500 V. The N2 pressure was set on 21 psi, with the sheath gas flow as 11 L/min and temperature at
379°C for the nebulizer. The data was acquired with MassHunter B.06.01 software (Agilent Technologies, Santa Clara, CA, the United States of America).
3.4 Data Pre-Processing
Lipids were processed separately from BA and PFAS data. Mass spectrometry data was pre-processed with MZmine 2.53 (Pluskal et al., 2010). Peak detection with a noise level of 1000 was performed first, following with ADAP chromatogram builder including group intensity threshold to be as noise level (1000), minimum highest intensity 10 and m/z tolerance 0.009 m/z or 8 ppm. Next, chromatogram deconvolution was performed with local minimum search as algorithm with a 70% chromatographic threshold, 0.05 minutes of minimum tR range, 5%
minimum relative height, 2250 minimum absolute height, 1 as minimum ratio of peak top/edge and peak duration range in minutes from 0.08 to 5.00. Isotopic peak grouper was done next with m/z tolerance of 0.05 m/z or 5 ppm, tR tolerance was set on 0.05 minutes and a maximum
charge of 2. For the alignment of peak lists, a Join alignment algorithm was applied with m/z tolerance as 0.006 or 10.0 ppm and a weight of m/z as 2 with a tR tolerance of 0.1 and a weight
of tR 1. Filtering with feature list rows filter was done next with 3 steps. First step rows that
match with all criteria were kept with a tR range from 2 to 12 and m/z of 369-1200. Second
filtering step removed rows that match with all criteria with a tR range of 2-4 and m/z of
800-1200. Third filtering step removed rows that match with criteria with a tR range of 4-8 and m/z
18 10.0 ppm, tR tolerance of 0.1 minutes and with intensity tolerance as 50%. Last, the
identification with a custom data base was done to identify the peak list with compound names.
3.5 Statistical analyses
Data processing was done in Microsoft Excel. Lipidomics data was processed separately from BA and PFAS data. Both datasets were filtered by removing compounds that were present at high concentrations in extraction blanks. Instrumental blanks were included but no internal contamination was detected. The ratio for that was calculated with median of all samples peak areas divided by median of extraction blanks peak areas. The samples were then sorted based on the ratio from lowest to highest, and for the compounds having the ratio less than 3 were removed. After sorting, the peak areas of the detectable compounds were then normalised with internal standards (ISTD). If ISTD was not available for a specific compound, then ISTD with closest retention time (tR) was used for normalisation. Calibration curves were calculated for
normalised values and used to calculate concentrations for identified compounds. Additional sorting was done for lipid samples. Relative standard deviation (RSD) was calculated for all QC samples and pooled samples. The RSD values were sorted from smallest to biggest and the compounds that had the RSD higher than 40% were removed from the list.
Statistical analyses were done by using MetaboAnalyst 4.0 (Chong, Wishart and Xia, 2019). The data matrix contained 196 samples and 45 measured variables in total in the Chinese dataset and 198 samples and 42 measured variables in total in the Finnish dataset. No additional sample normalisation was done in the system, but the data was normalised by Log transformation and the data went through auto-scaling. The results are presented in Figure 20- 23, Cytoscape 3.8.0 (Shannon et al., 2003) with Metscape 3.1.3 (Karnovsky et al., 2012) and Correlation Calculation 1.0.1 (Basu et al., 2017) were used for additional statistical analysis for building correlation networks with the Chinese samples both Beijing and Wuhan and with the Finnish samples. The results are presented in Figures 11 and 19.
4. Results
In this study, the focus was mainly on the evaluation of interactions between the identified compounds and here, the concentrations of individual compounds were less relevant. Quantitative measurements of individual compounds are still included and presented to show the variation of the concentrations in the study subjects.
