Cyanobacterial microcystins and human health – First steps towards a congener dependent risk assessment of microcystins
Dissertation zur Erlangung des
akademischen Grades
eines Doktors der Naturwissenschaften (Dr.rer.nat.)
vorgelegt von
Altaner, Stefan
an der
Mathematisch-Naturwissenschaftliche Sektion
Fachbereich Biologie
Konstanz, 2019
Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-2-uajiqfo13z2m0
Tag der mündlichen Prüfung: 19.07.2019
1. Referent: Prof. Dr. Daniel R. Dietrich
2. Referent: Prof. Dr. Valentin Wittmann
Für meine Familie. Angela, Bruno, Bernhard, Manuela und Marie.
Wer immer liegen bleibt, liegt falsch.
(Broilers – Stoßen wir an)
Danksagung
Danksagung Mein Dank gilt Daniel Dietrich für die Bereitstellung des Promotionsthemas, für die Unterstützung beim Schreiben verschiedener Stipendienanträge, für die fachlichen Diskussionen und für das Möglich-Machen meines Forschungsaufenthalts am Cawthron Institute in Neuseeland. Mein weiterer Dank gilt Valentin Wittmann für die Übernahme des Zweitgutachtens, aber auch für die Betreuung und Mitarbeit an mehreren Manuskripten. Danke auch an Aswin Mangerich für die Bereitschaft als Prüfungsvorsitzender bereit zu stehen und die Möglichkeit, meine UPLC-MS/MS-Arbeiten in der AG für Molekulare Toxikologie durchführen zu können. Besonderer Dank gilt der ganzen AG Dietrich! Danke an meine ‚Laborfrau‘ Heinke für all die fachlichen und nichtfachlichen Unterhaltungen, die mir die Zeit kurzweilig gemacht haben. Danke an Marci für die besten Protokolle. Ich habe mich immer gefreut, wenn ich etwas gemacht habe, dass du vor mir schon mal gemacht hast. Danke an Barbara und Phil für viele Stunden sportlicher Ertüchtigung und sonstiger Ablenkung vom Laboralltag. Danke an Kevin, Nadja, Feli und Lisanne für eine immer gute Atmosphäre in der AG. Danke an Sascha, Pia, Karin, Sabine und Alex für die ganze Arbeit im „Hintergrund“, die gemacht hat, dass die AG funktioniert. Weiterer Dank gilt allen Studenten, die ich betreuen durfte und die mir tatkräftig bei dieser Arbeit geholfen haben: Jahn, Felix, Regina, Eva, Kevin, Helena, Julia, Lisa, Michail. Ein spezielles Dankeschön soll hierbei Regina für die unzähligen Phosphatase-Assays gelten, Jahn für seine unglaubliche Eigeninitiative in einem Projekt, von dem ich eigentlich keine Ahnung hatte und Eva für all das Klonieren der widerspenstigen Phosphatasen. Auch möchte in den Mitarbeitern des Cawthron Instituts für die schöne Arbeitsatmosphäre im Institut danken und ganz speziell Jonathan Puddick, der sich die Zeit genommen hat, mich zu betreuen und mich in die Geheimnisse der UPLC-MS/MS-Analytik einzuweihen. Danke an Tabea Zubel, die in Konstanz immer für meine Fragen bezüglich UPLC- MS/MS ein offenes Ohr hatte. Darüber hinaus möchte ich allen Ko-Autoren der Paper und Manuskripte danken, die alle dazu beigetragen haben, diese Werke zu verbessern. Speziell sei hier Sabrina Jaeger erwähnt, die sich immer die Zeit nahm, um nochmal eine Berechnung laufen zu lassen. Ein weiterer spezieller Dank gilt der Arthur-und-Aenne-Feindt-Stiftung und all ihren Mitarbeitern für die finanzielle Unterstützung meinerseits und des ganzen Projektes. Ich bin
I Danksagung sehr dankbar über die unkomplizierte Art und Weise und die Verlängerungen der Stipendienzeit! Zuletzt möchte ich den wichtigsten und herzlichsten Dank an meine Familie aussprechen, denen ich diese Arbeit widmen möchte. Danke Mama und Papa für die bedingungslose Unterstützung, nicht nur während des Studiums, sondern auch allem was davor war und noch kommen wird. Danke an Bernhard für alle Gespräche über Wissenschaft oder sonstiges. Du warst und bist mir ein Vorbild, wissenschaftlich, wie auch im „richtigen Leben“. Danke an Manu, ohne dich wäre ich nicht, was ich bin und nicht komplett. Danke für deine Liebe und deine Unterstützung, für das Ertragen meiner Launen, für die Motivation in jeder Lebenslage, für wunderschöne zwölfeinhalb Jahre (es mögen noch ganz, ganz, ganz, (…) ganz viele weitere folgen) und vor allem Danke dafür, dass du mir die kleine Marie geschenkt hast! „You are my lighthouse in the dark. I can see you from a thousand miles” (In Flames – Here until forever). Danke auch an Marie, die mich immer wieder motiviert hat, wenn ich mal wieder die Schnauze voll hatte vom Schreiben .
II Table of content
Table of content
DANKSAGUNG I
TABLE OF CONTENT III
LIST OF FIGURES VII
LIST OF TABLES VIII
LIST OF SUPPLEMENTARY FIGURES IX
LIST OF SUPPLEMENTARY TABLES IX
ABBREVIATIONS XI
ABSTRACT XVI
ZUSAMMENFASSUNG XVIII
1. GENERAL INTRODUCTION 1 1.1 Cyanobacteria ...... 1 1.1.1 Ecology of cyanobacteria ...... 1 1.1.2 Cyanotoxins ...... 2 1.2 Microcystin ...... 4 1.2.1 Chemical structure ...... 4 1.2.2 Biosynthesis of microcystins and ecological role ...... 5 1.3 Toxicokinetics of microcystins ...... 7 1.4 Toxicodynamics of microcystins ...... 10 1.4.1 Excursus: The ser/thr-PPP family in mammals ...... 10 1.4.2 Consequences of microcystin exposure ...... 12 1.4.3 Proposed interaction sites of PPPs and MCs and mechanism of inhibition ...... 14 1.5 Risk assessment for microcystins and case reports of (human) intoxications ...... 15 1.5.1 Risk assessment ...... 15 1.5.2 Case reports ...... 17 1.6 Detection methods for microcystins ...... 19 2. AIM OF THE THESIS 21
3. MANUSCRIPT I (MICROCYSTIN ADSORPTION TO LAB-WARE) 23 3.1 Abstract ...... 23 3.2 Introduction ...... 24 3.3 Material and Methods ...... 26 3.3.1 Reagents and Laboratory-Ware ...... 26
III Table of content
3.3.2 Production of Microcystin Congener Stock Solution ...... 26 3.3.3 Adsorption of Microcystins to Common Pipetting Laboratory-Ware in Non-Acidified and Acidified Solvents ...... 27 3.3.4 Short-Term Storage in Glass or Polypropylene Vials ...... 27 3.3.5 Ultra-performance liquid Chromatography-Tandem Mass Spectrometry (UPLC- MS/MS) Detection of Microcystins ...... 27 3.3.6 Outlier Analysis ...... 28 3.3.7 Data Handling and Statistical Analyses ...... 28 3.4 Results ...... 29 3.4.1 Adsorption of Microcystin Congeners to Polypropylene Pipette Tips in Aqueous and High-Percentage Methanol Solutions ...... 29 3.4.2 Effect of Methanol Concentration on the Adsorption of Selected Microcystins to Polypropylene Pipette Tips ...... 31 3.4.3 Effect of Acidified Methanol Concentration on the Adsorption of Selected Microcystins (MC) to Polypropylene Pipette Tips ...... 32 3.4.4 Adsorption of Selected Microcystins (MC) in Acidified and Non-Acidified Aqueous Solutions to Glass-Ware (Pasteur Pipettes) ...... 33 3.4.5 Effect of Acidified Methanol Concentration on the Adsorption of Selected Microcystins to Glass-Ware (Pasteur Pipettes) ...... 34 3.4.6 Short Term Storage of MC Solutions in Glass or Polypropylene Vials ...... 35 3.5 Discussion ...... 36 3.6 Conclusions ...... 39 3.7 Supplementary material ...... 41 3.8 Acknowledgements ...... 47 4. MANUSCRIPT II (SYNTHESIS OF MC-LF AND DERIVATES) 48 4.1 Abstract ...... 48 4.2 Introduction ...... 49 4.3 Material and methods...... 52 4.3.1 General Experimental Methods ...... 52 4.3.2 General Procedures ...... 53 4.4 Results and discussion ...... 76 4.4.1 Retrosynthetic Analysis...... 76 4.4.2 Synthesis of Tetrapeptide 4...... 78 4.4.3 Synthesis of Dipeptides 5a-c...... 79 4.4.4 Synthesis of Linear Heptapeptides ...... 81 4.4.5 Macrocyclization and Final Steps ...... 82 4.5 Conclusion ...... 84
IV Table of content
4.6 Supplementary material ...... 85 4.7 Acknowledgements ...... 85 5. MANUSCRIPT III (UPLC-MS/MS DETECTION OF MICROCYSTINS) 86 5.1 Abstract ...... 86 5.2 Introduction ...... 87 5.3 Results ...... 90 5.3.1 Method establishment and optimization ...... 90 5.3.2 Use of internal standard ...... 94 5.3.3 Method validation ...... 96 5.3.4 MC levels in exposed mice ...... 99 5.4 Discussion ...... 100 5.5 Material and methods...... 104 5.5.1 Materials ...... 104 5.5.2 Sample generation for the establishment and validation ...... 105 5.5.3 Extraction method for MC from blood and liver tissue samples ...... 106 5.5.4 UPLC-MS/MS analysis ...... 106 5.5.5 Animal samples ...... 108 5.5.6 Data analyses and statistics ...... 109 5.6 Supplementary material ...... 110 5.7 Acknowledgements ...... 111 6. MANUSCRIPT IV (IN-SILICO PREDICTION OF MICROCYSTIN TOXICITY) 112 6.1 Abstract ...... 112 6.2 Introduction ...... 113 6.3 Material and methods...... 120 6.3.1 Materials ...... 120 6.3.2 Expression of 6xHis-hPPP5 in E.coli ...... 121 6.3.3 Purification of 6xHis-hPPP5 ...... 121 6.3.4 SDS-PAGE analysis ...... 122 6.3.5 Mass spectrometry ...... 122 6.3.6 Phosphatase activity assay ...... 123 6.3.7 Colorimetric protein phosphatase inhibition assay (cPPIA) ...... 123 6.3.8 Data analyses and statistics ...... 124 6.3.9 Machine learning (ML) ...... 124 6.4 Results ...... 127 6.5 Discussion ...... 132 6.6 Supplementary Material ...... 137 6.7 Acknowledgements ...... 149
V Table of content
7. GENERAL DISCUSSION 150 7.1 Recommendations for the handling of microcystin-containing samples ...... 150 7.2 Optimization of microcystin extraction from tissue ...... 151 7.3 UPLC-MS/MS analysis ...... 154 7.4 Serine-/Threonine-Proteinphosphatase (PPP) inhibition ...... 156 7.5 Computational chemistry and machine learning ...... 160 7.6 Risk assessment and determination of sample toxicity ...... 161 7.7 An example for the necessity of congener dependent observations ...... 163 8. CONCLUSIONS AND OUTLOOK 165
9. RECORD OF CONTRIBUTION 167
10. RECORD OF ACHIEVEMENTS 169
11. BIBLIOGRAPHY 171
VI List of figures
List of figures Figure 1-1: Cyanobacterial bloom in Lake Erie in 2003...... 2 Figure 1-2: A selection of cyanotoxins...... 3 Figure 1-3: General structure of microcystins...... 5 Figure 1-4: The families of ser/thr-phosphatases...... 11 Figure 1-5: Overlay of the catalytic subunits of PPP and PPP2A in complex with MC- LR...... 15 Figure 3-1: General structure of a microcystin...... 24 Figure 3-2: Reduction of various microcystin (MC) congeners in acidified and non-acidified solvents after increasing steps of pipetting using polypropylene pipette tips...... 30 Figure 3-3: Comparison of water and acidified water for the individual repeats of successive pipetting steps...... 31 Figure 3-4: Effect of methanol concentration on the adsorption of selected microcystins to polypropylene pipette tips...... 32 Figure 3-5: Effect of acidified methanol concentration on the adsorption of selected microcystins to polypropylene pipette tips...... 33 Figure 3-6: Multiple pipetting action using glass (Pasteur) pipettes...... 34 Figure 3-7: Effect of acidified methanol concentration on the adsorption of selected microcystins to glass-ware (Pasteur pipettes)...... 35 Figure 4-1: Graphical abstract to manuscript II...... 48 Figure 4-2: Microcystin-LF (MC-LF) and some amino acid variations (grey) of naturally occurring congeners...... 50 Figure 4-3: Proposed formation of aspartimide II and subsequent isomerization during the saponification of methyl ester protected MC-LA I ...... 51 Figure 4-4: Retrosynthetic analysis of MC derivatives 1a-c ...... 77 Figure 4-5: Synthesis of tetrapeptide fragment 4 ...... 78
Figure 4-6: Synthesis of Fmoc-D-MeAsp-Ot-Bu 7 ...... 79 Figure 4-7: A) Synthesis of alkyne labelled building block 8c. B) Synthesis of dipeptide fragments 5a-cb ...... 80 Figure 4-8: Synthesis of linear heptapeptides 22a-c by fragment couplingsb ...... 81 Figure 4-9: Deprotection and macrocyclization ...... 82 Figure 4-10: PPP1 inhibition assay with natural MC-LF and synthetic compounds 1a, 1c, and 23c...... 83
VII List of tables
Figure 5-1: General structure of microcystins where X and Z are variable l-amino acid positions...... 88 Figure 5-2: Scheme of the extraction highlighting different time points for spiking with MC congeners...... 91
Figure 5-3: Quantification of MC in spiked human serum using D5-MC-LF and D7-MC-LR as internal standards (IS)...... 96 Figure 5-4: Analysis of MC levels in plasma and livers of exposed mice...... 100 Figure 6-1: Consensus structure of microcystins and the synthetic variations produced for this study...... 114 Figure 6-2: Application of Word2vec to molecules and proteins...... 118 Figure 6-3: Workflow for feature generation for toxicity classification...... 119 Figure 6-4: After feature generation and pre-processing of the data, respective target values (toxicity class) are combined with the feature vector and used for a machine learning classification...... 120 Figure 6-5: Confusion matrix of microcystin toxicity prediction using 5-fold cross validation...... 131
List of tables Table 1-1: Distribution and substrates of human organic anion transporting polypeptides (OATP)...... 9 Table 1-2: Functions of PPP family members...... 12 Table 3-1: Microcystin congeners subjected to the experimental procedures described here...... 25 Table 3-2: Mass spectrometric parameters of the used microcystins (MC)...... 28 Table 3-3: Percentage of methanol needed to counteract the loss of microcystins from acidic and non-acidic solutions...... 39 Table 5-1: Congener dependent recovery during SPE and LLP steps of the extraction...... 92 Table 5-2: Total recovery of the extraction using different protein precipitation (PP) procedures...... 94 Table 5-3: Validation parameters of the established methods...... 97 Table 5-4: Intra-day and inter-day precision...... 98 Table 5-5: MS Parameters ...... 108
Table 6-1: IC50s of the tested MC congeners on rPPP1, hPPP2A and hPPP5...... 128 Table 6-2: “Toxicity” classes assigned to MC congeners...... 130
VIII
Table 6-3: MC congener toxicity equivalency factors (TEFs)...... 136 Table 7-1: Application of the TEF concept on a bloom sample from Hudson lake...... 163 Table 7-2: Application of the TEF concept on a bloom sample from Houghton lake...... 164
List of supplementary figures Figure S3-1: Reduction of all investigated microcystin (MC) congeners in solutions with decreasing methanol (MeOH) concentrations due to adsorption to polypropylene...... 41 Figure S3-2: Reduction of all investigated microcystin (MC) congeners in solutions with decreasing acidic methanol concentrations using polypropylene pipet tips...... 42 Figure S3-3: Reduction of various microcystin congeners in acidified and non-acidified solvents after increasing steps of pipetting using Pasteur pipettes...... 43 Figure S3-4: Reduction of all investigated microcystin (MC) congeners in solutions with decreasing acidic methanol (MeOH) concentrations using Pasteur pipettes...... 44 Figure S3-5: Effect of short term storage of microcystins in glass or polypropylene vials. .45 Figure S5-1: Column comparison using a middle-spike in serum...... 110 Figure S5-2: Distribution of microcystins in methanol and hexane fractions...... 110 Figure S5-3: Use of acetonitrile and methanol for protein precipitation...... 111 Figure S6-1: Calculation of prediction measurements. Precision is defined as the ability of a model to retrieve relevant results, recall as the ability of a model to retrieve relevant results and the F-Score as harmonic mean of precision and recall...... 137 Figure S6-2: Comparison of the hPPP5 expression with the empty expression...... 137 Figure S6-3: Inhibition curves for PPP1...... 138 Figure S6-4: Inhibition curves for PPP2A...... 138 Figure S6-5: Inhibition curves for PPP5...... 139
Figure S6-6: Comparison IC50s on the three tested phosphatases...... 140 Figure S6-7: Overlay of the structures of the used PPP...... 140
List of supplementary tables Table S3-1: Significance levels for Figure 3-2...... 46
IX List of supplementary tables
Table S3-2: Significance Levels for Figure S3-3...... 46 Table S6-1: Congener-dependent modifications...... 141 Table S6-2: Results of the mass spectrometric analysis...... 142 Table S6-3: Summary of settings of different machine learning models trained...... 148 Table S6-4: Evaluation of machine learning performance (mean ± standard deviation) on different classes...... 148
X Abbreviations
Abbreviations Aba Aminobutyric acid
Adda (2S,3S,8S,9S,4E,6E)-3-amino-9-methoxy-2,6,8-trimethyl-10-phenyl-4,6-decadienoic acid
AF Allocation factor
AFA Aphanizomenon flos-aqua
ALT Alanine aminotransferase
Amp Ampicillin
ANOVA Analysis of variance
AP Alkaline phosphatase
AST Aspartate aminotransferase
ATP Adenosintriphosphate
BGAS Blue-green algal supplements
Boc Tert-butyloxycarbonyl protecting group bw Body weight
Cam Chloramphenicol
CGN Cerebellar granular neurons
CV Cross validation
D5-MC-LF MC-LF labeled with five deuterium atoms
D7-MC-LR MC-LR labeled with seven deuterium atoms
Dhb Dehydrobutyrate
Dhb (E)-2-amino-2-butenoic acid
DIPEA N,N-Diisopropylethylamine
DMF Dimethylformamide dw Dry weight
ELISA enzyme-linked immuno-sorbent assay
ESI Electrospray ionisation
EtOAC Ethylacetate
FA Formic acid
FC Flash column chromatography
FCP/SCP Aspartate-based protein-phosphatases
XI Abbreviations fmoc Fluorenylmethyloxycarbonyl protecting group
GBM Gradient boosting machine
GF/C Glass fibre, class C
GGT γ-Glutamyl transaminase
GSH Glutathione
GV Guidance value
HAB Harmful algal bloom
HATU Hexafluorophosphate Azabenzotriazole Tetramethyl Uronium
HPLC High-performance/pressure liquid chromatography
HRMS High resolution mass spectrometry i.p. Intraperitoneal i.v. Intravenous
IARC International Agency for the Research on Cancer
IC50 Inhibitory concentration 50, concentration of a compound where 50% inhibition is reached
LC Liquid chromatography
LC-MS/MS Liquid chromatography coupled to tandem mass spectrometry
LHMDS Lithium bis(trimethylsilyl)amide
LLP Liquid-liquid partitioning
MF Morgan fingerprint
MC Microcystin
MC-LR Microcystin with Leucine and Arginine in positions X and Z
Mdha N-Methyldehydroalanine
MeOH Methanol
MMPB 3-methoxy-2-methyl-4-phenylbutyric acid
MRM Multiple reaction monitoring
MS Mass spectrometry
Ni-NTA Nickel-Nitrilotriacetic acid
NMeSecPh N-methylphenylselenocysteine
NMR Nuclear magnetic resonance
NOEL No-observed-effect-level
XII Abbreviations
OATP Organic anion transporting polypeptide
OD600 Optical density at 600 nm
Pac Phenacyl
PCB Polychlorinated biphenyls
PDA Photodiode array
Pfp Pentafluorophenyl
PhFl Phenylfluorenyl pNPP para-Nitrophenylphosphate
PP Protein precipitation
PPIA Proteinphosphatase inhibition assay
PPM Metal-dependent protein-phosphatases
PPP Serine/threonine phospho-protein-phosphatase
PPP1c Catalytic subunit of PPP1
Prg Propargyl rcf relative centrifugal force
RF Random forest rt Room temperature
SD Standard deviation
SDS-PAGE Sodium dodecyl sulfate polyacrylamide gel electrophoresis
SLCO Solute carrier organic anions (see OATP)
SMOTE Synthetic Minority Over-sampling Technique
SPE Solid phase extraction
SWATH Sequential window acquisition of all theoretical fragment ion spectra
TB Terrific broth
TDI Tolerably daily intake
TEF Toxicological equivalency factor
TEQ Toxic equivalency concept
TFA Trifluor acetic acid
THF Tetrahydrofuran
TLC Thin layer chromatography
XIII Abbreviations
TOF Time of flight
TPR Tetratricopeptide repeat
TrxA Thioredoxin A
TrxA-6×His-PPP5 Fusion protein consisting of Thioredoxin A, a Histidine-Tag and PPP5
UF Uncertainty factor
UPLC-MS/MS Ultra performance liquid chromatography coupled to tandem mass spectrometry
WHO World Health Organization
β-D-Masp Erythro-β-D-methylaspartate
XIV Abbreviations
This page is intentionally left blank.
