Investigating the Potential of Ion Mobility-Mass ... - ACS Publications

May 31, 2019 - from University of State of Rio de Janeiro and by Prof Sergio Lourenço from Fluminense. Federal University, respectively. Dunaliella s...
0 downloads 0 Views 7MB Size
Download PDF
Article Cite This: Anal. Chem. 2019, 91, 9266−9276

pubs.acs.org/ac

Investigating the Potential of Ion Mobility-Mass Spectrometry for Microalgae Biomass Characterization Maíra Fasciotti,*,†,⊥ Gustavo H. M. F. Souza,‡ Giuseppe Astarita,§ Ingrid C. R. Costa,† Thays. V. C. Monteiro,† Claudia M. L. L. Teixeira,∥ Marcos N. Eberlin,# and Amarijt S. Sarpal†

Downloaded via UNIV OF SOUTHERN INDIANA on July 23, 2019 at 09:34:19 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

†

National Institute of Metrology, Quality and Technology (INMETRO), Division of Chemical and Thermal Metrology, Laboratory of Organic Analysis, 25250-020, Duque de Caxias, Rio de Janeiro, Brazil ⊥ ThoMSon Mass Spectrometry Laboratory, Institute of Chemistry, University of Campinas − UNICAMP, 13083-970 Campinas, São Paulo, Brazil ‡ MS Applications and Development Laboratory, Waters Corporation, 06455-000 Barueri, São Paulo, Brazil § Department of Biochemistry and Molecular and Cellular Biology, Georgetown University, Washington DC 20007, United States ∥ Microalgal Biotechnology Laboratory, National Institute of Technology (INT), 20081-312 Rio de Janeiro, Rio de Janeiro, Brazil # Mackenzie Presbyterian University, School of Engineering, 01302-907 São Paulo, São Paulo, Brazil S Supporting Information *

ABSTRACT: Algae biomass is formed by an extremely complex set of metabolites, and its molecular characterization has been very challenging. We report the characterization of microalgae extracts via traveling wave ion mobility−mass spectrometry (TWIM-MS) by two different analysis strategies. First, the extracts were analyzed by direct infusion electrospray ionization (ESI) with no previous chromatographic separation (DI-ESI-TWIM-MS). Second, the samples were screened for metabolites and lipids using an untargeted high-throughput method that employs ultrahigh-performance liquid chromatography (UHPLC) using data-independent analysis (DIA) − MSE (UHPLC-HDMSE). Sixteen different microalgae biomasses were evaluated by both strategies. DI-ESI-TWIM-MS was able, via distinct drift times, to set apart different classes of metabolites, with the differences in the profiles of each microalga readily evident. With the UHPLC-HDMSE approach, 1251 different compounds were putatively annotated across 16 samples with 210 classified as lipids. From the normalized abundance for each annotated compound category, a detailed profiling in terms of metabolites, lipids, and lipid classes of each sample was performed. The reported workflow represents a powerful tool to determine the most suitable biotechnological applications for a given alga type and may allow for real-time monitoring of the algae composition distribution as a function of growth conditions, feedstocks, and the like. The determination of collision cross section results in improved confidence in the identification of triacylglycerols in samples, highly applicable to biofuels production. The two analysis strategies explored in this work offer powerful tools for the biomass industry by aiding in the identification of ideal strains and culture conditions for a specific application, saving analysis time and facilitating identification of a large number of constituents at once.

A

science. With the advent of modern biotechnology, algae have received even more attention, especially with the emergence of third and fourth generation fuels.5−7 After it was realized that

lgae are a diverse group of predominantly aquatic organisms that efficiently perform photosynthesis. They are responsible for close to 90% of the photosynthesis performed on the planet1−3 and form a diversified polyphyletic group; that is, they in fact form a group of an expanded set of organisms belonging to different phylogenetic groups.4 Algae have been widely studied for decades within several fields of © 2019 American Chemical Society

Received: May 8, 2019 Accepted: May 31, 2019 Published: May 31, 2019 9266

DOI: 10.1021/acs.analchem.9b02172 Anal. Chem. 2019, 91, 9266−9276

Article

Analytical Chemistry

size of the ions as measured by their collision cross section (CCS), ion−dipole interactions of the ions with the neutral drift gas molecules.57 Previous studies have demonstrated the excellent performance of the IM technique for separation of isomers and isobars and characterization of complex mixtures,58−61 as well as for increasing the selectivity of lipidomics and metabolomics analyses.50 In complex mixtures, different classes of compounds are separated by forming sets of characteristic tendencies, that is, drift time regions. To determine the best application for a specific biomass and also to evaluate the process of algae production, a simple, fast but as comprehensive as possible molecular characterization, for instance, of TAGs for biodiesel production, would be of great help. This work aimed, therefore, to explore the potential of ion mobility coupled to high-resolution mass spectrometry to comprehensively characterize different microalgae biomasses. Two different TWIM-MS analysis strategies were tested as detailed below: DI-ESI-TWIM-MS and UHPLCHDMSE.

