ANALYTICAL PROTOCOLS BASED ON HIGH- RESOLUTION MASS SPECTROMETRY FOR CHARACTERIZING EMERGING CONTAMINANTS AND THEIR DEGRADATION PRODUCTS IN FOODSTUFF AND ENVIRONMENT.
UNIVERSITÁ DEGLI STUDI DI PADOVA
DIPARTIMENTO DI SCIENZE CHIMICHE
SCUOLA DI DOTTORATO DI RICERCA IN SCIENZE MOLECOLARI INDIRIZZO SCIENZE CHIMICHE
CICLO XXVIII
ANALYTICAL PROTOCOLS BASED ON HIGH- RESOLUTION MASS SPECTROMETRY FOR CHARACTERIZING EMERGING CONTAMINANTS AND THEIR DEGRADATION PRODUCTS IN FOODSTUFF AND ENVIRONMENT.
Direttore della Scuola: Xx.xx Xxxx. Xxxxxxxx Xxxxxxxx
Supervisore: Xx.xx Xxxx. Xxxx Xxxxxxxx
Dottorando: Xxxxxxxxx Xxxxxxx
Anno 2013-2016
Ama e fa ciò che vuoi.
Table of contents
3.2 Screening analysis of emerging contaminants 13
3.3 HRMS data visualization tools 22
4 Aim of the project and structure of the thesis 25
4.1 Development of a two-step method for target and suspect analysis of freshwater cyanotoxins by LC/Q-TOF system 25
4.2 Development of “one-shot” analysis of PDE-5 inhibitors and analogues in natural products for the treatment of erectile dysfunction 26
4.3 Development of a workflow for HRMS analysis of PM2.5 organic fraction: post-run data analysis and the role of ionization sources 26
5.2 Samples and sample preparation 32
6 Development of a two-step protocol for target and suspect analysis of freshwater cyanotoxins by LC/Q-TOF system 41
6.2 Algal toxins identification in freshwaters 49
7 Development of “one-shot” analysis of PDE-5 inhibitors and analogues in natural products for the treatment of erectile dysfunction 61
7.1 Optimization of the instrumental conditions 61
7.2 Food supplement analysis 69
8 XXXXXX collaborative trial 75
8.3 Trial consideration and Conclusions 81
9 Development of a workflow for HRMS analysis of PM2.5 organic fraction: post-run data analysis and the role of ionization sources 83
9.1 APPI analysis optimization 83
9.3 Application of the protocol on real samples 109
9.4 Conclusions 122
10 General conclusion 125
11 Appendixes 127
12 References 143
13 Acknowledgment 151
In this PhD thesis, the capability of analytical systems based on high-resolution mass spectrometry (HRMS) has been investigated for the determination of emerging contaminants in environmental matrices and foodstuff. Since the molecular structures of the emerging contaminants could be know as well as unknown, target, suspect and non- target analyses have to be developed in order to propose a “mass-based” advanced screening. Attention has been focused on the scale-up process in the identification confidence by developing different specific protocols.
Two protocols based on HPLC/Q-TOF-MS have been developed for the simultaneous screening and confirmatory analysis of target and non-target cyanotoxins in freshwater intended for human consumption, PDE-5 inhibitors and analogues in food supplements marked as erectile dysfunction remedies. Both protocols have been optimized with the aim to obtain HRMS data of pseudomolecular ions and fragmentation patterns in tandem MS mode. In-house databases were implemented to simplify the data treatment. The application of these protocols in “non-target screening” mode has been attempted in real samples and in the frame of a collaborative trial organized by European XXXXXX foundation as regard as the analysis of water contaminants. The exercise was complex and time consuming, and it has highlighted the strengths and weaknesses of the developed protocols.
The crucial step in non-target screening was the assignment of reliable molecular formula to the m/z values. A specific workflow based on direct infusion and HRMS analysis by using an Orbitrap™ mass spectrometer has been developed for the characterization of PM2.5 organic fraction. The automatization of the data treatment using Mathematica based algorithms was accomplished for studying the chemical composition of PM2.5 organic fraction. Contextually, the possible use of the Atmospheric Pressure Photoionization source for characterizing PM2.5 organic fraction has been investigated on real samples.
1.1 Riassunto
In questa tesi di dottorato, le possibilità de l’uso de la spettrometria xx xxxxx ad xxxx risoluzione (HRMS) sono state indagate nella determinazione di contaminanti emergenti in matrici ambientali ed alimentari. Dal momento che le strutture molecolari dei contaminanti emergenti potrebbero essere ancora sconosciute, la loro determinazione richiede analisi di tipo target, di composti sospetti e non-target devono essere sviluppati al fine di proporre una metodologia di screening avanzata basata sulla spettrometria xx xxxxx. L'attenzione è stata focalizzata sul processo di scale-up nella confidenza di identificazione, sviluppando protocolli analitici specifici.
Due protocolli per la simultanea analisi target e di composti sospetti, basati sulla piattaforma HPLC-Q-TOF, sono stati sviluppati ad applicati xxxx'analisi di cianotossine in acqua dolce destinata al consumo umano e inibitori del PDE-5 negli integratori alimentari venduti come rimedi per la disfunzione erettile. Entrambi i protocolli sono stati ottimizzati con lo scopo di ottenere dai dati mass-spettrometrici in modalità tandemMS, gli ioni pseudomoleculari e lo spettro di frammentazione. Librerie sono state sviluppate e implementate per semplificare il trattamento dei dati.
La possibile applicazione di questi protocolli ne l’analisi di tipo non-target è stata tentata su campioni reali ne l’ambito di una prova collaborativa organizzata dall'associazione Europea XXXXXX, riguardante l'analisi di contaminanti xxxx'acqua. L'esercizio è stato complesso e richiedente molto tempo, e ha messo in evidenza i punti di forza e di debolezza dei protocolli sviluppati.
Il passaggio cruciale xxxx’analisi non-target è l'assegnazione di formule molecolari veritiere ai valori m/z. Un workflow di analisi basato sulla infusione diretta del xxxxxxxx e l'acquisizione xx xxxxx utilizzando un Orbitrap ™ è stato sviluppato e automatizzato utilizzando algoritmi basati sul linguaggio di programmazione Mathematica, per studiare la composizione chimica della frazione organica del PM2.5. Contestualmente, l'eventuale uso della fotoionizzazione a pressione atmosferica (APPI) per la caratterizzazione frazione organica del PM2.5 è stata indagata su campioni reali.
ACN | Acetonitrile |
Adda | 3-amino-9-methoxy-2,6,8-trimethyl-10-phenyl-4,6-decadienoic acid |
ANA-a | Anatoxin-a |
ANP | Anabaenopeptin |
APCI | Atmospheric pressure chemical ionization |
APPI | Atmospheric pressure photoionization |
B(α)P | Benzo[α]pyrene |
BMAA | β-N-methylamino- L-alanine |
BPC | Best peak chromatogram |
CE | Collision energy |
CID | Collision induced dissociation |
CRM | Certified reference material |
CYL | Cylindrospermopsin |
DBE | Double bond equivalent |
EC | Emerging contaminant |
EIC | Extracted ion current |
ELISA | Enzyme- linked immune assay |
ESI | Electrospray |
FA | Formic acid |
GC | Gas chromatography |
HPLC | High-pressure liquid chromatography |
HRMS | High-resolution mass spectrometry |
HWHM | Full width at half maximum |
XX | Xxxxxxxx mass |
KMD | Xxxxxxxx mass defect |
LC | Liquid chromatography |
LOD | Limit of detection and quantification |
LOQ | Limit of quantification |
LTQ | Linear Ion Trap |
m/z | Mass to charge ratio |
MC | Microcystin |
MME | Mass measure error |
MS | Mass spectrometry |
MS/MS | Xxxxxx Mass Spectrometry |
MW | Molecular weight |
NL | Noise level |
NMR | Nuclear magnetic resonance |
NOD | Nodularin |
OSc | Oxidation state of the carbon |
PAH | Polycyclic aromatic compounds |
PDE-5 | Phosphodiesterase type 5 |
PFPA | Pentafluoropropionic acid |
PM | Particulate matter |
POA | Primary organic aerosol |
PTFE | Polytetrafluoroethylene |
Q-TOF | Quadrupole-time-of- flight |
RSD | Relative standard deviation |
RTI | Retention time index |
S/N | Signal to noise ratio |
SD | Standard deviation |
SOA | Secondary organic aerosol |
STX | Saxitoxins |
TFA | Trifluoroacetic acid |
TIC | Total ion current |
XX | Xxx Xxxxxxxx |
VOC | Volatile organic compound |
The main purpose of an analytical chemist is to provide information about chemical constituents in a sample. In the case of quantitative information are request, the question he/she has to pone himself or herself is: “How much analyte is in the sample?”. A good answer obviously has to provide, before the quantitative value, all the unquestionable information indicating that the result refers to the considered analyte. When only qualitative evidences are requested, two different questions are anyway possible: “Is the analyte in the sample?” or “What is present in the sample?”.
The analytical approach to be followed depends on the different concepts under these two questions: while in the first case, a target analysis will be performed, in the second case the commonly defined “screening analysis” can be attempted [1].
The analysis of emerging contaminants is a topic where screening analysis is pivotal.
3.1 Emerging contaminants
With the term “emerging contaminants” (ECs) or “contaminants of emerging concern” we identify substances that are recently taken into account from the scientific community because they represent a potential risk factor for human health or environment. Xxxxxx and Shore 2001 [2]defined ECs as chemicals that have recently been shown to widely occur in water resources. Although adequate data to determine the correlated risk do not yet exist, these are anyway identified as potential toxicants. This definition is limited to water pollution, whilst a proper exhaustive categorization of ECs has to be extended to other compartments or matrices.
Starting from the previous definition, we would like to propose a definition of ECs substances that exhibit a somehow identified factor of risk, or whose hazard will be a factor of risk in the future for human health or environment.
From our definition, may be considered ECs compounds showing:
⮚ Known exposure route + newly identified hazard.
⮚ Known hazard + new exposure route.
⮚ Known hazard + known exposure + increasing in human’s sensitivity.
All the cited factors contribute to the risk associated to these compounds. This definition pays particular attention also on substances that might be a risk in the future. Running examples are far to be rare, e.g. human’s sensitization to a lergens.
Several compounds or class of compounds are currently been recognized as ECs in the water cycle: Perfluorinated compounds (PFCs) [3], brominated compounds and flame- retardants [4], personal care products (PCPs), (pharmaceuticals, hormones, cosmetics and sunscreens), drugs of abuse[2], disinfection by-products, nanomaterials[5], artificial sweeteners [6], together with their correlated transformation products [2, 7]. Other substances of concern are benzotriazoles, used as anticorrosive and for silver protection in dishwashing liquids, the naphthenic acids arising from crude oil, synthetic musk fragrances, prions, which are infectious particles composed of a protein in a misfolded form, and ionic liquids replacing traditional solvents used in industry [8].
Among the already known or potential ECs, three classes of compounds have been specifically taken into consideration in this PhD thesis: cyanotoxins, PM2.5 and a class of pharmaceuticals, the phosphodiesterase-5 inhibitors (PDE-5 inhibitors).
3.1.1 Cyanotoxins
Cyanobacteria are worldwide spread prokaryotic organisms present on Earth since early stages of life. Their role in the evolution of the Earth is of key importance, because these photosynthetic organisms, firstly released oxygen in the atmosphere. Cyanobacteria are ubiquitous present in eutrophic water reservoirs and they are notably associated to the unpleasant odour that may occur in drinking water, due to their ability in producing compounds as 2-methylisoborneol and geosmin [9]. However, the main issue of concern in surface waters is related to the ability of several species of cyanobacteria in producing secondary metabolites, toxic for several organisms including humans [10].
These cyanotoxins could be produced, when the biomass of cyanobacteria grows drastically in a short period, thus causing dense blooms on the water, and often with a remarkably colouring of the water surface e.g. “the red tide” of the Planktothrix rubescens. Although the presence of more than one genotype of cyanobacterium is rare [11], a co-blooming of different species could be experienced. In Italy the incidence of potentially toxic algal blooms is very high, with more than 60 basins interested [12].
Harmful cyanobacterial blooms are regulated by both genetic and environmental factors. Among the latter ones, water temperature is the most important, as many species of cyanobacteria prefers warm water (more than 25 °C). Incidentally, global warming is considered an indirect cause of the increasing occurrence of toxic algal
blooms [13][14]. The water concentration of macronutrients is also important: cyanobacteria efficiently grow in lentic aquatic ecosystems with relatively high concentrations of primary nutrients as nitrogen, phosphorus, and carbon [15], with a known correlation with the nitrogen/phosphorus concentration ratio[16]. The increasing release in the environment of nutrients coming from farming and agricultural activities, together with nutrient accumulation promoted by the long residence times in lakes and reservoirs, feed cyanobacteria blooms [16].The last environmental factor influencing cyanobacterial bloom is the light exposure. Most species can effectively make photosynthesis only in a limited range of light quality, intensity, and duration [9].
The Cyanotoxins are classified emerging contaminants since only few congeners are well known in terms of toxicity and in parallel to the discovery of new congeners, also new exposure routes have been identified.
3.1.2 Cyanotoxins classification
The potential toxicity of cyanobacteria is related to the biosynthesis of the harmful secondary metabolites produced. About 40 of the 150 known phytobacterium genera are able to produce toxins, classified according to their mode of action primarily into hepatotoxins, neurotoxins and skin irritants [9].
3.1.2.1 Hepatotoxic algal metabolites
Microcystins (MCs) and cylindrospermopsin [17-19] are the most diffused hepatotoxins in freshwaters, produced by various species within the genera Microcystis, Anabaena, Oscillatoria, Nodularia, Nostoc, Cylindrospermopsis, and Umezakia. Their occurrence has been reported in Asia, Europe, North Africa, North America and Scandinavian countries. MCs are monocyclic heptapeptides with relatively low molecular weight. More than 100 congeners of MCs with a general structure (-D-Ala1-X2-D-MeAsp3-Z4- Adda5-D-Glu6-Mdha7-) are known. This wide number of congeners is primarily due to the variability in composition of the amino acidic residues in positions X end Z (Figure 1). For example, the most studied MC, which has leucine (initial L), and arginine (initial R) in position 2 and 4 respectively, is identified as MC-LR.
Hepatotoxins show their toxicity in the liver where are quickly concentrated. The effect is dose dependent and mainly related to their interaction with the protein phosphatases (PP1A and PP2A), causing the inhibition of enzymatic activity with cell necrosis
followed by massive haemorrhages and death. These adverse effects seem to be ascribed to the unusual Adda amino acid, almost invariably shared by all MC variants.
Nodularins (NODs) are mainly associate with booms of N. spumigena, which occurrence have been reported Australia, New Zealand and the Baltic Sea [9]. NODs are cyclic pentapeptides structurally similar to MCs, including the Adda moiety but with only one variable amino acid. So far, nine variants have been identified, the most common being NOD-R with arginine as variable amino acid.
Whilst NODs are potential tumour promoters with hepatotoxic toxicity similar to MCs, no human intoxication has been reported and reliable toxicological data are not quoted. Cylindrospermopsin (CYL) was initially described as a tropical toxin because occurred in Australia, New Zealand and Thailand. Anyway, recent reports in temperate areas, such as Italy [19], Germany and France [8] have widened its ecological habitat. CYL was initially named from the algal specie producer, the cyanobacterium Cylindrospermopsis raciborskii, but nowadays Aphanizomenon ovalisporum, Raphidiopsis curvata and Umezakia natans have been described to perform the biosynthesis of this toxin.. CYL is a highly polar tricyclic alkaloid (Figure 1) with a guanidine moiety along with a uracil, which is described as potentially responsible for its toxicity. After ingestion, the toxin mainly affects the liver via the irreversible inhibition of protein synthesis leading to cell death. To date, two other variants have been reported, i.e. 7-epicylindrospermopsin and the non-toxic deoxycylindrospermopsin.
Anabaenopeptins (ANPs) is another class of hepatotoxic algal metabolites. They are
unique cyclic peptides that have the common cyclic peptide moiety linked with Tyr, Arg, Lys, and Phe through an ureido bond. The most representative congeners of this class are the Anabaenopeptin-A and Anabaenopeptin-B, but several congeners are reported in literature. [20].
The microginin FR1 was the first congener of the class isolated from a water bloom of a German lake [21]. Microginin FR1 structure is as linear peptide containing β-amino-α- hydroxy-decanoic acid (Ahda), alanine, N-methyl-leucine, and two tyrosine units (Ahda-Ala-N-Me-Leu-Tyr-Tyr). Microginin FR1 had angiotensin-converting enzyme inhibitory activity. Recently several microginin congeners have been identified and reported [22].
3.1.2.2 Neurotoxic algal metabolites
Anatoxins are well described neurotoxic algal metabolites. Anatoxin-a (ANA-a) and homoanatoxin-a (Figure 1) are the two main neurotoxic alkaloids produced by Anabaena, Aphanizomenon and Planktothrix cyanobacteria, whose occurrence was reported in USA, Africa, Asia and Europe [9, 17, 23].
ANA-a induces paralysis of the organism by interacting with acetylcholine receptors with the consequent death by respiratory arrest. The LD50 in mice is 375 μg/kg 24 h after the intra peritoneal injection. Animal poisoning by ANA-a causes vomit, convulsion and respiratory arrest. ANA-a(s) is the phosphate ester of a cyclic N- hydroxyguanine, with similar toxic behaviour of ANA-a, which has been identified associated to Anabaena strains in restricted areas of United States, Scotland, Denmark and Brazil [9].
The other main class of algal neurotoxins is saxitoxins (STXs), which have been detected in freshwaters of Australia and USA. Saxitoxins are biosynthesized by Anabaena circinalis and Aphanizomenon flos-aquae, but also Lyngbya wollei and C. raciborskii are known to be able to express these compounds. Saxitoxins are tricyclic compounds that can be non-sulphated, singly sulphated or doubly sulphated. These toxins can persist over 90 days in freshwater and can be converted into more toxic variants by high temperatures. As other neurotoxic metabolites, STXs are paralytic shellfish poisons, blocking sodium ion channels in membrane of nerve axons, and finally inducing death due to respiratory failure [17].
β-N-methylamino-L-alanine (BMAA) is a cyanotoxin recently identified in England, Peru South Africa, China and USA. BMAA is a non-protein amino acid acting on glutamate receptors and blocking motor neurons. In addition, BMAA could also cause intra-neuronal protein misfolding associated to neurodegeneration, and some studies connect the exposure to BMMA to the amyotrophic lateral sclerosis. This toxin has been reported to be produced by all known groups of cyanobacteria that possess genes encoding for cysteine synthase-like enzyme and methyl transferase, both involved in the BMAA biosynthesis. Anyway, several toxicological data are considered not reliable, because of possible misidentifications [20].
Figure 1. Widespread hepatotoxic and neurotoxic algal toxins.