XV Abstract
Abstract Over the recent years cyanobacterial blooms seem to increase in water bodies worldwide. As many cyanobacteria are able to produce secondary metabolites which are toxic to humans and animals, they represent a health threat. Among these so-called cyanotoxins, microcystins (MCs) are among the most prevalent. Upon uptake via contaminated drinking water or food items, MCs are transported into target cells via organic anion transporting polypeptides (OATPs), where they covalently and irreversible inhibit serine/threonine proteinphosphatases (PPPs). Due to variable positions in the cyclic peptide structure of MCs and (de)methylation sites, many congeners (> 200) are known to date. This complicates adequate risk assessment and thus the current guidelines by the World Health Organization are considered only ‘provisional’ as additional data is lacking. The present work deals with several aspects, which are necessary for a congener- dependent risk assessment of MCs. Firstly, laboratory handling of MC containing solutions was further improved, as ‘safe concentrations’ of solvents were determined at which no MC adsorption to plastic and glass surfaces under acidic and non-acidic conditions occurs. Generally, at least 40% of methanol should be used, irrespective of acidification, surface material or investigated congener. Secondly, extraction of MCs from complex matrices, such as blood or liver tissue material was improved. Here, a three-step extraction was employed using protein precipitation with methanol, liquid-liquid-partitioning with hexane and solid phase extraction. The procedure was streamlined, so that it can be finished during one day with analysis on the following day. Furthermore, an UPLC-MS/MS method was developed, which uses synthetic stable-isotope labelled internal standards (D5-MC-LF and D7-MC-LR) for the simultaneous analysis of 14 MC congeners. Moreover, the inhibitory capacity of 18 MC congeners against three representatives of the PPP family – PPP1, PPP2A and PPP5 – was determined, revealing differences in the toxicodynamic of individual MC congeners. This inhibition dataset was further used to train a machine learning algorithm that allows for the prediction of toxicity of MC congeners based on their 2D chemical structure. Taken together, the present work developed tools which allow for the determination of the individual toxicity of a MC containing sample. This is achieved through the possibility of stable detection of congeners from complex samples. This data further can be used to either look up the PPP inhibition capacity of the present congeners in a given sample from the inhibition dataset or toxicity may be predicted using the presented machine learning algorithm. Although this represents a first step towards congener dependent risk assessment for MCs, still some hurdles need to be overcome. Firstly, the protein-bound portion of the MC
XVI Abstract content of a given sample cannot be analysed with the presented method. To achieve this, the reported extraction procedure may be augmented with a step to deconjugate bound MCs. Secondly, the PPP inhibition dataset of course includes only the toxicodynamic portion of the observed MC toxicity. Organ distribution, cellular uptake and especially cellular export are still under research and play a major role during risk assessment of a certain compound.
XVII Zusammenfassung
Zusammenfassung Blüten von Cyanobakterien scheinen weltweit in den letzten Jahren zuzunehmen. Da viele Cyanobakterien in der Lage sind für Menschen und Tiere toxische Sekundärmetabolite zu produzieren, stellen sie eine Gesundheitsgefährdung dar. Unter diesen sogenannten Cyanotoxinen gehören Microcystine (MC) zu den am häufigsten vorkommenden. Nach der Aufnahme über kontaminiertes Trinkwasser oder Nahrungsmittel, werden MC über OATP (organic anion transporting polypeptides) in die Zielzellen transportiert, wo sie kovalent und irreversibel Serin-/Threonine-Proteinphosphatasen (PPP) inhibieren. Aufgrund variabler Positionen in der zyklischen Peptidstruktur und verschiedenen (De)Methylierungsstellen, sind derzeit über 200 Kongenere bekannt. Dies erschwert eine adäquate Risikoabschätzung für MC, weswegen die aktuellen Richtlinien der Weltgesundheitsorganisation nur als „vorläufig“ angesehen werden, da zusätzliche Daten fehlen. Die vorliegende Arbeit beschäftigt sich mit verschiedenen Aspekten, welche für eine kongenerspezifische Risikoabschätzung von MC notwendig sind. Zunächst wurde die Handhabung von MC-haltigen Lösungen weiter verbessert, indem „sichere Konzentrationen“ von Lösungsmitteln bestimmt wurden, bei denen keine Adsorption an Labormaterialien wie Plastik- und Glasoberflächen unter sauren und nichtsauren Bedingungen stattfindet. Generell sollten mindestens 40% Methanol verwendet werden, unabhängig von Ansäuerung, Oberflächenmaterial oder untersuchtem Kongener. Zweitens wurde die Extraktion von MC aus komplexen Matrizen, wie zum Beispiel Blut oder Lebergewebe verbessert. Hier wurde eine dreistufige Extraktionsmethode verwendet, welche Proteinfällung mit Methanol, flüssig-flüssig Extraktion mit Hexan und Festphasenextraktion verwendet. Darüber hinaus wurde eine UPLC-
MS/MS-Methode entwickelt, welche synthetische, isotop-markierte interne Standards (D5-MC-
LF und D7-MC-LR) für die gleichzeitige Analyse von 14 MC-Kongeneren verwendet. Außerdem wurde die inhibitorische Wirkung von 18 MC-Kongeneren auf drei Mitglieder der PPP-Familie – PPP1, PPP2A und PPP5 – bestimmt. Dies zeigte Unterschiede in der Toxikodynamik einzelner MC-Kongenere. Dieser Datensatz wurde daraufhin verwendet, um einen Algorithmus für maschinelles Lernen zu trainieren, welcher in der Lage ist, die Toxizität von MC-Kongeren basierend auf deren 2D-Struktur, vorherzusagen. Zusammengenommen können die, in dieser Arbeit entwickelten, Werkzeuge dazu benutzt werden, die individuelle Toxizität einer MC enthaltenden Probe zu bestimmen. Dies wird durch den Nachweis einzelner Kongenere aus komplexen Matrizen ermöglicht. Die so gewonnenen Daten können daraufhin entweder dafür benutzt werden, um die Toxizität der
XVIII Zusammenfassung anwesenden Kongenere in einer Datenbank abzurufen, oder deren Toxizität mit Hilfe des präsentierten Algorithmus vorhergesagt werde. Obwohl die vorliegende Arbeit einen ersten Schritt in Richtung einer kongenerspezifischen Risikoabschätzung für MC darstellt, sind dennoch weitere Hürden zu überwinden. Erstens kann der proteingebundene MC-Anteil einer Probe mit der vorliegenden Methode nicht analysiert werden. Um dies zu gewährleisten, könnte die vorgestellte Extraktionsmethode um einen Schritt zur Dekonjugation gebundener MC erweitert werden. Zweitens stellt der gezeigte Datensatz zur PPP-Inhibition nur den toxikodynamischen Anteil der beobachteten Toxizität dar. Organverteilung, zelluläre Aufnahme und insbesondere zellulärer Export werden weiterhin erforscht und spielen eine wichtige Rolle für die Risikoabschätzung eines Stoffes.
XIX
1. General Introduction
1. General Introduction
1.1 Cyanobacteria
1.1.1 Ecology of cyanobacteria
Cyanobacteria are a phylum of gram negative bacteria. Their ability to use photosynthetic pigments in the generation of adenosintriphosphate (ATP) makes them the first photoautotrophic organisms. It is also believed that they were the first producers of oxygen. Fossil records, so called stromatolites, are assumed to reach back around 3.5 billion years (Tomitani et al. 2006; Whitton and Potts 2007). Therefore cyanobacteria are supposed to be mainly responsible in producing an oxygen-rich atmosphere, which is the basis of life on earth as we know it today. Due to the presence of (photosynthetic) pigments (chlorophyll a, carotenoids, phycocyanin, allophycocyanin and phycoerythrocyanin) they often appear blue or green (Colyer et al. 2005), hence their misleading description as blue-green algae. The colour “blue-green” is also found in the prefix “cyano” in cyanobacteria, which describes them more accurately due to their bacterial nature. The so-called “endosymbiotic theory” states, that the chloroplasts in higher plants have been developed from cyanobacteria which were incorporated by other unicellular organisms. Although the identity of the host and the origin of mitochondria is still under debate, it is generally accepted that a cyanobacterial symbiote is the origin of the plastids in higher plants (Martin et al. 2015). Although the major portion of cyanobacteria are found in marine or freshwater habitats they are also found in almost all habitats, ranging from soil, to deserts, and even the polar regions, as well as in symbiosis with various partners, e.g. with some fungi to form lichens (Lacap-Bugler et al. 2017; Vincent 2002; Whitton and Potts 2007). Additionally to the production of oxygen, some species of cyanobacteria are able to fix nitrogen directly from the atmosphere. This allows them to grow in nutrient-poor environments and often act as pioneering species in a certain water body. Beyond that, many species of cyanobacteria are able to form so-called blooms. A bloom describes the sudden but massive appearance of one or few species in a particular water body (Figure 1-1). This is problematic, as it limits nutrients and photosynthetic light for algae and plants, but also may lead to a rapid decline of oxygen in the water, when a bloom collapses and cells are degraded. Additionally,
1 1. General Introduction many species are able to produce toxins, which are consequently called cyanotoxins. Due to that, such blooms are also called HAB (harmful algal blooms) or cyanoHAB.
Figure 1-1: Cyanobacterial bloom in Lake Erie in 2003. Lake Erie in the Northern part of the USA, at the Canadian border, on 12th April 2003. The major part of the lakes 25667 km² surface is covered with cyanobacterial mass. Picture taken by the MERIS spectrometer of the Envisat satellite of the European Space Agency (ESA).
Bloom formation is favoured in warmer waters and increased in nutrient rich (eutrophic, high phosphorous) water bodies. Eutrophication of lakes is increasing due to nutrient (mainly nitrogen and phosphorous) input of human origin (sewage run-off, increased use of fertilizers, groundwater discharge, etc.) (Paerl and Otten 2013). Due to fixation of atmospheric nitrogen by some cyanobacterial species, N is not necessarily limiting. Additionally, it is believed, that climate change and global warming will lead to increased bloom formation and also longer bloom periods (Kleinteich et al. 2012; Paerl and Huisman 2008). Furthermore, there is evidence, that warmer water temperatures favour toxin producing cyanobacterial species over non-toxic strains (Davis et al. 2009). Bloom formation is found in various species of cyanobacteria, of which some are able to fix atmospheric nitrogen, e.g. Anabaena spp., Aphanizomenon spp., Cylindrospermopsis spp. or Nodularia spp., as well as non-nitrogen-fixing species like Microcystis spp. and Planktothrix (Paerl and Otten 2013).
1.1.2 Cyanotoxins
Many cyanobacterial species are able to produce secondary metabolites which are harmful to other organism. Hence, they are called cyanotoxins. First scientific reports of adverse effects of cyanobacterial bloom material were already published in the 19th century, although
2 1. General Introduction they were not specifically attributed to toxins back then (Francis 1878). In the following, some cyanotoxins with an impact on human health are described briefly, although this is not a complete summary. Microcystin is described later in detail (see section 1.2). Saxitoxin (Figure 1-2), also known as paralytic shellfish poisoning toxin, is a neurotoxin acting via the blockage of voltage-gated sodium channels. Thus, it disturbs neuronal signal transduction resulting in acute cases in paralysis and death due to respiratory arrest. Non-acute doses may lead to other neuronal symptoms like tingling sensation or numbness in fingertips and toes, but also to headache and nausea (Buratti et al. 2017). The term saxitoxin actually comprises a family of > 30 structurally related alkaloid compounds with only minor alterations. Saxitoxins are often found in seafood, as they accumulate in shellfish which are then consumed by other animals or humans where the symptoms are exerted. Hence, their description as paralytic shellfish poisoning toxin. Interestingly, saxitoxins are produced by fresh water cyanobacterial species, but also by species of marine dinoflagellates, indicating a potential horizontal gene transfer (Kellmann et al. 2008).
Figure 1-2: A selection of cyanotoxins. Saxitoxin is actually a collective term for several structurally related toxins. A variant of Anatoxin, Homoanatoxin, displays the same structure, but has an additional methylene group after the carbonyl carbon. Chemical structures were taken from Wikipedia.
Anatoxin (Figure 1-2) and homoanatoxin are bycyclic alkaloids, which only differ in the presence of an additional methylene group (-CH2-) in homoanatoxin. Several cyanobacterial species are known to be producers of the compounds, but originally Anabaena species (now Dolichospermum) were identified as producers, hence the name Anatoxin. The producing species and thus the toxins are found in water bodies worldwide (James et al. 2007). Anatoxin
3 1. General Introduction is rapidly degraded in water due to photolysis and chemical instability, therefore accumulation in the food web is unlikely. Nevertheless, its high toxicity poses a health threat to humans and animals. Formerly, anatoxin was also called very-fast-death-factor, as it is an irreversible activator of the nicotinic acetylcholine receptor, thus leading to constantly active receptor and subsequently to blockage of neuronal transmission (Carmichael et al. 1975; Carmichael et al. 1979). Cylindrospermopsin is a tricyclic alkaloid whose main producer is Cylindrospermopsis raciborskii (Figure 1-2). It was first structurally described in 1992 (Ohtani et al. 1992), but is thought to have caused the so-called ‘Palm Island mystery disease’ in 1979 which described a sudden outbreak of gastroenteritis on Palm Island, Australia, after a cyanobacterial bloom of Cylindrospermopsis raciborskii was treated with copper and thus cells lysed thereby releasing the compound in the surrounding water body (Griffiths and Saker 2003). Cylindrospermopsin exerts its toxicity mainly through inhibition of protein synthesis. It is believed, that its structure as uracil analogue plays a major role in its mechanism of action (Reisner et al. 2004). Target organs are mainly the liver and kidneys, due to mostly oral exposure, but generally all organs could be affected (Buratti et al. 2017). Cylindrospermopsin has been observed globally and producing species are found in fresh and brackish water bodies (Rzymski and Poniedzialek 2014).
1.2 Microcystin
Although microcystins (MCs) are among the best studied cyanotoxins, there are still some research gaps, mainly regarding toxicokinetics, i.e. cellular export. Nevertheless, also research on toxicodynamics, e.g. about differential toxicities, is needed as the MC family comprises many individual toxins which differ in their activity (see below).
1.2.1 Chemical structure
MCs are cyclic peptides which comprise seven amino acids (Figure 1-3), among these “standard” but also unique and rare amino acids and derivatives. The consensus sequence for
MC, cyclo-[D-Ala1]-[L-X2]-[β-D-MeAsp3]-[L-Z4]-[Adda5]-[γ-D-Glu6]-[Mdha7], reveals that along with L-amino acids, also D-amino acids are incorporated. The amino acid in position 5, Adda, stands for (2S,3S,8S,9S,4E,6E)-3-amino-9-methoxy-2,6,8-trimethyl-10-phenyl-4,6- decadienoic acid and is a unique amino acid, which has only been found in MC and the related nodularins. In position 3 β-D-methylaspartate (β-D-MeAsp) is found, while
4 1. General Introduction methyldehydroalanine (Mdha) is present in position 7. Generally, Positions 1, 5 and 6 are quite conserved, while positions 3 and 7 sometimes display variations like a different methylation pattern. Positions 2 and 4, often also called X and Z, are hypervariable. Mostly standard L- amino acids are incorporated in these two positions. This variability, along with the small alterations observed in the other positions, leads to many possible combinations. Until now, around 250 congeners have been observed (Spoof and Catherine 2017). Timo Niedermeyer compiled a comprehensive, interactive table of around 100 structurally confirmed congeners which can be consulted online (Niedermeyer 2014). The one-letter code for the amino acids in the X and Z positions is used for nomenclature. Hence, a MC with leucine (L) and arginine (R) in the X and Z positions respectively, is called MC-LR, while leucine and phenylalanine will result in MC-LF. Additionally, a common nomenclature has been established to describe alterations from the consensus structure. These are given in square brackets at the beginning while indicating on which positions the difference is located, e.g. [Asp3]MC-LR describes a variant of MC-LR where in position 3 a standard aspartate residue is found instead of the consensus methylasparte.
Figure 1-3: General structure of microcystins. X and Z mark variable positions in which different amino acids can be found, thus generating many congeners. MC-LR for example displays leucine (L) in the X and arginine (R) in the Z position. Additional modifications (mostly de/methylations) are commonly found at the Adda side chain, the methylasparte (ß-D-MeAsp) and the methyldehydroalanine (Mdha), whereas alanine (D-Ala) and glutamic acid (γ-D-Glu) are altered only very rarely, if ever. (Figure generated with ChemDraw Professional 16.0 (PerkinElmer))
1.2.2 Biosynthesis of microcystins and ecological role
Although MCs are peptides, they are not synthesized with the ribosomal machinery like proteins, but by several enzymes which are organized in gene clusters (Tillett et al. 2000). The genes are called mcyA-L which all encode for different enzymes e.g. polyketide synthetases,
5 1. General Introduction non-ribosomal peptidesynthases, methyltransferases, etc. There seems to be core enzymes which are necessary, but some strains may have additional enzymes, which leads to rarer congeners during synthesis e.g. Adda-acetylated congeners when mcyL is present (Fewer et al. 2013) or demethylated Mdha residues when mcyA is inactive (Nishizawa et al. 1999). Some of the involved enzymes display relaxed substrate specificities and even variations in the genes of the cluster exist (Mikalsen et al. 2003). Additionally, it has been shown, that temperature and nitrogen/phosphorus availability seem to influence which congeners are produced (Amé et al. 2005; Puddick et al. 2016a; Xie et al. 2016), also contributing to the observed variety of congeners. The mcyH gene is discussed to be an ABC-transporter involved in the externalization of MC (Tillett et al. 2000). The mcy gene cluster is evolutionary very old and is present in organisms which lived before first eukaryotic cells evolved (Rantala et al. 2004). Therefore, the long discussed function of MCs in grazing defence cannot be the sole role of the compounds. More likely is, that MCs originally were produced for another reason and fulfilled a different ecological role. Nevertheless, the appearance of grazers later most likely has played part in the evolutionary conservation of the gene cluster. So far, the primary ecological role of MCs is still not completely understood, especially as production of MCs is costly in terms of energy and nitrogen usage. Several theories have been published and reviews exist which summarize the existing theories (Holland and Kinnear 2013; Omidi et al. 2018). In the following, some are presented briefly. Often the possible functions are classified as internal and external, describing functions which benefit the single cell or the community in whole, respectively. Internal functions include aid in photosynthesis and protection from light stress as well as oxidative protection. Additional roles as siderophores for iron acquisition and nutrient storage have been discussed. External functions could include quorum sensing e.g. for colony and/or bloom formation or of course as defence from grazing organisms or competitors. MC content in cyanobacterial cells has been demonstrated to be linked to chlorophyll a content indicating involvement in photosynthesis. Growth experiments with mutants lacking essential genes for MC production suggested involvement of MCs during light harvesting (Omidi et al. 2018). Additionally, non-producers display decreased fitness compared to toxin producers under high-light conditions, and even produce less photosystem-associated proteins, indicating a role in photo-oxidation protection (Omidi et al. 2018; Zilliges et al. 2011). Interestingly, also warmer growth conditions seem to increase the production of MCs per cell (Davis et al. 2009). It has been observed, that under low iron conditions, more MCs are
6 1. General Introduction produced, most likely due to binding of the transcription factor Fur (ferric uptake regulator) in the promoter region of the mcy gene cluster (Omidi et al. 2018). Nevertheless, it is not clear, whether MCs in case of low iron help the cells to acquire additional iron or if MC production is coupled to iron homeostasis for a different reason, e.g. shuttling of iron through cell membranes (Klein et al. 2013). Although it is known, that most of the MC content of producing cells is actually found within the cells (van Apeldoorn et al. 2007), it is still debated whether MCs might be signalling molecules, either towards other species (allelopathic signalling) or towards the same species (quorum sensing/quorum quenching). There are several publications dealing with growth inhibition of higher plants due to exposure to MCs (Omidi et al. 2018), however this might also be a secondary effect of MCs, similar to the protection against grazers. More likely, MCs are used to communicate between members of the same species, e.g. to determine population density (quorum sensing). MCs per cell increased in lake Rotorua (South Island, New Zealand) and mesocosm experiments during times of high cyanobacterial density (Wood et al. 2012; Wood et al. 2011). Although this might point to quorum sensing via MCs, the authors state, that it is also possible, that higher cell density and increased MC content per cell could both be reactions to environmental factors. Additionally, if MCs really would be quorum sensing molecule, then extracellular MC concentrations should increase before an increase in cell density is observed and not peak together with it, although it is possible that even a further density increase could be observed while MC levels are already declining. Nevertheless, quorum sensing has been observed in cyanobacteria, albeit using lactones and not the cyanobacterial toxins (Romero et al. 2011; Sharif et al. 2008; Zhai et al. 2012).
1.3 Toxicokinetics of microcystins
Human microcystin exposure can occur via various routes. It has been revealed, that wind activity over a water body may lead to aerosolized MCs (Wood and Dietrich 2011) and thus inhalation might be a possible route of exposure, although the authors calculated that aerosols produced by wind most likely would not reach critical concentrations. Nevertheless, inhalation should still be considered, especially in areas where recreational activities (swimming, rowing, jet-skiing) are performed on bloom-affected water bodies on a regular basis (Backer et al. 2008; Turner et al. 1990). Nevertheless, the main route of exposure by far still is oral ingestion. MCs are frequently found in drinking water (Bullerjahn et al. 2016; World Health Organization 2003a; World Health Organization 2017), fish and other seafood from contaminated water bodies (Ibelings and Chorus 2007), but also in spray-irrigated vegetables,
7 1. General Introduction when contaminated water is used (Codd et al. 1999b; Trifiro et al. 2016). Additionally, in the recent years, consumption of food supplements produced from dried cyanobacterial bloom material (BGAS, blue-green algal supplements) increased. Unfortunately, these are often contaminated with toxin-producing cyanobacterial species (Bautista et al. 2015; Gilroy et al. 2000; Heussner et al. 2012; Vichi et al. 2012). This is due to the manufacturing process, which involves harvesting bloom material of commonly Aphanizomenon flos-aqua (AFA), drying and then pressing into tablets without further analysis of co-occurring species or toxins. It has been demonstrated that AFA often actually co-occur with MC (and other cyanotoxin) producing species and are likely to be able to produce MCs by themselves (Lyon-Colbert et al. 2018). Additionally, two cases of intravenous exposure to MC have been reported. They are both related to the use of MC-contaminated water during dialysis treatment of patients suffering from renal insufficiency (Azevedo et al. 2002; Carmichael et al. 2001; Pouria et al. 1998; Soares et al. 2006; Yuan et al. 2006). Both cases occurred in Brazil: In 1996, 131 patients in Caruaru were accidently treated using water contaminated with MC. In the aftermath 76 patients died. This case also is the only case where human deaths could be directly attributed to microcystins (Azevedo et al. 2002; Carmichael et al. 2001; Pouria et al. 1998; Yuan et al. 2006) (see section 1.5.2). In 2001, 44 patients in Rio de Janeiro were exposed to sublethal doses of MC during renal dialysis (Soares et al. 2006). MC are, even though they possess the hydrophobic Adda side chain, generally hydrophilic molecules. Thus, to reach intracellular targets, they need to be transported over biological membranes. Organic anion transporting polypeptides (OATPs) have been demonstrated to be involved in the transmembrane transport of MCs in humans (Fischer et al. 2005), but also in rodents (Feurstein et al. 2009; Feurstein et al. 2010) and several fish species (Meier-Abt et al. 2007; Steiner et al. 2014; Steiner et al. 2015). OATPs are a group of transporters which display a broad substrate specificity which is also displayed in their other name SLCO (solute carrier organic anions). All OATPs display twelve transmembrane helices. They are involved in the transport of xenobiotics, bile salts and hormones (Hagenbuch and Meier 2004). In humans, eleven members are known, which display a distinct pattern of organ distribution, e.g. OATP1B1 and 1B3 are found in the liver, while OATP1A2 is expressed in the brain and blood-brain-barrier and OATP2A1 is found ubiquitously in the body (Table 1-1). Thus, it is clear that the OATP expression profile plays a non-negligible role in MC toxicity, especially as it has been proven that MC congeners are differentially transported by individual OATPs (Fischer et al. 2010). This also explains the fact that the liver is the main target organ of MCs: most exposure events happen via the oral route, where they end up in the portal vein.