algae were able to provide better yields than processes with other biomasses, biofuels from algae have been framed as more promising third-generation fuels.8 Intensive research followed in the quest of more efficient production of different types of biofuels and chemicals from algae.9,10 A very interesting advantage of algae as a source of biomass is that they can be grown in waste and salty waters,11,12 acting simultaneously as a bioremediation strategy.13,14 Currently, studies are focused on producing biohydrogen from algae via solar energy and water,15 as well as bioisoprene (2-methyl-1,3-butadiene)16 and bioethanol.17 Butanol and isobutanol have also been directly produced from CO2 via algae bioprocesses.18,19 Biodiesel is perhaps the most promising biofuel to be produced via algae, since some oleaginous algae have 30−60% (w/w) of triacylglycerols on a dry basis.20,21 To obtain better yields to produce biodiesel precursor lipids, mainly triacylglycerols, the algae species have been cultivated in different culture media, which strongly influence biomass composition.22 For instance, nitrogen limitation or nutrient imbalance increase lipid content in the microalgae cells, leading to high biodiesel potential production.23 Algae-based biofuels are indeed promising, hence major improvements in algae technology are being tested, especially for downstream and upstream processes optimization.24,25 The characterization of the resulting algae biomass is also extremely important to ensure that the growth conditions are actually affecting the biomass composition as desired,26 and can guide best biotechnological practices in the production of biofuels, biogas, biobased chemicals, bioremediation, or even in the area of nutraceutical and food supplements.27,28 The characterization of this extremely complex29 algae biomass has been performed using several analytical techniques, from the less expensive but less comprehensive classical physical-chemical methods such as elemental analysis, gravimetric determination of total lipid content, or classical methods for total protein or carbohydrate measurements;30 to more complex but more training and time-consuming instrumental techniques such as NMR,31,32 FTIR,33 laserbased techniques,34 and chromatography.35,36 Mass spectrometry (MS) generally coupled to previous chromatographic separation has been established as a powerful and versatile analytical technique for both quantitation of target compounds as well as identification of a massive number of unknown compounds at once in a wide variety of matrices and samples. MS-based analytical methods have already been successfully applied for algae characterization, and a great variety of chemicals have been identified such as peptides,37 proteins,38 toxins,39 isoflavones,40 lipids,41 fatty acids,42 sterols,43 polysaccharides44 hydrocarbons,45 aldehydes,46 and others.47−49 With the introduction of commercially available instruments, ion mobility coupled with high-resolution mass spectrometry (IM-HRMS) has recently become the technique of choice in omics sciences, especially in proteomics, lipidomics, and metabolomics studies.50−54 The technique is frequently coupled with liquid chromatography leading to multidimensional data, combining the compounds’ parameters of retention times versus ions m/z ratios versus ions drift times, the former being achieved in a few minutes and the latter on the millisecond time scale. Ion mobility (IM) is a technique that separates gaseous ions while traveling through a buffer gas under the influence of an electrical field.55,56 TWIM separation is based on the shape/

■

EXPERIMENTAL SECTION Chemicals. Methanol and chloroform (LC grade) were purchased from Tedia Brazil (Rio de Janeiro, Brazil). Analytical-grade phosphoric acid (85% w/w), formic acid (98−100%), LC-grade leucine enkephalin, poly-DL-alanine, and ammonium formate were purchased from Sigma-Aldrich (Merck Sigma-Aldrich, KGaA, Darmstadt, Germany). Mobile-phases acetonitrile and propan-2-ol (MS grade) were purchased from J. T. Backer (Sao Paulo, Brazil). Water was purified using a Millipore Milli-Q system (Merck Millipore, KGaA, Darmstadt, Germany). Samples. In total, 16 microalgae biomasses were analyzed, including 5 from different genera (Tetraselmis aff. chuii, (Arthrospira Spirulina) sp. , Dunaliella salina, Chlorella vulgaris, Scenedesmus ecornis), and these were obtained under different cultivation times and/or culture media composition or light source (Scheme 1). See Supporting Information (SI) for detailed cultivation conditions. For comparison with microalgae extracts, commercial samples of refined soybean, corn, rapeseed, sunflower, and olive oils were purchased in the local market. Sample Preparation. Lyophilized biomasses of different algae were extracted using the Bligh and Dyer-like method62 with some modifications.63 For Tetraselmis aff. chuii, Arthrospira (Spirulina) sp., Chlorella vulgaris, and Scenedesmus ecornis samples, 0.1 g of each lyophilized biomass was homogenized with 15 mL of chloroform:methanol:H2O (2:1:0.5, v/v) and placed in an ultrasonic bath (200W, 50 Hz) for 1 h at 30 °C. The lipid fractions were filtered using filter paper and then dried under water vapor heat set to 60 °C. The dried lipid fractions were determined gravimetrically. For the Dunaliella salina samples (ABC and DEF culture media), the protocol was different because of the high concentration of salts in the dried biomass. The lipid extraction was then performed in two steps: the first consisted in the dry biomass extraction using pure chloroform (20 mL) under ultrasonic agitation also for 1 h at 30 °C. The solid was decanted, and the liquid phase was transferred to a beaker. Then, in the second step, a volume of 20 mL of methanol was used with 300 μL of H2O to extract the solid residue of the first step (with pure chloroform), under the same extraction conditions. The chloroform and methanol extracts were also evaporated 9267

DOI: 10.1021/acs.analchem.9b02172 Anal. Chem. 2019, 91, 9266−9276

Article

Analytical Chemistry

i-Class as chromatograph (both Waters, U.K.). The separation was carried out using an ACQUITY CSH C18, 2.1 mm × 100 mm column kept at 55 °C during analysis. The mobile phase consisted of solvent A (acetonitrile:H2O 60:40 + 10 mM ammonium formate +0.1% formic acid) and solvent B (propan-2-ol:acetonitrile 90:10 + 10 mM ammonium formate +0.1% formic acid). The gradient elution program was as follows: 0−0.99 min from 60:40 A:B to 57:43 A:B; 0.99−1.24 min from 57:43 A:B to 50:50 A:B; 1.24−5.39 min from 50:50 A:B to 1:99 A:B; 5.39−5.64 min from 1:99 A:B to 60:40 A:B, then remaining constant until 8 min. Flow rate was 0.7 mL· min−1, and sample injection volumes were 4 μL (FTN AutoSampler). Total run time was 8 min. The mass spectrometer was equipped with an ESI source, also in positive ion mode, operating under 3.5 kV of capillary and 40 V of cone. The source and the desolvation temperature were 110 and 450 °C, respectively, and the nitrogen nebulizer gas was used at flow rate of 900 L/h at a pressure of 6 bar. The mass spectrometer was operated in ion-mobility (HDMSE) mode. In this method, the mass analyzer operates in both low-energy (for precursors ions) and high-energy (for fragments) mode to promote collision-induced dissociation (CID) of nonselected precursor ions, leading to spectra of product fragments aligned to the precursor ion. In this technique, the term “HD” comes from “high definition”, because in this case, the MSE acquisition technique is also coupled to the ion mobility. Trap collision energy (low) was 4 eV, and transfer collision energy (high) was a ramp from 25 to 65 eV, also using argon as collision gas. The IMS cell was filled with N2 under a pressure of 3.65 mbar. The IM cell was also calibrated using poly-DLalanine (10 mg/L) in H2O: acetonitrile (50:50 + 0.1% of formic acid) to allow the calculation of the analytes’ CCS values. Both MS calibration and lock mass correction during analysis was performed using a leucine enkephalin (theoretical m/z 554.2620) solution in H2O:acetonitrile (50:50 + 0.1% formic acid) at a concentration of 100 pg/μL and infused at a flow rate of 0.010 mL/min. The acquisitions in HDMSE mode ranged from m/z 50 to 1500. Default IMS conditions were used for IMS screening (T-wave velocity ramp start of 1000 m/s and end of 300 m/s; and T-wave pulse height of 40 V). Data Processing and Statistics. Both DI-ESI-TWIM-MS and UHPLC-HDMSE data were collected using the software MassLynx 4.1v (SCN 639 for Synapt HDMS and SCN 932 for Synapt G2-Si, Waters, U.K.). The DI-ESI-TWIM-MS spectral data were evaluated using the software DriftScope 2.7v and HDMS Compare 1.0v. The chemometric analysis for samples statistical comparison after data preprocessing (spectra normalization by height and Pareto scaling−default software parameters) was performed using the software EZinfo 2.0v (Umetrics, U.S.A.) appended into MarkerLynx (Waters, U.K.). For UHPLC-HDMSE data, processing and analysis was conducted by using Progenesis QI Informatics (Nonlinear Dynamics, Newcastle, U.K.) in a standard protocol50 in which each UHPLC−MS run was imported as a raw file to obtain the ion-intensity map, including m/z and retention time. As default, these ion maps were then aligned in the retention-time direction. From the aligned runs, an aggregate run representing the compounds in all samples was used for peak picking. This aggregate was then compared with all runs, to ensure that the same ions are detected in every run. Isotope and adduct deconvolution were applied in order to reduce the number of features detected. Data were normalized according to total ion intensity. Metabolites were identified by searches against their

