A robust method for performing 1H-NMR lipidomics in embryonic zebrafish

The embryonic zebrafish is an ideal system for lipid analyses with relevance to many areas of bioscience research, including biomarkers and therapeutics. Research in this area has been hampered by difficulties in extracting, identifying and quantifying lipids. We employed 1H-NMR at 700MHz to profile lipids in developing zebrafish embryos. The optimal method for lipidomics in embryonic zebrafish incorporated rapid lipid extraction using chloroform and an environment without oxygen depletion. Pools of 10 embryos gave the most acceptable signal-to-noise ratio, and the inclusion of chorions in the sample had no significant effect on lipid abundances. Embryos, bisected into cranial (head and yolk sac) and caudal (tail) regions, were compared by principal component analysis and analysis of variance. The lipid spectra (including lipid annotation) are available in the public repository MetaboLights (MTBLS2396).


23
Lipids are a diverse and abundant class of biomolecules within all living organisms [1]. Simple fats 24 and oils provide energy and have important roles in processes such as signalling and membrane 25 trafficking, whereas more complex, polar lipids such as sterols and phospholipids provide much 26 needed structures to biological membranes [2]. The lipid bilayer is a key component of skeletal 27 muscle membranes and here the lipids provide both structure and flexibility [3].

28
Disrupted lipid metabolism is a primary pathogenetic mechanism in a number of mammalian diseases 29 including myopathies such as lipid storage myopathy and HACD1-associated centronuclear myopathy 30 [4][5][6]. Despite the importance of lipids, the study of lipidomics is less advanced than that of genomics 31 or proteomics. Gene and protein structures are linear alignments of code (base pairs, amino acids)

32
whereas the structure of lipids is far more complex, making even the classification of these structure 33 more difficult [1].

34
There are a vast number of techniques available to study lipid profiles, all with their own benefits and 35 limitations. 1 H-NMR is an important tool for lipidomic analyses, it is time efficient, highly reproducible and has been shown to identify a number of lipid classes in samples, including mouse 37 liver [7]. NMR produces crowded spectra due to the lack of separation and so if separation is required 38 mass spectrometry (MS) as a class of technique may be more appropriate [8]. MS is highly sensitive, 39 however, the level of ionisation of lipids is variable and so lacks the reproducibility compared to 40 NMR [8]. MS is destructive to samples and therefore complementary analyses cannot be performed 41 on the same material, for these reasons' NMR can be a desirable procedure for initial explorations [8].

49
Robust methods for extraction and collection of 1 H-NMR spectra are essential. Therefore, in this 50 work, we established a working protocol that generates high-quality, reproducible spectra and 51 confirmed our ability to detect lipidomic differences in developing embryos. In this study we aimed to 52 carry out 1 H-NMR lipid profiling in developing zebrafish embryos to establish a reproducible 53 methodology for future studies, focusing on the protruding mouth stage (72 hours post fertilisation 54 (hpf)).

57
Adult AB strain wild-type (WT) zebrafish were housed in a multi-rack aquarium system at the 58 University of Liverpool. Zebrafish were maintained at 28.5 ± 0.5°C on a 12-hour light, dark cycle.

59
Husbandry and collection of embryos involving zebrafish embryos were performed in accordance 60 with both local guidelines (AWERB ref no. AWC0061) and within the UK Home Office Animals Scientific Procedures Act (1986), in a Home Office approved facility (University of Liverpool 62 Establishment License X70548BEB). Adults were bred and embryos incubated in the dark at 28°C, in 63 aquarium water and staged according to Kimmel et al., 1995 [20]. All experiments performed were 64 exempt from ethical approval due to embryonic age as stipulated in ASPA guidelines.

75
Whole embryos (all dpfs) were stored in pools of 10 embryos except where different embryo numbers 76 required for the sample size experiments. Bisected embryos were stored in groups of 20 cranial or 77 caudal portions per tube to counteract sample amount being decreased ( Figure 1).

78
Lipid extraction 79 500µl of ice-cold chloroform (C 1 HC l3 , Sigma Aldrich) was added to samples in microfuge tubes.

80
Samples were then sonicated at 50KHz using a microprobe in three x 30 second intervals including a 81 30 second pause in between intervals. Samples were vortexed for one minute before incubating for 10 82 minutes at 4°C prior to centrifugation for 10 minutes at 21,500g and 4°C. The supernatant was 83 transferred to a fresh microcentrifuge tube or glass via and snap frozen in liquid nitrogen before 84 lyophilisation. Chloroform is a highly volatile chemical and is known to react with plastic, for this 85 reason lipids were also extracted in glass mass spectrometry vials [21]. After plastic was deemed to be most suitable, lipids were extracted in both normoxic and hypoxic conditions under low oxygen levels parameters constant between samples) was used to acquire with a 25ppm spectral width and 256 scans 100 (thirty-minute experiment) at 15°C to offset the volatility of the chloroform. Pre-processing of spectra 101 proceeded using automated standard vendor routines (apk0.noe -Fourier transform, phasing and 102 window function) and spectra were aligned using the residual C 1 HC l3 peak at 7.26ppm.