4.1 Method development
The two methods, MeOH/MTBE/CHCl3 and MTBE/MeOH (explained in the 2.2 Methods),
which have been previously tested (Gil et al., 2018), were evaluated in this study,. The data was processed using raw values and corrected with a percentage of organic solvent volume used for the analysis. The results are presented as averages of overall peak area intensities and as well as the internal standards ISTD peak areas shown in Figure 1. Both methods were tested using the same amount of ISTD and the same sample volume. It is clear to see that ISTD recovery was much lower for the MTBE/MeOH method. Averages of overall peak areas are significantly higher using the MeOH/MTBE/CHCl3 method. Based on these results, the
19
Figure 1 – Comparison of the results from method development with two methods - MTBE/MeOH and MeOH/MTBE/CHCl3. The left graph on the figure presents the average values of ISTD peak area intensities
measured for both methods. The right graph on the figure presents the overall averages of peak area intensities, which show a clear result of the MeOH/MTBE/CHCl3 giving much better recovery. Based on this comparison,
this method was chosen for Chinese samples preparation.
A 3-day stability test was also conducted to evaluate any lipid degradation over time. QC-pooled samples were extracted according to the lipid extraction method (2.2.4 Lipid analysis) and evaporated to dryness. In total, 12 replicate samples were extracted for the test. Nine samples were kept in room temperature according to the plan, to test the stability in room temperature – Three for 1 day, three for 2 days and three for 3 days. Additional three samples were kept in the freezer from day one as comparison. After the test was over, all 12 samples were analysed by LC-qTOF and processed the data. No specific changes were detected. This test was needed to confirm stability of compounds of interest, if confronted with problems of sample transportation from China to Sweden.
4.2 Chinese samples
Initially, 204 samples were analysed. However, due to insufficient volume of sample for extraction, in total 8 samples were removed, and 196 samples were analysed.
4.2.1 Lipid profiles
A total number of 2880 of lipids were detected in the cord plasma samples, and 287 lipids, including four ISTDs, were identified with the help of a custom data base library. After sorting and cleaning the data (further explained in Chapter 4.2.4 Statistical analyses), 191 lipids were kept for further analysis. The lipid groups identified were free fatty acids (FFA), fatty acids (FA) from non-target approach analysis; while glycerolipids (GL), glycerophospholipids (GP), sterol lipids (ST) and sphingolipids (SP) from targeted lipid analysis. All measured lipid concentrations were median normalised and summed together as groups for further analysis. The lipids were grouped for statistical analysis based on their origin and named as Cer
20 (Ceramides), LPC (Lysophosphatidylcholines), LPE (Lysophosphatidylethanolamines), PC (Phosphatidylcholines), PC ether (Phosphatidylcholine ethers), PE (Phosphatidylethanolamines), PI (Phosphatidylinositols), PS (Phosphatidylserines), SM (Sphingomyelins), TG mufa (Triglyceride Monounsaturated fatty acids), TG pufa (Triglyceride Polyunsaturated fatty acids) and TG sfa (Triglyceride Saturated fatty acids). Median concentrations for lipid groups of Beijing and Wuhan samples were calculated separately. The error bars on the graphs present the maximum and minimum concentrations measured for both Beijing and Wuhan samples, representing possible biological variation of the lipids. The results are presented in Figure 2. Lipids are divided into different graphs based on concentration. Measured concentrations vary from 0.046 ng/mL to 958 ng/mL (Table 2).
Figure 2 A, B, C – Median values for all lipid groups with error bars presenting the maximum and minimum values measured. Beijing and Wuhan samples are presented in two different colours. Results are divided into three graphs due to measured values vary and different scales were needed.
4.2.2 Bile acids
A total number of 7639 compounds were detected, and 12 bile acids, including one novel BA 12-epiCA, were identified in cord plasma samples. The identified BAs were HCA (primary BA), CA, (primary BA), 7-oxo-DCA (secondary BA), TαMCA (primary BA, conjugated with taurine), GHCA (primary BA, conjugated with glycine), GCA (primary BA, conjugated with glycine), TCDCA (primary BA, conjugated with taurine), GCDCA (primary BA, conjugated with glycine), GLCA (secondary BA, conjugated with glycine) and 12-epiCA (3α, 7α, 12β-trihydroxy-5β-cholan-24-oic acid, novel secondary BA, epimerized). Median concentrations were calculated for all measured BAs and are presented in Figure 3. Error bars on the graphs represent measured maximum and minimum values. Values vary from 0.018 ng/mL to 2115 ng/mL (Table 11).
21
Figure 3 A,B – Bile acid median values with error bars presenting maximum and minimum values measured for both Beijing and Wuhan samples. The results are divided into different graphs due to measured values vary. Beijing and Wuhan samples are presented with different colours.