Scheme 1. Sixteen Different Microalgae Biomasses Analyzed in This Work

under 60 °C, and the dried extracts were gravimetrically determined. All the lipid extracts for all samples were then redissolved in pure methanol at an approximate concentration of 2 mg·mL−1. Then, 20 μL of the lipid fraction in methanol was diluted in 1 mL of methanol with 0.1% of formic acid for the direct infusion approach and in mobile phase B for the UHPLC-HDMSE analysis.

■

METHODS DI-ESI-TWIM-MS Approach. The direct infusion (DI) approach was performed on a Traveling Wave Ion Mobility Mass Spectrometer (TWIM-MS, Synapt HDMS, Waters, U.K.). The samples were direct infused at a flow rate of 20 μL/min using an external syringe pump into the ESI source operated in positive mode (3.5 kV), and cone and extractor cone were 40 and 5.0 V, respectively. The source and desolvation temperatures were 80 and 150 °C, and the nitrogen nebulizer gas flow rate was 400 L/h. Parameters such as the drift gas pressure (N2, 1.05 mbar), height (30 V), and wave velocity (650 m/s) were optimized to obtain the same spectra as when only QTOF mode is used, avoiding any distortion or sensitivity loss, in a range of m/z 50−1500. The mass spectrometer was calibrated in ESI(+) mode at 10,000 resolution (FWHM at m/z 882.7999) over an acquisition mass range of m/z 50−2000, using a solution of phosphoric acid 0.1% (v/v) in acetonitrile:H2O (1:1). The most abundant ions in the mass range from m/z 800 to 1000 (expected TAG region) had also their CCS values experimentally determined according to the procedure described by Paglia and Astarita in Nature Protocols (2017)50 for performing the manual CCS calibration, using poly-DL-alanine oligomers (10 mg/L, in H2O:acetonitrile 50:50 + 0.1% of formic acid) as reference values for calculation. A spreadsheet with the CCS calculations for isobaric ions is available as SI. UHPLC-HDMS E. UHPLC-HDMSE was performed in a Synapt G2-Si mass spectrometer with an ACQUITY UHPLC 9268

DOI: 10.1021/acs.analchem.9b02172 Anal. Chem. 2019, 91, 9266−9276

Article

Analytical Chemistry

Figure 1. DI- ESI(+)-TWIM-MS of the extract of Tetraselmis aff. chui cultivated in culture media G4 (sample “TG4”).

Figure 2. Principal component analysis (PCA) of the 16 algal extracts. Note the grouping by algae genus.

lipids precursors of fatty acid methyl esters (FAMEs) for biodiesel production. Figure 1 shows the full scan mass spectrum obtained by DIESI-TWIM-MS of the microalgae extract Tetraselmis aff. chui “TG4”. It highlights the sample complexity and some examples of expected lipid classes that could be present in specific m/z ranges. Figures S1−S16 show all other spectra. Quite different profiles are indeed observed, which shows that the composition of the algal biomass is considerably altered as a function of the culture media and species or strains and that these changes are easily detected via DI-ESI-TWIM-MS. Note that determining how specific classes of compounds vary more pronouncedly as a function of algae growth parameters is critical to optimize production conditions. Data from all 16 DI-ESI-TWIM-MS spectra for each microalgae extract were statistically evaluated by chemometrics

accurate masses in publicly available databases, including the LIPID MAPS64 database, the Human Metabolome database (HMDB),65 and MetaScope (CCS database provided in Progenesis QI software), as well as by fragmentation patterns, retention times, and CCS values, when available. The platform LipidCCS66 was also used for additional CCS comparison.

■

RESULTS AND DISCUSSION DI-ESI-TWIM-MS of Microalgae Extracts. To determine the best application for a specific biomass and also to evaluate the process of algae production, researchers and biorefineries must have robust analytical tools for the detailed analysis of these biomasses. We report two different TWIM-MS analysis strategies for performing a quite challenging molecular characterization of microalgae biomass extracts, focusing on triacylglycerols (TAGs), as they are the major transesterifiable 9269

DOI: 10.1021/acs.analchem.9b02172 Anal. Chem. 2019, 91, 9266−9276

Article

Analytical Chemistry

Figure 3. Representative two-dimensional drift time versus m/z of the MS detected ions for the samples extracts of (a) TG3; (b) SPIZK, (c) TG4, and (d) CHLOSPS.

Figure 4. Fusion map comparing the “DT versus m/z” spectrum of sample TG3 and a blend of commercial edible oils obtained with the DITWIM-MS approach. The inset highlights the superimposed drift time regions for GL_TAG class in both samples.