Spectra inclusion criteria
104 Strict quality control (QC) was conducted on all spectra. Phasing of baselines was checked and 105 manually corrected when required. The reference peak (C 1 HC l3 ) line width was measured at half 106 height to ensure it was a single peak of <1.1Hz. Any samples that failed QC were acquired again on 107 the spectrometer up to three times.

Spectra processing
Using a representative set of spectra, a pattern file defining bin boundaries was generated for all peak 110 positions. Where possible peaks were identified using and in-house Avanti Library or annotated

127
Sample collection was based on Beckonert et al., (2007) andFolch (1957) with the extraction 128 procedure simplified to focus solely on lipids and reduce sample losses (omitted two-phase extraction)

143
(TG) and total-cholesterol (TC) were selected due to their relevance in zebrafish development [9].

144
These bins show increasing signal in the spectra with increased embryo numbers within the sample.

145
We calculated signal-to-noise ratios to establish whether bins exceeded both the limits of detection 146 (LOD) and quantification (LOQ) ( Table 1). Table 1 shows that certain bins did not always meet the 147 LOQ or LOD (bins 118, 137 and 151) so these would not be appropriate bins to use for lipid 148 quantification therefore other bins representing the same lipid classes should be selected, for example 149 bin 150 was selected to quantify triacylglycerides over that of bin 151 (Table 1).

151
Bins that failed the LOD (limit of detection) threshold of 3:1 are highlighted in yellow and those failing the LOQ (limit of quantification) threshold of 10:1 in between the developmental stages, with all 10 of these differing between 1 and 3dpf ( Figure 3C).

180
The marginal separation of clusters seen between developmental stages ( Figure 3C) led us to assess 181 how much influence the yolk sac may be having in the spectra. Spectra acquired on dissected cranial

206
ANOVAs for these 3 bins found them to be significantly changed as a whole (p<0.05) with relative 207 abundance of these bins lower in the caudal sections. Tukey's multiple comparisons found TC to be 208 significantly lower in caudal sections than in either the cranial sections or whole embryos (p<0.05) 209 and all three groups caudal, cranial and whole significantly different for PC and TG (p<0.05).

211
Lipidomics in embryonic zebrafish is an emerging field and has potential for numerous applications 212 with far-reaching impact. This work provides a robust methodology for 1 H-NMR lipidomics using 213 embryonic zebrafish from which reproducible, informative spectra can be produced.

214
An increase in embryo number increased signal-to-noise ratio and allowed for the acquisition of a 215 quantitative lipid spectra (Figure 2). Pools of 10 embryos provided a spectra in which all bins tested 216 passed both the limit of detection and limit of quantification ( Figure 2 typically not experimentally feasible due to time constraints from commonly used techniques such as 220 microinjection. Therefore, our finding that it is possible to use pools of 10 embryos for 1 H-NMR 221 lipidomic analysis is important for studies of genetically modified embryos.

222
Dechorionation was deemed to be an unnecessary step because there was no significant difference in 223 the lipidome between samples with and without chorions ( Figure 3A). This was an expected result as 224 chorions are proteinaceous structures and not lipid based [30]. We were also able to conclude that 225 retaining the chorions benefits spectral consistency -a probable reason for the increased variability is 226 the possibility of an induced metabolic change in the sample. Embryos hatch from their chorions 227 between 2 -3 days post fertilisation (dpf) therefore when analysing different developmental changes,

228
we recommend collection of embryos without including this labour-intensive step [20].
We found the most robust method to extract lipids was rapid extraction in plastic Eppendorf tubes 230 without the depletion of oxygen by use of nitrogen gas (N 2 ). The use of increased variation seen in the 231 PCA plots ( Figure 3B) which led to a number of significant changes within the 1 H-NMR spectra 232 ( Figure 3D). We propose that the additional air-flow induced through N 2 rich working may have been 233 a factor in increased variance in the lipid metabolome and outweighs the benefit of oxygen depletion 234 on the lipidome in this study.

235
There was less cluster separation between embryos of different developmental stages than expected 236 ( Figure 3C) and we hypothesised this may be due to the yolk sac masking any effects seen from 237 within the rest of the embryo due to its high lipid composition [15]. To test this, we bisected the