4.2.3 The exposome
The cord plasma samples were extracted and analysed to detect environmental toxins with both a targeted and a non-targeted approach. Target PFAS analysis in the same samples have been measured and reported in another study (Wang et al., 2020). The detected and identified PFASs were PFOA, PFNA, PFDA, PFUnDA, PFTriDA, PFHxS, PFHpS, PFOS, 6:2 Cl-PFESA and 8:2 Cl-PFESA. Calculated median values for detectable PFAS are presented in Figure 4. In addition to PFASs, two environmental pollutants butylparaben and benzylparaben; nine free fatty acids C16:1, C22:6, C18:2 and C20:5, linolenic acid, oleic acid, palmitic acid, stearic acid and arachidic acid were identified with the non-target analysis approach. The results are all presented in Figure 5. Error bars on the graph are presenting minimum and maximum concentrations measured for the Beijing and Wuhan samples.
22
Figure 4 A,B,C – PFAS median values with error bars presenting maximum and minimum values for samples from Beijing and Wuhan presented in different colours. Results are divided into different scales.
Figure 5 A,B,C,D,E – Non-target analysis results including free fatty acids and parabens. Error bars present minimum and maximum values. Beijing and Wuhan samples are presented with different colours and results are divided into different scales.
All median values from PFAS and identified analytes from non-target analysis with minimum and maximum values for both Beijing and Wuhan samples are presented in Tables 12 and 13.
23
4.2.4 Statistical analyses
4.2.4.1 T-test
The T-test is used for predicting differences in two groups. It analyses if the means of two groups are different from each other and can be statistically used for dividing samples into two separate groups. Significant level is a risk percentage given for the test to make type 1 error. This gives the probability that the test would not be able to detect the difference. Type 1 error is usually interpreted as false positive. The p-values are calculated probabilities that describe the ability to reject the null hypothesis, which means that there is a difference in the samples between two groups.
A t-test was used to assess significant differences in concentrations of detectable analytes between samples from Wuhan and Beijing with significance level set with p<0.05. The p values obtained from the t-test were transformed by -log10 and plotted in Figure 6. In total, 39 significant features out of 45 variables were detected. Seven variables with the lowest p-values are named and circled on the graph. In addition, five compounds of interest are named on the graph; two novel chlorinated polyfluorinated ether sulfonates and triglycerides (monounsaturated, polyunsaturated, and saturated). Significance line runs on y-axis at 1.25. Seven compounds that stayed under the significant line are, in order, HCA, 12-epiCA, oleic acid, palmitic acid, Cer, LPC and PE. The full list of significant features is presented in Table 1.
The novel PFAS compounds were found in much higher concentrations in Wuhan samples, which are presented as group number one (1). The difference between two groups (Beijing and Wuhan) based on 6:2 Cl-PFESA and 8:2 Cl-PFESA compounds are presented in Figure 7.
Figure 6 – Graphical presentation of the p-values obtained from the T-test after -log10 between blood samples from Wuhan and Beijing analysed for markers of early life exposure.
Compounds on the x-axis are in the order as they are set in the file used for the statistical analysis.
24
Figure 7 – Two groups (Beijing and Wuhan) from Chinese samples compared by two different compounds – 6:2 Cl-PFESA and 8:2 Cl-PFESA
25 4.2.4.2 Correlations
A correlation heat map between all measured variables in Chinese samples is presented in Table 8. The scale is set to from -0.8 as negative correlation to 0.8 as positive correlation, perfect correlation is presented as 1. The correlations between compounds separately are presented in the Table 2 for PFASs vs. BAs, Table 3 for PFASs vs. Lipid classes and Table 4 for BAs vs. Lipid classes.
In Table 2, significant correlations were observed between PFAS and BA compounds. The strongest positive correlations are PFOA, PFNA, PFDA, PFUnDA, PFHxS, PFOS and 6:2 Cl-PFESA from the PFASs list with GHCA, CA and GLCA from the BAs list. Relatively strong positive correlation is also observed between TαMCA and GCA. The novel BA 12-epiCA has negative correlation with most of the PFAS compound. PFTriDA, PFHpS and 8:2 Cl-PFESA have the lowest positive correlation with the compounds presented.