(Figure 2) using principal component analysis (PCA). Grouping occurred primarily by the microalgae genus, confirming that microalga metabolite profile is indeed a mainly characteristic feature of the genus. Therefore, it follows that the genus could be the initial variable to be considered in a study of alga prospecting for a specific biotechnological application. For example, after defining the best genus, the culture medium and cultivation conditions could be studied and optimized. This approach may also be suitable to compare with reference samples, that is, microalgae that have been already shown to be

suitable for a particular application or whose lipid profile has already been well-characterized. It is a simple and fast approach and provides a good response for an initial profiling of a given set of samples. The algae extracts were also evaluated by their 2D drift time versus mass spectra, considering the TWIM separation of the ionized lipids and metabolites in samples. Figure 3 shows representative examples of 2D drift time versus m/z spectra (detected ions with their m/z values) for some microalgae species. Figures S17−S32 show the spectra for all samples. 9270

DOI: 10.1021/acs.analchem.9b02172 Anal. Chem. 2019, 91, 9266−9276

Article

Analytical Chemistry

Figure 5. Experimentally derived CCS values (Ω in Å2) for the most abundant ions in the sample TG4 obtained by DI-TIWM-MS, with the GL_TAG region highlighted.

wave velocity 650 m/s). Each selected ion has its CCS experimentally determined by calibration using substances whose CCS is known, thereby increasing data selectivity.67 Adding one more intrinsic physical-chemical ion property, such as the CCS, also significantly increases the confidence of metabolite/lipid identification.68 One of the most common calibrants for CCS values is poly-DL-alanine, whose CCS values for its singly charged oligomers have been well determined in N2 as drift gas. Figure 5 shows the experimentally determined CCS values (Ω in Å2) (electronically available in SI) for the most abundant ions from m/z 800 to 1000, exemplified for the sample TG4. Notably, a clear CCS separation exists for TAGs, resulting in a cluster possessing slightly higher CCS values, versus other isobaric ions likely probably from other metabolites classes. This highly interesting finding could be useful in assessing a qualitative response regarding TAG abundance in algal samples, as this region of drift time x m/z would appear to be selective for the TAG class. This useful spectral characteristic of TAGs can be used to inform the optimization of algae cultivation for biodiesel production. High-Throughput Characterization of Lipids and Metabolites by UHPLC-HDMSE. To ultimately perform a comprehensive characterization of algae extracts, additional MS/MS spectra are crucial in allowing full annotation. However, the manual interpretation of MS/MS data obtained by DI-ESI-TWIM-MS of the extracts would be extremely laborious and time-consuming because of the complexity of the samples that present different classes of lipids and other metabolites. We, therefore, applied the much more comprehensive UHPLC-HDMSE technique as a second approach to analyze the lipids and metabolites of the algae samples. In addition to the exact mass and CCS values obtained for precursor ions, MS/MS spectra are collected, generating information about fragments that can be compared to spectral databases and lead to a tentative annotation. Each chromatographic run, together with the MS, MS/MS and IMS data associated with it, was imported to Progenesis

Different drift time tendencies, even those possessing overlap in their m/z values, indicate that distinct classes of compounds are potentially separable. The differences in the profiles of each algae extract becomes more evident upon such visualization, and this approach allows for the profile screening of lipids. For example, to investigate the TAG content in a lipid extract, since this class of compounds will be distributed within a certain m/z range (which, however, may be superimposed with several other classes of lipids and metabolites, making the detection of this class difficult), the analyst can use the fact that the TAG ions will be distributed in a tendency of characteristic drift times (for a given experimental condition). Therefore, this approach provides rapid turnaround with respect to parameter changes and other decisions in algae culture studies, including continuation or termination of a biorefinery process depending on whether a desired product profile or other end point can be or has been met. Assessment of the lipid profile of algal samples can be confirmed by an additional approach, that is, the binary comparison of drift time x m/z spectrum through the HDMS compare software. In this approach, the ion mobility spectra are superimposed, highlighting the classes of compounds that are more abundant in each sample, generating a map in which each sample is marked with a different color, also named as fusion maps. Figure 4 shows the comparison of the spectrum of TG3 and a blend of commercial vegetable oils (a mixture of soybean, corn, rapeseed, sunflower, and olive oils in even proportions that are knowingly composed by TAGs). Clearly, the TAG abundance in sample TG3 is superimposed with the TAGs of the vegetable oil blend (Figure 4). Figure S33 also shows another binary comparison of samples TG4 and SPIKZ, being evident the differences in their drift plots, indicating that the sample SPIKZ has a notably lower abundance of TAGs. The principal TAGs ions that show characteristic mobility tendencies can also be further analyzed by their CCS values, manually determined via calibration with poly-DL-alanine standard solution infused in exactly the same TWIM conditions as the samples (N2 1.05 mbar, height 30 V, and 9271

DOI: 10.1021/acs.analchem.9b02172 Anal. Chem. 2019, 91, 9266−9276

Article

Analytical Chemistry

Figure 6. Sum of the normalized abundances of (a) each lipid classes and other metabolites annotated in the samples and (b) each lipid class annotated in the samples; (c) Relative abundance of lipids and metabolites for each sample; (d) Relative distribution of lipid classes in each sample. Legends: fatty acyls (FA), glycerolipids (GL), glycerophospholipids (GP), triacylglycerol (GL_TAG), sphingolipids (SP), sterol lipids (ST), prenol lipids (PR), saccharolipids (SL).