3.1.2.3 Cyanotoxins exposure and regulation
Humans could be chronically exposed to cyanotoxins via contaminated drinking water [9, 18, 24] or food, including dietary supplements. Assumption by drinking water could affect a large portion of population of the area served from a contaminated reservoir. Another exposure route is the possible contact through dermal and accidental inhalation/ingestion during recreational activities in waters subjected to a toxic bloom. Last route of exposure is the ingestion of cyanobacteria-based food ingredients or shellfish which previously bio-accumulated toxins through filtration of contaminated water.
The World Health Organization (WHO)[25] recommended a provisional guideline value of 1 µg/L of MC-LR equivalents in drinking water, and regulatory values were currently set in several countries specifically for MC-LR or MC-LR equivalents, anatoxins, CYL (0.1 to 15 μg/L) and saxitoxins (3 μg/L). Italy together with France and Turkey will be one of the first European countries to adopt a regulatory value of 1.0 µg/L for total content (intracellular +extracellular) of MCs in drinking water, intended as sum of all congeners that can be quantified. This very conservative approach with regard to health protection was inspired by a case study [12], and by a recent evaluation of the relative protein phosphatase (PP) inhibitory ability of several MCs variants compared to the MC-LR congener. Toxicological information relative to other cyanobacterial oligopeptides is not yet reliable, and no indication is currently available from WHO about their risk assessment.
New Zealand is the only country regulating simultaneously MCs, NOD, CYL, ANA-a, homoANA-a, ANA-a(s) and saxitoxins, due to the large number of case reports described in this country. However, emerging cyanotoxins like BMAA, aplysiatoxins and lyngbyatoxins are not considered, probably because the lack of data did not allow the calculation of a guideline [9].
3.1.3 PDE-5 inhibitors in food supplements
Food supplements and herbal remedies for the treatment of erectile dysfunction and for increasing sexual performance are getting from year to year more widespread [26, 27].
Various factors are responsible for the increased demand of these products:1) certainly consumers perceive natural products much safer and healthier than drugs containing synthetic active ingredients; 2) these products are available also without prescription outside the official health system, e.g. in herbalist’s shop, sex-shops and online market;
3) these products are often cheaper than official drugs [26]. As the Figure 2 shows, sexual performance enhancement remedies represent the most counterfeit products recognized on the market.
PDE-5 Inhibitors have a known toxicity but the new exposure route represented by the adulteration of food supplement classifies them as ECs.
Figure 3. a) sildenafil; b) homosildenafil; c) hydroxyhomosildenafil.
Sildenafil thio-derivatives have been synthesized by heating with P2S5, and have the same possible variety in analogues of their oxo-counterparts, with respect to whose they are described as more powerful.
Tadalafil (Figure 4-a) is known for the shorter synthesis and the advantage to have a different pharmacokinetic with respect to sildenafil. In fact, it exhibits a much longer time window (36 hours) than sildenafil (4 hours). Despite these advantages, analogues of tadalafil have been rarely found, mainly due to the availability of the starting reagents, requiring piperonal that is also used in the preparation of amphetamines, whose commercialization is strictly controlled.
Figure 4. a) tadalafil; b) nortadalafil; c) amino-tadalafil; d) butyl-tadalafil.
The Figure 4 shows the most reported analogues of tadalafil; the major route of modification of the structure is on the N-atom of the amide. This part of the structure is non-essential and the modification do not modify the action mode.
Figure 5 reports the most common vardenafil analogues used in adulteration of food supplements, although few compounds have been reported, probably due to the fact that no important pharmacological advantages are described.
Figure 5. a) vardenafil; b) norneovardenafil, c) N-desethylvardenafil, d) pseudo-vardenafil.
Other often-used unapproved active principle is the yohimbine (Figure 6) a natural tryptamine alkaloid, which can be extracted from the bark of a variety of plants mostly of African and Asian origin such as Pausinystalia Yohimbine. Yohimbine hydrochloride
is rapidly absorbed and the maximum plasma concentration is generally achieved in less than one hour after oral administration.
Figure 6. Yohimbine structure.
3.1.3.1 Health risk of counterfeit food supplements and normative
The possible adulteration of natural products with synthetic PDE5-Inhibitors and analogues [28-31][32-35][32-35] may be representing a risk for unaware customers [36][28][28], since PDE-5 Inhibitors show notable adverse effect such as headache, facial flushing, dyspepsia, visual disturbances and muscle pain [37][29][29]. Furthermore, analogues of the authorized PDE-5 inhibitors may have different and unknown side effect and pharmacokinetics. The health risk factor related to adulterated food supplements are mainly due to: a) the uncontrolled concentration, with cases with more than 170% of the normal dosage; b) the undefined toxicology of these compounds caused from variations of the pharmacokinetics and of the drug metabolism [26]; c) the unaware assumption of such drugs from consumers affected by incompatible sickness, as heart disease.
EU Directive 2002/46 concerning food supplements [38] allows the use of some vitamins and minerals, whilst presence of synthetic active compounds is forbidden. Therefore, any adulteration or cross contamination with PDE-5 inhibitors must be considered not in compliance with the enforced law.
3.1.4 Aerosols
The term aerosol identifies solid or liquid suspension in air, within the dimension range of 10-9-10-4m. Usually, in atmospheric science the term aerosols is limited only to the solid particulate suspended in the air and not to liquid, which is treated separately in the cloud science. A recent OECD report [39] forecasts that air pollution will be in the 2050 the primary environmentally cause of premature death worldwide, and identifying the increasing of the particulate matter (PM) concentration as the major contributor.
Although the link between the exposure to PM and adverse effects on human health and ecosystem is well known [40, 41], the chemistry and physic of airborne aerosol are still poorly understood; mainly because of the extreme complexity of the particulate composition and the absence of the analytical technology able to describe the whole chemistry present [42]. Aerosols are classified as emerging contaminates mainly for the increasing in the human sensitivity.
3.1.4.1 Sources, composition and size distribution of aerosols
While the 99.9% of the atmospheric content is related to the gases N2, O2, H2O and Ar, molecules present at trace levels, such as NO2, O3 SO2, *OH, *O2H and non-methane hydrocarbons, actually drive the atmospheric chemistry. Many different factors as gas- phase reaction, temperatures, etc. could condense low volatile species, thus forming aerosol [43, 44]. The average amount of PM in the atmosphere is in the order of 1 µg/m3 or 10-7% by mass, with a quite wide geographical and seasonal variability [45].
PM in atmosphere can be classified in many different ways taking into account different characteristic. Considering the origin of the aerosol, it can be divided in anthropogenic or biogenic. The first regards all the aerosols arising directly or indirectly from human activities and emission as for example fuel burning, while biogenic aerosols are originated from natural events as pollen, plant emission, and volcanoes emissions.
A PM mass fraction between 20-90% is represented by organic matter [46]. Organic aerosols can be emitted directly (primary organic aerosol, POA) or formed in the atmosphere through gas-to-particle conversion processes of volatile organic compounds (VOCs), thus generating what is known as secondary organic aerosol (SOA).
The aerosol composition is extremely complex because of the high chemical heterogeneity of the emissions (fuel and biomass burning) and reactions in atmosphere; a single VOC, interested by a series of complex reactions involving even oxidation by free radicals, can give thousands of different products. Furthermore, reactions inside the particles, promoted by water repartition and photochemical processes, provoke an “aging” of the aerosol with a consequent change of its composition [43].
The size of airborne matter is strongly dependent by its formation and cleaning process. There are three main modes of particulate size. The coarse mode, at high value of diameter (10 µm), is typical for the particles mechanically formed and sedimentation is the associated removal process. At the opposite in the diameter scale we find the nucleation mode, represented by ultrafine particles (<0.1µm) formed via homogenous
nucleation. These particles are generally lost by coagulation of them into the accumulation mode. Finally, the accumulation mode is described by values of diameter typically lower than 1µm (fine particles) and includes particles formed via nucleation and condensation from gas phase. Fine particles show a lifetime greater than the other two modes, and the principal removal processes are by rainout and washout.
Traditionally, in the contest of the effects of aerosol on the human health, another metric is used to describe the size of the airborne matter based on an operative concept: it is classified in PM10, PM2.5 and PM1, referring to the particles fraction having an aerodynamic diameter lower than 10, 2.5 and 1 µm respectively.
3.1.4.2 Health and environmental effects of particulate matter
The correlation between atmospheric aerosols and human health is well established and supported by historical events and many epidemiologic studies [47]. One of the first and well documented case clearly evidencing the connection between an increased death rate and urban smog was the “great smog” occurred in 1952 in London, when an extremely high concentration of PM arising from coal burning killed an estimated number of 12,000 people [48]. More recently, episodes of PM spikes with concentrations up to 10 times greater than those reported to cause adverse health effects have been studied in Calexico/Mexicali [49]and Beijing [50][49]. PM2.5 showed to have the heaviest impact on human health. In fact these particles can enter into the respiratory system and reach the alveoli [51]. Particles between approximately 5 and 10 μm are most likely deposited at the tracheobronchial level, while those between 1 and 5 μm are deposited at the respiratory bronchioles and the alveoli where gas exchange occurs [51]. The main human health adverse effects related to PM include premature mortality and high morbidity, asthma, cardiovascular and nervous diseases [52]. The PM toxicity shows different mechanisms: the oxidative stress is one of the major pathway affecting the respiratory system [52], involving an increased concentration of reactive oxygen species (ROS) such as superoxide radical (O2*-), hydroxyl radical (HO*), hydrogen peroxide and other organic hydro peroxide. The ROS increment is mainly caused by the interaction with respiratory system lining fluid, but also metals and other oxygenated species contribute to this alteration.
Exposure to PM2.5 is also correlated to the increased incidence of lung cancer, and it is estimated than 5% of lung cancer deaths are attributable to PM [53]. Carcinogenicity and mutagenicity of PM2.5 is primarily connected to the contextual presence of certain
classes of compounds as polycyclic aromatic compounds (PAHs), coming from vehicles and biomass burning, and a number of nitrogen-containing organic compounds (NOCs) that can form carcinogenic metabolites in the body [54, 55].
In addition to the health effect, PM causes damage to ecosystems, cultural heritages, reduces visibility and it is known to be linked to climate change [40], mainly acting on the radiation balance and on absorption phenomena, warming the atmosphere or scattering radiation with a cooling effect. Even more, PM strongly affects the cloud behaviours as it acts, for its hydrophilic characteristics, as cloud condensation nuclei. [56]. PM is involved in heterogeneous reactions promoted by the solar radiation and affecting the composition of the trace compounds present in the atmosphere. Deposition of PM containing black carbon on the snow or on the surface of glacier, promotes melting phenomena by absorption of radiation, thus changing the hydrogeological cycle.
3.2 Screening analysis of emerging contaminants
In literature, many approaches for screening analysis of emerging compounds are reported. Sampling and sample preparation represent in almost all the detection approaches the preliminary stages of the analysis [9].
Several biological or biochemical methods are currently used for screening a large variety of pollutants, since interactions with the animal metabolism both at micro (enzymes and proteins) and macro (physiological effects) levels represent the main concern about the toxic effect. Historically, mouse bioassay was the first in vivo test implemented for pollutants detection, and it is already used in some official methods tailored for toxins analysis in food. This method not allows the identification of the compound responsible of the observed toxicity, has low sensitivity and ethical issues respect the use of animals [57].
In ELISA (Enzyme-Linked Immune Assay) assays, analytes are detected through binding to specific antibodies. Many ELISA kits are commercially available for an extremely wide variety of compounds. Despite the performance in sensitivity ELISA tests have some limitations regarding the selectivity among different variants of the same class of compounds, and the identification capacity by the fact it not provide any structural information [58].
Among the chemo-physical techniques, vibrational spectroscopy, such as infrared (IR), near infrared (NIR) [59]or Raman spectroscopy [60]or their combinations have been used for screening known contaminants, e.g. for counterfeit pharmaceutical products, and for identifying new pollutants. These techniques when used in conjunction with a chemometric approach generate typical fingerprints that help to differentiate between authentic and fake or counterfeit samples. The great advantages of this simple and non- destructive technique are limited by the reduced selectivity and sensitivity.
NMR spectroscopy is one of the most powerful tool to unambiguously elucidate the structure of known and novel compounds [61]. This technique has the advantage of an easy sample preparation and high reproducibility. The technological advances in the field of magnetic resonance have dramatically improved in terms of the sensitivity and identification capabilities by developing new NMR experiments and data processing tools. Notwithstanding NMR can be used for quantitation purpose, it often requires a quite large amount of sample, due to the low sensitivity for the trace analysis that several emerging contaminants require [61].
Mass spectrometry (MS) became increasingly common over the last decades due to its high sensitivity and selectivity. Tandem MS (MS/MS) approach allows simultaneous detection of a larger amount of analytes with increasing easier sample preparation procedure. When coupled to chromatography, MS detection offers incomparable performance for trace analysis of organic compounds. Liquid chromatography (LC), usually with a reversed phase C18 column and methanol/water or water/acetonitrile as a mobile phase, is likely the most common separation method for the analysis of polar emerging contaminants, whilst GC has been used as a separation method for volatile and semi volatile pollutants. A large number of confirmatory methods based on mass spectrometric (MS) detection were developed for the determination of contaminants in environment, food and biological matrices [62-65]. Although emerging contaminants present a large number of different compounds, for only few of them are currently available analytical standards. This fact, together with the lack of an enforced regulation for some potential pollutants, or even toxicological and structural information, have recently moved the conventional confirmatory analysis of contaminants to on non-target approach. These MS methods are based on full-scan acquisition with GC or LC coupled to high resolution (HR) MS, mainly LC-quadrupole-Time of Flight (Q-TOF) and LC- Orbitrap™ [66, 67]. LC-HRMS full scan methods do not need a compound-specific tuning, and are prone to perform non-“a priori” post-run data analysis, mainly based on
mass accuracy [68, 69]. Anyway, well-known difficulties associated with the structural elucidation, a limited availability or reliability of mass spectral libraries and software [70, 71] for non-target and post-target analysis, represent important hindrances for a widespread application of LC-HRMS techniques for identification purposes [71]. HPLC-Q-TOF [72] or HPLC-Orbitrap™ systems have been used to perform multi- residue analysis, with often more than 100 target compounds and non-target analysis.
Emerging contaminants, such as cyanotoxins can also be detected by MS without preliminary chromatographic separation. For example, MALDI-TOF instruments can be used to perform toxin analysis in very small sample volume such as cell colonies [73]. Despite the rapidity and the possibility to avoid sample preparation, this approach is not suitable for all the matrices of interest, like water, and it is poor in sensitivity and selectivity, since no information about retention times are provided to differentiate compounds with a similar MS behaviour.
3.2.1 High-resolution mass spectrometry (HRMS) in characterization of emerging contaminants.
In the paradigm of screening analysis in mass spectrometry, as represented in Figure 7, we can classify three main approaches toward the substances identification by HRMS[74]:
⮚ Target analysis: This approach is focused on the confirmation and quantitation of a limited number of compounds. The number of compounds determinable depends on the specific detection system and experimental design. Target screening requests the use of analytical standards and mass spectrometry has to be coupled with a chromatographic separation step.
⮚ Suspect screening: this approach is performed on a relatively large number of selected compounds, whose presence in the sample is supposed, even if the corresponding certified standards are not necessary available. Usually, information about structures of the suspected compounds is included in a database in order to simplify data analysis. Through this screening approach is possible to reach identification or confirmation of substances when certified standards are available.
⮚ Non-target screening: this approach is followed when no structural information is available a priori and the analysis is virtually performed on all substances detectable in the sample analysed.
The use of standards, specific databases and software utilities are necessary depending on the screening approach followed.
3.2.2 Degree of identification
A consideration regarding the harmonization of the definitions in “identification” and “confirmation” has to be done, because the scientific community often uses them improperly. When acceptable evidences support the structure of an analyte, we can use the term “identification”, while a comparison with an analytical standard through orthogonal method has to be provided in order to have “confirmation” [75].
In literature, and generally in the field of environmental analysis, the identification of analytes by LC-MS mainly refers on the European guideline for the confirmation of veterinary drugs in food of animal origin, the EU Guideline 2002/657/EC [76]. In this guideline, the concept of identification points (IPs) is proposed for confirmatory methods that anyway require a comparison with analytical standards. The identification of a target analyte has to comply with the following constrictions:
⮚ Reference standard matching: analytical standards should be analysed contextually to the sample analysis. All signals and parameters related to the suspected compounds (e.g. retention time and mass spectra) have to match with
those obtained for the analytical standard in the same experimental conditions, with acceptable bias. The use of reference materials is encouraged, and if not available, the use of spiked blank matrix is desirable instead of a pure standard. When the matrix affects performance of the separation or detection system, the use of pure standard has to be avoided.
⮚ Chromatography: the chromatographic peak of a positive sample should exceed a signal to noise ratio (S/N) threshold of 3:1, and the retention time have to match with the peak of the corresponding standard with a maximum relative standard deviation of 2.5% for LC and 0.5% for gas chromatography.
⮚ Mass Spectral matching: the mass spectrum should include a number of signals satisfying the IPs requested, at least three. The number of mass signals requested depends on the MS system used, basically on low or high resolution mass analysers, and on single stage or tandem MS. Thus, the MS fingerprint has to match that obtained for the reference standard, also satisfying specifications on the S/N, and relative intensities.
In the framework of confirmatory analysis, the chromatographic separation is absolutely necessary for identification purpose, because it indirectly provides further information on substance structures, thus allowing a unique identification when standards are available. However, the retention time parameter could be useful even in absence of certified standards, if a suitable normalization is adopted. As example, retention time index (RTI) or Xxxxxx index could be used as further descriptor for databases in order to improve the reliability of an identification process.
Finally, when HRMS systems are employed, a separation system could be not necessary, if the aim of the analytical protocol is a qualitative characterization of the organic composition based on the raw formulas inferred by experimental measures.
Level 1. Confirmed structure. This level of identification confidence corresponds to the confirmatory analysis previously described. In order to provide a unique structural identification, orthogonal selective methods have to be used, and data obtained have to be compared with the corresponding ones obtained from analysis of certified standards. Mass spectrometry has to be necessarily coupled with a chromatographic separation step.
Level 2. Probable structure. An adequate amount of evidences, comprising MS/MS diagnostic fragments, indicates a unique structure of the analyte. The main difference with the level 1 is the absence of the comparison with a standard. This degree of confidence can arise from matching spectra with MS libraries or literature.
Level 3. Tentative candidates. In this level of confidence, structural information is significant but not sufficient to ensure a unique structure. MS/MS fragmentation is available, giving some indications for example, of the chemical class of compounds.
Level 4. Unequivocal molecular formula. In this case, a molecular formula can be unambiguously assigned by the MS data. No significant MS/MS spectra are recorded.
Level 5 Mass of interest. It is the lowest level of identification confidence, and it is limited to the knowledge of the m/z of the quasimolecular ion. In level 5 all the compounds are classified as “unknown”.