Table 2 – Measured values of PFASs compared with measured values of BAs in Chinese blood samples. Blue colour represents negative correlation and red colour represents positive correlation. White colour shows neutral correlation.
In Table 3, a correlation table between identified PFASs and lipid groups is shown. Very strong positive correlations are observed between most of the PFASs and all three TGs groups. PFTriDA and 8:2 Cl-PFESA and PFTriDA have the lowest positive correlation with TGs. Cer, LPC, PE, PI and PS are lipid groups that are in a negative correlation with all PFASs, with one exception, LPC, which has a low positive correlation with one PFAS, 8:2 Cl-PFESA. Lipid groups PC and PC ether have fairly low positive correlations with PFASs. SM and LPE have a significant positive correlation with all PFASs.
Table 3 - Measured values of PFASs compared with measured values of lipids in Chinese blood samples. Blue colour represents negative correlation and red colour represents positive correlation. White colour shows neutral correlation.
The correlations between all detected and identified BAs and lipid groups are presented in the Table 4. There are two BAs that stand out with a strong positive correlation with TGs – CA and GLCA. Three BAs, 7-oxo-DCA, HCA and 12-epiCA have low correlation with TGs. The rest of the bile acids have significant positive correlation with all three presented TG groups. Lipid groups Cer, PC and PC ether are fairly neutral compounds with few exceptions. LPC and
26 LPE have a low positive correlation with BAs. PE, PI, and PS have quite strong negative correlation with all BAs. Lipid group SM has a significant positive correlation with all BAs but not as strong as TGs.
Table 4 - Measured values of BAs compared with measured values of lipids in Chinese blood samples. Blue colour represents negative correlation and red colour represents positive correlation. White colour shows neutral correlation.
In Figure 8, the correlations between the top 24 compounds with mono- and polyunsaturated and saturated triglycerides are shown. These 24 compounds have either the strongest positive or negative correlations with chosen compound. Strong positive correlations are observed with most of the PFASs. Especially high correlation with PFOS, PFOA, PFNA and PFHxS. The BA that is strongly correlated with TGs is CA, which is one of the primary BAs that is synthesized first, directly in the liver.
Figure 8 – Triglycerides strongest positive or negative correlations with 24 compounds in the Chinese blood samples. Blue colour represents negative correlation and red colour represents positive correlation.
In Figure 9, the correlations between the top 24 compounds that show the strongest correlations, with four PFAS compounds, PFOA, PFNA, PFHxS and PFOS, are shown. All these PFASs are strongly correlated with each other but also with all TGs. In addition, we can see from the patterns, that three BAs (CA, GLCA and GHCA) have respectively strong correlation with these PFASs.
27
Figure 9 – Strongest positive or negative correlations of PFOA, PFNA, PFHxS and PFOS, with 24 compounds in the Chinese blood samples. Blue colour represents negative correlation and red colour represents positive correlation.
In Figure 10, the correlations between the top 24 compounds, that show the strongest correlations, and three BAs are shown. GLCA and CA have very strong correlation with previously presented PFASs – PFOA, PFNA, PFOS and PFHxS. These two BAs are also in a significant correlation with all three TG groups. GHCA is mainly correlating with different BAs, but is also in respectively strong correlation with PFOA, PFNA and PFOS.
28
Figure 10 – Strongest positive or negative correlations of GLCA, CA and GHCA, with 24 compounds in the Chinese blood samples. Blue colour represents negative correlation and red colour represents positive correlation.
Cytoscape 3.8.0 was used to build an additional correlation network showing the relationship between different compounds and the results are presented in Figure 11.
Figure 11 – Correlation network with all measured values for the whole sample list from China. On the network we can see compounds coloured based on their grouping – light blue as PFASs, light green as BAs, yellow as lipids and purple as compounds detected with a non-target approach. Dark blue lines represent negative correlation and if the line is bolder, the stronger is the negative correlation. Red lined represent positive correlation and similarly to blue lines, the bolder the line, the stronger is the positive correlation between compounds. The grey lines are presenting neutral correlation between all compounds. No correlation was detected if there are no lines between compounds.
29
4.3 Finnish samples – FinnBrain project
In total, two samples were removed during the data processing, due to technical difficulties during instrumental analysis.
