The development and validation of a fast and robust dried blood spot based lipid profiling method to study infant metabolism
- 2.4k Downloads
Early life exposures and metabolic programming are associated with later disease risk. In particular lipid metabolism is thought to play a key role in the development of the metabolic syndrome and insulin resistance in later life. Investigative studies of metabolic programming are limited by the ethics and practicalities of sample collection in small infants. Dried blood spots on filter paper, derived from heel pricks are considered as the most suitable option for this age group. We validated a novel lipid profiling method, based on high resolution mass spectrometry to successfully determine the lipid composition of infants using dried blood spots. The spotting and air drying of blood on paper has noticeable effects on many of the lipids, leading to lipid oxidation and hydrolysis, which demand careful interpretation of the obtained data. We compared the lipid profiles from plasma or whole blood samples and the results from dried blood spots to determine if these revealed the same inter-subject differences. The results from dried blood spots were no less reproducible than other lipid profiling methods which required comparatively larger sample volumes. Therefore, lipid profiles obtained from dried blood spots can be successfully used to monitor infancy lipid metabolism and we show significant differences in the lipid metabolism of infants at age 3 versus 12 months.
KeywordsLipidomics Dried blood spots DIMS FTMS Infant lipid metabolism
Lipidomics is commonly applied to study the aetiology of a range of metabolic diseases including the metabolic syndrome and type II diabetes. Multiple studies have shown that specific phosphatidylcholines and sphingolipids levels, as well as the triglyceride profile are associated with relative risk for type II diabetes (Floegel et al. 2012; Rhee et al. 2011). In addition, there is great interest in processes such as metabolic imprinting and programming, and how this influences the predisposition to insulin resistance in later life (Hattersley and Tooke 1999). This is based on associations between early life exposures, including low birth weight and rapid infancy growth, and increased risk for type II diabetes (Harder et al. 2007) and cardiovascular disease in adulthood (Barker et al. 1990) suggesting that modifying the early postnatal environment may have significant impact on later life disease risk.
Most mechanistic studies related to programming have been carried out using rodent models. Thus far it has been difficult to study lipid metabolism in infants and relate this to later health outcomes due to small cohort design and limited blood sampling in this age group. Many current methods used for detailed lipid or fatty acid profiling require considerable volumes of plasma, serum or blood, well surpassing the 30–50 μl per spot that can be collected from a heel prick. Although more blood can be collected from neonates (Escolano-Margarit et al. 2013; Kuipers et al. 2011; Sabel et al. 2009), the heel prick is generally considered as the least invasive, and therefore preferred by both parents and researchers. Dried blood spots have also become a standard biological medium especially in pharmacokinetic studies (Abu-Rabie and Spooner 2009) and in newborn screening, where mass spectrometry MS/MS methods have been developed for detection of inborn errors of metabolism (Chace et al. 1993; Pandor et al. 2004).
We aimed to develop a lipidomics approach that would allow us to rapidly measure the lipid metabolism of infants using dried blood spots. Initially we were skeptical about the application of dried blood spots as a sample format for the study of lipid metabolism. In particular we wanted to determine the stability of blood lipids on paper and to ascertain to what extent lipid oxidation and hydrolysis affect lipid concentrations generally and whether this was specific to particular classes of lipids. The final aim was to use the method on a sub-set of the Cambridge Baby Growth Study (CBGS) to determine the lipid profile of healthy, breast fed infants at 3 and 12 months of age, providing a proof of concept for our approach. This is the first report that uses dried blood spots for lipidomic profiling.
2 Materials and methods
2.1 Sample collection
2.1.1 Initial method development
For initial method development commercially available whole horse blood was used (Fisher Scientific UK Ltd., Loughborough UK) as well as anonymized adult human venous blood. From these whole EDTA blood samples, 50 μl aliquots were then immediately spotted onto standard untreated filter paper cards (Ahlstrom 226, ID Biological Systems; as used in national UK newborn bloodspot screening) and air dried for up to 4 h. From the same blood samples, 2 ml aliquots were additionally frozen as whole blood and 2 ml aliquots were processed to plasma, allowing us to determine if lipid profiles from these different sample types were comparable. All samples were then stored at or below −20 °C.
2.1.2 QC material
Dried blood spot quality control (QC) material was used in all experiments. QC1 spots were made from mixed anonymized adult human venous blood; QC2 spots from commercially obtained horse blood.