For both suspect screening and non-target analysis, it is possible reach the highest level of identification, gaining informationally. Incidentally, while the use of chromatography is mandatory for the confirmation, it is not necessary in the case of the lower level of identification.
3.2.3 Identification criteria in HRMS
3.2.3.1 Molecular and quasimolecular ion
When soft ionization are employed, molecular or quasimolecular ion is considered the MS signal giving the most important information in target and suspect screening analysis, as from its m/z value is possible the direct determination of the molecular formula. Molecular or quasimolecular ions have to be optimized in term of sensitivity and trueness. The last parameter results pivotal for the determination of unknown compounds, when HRMS is employed for screening purpose. Two parameters characterize the measure of the MS trueness: resolving power and mass accuracy. Resolving power or resolution (R) is the ability of a mass analyser to resolve
neighbouring signal. It is defined as m/Δm where m is the nominal mass of a given molecule and Δm is the difference in mass that the instrument is able to discriminate. This definition is nowadays subjects to many critical revisions in particular about how it is defined. An alternative definition, used even in the 2002/657/EC guideline is the full width at half maximum (HWHM).
The instrumental mass accuracy refers to the degree of closeness of a measured m/z to its true theoretical value. Accuracy better than 2 ppm and 5 ppm are nowadays quite common for the new generations of Orbitrap™ and Q-TOF mass analysers respectively and sub-ppm accuracy could be archive by FTICR.
The accurate mass of both molecular (and/or quasimolecular) and fragment ions is the parameter used in the libraries and online database in order to find the corresponding structure of possible unknown compounds. In the Table 1 the main libraries and databases available in MS are reported. While NIST, PubChem, ChemSpider and Wiley libraries can be considered general libraries of greater value, other databases as NIST MS2, MassBank, METLIN, mzCloud, are becoming powerful tools for identification in HRMS because a quite large amount of MS/MS xxxxxxx is currently available, with good perspectives of improvements.
Database or library name | total compounds present | Entries with MS xxxxxxx |
ChemSpider | 32000000 | - |
DAIOS | 1404 | >1000 |
PubChem | 68479719 | - |
STOFF-IDENT | 7864 | - |
MassBank MS/MS | 3350 | 3350 |
mzCloud | 2510 | 1956 |
NIST EI-MS | 242477 | 212961 |
NIST MS/MS | 8171 | 4628 |
Wiley Registry of Mass Spectral Data (EI-MS) | 638000 | 289000 |
Xxxxxxx Xxxxxxxx, Xxxxx & Pragst Toxicology/Forensicsf | 8998 | |
Agilent METLIN Pesticide Library | 1664 | |
Agilent METLIN Synthetic Substance Libraryg | 64092 | |
Agilent METLIN Veterinary Drug Library | 1684 | |
Bruker ToxScreener (incl. Pesticide Screener) | 1753 | |
Sciex / AB Sciex LC/MS/MS Meta Library | 2381 | |
Thermo Environmental Food Safety (EFS) with/without retention | 732 | |
Thermo toxicology | 654 | |
Waters database | 730 |
3.2.3.2 Isotopic pattern
The isotopic abundance can provide additional information on elemental formulas. The relative abundances of the various isotopomers are helpful for prioritizing and reducing the number of the possible formulas assigned, and this approach is implemented on many software tools for MS data elaboration, giving a weight at the final score on the basis of the isotope ratios and mass defects of isotopes [77].
Due to the low intensity of the isotopomer ions, high sensitivity and a low noise with few chemical and background interferences are required. Obviously, some chemical elements show a peculiar isotopic abundance, e.g. chlorine, bromine, silicon sulphur and several metals. Thus, such MS spectra result in a typical fingerprint that allows
some consideration about the presence of certain element, empowering a possible identification.
3.2.3.3 Fragmentation
MS/MS is a powerful technique ensuring information that increase the reliability of the molecular formula assignment and provide structural information for the identification. Official guidelines assign IPs for each fragment present matching with the fragmentation pattern of the corresponding pure standard. In addition to this approach, when commercial standards are not available, the matching with MS/MS spectra collected in libraries can be considered to strengthen the identification.
Instrumental approaches for acquiring fragmentation spectra can be different and dependent on the apparatus available and on the aim of the analysis performed. A recent trend in non-target analysis is the collection of MS/MS xxxxxxx of as many analytes as possible using data-dependent acquisition features. In other instrumental approach, the so-called MSE ™ involving the simultaneous acquisition of accurate mass data at low and high collision energy can provide this information.
In silico fragmentation is an alternative recent approach for structure identification. Mass Frontier and ACD/MS Xxxxxxxxxx are the two most popular MS fragmentation predictor’s rule-based programs. They are useful to predict MS/MS spectra of compounds when no reference mass spectra are present. Despite the potential capabilities, in-silico fragmentation has been proved to be efficient for only 56% of mass peaks, and often fails with analytes at trace level in noisy spectra. However, in- silico fragmentation remains the principal approach in protein identification.
3.3 HRMS data visualization tools
When HRMS is used in non-target analysis generates an extremely large amount of data as list of assigned molecular formulas [78]. The interpretation of any single compound is extremely time consuming and useless. Some commercial software’s provide tools for statistical evaluation of data, using all the possible parameters, like retention time, molecular and fragment ions, signal abundance, for blank subtraction and classification of results.
In the framework of the categorization of aerosol visualization methods are generally necessary for data evaluation. The most diffused visualization methods include the use of double bond equivalent (DBE), van Krevelen diagrams, carbon oxidation state and Xxxxxxxx mass analysis [79, 80].
DBE is an indicator of the hydrogen deficiency and represent the number of the double bonds and rings in a molecule structure [79]. For molecules with general formula CcHhOoNn, DBE can be calculated through the general formula:
DBE= c-0.5h +0.5n+1
Where c, h and n are the number of atoms of carbon, hydrogen and nitrogen respectively. This value is a useful tool to eliminate the molecular formula showing an unreasonable high number of unsaturations and high double rings. Usually, the data are visualized as DBE against number of carbon or m/z ratio. Anthropogenic emission are characterized by a high grade of unsaturations and values of DBE (>5) typical of aromatic hydrocarbons and their oxidized derives. DBE plots can provide useful information about the sources and precursors of aerosols.
Xxx Xxxxxxxx (VK) diagrams were initially developed for the study of coalification processes and then extended to the particulate matter [81]. In these diagrams, H/C ratio of formulas are plotted versus O/C ratios. VK diagrams are particular useful to classify aerosol samples by identification of the compound classes present on it. Different classes of compounds occupy different regions of the plot. Highly unsaturated compound (PAHs and derivatives) have low H/C and O/C ratios and lie close to the axes origin; aliphatic hydrocarbons (lipids as example) have again low O/C ratio but high H/C ratio and occupy a region above the previous ones. Highly functionalized compound are characterized by a high value of O/C ratio and the typical region of the
plot for those compound is the right side. Moreover, those plots are often used in three- dimension with signal intensity or the DBE value in the z-axis in order to maximize the information. Anyway, signal intensity is a parameter that has to be carefully evaluated when the direct infusion of sample is used for the analysis, as the direct correlation with compounds concentration is not possible.
Another useful parameter to characterize the extreme chemical complexity of organic compounds present in atmospheric aerosols is the oxidation state of carbons (OSc) Oxidative reaction plays a central role in the atmospheric chemistry and are involved in the removal of pollutants, O3 formation and SOA production. In fact, as reported in literature SOA is mainly formed by organic compound arising from the oxidation of gas-phase species. Thus, the oxidation state of carbons could be a metric to characterize SOA, studying its formation and temporal evolution. The trend in the atmospheric environment is the increasing of the oxidation state of carbon, through bonds formation with oxygen and breakage of hydrogen carbon bonds. OSc can be calculated taking in to account the oxidation state of each heteroatom present in the structure and applying the follow formula:
𝑂𝑆𝑐 = − ∑ 𝑂𝑆
𝑛𝑖
𝑐
𝑖 𝑛
𝑖
Where OSi is the oxidation state of i-element and ni/nc is the molar ratio of element i to carbon. The formula is accurate when the structure of the compound is exactly known [78, 82]. When it is applied on molecular formula it could be inaccurate if in the formula are present elements showing different oxidation states possible depending of the molecular structure. As example nitrogen in organic compound can be present as N (-3) in ammines, N (-1) in amine oxide compounds, N (+1) in nitrous-compounds and N (+3) in nitro-compounds. A similar variability is also showed by sulphur that can be present in the oxidation states -2 (thiols and sulphides), 0 (sulfoxides), +2 (sulfinic acid compounds), +4 (sulfonic acid compounds). This variability in oxidation states make difficult to predict OSc from formulas containing nitrogen and sulphur and its use should be avoided. In this study, OSc is calculated only for the species CcHhOo using the simplified formula [78]:
𝑂𝑆𝑐 = 𝑛ℎ − 2 𝑛𝑜
𝑛𝑐 𝑛𝑐
Despite the fact that oxygen shows multiple valence state, the (-2) is much more stable than the others and then the sporadic presence of formulas containing oxygen with different oxidation state does not significatively affect the overall analysis.
Xxxxxxxx mass defect (KMD) analysis is based on the fact that different nucleotides show different defects of mass from the nearest integer mass [83]. Therefore, different elemental compositions showing same integer mass have different exact mass. The addition or subtraction of two hydrogen atoms from a molecule means increase or decrease the number of unsaturations present on it by 1. In the same way addiction of a CH2 group, do not affect the unsaturation number but involve in a shift of mass and defect of mass (14.01565). In KMD analysis the masses of molecular formulas CcHhOoSsNn are rescaled converting the exact mass of CH2 to the nominal one applying the formula
𝐾𝑀 = 𝐼𝑈𝑃𝐴𝐶 𝑚𝑎𝑠𝑠 × (14⁄14.01565)
In these condition formulas, differing only by the number of CH2 will have identical mass defects. The KMD could be calculated with the formula:
𝐾𝑀𝐷 = 𝑛𝑜𝑚𝑖𝑛𝑎𝑙 𝑚𝑎𝑠𝑠 − Xxxxxxxx 𝑚𝑎𝑠𝑠
In the visualization usually KMD are plotted on y-scale versus Xxxxxxxx mass on x- scale. Molecular formulas differing by number of CH2 are displayed in horizontal series spaced by 14 Xxxxxxxx mass units. Instead, different homologue series are plotted at different KMD values and compound classes differing by 2H fall in lines spaced by 2.01564 * (14/14.01565) - 2 = 0.01340. KMD analysis is particularly useful to identify classes of compounds forming series differing by number of CH2 [83].
4 Aim of the project and structure of the thesis
In this PhD thesis, the capability of HRMS in screening analysis will be investigated for the study of emerging contaminants in environmental matrices and foodstuff. In details, we would focus the attention on the process of scale-up in the identification confidence by developing of different specific protocols.
Advantages in the screening analysis must be driven by the availability of analytical standards and the implementation of useful databases and software. These features will be treated taking into account the identification level we want obtain. Chapter “Development of a target and suspect two-step protocol for analysis of freshwater cyanotoxins by liquid chromatography-Q-TOF system” and chapter “and chapter “One- shot” analysis of PDE-5 inhibitors and analogues in natural products for the treatment of erectile dysfunction” will treat the highest levels of identification confidence (level 1 and level 2) and reports the developed methods to perform simultaneously the suspect and target screening.
Chapter “HRMS analysis of organic fraction in PM2.5: Post-run data analysis workflow and the role of ionization sources” will treating the low identification confidence levels (level 5, level 4) with emphasis on the crucial step of molecular formula assignment to m/z values.
Here below, the aims of the single chapters are introduced.
4.1 Development of a two-step method for target and suspect analysis of freshwater cyanotoxins by LC/Q-TOF system.
Occurrence of cyanotoxins in waters intended for human consumption and foodstuff requires a reliable analytical strategy able to support rapid decisions. Moreover, health agencies usually have to opt for a limitation of the drinking water distribution, without a clear legal orientation or a well-established risk assessment about the presence of MCs variants other than MC-LR, or other potentially toxic oligopeptides. Thus, specific
confirmatory analysis aimed to furnish as many information as possible on cyanotoxins risk management are valuable.
To our best knowledge, nor rational LC/HRMS-based analytical protocols neither specific databases devoted to determine standardless cyanotoxins in freshwater have been reported. Thus, our effort was focused in developing a reliable and rapid strategy, useful to risk assessment related to cyanotoxins.
4.2 Development of “one-shot” analysis of PDE-5 inhibitors and analogues in natural products for the treatment of erectile dysfunction.
This study started form a collaboration between our research group, the Italian agency for drugs (agenzia Italiana del farmaco, AIFA), the Ministry of Health, the National Institute of Health (Istituto Superiore xxxxx Xxxxxx; ISS) and the Department of Scientific investigations of the Police. The aim of the collaboration was the monitoring of commercial food supplements sold for the treatment of erectile dysfunction and the verification of the compliance of these products with the legal requirements. For this purpose, an analytical method for determining the presence of PDE-5 inhibitors and analogues able to perform simultaneously target and suspect screening of compounds with such pharmaceutical potential was needed. Seven compounds (yohimbine, sildenafil, vardenafil, tadanafil, homosildenafil, pseudovardenafil and hydroxyhomovardenafil) have been selected among PDE-5 inhibitors for the target analysis.
4.3 Development of a workflow for HRMS analysis of PM2.5 organic fraction: post-run data analysis and the role of ionization sources.
These activities were mainly carried out in the frame of six months of research conducted during the PhD at the Cambridge University, under the supervision of xxxx. Xxxxxx Xxxxxxxx. A method of analysis of organic fraction of PM2.5 based on a methanol extraction of samples and nanoESI-MS determination was previously
developed in the Xxxxxxxx’x group [84]. Two different algorithms were used for the data elaboration by using nanoESI in positive or negative acquisition mode. However, these protocols were time-consuming and affected by the presence of false positive results.
The purpose of the research activities was twofold:
⮚ the evaluation of the Atmospheric Pressure Photoionization (APPI) source for performing analysis of the PM2.5 organic fraction, which could be well ionized using this technique, since it is known to have highly unsaturated compounds, like polycyclic aromatic hydrocarbons (PAHs), nitro and oxidized PAHs.
⮚ the optimization of the algorithms used for the data treatment in order to consider different signals arising from the ionization sources, and to reduce both false positive and the elaboration time.
5.1 Reagents and samples
MC-RR, MC-YR, MC-LR, MC-LA, MC-LW, MC-LF, MC-LY, MC-HtyR, MC-HilR,
MC-WR [D-Asp3]-MC-RR, [D-Asp3]-MC-LR , and nodularin (NOD), used as internal standard (IS), were purchased from Alexis® Biochemicals.
Stock solutions of the twelve MCs and IS, were prepared by dissolving each compound with at least 2 mL of methanol. Subsequent dilutions to obtain working standard solutions were obtained by suitable diluting stock solutions with mobile phases. All solvents and chemicals were of analytical grade (Sigma Xxxxxxx) and all standard solutions and water samples were stored at -18°C in the dark to minimize analyte degradation.
Analytical standards (HPLC grade) of Yohimbine, sildenafil, Vardenafil, Tadalafil, Homosildenafil, Pseudovardenafil and Hydroxyhomovardenafil were purchased by LGC Standards . Formic acid (FA, reagent grade, min 99%), trifluoroacetic acid (TFA, reagent grade, min 99%), pentafluoropropionic acid (PFPA, reagent grade, min 99%), acetonitrile and methanol, both of LC-MS grade, were purchased from Sigma-Xxxxxxx. Water was purified using a Milli-Q Water System (Millipore) to 18 MΩcm. Homosildenafil, Vardenafil, Hydroxyhomosildenafil, Pseudovardenafil e Tadalafil stock solutions were prepared at concentration of 1.0 mg/mL in acetonitrile 0.1% FA, while Yohimbine e sildenafil stock solutions were prepared at the same concentration in a water: acetonitrile 50:50 (v/v) 0.1% FA solution.
Table 2) in the mixture is approximately 10μM dissolved in ACN:H2O 50:50 (v/v).The solution was stored at 4 °C in the fridge.
A standard mixture of endocrine disruptors containing the compounds reported in the Table 3 was used in the confirmation of some positive target compounds in XXXXXX trial. The stock solutions were in methanol:H2O 50:50 and stored al-18°C. Further dilutions were obtained using the same solvent.
Table 2. RTI standard mixture constituents.
CAS | Standard | monoisotopic MW [Da] | logP |
657-24-9 | Metformin | 129.1014 | -1.36 |
1698-60-8 | Chloridazon | 221.0355 | 1.11 |
00000-00-0 | Carbetamide | 236.1160 | 1.65 |
150-68-5 | Monuron | 198.0559 | 1.93 |
3060-89-7 | Metobromuron | 258.0004 | 2.24 |
00000-00-0 | Chlorbromuron | 291.9614 | 2.85 |
125116-23-6 | Metconazole | 319.1451 | 3.59 |
333-41-5 | Diazinon | 304.1010 | 4.19 |
124495-18-7 | Quinoxyfen | 306.9966 | 4.98 |
00000-00-0 | Fenofibrate | 360.1128 | 5.28 |
Table 3. Endocrine disruptors’ analytical standard and stock solution concentration.
Compound name | Abbreviation | molecular formula | [M-H]- m/z | Stock solution (mg/L) |
4-N-octylphenol | n-OP | C14H22O | 205.1598 | 1000 |
4-tert-Octylphenol | t-OP | C14H22O | 205.1598 | 1000 |
Nonylphenol | NP | C15H24O | 219.1754 | 1000 |
Bisphenol A | BPA | C15H16O2 | 227.1078 | 1000 |
Bisphenol A deuterate | BPA-D | C15D16O2 | 243.1329 | 1000 |
Perfluoro-n-pentanoic acid | PFPeA | C5HF9O2 | 262.976 | 1000 |
Estrone | E | C18H22O2 | 269.1547 | 1000 |
β-Estradiol | b-Eol | C18H24O2 | 271.1704 | 1000 |
β-Estradiol deuterate | b-Eol-d | C18D3H21O2 | 274.1852 | 1000 |
17α-ethinyl estradiol | a-Etinil E | C20H24O2 | 295.1704 | 1000 |
Perfluorobutanesulfonic acid | PFBS | C4HF9O3S | 298.943 | 1000 |
Perfluorohexanoic Acid | PFHxA | C6HF11O2 | 312.9728 | 1000 |
Perfluoroheptanoic Acid | PFHpA | C7HF13O2 | 362.9696 | 1000 |
perfluorohexane sulfonate | PFHxS | C6F13O3S | 398.9366 | 1000 |
Perfluorooctanoic acid | PFOA | C8HF15O2 | 412.9664 | 1000 |
Perfluoro-n-decanoic acid | PFNA | C9HF17O2 | 462.9632 | 1000 |
Perfluorooctanesulfonic acid | PFOS | C8HF17O3S | 498.9302 | 1000 |
Perfluoro-n-decanoic acid | PFDA | C10HF19O2 | 512.96 | 1000 |
Perfluoro-n-undecanoic acid | PFUnDA | C11HF21O2 | 562.9569 | 1000 |
Perfluoro-n-dodecanoic acid | PFDodA | C12HF23O2 | 612.9537 | 1000 |
A stock standard mixture of PAHs, Nitro-PAHs and oxidized PAHs (O-PAHs) (Supelco, grade TraceCERT®) was diluted in methanol:dichloromethane 1:1 to obtain the diluted standard solution reported in the Table 4. The concentrations were in the range 6-133 μg/mL for PAHs, 0.6-5.3 μg/mL for Nitro-PAHs and 0.13-13 μg/mL for O- PAHs. The solution was stored at -18°C to prevent degradation.
