Andrea
Bonicelli
a,
George
Taylor
b and
Noemi
Procopio
*a
aSchool of Law and Policing, Research Centre for Field Archaeology and Forensic Taphonomy, University of Central Lancashire, Preston, UK. E-mail: nprocopio@uclan.ac.uk
bBiological Mass Spectrometry (BioMS) Core Facility, Faculty of Biology, Medicine and Health, The University of Manchester, Manchester, M13 9PT, UK
First published on 22nd July 2024
Liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS) untargeted metabolomics has become the gold standard for the profiling of low-molecular-weight compounds. Recently, this discipline has raised great interest in forensic sciences, especially in the field of toxicology and for post-mortem interval estimation. The current study aims at evaluating three extraction protocols and two LC-MS/MS assays run in both positive and negative modes, to identify the most suitable method to conduct post-mortem metabolomic profiling of bone tissue. A fragment of the anterior tibia of a 82 years-old male sampled from a human taphonomy facility was powdered via freeze-milling. The powdered sub-samples were extracted in five replicates per protocol. Methods tested were (I) a biphasic chloroform–methanol–water protocol, (II) a single phase methanol–water protocol, and (III) a single phase methanol–acetonitrile–water protocol. LC-MS/MS analyses were carried out via high performance liquid chromatography, either on hydrophilic interaction (HILIC) or on reversed-phase (C18) columns in both positive and negative ionisation modes, coupled with a Q-TOF mass spectrometer. Results suggest that the highest consistency between replicates and quality control samples was obtained with the single phase extractions (i.e., methanol–acetonitrile–water), whilst the ideal combination of instrumental set up HILIC chromatography in positive ionisation mode and of C18 chromatography in negative ionisation mode. For the purpose of forensic investigations, a combination of a single phase extraction and the two aforementioned chromatographic and mass spectrometry modes could represent an ideal set up for obtaining bone metabolomic profiles from taphonomically altered bones.
Sample-preparation strategy plays a crucial role in planning an efficient and successful experiment.4–6 An efficient sample preparation protocol is characterised by four main attributes: (I) lack of selectivity, (II) simplicity, (III) reproducibility, and (IV) consideration of any chemical or enzymatic reactions that could affect the compound's stability after their extraction.3 The metabolomic workflow includes sample collection and extraction, experimental analysis, data pre-treatment, and statistical analysis.7 Proper storage of samples is essential to reduce post-collection instability of the metabolomic profiles. Quenching, which limits or removes chemical or enzymatic interactions after sampling, is a critical step, and the choice of quenching strategy depends on the matrix being analysed.7 The extraction process involves homogenization to increase the surface area exposed to the solvent and the selection of an appropriate solvent based on factors such as toxicity, solubilisation, selectivity, dissolution rate, chemical reactivity, and pH.3,7
High-throughput metabolomics approaches have gained popularity in bone research for understanding bone physiology and the connection between metabolite expression and tissue biomechanical properties.8 Elucidating bone metabolic pathways is valuable for investigating disease development and creating diagnostic and prognostic tools.9 For example, Zhao et al.10 studied the bone lipidome and metabolome in relation to bone mass changes in ovariectomised mice, and identified metabolic pathways associated with bone loss.
Recently, metabolic profiling of bone material has gained interest in the fields of archaeology and forensic science.11–14 Archaeological dental calculus analysis revealed changes in compounds and degradation products over time.11 Metabolomics applied to skeletal remains also aids to capture the significant correlations between the abundance and presence of specific compounds and the time elapsed since death (post-mortem interval, PMI) of the person.12 It is important to consider that the process of post-mortem decomposition, in fact, alters the metabolomic profile of the tissues ante mortem, due to the leaching of biomolecules in the environment surrounding the body, the biomolecular degradation of larger molecules into smaller metabolites due to taphonomic processes, and the introduction of new metabolites resulting from the decomposition process led by microbial decomposers. This results in profiles significantly different from the ante mortem ones, or from those extracted from preserved specimens from fresh cadavers, therefore requiring ad hoc protocols for their analysis and interpretation.