8 1. General Introduction
From there uptake into the liver takes place via OATP1B1 and 1B3 which are highly expressed in hepatocytes.
Table 1-1: Distribution and substrates of human organic anion transporting polypeptides (OATP). Name Tissue distribution Known substrate OATP1A2 Brain, kidney, liver Bile salts, organic anions and cations OATP1B1 Liver Bile salts, organic anions OATP1B3 Liver Bile salts, organic anions OATP1C1 Brain, Leydig cells (testis) T4, rT3, BSP OATP2A1 Ubiquitous Eicosanoids OATP2B1 Liver, placenta, ciliary body E-3-S, DHEAS, BSP OATP3A1 Ubiquitous E-3-S, prostaglandin OATP4A1 Ubiquitous Taurocholate, T3, prostaglandin OATP4C1 Unknown Unknown OATP5A1 Kidney Unknown OATP6A1 Testis Unknown The table shows the tissue distribution of the known eleven human OATP along their natural substrates. BSP and taurocholate are substrates which are often used in testing the function of OATP. BSP: bromosulfophthalein; DHEAS: dehydroepiandrosterone- sulfate; E-3-S: estrone-3-sulfate; rT3: (reverse tri-iodothyronine; T4: thyroxine. Adapted from (Hagenbuch and Meier 2004).
OATPs are clustered in families which are defined by at least 40% amino acid homology. A distinct subfamily is defined by more than 60% homology. Families are denominated using numbers, subfamilies with letters and individual transporters with another number, e.g. OATP1B3 (family 1, subfamily B, transporter 3). MCs can be conjugated via the C=C double bond of the Mdha residue using the cellular glutathione (GSH) machinery (Pflugmacher et al. 1998b). Conjugation with GSH is considered to be an important step of detoxification of xenobiotics because the addition of GSH increases hydrophilicity of a molecule and thus renders it susceptible to excretion via the urine. Nevertheless, conjugates of MCs are still toxic to the cells, because their inhibitory potential is still present, although reduced (Kondo et al. 1992; Smith et al. 2010). Apart from glutathione conjugates also cysteine-glycine and cysteine conjugates have been observed (Schmidt et al. 2014) which most likely are degradation products of the GSH conjugate. In order to be excreted, MCs and their conjugates need to be exported from the target cells. This cellular export still is under research and export-transporters have not yet been
9 1. General Introduction unambiguously identified, however it is discussed that multidrug-resistance related transporters are involved in the cellular export, e.g. MRP (Bieczynski et al. 2014; Schwarzenberger et al. 2014). So far, preliminary results corroborate these suggestions (Kaur and Dietrich 2018). Although involved transporters are not definitely identified, it is clear that they must exist, as MCs and MC conjugates have been identified in urine (9% of administered dose) of i.v. exposed animals (Robinson et al. 1991). Although MCs are hydrophilic, even before conjugation, they are also excreted via the faeces (15 % of administered dose) (Robinson et al. 1991). This may be explained by the expression of many multidrug-resistance related proteins in hepatocytes which also express many OATPs, thus serving as shuttle for MCs from the bloodstream into the bile.
1.4 Toxicodynamics of microcystins
1.4.1 Excursus: The ser/thr-PPP family in mammals
Protein phosphorylation and dephosphorylation is a dynamic process within cells, which regulates various different downstream processes. Kinases attach phosphate residues to other enzymes to either activate or deactivate their function, while phosphatases detach the phosphate residues, thereby counteracting the action of kinases. Phosphorylation can occur at tyrosine (tyr), serine (ser) and threonine (thr) residues. Here, the major portion occurs at serine (86.4 %) and threonine (11.8%) residues, while only 1.8% tyrosine-phosphorylations were observed in a representative sample (Olsen et al. 2006). Additionally, histidine phosphorylation is well- known in bacteria, fungi and plants, but only recently starts to be a field of research in mammals (Besant and Attwood 2005). There are four classes of protein-phosphatases: Serine/Threonine- phosphatases, tyrosine-phosphatases, so-called dual-specificity phosphatases, which are able to dephosphorylate serine, threonine and tyrosine residues and finally histidine phosphatases. Serine/threonine phosphatases are further divided into three families: the metal- dependent protein-phosphatase (PPM) family, the aspartate-based protein-phosphatase (FCP/SCP) family and the largest phospho-protein-phosphatase (PPP) family, which differ in their active centre and thus in their catalytic mechanism (Figure 1-4) (Shi 2009). Although the family names might suggest otherwise, both PPM and PPP members use metal ions during catalysis, while no metal ions are involved in the mechanism of FCP/SCP. The PPM family thereby uses manganese (Mn2+) and magnesium (Mg2+) ions, while PPP members use manganese (Mn2+) or zinc (Zn2+) and iron (Fe2+) (Barford et al. 1998; Egloff et al. 1995; Shi 2009).
10 1. General Introduction
Figure 1-4: The families of ser/thr-phosphatases. Members of the PPP family display a common catalytic core domain. PP2B and PP5 display additional domains, mostly involved in the regulation of the enzymes. The PPM family only has one member: PP2C. FCP and SCP share a common domain in which the catalytic centre is located. Taken from Shi (2009). PP1-7 in the figure correspond to PPP1-7 in the text.
The iron and manganese ions in PPP members are coordinated in the active centre by aspartate, asparagine and histidine residues, along with a water molecule which is important for the catalytic activity of the enzyme (Egloff et al. 1995). Along with the substrate phosphate, the water is coordinated by the metal ions and thus allows for a nucleophilic attack on the phosphate, ultimately resulting in the cleavage of the phosphate from the substrate (Egloff et al. 1995). Although the previous was described for PPP1 (and PPP2B), it is assumed that the mechanism is the same in all members of the PPP family, as they are closely related (Shi 2009). Indeed, the catalytic subunits of the members of the PPP family demonstrate a consensus catalytic core domain in which the metal coordinating residues are invariable (Barford et al. 1998; Shi 2009). In mammals, the family consists of seven members which are PPP1, PPP2A, PP2B (Calcineurin), PPP4, PPP5, PPP6 and PPP7 (Figure 1-4). Although they share the same catalytic mechanism and co-occur in most tissues (Pereira et al. 2011a), albeit with different expression levels, they have evolved different functions and substrate specificities (Table 1-2). This is possible because the catalytic subunits interact with different regulatory and scaffolding subunits which are the basis for substrate specificity. PPP1 is known to interact only with a regulatory subunit, while PPP2A, PPP4 and PPP6 additionally
11 1. General Introduction need a scaffolding subunit (Brautigan 2013). Interestingly, PPP2A, PPP4 and PPP6 are also more closely related to each other than to the other members (Cohen 2004). PPP2B is also called calcineurin, which indicates its predominant expression in the brain and its association to calcium signalling there. PPP2B interacts with a specific regulatory subunit called calmodulin, which reacts to Ca2+ in cells. Similarly, PPP7 is also calcium dependent and has a calmodulin binding motif. In contrast to the other catalytic subunits, both are inactive when not in complex with calmodulin (Cohen 2004; Shi 2009). PPP5 is a special case, as it is the only member of the PPP family, which does not interact with a regulatory subunit. It displays an additional autoregulatory tetratricopeptide repeat (TPR)-domain on the same peptide as the catalytic subunit (Chen et al. 1994; Yang et al. 2005). The TPR domain folds back into the active centre, thereby blocking its activity. Addition of polyunsaturated fatty acids e.g. arachidonic acid, releases this blockage and activates catalysis (Yang et al. 2005).
Table 1-2: Functions of PPP family members. Enzyme Functions known subunits PPP1 Muscle relaxation, synaptic transmission, gene expression, glycogen > 100 metabolism, RNA splicing, cell-cycle progression
PPP2A Cell-cycle regulation, cell growth control, cytoskeleton dynamics, cell mobility, metabolism, transcription, translation, RNA splicing, DNA > 15 replication, apoptosis, inflammation, differentiation
PPP2B Response to neurotransmitters and nerve impulses, NMDA-receptor 2 (Calcineurin) signalling pathway, T-lymphocyte activation
PPP4 Centrosome maturation, microtubule organization, histone > 5 phosphorylation, apoptosis
PPP5 Cell growth, ribosomal RNA transcription regulation, atrial natriuretic peptide (ANP) signalling, steroid signalling, blue-light signal none transduction, ion channels regulation
PPP6 Transcription, translation, morphogenesis, cell-cycle regulation > 4
PPP7 Control of phosphorylation status of G protein–coupled receptors unknown Table adapted from (Pereira et al. 2011a).
1.4.2 Consequences of microcystin exposure
The toxic effect of MCs has been investigated intensively over the past years using many different model organisms and even case studies with humans after involuntary, accidental exposure. As said earlier, the main route of exposure is via ingestion, thus the liver is the primary target organ as it expresses many OATPs. MCs are potent inhibitors of the catalytic
12 1. General Introduction subunits of members of the serine/threonine-phosphatases (PPP) family. Thus, in the affected cells, several signal transduction pathways are disturbed due to MC exposure. Affected cells remain in a hyperphosphorylated state, deregulating several important processes. Homeostasis of the cytoskeleton requires dynamic changes in phosphorylation and dephosphorylation (Ritchie and Battey 1996). Thus, inhibition of phosphorylation deregulates the cytoskeleton, affecting all three components (microtubules, intermediate filaments and actin) (Zhou et al. 2015), often resulting in a cell which cannot fulfil its specific task, e.g. tight junctions between cells get loose and cells are rounding up thereby damaging the barrier function of epithelia (Hooser et al. 1990; Zhou et al. 2017). Cytoskeleton disruption has been observed to be a trigger for apoptosis (Kulms et al. 2002; Suria et al. 1999). This is mainly mediated by mitogen- activated protein kinases (MAP-kinases) downstream from prominent PPPs, like PPP1 and PPP2A (McLellan and Manderville 2017; Tricker et al. 2011). Additionally, cellular hyperphosphorylation has been associated with the activation of p53 (Takumi et al. 2010) which is widely accepted as an additional trigger for apoptosis. Furthermore, apoptosis can be induced by reactive oxygen species (ROS), which are also reported during MC intoxication. The formation of ROS by MCs is discussed to be the result of two mechanisms: Firstly, GSH is used for the conjugation during phase-I-metabolism of MCs and is therefore not available as a ROS scavenger anymore (Bouaicha and Maatouk 2004). Secondly, it was hypothesized that the calmodulin-dependent protein kinase II (CaMK) is affected by MCs, thus increasing cellular ROS levels (Campos and Vasconcelos 2010; Fladmark et al. 2002). Several additional mechanism are discussed for MC toxicity (Campos and Vasconcelos 2010; Chen and Xie 2016; McLellan and Manderville 2017; Zhou et al. 2015), but apoptosis (through any mechanism) is generally accepted as the major mechanism of toxicity. This observed apoptosis correlates to the observed symptoms in the target organs, where (ultra)structural changes are observed. Rounding and apoptosis of hepatocytes results in disturbed tight junctions and thus in leakage of blood out from the vessels into the surrounding liver tissue and hence decreased blood pressure is observed (Hooser et al. 1990; Theiss et al. 1988). Ultimately, this can lead to coma or even death due to hypovolemic shock or in less acute cases to vomiting and nausea. These effects have been observed in several species, including rats, mice, dogs, sea otters and cattle (Backer et al. 2013; Miller et al. 2010; Puschner et al. 1998; Theiss et al. 1988). Furthermore, intraperitoneal injections in rats additionally revealed damage of the tubuli and glomeruli of the kidneys and aggregation of proteinaceous material, which might be blood and/or cell debris of surrounding tissue (Milutinović et al. 2013). A study with primary murine cerebellar granule neurons (CGN) revealed decreased
13 1. General Introduction neuronal connections and neurite length in cells exposed to MCs compared to control cells (Feurstein et al. 2011), indicating neuronal effects which might explain observed neuronal symptoms e.g. dizziness and tinnitus. Interestingly, MC-LF and MC-LW shows increased effects on neurites compared to MC-LR (Feurstein et al. 2011). In addition to the above mentioned acute effects, it is discussed whether low dose chronic exposure to MCs may lead to increased cancer rates. Indeed, the International Agency for the Research on Cancer (IARC, part of the WHO) classifies MCs as “possibly carcinogenic to humans“ (class 2B) (IARC 2011). This classification is based on animal studies with mice and rats, in which preneoplastic lesions have been observed (Ito et al. 1997; Nishiwaki- Matsushima et al. 1992). Additionally, human epidemiological data was considered, which associated hepatocellular carcinoma with regions in southwest China were MCs have been frequently detected (Yu 1995). In the meantime, even further studies emerged, which link exposure to MCs to increased risk for cancer (Chen et al. 2009; Li et al. 2011). Tumour promotion by low-dose chronic exposures might be occurring through several mechanisms which were nicely summarized by Zegura (2016). Firstly, the deregulation of cellular phosphorylation homeostasis may contribute to cellular growth and, in contrast to acute MC doses, to inhibition of apoptosis. Additionally, constant low grade cell death due to the toxic potential of MCs (ROS, PPP-inhibition etc.) could lead to enhanced cell proliferation to maintain organ function. Constant cell division without proper cell cycle checkpoints makes cells prone to mutations and thus increase for cancerous lesions (OpenStax 2013).
1.4.3 Proposed interaction sites of PPPs and MCs and mechanism of inhibition
The inhibition of PPPs by MCs is a two-step process which firstly involves non-covalent interaction of the toxin with the catalytic centre and secondly covalent attachment of the MC to the PPP via a cysteine residue. The covalent inhibition takes place only after a couple of hours (Craig et al. 1996) but inhibitory capacity is observed already before covalent attachment (Hastie et al. 2005). Interaction sites of MCs and PPPs could be determined via co- crystallization of PPP1 with MC-LR (Goldberg et al. 1995) and molecular dynamics simulations (Mattila et al. 2000a). The hydrophobic Adda side chain of the MC interacts with the catalytic subunit of PPP1 (PPP1c) via a hydrophobic grove. This step is considered as the first interaction between enzyme and inhibitor. The rest of the MC molecule then binds to the catalytic centre where it interacts with the metal ions via a carboxylgroup of the D-glutamate in position 6 and an arginine via the methylasparte residue in position 3. Finally, the MC is attached to a cysteine residue via the Mdha residue at position 7.
14 1. General Introduction
Generally, it is assumed, that the inhibition of PPP works similar in all the members, as the catalytic subunits are conserved. Comparison of the crystallized catalytic subunits of PPP1 and PPP2A, both complexed with MC-LR, reveal a very similar conformation of the toxin in the catalytic pocket (Figure 1-5).
Figure 1-5: Overlay of the catalytic subunits of PPP and PPP2A in complex with MC-LR. PyMOL was used using the ProteinDataBank (PDB) entries 1fjm (PPP1) and 2ie3 (PPP2A). The catalytic subunits align well which each other, as well as the complexed MC-LR molecules. Turquoise: PPP1, orange: PPP2A, green: MC-LR in complex with PPP1, blue: MC-LR in complex with PPP2A, red: Mn2+ ions.
1.5 Risk assessment for microcystins and case reports of (human) intoxications
1.5.1 Risk assessment
The World Health Organization (WHO) published a guideline for safe concentrations of MCs in drinking water and recreational waters (World Health Organization 2003a; World Health Organization 2003b; World Health Organization 2017). Guidance values for food and supplement tablets can be deviated using the established tolerably daily intake (TDI) (Dietrich and Hoeger 2005). The published TDI of 0.04 µg/kg bodyweight (bw) is based on animal data and several assumption illustrated in the following. First, a no-observed-effect-level (NOEL) was determined from two animal studies involving rats (Fawell et al. 1999) and pigs (Falconer et al. 1994). Fawell et al. (1999) used intraperitoneal (i.p.) injections of MC-LR in mice to determine liver damage. They observed no changes in blood chemistry (alkaline phosphatase (AP), alanine aminotransferase (ALT), aspartate aminotransferase (AST), γ-glutamyl transaminase (GGT)) at an injected concentration of 40 µg/kg bw, but changes appearing at the following dose of 200 µg/kg bw. In the pig study by Falconer et al. (1994), animals had free access to water containing cyanobacterial bloom material. The identity of the present MC
15 1. General Introduction congeners in the material was not determined unambiguously, but according to the authors contained nine different congeners not including MC-LR or -RR (as measured by high performance/pressure liquid chromatography (HPLC)). Thus, extracting definite data for risk assessment is difficult. Given the toxin content of on average 2.25 µg/g bloom material and the 80 mg/kg bw/day ingestion of bloom material in the low dose group, around 180 µg/kg bw/day MC were ingested by the animals. In that group only one of the five animals developed mild symptoms, therefore the NOEL was expected to be slightly below 180 µg/kg bw MC. Additionally, Fawell et al. (1999) showed, that oral ingestion in mice is less toxic than i.p. administration with a factor of at least 30. Therefore, and because in the pig study no toxicity for single congeners could be elucidated, the NOEL of 40 µg/kg bw from Fawell et al. (1999) 푁푂퐸퐿 was considered for generating a TDI for drinking water according to the formula 푇퐷퐼 = , 푈퐹 where an uncertainty factor (UF) derived from extrapolation from animal data (factor 10), variability in the population (factor 10) and other factors (data set inadequacies, tumour promotion, extrapolation to chronic effects, factor 10) of 1000 was used, thereby arriving at a TDI of 0.04 µg/kg bw (Dietrich and Hoeger 2005). Using the TDI a guidance value (GV) for a safe concentration in drinking water (or other food items) can be determined. For this purpose, 푇퐷퐼 푥 푏푤 푥 퐴퐹 the formula 퐺푉 = with an allocation factor (AF, percentage of exposure by a given 퐶 source) of 0.8, the bodyweight (bw) of a standard “international adult” of 70 kg and the total amount of the resource (C) consumed of 2 L, a GV for drinking water of 0.96 µg/L (= 1 µg/L) was derived by the WHO (Dietrich and Hoeger 2005; World Health Organization 2017). Similarly, 10 µg/g MC-LR in supplement tablets (2 g of tables, AF 1) can be derived (Dietrich and Hoeger 2005). This guideline value is part of the WHO’s assessment since 1999 and remains unchanged since then, including the caveats and pitfalls already present back then. The major deficit of the assessment is, that all values are based only on the toxicity of MC-LR, although more than 200 congeners are known to date (Spoof and Catherine 2017). Especially, at it is known, that congeners display differential toxicity (Feurstein et al. 2011; Fischer et al. 2010; Monks et al. 2007). It was proposed to use toxicological equivalency factors (TEF) of MC-LR to describe the toxicity of other congeners (Dietrich and Hoeger 2005), similar to the approach used for polychlorinated biphenyls (PCB) previously (Ahlborg et al. 1992). In such an approach the toxicity of a certain congener would be described in relation to toxicity of a known congener. Although, in general this approach is promising, the caveat is, that every congener then should be tested individually and subsequently compared to, e.g. MC-LR, which implies a major effort
16 1. General Introduction of identifying and purifying individual congeners. The current assessment does not include or even discuss human data, which unfortunately is actually available due to accidents and involuntary exposures that are well researched, e.g. the Caruaru case (see below). Using the available data would strengthen the guideline values and therefore improve the risk assessment. Furthermore, with rapidly increasing computing capacities and machine learning approaches, additional improvements could be achieved. Structural classification of congeners in conjunction with machine learning may provide for a platform that is able to predict toxicity of unknown or not tested congeners based on their structural features.
1.5.2 Case reports
Over the years many, mostly involuntary exposures of humans (and animals) to MCs have occurred and some of them have been even researched further. Here, a small (but not complete) summary of some cases will be given. More cases and a list of references can be found in the guidance document by the WHO (World Health Organization 2003b). In August 2014 the city of Toledo (Ohio, USA) issued a ‘do-not-drink’ advisory for the drinking water of around 500,000 people (ToledoBlade 2014). This was due to MCs found in the processed drinking water above the WHO’s guideline of 1 µg/L in samples taken for routine analysis. During that period a large algal bloom was present in Lake Erie, which serves as drinking water reservoir of the city. The water ban was lifted on the third day as measurements showed MC levels below 1 µg/L. Ultimately, the water ban caused a major commotion as the inhabitants could only rely on bottled water which was sold out quickly. Later on, it was discussed whether the water ban was justified. It was problematic that total microcystin levels were analysed using ELISA (enzyme-linked immune-sorbent assay) techniques, which give no insight into the toxin composition and thus hazard and risk assessment was difficult (Wilson 2014). This demonstrates the previously discussed caveats in the current guidelines for MCs. A tragic accident occurred 1996 in Caruaru, Brazil. Here, hemodialysis patients were treated with MC-contaminated water of which 116 developed symptoms, which are now known as the ‘Caruaru-Syndrome’ (Carmichael et al. 2001). These symptoms occurred, as MCs could directly enter the bloodstream of the patients and thus were exacerbated because no first-pass detoxification via the liver could take place. Among these were classical hepatological symptoms like hepatomegaly, jaundice, cholestasis, cell death of hepatocytes and elevation of liver injury markers e.g. ALT, AST and GGT, as well as neurological symptoms like headache, blurred vision, tinnitus, dizziness and muscle weakness and others e.g. nausea and vomiting. Although neuronal symptoms could be observed, the hepatic symptoms were the graver ones,
17 1. General Introduction leaving 100 patients with acute liver failures in need of liver transplants. Ultimately, 76 patients died over the course of around 1.5 years (Azevedo et al. 2002; Carmichael et al. 2001). Liver biopsies of some patients revealed on average 223 ng MCs per g tissue (measured by ELISA), which in total is about 335 µg per liver. Due to limited uptake from blood and protein bound MCs, it was estimated that only 1/7th of the total MC was available for detection resulting in an estimated total MC load of 2345 µg per liver. If 120 L is assumed as the amount of water needed for hemodialysis per patient, it can be calculated that around 19.5 µg MCs per L were found in the water, if assumed that all MCs are found in the liver, which is not the case. Thus, the WHO’s guideline value for oral consumption was exceeded by at least 20 fold (Carmichael et al. 2001). As it is known from animal experiment, that intravenous (i.v.) exposure is more toxic than oral exposure (see above), the observed symptoms and deaths are easily explainable. Here again, the ELISA detection was performed with an antibody raised against MC-LR, but in a HPLC- PDA analysis also MC-YR and MC-AR could be detected (Carmichael et al. 2001), again highlighting the need of congener-dependent risk assessment. Blue-green algal supplements (BGAS) can contain MCs (Gilroy et al. 2000; Heussner et al. 2012; Lyon-Colbert et al. 2018; Roy-Lachapelle et al. 2017; Vichi et al. 2012), sometimes above the guidance value of 1 µg/g dw (dry weight) given by the Oregon Health department (Gilroy et al. 2000) or the 10 µg/g dw given by Dietrich and Hoeger (2005). Thus, the consumption of BGAS can pose a threat to humans, especially as they are marketed for many purposes, to all kinds of customers including children and manufacturers encourage taking many grams per day. Fortunately, the only documented case of toxicity due to BGAS is reported in a dog (Bautista et al. 2015), nevertheless human intoxications most probably occurred (personal communication with Prof. Daniel Dietrich). There is evidence that liver damage is increased in populations frequently exposed to cyanobacterial blooms (Chen et al. 2009; Li et al. 2011). Chen et al. (2009) detected serum levels of 0.39 ng/ml MCs (MC-RR, MC-YR and MC-LR) in the serum of Chinese fishermen, consuming an estimated amount of 2.2 - 3.9 mg MCs per day due to contamination of water- dwelling food animals. The authors found correlations between MC-levels and liver damage markers (ALT, AST, AP) indicating liver damage. A similar study with Chinese children also detected elevated AST and AP levels in children consuming an estimated amount of around 2 mg MC per day via drinking water and contaminated food items (carp and duck) (Li et al. 2011). Both studies demonstrated that low dose chronic exposure in the range of the WHO’s guideline of 0.04 mg/kg bw per day (2.8 mg for a 70 kg adult) may result in liver damage.