Table 4. PAHs, oxo-PAHs and nitro-PAHs standard solution composition.
Compound | Molecular formula | Stock solution (μg/mL) | Diluted St. MIX (µg/mL) |
Acenaphthene | C12H10 | 1000 | 66.67 |
Acenaphthylene | C12H8 | 2000 | 133.33 |
Anthracene | C14H10 | 100 | 6.67 |
Benz[a]anthracene | 100 | 6.67 | |
Benzo[b]fluoranthene | 100 | 6.67 | |
Benzo[k]fluoranthene | 100 | 6.67 | |
Benzo[ghi]perylene | 200 | 13.33 | |
Benzo[a]pyrene | 100 | 6.67 | |
Chrysene | 100 | 6.67 | |
Dibenz[a,h]anthracene | 200 | 13.33 | |
Fluoranthene | 100 | 6.67 | |
Fluorene | 200 | 13.33 | |
Indeno[1,2,3-cd]pyrene | 100 | 6.67 | |
1-Methylnaphthalene | 1000 | 66.67 | |
2-Methylnaphthalene | 1000 | 66.67 | |
Naphthalene | 1000 | 66.67 | |
Phenanthrene | 100 | 6.67 | |
Pyrene | 100 | 6.67 | |
9-nitroanthracene | 1000 | 0.67 | |
4-nitrocatechol | 1000 | 5.33 | |
4-nitrophenol | 1000 | 2.67 | |
9,10-antraquinone | 1000* | 13.33 | |
9-phenanthrenecarboxaldehyde | 1000 | 2.67 | |
9-fluorenone | 1000 | 2.67 | |
1-naphthaldeyde | 1000 | 5.33 | |
9-hydroxyphenanthrene | 500 | 0.13 | |
9-hydroxyfluorene | 500 | 0.13 |
5.2 Samples and sample preparation
5.2.1 Freshwater samples
All water samples (27) analysed for cyanotoxins determination were selected among freshwaters intended for human consumption or drinking water affected by cyanobacterial blooms. Samples have been collected and processed by the Department of Inland Water of the National Institute of Health (ISS), accordingly to the analytical procedures described in a previous research [12]. The procedure consist in a preliminary freezing of the sample; this step directly damage cell membranes and release intracellular toxins. Cell lysis prior to filtration allows the simultaneous detection of both extracellular and intracellular toxins. The sample concentration was carried out by extraction and clean-up through a Graphitized Carbon Black cartridge and elution with organic solvent.
A list of samples, together with the cyanobacteria identified during the morphological analysis is reported in Table S 1. Twenty µL of the sample extracts were injected onto the LC/MS system. Conversely, the water sample named Bidighinzu, was injected directly into the detection system (40 µL), since the concentration of cyanotoxins was expected to be quite large.
5.2.2 Food supplements.
The following commercial pharmaceutical formulations were purchased in drugstores: Cialis® from EliLilly (Sesto Xxxxxxxxxx , Italy), Levitra® from Bayer Pharma AG (Berlin, Germany) and sildenafil from DOC Generici (Milan, Italy). Two sets of food supplements and herbal dietary products were collected in this project. The first one, consisting in 21 bulk material for herbal products were supplied by a local seller, while the second set for a total of 5 samples were collected during a market monitoring campaign coordinated by the Italian Medicine Agency (AIFA) and in collaboration with the special anti-adulteration police force (N.A.S). These samples were bought in various shops (sexy-shops, herbalist's shop and ethnic shops) in four Italian cities.
Figure 8. Examples of food dietary supplement analysed.
After grinding solid samples in a ceramic mortar, about 10 mg of each matrix were weighted into an Eppendorf tube. Each sample was extracted with 1.5 mL of water: acetonitrile 50:50 (v/v) acidified with 0.1% FA for 10’ in a sonic bath. After centrifugation at 10000 RPM for 10’ (Mikro120, Xxxxxxx), an aliquot of 0.5 mL of the supernatant was transferred in a 1.5 xX xxxx for auto-sampler. For pharmaceutical formulations, the extracts were diluted by a factor 1000 to bring the analyte concentrations inside the range of the calibration curve.
5.2.3 XXXXXX trial sample.
The sample used in the collaborative trial was collected from location JDS57 on the Danube River, downstream of Ruse/Giurgiu on the 18th of September 2013. The sample had been prepared by large-volume solid-phase extraction (LVSPE) of 1000 litres of river water . The sampler cartridge was composed by 160 g of Macherey Xxxxx Chromabond® HR-X (neutral resin), 100 g each of Chromabond® HR-XAW (anionic) and HR-XCW (cationic exchange resin). The resins were extracted with 500 mL each of ethyl acetate and methanol (HR-X), 500 mL methanol with 2% of 7 M ammonia in methanol (HR-XAW) or 500 mL methanol with 1% formic acid (HR-XCW). The extracts were then combined, neutralized, filtered (Whatman GF/F) and reduced to a final volume of 1 L using rotary evaporation. Aliquots of 1.5 mL, equivalent to 1.5 L of river water, were transferred into vials and evaporated to dryness under nitrogen. The sample war reconstructed with 1.0mL of H2O:AcN 50:50, and an aliquot of 40µL was injected into the LC-MS apparatus.
5.2.4 Urban PM2.5
Ten Teflon filters (47 mm, 0.2 µm) were pre-treated for removing organic contaminants. Filters were washed for 30 minutes in ultrasound bath successively with 2 X 20mL of deionized water, 2 X 20mL of acetonitrile and 2 X 20 mLof methanol. Finally, filters were dried under vacuum for one hour and stored in a clean desiccator.
Six samples of PM2.5 were collected for 24 hours in the sampling room of the Department of Chemical Sciences of Padua (45.41°N, 11.88°E), while four filters were left unused as procedural blanks. A Xxxxxxxx Explorer Plus PM sampler equipped with proper inertial impactors was placed in a sampling room equipped with a wall fan which sucks air from outside making the room representative of Padua urban background air. The sampling system was fitted with PM2.5 certified selectors (in 2006, CEN standard methods UNI-EN 12341 and UNI-EN 14907) working at a constant flow rate of 38.3 L/min (2.3 m3/h) and equipped with Ø 47 mm Teflon (PALL, Teflon Membrane, 1 µm pore size).
Further six samples, collected on quartz filters in Mandria, a suburban area of Padua, were obtained from ARPAV (Regional Agency for the environmental protection), together with their respective blank samples. Table 5 reports the specification of the samples collected.
Table 5: Sampling specification.
Date | Filter’s code | Volume (L) | PM 10 Conc. (μg/m3) | Rain precipitation (mm) | Average Temperatur e (°C) | Filter type |
08/01/2015 | FP1-080115 | 54947 | 96 | 0 | 11.1 | PTFE |
09/01/2015 | FP2-090115 | 55022 | 113 | 0 | 10.2 | PTFE |
10/01/2015 | FP3-100115 | 55192 | 113 | 0 | 10.0 | PTFE |
12/01/2015 | FP4-120115 | 54943 | 87 | 0 | 10.5 | PTFE |
13/01/2015 | FP5-130115 | 55047 | 80 | 0 | 10.9 | PTFE |
14/01/2015 | FP6-140115 | 55051 | 91 | 0 | 10.7 | PTFE |
02/06/2014 | Q2-020614 | N/A | 16 | 0 | 19 | Quartz |
04/06/2014 | Q3-040614 | N/A | 17 | 8 | 19 | Quartz |
06/06/2014 | Q4-060614 | N/A | 33 | 0 | 22 | Quartz |
09/06/2014 | Q5-090614 | N/A | 34 | 0 | 26 | Quartz |
17/06/2014 | Q6-170614 | N/A | 11 | 0 | 21 | Quartz |
19/06/2014 | Q7-190214 | N/A | 17 | 0 | 21 | Quartz |
All glassware and taps used in the extraction procedure were accurately cleaned using at least three washing with HPLC grade methanol (Fluka).
A quarter of filter was manually cut and the portions of filter on the borders not containing sample were removed. The samples were extracted three times either with 5mL of methanol in ultrasound bath at 0°C for 30 minutes. Low temperature is needed to avoid the methylation of carboxylic groups by reaction with methanol. The extracts were combined and filtered through PTFE filter 0.45µm and 0.22µm and then evaporated at 29°C under gentle nitrogen steam until the final volume of 1.0 mL.
5.3 Instrumental analysis
5.3.1 Cyanotoxins in drinking water
LC-Q-TOF-MS analysis were performed with an ultra-high pressure liquid chromatography (UHPLC) system (Agilent Series 1200; Agilent Technologies, Palo Alto, CA, USA), consisting of vacuum degasser, auto-sampler, a binary pump and a column oven coupled to both Diode-Array Detection and Q-TOF-MS mass analyser (Agilent Series 6520; Agilent Technologies, Palo Alto, CA, USA).
The analytical column was a Kinetex C-18 (2.6µm 100 mm x 2.1 mm i.d., Phenomenex, Italy) and it was thermostated at 30°C. The sample-injected volume was 20 µL. The mobile phase components A and B were water and acetonitrile respectively, both acidified with 10 mM formic acid. The eluent flow rate was 0.3 mL/min. The mobile phase gradient profile was as follow (t in min): t0, B= 20%; t10, B= 55%; t11, B= 80%; t15, B= 100%; t19, B= 100%; t20, B=20%; t25, B=20%.
The Q-TOF system was equipped with an ESI, operating in dual ESI mode and positive ESI acquisition, with the following operation parameters: capillary voltage, 3500 V; nebulizer pressure, 35 psi; drying gas, 8 L/min; gas temperature, 350°C; fragmentor voltage, 180V; skimmer 65 V.
5.3.1.1 Two-step confirmation analysis and suspect screening protocol
This protocol based on HPLC-Q-TOF platform, consists in a preliminary analysis in full scan MS mode followed by a data processing step involving a customized library for detecting the MS signals of interest. A further acquisition analysis in target MS/MS xxxx provided the structural information needed for the identification of the cyanotoxin
variants. The use of standards in both the analysis mode allowed the confirmation of the analytes.
Full scan mass spectrum was recorded as centroid over the range 50–2000 m/z with a scan rate of 2 spectra/s and analysed by the software Masshunter to find the suspect analytes. A Metlin database was generated and used within the Molecular Features Extraction (MFE), setting the following parameters and thresholds: a) ion compound filters ≥ 1000 in MS level, b) retention time ≤ 2.5% of tolerance with respect to selected cyanotoxins standards and c) MME ≤ 20 ppm.
Target MS/MS analysis were performed at 8 spectra/s over the m/z range 50-2000 and CID energy of 45eV for quasimolecular ions and 20eV for double-charged species.
The Q-TOF calibration was daily performed by using the manufacturer’s solution. For all chromatographic runs, the m/z 391.2843 relative to the diisooctylphthalate molecular ion, always present as impurity, was set as lock mass for accurate mass analysis. The instrument provided a typical resolving power (FWHM) of about 18000 at m/z 311.0805.
5.3.1.2 Auto MS analysis and suspect screening protocol.
This protocol consists in recording in the same chromatographic run the full scan and MS/MS xxxxxxx of the suspect analytes by the “autoMS/MS” feature of the mass analyser.
The same database used in the two-step protocol described above, was converted into a
.csv file to be used as a preference list for auto MS scan acquisition mode.
The same HPLC conditions and source parameters as described above were used in this screening approach. In this MS feature the following parameters were set: MS and MS/MS scan speed 6 and 8 spectra/s respectively, for a total cycle of 1.26 s, 8 compounds per cycle, isolation width of 4 m/z, active mass exclusion enabled after 100 spectra and 0.3 min, absolute and relative precursor threshold 200 counts and 0.001 % respectively.
5.3.2 PDE-5 inhibitors in food supplements
Q-TOF-MS analysis were performed with an HPLC system (Agilent Series 1200; Agilent Technologies), consisting of vacuum degasser, auto-sampler, a binary pump and
a column oven coupled to a Q-TOF-MS mass analyser (Agilent Series 6520; Agilent Technologies).
Separation was achieved with a PolymerX™ (5 µm, 150 mm x 2 mm i.d., Phenomenex, Italy) thermostated at 30°C. The mobile phase components A and B were water and acetonitrile, respectively, both acidified with 0.1% of FA. The eluent flow rate was 0.3 mL/min. The mobile phase gradient profile is the following: t0, B= 5%; t15, B= 65%; t16, B=100%; t22, B= 100%; t23, B= 5%; t30, B= 5%. Injected sample volumes were 5 µL.
In this MS feature the following parameters were set: MS and MS/MS scan speed 6 and
8 spectra/s respectively, for a total cycle of 1.26 s, 8 compounds per cycle, isolation width of 4 m/z, active mass exclusion enabled after 100 spectra and 0.3 min, absolute and relative precursor threshold 200 counts and 0.001 % respectively.
Simultaneous MS and tandem MS analysis were performed by using the AutoMS/MS feature. Centroid MSMS spectra were recorded over the range 50-2000 m/z with a scan rate of 3 spectra/s in both MS and tandem MS mode for a total cycle time of 1.26 s, 2 compounds per cycle, isolation width of 4 m/z, absolute and relative precursor threshold of 200 counts and 0.001 % respectively. Collision energies are reported in Table 9.
A “.csv” file including 82 PDE-5 inhibitor analogues was used as preferred list of precursor ions in the AutoMS/MS experiments. The corresponding Metlin database was generated and loaded within the Molecular Features Extraction (MFE), setting the following parameters and thresholds: a) ion compound filters ≥1000, b) retention time ≤ 2.5% of tolerance with respect to the seven selected PDE-5 Inhibitors standards and c) MME ≤ 20 ppm.
Mass spectra acquisition and data analysis was processed with Masshunter Workstation B 04.00 software (Agilent Technologies).
5.3.2.1 Quantification of PDE-5 and evaluation of the method performance.
Due to the lack of an internal standard, the quantification for target analytes was performed using the corresponding external calibration curve, using least squares regression. Precursor ions at the MS level were chosen as quantifier ions, after checking the simultaneously presence of at least one fragment ion with S/N >3 in the MS/MS acquisition mode.
The method performance was evaluated in terms of linearity, limit of detection (LOD) and quantification (LOQ), recovery, repeatability, and matrix effect following the guidelines reported in SANCO/10684/2009 [85] and Decision 2002/657/EC [86].
Five-point calibration curves were constructed by injecting standard solutions prepared in the extractant solution. Concentrations ranged between the limit of quantification to 1000 pg injected. LODs and quantification LOQs were experimentally estimated through a specific calibration with at least four levels at concentration close to the LOD, and applying a recent statistical method reported in literature [87, 88].
For herbal bulk analysis, trueness was calculated from replicate analysis (n = 3) of blank matrix spiked at three different concentrations levels 1.0, 5.0, 10.0 µg/g, comparing sample peak areas with those of standard solutions at the same concentration. Repeatability of the method was assessed as intra-day precision (n=3) expressed as relative standard deviation (RSD) of the peak areas, while reproducibility as inter-day precision (N=15).
Matrix effects were evaluated by comparing the slopes of three-point calibration curves obtained fortifying each sample before sample injection, with those of the standard calibration. Solvent and procedural blank samples were used to check selectivity of the method, and in particular, the absence of any carryover effect was verified by repeating injections of solvent blank after analysis of 1.0 ng of standard solution.
5.3.3 Direct infusion full scan protocol for molecular formulas identification in PM2.5.
This protocol allowed the determination of the chemical composition of the sample through a fast data treatment.
Instrumental analyses were performed using high-resolution LTQ Orbitrap™ Velos mass spectrometer (Thermo Xxxxxx, Bremen, Germany). The mass analyser was calibrated with the manufacturer calibration solution before each analysis. The mass accuracy of the instrument, checked before analysis, was always below 0.5 ppm. The instrument mass resolution was set at 100 000 (measure at m/z 400). Each sample was analysed in the ranges of m/z 100−650 and m/z 150-900, acquiring each range three times for 60seconds. The acquisition was considered acceptable only if the spray resulted sufficiently stable, with variations of the total ion current (TIC) profile versus time within 80- 100%.
NanoESI mass analyses were performed using a TriVersa Nanomate robotic nano-flow chip-based ESI (Advion Biosciences, Ithaca NY, USA) source. The direct infusion negative nanoESI parameters were as follows: ionization voltage 1.6 kV, back pressure
0.8 psi, capillary temperature 275 °C, S-lens RF level 60%, sample volume 8µL. For the positive mode the same parameter were used except ionization voltage and back pressure set at 1.4kV and 0.3 psi respectively.
APPI analyses were performed using an Ion Max™ source (Thermo Xxxxxx, Bremen, Germany) set to work in APPI mode with a Syagen Krypton lamp emitting photons at
10.0 eV and 10.6 eV. Source parameters were: temperature 200C, auxiliary gas flow 5a.u. (arbitrary units) and sweep gas 10 a.u. The flow used in direct injection was 10µL/min.
The mass analyser was calibrated before the analysis on the samples using the commercial calibration solution. The mass accuracy of the instrument was checked before the analysis and was below 0.5 ppm.
5.3.3.1 Post run data analysis
The post-run data analysis workflow for the assignation of unique molecular formulas to each m/z values involves four steps here summarized:
1. Assignation of molecular formulas to the experimental MS signals.
2. Determination of noise, mass precision and mass accuracy of the measures.
3. Molecular formulas filtering.
4. Determination of the common ions in the replicate acquisitions.
The first step was carried out using the qualitative browser of the software XcaliburTM
2.1 by Thermo Scientific, which can generate molecular formulas for the peaks present in the MS spectrum, enforcing the results to observe some constriction described below. The steps 2 and 3 were implemented through two separate algorithms (“Mass Drift v1.11” and “MassSpecProcessing v1.0” respectively) wrote in Mathematica programming language and implemented in Mathematica 10.0 by Wolfram Research. Steps 2 and 3 were separated in order to perform a manual control of the process and identify possible failure in the “Mass Drift v1.11” algorithm. Fina ly, using Excel, mass ranges are merged in the same spreadsheet and the common ions determined in the replicates are obtained. Each mass spectrum was obtained by the average of one minute of acquisition, corresponding at 40 single spectra. In the generation of molecular formulas, carried out in Xcalibur 2.1 qualitative software, the follow constrains on the elemental composition were applied: 1 ≤ 12C ≤ 75; 13C ≤ 1; 1 ≤1 H ≤ 180; 1 ≤ 16O ≤ 50; 14N ≤ 30; 32S ≤ 2; 34S ≤ 1, mass tolerance 6ppm and maximum number of formulas per peak 10. For positive nanoESI acquisitions, the presence of one sodium atom is allowed in the molecular formula generation. Limiting the number of generated possibility was
needed to avoid extremely long processing time. The formulas list associated to each peaks, are exported in CSV file containing the accurate m/z value (five decimal digits), the intensity of the signals, the associated possible molecular formulas and the mass errors referred to the exact mass of the formulas expressed in ppm.