2.1.3 Infant dried blood spot samples
Infant samples were collected as part of CBGS (Ong et al. 2009), using heel prick blood sampling and dried blood spots were then stored at −20 °C. The CBGS was approved by the Cambridge local research ethics committee, UK and all mothers gave informed written consent. Dried blood spots were obtained from healthy infants at age 3 months (n = 40) and 12 months (n = 36), all exclusively breast-fed until at least 3 months of age, average growth characteristics are in electronic supplementary material, Table S1. These samples were collected over a 6 year period with the oldest samples collected in March 2004. The samples were balanced for sex and birth weight (details in supplementary data).
For all samples a round disc with diameter of 3.2 mm was punched out and used for analysis. When the text refers to dried blood spot samples, it means 3.2 mm diameter discs punched out from larger blood spots.
All solvents of LC–MS grade or better were ordered from Sigma Aldrich (Gillingham, UK). All internal standards were obtained from Avanti Polar Lipids (through Instruchem, Delfzijn, NL) with the exception of undecanoic acid and trilaurin (Sigma Aldrich, Gillingham, UK).
2.3 Sample extraction
After initial method development we standardized the sample extraction to the following protocol. An automated method for the extraction of lipids was developed using an Anachem Flexus automated liquid handler (Anachem, Milton Keynes, UK). A total of 80 dried blood spot samples, 4 blanks, 8 QC1 and 4 QC2, all as 3.2 mm diameter discs in 1.2 ml Cryovials, were placed in the wells of a glass coated 2.4 ml deep well plate (Plate+™, Esslab, Hadleigh, UK). Firstly, 100 μl of MilliQ H2O was added to each of the wells. Then 250 μl of MeOH was added containing six internal standards (0.6 μM 1,2-di-o-octadecyl-sn-glycero-3-phosphocHoline, 1.2 μM 1,2-di-O-phytanyl-sn-glycero-3-phosphoethanolamine, 0.6 μM C8-ceramide, 0.6 μM N-heptadecanoyl-d-erythro-sphingosylphosporylcholine, 6.2 μM undecanoic acid, 0.6 μM trilaurin), followed by 500 μl of methyl tert-butyl ether (MTBE). The plates were then sealed (using Corning aluminium microplate sealing tape Sigma Aldrich Company, UK) and shaken for 10 min at 600 rpm, after which the plate was transferred to a centrifuge and spun for 10 min at 6,000 rpm. Each well in the resulting plate had two layers, with an aqueous layer at the bottom and an organic layer on top.
A 96 head micro-dispenser (Hydra Matrix, Thermo Fisher Ltd., Hemel Hampstead, UK) was used to transfer 200 μl of the organic layer to a glass coated 240 μl low well plate (Plate+™, Esslab, Hadleigh, UK). The plate was transferred to a Genevac EZ-2 evaporator (Genevac Ltd., Ipswich, UK) and dried down. The samples were reconstituted in 25 μl of MTBE and with 90 μl of MS-mix [7.5 mM NH4Ac IPA:MeOH (2:1)] using a Hydra Matrix, after which the plate was sealed and stored at −20 °C until analysis.
2.4 Mass spectrometry
2.4.1 Direct infusion high resolution MS (DIHRMS)
All samples were infused into an Exactive Orbitrap (Thermo, Hemel Hampstead, UK), using a Triversa Nanomate (Advion, Ithaca US). The Nanomate infusion mandrel was used to pierce the seal of each well before analysis, after which, with a fresh tip, 5 μl of sample was aspirated, followed by an air gap (1.5 μl). The tip was pressed against a fresh nozzle and the sample was dispensed using 0.2 psi nitrogen pressure. Ionization was achieved with a 1.2 kV voltage. The Exactive started acquiring data 20 s after sample aspiration began. After 72 s of acquisition in positive mode the Nanomate and the Exactive switched over to negative mode, decreasing the voltage to −1.5 kV. The spray was maintained for another 66 s, after which the analysis was stopped and the tip discarded, before the analysis of the next sample began. Throughout the analysis the sample plate was kept at 15 °C. Samples were run in row order.