18 1. General Introduction
1.6 Detection methods for microcystins
Microcystin detection can be achieved using different methods with varying costs, sensitivities, and necessary efforts. Here only a brief summary is given which quickly explains the advantages and disadvantages of commonly used methods. A comprehensive review of many methods has been given by the Joint Research Council of the European Commission (Sanseverino et al. 2017). The capability of MC to inhibit PPPs can be used to determine MC content in a sample, when compared to a standard curve produced with known amounts of pure toxin (proteinphosphatase inhibition assay, PPIA). Typically a colorimetric substrate for the PPPs is used, and decrease of enzymatic activity expressed by change in colour is used for toxin detection and quantification (Heresztyn and Nicholson 2001). Previously, also the release of heavy phosphorous labelled phosphates was used as a read-out (Lambert et al. 1994). The method is cheap and gives information not only about the ability to actually inhibit PPPs but also about quantities present in a sample. Obviously, the chemical composition of the PPP-inhibiting compound cannot be determined. Thus, no information about the present congener can be gathered, nor if the observed inhibition is actually due to MC or other compounds. Many antibodies against MCs have been raised to be used in ELISA approaches. Optimally, an antibody would react to all MC congeners, thus detecting the total MC amount of a sample. Therefore, antibodies against the Adda-chain, which is present in all MC, although sometimes slightly modified, were raised e.g. by Fischer et al. (2001) which cross-react with most known congeners, i.e. all congeners displaying the Adda side chain (around 85%). With this, total MC content can be determined, although no information about congener identity or actual inhibitory potential is investigated. Nevertheless, ELISA is often used in routine analysis, due to its easy handling, its low prize and fast turnaround time. Methods employing liquid-chromatography have the ability to distinguish between MC congeners and are used frequently. To unambiguously identify congeners in a sample, corresponding pure standards are necessary to compare retention times (and possible spectra). Using HPLC, MCs in various matrices have been analysed in the past and several methods exist e.g. the method of Lawton et al. (1994) was used also for extracted samples from Caruaru patients. More recently, new chromatographic tools have been coupled to mass spectrometry (LC-MS/MS, liquid chromatography tandem mass spectrometry) which represents a major improvement in sensitivity, allowing detection limits similar or even below ELISA methods,
19 1. General Introduction but also maintaining the power to discern different congeners in a sample. UPLC-MS/MS (ultra performance liquid chromatography coupled to tandem mass spectrometry) methods additionally have the advantage of reduced time needed per analysis. For quantitation, still standards of pure reference material are needed, but using information about the mass of an analyte, e.g. a rare MC congener, these can be detected anyway. A drawback with commonly used LC-MS/MS MRM (multiple reaction monitoring) analyses still is, that only analytes are detected for which the instrument is tuned and actively looks for. Although, identification of new analytes is also possible using UPLC-MS/MS instruments e.g. in Puddick et al. (2014) or Puddick et al. (2015), this cannot be done in routine analysis. To overcome this problem, a mass spectrometric based method has been established, which detects a unique fragment of MCs (part of the Adda-chain) which is present in most congeners (MMPB, 3-methoxy-2-methyl-4- phenylbutyric acid), however this method again is not able to distinguish congeners anymore, although it gives a good measurement for the total MC content of a sample (Neffling et al. 2010a; Sano et al. 1992) although it has been shown that only around 55% of the MC content is actually transformed to MMPB in terms of molarity (Greer et al. 2017). It is unknown whether these 55% are repeatable or are true for every kind of sample. Common to most methods, is the need for extraction of MCs, especially from tissue samples or blood/serum, while detection from bloom material and water samples requires much less effort. Extraction is commonly achieved with solvents like methanol, acetonitrile and sometimes butanol, or mixtures of those. For samples from tissue, often solid phase extraction (SPE) procedures are subsequently added to get rid of contaminants which might interfere with the detection methods thus producing false signals. In the past, extraction was mostly optimized for few congeners like MC-LR, -RR and -YR, however also rather hydrophobic congeners like MC-LF, -LA or -LW are often found in blooms, highlighting the need for universally applicable extraction methods. Furthermore, in recent years it has been demonstrated that MCs are often lost to plastic material during laboratory handling (Heussner et al. 2014a; Hyenstrand et al. 2001a; Hyenstrand et al. 2001b; Rogers et al. 2015), thus complicating extraction and resulting in probably underestimated MC levels in the past.
20 2. Aim of the thesis
2. Aim of the thesis
Microcystins are a diverse family of cyanobacterial toxins as many congeners are known which differ in their amino acid composition due to variable positions, but also due to the possibility of de/methylation at various residues (Niedermeyer 2014; Spoof and Catherine 2017). Although the general composition is always a cyclic heptapeptide, the different congeners display different toxicity (Garibo et al. 2014b; Hoeger et al. 2007). Toxicity is exerted via the inhibition of serine/threonine protein phosphatases (PPP), while the major route of exposure is oral ingestion due to contaminated drinking water or food sources. Mass- occurrences of cyanobacteria in water bodies worldwide, so-called blooms, seem to increase (Huisman et al. 2018a), either by increased awareness or due to increased frequency. Furthermore, it is known that these blooms often contain more than one congener at a time (Preece et al. 2017; Turner et al. 2018b). The current risk assessment for MCs is outdated, as it is based on the knowledge that is around 20 years old. The major drawback however is that it solely focusses on MC-LR which was considered to be the most prevalent and most toxic congener. More recent data however suggests, that both assumptions are not correct, as e.g. MC-LF shows increased neurotoxic effects compared to MC-LR (Feurstein et al. 2009; Feurstein et al. 2010; Feurstein et al. 2011) and many blooms are observed in which e.g. MC-RR and MC-LA are found more frequently than MC-LR (Birbeck et al. 2019). Advances in the field of chromatography and mass-spectrometry in the recent years, e.g. UPLC-MS/MS provide more accurate quantification of samples and allow for the simultaneous detection of many analytes during one single analytical run. Consequently, methods have been developed which are able to quantify a range of MC congeners at once (Greer et al. 2016; Puddick et al. 2014; Rogers et al. 2015; Turner et al. 2017). Furthermore, as the calculation rate of computers steadily increases, bioinformatics become more widely used for the prediction of toxicity of various compounds (Krusemark 2012). Machine learning approaches, based on experimental data, could therefore be a promising tool for the toxicity classification of MC congeners, especially when the big number of known congeners are considered. Thus, the present work aims at improving individual aspects which are necessary for an up-to-date risk assessment of MCs, using the above mentioned laboratory and in-silico tools. Manuscript I deals with further recommendations about the handling of MC containing solutions to reduce adsorption of sample material to commonly used laboratory-ware such as
21 2. Aim of the thesis plastics and glass. Proper handling is prerequisite for downstream applications and provides the basis for reliable quantitation. Manuscript II describes the full synthesis of MC-LF which allows for the incorporation of modifications. This on the one hand allows for investigations of important substructures regarding their main mode of action, the inhibition of PPPs, but on the other hand gives the possibility of producing stable-isotope labelled internal standards for quantitation analysis. Consequently, Manuscript III aims at the improvement of current UPLC-MS/MS based methods for the detection and quantitation of MC from complex matrices, such as tissue and blood. Here, the previously mentioned internal standard is employed to allow for inaccuracies during handling and analysis. Together with improved extraction methods and incorporation of many congeners into the UPLC-MS/MS method this allows for stable detection of MCs from complex matrices. Finally, Manuscript IV aims at the generation of a bigger database concerning toxicodynamics of MCs (PPP inhibition), thereby employing natural and artificial MC congeners on PPP1, PPP2A and PPP5. Subsequently, the obtained data was used to train a machine learning algorithm which has the ability to classify MC congeners into toxicity classes based on their chemical structure and their PPP inhibition, which could be further used for toxicity prediction of yet non-tested MC congeners.
22 3. Manuscript I (Microcystin adsorption to lab-ware)
3. Manuscript I (Microcystin adsorption to lab-ware)
Adsorption of Ten Microcystin Congeners to Common Laboratory-Ware Is Solvent and Surface Dependent
Stefan Altaner1, Jonathan Puddick2, Susanna A. Wood3 and Daniel R. Dietrich1
1 Human and Environmental Toxicology, University of Konstanz, P.O. Box 662, 78457 Konstanz, Germany 2 Cawthron Institute, Private Bag 2, Nelson 7010, New Zealand; 3 Environmental Research Institute, University of Waikato, Private Bag 3105, Hamilton 3240, New Zealand
Published in: Toxins, 2017, 9, 129; doi:10.3390/toxins9040129
3.1 Abstract
Cyanobacteria can produce heptapeptides called microcystins (MC) which are harmful to humans due to their ability to inhibit cellular protein phosphatases. Quantitation of these toxins can be hampered by their adsorption to common laboratory-ware during sample processing and analysis. Because of their structural diversity (>100 congeners) and different physico-chemical properties, they vary in their adsorption to surfaces. In this study, the adsorption of ten different MC congeners (encompassing non-arginated to doubly-arginated congeners) to common laboratory-ware was assessed using different solvent combinations. Sample handling steps were mimicked with glass and polypropylene pipettes and vials with increasing methanol concentrations at two pH levels, before analysis by liquid chromatography- tandem mass spectrometry. We demonstrated that MC adsorb to polypropylene surfaces irrespective of pH. After eight successive pipet actions using polypropylene tips ca. 20% of the MC were lost to the surface material, which increased to 25%–40% when solutions were acidified. The observed loss was alleviated by changing the methanol (MeOH) concentration in the final solvent. The required MeOH concentration varied depending on which congener was present. Microcystins only adsorbed to glass pipettes (loss up to 30% after eight pipet actions) when in acidified aqueous solutions. The latter appeared largely dependent on the presence of ionisable groups, such as arginine residues.
23 3. Manuscript I (Microcystin adsorption to lab-ware)
3.2 Introduction
MC are a family of structurally related heptapeptides, produced by some freshwater cyanobacteria. They are potent inhibitors of cellular serine/threonine-protein-phosphatases (ser/thr-PP) (Mackintosh et al. 1990). Consequently, when humans and animals are exposed (mainly via ingestion), the liver is primarily affected due to the first-pass effect and the expression of OATP (Fischer et al. 2005). Toxicodynamics are characterized by irreversible PPP inhibition and subsequent protein hyper-phosphorylation leading to loss of cell structure, apoptosis, and necrosis (Falconer and Yeung 1992; Fladmark et al. 1999). Acute intoxications result in human morbidity and occasional mortality (WHO et al. 1999a), while chronic exposure is associated with increased incidences of liver tumours (Fujiki and Suganuma 1993;
Nishiwaki-Matsushima et al. 1992). Structurally, MC consist of proteinogenic (L-) and non- proteinogenic (D-) amino acids with a generalized sequence of cyclo-[D-Ala]-[X]-[D-Masp]-
[Z]-[Adda]-[D-Glu]-[Mdha] (Figure 3-1). Here, [X] and [Z] are variable L-amino acids, Mdha is N-methyldehydroalanine, and D-Masp is D-methylaspartic acid. The amino acid Adda, (2S,3S,8S,9S,4E,6E)-3-amino-9-methoxy-2,6,8-trimethyl-10-phenyl-4,6-decadienoic acid, is a unique amino acid, a priori only observed in MC and the related nodularins (Carmichael 1997). Along with the two variable positions X and Z, structural differences in the congeners arise from variable methylation/demethylation patterns (reviewed in (Codd et al. 1999a)). The amino acids in the X- and Z-positions define the name of the respective congener, for example, MC- LR has leucine (L) in the X-position and arginine (R) in the Z-position (Figure 3-1, Table 3-1). To date, more than 100 congeners have been identified (Niedermeyer 2014). This high number of congeners makes analysis challenging, as physicochemical parameters, such as hydrophobicity or ion formation, differ according to the functional groups of the amino acids in the variable positions.
Figure 3-1: General structure of a microcystin. The structure contains the unusual amino acids Adda ((2S,3S,8S,9S,4E,6E)-3-amino- 9-methoxy-2,6,8-trimethyl-10-phenyl-4,6-decadienoic acid) and Mdha (N- methyldehydroalanine), as well as (L-) and (D-) amino acids.
24 3. Manuscript I (Microcystin adsorption to lab-ware)
Table 3-1: Microcystin congeners subjected to the experimental procedures described here. MC X2 Z4
MC-RR L-Arginine L-Arginine
MC-YR L-Tyrosine L-Arginine
MC-LR L-Leucine L-Arginine
MC-FR L-Phenylalanine L-Arginine
MC-WR L-Tryptophan L-Arginine
MC-RA L-Arginine L-Alanine
MC-RAba L-Arginine L-Aminobutyric acid
MC-LA L-Leucine L-Alanine
MC-FA L-Phenylalanine L-Alanine
MC-WA L-Tryptophan L-Alanine The amino acids in the X2 and Z4 positions denote the abbreviated name for the individual congeners. For example MC-LR has L-Leucine in position X2 and L- Arginine in position Z4.
Eutrophication and rising temperatures have been associated with increased mass occurrences (blooms) of toxin-producing cyanobacteria globally (Harke et al. 2016; Paerl and Huisman 2008). Consequently, and in conjunction with an ever increasing world population and enhanced use of water per capita (UN 2008; UN 2015), the risk for human intoxication by MC is also increasing. Thus, reliable analysis (identification and quantification) of these toxins is critical to sustain safe drinking and recreational waters. Potential hazard and exposure assessment currently employs the congener addition concept, whereby all microcystin congeners present in a sample are expressed as MC-LR equivalents (reviewed in (Ibelings et al. 2015)). However, the latter approach not only wrongly assumes that all microcystins have the same toxicokinetics and dynamics (Feurstein et al. 2010; Feurstein et al. 2011; Fischer et al. 2010), but also presumes that all congeners behave similarly during sample processing and handling. Indeed, previous analyses have demonstrated that losses of different congeners can occur during sample handling (Heussner et al. 2014a; Hyenstrand et al. 2001a; Hyenstrand et al. 2001b; Rogers et al. 2015). Thus, it is crucial to analyse all congeners present in a certain sample, without loss due to pre-analysis handling. Losses during handling can happen during transfer steps or storage through adsorption to commonly used surfaces like polypropylene or various kinds of glass material. Studies have shown that MC adsorption to plastic-ware was dependent on the solvent used (i.e., methanol or acetonitrile concentration) and differed according to the structural characteristics of the MC congeners (Heussner et al. 2014a; Hyenstrand et al. 2001a; Hyenstrand et al. 2001b). The pH of the solvent can affect adsorption, with acidified methanol
25 3. Manuscript I (Microcystin adsorption to lab-ware) leading to less adsorption of MC to GF/C filters, which are commonly used during sample preparation, compared to non-acidified solvents (Rogers et al. 2015). However, currently there are no uniform suggestions on how to handle MC-containing samples and solutions. Microcystin congeners might behave differently in various solvents, depending on the amino acids present in positions X and Z. Due to the presence of certain functional groups, such as carboxyl groups or guanidine moieties, in the variable positions of the MC structure, differences in hydrophobicity could affect the extent of MC loss. In the present study, the influence of acidification and methanol (MeOH) concentration on the loss of ten structurally-diverse MC congeners to polypropylene and glass surfaces was assessed.
3.3 Material and Methods
3.3.1 Reagents and Laboratory-Ware
The Milli-Q water used was deionized water with 18.2 MΩ purified using a Milli-Q Plus ultra-pure water system still (Millipore, Billerica, MA., United States of America; USA). Formic acid (FA; Merck, Kenilworth, NJ, USA) and acetonitrile (Honeywell, Morris Plains, NJ, USA) were of MS grade. Analytical standards for MC-RR, MC-LR, and MC-YR were from DHI LAB products (Hoersholm, Denmark). Liquid Chromatography vials were made from clear glass or polypropylene with lids having pre-split septa (Phenomenex; Torrance, CA, USA).
3.3.2 Production of Microcystin Congener Stock Solution
A lyophilised extract of Microcystis CAWBG11, containing di-arginated, mono- arginated, and non-arginated MC congeners (Puddick et al. 2014), was dissolved in water. For a working stock, the extract was diluted in water to a nominal concentration of 100 ng/mL of MC-RR whilst other congeners ranged from 6.5 ng/mL (MC-RAba) to 314.5 ng/mL (MC-LR). Only the results for congeners MC-RR, -YR, -LR, -FR, -WR, -RA, -RAba, -LA, -FA, and -WA are presented. These were chosen as they were present in sufficient amounts to be reliably quantified with the UPLC-MS/MS method used. Additional congeners present in the extract (desmethyl and didesmethyl variants of MC-RR and MC-LR and Aba containing variants of MC-LA, -FA, and -WA) were also detected, but the amounts were too low to reliably assess the loss of those congeners to adsorption. In general, we observed that these modified congeners followed a similar adsorption pattern as their basic chemical entity (e.g., desmethyl-MC-LR showed a similar trend as MC-LR).
26 3. Manuscript I (Microcystin adsorption to lab-ware)
3.3.3 Adsorption of Microcystins to Common Pipetting Laboratory-Ware in Non- Acidified and Acidified Solvents
The MC working stock was diluted 1:4 with solutions of differing methanol concentrations (0%, 25%, 50%, 75%, and 100%) which were either neutral or acidified with 0.1% formic acid (FA) resulting in solutions with 0%, 20%, 40%, 60%, and 80% methanol. Using glass Pasteur pipettes, the different MC solutions were distributed into the individual glass vials containing 0.3 mL inserts, preparing triplicates for each solution (volume 0.2 mL). Afterwards the individual samples were subjected to either 0, 1, 2, 4, or 8 pipetting actions using a 200 µL auto-pipettor with polypropylene pipette tips (AXYGTR-222-C-L; Axygen, Corning, NY, USA) or glass Pasteur pipettes. A new pipette tip/Pasteur pipette was used for each pipetting action to ensure no saturation of the tip binding capacities. The solutions were then analysed by UPLC-MS/MS to determine the MC concentration. Loss due to adsorption to the pipette tips was defined as the difference in concentrations found between the control and the resulting MC concentration in the solvent after pipetting.
3.3.4 Short-Term Storage in Glass or Polypropylene Vials
The MC working stock was again diluted 1:4 into the same matrices as above to achieve a nominal concentration of 20 ng/mL MC-RR. The solution was distributed into the individual samples in either glass LC-vials (Verex AR0-3010-13; Phenomenex) or polypropylene LC- vials (Verex AR0-9994-13-C; Phenomenex) using a three times equilibrated (take in and blow out of solution) polypropylene pipette tip. A sample volume of 600 µL was used in the glass vials and 512 µL for polypropylene vials. The different volumes were used to ensure that the solutions cover the same surface area in the respective vials (calculated from the inner diameter of the respective vials). Vials were then incubated for 1 h at 100 rpm on a rotator to mimic sample handling and stored at 4 °C until analyses (around 1 h later).
3.3.5 Ultra-performance liquid Chromatography-Tandem Mass Spectrometry (UPLC- MS/MS) Detection of Microcystins
For MC analyses, an Acquity I-class liquid chromatograph (Waters, Milford, MA.,
USA) equipped with a Waters Acquity BEH C18 column (1.6 µm, 2.1 × 50 mm) coupled to a XEVO TQ-S mass spectrometer was used (Waters). Column temperature was set to 40 °C. The used solvents were A: 10% acetonitrile + 100 mM FA and 6 mM ammonia; B: 90% acetonitrile
27 3. Manuscript I (Microcystin adsorption to lab-ware)
+ 100 mM FA and 6 mM ammonia. The used gradient was the same as published previously (Puddick et al. 2016b). The injected sample volume was 5 µL. The used masses and transitions for the investigated congeners can be found in Table 3-2. Compounds entering the mass spectrometer were ionized using a capillary voltage of 1.5 kV and a nebulizer pressure of 7.0 bar. For desolvation, a nitrogen flow of 1000 L/h at 500 °C was used.
Table 3-2: Mass spectrometric parameters of the used microcystins (MC). Parent Daughter Cone Collision Congener (m/z) (m/z) Voltage (V) Energy (V) MC-RR 519.7 135.1 40 27 MC-YR 1045.5 135.1 40 70 MC-LR 995.5 135.1 40 65 MC-FR 1029.5 135.1 40 65 MC-WR 1068.5 135.1 40 65 MC-RA 953.5 135.1 40 65 MC-RAba 967.5 135.1 40 65 MC-LA 910.6 135.1 40 65 MC-FA 944.6 135.1 40 65 MC-WA 983.6 135.1 40 65
3.3.6 Outlier Analysis
Outlier analysis was performed by two-sided Grubb’s test using an online tool provided on the homepage of GraphPad Prism software at a significance level of p < 0.05 (GraphPad). In order to define outlier(s) and to determine their potential effects on the results and interpretation, data were statistically analysed (see below) with and without the inclusion of the presumed outlier(s).