4.3.1 Lipid profiles
In total, 1134 compounds were detected from plasma samples, and 278 lipids, including 8 ISTD, from the whole list were identified with a help of the same custom data base library used for the Chinese samples. The lipid groups identified were FFA, FA, GL, GP, ST, and SP. All lipid concentrations were median normalised and summed together as groups based on their origin for statistical analysis. The groups created for further analysis were Cer, LPC, PC, PC ether, PE, PE ether, PI, PS, SM, TG mufa, TG pufa and TG sfa. Median values were calculated for compounds and are presented with minimum and maximum values as error bars in Figure 12. Values are divided into three different scales due to values varying from 0.273 ng/mL to 1466 ng/ml. The median concentrations with minimum and maximum values for FinnBrain lipid analysis are presented in Table 14.
Figure 12 A,B,C – Lipid median values with error bars presenting maximum and minimum values measured for all compounds in Finnish samples. Results are divided into different scales.
4.3.2 Bile acids
BA and PFAS extraction and instrumental analysis was performed simultaneously. In total, 1142 compounds were detected from plasma samples and 20 BAs were identified. The identified BAs were CA (primary BA), DCA (primary BA), CDCA (primary BA), GCA (primary BA, conjugated with glycine), GDCA (primary BA, conjugated with glycine), GCDCA (primary BA, conjugated with glycine), GHDA (primary BA, conjugated with glycine), TαMCA (primary BA, conjugated with taurine), TCA (primary BA, conjugated with taurine), TDCA (primary BA, conjugated with taurine), TCDCA (primary BA, conjugated with taurine), HDCA (secondary BA), LCA (secondary BA), UDCA (secondary BA), GHDCA (secondary BA, conjugated with glycine), GLCA (secondary BA, conjugated with glycine), GUDCA (secondary BA, conjugated with glycine), THDCA (secondary BA, conjugated with
30 taurine), TLCA (secondary BA, conjugated with taurine), TUDCA (secondary, conjugated with taurine). Median concentrations were calculated for all identified BAs and are presented in Figure 13 with error bars as minimum and maximum values. The compounds have been spread out on three different scales due to concentrations varying from 0-189 ng/mL.
Figure 13 A,B,C,D – BA median values with error bars presenting minimum and maximum values measured for all BAs in Finnish sample list. Results are divided into different scales.
4.3.3 Exposome
In total, five PFASs were identified from the 1142 compounds detected. In addition to five PFASs, one paraben and one amino acid and free fatty acids were identified in the non-target analysis to find environmental toxins and other substances. The five identified PFASs were PFHxS, PFNA, PFOA, PFOS and PFPeA. The other identified substances were propylparaben, tyrosine, stearic acid, palmitic acid, myristic acid and linolenic acid. All median concentrations with minimum and maximum values for both PFASs and identified analytes from non-target analysis results are presented in the Figure 14 and Figure 15. With one exception, PFPeA was excluded, due to high blank contamination and issues with ISTD.
31
Figure 14 A,B – PFAS median values with error bars as minimum and maximum values for all compounds in Finnish samples. Results are divided into different scales.
Figure 15 A,B,C – Median values for non-target analysis with error bars as minimum and maximum values for all compounds. Results are divided into different scales.
4.3.4 Statistical analyses
4.3.4.1 Correlation
The correlation heat map between all measured variables in Finnish samples are presented in Table 9. The scale is set similarly to that of the Chinese samples, from -0.8 as negative correlation to 0.8 as positive correlation, where 1 shows the perfect correlation. The
32 correlations between compound groups separately are presented in the Table 5 for PFASs vs BAs, Table 6 for PFASs vs. Lipid classes and Table 7 for BAs vs. Lipid classes.
In Table 5, significant correlations are observed between half of the PFASs and BAs. CA, DCA, GLCA, HDCA, LCA and UDCA have quite neutral correlation with PFASs. PFNA has the strongest positive correlation with BAs, especially with taurine conjugated BAs. GDCA, GUDCA, TαMCA, TCA, TDCA, TDCA, THDCA, TLCA and TUDCA have all the strongest positive correlation with all PFASs, but the correlations are much lower when comparing the Finnish results with the Chinese results.
Table 5 - Measured values of PFASs compared with measured values of BAs in Finnish blood samples. Blue colour represents negative correlation and red colour represents positive correlation. White colour shows neutral correlation.