A selected subset of samples was analysed using a LTQ-Velos Orbitrap Elite coupled to a Surveyor (both Thermo, Hemel Hampstead, UK) LC system for lipid identification. Ionization was achieved using ion max electrospray source (voltage 4.0 Kv, capillary temperature 275 °C, sheath gas 20 arbitrary (arb) units, aux gas 5 arb, sweep gas 10 arb). For each sample, 10 μl was injected onto a Hypersil Gold C18 (50 × 2.1 mm, 1.9 μm particle size) column and lipids were eluted with a gradient of MeOH/H2O and MeOH/IPA (Bird et al. 2011). The LTQ-Orbitrap Velos was controlled either manually or set-up using the data-dependent precursor selection. Generally the selected masses were isolated with a 1.5 m/z width in the linear ion trap and then fragmented using either linear ion trap with 35 % relative collision energy or in the higher-energy collision-induced dissociation collision cell, with a range of collision energies from 5 to 75 % relative collision energy. All spectra were recorded in the Orbitrap set at 100,000 resolution.
2.5 Data and statistics
The raw data were converted to .mzXML (MS convert), parsed to R and 50 spectra averaged per mode using XCMS. Further processing was conducted using Peakpicker v2.0, an in-house R script for peak picking, deisotoping and annotating DIMS data sets. Univariate analysis was performed in Prism 6.01 (GraphPad Software, Inc.). All significance thresholds were set using Bonferroni multi-comparison correction. Multivariate analysis was performed in SIMCA (v13.0) (Umetrics). When replicate samples are described in the manuscript it means that separate aliquots of the same sample, or separate 3.2 mm disc of the same spot, were analysed independently.
2.6 Lipid identification
All lipids mentioned in the text were preliminary identified based on accurate mass and subsequently based on accurate mass MS2 spectra and retention time from the LC–MS experiment. The identifications, with ChEBI IDs where possible are summarized in electronic supplementary material, Table S2.
3 Results and discussion
We have adapted a lipidomic method (Graessler et al. 2009), based on work with human plasma, to assess the suitability of dried blood spots for lipid profiling, and determined the extent of lipid degradation and oxidation during this sampling process. The use of glass-coated well plates was essential to minimize the introduction of contaminants, although non-glass pipette tips still were a possible source of contamination, we did not find that the levels impaired the reproducibility of the lipids of interest (see example spectra in Fig. S1 of electronic supplementary material). We aimed to determine if the dried blood spot samples yielded data of sufficient signal and robustness such that it was comparable to that from plasma. Our novel method only required one 3.2 mm diameter disc, containing 2.5–5 μl of blood, punched from a dried blood spot obtained by heel prick capillary blood sampling. This approach substantially reduced the sample volume required compared to alternative open profiling methods (Dénes et al. 2012). Signal intensities corresponding to specific lipids (see Table S1 of electronic supplementary material) were extracted and summed. The intensity of each lipid species was expressed relative to the summed intensities of selected signals, in the same way as fatty acid profiling by GC is often expressed (Wang et al. 2013). This relative quantitation is therefore independent of the blood volume in the spot and hematocrit level. For the method development and validation, oxidized lipids were also included to the list of target lipids. The purpose of the internal standards added was primarily to check mass accuracy and post run calibration. Our method does not provide absolute concentrations but only relative levels as commonly used in other lipidomics approaches (Graessler et al. 2009; Rhee et al. 2011).
The samples were extracted and processed using an automated liquid handler. A plate containing 80 analytical samples and 16 QC and blank samples could be processed in under 2 h. Lipid extracts were directly infused into a HRMS using chip based nanospray. The analysis time was ca. 2.5 min per sample, which is over 10 times faster than previously reported methods for the profiling of lipids (or fatty acids) from infant blood samples (Escolano-Margarit et al. 2013; Nikkila et al. 2008) and on par with other high-throughput methods for lipid profiling and suitable for the profiling of large sample sets of large epidemiological cohorts.
Dried blood spots are a convenient sample format to store over long time periods and different cohorts have done so. A fast and robust lipid profiling method would therefore provide the opportunity to study lipid metabolism in large epidemiological cohorts. However, some lipids are known to be prone to oxidation and hydrolysis, and it has never been documented how storage time affects the recovery of lipids from dried blood spot samples. It was therefore important to validate the method and determine lipid profile variation due to blood spotting, drying and prolonged storage.