Mass error offset (mass accuracy, µ) and standard deviation of mass errors (mass precision, SD) were automatica ly calculated using the algorithm “Mass Drift v1.11”, on the basis of known contaminants or substances likely to be present in the sample and previously confirmed via MS/MS experiments.
The algorithm “MassSpecProcessing v1.0” filters the assignments on the bases of heuristic rules. The elemental ratios were set at H/Cmin 0.3; H/Cmax 2.5; O/C max 2; O/Cmin 0; N/Cmax 1.3; S/Cmax 0.8. Signal to noise level was set at 5.
The common ions in the replicates of the two ranges are then merged in the same excel spreadsheet and the duplicates entries were removed.
6.1 Protocol optimization
6.1.1 Creation of database.
When LC/HRMS is used for suspect screening and confirmation of target compounds with reference standards, the positive candidate has to comply with general queries, i.e. matching retention times and/or fragmentation pattern with acceptable tolerance, and a signal or S/N higher than a certain threshold[89]. The possibility of identify and/or quantify suspect contaminants is limited by the creation of databases as large as possible, with retention times and accurate MS and MS/MS spectra of target compounds. Because the non-availability of a fit-to-purpose commercial or online database, the development of an in-house library was required.
Molecular formulas of selected cyanotoxins with relative monoisotopic mass and chemical structures, uploaded as .mol files, were inserted in the library, for a total of
210 entries, belong to MCs (110), anabaenopeptins (ANPs) (23), cyanopeptolins (13), microginins (9), saxitoxins, anatoxins, cilindrospermopsins and other oligopeptides (the full list is not reported). Some MC-related compounds reported as transformation products by water treatment [90] were added to explore their contribute to drinking waters contamination. All entries were found in the literature and all the references are not reported because the great number. Moreover, the database could be improved in terms of number of entries; retention times and fragments emerged from cyanotoxins characterization.
6.1.2 Optimization of the instrumental conditions
An initial instrumental condition was adopted according a previous study [12]. Then, two different columns, both based on the “solid core technology”, namely Kinetex C-18 and Accucore (2.6µm 100 mm x 2.1 mm i.d, Superchrom, Italy) were evaluated in terms
of peak width and resolution on selected standards of cyanotoxins, using ESI source in positive acquisition mode. Mobile phase composition was modified to improve peak shapes and MS signals intensity of the selected cyanotoxins. Water, acetonitrile and methanol all containing formic acid at different concentrations (10 mM and 0.1% v/v) were tested on both columns. Flow rate, initially set at 0.2 mL/min, was increased up to
0.3 mL/min in order to improve S/N and consequently the Limits of Detection (LODs) of the methods, without incurring in the signal depletion due to a poor efficiency of the ionization process. Results of these experiments have shown that the best conditions in terms of sensitivity, S/N and chromatographic efficiency were obtained with the Kinetex C-18 column, using water and acetonitrile; both contains formic acid at 10 mM, at flow rate of 0.3 mL/min.
For MC variants that exhibit both the mono charged molecular ion [M+H]+ and the doubly charged molecular ion [M+2H]2+, the most intense between the two signals was used in the subsequent processing. Figure 9 shows the extracted ion current profile for a working standard solution of available MCs injecting 250pg of each compounds. The only remarkable observation about ESI-MS pattern of the MCs analysed, is the presence of the double charge protonated molecular ion [M+2H]2+ as base peak for the MC-LR (m/z 498.2817), [D-Asp3] MC-LR (m/z 491.2738), and MC-YR (m/z 523.2713), while
the [M+H]+ ones account for about 40%. Conversely, MC-RR variants exhibit almost exclusively [M+2H]2+and the other selected congeners only [M+H]+ ones [91]. This general behaviour has been useful for the characterization of the non-target MCs by using MS fragmentation experiments.
Negative acquisition mode was evaluated using ammonium as modificant in the range 0.2-10 mM in both mobile phases, since the Accucore column can be used with a quite large range of pH (1.5-11). Although signals of acid variants of MCs, like MC-LA, MC- LY and MC-LW, have shown a significant improvement with the increasing concentration of ammonium, nevertheless this conditions were not adapt for analysing other cyanotoxins with basic amino acids (signals about one order of magnitude lower). Even using mixed mobile phases (acidic water and basic acetonitrile), and switching the acquisition mode from ESI+ to ESI- during the chromatographic run, results were not suitable to a comprehensive target method.
Identification and confirmation purposes need the presence for each structure present in the database of fragmentation spectra. As fist attempt, in-source fragmentation was studied. High transmission voltage (fragmentor voltage) have been used to produce the
in-source fragmentation in order to obtain exact masses of both precursor and daughter ions in a single chromatographic run. Experiments have been performed with some representative MCs, after the variation of the fragmentor voltage (range 100-350 V) to obtain both molecular ions and the characteristic ion fragment at 135.0804 m/z, related to the [C9H11O]+ ion generated by the breakage of the common Adda amino acid. Data obtained have showed that this approach is not effective, since a unique value of fragmentor voltage was not suitable to produce molecular and daughter ions with the adequate sensitivity, for analysing all target and suspect cyanotoxins. In fact, different fragmentor voltages are required, due to the generation of mono and double charged molecular ions related to the amino acid composition, resulting in a different energy involved in the MS fragmentation process. Moreover, the unpredictable co-elution of MC variants could generate merged fragment ions related to the Adda-moiety that results in a non-specific assignment of the structure.
Due to the issues present in the use of in-source fragmentation, we have preferred to perform the fragmentation in the collision cell. The optimization of collision energies for the target MS/MS analysis, were made by using MC standards. Two ranges of collision energies were outlined: 15 eV for double charged and 60 eV for mono-charged molecular ions respectively. The general fragmentation behaviour of the MC congeners is the α-cleavage at the methoxy-group of the Adda β-amino acid moiety, with the production of the previously described ion at 135.0804 m/z, often together with its complementary [M+H-134]+ [12].
Compound | Theoretic molecular ion, m/z | Tr, min | LOD, pg inj | Collision Energy, eV | Theoretic fragment ion, m/z | Fragment structure |
[D-Asp3]MC- RR | 512.7827* 1024.5574 | 890.4724 | [M+H-134]+ | |||
6.05 | 20 | 15 | 135.0804 | [PhCH2CH(OMe)]+ | ||
519.7905* 1038.5731 | 904.4981 | [M+H-134]+ | ||||
MC-RR | 6.29 | 4 | 15 | 135.0804 | [PhCH2CH(OMe)]+ | |
375.2011 | [C11H13O+Glu+Mdha+H]+ [PhCH2CH(OMe)]+ | |||||
MC-YR | 1045.536 | 7.9 | 50 | 60 | 135.0804 | |
375.2011 | [C11H13O+Glu+Mdha+H]+ [PhCH2CH(OMe)]+ | |||||
MC-HtyR | 1059.548 | 7.99 | 50 | 60 | 135.0804 | |
213.0828 | [Glu+Mdha]+ [PhCH2CH(OMe)]+ | |||||
MC-LR | 995.5564 | 8.11 | 30 | 60 | 135.0804 | |
[D-Asp3]MC- LR | 375.2011 | [C11H13O+Glu+Mdha+H]+[P hCH2CH(OMe)]+ | ||||
981.5402 | 8.16 | 50 | 60 | 135.0804 | ||
213.0859 | [Arg+NH3CO]+ [PhCH2CH(OMe)]+ | |||||
MC-HilR | 1009.574 | 8.47 | 50 | 60 | 135.0804 | |
505.2920* 909.4848 | 375.2011 | [C11H13O+Glu+Mdha+H]+ [PhCH2CH(OMe)]+ | ||||
MC-LA | 8.55 | 30 | 15 | 135.0804 | ||
375.2011 | [C11H13O+Glu+Mdha+H]+ [PhCH2CH(OMe)]+ | |||||
MC-WR | 1068.553 | 8.71 | 40 | 60 | 135.0804 | |
375.2011 | [C11H13O+Glu+Mdha+H]+ [PhCH2CH(OMe)]+ | |||||
MC-LY | 1002.517 | 11.62 | 30 | 60 | 135.0804 | |
375.2011 | [C11H13O+Glu+Mdha+H]+ [PhCH2CH(OMe)]+ | |||||
MC-LW | 1025.536 | 12.98 | 40 | 60 | 135.0804 | |
375.2011 | [C11H13O+Glu+Mdha+H]+ [PhCH2CH(OMe)]+ | |||||
MC-LF | 986.5226 | 13.21 | 30 | 60 | 135.0804 |
LR 8) MC-HilR 9) MC-LA 10) MC-WR 11) MC-LY 12) MC-LW 13) MC-LF.
6.1.2.1 Post-run data analysis: full scan acquisition
A post-run data analysis approach allows the identification of suspect contaminants re- processing data acquired in full scan mode, without analyte-specific optimizations. This feature is suitable to be used, even in parallel with more robust quadrupoles, for simultaneously target confirming and suspect screening purpose. Consequently, data will be useful for selecting emerging contaminants, in the frame of food and environmental safety.
Thus, a database was created using the Metlin feature embedded with the Masshunter software that takes into consideration automatically multi-charged molecular ions.
Threshold values and queries were set within the MFE functions to narrow the list of positive candidates (see experimental section). Chromatographic signals with retention times lower than two minutes were rejected as a further query, since in this range the accuracy and consequently, the reliability of results, was strongly influenced by matrix compounds. It is worth noticing that the tolerance of 20 ppm is not so large, if we consider that 10 mDa value, often used for the analysis of low-weight pesticides, corresponds to about 15-150 ppm for that molecular weights. The value chosen by us, corresponding to about 20 mDa for the majority of cyanotoxins, includes all the sources of uncertainty related to non-target analysis in real samples, e.g. analyses at trace levels, saturation of the detector and matrix effects.
Nevertheless, the number of total positive results was huge, even because all water samples were selected among ones highly contaminated by cyanobacteria. Thus, about two hundreds of results were carefully checked to eliminate false positive results due to
a) ions recognized as salts adduct of the databases entries, but actually related to different signals; b) ions clearly referable to series of homologues compounds, e.g. differing for an ethoxylate group; c) ions belonging to the isotopic pattern of other signals; d) ions ascribable to other compounds loosing water or ammonia. Finally, the ratio between [M+H]+ and [M+2H]2+ was considered a remarkable criterion for the identification of toxins containing basic structures like Arg (R-series of MCs). Following this scheme, the initial list was substantially reduced to a total of about a hundred hits for all samples processed. Then, all results related to algal toxins for which standards are not available were subjected to a structural confirmation by means of tandem MS analysis in target acquisition mode.
6.1.2.2 Post-run data analysis: MS fragmentation
CID fragmentation was used to confirm the toxins which the standard is present and the standardless metabolites. Two ranges of collision energies were outlined, slightly different from those used for MCs optimization: 10-20 eV for [M+2H]2+ and 40-70 eV for [M+H]+ respectively. Only compounds confirmed by the presence of two or more characterizing fragment ions were considered positively identified (Table 8).
A general behaviour of the MCs congeners is the α-cleavage at the methoxy group of the Adda β-amino acidic moiety, with the production of the previous described ion at 135.0804 m/z, often together with its complementary [M+H-134]+ [92]. Similarly, ANPs exhibit a typical signal at 263.1390 m/z , corresponding to the |MAla2-Hty3| charged residue (through all the text dm stands for demethyl, m for O-methylation, while M for N-methylation). Figure 10 and Figure 11 report the general structures of MCs and anabanopeptins with their corresponding class-fragment ions.
Other significant signals were ascribed to the typical fragmentation pattern, mainly b and y-type ions, with relevant difference due cyclic nature of these peptides and the involving of the peculiar Adda residue [93][94]. Further information has been supplied in the specific discussion of algal metabolites encountered in water samples.
Figure 10. General structure of microcystins with related class-fragment ions.
Figure 11. General structures of Anabaenopeptin with related class-fragment ions
6.1.3 Auto MS scan
An alternative approach for cyanotoxins identification is represented by the possibility to perform an automatic fragmentation of relevant signals acquired in HRMS full scan mode within a single chromatographic run. When the Metlin file was converted into a
.csv file to be used as a preference list, attention was taken at adding both [M+H]+ and [M+2H]2+ according to the observed MS behaviour. The instrumental parameters described in the experimental section have been optimized using standard solutions and real samples. In particular, the scan rate in MS level and MS/MS level and number of ions processed in each cycle were selected in order to have a good peak reconstruction (needed for the quantitation) and the presence of the fragmentation spectra maintaining
at the same time a good sensitivity for all target and non-target compounds. Active exclusion was introduced to in the method because, in the first attempts, several peaks were lost.
Since the presence of several algal toxins in a limited range of retention times, requires the rapid acquisition of a high number of MS and MS/MS spectra. In this circumstance, the typical MCs fragment ion is not sufficient to guarantee a specific characterization, since co-eluted MCs variants with the same precursor ion (within the isolation width range), could result in merged MS/MS signals, not easy to be interpreted. The auto MS scan feature was evaluated in terms of number of positive results at the MS level, presence of diagnostic daughter ions, chromatographic peaks definition and detection capability. Results have been compared with the two-steps acquisition mode: auto MS profiles obtained, appeared qualitatively equivalent to those arisen from the two-steps analysis, ensuring the accurate determination of both molecular and fragment ions. Moreover, MS/MS fragmentation is often available from both mono and double charged precursor ions, giving useful information on peptide residues. Anyway, a lower sensitivity was experienced for the auto MS scan, being its MS signals 50-80% lowers than corresponding ones in MS mode. This fact influenced the number of positive results, when the supposed algal toxins were present at very low concentrations (low XX xxxxxx). Consequently, up to 10% of identification ability was lost for water samples contaminated at so critical concentrations. Comparable results were obtained by applying the internal database in post-run analysis of the autoMS/MS file. Notwithstanding the described drawbacks, auto MS approach seems to be more friendly in terms of time of analysis and data interpretation, because the presence of the typical class-fragments for MCs and ANPs is useful to identify these compounds, without any other consideration about possible interferences.
6.1.4 Quality control and quantitation
Quality control has been mainly focused on the instrumental accuracy. The use of a continuous infusion of the manufacturer calibrant solution is not suitable, because a significant matrix effect can affect identification and quantitation [95]. Thus, the use of a lock mass (diisooctyl phthalate) has been preferred. After an initial tune calibration, the MME of nodularin, always added as IS at 1 µg/L [12], was 5.67±3.22 ppm assessed on 21 measures over three working-days. This performance, without any drift observed,
suggested that this raw approach is suitable for confirmatory-screening purpose, if further criteria, like MS fragmentation and retention times for target analytes, were considered. Retention times variations for IS and cyanotoxins with standards were lower than 0.1 min for all analyses performed.
Calibration of selected cyanotoxins has been performed on MS levels to explore the linear dynamic range of the LC-MS system. Coefficients of variation R2, obtained injecting in duplicate 20 µL of standard solutions in the 0.25-100 ppb range, were always close to the unit (Table 7), while overall RSD were lower than 5%. No matrix effects have been observed.
For the available standards, LODs has been estimated by evaluating S/N of the extracted ion current related to the exact molecular ion mass with a tolerance of 20 ppm from the full-scan spectra. LODs, assessed as the concentration equal to S/N=3 from injecting 50 pg of a standard solution, were reported in Table 7. Devoted validation experiments and a robust statistical analysis are necessary for any other specific application to be implemented, i.e. food analysis. These values were proposed only as an indication of the instrumental possibility of this approach.
Table 7. Calibration parameters and LODs of selected toxins.
External calibration | Internal Standard calibration | LOD (ppb) | |||
regression equation | R2 | regression equation | R2 | ||
[D-Asp3]MC-RR | y = 19196x - 28768 | 0.9989 | y = 0.0036x - 0.0043 | 0.9979 | 20 |
MC-RR | y = 9282x - 8454.5 | 0.9953 | y = 0.0017x + 0.0004 | 0.9969 | 4 |
MC-YR | y = 949x - 475.63 | 0.9983 | y = 0.0002x + 0.0002 | 0.9973 | 50 |
[D-Asp3]MC-LR | y = 1864x - 2017.8 | 0.999 | y = 0.0003x + 0.0001 | 0.9986 | 50 |
MC-LR | y = 1474x - 1511.7 | 0.9989 | y = 0.0003x + 0.0002 | 0.9982 | 30 |
MC-LA | y = 2058x - 330.17 | 0.9988 | y = 0.0004x + 0.0004 | 0.9995 | 30 |
MC-LY | y = 1897x - 1225 | 0.9985 | y = 0.0003x + 0.0003 | 0.9996 | 30 |
MC-LW | y = 1655x - 4031.4 | 0.998 | y = 0.0003x - 0.0003 | 0.9986 | 40 |
MC-LF | y = 2107x - 2934.7 | 0.9985 | y = 0.0004x – 0.0001 | 0.9994 | 30 |
6.2 Algal toxins identification in freshwaters
We have to point out that results on suspect identification of cyanotoxins here described are preliminary, since the analyses of surface and drinking waters were performed on a limited samples volume. Quantitation of the cyanotoxins with standards, already performed by ISS, was out of our scope, and the MS characterization of the suspect cyanotoxins has been restricted to the most relevant compounds, emerged from database
or auto MS scan processing. Anyway, the sample named Bidighinzu, collected in a lake of t Sardinia isle, was hardly contaminated by Microcystis spp and Aphanizomenon spp (about 1011 cells/L, reported by official analysis). The direct analysis of an aliquot of 40 µL of the filtered water sample, allowed to achieve very high MS signals (often >105) of compounds subsequently recognized as cyanotoxins. Thus, multiple MS/MS analyses for characterizing algal toxins were conducted mainly on this sample.
6.2.1 Microcystis toxins congeners
All water samples collected were affected by cyanobacteria, belonging to the Microcystis aeruginosa, Planktothrix rubescens and Aphanizomenon flos aquae. A description of water samples with results obtained with LC/HRMS analysis is reported in Table S 1, whilst Table 8 shows a detailed elucidation of the ions significant for the structural elucidation of selected target algal toxins.
Twelve uncommon MC congeners were identified using both LC/HRMS protocols proposed, derived from MFE processing followed by target MS experiments, or from auto MS scan. We have here focused our attention on product ions characterizing the specific MC variant.