TAG are also lipids (glycerolipids), this class was categorized separately as “GL_TAG”, as we discuss and emphasize its importance for biodiesel production. From the normalized abundances of each deconvoluted annotated compound per sample, several interpretations can be made. Figure 6 shows the lipidomics and metabolomics results interpretations for all the samples. Figure 6a shows the sum of the normalized abundance for each class of lipids and other metabolites, and Figure 6b shows only the normalized abundances for different lipid classes in the samples. Figure 6c shows the percentage abundance of compounds identified as lipids and as other metabolites, and Figure 6d shows the percentage abundance of lipid classes distributed in each sample. Figure 6 data are not quantitative, meaning that they cannot be considered as a “weight by weight %” measurement in the sample extracts as not all the reported compounds have the same ionization behavior, implying different response factors. Note that lipidomics or metabolomics analysis are comparative methods and would not cover 100% of the compounds that are indeed present in a given sample. It is expected that several other compounds remained undetected and/or unidentified, but a comprehensive profile is not essential for comparisons and predictions. Note also that, as recently reported by Dorrestein and co-workers49 in a recent metabolomics study of algae and cyanobacteria, approximately 86% of the metabolomics signals detected were not found in other available data sets, concluding that even using a very diverse data sets, algae and cyanobacteria have unique features and families of metabolites that are still uncharacterized. However, the normalized abundance of each class of known compounds

QI software, which via comparison with the LIPID MAPS, HMDB, and a search engine with a CCS database provided by Waters (MetaScope) performs the identification by theoretical/experimental comparison of the precursor and fragment exact masses, isotope distribution, fragmentation profile (in silico fragmentation or known experimental fragmentation pathways), and CCS values. The combination of all these identification parameters results in an overall compound identification score for each identified compound. From the 16 analyzed samples, all in the same conditions of analysis, 1251 different compounds were putatively annotated, within an m/z error of 10 ppm for both precursor ion and fragments. Among these 1251 compounds: (1) only 158 could not be identified by their fragments; (2) 30 of them were also identified by their CCS, assuming a theoretical/experimental error lower than 4%; (3) the isotope similarities were from 30.4 to 99.4, with an average of 82.1; (4) for the fragmented ions, the average fragmentation score was 41.9 (ranging from 1 to 99.4); and finally, (5) the identification score ranged from 30 to 59.1, with an average of 39.5. The annotations with the highest scores were accepted and described in the accompanying spreadsheet with the metabolomics and lipidomics description of this study, provided as SI. The number of compounds annotated by their CCS values is relatively low (30 compounds) due to limitation of the database used, as it is not as broad as the LIPID MAPS or the HMDB. Among the 1251 compounds that were putatively annotated, 210 were lipids, with the others categorized as “other metabolites”. The lipids were also categorized into fatty acyls (FA), glycerolipids (GL), glycerophospholipids (GP), sphingolipids (SP), sterol lipids (ST), prenol lipids (PR), and saccharolipids (SL). Although 9272

DOI: 10.1021/acs.analchem.9b02172 Anal. Chem. 2019, 91, 9266−9276

Article

Analytical Chemistry

Figure 7. Comparison of the two-dimensional drift time versus m/z plot for sample TG3 with the CCS experimentally determined by UHPLCHDMSE for all the 1251 compounds, highlighting the areas I, II, III, and IV that could be detected in both plots.

detect misleading identifications if some of them appear in a very distinct region. It can also help to choose classes of compounds and take a closer look at their identification. The separation of distinct tendencies in the sample TG3 are more evident because not all of the 1251 compounds are present in sample TG3. The use of more polarizable drift gases such as CO2 has been proven to be an alternative to increase the selectivity and classes separations in complex mixtures such as crude oil.59 On the other hand, databases of CCS values in CO2 or CCS calibrants (such as poly-DL-alanine) values in this gas are not yet available, as the determination of this parameter in CO2 is much more complex. However, if only the separation of classes of different compounds are needed and no CCS values must be determined, the use of CO2 can be a good alternative to better detect different classes of lipids and metabolites in complex mixtures such as microalgae extracts. Additional TAGs confirmation was made by comparing the experimental data of measured CCS values with the ones available in the recently released CCS database, LipidsCCS,66 (Table S1). Both the measured and the LipidsCCS theoretical values were determined in N2, as drift gas, so they are comparable. Only two compounds annotated as TAG had no theoretical CCS value available in LipidCCS data set, which were TG(10:0_8:0_8:0) and TG(21:0_22:6_22:6). For the other TAGs, the experimental versus theoretical CCS error was mostly below 10%, with a 6.7% average. The experimental values tended to be around 20 Å2 positively biased, that is, 20 Å2 on average larger than the theoretical values, but this minor systematic error probably comes from the Synapt G2-Si CCS calibration with poly-DL-alanine step. In general, therefore, the CCS were mostly in accordance with LipidCCS database, which shows that this parameter is actually very useful and helps with further identification and confirmation of compounds. The calculated CCS values obtained with DITWIM-MS (Figure 5) seem to be in accordance with theoretical data, with the CCS values of TAGs mostly falling within 300 to 330 Å. Another interesting approach that could be used for CCS determinations in both DI-TWIM-MS and UHPLC-HDMSE is the use of a standard mixture of lipids of interest with available theoretical CCS values.66 In this way, it is expected that experimental CCS values could be determined more

should serve as suitable screening for the alga biomass composition and to understand how different processes may be affecting alga composition. Figure 6a,c point out that the abundance of compounds identified as metabolites is higher than the compounds identified as lipids, as indeed expected, as the number of potential metabolites is indeed much higher than the number of compounds of the lipid class. The sample with the highest abundance of lipids was the sample TG2 (Figure 6c), mainly distributed between GL (excepting GL_TAG, which is highlighted separately) and GP. The sample with the highest % of GL_TAG was DUNDEF_CHCl3 (Figure 6d), but when comparing the sum of the normalized abundance (Figure 6b). It can be noted also that the samples DUNDEF_CHL3, TG3, and TG4 were the ones presenting the highest abundances of compounds identified as TAGs. It indicates that they would at least in theoryprovide better yields for the biodiesel production, if it was a real case of a study of algae prospecting for biodiesel production. The detection of some metabolites and ST compounds may have important implications when alga biomass is used for nutraceutical products. When the untargeted approach detects toxins or sterol compounds (i.e., testosterone) or other classes of biologically active compounds, it is important to state at this point that further studies must be conducted, and these compounds must be characterized using targeted analysis to confirm the primary assumption of the untargeted analysis. In fact, the DI-ESI-TWIM-MS data really converge with the CCS values determined by UHPLC-HDMSE. Comparing the drift time versus m/z plot for a sample obtained by the DI approach with the plot of all the experimental CCS values versus m/z measured for all the annotated compounds, we can observe the same drift time tendencies showing up in both cases. To illustrate that, Figure 7 compares the TG3 sample drift time plot and the experimental CCS values determined for all the 1251 annotated compounds. We can highlight at least four different areas: I − a low abundance very separated tendency, probably minor metabolites; II − the compounds identified as TAGs, mainly from m/z 700−1000 and slightly higher CCS values; III − a very busy region with several superimposed lipids and other metabolites; and IV − also a low abundant separated tendency. We can also observe in what regions the lipid classes tend to appear, which is helpful to 9273