3.3.7 Data Handling and Statistical Analyses
Concentrations of MC were quantified using MassLynx 4.1 software (Waters). Microsoft Excel was used to process the data of the individual samples. Means and standard deviations (SD) were generated from the triplicates. Means were normalized to control values, which were set as 100%. Percentile standard deviation was calculated by normalizing standard deviations to the mean of the control. Statistical analyses were performed using GraphPad Prism 5 software (GraphPad Software Inc., La Jolla, CA, USA). Statistical tests were either a 2-way-
28 3. Manuscript I (Microcystin adsorption to lab-ware)
ANOVA with Bonferroni post-test, or a Student’s t-test and were indicated in the figure legends of the individual graphs or tables. Means ± SD of triplicate analyses were depicted.
3.4 Results
3.4.1 Adsorption of Microcystin Congeners to Polypropylene Pipette Tips in Aqueous and High-Percentage Methanol Solutions
A structurally-diverse range of MC congeners including di-arginated, mono-arginated, and non-arginated variants were used to assess the effect of solvent composition on MC adsorption to polypropylene laboratory-ware. MC were dissolved in either water, 80% methanol, or acidified versions (0.1% formic acid added, pH around 2.7) thereof and distributed into identical LC-MS/MS vials for each MC congener–solvent combination. While one set served as a control (assumed 100% MC recovery) the others were used to investigate adsorption to the pipette tips during pipetting (1, 2, 4, or 8 pipette aspirations and releases). When the MC congeners were dissolved in non-acidified water, no statistically significant loss due to adsorption was observed after one pipetting step (Figure 3-2A, Table S3-1). However, two pipetting steps resulted in a significant (p < 0.05) adsorption (10% of total) for all congeners (except for MC-LA). Similarly, a loss of ca. 20% of their original concentration (p < 0.001, Figure 3-2A, Table S3-1) was observed after four and eight pipetting steps for all MC congeners. A significant (p < 0.01) loss of >5% resulted for all MC congeners, except for MC-RR and MC-YR, when they were dissolved in acidified water (pH = ~2.7) already after one pipetting action. Most MC congeners encountered further losses due to adsorption with increasing pipetting steps, whereby after the eighth step, the total loss for most congeners reached 40%, resulting in 60% of the concentration at the outset (Figure 3-2B, Table S3-1). In contrast to other MC congeners, MC-RR showed a significant (p < 0.01) loss to pipette adsorption only after four (and eight) pipetting steps. Overall, the use of acidified water resulted in approximately double the loss than observed for non-acidified aqueous solutions. No significant reduction in MC concentration was observed with one to eight pipetting actions when MC were dissolved in non-acidified 80% methanol, (Figure 3-2C). This was also apparent when the MC were dissolved in 80% methanol acidified with 0.1% FA (Figure 3-2D). Within these experiments, MC-RA when four pipetting were used (Figure 3-2D), showed surprisingly large data variability (large standard deviation). The latter was tested for outliers using the Grubb’s test, which indicated that one of the raw data points could be an outlier, an
29 3. Manuscript I (Microcystin adsorption to lab-ware) interpretation that is supported by the fact that all other values for MC-RA, irrespective of the number of pipetting steps, did not differ significantly from the 100% control. The side-by-side comparison of aqueous and acidified aqueous solutions (Figure 3-3) moreover demonstrated that non-arginine-containing congeners experienced significant adsorption loss in the acidified solution already after the first 1–4 pipetting steps (p < 0.01, Figure 3-3A–C), whereas this was the case for all MC congeners (except for MC-RR) after eight pipetting steps (Figure 3-3D).
Figure 3-2: Reduction of various microcystin (MC) congeners in acidified and non-acidified solvents after increasing steps of pipetting using polypropylene pipette tips. MC were spiked into water (A); acidified water (B); 80% methanol (C); or acidified 80% methanol (D); and subjected to increasing numbers of pipetting steps. Acidification was achieved using 0.1% formic acid. Significance levels are not inserted in the graph for clarity reasons, but are available in Table S3-1.
30 3. Manuscript I (Microcystin adsorption to lab-ware)
Figure 3-3: Comparison of water and acidified water for the individual repeats of successive pipetting steps. The data used for Figure 3-2 was reanalysed comparing only water (black bars) and acidified water (grey bars) directly with each other following additional pipetting steps, one (A); two (B); four (C); and eight additional steps (D). * p < 0.05, ** p < 0.01, *** p < 0.001.
3.4.2 Effect of Methanol Concentration on the Adsorption of Selected Microcystins to Polypropylene Pipette Tips
Adsorption of MC-RR, -FR, -WR, -RAba, and -RA was completely avoided in aqueous solutions containing ≥40% MeOH, whereas an adsorption loss was observed after eight pipetting steps in solutions containing 20% MeOH (Figure 3-4A/C and Figure S3-1). Similarly, all other congeners tested showed a lack of adsorption in all solutions containing MeOH, irrespective of the number of pipetting steps (Figure 3-4 and Figure S3-1).
31 3. Manuscript I (Microcystin adsorption to lab-ware)
Figure 3-4: Effect of methanol concentration on the adsorption of selected microcystins to polypropylene pipette tips. Individual samples were subjected to eight successive pipetting steps. Representative congeners (double-, single-, non-arginated) include MC-RR (A); MC-LR (B); MC- RA (C); and MC-LA (D). Controls were performed without using polypropylene pipette tips (0 pipetting steps) in triplicates. Small letters represent significance levels at the individual pipetting steps: a, water vs. 20% MeOH; b, water vs. 40% MeOH; c, water vs. 60% MeOH; d, water vs. 80% MeOH; e, 20% MeOH vs. 40% MeOH; f, 20% MeOH vs. 60% MeOH; g, 20% MeOH vs. 80% MeOH. Significance levels are represented by the repetition of the letters, e.g., a, p < 0.05; aa, p < 0.01; aaa, p < 0.001.
3.4.3 Effect of Acidified Methanol Concentration on the Adsorption of Selected Microcystins (MC) to Polypropylene Pipette Tips
Acidification of aqueous and aqueous-MeOH solutions resulted in increasing adsorption of all MC congeners with increasing pipetting steps, except for the doubly-arginated MC-RR which showed no decrease in 20% MeOH (Figure 3-5 and Figure S3-2). However, for all congeners no adsorption was observed when the MeOH concentrations were increased to ≥40%.
32 3. Manuscript I (Microcystin adsorption to lab-ware)
Figure 3-5: Effect of acidified methanol concentration on the adsorption of selected microcystins to polypropylene pipette tips. Individual samples were subjected to eight successive pipetting steps. Representative congeners (double-, single-, non-arginated) include MC-RR (A); MC-LR (B); MC- RA (C); and MC-LA (D). Solvents were acidified by adding 0.1% (v/v) of formic acid (FA) to the individual solvents. Controls were performed without using polypropylene pipette tips (0 pipetting steps) in triplicate. Small letters represent the significance levels at the individual pipetting steps: a, water vs. 20% MeOH; b, water vs. 40% MeOH; c, water vs. 60% MeOH; d, water vs. 80% MeOH; e, 20% MeOH vs. 40% MeOH; f, 20% MeOH vs. 60% MeOH; g, 20% MeOH vs. 80% MeOH. Significance levels are represented by the repetition of the letters, e.g., a, p < 0.05; aa, p < 0.01; aaa, p < 0.001.
3.4.4 Adsorption of Selected Microcystins (MC) in Acidified and Non-Acidified Aqueous Solutions to Glass-Ware (Pasteur Pipettes)
No adsorption was observed in non-acidified aqueous solutions, irrespective of the number of pipetting steps with glass pipettes (Figure 3-6 and Figure S3-3A). However, similar to the observations in polypropylene pipette tips, acidification of the aqueous solutions resulted in a significant adsorption of all MC congeners observed at ≥2 pipetting steps with glass pipettes
33 3. Manuscript I (Microcystin adsorption to lab-ware)
(Figure 3-6B–D). The degree of adsorption appeared more pronounced with glass than with polypropylene pipettes (Figures 3 and 6).
Figure 3-6: Multiple pipetting action using glass (Pasteur) pipettes. Microcystin (MC) amounts in water (water, black bars) and acidified water (water + 0.1% Formic Acid (FA), grey bars) after one (A); two (B); four (C); and eight steps (D) of pipetting actions using glass (Pasteur) pipettes. * p < 0.05, ** p < 0.01, *** p < 0.001.
3.4.5 Effect of Acidified Methanol Concentration on the Adsorption of Selected Microcystins to Glass-Ware (Pasteur Pipettes)
As non-acidified aqueous solutions resulted in no detectable adsorption of the ten MC congeners tested (Figure S3-3 and Table S3-2), the question was raised whether acidification would increase the adsorption of MC congeners even in solutions containing MeOH. Indeed, as expected, acidified solutions containing no MeOH resulted in increasing adsorption with increasing number of pipetting steps, whereby this adsorption was alleviated by the addition of MeOH to the solutions (Figure 3-7). Of interest, however, was the finding that arginine containing MC congeners appeared also affected by the acidification of aqueous solutions even
34 3. Manuscript I (Microcystin adsorption to lab-ware) when 20% MeOH was added, whereas this was not the case for MC congeners without arginine residues, e.g., MC-LA (Figure 3-7 and Figure S3-4).
Figure 3-7: Effect of acidified methanol concentration on the adsorption of selected microcystins to glass-ware (Pasteur pipettes). Individual samples were subjected to eight successive pipetting steps. Representative congeners (double-, single-, non-arginated) include MC-RR (A); MC-LR (B); MC- RA (C); and MC-LA (D). Solvents were acidified by adding 0.1% (v/v) of formic acid (FA) to the individual solvents. Controls were performed without additional pipetting steps after distribution to the sample vials. Small letters represent significance levels at the individual pipetting steps: a, water vs. 20% MeOH; b, water vs. 40% MeOH; c, water vs. 60% MeOH; d, water vs. 80% MeOH; e, 20% MeOH vs. 40% MeOH; f, 20% MeOH vs. 60% MeOH; g, 20% MeOH vs. 80% MeOH; h, 40% MeOH vs. 60% MeOH. Significance levels are represented by the repetition of the letters, e.g., a, p < 0.05; aa, p < 0.01; aaa, p < 0.001.
3.4.6 Short Term Storage of MC Solutions in Glass or Polypropylene Vials
The comparison between glass or polypropylene vials demonstrated no difference in congener adsorption (Figure S3-5) after 2 h, irrespective of whether the solutions were acidified or not. However, the addition of MeOH to the aqueous solution resulted in lower adsorption of
35 3. Manuscript I (Microcystin adsorption to lab-ware) all congeners when compared to the aqueous solution without MeOH, suggesting that the use of MeOH will result in lowered losses during the analytical process.
3.5 Discussion
Microcystins are known to be present in different charge states according to the pH of the environment they are in (de Maagd et al. 1999; Liang et al. 2011; Rivasseau et al. 1998), and the different ionisable functional groups they possess. Most MC congeners have two carboxylic groups (on Masp2 and Glu6, Table 3-1) and zero to two guanidine groups, depending on whether there are arginine residues present in the two variable positions X and Z. The fact that acidification of the solution (whether water or MeOH-water mixtures) resulted in an increased loss of MC congeners, primarily in glass but also to some extent in polypropylene pipettes, and is most likely due to different ionization states of the MC congeners at different pH. Indeed, the differences observed appear pronounced between MC congeners containing no, one, or two arginine residues (Table 3-3). At very low pH, MC are likely to be positively charged, as carboxyl groups and guanidine groups, if present, are protonated (de Maagd et al. 1999; Liang et al. 2011; Rivasseau et al. 1998). With increasing pH, the carboxylic groups initially become deprotonated yielding neutral (MC-RR) or single-negatively charged compounds (MC-XRs), followed later by the guanidine groups, resulting in double-negatively charged compounds in basic solutions. MC congeners not containing arginine residues most likely are double-negatively charged at lower pH values than arginated ones. At which pH MC are present in a neutral state or single charged (+ or −) state is not easily discerned. The calculated pKa values of MC-RR, MC-LR, and MC-YR are all similar and lie around pKa 3.5 (Rivasseau et al. 1998). For non-arginated congeners, no pKa values are known, but they could be higher as there is no ionisable guanidine group present. The adsorption to polypropylene surfaces can be seen similar to the mechanism of reversed-phase chromatography. In this case, the polypropylene surface of the used pipette tip resembles the stationary phase, to which analytes may adhere and therefore are not able to be analysed via liquid chromatography-tandem mass spectrometry (LC-MS/MS), as they are no longer present in solution. The solvent used resembles the elution solution (mobile phase). The higher the elution strength of a solvent towards the respective MC congener, the more likely it will stay in solution, rather than adhering to the pipette tips. To successfully retain the MC congeners in solution, the polarity of the solvent must be decreased. Water has a polarity index around twice as high as methanol (10.2 vs. 5.1) (Snyder 1978). By adding methanol to water
36 3. Manuscript I (Microcystin adsorption to lab-ware) the polarity of the solution decreases and provides the MC congener with an environment which is more favourable than the surface of the pipette tip. In acidic environments the polarity of MC changes. After acidification (the pH of 0.1% FA is ca. 2.7), MC are expected to be ionized due to the deprotonation of their carboxylic functions (pKa ca. 3.5 (Rivasseau et al. 1998)). It is difficult to explain the higher retention to polypropylene in this situation, but an ion-pair retention mechanism could be causing the increased adsorption in acidic solutions (Rivasseau et al. 1998). Negatively charged ions (here: − formate, HCO2 ) would interact with positively charged guanidine groups forming uncharged complexes. These subsequently may interact with the polypropylene (“stationary phase”) more favourably and require higher methanol concentrations to replace them. Additionally, the guanidine groups may also interact with carboxyl groups (from the same MC molecule or from another one in proximity) again forming complexes which are uncharged, thus increasing their adsorption to the uncharged polypropylene surface. Corroborating observations from previous studies (Heussner et al. 2014a; Hyenstrand et al. 2001a; Hyenstrand et al. 2001b), the lipophilicity/hydrophilicity of the MC congeners (Table 3-3) in our study also appeared to be governed by their adsorptivity to non-charged pipetting material, especially as MC congeners dissolved in water, and adsorbed to polypropylene laboratory-ware increasingly with the number of pipetting steps. The microscopic structure of the pipette tips used in this study (Axygen, Maxymum Recovery) is a relatively smooth surface, as protruding polypropylene chains are removed using acid treatment by the manufacturer. Because of this, the uncharged hydrophobic MC congeners might adhere to this smooth uncharged surface due to hydrophobic interactions. During the adsorption to glass surfaces, the acidity/alkalinity of the surface may play a major role. Glass consists of glass formers (SiO2, B2O3, P2O5, etc.), glass modifiers (Na2O,
K2O, CaO, etc.), and intermediates (BeO, MgO, Al2O3, etc.). Generally speaking, the formers are Lewis acids, the modifiers are Lewis bases, and the intermediates are amphoteric (Sun and Silverman 1945). As glass formers are the major part of (soda lime) glass, the surface is most likely negatively charged in aqueous solutions, as dissociation occurs at the silanol groups (Behrens and Grier 2001). In acidic environments, as used during the present study (around pH 2.7), the silanol groups are most likely still in an ionic (negative) state, as they would require extremely low pH to become protonated (Hiemstra et al. 1989). Therefore, the increased adsorption which was observed in acidic solvents, most likely occurred through the differential ionization of the MC congeners as opposed to that of the glass surface. As discussed earlier, in acidic environments MC are likely to be present in protonated form. As the surface of the soda-
37 3. Manuscript I (Microcystin adsorption to lab-ware) lime glass is negatively charged and MC are positively charged in acidic environments, interactions between silanol groups of the glass surface and the guanidine group of MC are most likely leading to increased adsorption. A previous study (Rogers et al. 2015) observed an effect from solvent acidification on the amounts of MC binding to GF/C filters (glass-fibre, class C). The researchers found that arginine-containing MC adhered to the GF/C material when in neutral methanol. In acidified methanol, the adsorption was absent for singly-arginated MC and only partially reduced for MC-RR. When we tested the effects of solvent acidification on MC adsorption to soda-lime glass (as opposed to the borosilicate glass used in GF/C filters), the MC adsorption was enhanced in acidified water. However, MC adsorption to soda-lime glass was not observed in non-acidified water or methanol solutions. When acidified solutions were supplemented with 40% methanol, the loss of MC was negated. Heussner et al. observed no loss of arginine-containing MC congeners when using borosilicate glass vials as storage containers for MC (Heussner et al. 2014a). In their study, solutions were stored in 5% non-acidified methanol and there was no control using an acidified storage solution. The researchers assessed the recovery after up to two months and did not observe any losses of MC-RR, MC-LR, or MC-YR, but losses were noted for MC-LA, MC- LF, and MC-LW. In light of the results presented here, it might be possible that Heussner et al. did not actually see loss due to storage in glass, but due to the repeated contact with polypropylene tips during sample preparation for the ELISA analysis used to measure MC in the study. When using ELISA detection for MC, the maximum amount of methanol in the final solution is less than 5%. This may have led to the apparent loss of non-arginated congeners but not arginine-containing MC. In our study, short-term storage of MC congeners in solution with low methanol concentrations (0% – 40%) did not differ when polypropylene or glass vials were used. A recent study also using ELISA reported that MC-LR, MC-LA, and MC-LF adsorb to various surfaces like polypropylene, polystyrene, high-density polyethylene, and polycarbonate when stored up to 120 h (Kamp et al. 2016). The amount of absorbed toxin was dependent on the type of water the toxins were in. The authors saw the highest adsorption in deionized water, followed by chlorine-quenched drinking water and surface water. The quenching of chlorine in the drinking water was crucial, as MC were rapidly degraded without quenching. Only glass and polyethylene terephthalate could stop adsorption effectively from the water types tested. As all materials of the storage containers were compared to glass and not to absolute recovery, it cannot be excluded that loss occurred (through the pipetting steps necessary for the ELISA).
38 3. Manuscript I (Microcystin adsorption to lab-ware)
It is also possible that the quenching did not actually lead to degradation, but altered the polarity of the present congeners and increased their adsorptive behaviour. Collectively, our results and those from previous studies show that under non-acidified conditions, 5% methanol is sufficient to limit the loss of arginated MC congeners to polypropylene surfaces, but at least 40% methanol should be used, if possible, to reduce the loss of other (non-arginated) congeners. The abundance of non-arginated congeners present in environmental samples might be higher than reported to-date, as they are more easily lost during sample preparation procedures.
Table 3-3: Percentage of methanol needed to counteract the loss of microcystins from acidic and non-acidic solutions. Microcystin Variant Non-Acidified Acidified Glass Polypropylene Glass Polypropylene Doubly-arginated 0 (Milli-Q 20 40 20 (more hydrophilic) water) Singly-arginated 0 (Milli-Q 40 20 40 (amphiphilic) water) Non-arginated 0 (Milli-Q 40 20 40 (more lipophilic) water)
3.6 Conclusions
In the present study, MC congeners were shown to adsorb to two common materials used during laboratory handling of samples: polypropylene and soda-lime glass. The level of adsorption was dependent on the structure and thus the physico-chemical properties of the MC congeners, the acidity/polarity of the solvent in which the MC were dissolved, and how many pipetting actions were performed. Microcystin congeners dissolved in water adsorbed to polypropylene and this was more severe when the solutions were acidified. The adsorption of MC to soda-lime glass was only apparent in acidified solutions. Under acidic conditions, the number of guanidine moieties (in the arginine residues) in the individual congeners is critical for the extent of adsorption (Table 3-3). Addition of methanol rectifies the observed loss. In order to avoid adsorption when working with MC-containing solutions, it is recommended that a methanol concentration of ≥40% and a neutral pH is used. The latter recommendation differed from those provided in previous studies (Heussner et al. 2014a; WHO et al. 1999b; WHO et al. 1999c; WHO et al. 1999d). Of importance for researchers working with MC congeners is, to recognize the fact that congeners differ in their ability to be ionized. The latter is a critical determinant for their polarity, behaviour in solution, and thus adsorption.
39 3. Manuscript I (Microcystin adsorption to lab-ware)
In consequence, disregarding the physico-chemical properties of MC congeners and thus not using the appropriate solvents will result in massively different recoveries of the original sample concentrations and therefore possibly to analytical results misrepresenting the true situation in the sample.
40 3. Manuscript I (Microcystin adsorption to lab-ware)
3.7 Supplementary material
Figure S3-1: Reduction of all investigated microcystin (MC) congeners in solutions with decreasing methanol (MeOH) concentrations due to adsorption to polypropylene. Individual samples were subjected to eight successive pipetting steps. A: MC-RR, B: MC-YR, C: MC-LR, D: MC-FR, E: MC-WR, F: MC-RA, G: MC-RAba, H: MC-LA, I: MC-FA, J: MC-WA. Controls were performed without using polypropylene pipette tips in triplicates. Small letters represent significance levels at the individual pipetting steps: a, water vs. 20% MeOH; b, water vs. 40% MeOH; c, water vs. 60% MeOH; d, water vs. 80% MeOH; e, 20% MeOH vs.40% MeOH; f, 20% MeOH vs. 60% MeOH; g, 20% MeOH vs.80% MeOH; h, 40% MeOH vs. 60% MeOH; i, 40% MeOH vs. 80% MeOH. Significance levels are represented by repetition of the letters, e.g.: a, p < 0.05; aa, p < 0.01; aaa, p < 0.001.
41 3. Manuscript I (Microcystin adsorption to lab-ware)
Figure S3-2: Reduction of all investigated microcystin (MC) congeners in solutions with decreasing acidic methanol concentrations using polypropylene pipet tips. Individual samples were subjected to eight successive pipetting steps. A: MC-RR, B: MC-YR, C: MC-LR, D: MC-FR, E: MC-WR, F: MC-RA, G: MC-RAba, H: MC-LA, I: MC-FA, J: MC-WA. Controls were performed without using polypropylene pipette tips in triplicates. Small letters represent significance levels at the individual pipetting steps: a, water vs. 20% MeOH; b, water vs. 40% MeOH; c, water vs. 60% MeOH; d, water vs. 80% MeOH; e, 20% MeOH vs.40% MeOH; f, 20% MeOH vs. 60% MeOH; g, 20% MeOH vs.80% MeOH; h, 40% MeOH vs. 60% MeOH; i, 40% MeOH vs. 80% MeOH. Significance levels are represented by repetition of the letters, e.g.: a, p < 0.05; aa, p < 0.01; aaa, p < 0.001.