In Table 6, correlations are observed between four PFASs and twelve lipid groups. Cer, LPS, PE ether and SM have the strongest positive correlation with all PFASs presented in this table. PC, PC ether, and TG pufa have significant correlation with PFASs. The most neutral correlation is between PE, TG mufa, TG sfa and all PFASs.
Table 6 - Measured values of PFASs compared with measured values of lipids in Finnish blood samples. Blue colour represents negative correlation and red colour represents positive correlation. White colour shows neutral correlation.
In Table 7, correlations are observed between all identified BAs and the lipid groups. Most of the compounds do not have strong correlation with each other, with some exceptions. Two lipid groups, LPC and TG sfa, have the strongest positive correlation with most of the BAs. The strongest with CA, DCA, GDCA, GHDCA, GUDCA and TDCA. Another significant correlation is between GHDCA and lipid group PS. Cer, PC, PC ether, PE ether, PI, and TG pufa have mostly negative correlation with all BAs. PE, SM and TG mufa have little correlation with most of the BAs.
33
Table 7 - Measured values of BAs compared with measured values of lipids in Finnish blood samples. Blue colour represents negative correlation and red colour represents positive correlation. White colour shows neutral correlation.
In Figure 16, correlations between top 24 compounds that show the strongest correlations with four BAs are presented. The highest positive correlation is between these BAs and all the rest of BAs. As shown above, PFNA is one of the PFASs that is showing significant positive correlation with all four BAs. Other two PFASs, PFOA and PFOS have a bit lower correlation, but still significant with these BAs. One lipid group, TG sfa, has a low positive correlation with presented compounds and rest of the lipids have either negative correlation or do not show any strong correlations and thus, are not shown here. There is one compound, TDCA, that has a strong positive correlation with most of the compounds, although in the Table 7 we can see that TDCA is mostly in negative correlation with lipid groups.
33
Figure 16 – Strongest positive or negative correlations of TCA, TCDCA, TDCA and TUDCA, with 24 compounds in the Finnish blood samples. Blue colour represents negative correlation and red colour represents positive correlation.
Correlation patterns of top 24 compounds with PFASs are presented in the Figure 17. All four compounds have strong positive correlation with each other and much lower positive correlation with all the other compounds. All four PFASs have very similar correlation patterns with a few exceptions. Lipid groups SM, Cer, LPC and PE ether, have the strongest positive correlations with PFASs from all the lipid groups, as also seen in Table 6. In addition, all four compounds have significant negative correlation with one detected environmental chemical – propylparaben.
34
Figure 17 – Strongest positive or negative correlations of PFNA, PFOA, PFOS and PFHxS, with 24 compounds in the Finnish blood samples. Blue colour represents negative correlation and red colour represents positive correlation.
Correlation patterns of top 24 compounds with lipid groups Cer, LPC, PE ether, SM and TG sfa are presented in Figure 18. These lipid groups are chosen based on having significant numbers on the correlation heat maps and are presented separately to show more specific correlations. Cer, LPC, PE ether and SM have strong positive correlations with other lipid groups but also with all PFAS substances, as also seen in Table 6. One group, TG sfa, is in very neutral, slightly negative correlation with PFASs, but instead is in quite strong positive correlation with almost all BAs. In addition to TG sfa, LPC has a significant positive correlation with most of the BAs.
35
Figure 18 – Strongest positive or negative correlations of Cer, LPC, PE ether, SM and TG sfa, with 24 compounds in the Finnish blood samples. Blue colour represents negative correlation and red colour represents positive correlation.
36 Cytoscape 3.8.0 was used to build an additional correlation network showing the relationship between different compounds and the results are presented in Figure 19.
Figure 19 - Correlation network with all measured values for the whole sample list from Finland. Compounds are coloured in the same order on the correlation network as with China samples on the Figure 11 – light blue as PFASs, light green as BAs, yellow as lipids and purple as compounds detected with non-target approach. Dark blue lines are presenting negative correlation and if the line is bolder, the stronger is the negative correlation. Red lined are showing positive correlation and similarly to blue lines, the bolder the line, the stronger is the positive correlation between compounds. Light grey lines present neutral correlation between all compounds. No correlation was detected if there are no lines between compounds.