DOI: 10.1021/acs.analchem.9b02172 Anal. Chem. 2019, 91, 9266−9276

Article

Analytical Chemistry

as lipids of several classes. Among the 16 samples evaluated in this study, Tetraselmis aff. chuii, cultivated in F/2 medium plus fertilizers for 12 (TG3) and 16 days (TG4), as well as biomass of Dunaliella salina produced in nitrogen-reduced Shaish medium and extracted with CHCl3 (DUDEF_CHL3), were the ones with the highest TAGs abundances and are, therefore, more promising for biodiesel production within the evaluated samples. The methodologies explored in this work could be applied to microalgae biorefineries to predict whether some specific algae will be viable for biodiesel production or for other biotechnology applications, as well as help in studies of microalgae cultivation improvements.

accurately, as the CCS calibrants and the targeted lipid classes for a set of experimental conditions would exhibit the same gas-phase behavior. Finally, comparing the two analysis strategies presented in this work, it is important to emphasize that the DI-TWIM-MS allows a rapid fingerprinting of microalgae extracts, helping the visualization of different classes of lipids and metabolites via mobility mass correlation curves such as the data presented in Figure 7, in which lipid classes such as TAGs can be detected in a specific m/z versus ion mobility tendency. Moreover, from the DI-TWIM-MS data, the binary comparison of the twodimensional drift time versus m/z spectrum of different samples can be applied, for instance, to compare two cultivation conditions or two microalgae of different species, highlighting the differences between them, and it can be very useful in the biotechnology industry. However, a comprehensive molecular description with DI-TWIM-MS is an extremely laborious task because of sample complexity, and it is not suitable for routine analysis, as MS/MS data would have to be manually acquired. For that, if a more detailed molecular characterization is required to understand the metabolites or lipids that are most affected by some specific cultivation condition, using an additional anlysis, for example, the UHPLC-HDMSE strategy, would allow annotation of metabolites and lipids in samples to also consider their CCS values as a specific molecular descriptor to improve identification confidence.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.9b02172. Detailed cultivation conditions of each microalgae biomass, direct infusion ESI(+)-QTOF (Synapt HDMS) mass spectra for all samples, two-dimensional drift time versus m/z for all algae samples, fusion map comparing samples TG4 and SPIZK, additional TAGs confirmation by comparing the experimental data of measured CCS values with LipidsCCS theoretical values (PDF) Additional characterization data (XLSX) CCS calculations (XLSX)

■

CONCLUSIONS All the approaches demonstrated here have the potential to be applied as methodologies for screening of lipids and metabolites in real cases of microalgae prospecting and can be implemented as routine both for research and in microalgae biorefineries. Ion mobility mass spectrometry provided a fast profile of microalgae extracts, in which the new dimension of ion mobility separation reveals distinct classes of compounds, especially triacylglycerols. The variability of the lipids in the algae was found to be incredibly high and sensitive to the cultivation parameters, but the statistical classification among the samples indicated that the main parameter that influences the composition of the extracts is the genus of the algae. Genus seems, therefore, to be an initial variable to be considered in a study of algae prospecting for a specific biotechnological application. For example, after defining the best genus, the culture medium and cultivation conditions could be studied and optimized. Statistical or binary comparison of the samples drift time plots spectrum seems necessary, especially for comparison with reference samples, that is, with a microalga that has been already shown to be suitable for a particular application, or whose lipid profile has already been characterized. DI-TIWIMMS has been shown, therefore, to be a simple and fast approach and provides a good response for an initial profiling of a given set of samples, but performing comprehensive identification of all the ions of the samples can be quite laborious. The high level of complexity of the samples requires a more powerful methodology to perform a more global and detailed sample screening. For this, the UHPLC-HDMSE technique was used to putatively annotate over 1200 compounds, collectively in the 16 samples, based on intrinsic parameters of the detected ions (exact mass, fragmentation pattern, isotope distribution, and CCS). More than 200 were annotated

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Maíra Fasciotti: 0000-0002-6358-6067 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank the State of São Paulo Research Foundation (FAPESP), State of Rio de Janeiro Research Foundation (FAPERJ), the Brazilian National Council for Scientific and Technological Development (CNPq) and the Financing Agency of Studies and Projects (FINEP) for financial assistance. We gratefully acknowledge Waters Brazil managers for opening their demo laboratory (Barueri, Brazil) to run the UHPLC-HDMSE experiments. We also thank the National Institute of Technology for microalgae biomasses donation and Dr. M. D. Bartberger for the scientific opinion about the final work. Finally, we thank the referees of this work for their contribution to its quality improvement.

■

REFERENCES

(1) Raven, J. A.; Giordano, M. Curr. Biol. 2014, 24 (13), R590− R595. (2) Lloyd, N. D. H.; Canvin, D. T.; Culver, D. A. Plant Physiol. 1977, 59 (5), 936−940. (3) Pedersen, O.; Colmer, T. D.; Sand-Jensen, K. Front. Plant Sci. 2013, 4, 1−19. (4) Bhattacharya, D.; Medlin, L. Plant Physiol. 1998, 116 (1), 9−15. (5) Fasciotti, M. Sustainable Chemistry and Pharmacy 2017, 6, 82− 89. 9274