Another limitation encountered by forensic and archaeological experts dealing with bones samples from curated skeletal collections (e.g., from forensic osteological collections at human taphonomy facilities, HTFs, or from museums) is the processing that bones undergo prior to their long term storage. 'Bone maceration', the way in which such bone processing is defined, consists in the submersion of skeletal elements in high temperature water baths with the addition of chemical agents to degrease the bone surface, and was shown to significantly impact the bone metabolomic profiles by reducing the compounds coverage and by introducing contaminating features.13,14
This study aims to investigate three different extraction protocols, selected amongst existing ones for murine bone samples,10 plasma15 and bacteria,16 on a taphonomically altered and macerated bone sample collected from an HTF: an adapted biphasic extraction method (Chlor_Meth),15 and two single-phase extraction methods using methanol and water (Meth_Water)10 and methanol, acetonitrile, and water (Meth_ACN)16 extraction solvents. Two LC-MS/MS assays (HILIC and reversed-phase (C18) chromatography run in both positive and negative ionization modes) are investigated to determine the most appropriate protocol for undemineralised bone metabolomics of challenging forensic samples.
Mass spectrometry data was acquired in a data-dependent manner. Features were selected for fragmentation automatically on a basis of the top 10 most intense ions with a charge state of 1–2 and minimum threshold of 10 cps. Isotopes within 4 Da were excluded from the scan. The accumulation time for each scan was 100 ms and the accumulation time for the TOF survey scan was 250 ms. Total cycle time was 1.3 s. Collision energy was determined using the formula CE (V) = 0.084 × m/z +12 up to a maximum of 55 V. Isotopes within 4 Da were excluded from the scan. Acquired data were checked in PeakView 2.2 and imported into Progenesis QI 2.4 for metabolomics, where they were aligned, peaks were picked, normalised to all compounds and deconvoluted according to standard Progenesis QI workflows. Peak picking parameters were set to automatic with default sensitivity level and a minimum peak width of 0.1 min. Ions were ignored before 1.3 min and after 24 min for HILIC runs and before 0.9 min and after 10 min for C18 runs. Adducts and specific of all assays are given in ESI† Table S1. MSI level 2 annotations were made by searching the accurate mass, MS/MS spectrum and isotope distribution ratios of acquired data against the NIST MS/MS metabolite library, and additionally by searching retention times and accurate masses against an in-house made library of chemical standards using Progenesis QI (using a 0.5 min retention time tolerance, Table S2, ESI†). Metabolites with a score higher than 40 were accepted. MSI level 1 identifications were made when both libraries were in agreement and MS/MS spectra matching the NIST library entries were present. Table S3 (ESI†) includes the list of putatively annotated compounds and their annotation scores.
It is appropriate to introduce certain limitations of the current study. First, QC samples were injected only before and after the 15 samples injected, in every assay. Ideally, QC samples should be analysed at regular intervals in the run to better control for instrumental drifts. Another limitation is that samples from each class were analysed consecutively and not in a randomised order, which could have introduced a bias. Despite the above mentioned limitations, we believe that the differences observed among extraction protocols are genuine and related with the procedure chosen and the solvents employed.