Three high signals at 500, 507 and 532 m/z, ascribing at [M+2H]2+ species by the isotope patterns, and with correlated MS/MS ions, retained between RR and LR series, were recognized as two MSer7 variants of the MC-LR and one of the MC-YR [96]. Fragment ions accountable to MC-LR having the amino acid in position 3 methylated (m/z 399.2350, 416.2616 and 841.4818) or demethylated (272.1353, 385.2194, 402.2459 and 637.3596) were observed. The lack of the major signals involving
|Mdha7-Ala1| residues (e.g. 155.0815, 397.2082), always observed by us with MC standard, confirmed a modification in positions 1-7. The assignment to MSer7-LC variants was supported from several fragments experienced by us, i.e. those at m/z 173.0921, 771.4400, attributed to the sequence |MSer7-Ala1| and |Arg4-Adda5-Glu6- MSer7-Ala1| respectively. Anyway, we found signals elsewhere attributed to the conventional sequence Mdha7-Ala1 (m/z 213.0870 and 375.1915), in spite of those considered diagnostic for the MSer7 residue, i.e. ions 393 and 231 [93], which were not observed by us, neither using high nor in low collision energies. These fragments could be ascribed, even employing the useful freeware software mmass, to a dehydration of the corresponding b-type ion.
The same fragmentation pattern was observed for a compound eluted early before the described [MSer7] MC-LR variants, which we have characterized as the [MSer7] MC- YR. We have to point out that its molecular ion could be hypothetically referable to the [Ser7] MC-HtyR, as indicated by records of our database, but the MS behaviour was not coherent with this structure. This fact often occurred with putative MC signals.
Two signals corresponding to a demethylation of the MC-YR was recorded at tr=6.33 and 7.80 minutes. For the former peak, the presence of the fragment at m/z 121.0648 instead of the conventional 135.0804, and other signals indicating the methylation in position 7 and 3 (ions 213.0870, 375.1899 and 269.1244 respectively, data not shown), suggests a dmAdda5 variant. The latter was identified as the [Asp3] MC-YR, mainly for the presence of previously mentioned residues related to the sequence |Glu6-Mdha7- Ala1| and to |Arg4-Asp3| (measured m/z 272.1364, Δ = +1.1 mDa). Both compounds have produced Tyr immonium ions [97].
In this case, the comparison with the retention time related to the standard of the [D- Asp3] MC-LR, help us to assign this variant to the peak with higher intensity at tr=8.17. The presence or absence of the diagnostic fragments ions [93] related to the Adda- cleavage (m/z 135.0804 and 847.4672 or 121.0648), Mdha7 (m/z 127.0866, 213.0870 and 375.1899), Asp3 (m/z 272.1353) or Dha7 (m/z 475.2187), allowed us to characterize
these variants. Thus, we have assigned via MS fragmentation the structure of [dmAdda] MC-LR and [Dha7] MC-LR to peaks at tr=6.59 and 8.26 respectively, coherently with the chromatographic behaviour.
MC-M(O)R, with theoretical mass of 1028.5001 Da corresponding to C48H72N10O13S, was identified at tr =7.26, from both [M+H]+ and [M+2H]2+ with of MME <3 mDa. The daughter ions at m/z 965.5091 and 803.4120, obtained from fragmentation of the [M+H]+ at 60 eV, were ascribed to a characteristic loss from the methionine-S-oxide2 (MetO2) and amino acidic residue of |Glu6-Mdha7| respectively. Another MS signal at m/z 682.3803 could be assigned to the sequence |Arg4-Adda5-Glu6-Mdha7| [98]. No signals referred to demethylation in position 3 or 7 were registered. These MS signals, together with those generally produced by the MC class, allow the characterization of this MC congener.
Although present with low counts, the structure of MC-FR has been assigned to a peak present in the Bidighinzu sample at tr=8.72. The [M+H]+ measured at 1029.5400 m/z (-
0.4 mDa) was significantly different from that described for the MC-M(O)R, and it is
potentially referable to other two variants of MCs, i.e. [Asp3] MC-HphR and [Dha7] MC-HphR. The Adda-related MS fragment at m/z 375.1899 showed the presence of amino acidic residues of |Glu6-Mdha7|. Besides, signals corresponding to Arg4 and Phe- immonium ion with theoretical 120.0808 m/z [99] have allowed us to assign this compound to the MC-FR variant in place of the isobaric demethylated forms of MC- HphR.
Other relevant signals that showed several MC typical daughter ions with good accuracy suggest the presence of several MC variants reported in literature [100]. Anyway, the fragmentation pattern was quite different from that expected, so that a deeper investigation is necessary in order to identify these compounds.
No signals corresponding to cyanotoxins by-products reported elsewhere [90] were evidenced, although several samples of contaminated drinking water treated with hypochlorite or chloride dioxide have been examined.
Figure 13. Molecular structures of identified uncommon Microcystis.
6.2.2 Anabaenopeptins.
About 30 different ANPs, having a general hexapeptidic structure of five cyclic amino acids, with main variations in residues 6 and 4, and a characteristic uredo linkage have been described (Figure 10). Their toxicity, at concentrations occurring in environment, is currently unknown, although their biological activity includes inhibition of PP and serine protease [93, 101, 102]. ANP–type peptides have been produced from several cyanobacterial species, including Microcystis spp, Planktothrix and Aphanizomenon spp
[93] occurred in water samples analysed by us. Data are reported in Table S 1 and Table 6.
ANP-A and B were found in several water samples at tr =7.6 and tr =4.7 minutes respectively, as both [M+H]+ and [M+2H]2+, with different relative intensities. In fact,
ANP-B, having an Arg instead of a Hty in position 6, gives a [M+2H]2+ as base peak. Fragmentation of the two substances, performed on the [M+H]+ precursor ion at 40 eV for ANP-A and on [M+2H]2+ at 10 eV for ANP-B, gave the class-fragment at m/z 263.1390 and a number of product ions already described in literature [101, 103]. Their presence in water samples could be confirmed by comparing retention times and MS data with those of respective standards, which are nowadays commercially available, but did not for us at the time of this research.
Three isobaric ANPs variants (F, E, B1 or MM850) with molecular formula C42H62N10O9 and a molecular weight of 850.4701 Da, that cannot be distinguished with a MALDI-TOF system in full scan analysis [93], have been reported [98, 104, 105] (Figure 15). We often found a signal at tr =5.5 min related to their [M+H]+, that we assigned to ANP-F via fragmentation pattern. Instead, two distinct signals resolved with high efficiency (tr =5.51 and tr =5.99 in Figure 3), were observed in surface samples contaminated by Planktothrix rubescens and Microcystis aeruginosa or Plankthotrix rubescens and Aphanizomenon flos-aquae (Table S 1). Both compounds exhibit [M+H]+ and [M+2H]2+. Subsequently, in the auto MS scan mode, it was possible to obtain the fragmentation pattern of the two isobaric compounds in one step, using [M+H]+ (CE 40 eV) and [M+2H]2+ (CE 10 eV) as precursor ions. The assignment of the second signal to ANP-E having a MeHty in position 3 [101] was excluded, since fragments related to the Hty immonium ion and its amino acidic loss (theoretical m/z 674.3984) were always present. Moreover, this compound showed MS/MS pattern similar to ANP-B, having Val4 instead of Ile4 (theoretical m/z 637.3708), but with an increment corresponding to a methylation for all fragments containing the Arg residue (Figure 8 and Figure 14). The existence of an ANP named B1, presumably containing Har (homoarginine) in position 6, was supposed by Xxxxxxxx et al. [104], but a structure coherent with such fragmentation pattern was previously described for the ANP-MM850. The structure of this ANP-MM850, exhibiting a methyl ester of the Arg6, was confirmed by NMR analysis. Thus, we are prone to sustain the last description for the ANP variant experienced by us. Coupling the chromatographic separation to a “double” fragmentation process consented the unambiguous discrimination of these isobaric ANPs.
Another ANP, with fragment ions referable to the presence of Ile4 (m/z 651.3794) and Har6/mArg6 (m/z 189.1337 and 215.1131) was characterized at MS and MS/MS level. A similar structure with Har6 has been previously named F1. Anyway, even in this case,
the structure could be ascribed to a modification of the ANP-F, with a mArg6. Until the molecular structure will be confirmed with NMR data, this algal toxin will be here named ANP-MM864.
Oscillamide Y, a well known ANP-analogue, with formula C45H59N7O10 was present in several samples. In this case, structural information is available from fragmentation of the [M+H]+ with a very low MME.
Figure 14. Molecular structures of the uncommon Anabaenopeptins identified.
56
Figure 15. Extracted Ion Chromatogram (EIC) and relative mass spectra assigned to the isobaric anabaenopeptin F (left panel) and MM 850 (right panel), obtained from the analysis of a surface water sample from the lake Occhito: a) EIC at the MS level of the ion at at m/z 851.4774 with a tolerance of 20 mDa; b) mass spectrum at the MS level showing [M+H]+ and [M+2H]2+ ions; c) product ion scan obtained by using autoMS/MS features and [M+H]+ as precursor ion; d) product ion scan obtained by using autoMS/MS features and [M+2H]2+ as precursor ion.
57
Compound | Tr, min | Molecular formula | Molecular ion, m/z (MME, mDa)1 | Fragment ions, m/z MME, mDa) | Fragment formula2 | Assigned structures |
135.0798(-0.6) | [C9H11O]+ | [PhCH2CH(OMe)]+ | ||||
532.27713 (+0.5) | 929.4687 (-4.0) | [C43H65N10O13]+ | [M+H-1344]+ | |||
C52H74N10O1 4 | 173.09116 (-1.0) | [C7H13N2O3]+ | |MSer7-Ala1| | |||
[MSer7] MC-YR | 6.18 |
| ||||
891.4570 (-4.1) | [C45H63N8O11]+ | |Tyr2-MAsp3-Arg4-Adda5-Glu6| | ||||
1063.5439 (-2.0) | 771.4382 (-1.8) | [C38H59N8O9]+ | |Arg4-Adda5-Glu6-MSer7-Ala1| | |||
466.2389 (-2.0) | [C20H32N7O6]+ | |Arg4-MAsp3-Tyr2| | ||||
500.2812 (+2.1) | 135.0802 (-0.2) | [C9H11O]+ | [PhCH2CH(OMe)]+ | |||
865.4775 (-0.3) | [C39H65N10O12]+ | [M+H-134]+ | ||||
[Asp3,MSer7]MC-LR | 6.41 | C48H74N10O1 3 | 402.2440 (-1.9) | [C16H32N7O5]+ | |Leu2-Asp3-Arg4| | |
173.0922 (+0.1) | [C7H13N2O3]+ | |MSer7-Ala1| | ||||
771.4347 (-5.3) | [C38H59N8O9]+ | |Arg4-Adda5-Glu6-MSer7-Ala1| | ||||
999.5511 (+0.1) | ||||||
385.2127 (-6.7) | [C16H29N6O5]+ | |Leu2-Asp3- Arg4| | ||||
135.0810 (+0.6) | [C9H11O]+ | [PhCH2CH(OMe)]+ | ||||
C49H76N10O1 3 | 1013,5626 (-4.0) | 771.4408 (+0.8) | [C38H59N8O9]+ | |Arg4-Adda5-Glu6-MSer7-Ala1| | ||
[MSer7] MC-LR | 6.51 | 399.23816 (+3.1) | [C17H31N6O5]+ | |Leu2-MAsp3-Arg4| | ||
375.19536 (+3.9) | [C20H27N2O5]+ | |C11H13O-Glu6-MSer7(-H2O)| | ||||
507.2869 (0) | 000.0000 (+7.1) | [C40H67N10O12]+ | [M+H-134]+ |
173.09276 (+0.6) | [C7H13N2O3]+ | |MSer7-Ala1| | ||||
269.1186 (-5.8) | [C11H16N4O4]+ | |MAsp3-Arg4| | ||||
1031.5139 (-4.8) | 902.4784 (+1.3) | [C46H63N9O10]+ | |Arg4-dmAdda5-Glu6-Mdha7-Ala1-Tyr2| | |||
[dmAdda]MC-YR | 6.33 | C51H70N10O1 3 | 537.2762 (-1.8) | [C23H37N8O7]+ | |Ala1-Tyr2-MAsp3-Arg4| | |
456.2296 (-5.1) | [C43H64N10O12]2+ | [M+2H-1204]2+ | ||||
516.2692 (-5.2) | 136.0740 (-1.7) | [C8H10NO]+ | [imTyr5]+ | |||
121.0612 (-3.6) | [C8H9O]+ | [PhCH2CH(O)+H]+ | ||||
965.5102 (+1.1) | [C47H69N10O12]+ | [M+H-CH4SO]+ | ||||
1029.5053 (-2.1) | ||||||
C48H72N10O1 3S | 803.4136 (+1.6) | [C38H59N8O9S]+ | |Ala1-MetO2-MAsp3-Arg4-Adda5| | |||
MC-M(O)R | 7.26 |
| ||||
895.4334 (-0.8) | [C39H63N10O12S]+ | [M+H-134]+ | ||||
515.2597 (+0.4) | ||||||
135.0800 (-0.4) | [C9H11O]+ | [PhCH2CH(OMe)]+ | ||||
895.4688 (+1.6) | [C43H63N10O11]+ | [M+H-134]+ | ||||
1029.5400 (-0.4) | ||||||
135.0797 (-0.7) | [C9H11O]+ | [PhCH2CH(OMe)]+ | ||||
MC-FR | 8.72 | C52H72N10O1 2 | 515.2766 (+2.8) | 375.1899 (-1.5) | [C20H27N2O15]+ | |C11H13O-Glu6-Mdha7| |
174.1326 (-2.3) | [C6H16N5O]+ | [Arg4+NH3+H]+ | ||||
120.0785 (-2.3) | [C8H10N]+ | [imPhe]+ | ||||
667.3396 (-5.4) | [C34H47N6O8]+ | |Val4-Lys5-CO-Tyr6-Phe1-MAla2| | ||||
ANP-A | 7.56 | C44H57N7O10 | 844.4232 (-0.8) | 362.2036 (-3.8) | [C19H28N3O4]+ | |MAla2-Hty3-Val4| |
263.1371 (-1.9) | [C14H19N2O3]+ | |MAla2-Hty3| | ||||
837.4616 (-0.2) | 201.0970 (-1.2) | [X0X00X0X0]+ | |CO-Arg| | |||
ANP-B | 4.73 | C41H60N10O9 | 637.3685 (-2.3) | [C34H49N6O6]+ | |Phe1-MAla2-Hty3-Val4-Lys5| | |
419.2342 (-0.3) | 263.1367 (-2.3) | [C14H19N2O3]+ | |MAla2-Hty3| |
58
752.4069 (-2.1) | [C37H54N9O8]+ | |Hty3-Val4-Lys5-CO-Arg6-Phe1| | ||||
426.2399 (-2.4) | 674.3970 (-1.4) | [C32H52N9O7]+ | |Ile4-Lys5-CO-Arg6-Phe1-MAla2| | |||
ANP-F | 5.51 | C42H62N10O9 | 201.0967(-1.5) | [X0X00X0X0]+ | |CO-Arg6| | |
851.4793 (+1.9) | ||||||
651.3835(-3.0) | [C35H51N6O6]+ | |Phe1-MAla2-Hty3-Ile4-Lys5| | ||||
637.3715 (+0.7) | [C34H49N6O6]+ | |Phe1-MAla2-Hty3-Val4-Lys5| | ||||
ANP-MM850 | 5.99 | C42H62N10O9 | 851.4757(-1.7) | 215.1125 (-1.4) | [C8H15N4O3]+ | |CO-mArg| |
189.1383(-3.7) | [C7H17N4O2]+ | |mArg| | ||||
426.2311 (-11.2) | 674.3920 (-6.4) | [C32H52N9O7]+ | |Val4-Lys5-CO-mArg6-Phe1-MAla2| | |||
651.3794 (-7.1) | [C35H51N6O6]+ | |Phe1-MAla2-Hty3-Ile4-Lys5| | ||||
865.4891 (-4.0) | 215.1131 (-0.8) | [C8H15N4O3]+ | |CO-XArg6| | |||
ANP-MM864 | 6.54 | C43H64N10O9 | 189.1337 (-0.9) | [C7H17N4O2]+ | |XArg| | |
603.3596 (-1.7) | [C29H47N8O6]+ | |Ile4-Lys5-CO-XArg6-Phe1| | ||||
433.2469 (-5.3) | ||||||
688.4079 (-6.2) | [C33H54N9O7]+ | |Ile4-Lys5-CO-XArg6-Phe1-MAla2| | ||||
681.3577 (-2.9) | [C35H49N6O8]+ | |Ile4-Lys5-CO-Tyr6-Phe1-MAla2| | ||||
Oscillamide Y | 8.18 | C45H59 N7O10 | 858.4377 (-1.9) | 376.2229 (-0.2) | [C20H30N3O4]+ | |MAla2-Hty3-Ile4| |
263.1396 (+0.6) | [C14H19N2O3]+ | |MAla2-Hty3| |
59
6.3 Conclusion
The ability of the proposed protocols to work both as a structural-based screening and as confirmatory method has been proved on surface and drinking waters. In all samples processed, the target determination of cyanotoxins with certified standards was successfully accomplished. The database specifically implemented demonstrated its potentiality for the identification and characterization of suspect algal metabolites. Auto MS scan and the two-steps post-run data analysis could be efficiently used, depending on the different planning and purpose of the water monitoring: the former, as one-shot analysis specifically devoted to elucidation of the cyanotoxins contamination; the latter, as a possibility to acquire general LC/HRMS data that can be processed a posteriori to find emerging contaminants, even different from algal toxins. If carefully optimized, the auto MS feature seems to be able in achieving both objectives, with an acceptable loss of sensitivity. Both approaches are suitable to be improved enlarging the database entries, even with characteristic fragment ions, and used at ultra-trace levels in whatever matrix involved with cyanobacteria (environmental, biological or food samples). For such purpose, a full validation will be necessary for quantitative analysis. Preliminary results have evidenced a complex scenario of contamination for freshwaters affected by cyanobacterial blooms, enriched by unrecognized potential toxic MCs, and oligopeptides with a doubt harmful bioactivity. Features of an LC/Q-TOF-MS coupled to information automatically gained by databases or libraries, can effectively improve knowledge about cyanobacteria metabolites occurring, and thus restyle the analytical tools available for risk management related to cyanotoxins exposure.
7.1 Optimization of the instrumental conditions.
The optimization of MS parameters for the seven PDE-5 inhibitors (fragmentor voltage, source parameters and collision energies) was performed in flow injection analysis (FIA) using individual standard solutions at 100 ng/mL at flow rate of 0.3 mL/min water:acetonitrile 20:80 with 0.1% FA
Different ion sources were tested both in positive and negative operating mode. The ESI probe operating in positive mode was preferred to APPI (Atmospheric Pressure Photo- Ionization) and APCI (Atmospheric Pressure Chemical Ionization) sources, owing to its better performances in terms of signals to noise ratio and its robustness.