Importantly, the precision of the dried blood spot based lipid profiling was not different to that based on whole blood samples for the known identified lipid species (Fig. 1b), such as triglycerides, sphingomyelins and phosphatidylcholines, encompassing many lipids associated with risk for developing type II diabetes (Floegel et al. 2012; Rhee et al. 2011). Generally the coefficient of variance (CV) was below 20 %, except for those lipids with concentrations around the limit of detection, which was around a 1,000-fold lower level than the most dominant lipids (Fig. 1b). The lipid profiles of dried blood spots revealed the same pattern of inter-subject differences as plasma or whole blood samples (Figs. 1c, S3 of electronic supplementary material) with comparable or better precision (Fig. 1b). This showed that any degradation processes did not introduce extra variability in the major lipid species. The average CV for the lipids measured in the dried blood spot samples was 8.7 % while it was 10.6 for the plasma samples, which was a significant difference (p < 0.05). The reason that dried blood spots yielded more precise measurements could be that for each sample, oxidative and other degrading processes are stopped once the spots are fully dried, while in plasma samples the enzymatic and chemical processes can occur once the sample is defrosted, and therefore more difficult to control.
We also wanted to simulate longer term storage, which we did using an accelerated aging experiment; dried blood spots were stored at 40 °C for up to a week. This did not lead to any further significant changes, suggesting that the major changes observed take place during initial blood spotting and drying, with no further significant effect on the lipid levels with prolonged storage (see Fig. 1a). Since most lipid profiling studies based on plasma or serum use venepuncture (Floegel et al. 2012; Rhee et al. 2011), we determined if there was a difference in whole blood collected by venepuncture or collected from the capillary bed (heel prick or finger prick). We found that there were no significant relative changes in the lipid profile between dried blood spots of venous versus blood of the arterial bed (see Fig. S4 in electronic supplementary material).
When we were confident that the dried blood spots were able to provide a representative and robust lipid profile, we applied our new method to study the lipid profiles of healthy infants using a subset of samples from the CBGS (Ong et al. 2009). Of note, we had no difficulty in obtaining lipid profiles from the samples and the sample age did not show a significant correlation with the total lipid extracted or with any individual lipid classes, in accordance with our validation results.
The major differences between the two age groups were the relative amount of palmitate versus stearate in the phosphatidylcholines fraction. A previous study showed a similar difference when comparing the fatty acid profile of plasma phospholipids between mothers and preterm infants (Sabel et al. 2009). The differences we found were dramatic over the first year of life. PC(34:2), mainly containing palmitate at sn-1 position, was increased in the 12 month old infants and PC(36:2), mainly containing stearate at sn-1 position, was increased in the 3 month old infants. The ratio between the two phosphatidylcholines was an almost perfect predictor for the two age groups (p = 6.9e−22), and at 12 months this was already much closer to that observed in adults (see inset of Fig. 2). It was not possible to build a robust model based on sex, although there was a trend in the data towards increased triglycerides in boys, particularly at 12 months of age, though not significant.
The other striking difference was the level of sphingomyelins (see Fig. 3c) which was 25–50 % higher in the 3 month old, compared to adults and still 10 % higher in the samples of 12 month old [only SM(41:1) was significantly higher in the 12 month old infants]. Sphingomyelins are the most abundant phospholipids in human breast milk (Jensen et al. 1978) and the higher concentration in the 3 month old infants might be due to the metabolism of the breast milk fats (Nilsson and Duan 2006). Previous studies suggested that lipid metabolism is rapidly changing within childhood (Nikkila et al. 2008) but to our knowledge, it has not been previously documented that these differences are significant within the first year of life.
4 Concluding remarks
We believe that lipid profiles of dried blood spots are an important new tool to study the metabolism of infants, and have successfully shown novel methodology to do this. The developed method is fast and requires only a 3.2 mm diameter disc of a dried blood spot. It is therefore compatible with simultaneous biochemical or genetics analyses of the same sample. Although the spotting and drying of the blood has a distinct effect on the lipid profile it is still possible to obtain information on global lipid metabolism changes, comparable to lipid profiles obtained from plasma collected through venepuncture. This now gives us the unique opportunity to explore infancy lipidomics and relationships with later anthropometric and metabolic outcomes in large epidemiological study cohorts using samples already gathered, for example during the newborn bloodspot test. We hypothesize this will lead to better understanding into the early biological mechanisms of insulin resistance and related metabolic diseases.
This study received financial support from the Medical Research Council (UD99999906 and Cambridge Lipidomics Biomarker Research Initiative G0800783). The Cambridge Baby Growth Study was supported by the European Union, the World Cancer Research Foundation International, the Medical Research Council (including a Centenary Award), the Mothercare Charitable Foundation and the NIHR Cambridge Comprehensive Biomedical Research Centre. We thank all the families who participated in the study, the paediatric research nurses, the rest of the study team, the staff at the Addenbrooke’s Wellcome Trust Clinical Research Facility and the midwives at the Rosie Maternity Hospital, Cambridge, UK.