42 3. Manuscript I (Microcystin adsorption to lab-ware)
Figure S3-3: Reduction of various microcystin congeners in acidified and non- acidified solvents after increasing steps of pipetting using Pasteur pipettes. Microcystins were spiked into different solvents and submitted to the indicated pipetting steps. A: water; B: water + 0.1% formic acid, C: 80% methanol, D: 80% methanol + 0.01% formic acid. Significance levels are not inserted in the graph for clarity reasons but can be found in Table S3-2.
43 3. Manuscript I (Microcystin adsorption to lab-ware)
Figure S3-4: Reduction of all investigated microcystin (MC) congeners in solutions with decreasing acidic methanol (MeOH) concentrations using Pasteur pipettes. Individual samples were subjected to eight successive pipetting steps. A: MC-RR, B: MC-YR, C: MC-LR, D: MC-FR, E: MC-WR, F: MC-RA, G: MC-RAba, H: MC-LA, I: MC-FA, J: MC-WA. Controls were performed without using polypropylene pipette tips in triplicates. Small letters represent significance levels at the individual pipetting steps: a, water vs. 20% MeOH; b, water vs. 40% MeOH; c, water vs. 60% MeOH; d, water vs. 80% MeOH; e, 20% MeOH vs.40% MeOH; f, 20% MeOH vs. 60% MeOH; g, 20% MeOH vs.80% MeOH; h, 40% MeOH vs. 60% MeOH; i, 40% MeOH vs. 80% MeOH. Significance levels are represented by repetition of the letters, e.g.: a, p < 0.05; aa, p < 0.01; aaa, p < 0.001. FA, formic acid.
44 3. Manuscript I (Microcystin adsorption to lab-ware)
Figure S3-5: Effect of short term storage of microcystins in glass or polypropylene vials. Microcystins were spiked into non-acidified (A, C, E) and acidified (B, D, F) water (A+B), 20% methanol (C+D), 40% methanol (E+F) and distributed to either glass (black bars) or polypropylene LC vials (grey bars) before short term storage (∼2 h). * p < 0.05, ** p < 0.01, *** p < 0.001
.
45 3. Manuscript I (Microcystin adsorption to lab-ware)
Table S3-1: Significance levels for Figure 3-2. MC- Congener MC-RR MC-YR MC-LR MC-FR MC-WR MC-RA MC-LA MC-FA MC-WA RAba Experiment A B A B A B A B A B A B A B A B A B A B 1 - - - - - ** - *** - *** - *** - ** - *** - *** - *** Pipette 2 *** - *** ** ** - ** ** ** ** *** - *** - - *** * *** ** ***
4 *** ** *** *** *** *** *** *** *** *** *** *** *** ** *** *** *** *** *** *** actions 8 *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** A (water) and B (water + 0.1% v/v formic acid) correspond to the individual graphs of Figure 3-2. Experiments C (80% MeOH) and D (80% MeOH + 0.1% v/v formic acid) are not shown, as there were no significant differences to controls detected. Percentages of microcystins (MC) after the individual repetitions of pipetting steps are compared to the control (0 pipette actions). * p < 0.05, ** p < 0.01, *** p < 0.001, - no significance.
Table S3-2: Significance Levels for Figure S3-3. MC-RR MC-YR MC-LR MC-FR MC-WR MC-RA MC-RAba MC-LA MC-FA MC-WA
Experiment A B C D A B C D A B C D A B C D A B C D A B C D A B C D A B C D A B C D A B C D
1 ------* - - - *** ------* ------* * - Pipette 2 ------** - - - *** - - - *** ------*** - * - * - - - *** - - - *** - - actions 3 - *** - - - *** - - - *** - - - *** - - - *** - - - *** - - - *** - - - *** - - * *** - - - *** - ** 4 - *** - - - *** - - - *** - - - *** - - - *** - - - *** - - * *** - - - *** - - - *** - - - *** - - A (water), B (water + 0.1% v/v formic acid), C (80% methanol) and D (80% methanol + 0.1% FA) correspond to the individual graphs of Figure S3-3. Percentages of microcystins (MC) after the individual repetitions of pipetting steps are compared to the control (0 pipette actions). * p < 0.05, ** p < 0.01, *** p < 0.001, - no significance.
46 3. Manuscript I (Microcystin adsorption to lab-ware)
3.8 Acknowledgements
Thanks to Roel van Ginkel (Cawthron) for useful discussions about the mechanism of adsorption to the used surfaces. This work was made possible through the financial support of the Arthur-und-Aenne-Feindt Foundation (Hamburg, Germany), the Marie Curie International Research Staff Exchange Scheme Fellowship (PIRSES-GA-2011-295223), the Royal Society of New Zealand International Research Staff Exchange Scheme Fellowship (MEAT Agreement 295223), and the Marsden Fund of the Royal Society of New Zealand (12-UOW-087).
47 4. Manuscript II (Synthesis of MC-LF and derivates)
4. Manuscript II (Synthesis of MC-LF and derivates)
Total Synthesis of Microcystin-LF and Derivatives Thereof
Ivan Zemskov1, Stefan Altaner2, Daniel R. Dietrich2, and Valentin Wittmann1
1 Department of Chemistry and Konstanz Research School Chemical Biology (KoRS-CB), University of Konstanz, 78457 Konstanz (Germany) 2 Department of Biology and Graduate School Biological Sciences (GBS), University of Konstanz, 78457 Konstanz (Germany)
Published in: Journal of Organic Chemistry, 2017, 82 (7), doi:
10.1021/acs.joc.7b00175
4.1 Abstract
Figure 4-1: Graphical abstract to manuscript II. Microcystins (MC) are highly toxic natural products which are produced by cyanobacteria. They can be released to the water during harmful algal blooms and are a serious threat to animals and humans. Described is the total synthesis of the cyanotoxin microcystin-LF
(MC-LF, 1a) and two derivatives thereof. Deuterated derivative 1b is of interest as an internal
48 4. Manuscript II (Synthesis of MC-LF and derivates) standard during MC quantification in biological samples by mass spectrometry and alkyne- labelled 1c can be employed for toxin derivatization by click chemistry with an azide-containing reporter molecule or as an activity-based probe to identify interaction partners. Application of tert-butyl ester protecting groups for erythro--D-methylaspartic acid and -D-glutamic acid were key for an isomerization-free synthesis. The analytical data of synthetic MC-LF were identical to those of an authentic sample of the natural product. All derivatives 1a-c were determined to be potent inhibitors of protein phosphatase-1 with similar activity.
4.2 Introduction
Microcystins (MC) are highly toxic natural products which are produced by cyanobacteria and, according to the WHO, are amongst the most dangerous water pollutants
(Chorus and Bartram 1999). Widespread harmful algal blooms in densely populated areas, which are favoured by global warming along with eutrophication of surface waters (Michalak et al.
2013; Pelley 2016), resulted in a shutdown of the public water supply for millions of people, for example in Toledo, Ohio, USA, (August 2014) and around lake Taihu in China (July 2007)
(Michalak et al. 2013; Paerl et al. 2014; Pelley 2016; Qu et al. 2014). The sole analytical method for congener-specific MC detection enabling the identification and quantification of MC is LC-
MS/MS (Pelley 2016; Shoemaker et al. 2015). However, the lack of certified reference compounds limits the applicability of this method (Pelley 2016). Although research on MC has been conducted since 1878 (Francis 1878), their physiological function in cyanobacteria is still under scientific debate (Konst et al. 1965; Pearson et al. 2016; Rinehart et al. 1988).
Consequently, the total synthesis of MC would provide access to a yet missing variety of reference compounds as well as microcystin derivatives for structure-activity relationship studies enabling future studies into the biological roles of MC. Furthermore, the high cytotoxicity of
MC makes this scaffold an attractive lead structure for the development of novel potent anticancer drugs.
49 4. Manuscript II (Synthesis of MC-LF and derivates)
Figure 4-2: Microcystin-LF (MC-LF) and some amino acid variations (grey) of naturally occurring congeners. The two amino acids indicated in the name (LF in the current example) denote the natural L-amino acids in positions 2 and 4. Adda = (2S,3S,4E,6E,8S,9S)-3-amino-9- methoxy-2,6,8-trimethyl-10-phenyldeca-4,6-dienoic acid, D-Ala = D-alanine, -D-Glu = -D-glutamic acid, Mdha = N-methyldehydroalanine, -D-MeAsp = erythro--D- methylaspartic acid.
As shown in Figure 4-2, MC consist of three D-amino acids in positions 1, 3, and 6, the
-amino acid Adda5, and N-methyldehydroalanine (Mdha7) which is involved in covalent binding of MC to ser/thr-protein phosphatases by Michael addition of a proximate cysteine residue (Craig et al. 1996; Zemskov et al. 2016). Two L-amino acids in positions 2 and 4 are variable and account for the major differences amongst the more than 100 individual congeners that have been reported so far (Campos and Vasconcelos 2010; Puddick et al. 2014). Despite numerous attempts (Clave et al. 2010; Fontanillo et al. 2016; Mehrotra et al. 1997; Tappan and
Chamberlin 2008; Taylor et al. 1996a; Taylor et al. 1996b; Webster et al. 2001) the total synthesis of a natural microcystin (MC-LA) was accomplished only once (Humphrey et al.
1996). This landmark achievement featured protection of -D-MeAsp3 and -D-Glu6 by methyl esters that were cleaved under basic conditions in the last synthetic step. Unfortunately, this treatment led to the formation of several uncharacterized isomers of the final product (Humphrey et al. 1996). In our interpretation the isomerization can be attributed to a cyclization involving
-D-MeAsp3 (I, Figure 4-3) leading to aspartimide II which is prone to isomerization at the
50 4. Manuscript II (Synthesis of MC-LF and derivates) chiral centres in the - and -positions. Ring opening by basic hydrolysis can lead to the stereoisomers III of desired MC-LA as well as constitutional isomers IV. Despite extensive investigation of different methyl ester cleavage conditions, the isomerization could not be avoided and resulted in significant loss of material (Humphrey et al. 1996). To identify natural
MC-LA, the isomeric mixture was separated by HPLC and the isomers were compared to an authentic sample of the cyanotoxin (Humphrey et al. 1996). The identification of the desired isomers was particularly intricate during the synthesis of unnatural derivatives for which no standard was available (Aggen et al. 1999).
Figure 4-3: Proposed formation of aspartimide II and subsequent isomerization during the saponification of methyl ester protected MC-LA I Here, we present an isomerization-free synthetic strategy and its application to the first total synthesis of MC-LF (1a) and its deuterated derivative 1b (Figure 4-4) which is of interest as an internal standard during MC quantification by mass spectrometry. Since this approach delivers only one isomer, the synthesis of unnatural MC derivatives becomes feasible which we demonstrate by the preparation of derivative 1c. The alkyne-labelled 1c can be employed for toxin derivatization by click chemistry with an azide-containing reporter molecule, e.g. biotin or a fluorescent dye and, furthermore, be used as an activity-based probe to identify interaction partners (Böttcher et al. 2010; Shreder et al. 2004). According to crystal structures of MC in
51 4. Manuscript II (Synthesis of MC-LF and derivates) complex with ser/thr-protein phosphatases (Dawson and Holmes 1999; Goldberg et al. 1995) the side chain at position 4 of the MC backbone is not involved in binding and thus represents an optimal position for synthetic modifications.
4.3 Material and methods
4.3.1 General Experimental Methods
Technical solvents (petroleum ether and EtOAc) were distilled prior to use. THF and
CH2Cl2 were distilled from Na/K or CaH2 under inert atmosphere immediately prior to use.
Peptide grade DMF was used for peptide couplings. A sample of natural microcystin-LF was obtained from Enzo Life Sciences (ALX-350-081-C100, 100 g) and used as received. The compounds 3 (Pearson et al. 2000), 6 (Riggs Costerison 2002), 8a (Kokinaki et al. 2005), and
10 (Riggs Costerison 2002), were synthesized according to the literature. For the reactions performed under inert gas conditions (nitrogen) Schlenk technique and oven dried glassware were used. Analytical thin layer chromatography (TLC) was performed using TLC silica gel 60
F254 coated aluminium sheets (Merck). Spots were visualized either by UV light (= 254 nm) or by dipping and heating using ethanolic ninhydrin solution (3 % w/v), aqueous potassium permanganate (1 % w/v), anisaldehyde solution (135 mL EtOH, 5 mL conc. H2SO4, 15 mL glacial acetic acid, and 3.7 mL p-anisaldehyde), or Seebach’s stain (25 g phosphomolybdic acid,
10 g Ce(SO4)2 · 4H2O, and 60 mL conc. H2SO4 in 1 L H2O). Preparative flash column chromatography (FC) was performed using Geduran 60 silica gel (40-60 m, Merck). NMR spectra were recorded on Bruker Avance III 400 or Bruker Avance III 600 instruments. Chemical shifts are reported in ppm relative to solvent signals (CDCl3 H = 7.26 ppm, C = 77.2 ppm;
CD3OD H = 3.34 ppm; C = 49.0 ppm; DMSO-d6 H = 2.50 ppm, C = 39.5 ppm). For all new compounds two-dimensional NMR experiments (COSY, HSQC, and HMBC) were used for signal assignments. For numbering of carbons see supporting information.
52 4. Manuscript II (Synthesis of MC-LF and derivates)
Semi-preparative and analytical RP-HPLC was performed using a Shimadzu LC-20A prominence system (LC-20AT pumps, SIL-20A auto sampler, CTO-20AC column oven, SPD-
M2OA PDA detector, CBM-20A communication bus module and LC-Solution software). A
Kinetex 5u C18 100A, AXIA (250 x 21.2 mm, Phenomenex) column was used as a stationary phase at a flow rate of 10 mL min–1 unless mentioned otherwise. LC-MS measurements were performed on a Shimadzu LCMS-2020 system (LC-20 AD high pressure pumps, SIL-20AT HAT autosampler, CTO-20AC column oven, SPD-20A UV-Vis detector, CBM-20A communication bus module, LCMS-2020 ESI-detector and LC-MS-Solution software) using a Nucleodur 100-
3 C18ec (4 x 125 mm, Macherey-Nagel) reversed phase column as stationary phase at a flow rate of 0.4 mL min–1. A gradient of water with 0.1 % formic acid (mobile phase A) and acetonitrile with 0.1 % formic acid (mobile phase B) was used for the HPLC and LC-MS separations. Elemental analyses were performed by the microanalytical laboratory of the
University of Konstanz using an Elementar vario EL instrument. HRMS measurements were performed on a Bruker micrOTOF II (ESI-TOF) or a Thermo LTQ Orbitrap Discovery (ESI-
Orbitrap) instrument.
4.3.2 General Procedures
Phenacyl protection (Kokinaki et al. 2005). The amino acid derivative (1.0 equiv) is dissolved in EtOAc. Then Et3N (1.1 equiv) and phenacyl bromide (1.1 equiv) are added and the reaction mixture is stirred at rt for 4h. During the reaction, the formation of a white precipitate can be observed. After 4h the mixture is diluted with EtOAc and washed with brine (1x), sat. NaHCO3 solution (1x), and brine (2x). The organic phase is dried over Na2SO4, filtered, and concentrated under reduced pressure.
Phenacyl deprotection. The protected peptide is dissolved in MeOH/DMF (8:2, 8 L mg-1 peptide). Then Mg-turnings (0.16 mg/mg peptide) and acetic acid (0.8 L mg-1 peptide) are added and the reaction mixture is cooled to 0 °C. After 20 min cooling is removed and the
53 4. Manuscript II (Synthesis of MC-LF and derivates) reaction mixture is stirred at rt for 30 min. The reaction mixture is filtered, the volatiles are removed under reduced pressure, and the residue is taken up in CH2Cl2/MeOH (9:1). The mixture
3 is filtered through 3 cm silica column which is extensively washed with CH2Cl2/MeOH (9:1).
The solvents are removed under reduced pressure and the residue is lyophilized. The obtained deprotected peptide is used without further purification.
Fmoc deprotection. The peptide is dissolved in 5% piperidine in DMF (v/v). After 3 min, the solution is diluted with DMF and the volatiles are removed under reduced pressure. The residue is coevaporated with toluene (3x) and used without further purification.
Boc/t-Bu deprotection. The starting material is dissolved in neat TFA and stirred at rt for 30 min. Subsequently the TFA is removed and the residue is coevaporated with toluene, EtOAc and petroleum ether (each 1x) and used without further purification.
- Fmoc protection. The amino acid (1.0 equiv) is suspended in Na2CO3 (10% w/v, 1.25 mL mmol
1 amino acid). Then Fmoc-OSu (1.3 equiv) in dioxane (1.875 mL mmol-1 amino acid) is added, and the mixture is stirred at rt overnight. The reaction mixture is diluted with H2O and EtOAc, acidified with aqueous NaHSO4 (1M) to pH < 3, and extracted with EtOAc (3x). The combined organic phases are washed with brine (1x), dried over Na2SO4, filtered, and the volatiles removed under reduced pressure. The crude product is purified with FC.
Peptide coupling. The amine and carboxylic acid are dissolved in DMF. The resulting solution is cooled to 0 °C. Then base (DIPEA or collidine) and after 5 min the coupling reagent (HATU or HBTU) are added. The resulting yellowish solution is stirred at 0 °C for 45 min and at rt overnight. The reaction mixture is diluted with EtOAc and washed subsequently with H2O, 1:1 diluted saturated citric acid solution, 1:1 diluted saturated NaHCO3 solution, and brine (each 1x).
The organic phase is dried over Na2SO4, filtered, and purified with FC.
Macrocylization. The HPLC-purified C-terminally deprotected heptapeptide (1.0 equiv) is placed in a 4 mL glass vial. Then pentafluorophenol (1.95 equiv), dissolved in freshly distilled
EtOAc (24.2 mL mmol-1 peptide) and solid DCC are added. The reaction mixture is cooled to 0
54 4. Manuscript II (Synthesis of MC-LF and derivates)
°C, stirred for 90 min and then at rt for 8-10h. Afterwards, the solvent is removed and the activated peptide is treated with TFA according to GP4. The resulting C-terminally activated
-1 unprotected heptapeptide is dissolved in CHCl3 (515 mL mmol peptide). The resulting solution is added dropwise over a period of 10 min to a vigorously mechanically stirred mixture of CHCl3
(875 mL mmol-1 peptide) and pH = 9.5 phosphate buffer (1M, 875 mL mmol-1 peptide). After
30 min, the reaction mixture is diluted with a small amount of CHCl3 and H2O. The phases are separated, the aqueous phase is acidified with NaHSO4 (1M) to pH < 3 and extracted with EtOAc
(3x). The organic phases are combined, washed with brine, and dried over Na2SO4. The crude macrocyclic heptapeptide is lyophilized and purified using HPLC.
Selenoxide elimination. The macrocyclic peptide is dissolved in MeCN/H2O (3:2, 1 mL) and
30% aqueous H2O2 (4 L) is added. After 1h at rt, the reaction mixture is quenched with Me2S
(50 L) and purified by HPLC.
Synthesized Compounds (Sorted According to Compound Number)
Microcystin-LF (1a). The C-terminally unprotected heptapeptide 22a (15 mg, 10.9 mol) was macrocyclized using pentafluorophenol (4.0 mg, 21.3 mol) and DCC (2.8 mg, 13.3 mol) according to GP7. The crude macrocycle 23a was purified by semi-preparative RP-HPLC
(gradient: 60-90% B in 30 min, tR = 13.3 min). The cyclic peptide 23a was transformed to 1a according to GP8 and the reaction mixture was separated by semi-preparative RP-HPLC
(gradient: 50-70% B in 20 min, tR = 13.2 min). MC-LF 1a was obtained as a white amorphous solid (2.0 mg, 19%). LC-MS analysis of synthetic 1a coinjected with natural MC-LF resulted in
1 a single peak with the expected mass (Figure S3). H NMR (600 MHz, CD3OD, 300 K) 8.89
(d, J = 9.6 Hz, 1H, NH MeAsp), 8.17 – 8.08 (m, 2H, NH Leu, NH Phe), 7.46 (d, J = 8.2 Hz, 1H,
NH Ala), 7.31 (d, J = 9.2 Hz, 1H, NH Adda), 7.27 – 7.25 (m, 2H, Ar), 7.23 – 7.14 (m, 8H, Ar),
6.35 (d, J = 15.5 Hz, 1H, H-5 Adda), 5.89 (s, 1H, C=CH2), 5.51 (d, J = 9.8 Hz, 1H, H-7 Adda),
5.45 (s, 1H, C=CH2), 5.43 (dd, J = 15.5, 8.6 Hz 1H, H-4 Adda), 4.74 – 4.69 (m, H-3 Adda), 4.58
– 4.49 (m, 3H, H- MeAsp, H- Phe, H- Ala), 4.45 (t, J = 7.4 Hz, 1H, H- Glu), 4.22 – 4.18
55 4. Manuscript II (Synthesis of MC-LF and derivates)
(m, 1H, H- Leu), 3.50 (dd, J = 14.1, 3.2 Hz, 1H, H- Phe), 3.38 (s, 3H, -NCH3), 3.30 – 3.27
(m, 1H, H-9 Adda), 3.26 (s, 3H, -OCH3), 2.93 (dq, J = 7.2, 3.3 Hz, 1H, H- MeAsp), 2.84 (dd,
J = 14.0, 4.8 Hz, 1H, H-10 Adda), 2.74 – 2.68 (m, 2H, H-10 Adda, H-2 Adda), 2.66 – 2.55 (m,
4H, H-8 Adda, 2x H- Glu, H- Phe), 2.19 – 2.11 (m, 1H, H- Glu), 1.91 – 1.83 (m, 1H, H-
Leu), 1.80 – 1.68 (m, 2H, H- Glu, H- Leu), 1.65 (s, 3H, 3xH-6’ Adda), 1.53 (ddd, J = 13.9,
9.8, 4.2 Hz, 1H, H- Leu), 1.11 (d, J = 6.9 Hz, 3H, 3xH-2’ Adda), 1.04 (d, J = 6.7 Hz, 3H, 3xH-
8’ Adda), 0.99 (d, J = 7.4 Hz, 3H, CH3 Ala), 0.88 (d, J = 6.6 Hz, 3H, 3xH- Leu), 0.86 (d, J =
13 6.5 Hz, 3H, 3xH- Leu), 0.76 (d, J = 7.2 Hz, 3H, CH3 MeAsp); C NMR (151 MHz, CD3OD,
300 K) 178.4 (C=O MeAsp), 176.5 (C=O), 176.4 (C=O), 176.0 (C=O), 175.7 (C=O), 175.2
(C=O), 175.2 (C=O), 171.5 (C=O Phe), 166.0 (C=O Mdha), 146.2 (-C=CH2), 140.6 (C Ar),
139.4 (C-5 Adda), 139.2 (C Ar) , 137.6 (C-7 Adda), 133.7 (C-6 Adda), 130.5 (C Ar), 130.1 (2xC
Ar), 129.5 (2xC Ar), 129.2 (C Ar), 127.9 (C Ar), 127.1 (C Ar), 125.9 (C-4 Adda), 114.4 (-
C=CH2), 88.4 (C-9 Adda), 58.8 (-OCH3), 56.1 (C-7 Adda), 55.3 (C- MeAsp), 55.3 (C- Phe),
55.2 (C- Leu), 53.3 (C- Glu), 49.6 (C- Ala), 46.0 (C-2 Adda), 40.8 (C- Leu), 40.6 (C-
MeAsp), 39.0 (C-10 Adda), 38.5 (-NCH3), 38.1 (C- Phe), 37.8 (C-8 Adda), 33.4 (C- Glu),
29.5 (C- Glu), 25.7 (C- Leu), 23.5 (C- Leu), 21.2 (C- Leu), 17.4 (CH3 Ala), 16.5 (C-8’
Adda), 16.4 (C-2’ Adda), 15.0 (CH3 MeAsp), 12.9 (C-6’ Adda). HRMS (ESI-Orbitrap) m/z:
+ [M+H] Calcd for C52H72N7O12 986.52335; Found 986. 52418.