Fig. 1A–D reports the RSD for both extraction replicates and QCs across the four experiments (RSD for single compounds are available in ESI† Table S4). In HILIC ESI+ the RSD values for the total peak area amongst the three extractions showed acceptable median values for Meth_ACN (5.93%) and QCs (9%). Higher values were found for Meth_Water and Chlor_Meth, respectively 14.6% and 18.9%. For HILIC ESI− the highest RSD remained for Chlor_Meth (57.6%), followed by Meth_Water (30.0%) and Meth_ACN (21.7%). The lowest RSD was calculated for QCs (18.9%). All these values for HILIC ESI− are considered to be unacceptable in terms of replicates' agreement. C18 ESI+ gave the highest RSD with Chlor_Meth (18.4%), followed by QCs (11.9%), Meth_Water (8.36%), and Meth_ACN (5.28%). C18 ESI− showed similar results to HILIC ESI+ with the highest RSD being the one for Chlor_Meth (10.1%), followed by Meth_Water (8.36%). Lower RSD were those of QCs (6.66%) and Meth_ACN (6.04%). Fig. S1 (ESI†) shows the distribution of single metabolite's RSD for each extraction protocol and for the QCs, confirming the better suitability of Meth_ACN in combination with HILIC chromatography and the overall appropriate replicability of both single phase extractions with C18. QCs RSD distributions confirmed the repeatability of the all the assays. The plots in Fig. 1E–H provide an overview of the instrumental stability of the each assay across the entire run. It is clear that the two HILIC experiments presented the same trend with a drift towards the end of the experiment. This drift was effectively corrected via data normalisation (see also Fig. S2, ESI†) and transformation as showed in Fig. 1I–L. In contrast, higher stability in the system was shown for C18 ESI+ and ESI− as noticeable both in extraction replicates and in QCs. These results suggest that overall Meth_ACN is the most repeatable extraction protocol, and that HILIC ESI− and C18 ESI+ has the lowest instrumental stability across the QCs. All the remaining assays were characterised by an acceptable RSD range.
Considering the number of annotated compounds profiled with the three different protocols post data processing performed separately, for HILIC ESI+ 74 compounds were retained when extracted with Meth_ACN, 77 with Chlor_Met and 71 with Meth_Water, with 63 of those being shared between the three extraction protocols as shown in ESI† Fig. S3. A much lower number of compounds was obtained using HILIC ESI−, where a maximum of 11 compounds were found with Chlor_Meth and Meth_ACN extractions, and 8 with Meth_Water. Coverage for C18 ESI+ was deeper, with a maximum of 97 compounds obtained using Meth_ACN, 95 using Chlor_Meth, and 87 using Meth_Water. This assay allowed the profiling of the highest number of compounds. C18 ESI− allowed the profiling of 71, 70, and 64 compounds respectively for Meth_ACN, Chlor_Met, and Meth_ACN. After data processing, none of the assays were able to profile more than 97 putatively annotated compounds. The limited number of compounds identified in this study stems from the unique composition of the bone tissue and on the nature of the bone under analysis (decomposed and macerated). Bones are constituted approximately by 60–70% of mineral matrix, and the remaining organic fraction is made up by approximately 90% type I collagen, 5% non-collagenous proteins (NCPs), 2% lipids and metabolites by weight.20 Additionally, bones subjected to taphonomic processes (such as harsh climate conditions) and maceration are further depleted of small molecules and organic components, therefore explaining the reduced number of compounds identified in this analysis.13
The three extraction methods tested were then processed together in order to evaluate the overall number of compounds identified. Prior to data processing, 110 compounds were putatively annotated in HILIC ESI+, 18 in HILIC ESI−, 228 in C18 ESI+, and 106 in C18 ESI−. After data processing, results showed the highest number for C18 ESI− (n = 87), followed by HILIC ESI+ (n = 66), C18 ESI− (n = 64), and HILIC ESI− (n = 9). Fig. 2 shows the compounds shared by the four experimental modes. HILIC ESI+ allowed the profiling of 54 unique compounds, 11 shared with C18 ESI+ and one only shared with C18 ESI−. HILIC ESI− had only 6 unique compounds, one shared with C18 ESI+ and two shared with C18 ESI−. C18 ESI+ had 73 unique compounds with two compounds being exclusively shared with C18 ESI−. Finally, C18 ESI− had 58 compounds profiled only with this assay. No putatively annotated compounds were shared across all four assays. This suggests that no clear advantage in the identification and relative quantification of polar compounds can be obtained by the removal of the lipids from the polar phase.21