DOI: 10.1021/acs.analchem.9b02172 Anal. Chem. 2019, 91, 9266−9276

Article

Analytical Chemistry (6) Lü, J.; Sheahan, C.; Fu, P. Energy Environ. Sci. 2011, 4, 2451− 2466. (7) Alam, F.; Mobin, S.; Chowdhury, H. Procedia Eng. 2015, 105, 763−768. (8) Singh, A.; Nigam, P. S.; Murphy, J. D. Bioresour. Technol. 2011, 102 (1), 10−16. (9) Scott, S. A.; Davey, M. P.; Dennis, J. S.; Horst, I.; Howe, C. J.; Lea-Smith, D. J.; Smith, A. G. Curr. Opin. Biotechnol. 2010, 21 (3), 277−286. (10) Giordano, M.; Wang, Q. Microalgae for Industrial Purposes. In Biomass and Green Chemistry; Vaz, S., Jr., Eds; Springer: Cham, 2018. (11) Sun, X.; Wang, C.; Li, Z.; Wang, W.; Tong, Y.; Wei, J. Bioresour. Technol. 2013, 143, 499−504. (12) Miazek, K.; Kratky, L.; Sulc, R.; Jirout, T.; Aguedo, M.; Richel, A.; Goffin, D.; et al. Int. J. Mol. Sci. 2017, 18 (7), No. 1429. (13) Cuellar-Bermudez, S. P.; et al. Algal Res. 2017, 24, 438−449. (14) Delrue, F.; Á lvarez-Díaz, P. D.; Fon-Sing, S.; Fleury, G.; Sassi, J.-F. Energies 2016, 9, 132. (15) Kruse, O.; Hankamer, B. Curr. Opin. Biotechnol. 2010, 21, 238− 243. (16) Yang, J.; Xian, M.; Su, S.; et al. PLoS One 2012, 7 (4), No. e33509. (17) John, R. P.; Anisha, G. S.; Nampoothiri, K. M.; Pandey, A. Bioresour. Technol. 2011, 102 (1), 186−193. (18) Atsumi, S.; Cann, A. F.; Connor, M. R.; Shen, C. R.; Smith, K. M.; Brynildsen, M. P.; Chou, K. J. Y.; Hanai, T; Liao, J. C.; et al. Metab. Eng. 2008, 10, 305−311. (19) Atsumi, S.; Higashide, W.; Liao, J. C. Nat. Biotechnol. 2009, 27, 1177−1180. (20) Mata, T. M.; Martins, A. A.; Caetano, N. S. Renewable Sustainable Energy Rev. 2010, 14 (1), 217−232. (21) Chisti, Y. Trends Biotechnol. 2008, 26 (3), 126−131. (22) Kerkhoven, E. J.; Pomraning, K. R.; Baker, S. E.; Nielsen, J. npj Systems Biology and Applications 2016, 2, 16005. (23) Ma, N.-L.; Aziz, A.; Teh, K.-Y.; Lam, S. S.; Cha, T.-S. Sci. Rep. 2018, 8, 9746. (24) Sharma, A. K.; Sahoo, P. K.; Singhal, S.; Joshi, G. Bioresour. Technol. 2016, 216, 793−800. (25) Ashokkumar, V.; Salam, Z.; Sathishkumar, P.; Hadibarata, T.; Mohd Yusoff, A. R.; Ani, F. N. J. Cleaner Prod. 2015, 92, 162−167. (26) Laurens, L. M. L.; Van Wychen, S.; McAllister, S. A.; Arrowsmith, S.; Dempster, T. A.; McGowen, J.; Pienkos, P. T. Anal. Biochem. 2014, 452, 86−95. (27) Singh, S.; Kate, B. N.; Banerjee, U. C. Crit. Rev. Biotechnol. 2005, 25, 73−95. (28) Spolaore, P.; Joannis-Cassan, C.; Duran, E.; Isambert, A. J. Biosci Bioeng 2006, 101 (2), 87−96. (29) Collard, F.; Blin, J. Renewable Sustainable Energy Rev. 2014, 38, 594−608. (30) Slocombe, S. P.; Zhang, Q.; Ross, M.; Anderson, A.; Thomas, N. J.; Lapresa, A.; Rad-Menéndez, C.; Campbell, C. N.; Black, K. D.; Stanley, M. S.; Day, J. G. Sci. Rep. 2015, 9844, 1−17. (31) Pollesello, P.; Toffanin, R.; Murano, E.; Paoletti, S.; Rizzo, R.; Kvam, B. J. J. Appl. Phycol. 1992, 4, 315−322. (32) Sarpal, A. S.; Silva, P. R. M.; Martins, J. L.; Amaral, J. J.; Monnerat, M. M.; Cunha, V. S.; Daroda, R. J.; de Souza, W. Energy Fuels 2014, 28 (6), 3766−3777. (33) Dean, A. P.; Sigee, D. C.; Estrada, B.; Pittman, J. K. Bioresour. Technol. 2010, 101, 4499−4507. (34) Pořízka, P.; Prochazkova, P.; Prochazka, D.; Sladkova, L.; Novotny, J.; Petrilak, M.; Brada, M.; Samek, O.; Pilat, Z.; Zemanek, P.; et al. Sensors 2014, 14 (9), 17725−17752. (35) Sarpal, A. S.; Teixeira, C. M.; Silva, P. R.; Lima, G. M.; Silva, S. R.; Monteiro, T. V.; Cunha, V. S.; Daroda, R. J. Anal. Bioanal. Chem. 2015, 407 (13), 3799−3816. (36) Chopra, A.; Tewari, A. K.; Vatsala, S.; Kumar, R.; Sarpal, A. S.; Basu, B. J. Am. Oil Chem. Soc. 2011, 88, 1285−1296. (37) Bogialli, S.; Bruno, M.; Curini, A.; Di Corcia, C.; Fanali, C.; Lagana, A. Environ. Sci. Technol. 2006, 40 (9), 2917−2923.