4 [Phe-d5 ]-Microcystin-LF (1b). The macrocyclic derivative 23b (5 mg, 4.35 mol) was transformed to 1a according to GP8, and the reaction mixture was separated by semi-preparative
RP-HPLC (gradient: 50-70% B in 20 min, tR = 13.6 min). The microcystin 1b was obtained as a white amorphous solid (3.3 mg, 76%). LC-MS analysis of synthetic 1b coinjected with MC-LF resulted in a single peak with the expected masses of both deuterated 1b and undeuterated MC-
1 LF (Figure S4). H NMR (600 MHz, CD3OD, 284 K) 8.16 (d, J = 6.8 Hz, 1H, NH Leu), 8.07 (d, J = 9.1 Hz, 1H, NH MeAsp), 7.56 (d, J = 8.4 Hz, 1H, NH Ala), 7.27 – 7.25 (m, 2H, Ar), 7.23
– 7.15 (m, 3H, Ar), 6.35 (d, J = 15.5 Hz, 1H, H-5 Adda), 5.89 (s, 1H, -C=CH2), 5.51 (d, J = 9.9
56 4. Manuscript II (Synthesis of MC-LF and derivates)
Hz, 1H, H-7 Adda), 5.46 (s, 1H, -C=CH2), 5.45 (dd, J = 15.6, 8.6 Hz, 1H, H-4 Adda), 4.69 (dd,
J = 11.0, 8.7 Hz, 1H, H-3 Adda), 4.58 – 4.49 (m, 3H, H- MeAsp, H- Phe-d5, H- Ala), 4.40
(t, J = 7.5 Hz, 1H, H- Glu), 4.23 – 4.13 (m, 1H, H- Leu), 3.49 (dd, J = 14.1, 3.4 Hz, 1H, H-
Phe-d5), 3.37 (s, 3H, -NCH3), 3.30 – 3.28 (m, 1H, H-9 Adda), 3.26 (s, 3H, -OCH3), 2.95 (dq,
J = 7.2, 3.5 Hz, 1H, H- MeAsp), 2.85 (dd, J = 14.0, 4.7 Hz, 1H, H-10 Adda), 2.80 (dd, J =
10.9, 6.9 Hz, 1H, H-2 Adda), 2.69 (dd, J = 13.9, 7.3 Hz, 1H, H-10 Adda), 2.65 – 2.52 (m, 4H,
H-8 Adda, H- Phe-d5, 2xH- Glu), 2.17 – 2.07 (m, 1H, H- Glu), 1.90 (ddd, J = 15.5, 12.8,
4.1 Hz, 1H, H- Leu), 1.80 – 1.69 (m, 2H, H- Glu, H- Leu), 1.65 (s, 3H, 3xH-6’ Adda), 1.52
(ddd, J = 13.8, 10.2, 3.9 Hz, 1H, H- Leu), 1.10 (d, J = 6.9 Hz, 3H, 3xH-2’ Adda), 1.04 (d, J =
6.7 Hz, 3H, 3xH-8’ Adda), 0.97 (d, J = 7.4 Hz, 3H, -CH3 Ala), 0.88 (d, J = 6.6 Hz, 3H, 3xH-
13 Leu), 0.85 (d, J = 6.6 Hz, 3H, 3xH- Leu), 0.75 (d, J = 7.2 Hz, 3H, CH3 MeAsp); C NMR (151
MHz, CD3OD, 284 K): = 178.5 (C=O MeAsp), 176.6 (C=O), 176.5 (C=O), 176.2 (C=O), 175.6
(C=O), 175.3 (C=O), 175.2 (C=O), 171.6 (C=O Phe), 166.0 (C=O Mdha), 146.2 (-C=CH2),
140.5 (C Ar), 139.3 (C-5 Adda), 139.0 (C Ar), 137.5 (C-7 Adda), 133.7 (C-6 Adda), 130.6 (C
Ar), 129.2 (C Ar), 127.1 (C Ar), 126.0 (C-4 Adda), 114.5 (-C=CH2), 88.3 (C-9 Adda), 58.7 (-
OCH3) , 56.2 (C-7 Adda), 55.6 (C- MeAsp), 55.2 (C- Phe-d5), 55.1 (C- Leu), 53.6 (C-
Glu), 49.6 (C- Ala), 45.7 (C-2 Adda), 40.8 (C- Leu), 38.9 (C-10 Adda), 38.5 (-NCH3), 38.0
(C- Phe-d5), 37.7 (C-8 Adda), 33.4 (C- Glu), 29.2 (C- Glu), 25.7 (C- Leu), 23.6 (C- Leu),
21.2 (C- Leu), 17.3 (-CH3 Ala), 16.6 (C-8’ Adda), 16.3 (C-2’ Adda), 15.0 (CH3 MeAsp), 12.9
+ (C-6’ Adda); HRMS (ESI-TOF) m/z: [M+H] Calcd for C52H67D5N7O12 991.5547; Found
991.5553.
Microcystin-LY(Prg) (1c). The macrocyclic derivative 23c (6.2 mg, 5.18 mol) was transformed to 1c according to GP8 and the reaction mixture was separated by semi-preparative
RP-HPLC (gradient: 50-70% B in 20 min, tR = 14.1 min). The microcystin 1c was obtained as a
1 white amorphous solid (2.3 mg, 43 %). H NMR (600 MHz, CD3OD, 300 K) 8.89 (d, J = 9.6
Hz, 1H, NH Tyr(Prg)), 8.21 (d, J = 8.7 Hz, 1H, NH MeAsp), 8.11 (d, J = 6.7 Hz, 1H, NH Leu),
57 4. Manuscript II (Synthesis of MC-LF and derivates)
7.43 (d, J = 8.0 Hz, 1H, NH Ala), 7.28 – 7.26 (m, 2H, Ar), 7.22 – 7.16 (m, 3H, Ar), 7.09 (d, J =
8.6 Hz, 2H, Ar), 6.84 (d, J = 8.7 Hz, 2H, Ar), 6.35 (d, J = 15.5 Hz, 1H, H-5 Adda), 5.90 (s, 1H,
-C=CH2), 5.51 (d, J = 9.8 Hz, 1H, H-7 Adda), 5.46 (s, 1H, -C=CH2), 5.42 (dd, J = 15.5, 8.7 Hz,
1H, H-4 Adda), 4.71 (dd, J = 11.0, 8.6 Hz, 1H, H-3 Adda), 4.65 (d, J = 2.3 Hz, 2H, -OCH2-
Tyr(Prg)), 4.57 – 4.50 (m, 3H, H- MeAsp, H- Tyr(Prg), H- Ala), 4.46 (dd, J = 8.8, 6.1 Hz,
1H, H- Glu), 4.22 – 4.15 (m, 1H, H- Leu), 3.44 (dd, J = 14.1, 3.1 Hz, 1H, H- Tyr(Prg)),
3.37 (s, 3H, -NCH3), 3.30 – 3.26 (m, 1H, H-9 Adda), 3.26 (s, 3H, -OCH3), 2.96 – 2.91 (m, 1H,
H- MeAsp), 2.91 (t, J = 2.4 Hz, 1H, -C≡CH), 2.84 (dd, J = 14.0, 4.7 Hz, 1H, H-10 Adda), 2.73 – 2.66 (m, 2H, H-2 Adda, H-10 Adda), 2.63 (dq, J = 9.9, 6.6 Hz, 1H, H-8 Adda), 2.61 – 2.56 (m,
1H, 2xH- Glu), 2.51 (dd, J = 14.1, 11.9 Hz, 1H, H- Tyr(Prg)), 2.19 – 2.11 (m, 1H, H- Glu),
1.89 – 1.83 (m, 1H, H- Leu), 1.79 – 1.73 (m, 1H, H- Leu), 1.73 – 1.65 (m, 1H, H- Glu),
1.65 (d, J = 1 Hz, 3H, 3xH-6’ Adda), 1.52 (ddd, J = 13.7, 9.8, 4.1 Hz, 1H, H- Leu), 1.11 (d, J
= 6.9 Hz, 3xH-2’ Adda), 1.04 (d, J = 6.7 Hz, 3xH-8’ Adda), 1.01 (d, J = 7.4 Hz, 3H, CH3 Ala),
0.89 (d, J = 6.6 Hz, 3H, 3xH- Leu), 0.86 (d, J = 6.6 Hz, 3H, 3xH- Leu), 0.81 (d, J = 7.2 Hz,
13 3H, CH3 MeAsp); C NMR (151 MHz, CD3OD, 300 K) 178.4 (C=O MeAsp), 176.4 (C=O),
175.9 (C=O), 175.8 (C=O), 175.2 (C=O), 175.0 (C=O), 171.5 (C=O), 166.0 (C=O), 165.9
(C=O), 158.0 (-C-O-CH2-), 146.2 (-C=CH2), 140.5 (C Ar Adda), 139.4 (C-5 Adda), 137.7 (C-7
Adda), 133.7 (C-6 Adda), 131.8 (C Ar Tyr(Prg)), 131.1 (2xC Ar Tyr(Prg)), 130.5 (2xC Ar Adda),
129.2 (2xC Ar Adda), 127.1 (CH Ar Adda), 125.8 (C-4 Adda), 116.0 (2xC Ar Tyr(Prg)), 114.4
(-C=CH2), 88.4 (C-9 Adda), 79.8 (-C≡CH), 76.8 (-C≡CH), 58.7 (-OCH3), 56.5 (-OCH2-), 56.0
(C-3 Adda), 55.4, 55.3, 55.2 (m, 3C, C- Leu, C- MeAsp, C- Tyr(Prg)), 53.2 (C- Glu), 49.6
(C- Ala), 46.0 (C-2 Adda), 40.8 (C- Leu), 40.5 (C- MeAsp), 39.0 (C-10 Adda), 38.5 (-
NCH3), 37.7 (C-8 Adda), 37.2 (C- Tyr(Prg)), 33.4 (C- Glu), 29.6 (C- Glu), 25.7 (C- Leu),
23.5 (CH3 Leu), 21.2 (CH3 Leu), 17.4 (CH3 Ala), 16.5 (C-8’ Adda), 16.4 (C-2’ Adda), 15.2 (CH3
+ MeAsp), 12.9 (C-6’ Adda); HRMS (ESI-TOF) m/z: [M+H] Calcd for C55H74N7O13 1040.5339;
Found 1040.5353.
58 4. Manuscript II (Synthesis of MC-LF and derivates)
Fmoc--D-Glu(Ot-Bu)-N-MeSecPh-D-Ala-Leu-OPac (4). Tripeptide 12 (297 mg, 449.5 mol) was deprotected according to GP4. The N-terminally deprotected tripeptide (91 mg, 107.5 mol) and Fmoc-D-Glu-Ot-Bu (249 mg, 584.4 mol) were dissolved in DMF (2 mL). Peptide coupling was performed according to GP6 using HATU (222 mg, 584.4 mol) and DIPEA (313 L, 1.8 mmol). The crude product was purified by FC to give tetrapeptide 4 (390 mg, 90%) as a white
1 amorphous solid: Rf = 0.77 (EtOAc); H NMR (400 MHz, CDCl3, 300 K) 7.80 – 7.70 (m, 4H,
4x H-Ar), 7.68 – 7.54 (m, 3H, 3x H-Ar), 7.53 – 7.47 (m, 1H, 2x H-Ar), 7.45 – 7.35 (m, 4H, 4x
H-Ar), 7.33 – 7.25 (m, 2H, 2x H-Ar), 7.24 – 7.19 (m, 4H, 3x H-Ar, NH Ala), 6.84 (d, J = 8.3
Hz, 1H, NH Leu), 5.57 (d, J = 8.0 Hz, 1H, NH Glu), 5.35 (dd, J = 10.2, 5.3 Hz, 1H, H-
NMeSecPh), 5.19 (d, J = 16.5 Hz, 1H, -C(O)CH2-O), 4.94 (d, J = 16.5 Hz, 1H, -C(O)CH2-O),
4.65 – 4.53 (m, 2H, H- Leu, H- Ala), 4.44 – 4.38 (m, 1H, -OCH2-CH- Fmoc), 4.35 – 4.25 (m,
2H, -OCH2-CH- Fmoc, H- Glu), 4.17 (t, J = 6.9 Hz, 1H, -OCH2-CH- Fmoc), 3.63 (dd, J = 13.0,
5.2 Hz, 1H, H- NMeSecPh), 3.19 (dd, J = 12.9, 10.5 Hz, 1H, H- NMeSecPh), 2.76 (s, 3H, -
NCH3), 2.47 – 2.32 (m, 2H, 2x H- Glu), 2.18 (dd, J = 15.6, 4.4 Hz, 1H, H- Glu), 1.82 – 1.73
(m, J = 3.1 Hz, 2H, H- Leu, H- Leu), 1.70 – 1.60 (m, 2H, H- Leu, H- Glu), 1.48 (s, 9H, t Bu), 1.36 (d, J = 7.1 Hz, 3H, CH3 Ala), 0.97 (d, J = 6.3 Hz, 3H, 3xH- Leu), 0.95 (d, J = 6.3
13 Hz, 3H, 3xH- Leu); C NMR (101 MHz, CDCl3, 300 K) 173.4 (C=O), 172.6 (C=O), 172.1
(C=O), 171.7 (C=O), 171.3 (C=O), 170.5 (C=O), 156.5 (C=O), 144.3 (C Ar), 143.8 (C Ar),
141.5 (C Ar), 141.4 (C Ar), 134.1 (C Ar), 134.1 (C Ar), 133.2 (C Ar), 129.8 (C Ar), 129.3 (2C
Ar), 129.0 (C Ar), 127.8 (C Ar), 127.8 (C Ar), 127.4 (C Ar), 127.3 (C Ar), 127.3 (C Ar), 125.5
(C Ar) 125.3 (C Ar), 120.0 (2C Ar), 82.7 (-C(CH3)3), 67.0 (-OCH2-CH- Fmoc), 66.3 (CH2 Pac),
58.0 (C- NMeSecPh), 53.4 (-OCH2-CH- Fmoc), 50.7 (C-Leu), 48.6 (C-Ala), 47.2 (C-
Glu), 41.2 (C-Leu), 32.2 (-NCH3), 28.84 (C-Glu), 28.75 (C-Glu), 28.2 (3C, -C(CH3)3),
26.4 (C- NMeSecPh), 25.0 (C-Leu), 23.1 (C- Leu), 21.8 (C- Leu) 16.1 (CH3 Ala); HRMS
+ (ESI-TOF) m/z: [M+H] Calcd for C51H61N4O10Se 969.35474; Found 969.35351.
59 4. Manuscript II (Synthesis of MC-LF and derivates)
Fmoc--D-MeAsp(Ot-Bu)-Phe-OH (5a). Boc-Phe-OPac (Kokinaki et al. 2005) 8a (120 mg,
313 mol) was deprotected according to GP4. The obtained H-Phe-OPac and Fmoc-D-MeAsp-
Ot-Bu 7 (140 mg, 329 mol) were dissolved in DMF (3 mL). Peptide coupling was performed according to GP6 using HATU (125 mg, 329 mol) and DIPEA (165 L, 329 mol). The crude product was purified by FC (petroleum ether/EtOAc 2:1) to give dipeptide 5a (153 mg, 71%) as
1 a white amorphous solid: Rf = 0.77 (petroleum ether/EtOAc 1:1); H NMR (400 MHz, CDCl3,
300 K) 7.95 – 7.87 (m, 2H, Ar), 7.76 (d, J = 7.6 Hz, 2H, Ar), 7.65 – 7.60 (m, 2H, Ar), 7.52 –
7.48 (m, 2H, Ar), 7.40 – 7.37 (m, 2H, Ar), 7.34 – 7.20 (m, 8H, Ar), 6.13 – 6.05 (m, 1H, NH
MeAsp, NH Phe), 5.50 (d, J = 16.3 Hz, 1H, CH2 Pac), 5.34 (d, J = 16.3 Hz, 1H, CH2 Pac), 5.01
(ddd, J = 13.0, 7.4, 5.5 Hz, 1H, H- Phe), 4.41 (dd, J = 10.2, 7.2 Hz, 1H, -OCH2-CH- Fmoc),
4.36 – 4.28 (m, 2H, -OCH2-CH- Fmoc, H- MeAsp), 4.24 (t, J = 7.2 Hz, 1H, -OCH2-CH- Fmoc),
3.42 (dd, J = 14.2, 5.6 Hz, 1H, H-Phe), 3.19 (dd, J = 14.2, 7.1 Hz, 1H, H- Phe), 3.04 (qd, J
13 = 7.3, 3.8 Hz, 1H, H- MeAsp), 1.41 (s, 9H, t-Bu), 1.15 (d, J = 7.3 Hz, 3H, CH3 MeAsp); C
NMR (101 MHz, CDCl3, 300 K) 191.4 (C=O), 173.6 (C=O), 171.2 (C=O), 170.1 (C=O), 157.1
(C=O), 144.2 (C Ar), 144.0 (C Ar), 141.4 (C Ar), 135.9 (C Ar), 134.3 (C Ar), 134.1 (C Ar),
129.6 (2C Ar), 129.1 (2C, Ar), 128.8 (2C Ar), 127.9 (2C Ar), 127.8 (2C Ar), 127.4 (C Ar), 127.2
(2C Ar), 127.2 (C Ar), 125.5 (C Ar), 125.4 (C Ar), 120.1 (C Ar) 120.0 (C Ar), 82.5 (-C(CH3)3),
67.3 (-OCH2-CH- Fmoc), 66.7 (CH2 Pac), 57.1 (-OCH2-CH- Fmoc), 53.0 (C- Phe), 47.3 (C-
MeAsp), 41.6 (C- MeAsp), 37.8 (C- Phe), 28.0 (3C, -C(CH3)3), 15.2 (CH3 MeAsp); HRMS
+ (ESI-TOF) m/z: [M+H] Calcd for C41H43N2O8 691.30139; Found 691.30123.
Fmoc--D-MeAsp(Ot-Bu)-Phe-d5-OH (5b). Boc-Phe-d5-OPac 8b (43 mg, 109.7 mol) was deprotected according to GP4. The obtained H-Phe-d5-OPac and Fmoc-D-MeAsp-Ot-Bu 7 (49 mg, 115.0 mol) were dissolved in DMF (1 mL). Peptide coupling was performed according to
GP6 using HATU (44 mg, 115.0 mol) and DIPEA (57.2 L, 329.0 mol). The crude product was purified by FC (petroleum ether/EtOAc 2:1) to give dipeptide 5b (66 mg, 86%) as a white
1 amorphous solid: Rf = 0.77 (petroleum ether/EtOAc 1:1); H NMR (400 MHz, CDCl3, 300 K)
60 4. Manuscript II (Synthesis of MC-LF and derivates)
7.93 – 7.87 (m, 2H, Ar), 7.76 (d, J = 7.6 Hz, 2H, Ar), 7.66 – 7.59 (m, 3H, Ar), 7.52 – 7.48 (m,
2H, Ar), 7.40 – 7.48 (m, 2H, Ar), 7.32 – 7.27 (m, 2H, Ar), 6.12 – 6.03 (m, 2H, NH MeAsp, NH
Phe-d5), 5.50 (d, J = 16.3 Hz, 1H, CH2 Pac), 5.34 (d, J = 16.3 Hz, 1H, CH2 Pac), 5.01 (ddd, J =
13.0, 7.2, 5.5 Hz, 1H, H- Phe), 4.41 (dd, J = 10.2, 7.2 Hz, 1H, -OCH2-CH- Fmoc), 4.36 – 4.28
(m, 2H, -OCH2-CH- Fmoc, H- MeAsp), 4.24 (t, J = 7.3 Hz, 1H, -OCH2-CH- Fmoc), 3.42 (dd,
J = 14.2, 5.6 Hz, 1H, H- Phe), 3.19 (dd, J = 14.2, 7.1 Hz, 1H, H- Phe), 3.04 (qd, J = 7.1, 3.9
13 Hz, 1H, H- MeAsp), 1.41 (s, 9H, t-Bu), 1.15 (d, J = 7.2 Hz, 3H, CH3 MeAsp); C NMR (101
MHz, CDCl3, 300 K) 191.4 (C=O), 173.6 (C=O), 171.2 (C=O), 170.1 (C=O), 157.1 (C=O), 141.4 (C Ar), 134.3 (C Ar), 134.1 (C Ar), 129.1 (2C Ar), 127.9 (2C Ar), 127.8 (2C Ar), 127.2
(2C Ar), 125.5 (C Ar), 125.4 (C Ar), 120.1 (2C Ar), 82.5 (-C(CH3)3), 67.4 (-OCH2-CH- Fmoc),
66.7 (CH2 Pac), 56.1 (-OCH2-CH- Fmoc), 53.0 (C- Phe-d5), 47.4 (C- MeAsp), 41.6 (C-
MeAsp), 37.7 (C- Phe-d5), 28.0 (3C, -C(CH3)3), 15.2 (CH3 MeAsp); HRMS (ESI-TOF) m/z:
+ [M+H] Calcd for C41H38D5N2O8 696.3328; Found 696.3313.