The full scan acquisition of each analyte was made by varying the fragmentor voltage in order to 1) establish the most abundant ion, 2) obtain the best response and 3) investigate if an in-source fragmentation is possible. Quasimolecular ions [M+H]+ were found to be the most abundant ones, being sodium and potassium adducts virtually absent. In two cases (Sildenafil and Pseudovardenafil) double charged molecular ions [M+H]2+ were present, but [M+H]+ were always chosen as precursor ions. No fragmentation was experienced in the range of fragmentor voltages tested, and a series of MS/MS experiments were implemented by varying the collision energy (CE) by 5 units (Figure 16) with the aim to ensure the largest response for both precursor and qualifier ions. The best collision energy was selected by the intersection of the fitting line of the quasimolecular ion to the one belonging to the qualifier ion. MS and MS/MS scan rate was also optimized accordingly in order to have at the same time the best response and a sufficient peak reconstruction. Table 9 shows the optimized fragmentor voltages, collision energies and related precursor and qualifier ions in MS/MS acquisition for the target compounds.
Table 9. Retention times, theoretical m/z for molecular and qualifier ions and optimized collision energies.
Compound | Molecular formula | RT | Molecular ion | Qualifier ion | Optimize d CE |
(min) | (m/z) | (m/z) | (eV) | ||
Yohimbine | C21N26N2O3 | 9.9 | 355.2016 | 144.0817 | 24 |
Tadalafil | C22N19N3O4 | 15.5 | 390.1448 | 268.1012 | 3 |
Pseudovardenafil | C22N29N5O4S | 16.6 | 460.2013 | 151.0826 | 36 |
Sildenafil | C22N30N6O4S | 11.4 | 475.2122 | 58.0666 | 32 |
Vardenafil | C23N32N6O4S | 10.5 | 489.2279 | 151.0874 | 37 |
Homosildenafil | C23N32N6O4S | 11.7 | 489.2279 | 72.0813 | 34 |
Hydroxyhomosildenafil | C23N32N6O5S | 11.2 | 505.2228 | 99.0924 | 30 |
7.1.1 Optimization of the chromatographic conditions.
The optimization of chromatographic conditions was initially performed using a C18 column (Kinetex, 150mm × 2.0 mm I.D.; 2.6 μm particle size from Phenomenex, USA), thermostated at 30°C and using water and acetonitrile acidified eluent (25 mM FA) and a linear gradient from 10% of acetonitrile to 100% in 24 min. Signals of the composite working standard solution at 100 ng/mL were registered in full scan mode.
Heavy and persistent tailing end carryover phenomena limited to Vardenafil were always experimented, evidencing the presence of specific interactions between this analyte and the stationary phase, probably arising from the presence of free silanolic groups or metal impurities on the surface of the C18-phase. The extent of carryover phenomena was estimated to be 3 ng/mL equivalent in this condition.
In order to evaluate and reduce these interactions, the concentration of formic acid was varied in both water and acetonitrile from 10 mM to 40 mM, without success. Thus, trifluoroacetic acid (TFA) 0.1% v/v and pentafluoroproprionic acid (PFPA) 5mM was employed as ion pair agents in mobile phases. Anyway, while the peak tailing was eliminated, both modifiers did not reduce the extent of the carryover effect. Moreover, ESI response significantly decreased of about 15-40% for all analytes.
The active interactions between Vardenafil and the stationary phase were completely overcome by the use of a polymeric column PolimerX™ (Figure 17). The absence of silanols ensured no carryover phenomena using both TFA and FA as acidifiers in mobile phases. Formic acid was however chosen because it produces a greater response in ESI. In this optimized chromatographic conditions, the highest sensitivity and selectivity, in terms of chromatographic separation of the isobaric Homosildenafil and
Vardenafil, were reached. Figure 18 shows the chromatographic profiles and relative MS/MS xxxxxxx for target analytes at concentrations of 10 times LODs estimated.
Figure 16. Relative daughter ion abundances vs collision energy (CE) energy for MS/MS experiments conducted with individual standard solutions at 100 ng/µL.
Figure 17. Chromatographic profiles for Vardenafil and Homosildenafil (isobaric, m/z 489.2279) using different columns and eluent modificants:
7.1.2 Optimization of the extraction conditions.
The sample preparation for trace analysis often involves several steps of clean-up, and enrichment with solid phase extraction. In the proposed method, the procedure is drastically simplified as it involves only a solvent extraction followed by sonication and centrifugation. The extraction parameters were tested in different conditions with a twofold purpose: at large concentration levels, typically encountered in the pharmacologic dosage, to cope with the counterfeit drugs, and at trace levels for analysis of cross contaminations.
Since certified reference materials were not available, the optimization of the extraction was made by using five commercial pharmaceutical formulations (see sample collection and preparation) of PDE-5 inhibitors, containing vardenafil, tadalafil, and sildenafil as API. Recoveries were calculated on the basis of the reported nominal amount of API and were expressed as concentration in the weighed matrix. The extracting solution (water, water:methanol 50:50, water:acetonitrile 50:50, all acidified with 0.1% FA), the sample amount (from 10 to 30 mg) and the extraction volume (from 1 to 2 mL) were optimized. Fig. 2 shows the percentage of recovery of the three synthetic API coming from the extraction of 10 mg of the relative milled tabs, varying the extraction solvent. The extraction of the most lipophilic tadalafil resulted very poor in acidified water, reaching the best yields with the water: acetonitrile solution. No difference in the extraction efficiency was observed for the different formulations of API. The optimization of the amount of the matrix and extraction volume was mainly evaluated in terms of efficiency and matrix effect. In ESI-MS technique, the matrix effect is described as the signal variation, mainly a suppression, of a compound in the ionization process due to the competition with matrix endogenous compounds. No further improvement of the extraction yields was obtained increasing the volume over 1.5 mL, whilst a significant matrix effect occurred when the amount of the sample extracted exceeded 15 mg (data not shown). Thus, the best conditions in terms of MME were obtained by extracting 10 mg of material with 1.5 mL of 50:50 water: acetonitrile solution acidified with 0.1% FA. However, when we tried to transfer this protocol for analysing bulk materials spiked at trace levels the same experimental conditions produced a quite relevant matrix effect. The slopes of the three-point calibration curves (1.0, 5.0 and 10.0 µg/g in matrix) obtained by spiking the five bulk materials after extraction using standard solutions have been compared. Results indicated that at low
concentrations, matrix components do interfere in the ESI ionization process, causing signal depletion, mainly for vardenafil and for the early eluted yohimbine in some bulk materials. Moreover, the extent of the matrix effect was sometimes dependent on the matrix composition. Extracts of pharmaceutical formulations did not produce that behaviour because the relative concentrations of the target analytes were much larger and taking into consideration the diluting factor. This drawback was anyway overcome using an extraction solution acidified with 1% FA, resulting both in a reduction of the ESI matrix effect and in the influence of the bulk composition. Therefore, it was possible to quantify the target analytes using a simple external calibration curve for all the considered herbal materials. Finally, no relevant changes in the extraction efficiency of the pharmaceutical formulations were experienced by using the extractant acidified with 1% FA instead of 0.1%. Figure 19 shows the percentage of recovery of the three synthetic API coming from the extraction of 10 mg of the relative milled tabs, varying the extraction solvent.
7.1.3 Validation.
The method selectivity was tested by comparison between five blank herbal matrices, not spiked and spiked at the lowest calibration level of the calibration curves. Mass accuracy obtained from the extracted currents of the quantifier [M+H]+ ions were always better than 5 ppm without interferences of the matrix. Selectivity is also guaranteed by the presence of the qualifier ions for all the concentration levels in the fragmentation spectra. As previously reported, no significant matrix effect was
experimented in the optimized conditions, so that any consideration about linearity and LODs were made on conventional standard solutions.
In ESI-MS technique, the matrix effect is described as the signal variation, mainly a suppression, of a compound in the ionization process due to the competition with matrix endogenous compounds. Furthermore, a general matrix effect is generally experimented in the extraction procedure. In the present work, the possible presence of ESI-matrix effects was studied by comparing the regression slopes related to standard calibration and matrix-matched calibration procedures, obtained spiking blank samples of herbal bulk after extraction at the same concentration reported above. Comparable slopes within the statistical errors indicated that target analytes were free of matrix ionization suppression effect, so that quantification of target analytes can be made by using conventional external standard calibration.
The linearity for the proposed method was evaluated for each target compound by making a seven point calibration curve with injected amounts of 2, 5, 10, 50, 100, 500, 1000 pg (n=3), respectively. Calibration plots were made in terms of peak areas vs amount of injected analyte. The main parameters of the calibration curves were very satisfying (Table 10), with R2 greater than 0.9991 for all target compounds and the relative residues always below 20%.
Table 10. Calibrations parameters and LODs for the proposed method.
Compound | calibration curve | R2 | LOD | LOQ |
(pg injected) | (pg injected) | |||
Yohimbine | y = 4794.2x + 5259 | 0.9999 | 1.04 | 3.11 |
Tadalafil | y = 372.5x - 598 | 0.9997 | 4.24 | 12.71 |
Pseudovardenafil | y = 6616.5x - 2867 | 0.9999 | 1.94 | 5.83 |
Sildenafil | y = 1808.2x - 198 | 0.9998 | 2.84 | 8.52 |
Vardenafil | y = 805.6x - 1548 | 0.9997 | 1.95 | 5.84 |
Homosildenafil | y = 2323.2x - 1920 | 0.9997 | 2.15 | 6.44 |
Hydroxyhomosildenafil | y = 1244.3x - 4332 | 0.9991 | 1.60 | 4.81 |
Limit of detection was rigorously calculated on the basis of a four-point calibration curve at low concentration (n=3) according to the Voigtmann method [87, 88]. As usual, LOQ was set = 3 LOD. LODs varying from 1 to 4 pg injected (corresponding to
30-120 ng/g in matrix) (Table 10) were considered satisfactory and suitable to detect target analytes even at trace levels.
Because of the unavailability of suitable CRMs, accuracy of the method was evaluated by spiking five different blank herbal formulations (Herbal powder formulation, gel for topic usage, and Herbal extract) at three concentration levels 1.0, 5.0, 10.0 µg/g of each target analytes (n=3, N=15). Trueness was computed as the mean value of thee replicates of each concentration level and ranged between 80.9% and 108.1% (Table 11). These values fulfil the criteria of 2002/657/EC and SANCO/10684/2009 guidelines [85, 86] that require recoveries in the range of 80–110% and 70–120%, respectively. Intraday repeatability was estimated by RDS values obtained from recoveries. Overall RSD values ranging from 2.7% and 10.8 % were considered very good, taking into consideration the variability of the five spiked matrices, and in accordance with validation guidelines.
1.0 µg/g | 5.0 µg/g | 10.0 µg/g | ||||
compound | Recovery (%) | Overall RSD (%) | Recovery (%) | Overall RSD (%) | Recovery (%) | Overall RSD (%) |
Yohimbine | 90.6% | 10.1% | 89.5% | 8.7% | 91.1% | 10.8% |
Tadalafil | 101.6% | 9.5% | 106.5% | 8.6% | 92.6% | 7.4% |
Pseudovardenafil | 91.0% | 4.3% | 96.1% | 3.2% | 93.6% | 3.4% |
Sildenafil | 92.6% | 4.7% | 95.0% | 10.7% | 95.4% | 9.1% |
Vardenafil | 94.8% | 2.7% | 80.9% | 3.1% | 87.4% | 8.0% |
Homosildenafil | 97.7% | 9.1% | 99.7% | 2.8% | 93.1% | 8.7% |
Hydroxyhomosildena fil | 108.1% | 5.2% | 102.1% | 4.1% | 100.3% | 5.5% |
7.2 Food supplement analysis
The proposed and validated method was applied to the analysis of 26 food supplement samples present in the Italian market. Two tablet samples resulted contaminated with sildenafil at the concentration barely above the respective LODs. EIC profiles for detected target analytes in experienced counterfeit samples with related MS/MS spectra are reported in Figure 21. The concentration levels are consistent with a cross contamination by authorized drugs, during storage or carriage, and are not able to cause
relevant adverse effects on human health. A capsule of a dietary supplement named Hero, bought in a web-store, was previously analysed using a LC-DAD system, resulting contaminated mainly by a compound with peak wavelengths at about 290 and
360 nm, characteristic of a thio derivative of Homosildenafil [31]. In 2008, FDA warned consumers this product containing an unapproved substance similar in chemical structure to sildenafil (xxxx://xxx.xxx.xxx/XxxxXxxxxx/Xxxxxxxx/XxxxxXxxxxxxxxxxxx/0000/xxx000000.xx
m). At that time, that compound was not identify also because no certified standards were available for the quantitative analysis. Further analysis was made on two other capsules of the same lot only in the 2014 after the optimization of the presented LC- HRMS method. In this case, the MFE post run analysis showed two relevant signals identified as Homosildenafil (experimental m/z 489.2273, Δm=-1.2 ppm) or an its isomer, and one of the three isobaric compounds having a theoretical m/z 505.2050, namely thiodimethylsildenafil, thiomethisosildenafil and thiohomosildenafil. The signal of this second compound was two orders of magnitude larger than that obtained for hypothetic Homosildenafil. Anyway, although retention time of the supposed Homosildenafil, which standard was available, was within the query set, the observed fragmentation pattern in the AutoMS/MS acquisition did not correspond exactly. Thus, we found a compound in the web-free database m/z Cloud™, (xxxxx://xxx.xxxxxxx.xxx/) exhibiting the same fragmentation pattern and identified as the isobaric dimethylsildenafil (compound N°699 fragmented at 40 NCE in HCD mode). Analogously, the compound with experimental [M+H]+ at m/z 505.2053 ( Δm=+0.6 ppm) could be tentatively ascribed to both thiodimethylsildenafil (compound N°955) or thiomethisosildenafil (compound N°957) by comparing the available m/z Cloud™-MS/MS spectra registered at 40 NCE in HCD mode. Since the MS/MS fingerprints of the two compounds were unnoticeable, and although the acquired UV spectra was identical to that described for thiodimethylsildenafil [106], it was not possible discriminating among these two isomers. EIC profiles for detected target analytes in experienced counterfeit sample of Hero with related MS/MS xxxxxxx are reported in Figure 20.
Figure 20. LC-MS xxxxxxxxxxxx acquired for the analysis of a counterfeit sample of Hero with the related MS/MS xxxxxxx. EIC profiles of the compounds with theoretical m/z at a 489.2279 and b 505.2050.
After ascertaining that contamination was due to the inner bulk and not to the capsule shell, a semi-quantitation was made by diluting the extract by a factor 1000 like to the
authentic pharmaceutical formulations, and assuming the same molar response of the Homosildenafil. The assessed concentration for this single capsule is 0.25±0.02 µg/mg for dimethylsildenafil and 59.02±0.36 µg/mg for thiodimethylsildenafil or thiomethisosildenafil. The correlated dosage (0.11 mg/capsule of dimethylsildenafil +
25.12 mg/capsule of the thiodimethylsildenafil or thiomethisosildenafil), coherent with a pharmacological one and therefore may represent a concrete health risk for the unaware customers assuming this “natural herbal” food supplement. Fina ly, no other suspect PDE-5 inhibitor analogues resulted from data analysis made with the internal database, using the described MFE features.
7.3 Conclusion
The proposed method was proved to be very simple and robust, suitable for both screening and confirmatory purpose, allowing the identification of suspect PDE-5 analogues and the quantitation of the analytes with a single chromatographic run. The entire analytical process can be performed in less than one hour. A complete validation of the procedure was carried out, demonstrating a very high extraction efficiency, reproducibility and selectivity without incurring in any relevant matrix effect. The
performance of the extraction method makes possible both the determination of PDE-5 inhibitor analogues at high concentration for counterfeit analysis, and at trace levels for cross contaminations. The optimized method was successfully applied to the analysis of
26 real samples of natural dietary supplements and herbal remedies marketed for erectile dysfunctions. Three samples were found to be contaminated with synthetic PDE-5 inhibitors, both approved and unregistered. The possibility of further
improvements of this procedure enlarging the database, even with fragmentation spectra, allows this method to be easily adopted by health agencies to contrast the illicit misuse of synthetic PDE-5 analogues.
8.1 Introduction
XXXXXX network is a non-profit European association that through several initiatives would provide a help to the legislative organs in the field of environmental contaminants. It started its activities in September 2005 with the financial support of the European Commission.
The main objective are:
⮚ Measurement methods harmonization for a better monitoring and risk assessment.
⮚ Enhance the exchange of information and data on environmental emerging substances.
⮚ Promote the maintaining and developing of knowledge of emerging pollutants
stimulating interdisciplinary projects on problem-oriented research and knowledge transfer.
XXXXXX network organizes many activities spacing throughout expert group meetings, workshops, databases and methods validation trials. As long-term goal, the association is also active in improving the identification of environmental unknown compounds and prioritizing of emerging substances. In August 2012, XXXXXX started a successful cooperation with the web-mass library “MassBank” (xxxx://xxx.xxxxxxxx.xx) in order to fund a mass spectrum library focused on environmental pollutants. Coherently with the web vision, all the spectral information included in the database are addressed to improve the identification of unknowns and the access is free-of-charge.
In recent years, improvements of the analytical techniques have driven the interest of aquatic environmental scientist to the determination of organic pollutants at ultratrace levels. The scientific literature has depicted a scenario with a more and more number of compounds recognized in water, implying that the target analysis will be no longer sufficient to provide an exhaustive representation of the pollution status of the water bodies. On the other hand, the non-target analysis, needed to detect as many harmful substances as possible, is not harmonized and affected by a wide variability in the method implementation, making difficult the comparison of the obtained results.
In order to respond to this drawback, in 2013 XXXXXX organized the first collaborative trial on non-target screening. The exercise involved many laboratories
across Europe (including the University of Padua) and contemplated the non-target analysis of the compounds present on a river water sample. A workshop has followed the trial with the aim of share the experiences, discuss the results, get to an agreement on harmonized terminology and workflows and, finally, get a proposal for further actions to be promoted in the field of non-target screening.
8.2 Trial results.
8.2.1 Target approach and suspect approach.
The analysis of the water extract was conducted using the “two step” protocol described in the section materials and methods. The column used was a Kinetex C18 (2.1X10 mm,
2.6 μm, Phenomenex) end the eluents were H2O and AcN with 0.1% formic acid for positive mode and 0.1% NH4OH for negative mode. The use of acidic and basic modificant was chosen to promote the ionization efficiently in the corresponding polarity. Injection of 40µL was perform to detect compounds at trace level.
A preliminary chromatographic run was performed in both positive and negative full scan MS, and the pseudomolecular m/z values of the available standards were used to extract EIC profiles with 20 ppm of mass tolerance. The presence of chromatographic peaks with significant S/N ratio in the EIC profiles was considered as a possible positive result.