PCA for HILIC ESI+ (Fig. 3) showed extremely good instrumental stability, with QCs forming a close cluster and principal component (PC) one explaining 66.9% of the sample variance responsible for the separation between single and biphasic extractions. Furthermore, it can be seen in Fig. 3 that, with the exception of one outlier for the Meth_Water extraction (B4), the two clusters for the single phase extractions almost overlap, suggesting great similarity in the two profiles. HILIC ESI− results are notably different, with PC1, accounting for 68.2% of the total variance, failing to explain the difference between the different types of extractions. This components seems to be heavily influenced by the presence of two outliers for Chlor_Meth and Meth_Water. In contrast, 18.0% of the variance explained by PC2 seems to capture the variation between single phase and biphasic extraction protocols. C18 experiments in both ESI+ and ESI− showed very similar trends. In ESI+, 88.8% of the variance described the separation between Chlor_Meth and the two single phase extractions, while in ESI− the variance accounted for 92.7%. There was one outliers for Chlor_Meth. Due to the acceptable RSD values for the QC samples, outliers for all assays may be attributed to issues related to the extraction phase rather than with instrumental instability. Overall, the profiles obtained by the two single phase extractions across all experiments seem to have minimal differences. The higher volume for the extraction solvent for the single phase protocols (1:
15 compared to 1
:
8 for the biphasic protocol) could have aided the operator into more accurate pipetting.22 Furthermore, the biphasic protocol is considerably slower than then single phase one, although it might be ideal in cases of low amounts of starting material. Furthermore, the interaction between the two solvents allows the selective removal of less polar lipid compounds and optimises deproteinisation.22 However, the present study shows that when applied to bone powder material and it might increase technical variance.22 Furthermore, due to the the complexity of the protocol and the use of bead homogenisation, the biphasic extraction is not advisable for large scale studies as suggested also by the low degree of agreement obtained between the biological replicates. Finally, several classes are consistently found across multiple assays (such as amino acids and fatty acyls), some others are found in two out of three protocols (alkaloids and bases for HILIC ESI+ and C18 ESI+, carboxylic acids and vitamins for C18 ESI+ and C18 ESI−), and some classes are unique for specific protocols (glycerophospholipids and phenylpropanoids for HILIC ESI+, sphingolipids for C18 ESI+ and monosaccharides and sterol lipids for C18 ESI−). For the HILIC ESI+ runs, we detected significant differences between the biphasic protocol and the two monophasic ones for glycerophospholipids, where the monophasic extractions gave higher abundances than the biphasic one, and for phenylpropanoids, where all three protocols were significantly different with the higher intensity being the one obtained using the biphasic method. In C18 ESI+ runs, alkaloids abundance was higher in the biphasic protocol and significantly lower in the Meth_ACN one, carboxylic acids intensities were lower in the two monophasic protocols, while for sphingolipids and vitamins the abundance was significantly higher with the two monophasic protocols than in the biphasic one. In C18 ESI− runs, higher intensities for carboxylic acids, monosaccharides, sterol lipids and vitamins were found using the biphasic protocol in comparison with the monophasic ones, whereas fatty acyls were more abundant in the monophasic protocols.
Footnote |
† Electronic supplementary information (ESI) available: (1) Supplementary tables and figures (PDF); (2) Table S2 (Microsoft Excel file): retention time and neutral mass values for in-house standard library; (3) Table S3 (Microsoft Excel file): additional information on the annotation quality for each experiment (assigned annotation, compound, neutral mass (Da), m/z, retention time (min), chromatographic peak width (min), identifications, identifier, MSI classification, adducts, formula, score, fragmentation score, absolute mass error (ppm), isotope similarity); (4) Table S4 (Microsoft Excel file): RSD values for all compounds across extraction protocols and for quality control samples for each experiment. All raw data is available on-line at Metabolomics Workbench1 using identifier PR001650. See DOI: https://doi.org/10.1039/d4mo00015c |
This journal is © The Royal Society of Chemistry 2024 |