(38) Khan, A.; Eikani, C. K.; Khan, H.; Iavarone, A. T.; Pesavento, J. J. J. Proteome Res. 2018, 17 (1), 23−32. (39) Zendong, Z.; McCarron, P.; Herrenknecht, C.; Sibat, M.; Amzil, Z.; Cole, R. B.; Hess, P. Journal of Chromatography A 2015, 1416, 10−21. (40) Klejdus, B.; Lojková, L.; Plaza, M.; Š nóblová, M.; Š těrbová, D. Journal of Chromatography A 2010, 1217 (51), 7956−7965. (41) Jones, J.; Manning, S.; Montoya, M.; Keller, K.; Poenie, M. J. Am. Oil Chem. Soc. 2012, 89 (8), 1371−1381. (42) Samburova, V.; Lemos, M. S.; Hiibel, S.; Kent Hoekman, S.; Cushman, J. C.; Zielinska, B. J. Am. Oil Chem. Soc. 2013, 90 (1), 53− 64. (43) Ř ezanka, T.; Vyhnálek, O.; Podojil, M. Folia Microbiol. 1986, 31, 44. (44) Anastyuk, S. D.; Shevchenko, N. M.; Nazarenko, E. L.; Dmitrenok, P. S.; Zvyagintseva, T. N. Carbohydr. Res. 2009, 344 (6), 779−787. (45) Han, J.; Calvin, M. Proc. Natl. Acad. Sci. U. S. A. 1969, 64 (2), 436−443. (46) Ma, J.; Xiao, R.; Li, J.; Li, J.; Shi, B.; Liang, Y.; Lu, W.; Chen, L. J. Sep. Sci. 2011, 34 (12), 1477−83. (47) Klejdus, B.; Lojková, L.; Plaza, M.; Š nóblová, M.; Š těrbová, D. Journal of Chromatography A 2010, 1217 (51), 7956−7965. (48) Kadam, S. U.; Tiwari, B. K.; O’Donnell, C. P. J. Agric. Food Chem. 2013, 61 (20), 4667−4675. (49) Luzzatto-Knaan, T.; Garg, N.; Wang, M.; Glukhov, E.; Peng, Y.; Ackermann, G.; Amir, A.; Duggan, B. M.; Ryazanov, S.; Gerwick, L.; Knight, R.; Alexandrov, T.; Bandeira, N.; Gerwick, W. H.; Dorrestein, P. C. eLife 2017, 6, No. e24214. (50) Paglia, G.; Astarita, G. Nat. Protoc. 2017, 12, 797−813. (51) Paglia, G.; Angel, P.; Williams, J. P.; Richardson, K.; Olivos, H. J.; Thompson, J. W.; Menikarachchi, L.; Lai, S.; Walsh, C.; Moseley, A.; Plumb, R. S.; Grant, D. F.; Palsson, B. O.; Langridge, J.; Geromanos, S.; Astarita, G. Anal. Chem. 2015, 87 (2), 1137−1144. (52) Paglia, G.; Williams, J. P.; Menikarachchi, L.; Thompson, J. W.; Tyldesley-Worster, R.; Halldórsson, S.; Rolfsson, O.; Moseley, A.; Grant, D.; Langridge, J.; Palsson, B. O.; Astarita, G. Anal. Chem. 2014, 86 (8), 3985−3993. (53) Paglia, G.; Kliman, M.; Claude, E.; Geromanos, S.; Astarita, G. Anal. Bioanal. Chem. 2015, 407 (17), 4995−5007. (54) Souza, G. H. M. F.; Guest, P. C.; Martins-de-Souza, D. Methods Mol. Biol. 2017, 1546, 57−73. (55) Kanu, A. B.; Dwivedi, P.; Tam, M.; Matz, L.; Hill, H. H., Jr. J. Mass Spectrom. 2008, 43, 1−22. (56) Cumeras, R.; Figueras, E.; Davis, C. E.; Baumbach, J. I.; Gràcia, I. Analyst 2015, 140, 1376−1390. (57) Fasciotti, M.; Lalli, P. M.; Heerdt, G.; Steffen, R. A.; Corilo, Y. E.; Sá, G. F.; Daroda, R. J.; Reis, F. A. M.; Morgon, N. H.; Pereira, R. C. L.; Eberlin, M. N.; Klitzke, C. F. Int. J. Ion Mobility Spectrom. 2013, 16, 117. (58) Valentine, S. J.; Kulchania, M.; Barnes, C. A. S.; Clemmer, D. E. Int. J. Mass Spectrom. 2001, 212, 97−109. (59) (A) Fasciotti, M.; Lalli, P. M.; Klitzke, C. F.; Corilo, Y. E.; Pudenzi, M. A.; Pereira, R. C. L.; Bastos, W.; Daroda, R. J.; Eberlin, M. N. Energy Fuels 2013, 27 (12), 7277−7286. (B) Santos, J. M.; Galaverna, R. S.; Pudenzi, M. A.; Schmidt, E. M.; Sanders, N. L.; Kurulugama, R. T.; Mordehai, A.; Stafford, G. C.; Wisniewski, A., Jr; Eberlin, M. N. Anal. Methods 2015, 7, 4450−4463. (60) McCullagh, M.; Douce, D.; Van Hoeck, E.; Goscinny, S. Anal. Chem. 2018, 90 (7), 4585−4595. (61) Benigni, P.; Thompson, C. J.; Ridgeway, M. A.; Park, M. A.; Fernandez-Lima, F. Anal. Chem. 2015, 87 (8), 4321−4325. (62) Bligh, E. G.; Dyer, W. J. Can. J. Biochem. Physiol. 1959, 37, 911−917. (63) Sarpal, A. S.; Teixeira, C. M. L. L.; Silva, P. R. M.; da Costa Monteiro, T. V.; da Silva, J. I.; da Cunha, V. S.; Daroda, R. J. Appl. Microbiol. Biotechnol. 2016, 100 (1), 2471. (64) LIPID MAPS Lipidomics Gateway. See the following: http:// www.lipidmaps.org. 9275

DOI: 10.1021/acs.analchem.9b02172 Anal. Chem. 2019, 91, 9266−9276

Article

Analytical Chemistry (65) (a) Wishart, D. S.; Tzur, D.; Knox, C.; et al. Nucleic Acids Res. 2007, 35 (Databaseissue), D521−6. (b) Wishart, D. S.; Knox, C.; Guo, A. C.; et al. Nucleic Acids Res. 2009, 37 (Databaseissue), D603− 610. (c) Wishart, D. S.; Jewison, T.; Guo, A. C.; Wilson, M.; Knox, C.; et al. Nucleic Acids Res. 2012, 41 (D1), D801−7. (d) Wishart, D. S.; Feunang, Y. D.; Marcu, A.; Guo, A. C.; Liang, K.; et al. Nucleic Acids Res. 2018, 46 (D1), D608−17. (66) Zhou, Z.; Tu, J.; Xiong, X.; Shen, X.; Zhu, Z.-J. Anal. Chem. 2017, 89, 9559−9566. (67) Campuzano, I.; Bush, M. F.; Robinson, C. V.; Beaumont, C.; Richardson, K.; Kim, H.; Kim, H. I. Anal. Chem. 2012, 84 (2), 1026− 1033. (68) Zhou, Z.; Tu, J.; Zhu, Z.-J. Curr. Opin. Chem. Biol. 2018, 42, 34−41.

9276

DOI: 10.1021/acs.analchem.9b02172 Anal. Chem. 2019, 91, 9266−9276