At the moment of the collaborative trial, all compounds hade to be reported as target, suspect, non-target or unknown in a common excel spreadsheet (Table S 4) with all experimental and MS information useful for supporting the identification and comparing the results. This classification resulted vague when a comparison among different laboratories was made, and for the sake of simplicity the harmonized definition introduced in the Introduction chapter and proposed after the XXXXXX workshop was here used.
The identification at level 1 was confirmed by comparison of retention times, and fragmentation spectra with the pure standards analysed in the same conditions. All the other non-confirmed possible positive results were subjected to MS/MS analysis, and then they were classified as non-target compounds with an identification confidence level from 2 to 5, depending on the structural information disused by the fragmentation spectra and on the proposal of a unique molecular formula.
Positive and confirmed target compounds were for perfluoroalkyl compounds, described as endocrine disruptors: perfluorooctanoic acid, perfluorooctansulfonic acid, perfuorohexanesulfonic acid and perfluorobutanesulfonic acid. As it was not possible to use surrogate standards nor to evaluate the matrix effect, a semiquantitation was attempted showing they were in the range of ng/L or sub-ng/L.
The suspect screening was approached on full scan analysis, searching for a limited number of compounds suggested by the XXXXXX association (Figure 22) on the basis of previous monitoring campaigns, and employing libraries available for us: Pesticides (1600 entries), Forensic Tox (7300 entries), Synthesis (16,000 entries) from vendors, and in-house databases, i.e. cyanotoxins and PDE-5 inhibitors.
The libraries have been implemented in Molecular Features Extraction (MFE) setting the following parameters and thresholds: ion compound filters ≥ 1000 in MS level, and MME ≤ 20 ppm.
Positive results were carefully checked to eliminate false positive results due to:
a) ions recognized as salts adduct of the databases entries, but actually related to different signals;
b) ions clearly referable to series of homologues compounds, e.g. differing for an ethoxylate group;
c) ions belonging to the isotopic pattern of other signals;
d) ions ascribable to other compounds loosing water or ammonia.
f) signals close to dead time of the chromatographic run (tR=0-2 min)
Compounds unlikely to be present were deleted. The software outputted as sum of the suspect of the different libraries and polarity, 167 possible structures. Notable was the failing of the library “Synthetic” because of the low performance of the computer in processing such big amount of data, thus limiting our suspect screening capability.
The m/z values of the positive results were first used to extract EIC profile from the blank sample to manually check the presence of the same signal detected in the sample. Then, the m/z values converted in an inclusion list in AutoMS/MS experiments. The CID energy was set at 40eV, which was relatively high and ensured a high rate of fragmentation with several low mass fragments. A maximum number of MS/MS experiment per cycle of 8 and an exclusion time of 30 s was adopted to detect as many target as possible and ensure sensitivity and a good peak reconstruction.
The MS/MS spectra were evaluated by comparison with the spectra present in on-line web databases, especially MassBank and MzCloud. If sufficient structural evidences were provided by the fragmentation spectra, the substance was identified as level 2. Belonging to this class were 13 results (Table 12). When MS/MS spectra interpretation was not able to identify a unique structure, providing only information about the chemical class, the proposed identification was at level 3. Three structures were tentatively identified in this way: dihydroxy-octadecenoic acid, hydroxy-octadecenoic acid and dihydroxy-linoleic acid. For these compounds, despite the presence of characteristic fragments identifying the functional groups, the relative position of the hydroxylation was not inferable by the MS/MS analysis.
Finally, positive results whose assignments were not confirmed by MS/MS, were declassified as non-target compound and treated with a different approach.
level 1 | level 2 | level 3 | level 4 | level 5 | |
target | 4 | ||||
suspect | 13 | 3 | |||
non-target | 17 | 6 | 57 |
8.2.2 Non-target approach
Non-target analysis was primary focused on the best peak chromatogram (BPC) in which the profile is the current ion of the most intense signals. This approach was chosen for the manually detection of the ions present. From the MS spectra of the chromatography peaks, the most intense signals were selected and, after a raw evaluation of the isotopic pattern, for each ones:
a) The molecular formula was generated, setting the software algorithm with a limited number of carbon, hydrogen, oxygen, nitrogen, sulphur, chlorine and bromine.
b) The m/z value was input in an inclusion list to perform the target MS/MS experiments.
Target MS/MS was conducted setting the same parameters as suspect screening. The majority of non-target results were at level 5 (57 mass of interest) because the formula generated had an non acceptable high mass error or was likely to be non-natural.
Twenty-one unique molecular formulas were successful characterized and for 17 of them a tentative identification was proposed at level 3 by evidences arising by MS/MS experiments. Some tentative identifications were achieved taking advantage of the retention time standardization through the RTI, and comparing results with the RTI database implemented by some researchers belonging to the XXXXXX group.
8.2.3 Retention time index (RTI)
An important objective of the trial was to assess the use of retention time information in the LC screening approaches.
Table 13. Retention times for the analyte for the RTI calculation in positive and negative modes.
substance name | Monoisotopic MW (Da) | logP | Tr (min) Positive detection | Tr (min) Negative detection |
Metformin | 129.10143 | -1.36 | 1.1 | |
Chloridazon | 221.03558 | 1.11 | 10.6 | 9.5 |
Carbetamide | 236.11608 | 1.65 | 12.5 | 11.28 |
Monuron | 198.05598 | 1.93 | 13.0 | 11.86 |
Metobromuron | 258.00043 | 2.24 | 15.4 | 14.18 |
Chlorbromuron | 291.96146 | 2.85 | 17.3 | 16.06 |
Metconazole | 319.96146 | 3.59 | 18.5 | |
Diazinon | 304.10104 | 4.19 | 19.9 | |
Quinoxyfen | 306.99669 | 4.98 | 20.9 | |
Fenofibrate | 360.11282 | 5.28 | 22.4 |
We measured the substances mixture provided by the XXXXXX foundation in the described experimental conditions in both the ESI polarities. Table 13 reports the detected compounds and respective retention times. It is possible to note that the majority of substances used were ionized only in positive mode, while in negative mode only five of these have been efficiently ionized. The retention times were plotted against the logP (Figure 23) of the respective compounds shared within the XXXXXX material, and the linear fitting of the data was done. The equation obtained was used to calculate a retention time index (RTI) for the LC compounds identified. These data was used according the instruction furnished in the RTI database, which performance was tested in the frame of this collaborative trial. The database collects some thousands of environmental substances with RTI, CAS number, structures, commercial use and other information. On the basis of the experimental chromatographic conditions used and RTI
inserted, the database propose some compounds listed with the respective MS accuracies, which have to be anyway evaluated in terms of MS/MS fragmentation (Table S 4)
Figure 23. RTI calibration curve for positive and negative mode and linear fitting (dotted line).
8.3 Trial consideration and Conclusions
The laboratories participants to the trial had different backgrounds; either experienced in non-target analysis either novice in the field, attending the screening analysis for the first time. For our group was in absolute the first attempt in such comprehensive non- target analysis, and this fact was reflected in the relative low number of submitted results of non-target compound.
The used instrumentation setting was roughly comparable with the rest of the trial’s participants (Table S 2 and Table S 3). The adopted “two-step” protocol was exce lent for the target analysis. In fact, this approach allowed a fast, easy and reliable confirmation of the structure. The unique limitation of the target screening we have experimented was the very limited availability of analytical standards.
For the suspect screening the limitation was the non-availability of a database containing specific water contaminants, so that we used vendor’s libraries in which most of the compounds were non likely to be present in river water. Structural investigation using target MS/MS feature was resulted critic to increase the identification level of the structure. Q-TOF system is known for the speed in acquisition and this allows the acquisition of many MS/MS spectra at the same time. Despite this, a partial sensitivity loss was experimented with the increasing number of fragmentation per cycle, thus limiting the overall capability in structural confirmation, especially when analytes were at ultratrace levels. To avoid this drawback, similarly to all the trial participants, we adopted a quite long chromatographic run in order to give a sufficient time to the analyser for performing the MS/MS experiments and having a good peak reconstruction in EIC spectra.
In non-target analysis, in addition to the same issues as suspect screening, the generation of a unique and reliable molecular formula seems to be the limiting step. As expressed by most of the participants, non-target analysis require specific software that are non-commercially available able to treat big dataset and produce reliable results. Finally, the non-target screening was tedious and highly time-consuming because, without any previous experience in the field, and the non-availability of specific software required a big amount of manual work. The needing of much more time for the data treating was a common feeling of the trial participants, suggesting that it is currently impossible adopting such non-target approach for the routinely analysis of the water contaminants, although a rough harmonization of the analytical protocol could be certainly indicated.
9 Development of a workflow for HRMS analysis of PM2.5 organic fraction: post-run data analysis and the role of ionization sources.
9.1 APPI analysis optimization.
9.1.1 Source parameters optimization
Initial optimization of the APPI measures was focused on the source parameters using a standard mixture of PAHs, Nitro-PAHs and O-PAHs in a methanol:dichloromethane 1:1 solution. The concentration levels (Table 4) were in the range 6-133μg/mL for PAHs, 0.6-5.3μg/mL for Nitro-PAHs and 0.13-13 μg/mL for O-PAHs.
The standard mixture was analysed in direct infusion at different temperatures from 50 to 350ºC setting the mass range to m/z 100-650 and the flow at 10 μL/min. For each temperature, after the stabilization of conditions, the mass spectra were recorded for 30 seconds. For each measure, the overall intensity and spray stability were assessed through the average of the single TIC values and its standard deviation respectively. As Figure 24 shows, with the increasing of the source temperature both TIC values and related variances increased. More specifically, looking at the responses of single analytes of the mixture (data not reported), for two nitro-PAHs (4-nitrocatechol and 4- nitrophenol) and one O-PAH (4-phenanthrenecarboxaldehyde) a decreasing in intensity was experimented over 200-250°C, indicating a thermal decomposition (Figure 25). The spray stability dramatically decreased over 200°C as represented by the standard deviations values. Thus, a temperature of 200°C was chosen for the APPI analysis in order to prevent the loss of Nitro and oxidized PAH, having and at the same time a good overall sensitivity and spray stability.
5.0E+08
4.5E+08
4.0E+08
3.5E+08
3.0E+08
2.5E+08
2.0E+08
1.5E+08
1.0E+08
5.0E+07
0.0E+00
• TIC
0 50 100 150 200
250
300
350
400
Source temperature (°C)
1.E+07
9.E+06
8.E+06
7.E+06
6.E+06
5.E+06
4.E+06
3.E+06
2.E+06
1.E+06
0.E+00
4-nitrocatechol (M+H)
4-nitrophenol (M+H)
9-phenanthrenecarboxaldehyde (M+H)
25 75 125 175 225 275
325
375
Source emperature (°C)
Intensity (counts)
Intensity (counts)
Figure 24. TIC intensity trend with the APPI source temperature.
Figure 25. EIC intensity trend of selected nitro-PAHs and O-PAHs with the APPI source temperature.
An analogue procedure was followed for the gas flows optimization. The histogram plot in
Figure 26 shows the TIC average values with standard deviations for each tested pair of auxiliary gas and sweep gas. Flow rate of 5 and 10 arbitrary units for auxiliary gas and sweep gas were chosen respectively because representing a good compromise between response and spray stability.
3.E+09
2.E+09
2.E+09
1.E+09
5.E+08
0.E+00
5;5 10;5 15;5 20;5 5;5 5;10 5;15 7;5 7;10
Gas Flow Rate (Aux; Sweep)
Total current (counts)
9.1.2 Dopant optimization
Toluene and acetone were tested as dopant agents at concentration of 5 and 10% (v/v) in the mobile phase for improving APPI ionization efficiency. The best results were showed by using toluene at 10%. A complete table of intensities is reported in the appendices (Table S 6), and the Table 14 reports the sums of the intensities of the [M+H]+ and [M∙]+• ions and the relative percentage with respect to the higher signal registered. As general observations:
⮚ Toluene at concentration of 10% produced the best results in terms of intensity of the standard’s signals for a l the class compounds.
⮚ PAHs formed manly molecular ions, while Nitro-PAHs and O-PAHs preferentially generated quasimolecular ions [M+H]+. 4-nitrocatechol and 4-nitrophenol formed only quasimolecular ions (Figure 27).
⮚ 4-nitrocatechol signal was registered only using toluene, with the best results experienced with 10% toluene with respect to 5% (about an half of the signal), whilst acetone never produced an appreciable signal.
Compound class | Toluene 10% | Toluene 5% | Acetone 10% | Acetone 5% | |
Quasimolecular ions (M+H+) | PAH | 1.08*108 | 7.66*107 | 1.04*107 | 1.28*107 |
100% | 71% | 10% | 12% | ||
Nitro-PAH | 1.23*107 | 7.72*106 | 3.21*105 | 4.80*105 | |
100% | 63% | 3% | 4% | ||
O-PAH | 1.34*108 | 8.55*107 | 6.29*107 | 6.26*107 | |
100% | 64% | 47% | 47% | ||
Total | 2.54*108 | 1.70*108 | 7.36*107 | 7.58*107 | |
100% | 67% | 29% | 30% | ||
Molecular ions (M+) | PAH | 6.56*108 | 3.79*108 | 3.11*107 | 3.65*107 |
100% | 58% | 5% | 6% | ||
Nitro-PAH | 1.26*107 | 5.18*106 | 4.39*105 | 5.97*105 | |
100% | 41% | 3% | 5% | ||
O-PAH | 3.27*106 | 1.58*106 | 9.84*105 | 9.09*105 | |
100% | 48% | 30% | 28% | ||
Total | 6.72*108 | 3.86*108 | 3.26*107 | 3.80*107 | |
100% | 57% | 5% | 6% | ||
All Ions | 9.26*108 | 5.56*108 | 1.06*108 | 1.14*108 | |
100% | 60% | 11% | 12% | ||
100%
Toluene 10%
Acetone 10%
80%
60%
40%
20%
0%
M+H/M Rate
Figure 27. Intensity ratio between quasimolecular [M+H]+ and molecular ions [M]+• using toluene 10% and acetone 10%
9.1.3 Recovery study on spiked blank filters.
The study of the recovery for the extraction procedure was carried out spiking Teflon blank filters at concentrations of PAHs close to those expected in real samples. As reported in literature, the average winter concentration of the sum of PAHs in north Italy is in the range 20-50 ng*m-3 and Benzo[α]pyrene (B(α)P) represent the 17% of the total amount (3.6-8.5 ng m-3).
Two pieces of 1/8 of filter were spiked with 20 µL of the stock solution and then one was extracted using 5X15 mL of methanol and the other using the same volume of methanol:dichloromethane 1:1. The 15 mL of extracts were evaporated under a gentle nitrogen flow until a final volume of about 3 mL. The obtained extracts were filtered using 0.22 µm Teflon filter and divided in two aliquots of 1.5 mL. For each aliquot, extract one aliquot is evaporated to dryness and then reconstructed using 1.5 mL of the same solvent. The standard solution was obtained by dilution of 20 µL in a final volume of 3 of the corresponding solvent.
The Table S 7 reports all the intensities of the detected ions obtained with the different extraction solvents, while Table 15 summarizes the experimental highlights.
Number of detected Ions | |||||
St. Mix Methanol | Spiked sample Methanol | Spiked sample Methanol reconstituted | St. Mix Methanol:CH2Cl2 | Spiked sample Methanol:CH2Cl2 | Spiked sample Methanol: CH2Cl2 reconstituted |
26 | 29 | 21 | 28 | 25 | 21 |
average values of recoveries | |||||
Ratio St. Mix Methanol/ CH2Cl2: Methanol | Recovery Methanol extraction | Recovery CH2Cl2: Methanol extraction | Recovery Methanol reconstituted extraction | Recovery Methanol: CH2Cl2 reconstituted extraction | Ratio Sample extraction Methanol/ CH2Cl2:Methanol |
78% | 105% | 57% | 47% | 31% | 122% |
Considering at first instance the number of compounds found in the mass spectra of the different sample preparations, the extraction using methanol allowed to recognize a greater numbers of ions with respect of the extraction using the mixture methanol:dichloromethane 1:1 (29 instead 28). Although the standard solution gave a
better response in methanol:dichloromethane 1:1, the mean recoveries was higher for the methanol extract.
1000
900
800
700
600
500
400
300
200
100
0
Total
CH
XXX XXXX CHN
Positive
Total
CH
XXX XXXX
negative
CHN
Compound class
CH2Cl2-Methanol recostructed
CH2Cl2-Methanol
Methanol
Number of detected molecular formulas
The same extraction conditions were used on aliquots of one quarter of filter from a winter sample (code FP-1) and analysed both in positive and negative modes in order to assess the best protocol in terms of number of molecular formulas obtained from data analysis, as common ions of triplicate measures (Figure 28). Among the three organic extractants evaluated, methanol showed the highest performance in terms of ability to elute compounds generating a larger number of molecular formulas. These results can be explained by a better solubility of the compounds present in the aerosols.
Figure 28. Number of inferred molecular formulas using different extractant solution.
9.2 Algorithm development
9.2.1 Previous algorithms and criticisms
Two different algorithms for positive and negative nanoESI molecular formula filtering were previously developed in the centre of Atmospheric Science group of the University of Cambridge (xxxx://xxx.xxx.xx.xxx.xx.xx/). As inputs of the algorithms, a table containing the name of the sample, its mass drift range, noise level and “signal to noise” had to be provided. The mass error ranges were calculated looking for at least five known contaminants present in the MS measures, taking their mass errors and adding and subtracting 0.5 ppm respectively to the maximum and minimum value of the set, obtaining in this way the “lowerppmLimit” and “upperppmLimit”.
The noise was calculated sampling it manually in three different regions of the mass spectra, calculating the average and standard deviation and finally adding three times the standard deviation to the mean. The “signaltonoise” value was used for discriminating signals to be ascribed to samples from blanks, and entries with values below five was deleted.
The original algorithm performed the following steps both on the sample and on the blank:
1. Importation of the raw .csv table and addition of 17 columns named: C; H; N, S, O, 13C, 34S, Na, Xxxx x.Mass, DBE, N comp, C, H, N, S, O, 13C, 34S. The new table formed has 25 columns and it was called xcalraw.
2. The algorithm for each row takes the molecular formulas, that have the general structure of XXXXXXx, and splits them writing in the corresponding columns the number each elements present in the formula.
3. The algorithm selected only the molecular formulas with: mass errors >ppm-lower; mass error <ppm-upper;12C+13C > 0; 4) H > 1; H(+1 if Na adduct)/( 12C+13C) < H/Cmax; H(+1 if Na adduct)/( 12C+13C) > H/Cmin; N/(12C+13C) < N/Cmax; N/(12C+13C) > N/Cmin; O/(12C+13C) <O/Cmax; O/(12C+13C) >O/Cmin; S/(C+13C)
<S/Cmax.
4. The algorithm copied the elemental composition in the columns corresponding to the neutral molecular formulas. For the concerning of hydrogen the number is decreased by 1 or in the case of Na is present no subtraction is performed. The model considers all the ions M+H or M+Na.