- Balazs, Z., Panzenboeck, U., Hammer, A., Sovic, A., Quehenberger, O., Malle, E., et al. (2004). Uptake and transport of high-density lipoprotein (HDL) and HDL-associated alpha-tocopherol by an in vitro blood–brain barrier model. Journal of Neurochemistry, 89, 939–950. doi: 10.1111/j.1471-4159.2004.02373.x.CrossRefPubMedGoogle Scholar
- Bird, S. S., Marur, V. R., Sniatynski, M. J., Greenberg, H. K., & Kristal, B. S. (2011). Serum lipidomics profiling using LC–MS and high-energy collisional dissociation fragmentation: Focus on triglyceride detection and characterization. Analytical Chemistry, 83, 6648–6657. doi: 10.1021/ac201195d.CrossRefPubMedPubMedCentralGoogle Scholar
- Dénes, J., Szabo, E., Robinette, S. L., Szatmari, I., Szonyi, L., Kreuder, J. G., et al. (2012). Metabonomics of newborn screening dried blood spot samples: A novel approach in the screening and diagnostics of inborn errors of metabolism. Analytical Chemistry, 84, 10113–10120. doi: 10.1021/ac302527m.CrossRefPubMedGoogle Scholar
- Escolano-Margarit, M. V., Campoy, C., Ramirez-Tortosa, M. C., Demmelmair, H., Miranda, M. T., Gil, A., et al. (2013). Effects of fish oil supplementation on the fatty acid profile in erythrocyte membrane and plasma phospholipids of pregnant women and their offspring: A randomised controlled trial. The British Journal of Nutrition, 109, 1647–1656. doi: 10.1017/S0007114512003716.CrossRefPubMedGoogle Scholar
- Hutchins, P. M., Moore, E. E., & Murphy, R. C. (2011). Electrospray tandem mass spectrometry reveals extensive and non-specific oxidation of cholesterol esters in human peripheral vascular lesions. Journal of Lipid Research, 52, 2070–2083. doi: 10.1194/jlr.M019174.CrossRefPubMedPubMedCentralGoogle Scholar
- Kuipers, R. S., Luxwolda, M. F., Sango, W. S., Kwesigabo, G., Velzing-Aarts, F. V., Dijck-Brouwer, D. A., et al. (2011). Postpartum changes in maternal and infant erythrocyte fatty acids are likely to be driven by restoring insulin sensitivity and DHA status. Medical Hypotheses, 76, 794–801. doi: 10.1016/j.mehy.2011.02.020.CrossRefPubMedGoogle Scholar
- Ong, K. K., Langkamp, M., Ranke, M. B., Whitehead, K., Hughes, I. A., Acerini, C. L., et al. (2009). Insulin-like growth factor I concentrations in infancy predict differential gains in body length and adiposity: The Cambridge Baby Growth Study. The American Journal of Clinical Nutrition, 90, 156–161. doi: 10.3945/ajcn.2008.27408.CrossRefPubMedGoogle Scholar
- Rhee, E., Cheng, S., Larson, M., Walford, G., Lewis, G., McCabe, E., et al. (2011). Lipid profiling identifies a triacylglycerol signature of insulin resistance and improves diabetes prediction in humans. The Journal of Clinical Investigation, 121, 1402–1411. doi: 10.1172/jci44442.CrossRefPubMedPubMedCentralGoogle Scholar
- Sabel, K. G., Lundqvist-Persson, C., Bona, E., Petzold, M., & Strandvik, B. (2009). Fatty acid patterns early after premature birth, simultaneously analysed in mothers’ food, breast milk and serum phospholipids of mothers and infants. Lipids in Health and Disease, 8, 20. doi: 10.1186/1476-511X-8-20.CrossRefPubMedPubMedCentralGoogle Scholar
- Wang, L. Y., Summerhill, K., Rodriguez-Canas, C., Mather, I., Patel, P., Eiden, M., et al. (2013). Development and validation of a robust automated analysis of plasma phospholipid fatty acids for metabolic phenotyping of large epidemiological studies. Genome Medicine, 5, 39. doi: 10.1186/gm443.CrossRefPubMedPubMedCentralGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.