Fmoc--D-MeAsp(Ot-Bu)-Tyr(Prg)-OH (5c). Boc-Tyr(Prg)-OPac 8c (137 mg, 313 mol) was deprotected according to GP4. The obtained H-Tyr(Prg)-OPac and Fmoc-D-MeAsp-Ot-Bu 7
(140 mg, 329 mol) were dissolved in DMF (3 mL). Peptide coupling was performed according to GP6 using HATU (125 mg, 329 mol) and DIPEA (165 L, 329 mol). The crude product was purified by FC to give dipeptide 5c (181 mg, 84%) as a white amorphous solid: Rf = 0.3
1 (petroleum ether/EtOAc 2:1); H NMR (400 MHz, CDCl3, 300 K) 7.94 – 7.86 (m, 2H, Ar),
7.75 (d, J = 7.5 Hz, 2H, Ar), 7.67 – 7.58 (m, 3H, Ar), 7.52 – 7.48 (m, 2H, Ar), 7.40 – 7.36 (m,
2H, Ar), 7.31 – 7.27 (m, 2H, Ar), 7.17 (d, J = 8.6 Hz, 2H, Ar), 6.94 – 6.90 (m, 2H, Ar), 6.11 –
6.03 (m, 2H, NH MeAsp, NH Phe), 5.50 (d, J = 16.3 Hz, 1H, CH2 Pac), 5.33 (d, J = 16.3 Hz,
1H, CH2 Pac), 4.97 (ddd, J = 13.3, 6.8, 6.0 Hz, 1H, H- Tyr(Prg)), 4.65 (d, J = 2.3 Hz, 2H, -
OCH2- Tyr(Prg)), 4.42 (dd, J = 10.1, 7.2 Hz, 1H, -OCH2-CH- Fmoc), 4.37 – 4.30 (m, 2H, H-
MeAsp, -OCH2-CH- Fmoc), 4.29 – 4.21 (m, 1H, -OCH2-CH- Fmoc), 3.36 (dd, J = 14.3, 5.8 Hz,
1H, H- Tyr(Prg)), 3.16 (dd, J = 14.3, 6.8 Hz, 1H, H- Tyr(Prg)), 3.05 (qd, J = 7.3, 3.5 Hz, 1H,
61 4. Manuscript II (Synthesis of MC-LF and derivates)
H- MeAsp), 2.49 (t, J = 2.2 Hz, 1H, -C≡CH), 1.41 (s, 9H, t-Bu), 1.17 (d, J = 7.2 Hz, 3H, CH3
13 MeAsp); C NMR (101 MHz, CDCl3, 300 K) 191.4 (C=O), 173.6 (C=O), 171.2 (C=O), 170.1
(C=O), 157.1 (C=O) 156.9 (C Ar), 141.4 (C Ar) , 134.3 (C Ar), 134.1 (C Ar), 130.7 (2C, Ar),
129.1 (2C, Ar), 128.8 (C Ar), 127.9 (2C, Ar), 127.8 (2C, Ar), 127.2 (C Ar), 127.2 (C Ar), 125.5
(C Ar), 125.4 (C Ar), 120.1 (C Ar), 115.2 (2C, Ar), 82.5 (-C(CH3)3), 78.7 (-C≡CH), 75.7 (-
C≡CH), 67.4 (-OCH2-CH- Fmoc), 66.7 (CH2 Pac), 57.1 (C- MeAsp), 56.0 (-OCH2- Tyr(Prg)),
53.1 (C- Tyr(Prg)), 47.3 (m, 1H, -OCH2-CH- Fmoc), 41.6 (C- MeAsp), 36.9 (C- Tyr(Prg)),
+ 28.0 (3C, -C(CH3)3), 15.3 (CH3 MeAsp); HRMS (ESI-TOF) m/z: [M+H] Calcd for C44H45N2O9
745.31196; Found 745.31092.
Fmoc-D-MeAsp-Ot-Bu (7). The amino acid erythro-16 (1.00 g, 1.6 mmol) was suspended in methanol (13 mL) and 10% Pd/C catalyst (410 mg, wet, 53.7% water) was added. The reaction mixture was hydrogenated overnight at slightly positive hydrogen pressure and filtered through a Celite pad. The residue was further treated according to GP5 using Fmoc-OSu (703 mg, 2.084 mmol), dioxane (3 mL) and 10% w/v aqueous Na2CO3 (2 mL). The crude product was purified with FC (petroleum ether/EtOAc/AcOH 74:25:1) to give 7 (572 mg, 84%) as a white amorphous
27 solid: Rf = 0.5 (petroleum ether/EtOAc/AcOH 49:50:1); mp 55-57 °C; []D +11.8 (c 1.0,
1 MeCN); H NMR (400 MHz, CDCl3, 300 K) 7.76 (d, J = 7.5 Hz, 2H, Ar), 7.62 (d, J = 7.4 Hz,
2H, Ar), 7.42 – 7.38 (m, 2H, Ar), 7.34 – 7.29 (m, 2H, Ar), 5.75 (d, J = 8.9 Hz, 1H, NH), 4.59
(dd, J = 8.9, 3.6 Hz, 1H, H-), 4.49 – 4.34 (m, 2H, CH2 Fmoc), 4.25 (t, J = 7.1 Hz, 1H, CH
t 13 Fmoc), 3.30 (qd, J = 7.3, 3.6 Hz, 1H, H-) 1.46 (s, 9H, Bu), 1.27 (d, J = 7.3 Hz, 3H, CH3); C
NMR (101 MHz, CDCl3, 300 K) 178.7 (C=O), 169.7 (C=O), 156.7 (C=O Fmoc), 144.1 (C
Ar), 143.9 (C Ar), 141.5 (2C Ar), 127.9 (2C Ar), 127.2 (2C Ar), 125.3 (2C Ar), 120.1 (C Ar),
120.1 (C Ar), 83.2 (-C(CH3)3), 67.4 (CH2 Fmoc), 56.2 (C-), 47.3 (CH Fmoc), 41.5 (C-), 28.0
+ (3C, -C(CH3)3), 13.0 (CH3); HRMS (ESI-TOF) m/z: [2M+H] Calcd for C48H55N2O12 851.3750;
Found 851.3708; Anal. Calcd for C24H27NO6: C, 67.75; H, 6.40; N, 3.29. Found: C 67.64; H,
6.50; N, 3.38.
62 4. Manuscript II (Synthesis of MC-LF and derivates)
Boc-Phe-d5-OPac (8b) Boc-Phe-d5-OH (390 mg, 1.44 mmol) was dissolved in EtOAc (6 mL) and reacted according to GP1. The product 8b was obtained as a white amorphous solid (260
1 mg, 46%): Rf = 0.55 (petroleum ether/EtOAc 7:3); H NMR (400 MHz, CDCl3, 300 K) 7.91
(d, J = 7.7 Hz, 2H, Ar), 7.62 (t, J = 7.4 Hz, 1H, Ar), 7.50 (t, J = 7.7 Hz, 2H, Ar), 5.49 (d, J =
16.3 Hz, 1H, CH2 Pac), 5.31 (d, J = 16.4 Hz, 1H, CH2 Pac), 4.97 (d, J = 7.7 Hz, 1H, NH), 4.77
– 4.72 (m, 1H, H-), 3.35 (dd, J = 14.1, 5.3 Hz, 1H, H-), 3.15 (dd, J = 14.0, 7.0 Hz, 1H, H-),
t 13 1.40 (s, 9H, Bu); C NMR (101 MHz, CDCl3, 300 K): = 191.7 (C=O), 171.8 (C=O), 155.3
(C Ar), 134.2 (C Ar), 129.1 (2C, Ar), 127.9 (2C, Ar), 80.1 (-C(CH3)3), 66.6 (CH2 Pac), 54.5 (C-
+ ), 38.2 (C-), 28.4 (3C, -C(CH3)3); HRMS (ESI-TOF) m/z: [2M+H] Calcd for
C44H41D10N2O10 777.4166; Found 777.4151.
Boc-Tyr(Prg)-OPac (8c). The carboxylic acid 18 (350 mg, 1.10 mmol) was dissolved in EtOAc
(5.5 mL) and reacted according to GP1. The product 8c was obtained as a white amorphous solid
25 (421 mg, 90%): Rf = 0.55 (petroleum ether/EtOAc 7:3); mp 94-95 °C; []D -21.2
1 (c 0.5, MeCN); H NMR (400 MHz, CDCl3, 300 K) 7.95 – 7.88 (m, 2H, Ar), 7.65 – 7.60 (m,
1H, Ar), 7.50 (t, J = 7.7 Hz, 2H, Ar), 7.19 (d, J = 8.5 Hz, 2H, Ar), 6.94 – 6.90 (m, 2H Ar), 5.50
(d, J = 16.4 Hz, 1H, CH2 Pac), 5.30 (d, J = 16.3 Hz, 1H, CH2 Pac), 4.95 (d, J = 7.9 Hz, 1H, NH),
4.74 – 4.69 (m, 1H, H-), 4.67 (d, J = 2.4 Hz, 2H, -OCH2- Tyr(Prg)), 3.29 (dd, J = 14.2, 5.4 Hz,
1H, H-), 3.11 (dd, J = 14.0, 6.7 Hz, 1H, H-), 2.51 (t, J = 2.4 Hz, 1H, -C≡CH), 1.41 (s, 9H, t 13 Bu); C NMR (101 MHz, CDCl3, 300 K) 191.7 (C=O), 171.8 (C=O), 156.8 (-C-O-CH2-),
155.3 (C=O), 134.2 (C Ar), 130.7 (2C, Ar), 129.2 (C Ar), 129.1 (2C, Ar), 127.9 (2C, Ar), 115.1
(C Ar), 80.1 (-C(CH3)3), 78.8 (-C≡CH), 75.6 (-C≡CH), 66.5 (CH2 Pac), 56.0 (-OCH2- Tyr(Prg)),
+ 54.5 (C-), 37.5 (C-), 28.4 (3C, -C(CH3)3); HRMS (ESI-TOF) m/z: [M+H] Calcd for
C25H28NO6: 438.19111; Found 438.19216; Anal. Calcd for C25H27NO6: C, 68.64; H, 6.22; N,
3.20. Found: C, 68.56; H, 6.11; N, 3.32.
Boc-N-MeSecPh-D-Ala-Leu-OPac (12) (Riggs Costerison 2002). Boc-Leu-OPac (Kokinaki et al. 2005) (500 mg, 1.43 mmol) was deprotected according to GP4. Then H-Leu-OPac, Boc-D-
63 4. Manuscript II (Synthesis of MC-LF and derivates)
Ala-OH (325 mg, 1.72 mmol) and HOBt (232 mg, 1.72 mmol) were dissolved in CH2Cl2 (5 mL) and coupled according to GP6 using HBTU (651 mg, 1.72 mmol) and DIPEA (980 L, 5.72 mmol). The crude product was purified by FC to give Boc-D-Ala-Leu-OPac (Riggs Costerison
2002) (500 mg, 83%) as a white solid: Rf = 0.66 (PE/EtOAc 1:1); The analytical data for Boc-D-
Ala-Leu-OPac were in agreement with the published ones (Riggs Costerison 2002). The dipeptide Boc-D-Ala-Leu-OPac (502 mg, 1.195 mmol) was deprotected according to GP4 to give
H-D-Ala-Leu-OPac 11 that was immediately used in the next step. Crude deprotected dipeptide
11 (1.195 mmol) and Boc-NMeSecPh-OH 6 (471 mg, 1.314 mmol) were dissolved in DMF (5 mL) and coupled according to GP6 using HATU (500 mg, 1.314 mmol) and DIPEA (832 L,
4.778 mmol). The crude product was purified by FC (petroleum ether/EtOAc 1:1) to give title compound 12 (683 mg, 90%) as a white amorphous solid. The analytical data for 12 were in
1 agreement with the published ones (Riggs Costerison 2002). H NMR (400 MHz, CDCl3) 7.92 – 7.81 (dd, J = 8.5, 1.3 Hz, 2H, Ar), 7.62 (dt, J = 7.5, 1.3 Hz, 1H, Ar), 7.54 – 7.45 (m, 4H, Ar),
7.27 – 7.23 (m, 3H, Ar), 6.79 (b, 1H, NH Leu), 6.53 (b, 1H, NH Ala), 5.47 (d, J = 16.3 Hz, 1H,
CH2 Pac), 5.22 (d, J = 16.3 Hz, 1H, CH2 Pac), 4.69 (b, 1H, H-Leu), 4.50 (b, 2H, H-Ala, H-
NMeSecPh), 3.60 (dd, J = 13.0, 5.9 Hz, 1H, H- NMeSecPh), 3.16 (b, 1H, H- NMeSecPh),
2.80 (s, 3H, -NMe), 1.90 – 1.66 (m, 3H, 2xH-Leu, H-Leu), 1.45 (s, 9H, tBu), 1.38 (d, J = 6.9
Hz, 1H, Ala), 0.99 (d, J = 6.3, Hz, 3H, CH3 Leu), 0.97 (d, J = 6.3 Hz, 3H, CH3 Leu).
Bn-D-Asp(OBn)-OH (14). Starting from Bn-D-Asp-OH (8.61 g, 38.6 mmol), compound 14
(10.16 g, 84%) was prepared according to a procedure published for the synthesis of Bn-L-
Asp(OBn)-OH (Dunn et al. 1990). The analytical data of 14 were in agreement with the data
1 published for its enantiomer (Dunn et al. 1990). H NMR (400 MHz, DMSO-d6, 300 K) 7.43
– 7.18 (m, 10H, Ar), 5.10 (s, 2H, 1H, -COOCH2-), 3.92 (d, J = 13.4 Hz, 1H, N-CH2-), 3.80 (d, J
= 13.4 Hz, 1H, N-CH2-), 3.55 (t, J = 6.6 Hz, 1H, H-), 2.80 (dd, J = 16.0, 6.2 Hz, 1H, H-),
13 2.70 (dd, J = 16.0, 7.1 Hz, 1H, H-); C NMR (101 MHz, DMSO-d6, 300 K) 170.9 (C=O),
170.1 (C=O), 150.0 (C Ar), 136.8 (C Ar), 136.2 (C Ar), 129.9 (C Ar), 129.0 (C Ar), 128.9 (2C,
64 4. Manuscript II (Synthesis of MC-LF and derivates)
Ar), 128.6 (C Ar), 128.4 (2C, Ar), 128.0 (C Ar), 124.4 (C Ar), 66.5 (-COOCH2-), 56.1 (C-),
50.4 (N-CH2-), 35.4 (C-).
Bn-D-Asp(OBn)-Ot-Bu. Bn-D-Asp(OBn)-OH 14 (4.0 g, 12.8 mmol) was suspended in t-BuOAc (72 mL) and 70% aqueous perchloric acid (1.840 mL, 15.2 mmol) was added dropwise.
After stirring for 18h at rt, water (40 mL) was added and the phases were separated. The aqueous phase was extracted with EtOAc (1x 30 mL). The organic phases were combined and washed with saturated NaHCO3 solution (3x). During the workup, formation of a white precipitate could be observed. The organic phase was filtered, washed with brine and dried with Na2SO4. The volatiles were removed and Bn-D-Asp(OBn)-Ot-Bu was obtained as a pale yellow oil (3.168 g,
67%). The analytical data for Bn-D-Asp(OBn)-Ot-Bu were in agreement with the data published
26 1 for its enantiomer (Dunn et al. 1990). []D +17.5 (c 1.0, MeCN); H NMR (400 MHz, CDCl3,
300 K) 7.38 – 7.21 (m, 10H, Ar), 5.15 (d, J = 12.3 Hz, 1H, -COOCH2-) 5.11 (d, J = 12.3 Hz,
1H, -COOCH2-), 3.87 (d, J = 12.9 Hz, 1H, N-CH2-), 3.71 (d, J = 12.9 Hz, 1H, N-CH2-), 3.59
(dd, J = 7.1, 5.9 Hz, 1H, H-), 2.75 (dd, J = 15.6, 5.9 Hz, 1H, H-), 2.67 (dd, J = 15.6, 7.2 Hz,
13 1H, H-), 1.45 (s, 9H, t-Bu); C NMR (101 MHz, CDCl3, 300 K) 172.8 (C=O), 171.0 (C=O),
139.9 (C Ar), 135.9 (C Ar), 128.7 (C Ar), 128.5 (2C Ar), 128.4 (2C Ar), 128.4 (C Ar), 128.4 (C
Ar), 127.2 (C Ar), 81.8 (-C(CH3)3), 66.6 (-COOCH2-), 58.0 (C-), 52.2 (N-CH2-), 38.6 (C-),
28.2 (3C, -C(CH3)3).
(2R)-4-Benzyl-1-tert-butyl-N-benzyl-N-(9-phenylfluoren-9-yl)-aspartate (15). To the solution of Bn-D-Asp(OBn)-Ot-Bu (1.587 g, 4.274 mmol) in dry acetonitrile (47 mL) anhydrous
K3PO4 (1.088 g, 5.192 mmol) and phenylfluorenyl bromide (1.324 g, 4.122 equiv) were added.
The resulting heterogenic mixture was mechanically stirred for 24h at rt, filtered, and the solvent was removed. The crude product was purified by FC to give 15 (1.870 g, 72%) as a white solid.
The analytical data for 15 were in agreement with the data published for its enantiomer (Dunn
29 1 et al. 1990). Rf = 0.25 (petroleum ether/EtOAc 15:1); []D -25.0 (c 1.0, MeCN); H NMR (400
MHz, CDCl3, 300 K) 7.82 (d, J = 7.1 Hz, 2H, Ar), 7.74 (d, J = 7.5 Hz, 1H, Ar), 7.66 (d, J =
65 4. Manuscript II (Synthesis of MC-LF and derivates)
7.4 Hz, 1H, Ar), 7.60 (d, J = 7.5 Hz, 1H, Ar), 7.56 (d, J = 7.5 Hz, 1H, Ar), 7.47 (d, J = 7.0 Hz,
2H, Ar), 7.37 (td, J = 7.5, 1.0 Hz, 1H, Ar), 7.34 – 7.16 (m, 11H, Ar), 7.11 – 7.05 (m, 2H, Ar),
4.87 (d, J = 12.5 Hz, 1H, -COOCH2-), 4.80 (d, J = 12.5 Hz, 1H, -COOCH2-), 4.22 (d, J = 13.8
Hz, 1H, N-CH2-), 3.93 (dd, J = 10.9, 2.6 Hz, 1H, H-), 3.85 (d, J = 13.9 Hz, 1H, N-CH2-), 2.61
(dd, J = 15.8, 10.9 Hz, 1H, H-), 1.98 (dd, J = 15.9, 2.7 Hz, 1H, H-), 1.08 (s, 9H, t-Bu); 13C
NMR (101 MHz, CDCl3, 300 K) 171.3 (C=O), 171.2 (C=O), 147.7 (C Ar), 146.4 (C Ar), 143.8
(C Ar), 141.0 (C Ar), 140.3 (C Ar), 139.0 (C Ar), 136.0 (C Ar), 129.7 (2C Ar), 128.7 (2C Ar),
128.5 (2C Ar), 128.5 (2C Ar), 128.2 (2C Ar), 128.1 (C Ar), 128.0 (2C Ar), 127.7 (C Ar), 127.7
(C Ar), 127.4 (2C Ar), 127.3 (C Ar), 127.1 (C Ar), 126.7 (C Ar), 120.6 (C Ar), 120.0 (2C Ar),
80.8 (-C(CH3)3), 79.7 (C-Ph), 66.0 (-COOCH2-), 57.6 (C-), 51.8 (N-CH2-), 34.4 (C-), 27.8
(3C, -C(CH3)3); Anal. Calcd for C41H39NO4: C, 80.76; H, 6.45; N, 2.30. Found: C, 80.58; H,
6.53; N, 2.46.
(2R,3S)-4-Benzyl-1-tert-butyl-N-benzyl-N-(9-phenylfluoren-9-y1)-3-methylaspartate
(eryhtro-16). A solution of LHMDS in THF (1.0 M, 8.25 mL, 8.25 mmol) was placed in a
Schlenk flask under nitrogen atmosphere and cooled to < -20 °C. First a solution of aspartic acid derivative 15 (2.0 g, 3.28 mmol in dry THF (15 mL) and then a solution of methyl iodide (707
L, 11.35 mmol) in dry THF (11 mL) were added slowly. The mixture was stirred for 3h at < -
20 °C and 30 min at room temperature and quenched with sat. aqueous NH4Cl (10 mL) and water
(10 mL). The phases were separated and the aqueous phase was extracted with EtOAc (100 mL).
The combined organic phases were washed with brine (1x), dried over Na2SO4, filtered and the solvents were removed. The crude product was purified by FC to give eryhtro-16 (1.17 g, 57%)
26 as a white solid: Rf = 0.4 (petroleum ether/EtOAc 15:1) mp 115.5 °C; []D +250.0 (c 1.0,
1 MeCN); H NMR (400 MHz, CDCl3, 300 K) 7.83 (d, J = 7.4 Hz, 1H, Ar), 7.73 (d, J = 7.4 Hz,
1H, Ar), 7.62 (d, J = 6.9 Hz, 3H, Ar), 7.56 (d, J = 7.7 Hz, 1H, Ar), 7.47 (d, J = 7.0 Hz, 2H, Ar),
7.44 – 7.08 (m, 15H, Ar), 5.14 (d, J = 12.4 Hz, 1H, -COOCH2-), 5.02 (d, J = 12.4 Hz, 1H, -
COOCH2-), 4.67 (d, J = 14.3 Hz, 1H, CH2 N-Bn), 4.26 (d, J = 14.3 Hz, 1H, CH2 N-Bn), 3.97 (d,
66 4. Manuscript II (Synthesis of MC-LF and derivates)
J = 9.9 Hz, 1H, H-), 2.75 (dq, J = 9.9, 7.0 Hz, 1H, H-), 1.06 (s, 9H, t-Bu), 0.79 (d, J = 7.1
13 Hz, 3H, CH3); C NMR (101 MHz, CDCl3, 300 K) 174.2 (-COOBn), 170.3 (-COOt-Bu), 147.1
(C Ar), 146.0 (C Ar), 145.3 (C Ar), 142.2 (C Ar), 142.0 (2C, Ar), 139.6 (C Ar), 136.2 (C Ar),
129.2 (2C, Ar), 128.7 (C Ar), 128.6 (2C, Ar), 128.4 (2C, Ar), 128.3 (C Ar), 128.3 (C Ar), 128.2
(2C, Ar), 128.2 (2C, Ar), 128.0 (2C, Ar), 127.7 (C Ar), 127.6 (C Ar), 127.5 (2C, Ar), 127.4 (C
Ar), 127.0 (C Ar), 126.7 (C Ar), 80.9 (-C(CH3)3), 80.5 (C-Ph), 66.1 (CH2 Bn), 64.0 (C-), 51.7
(CH2 Bn), 42.5 (C-), 27.8 (3C, -C(CH3)3), 15.4 (CH3); Anal. Calcd for C42H41NO4: C, 80.87;
H, 6.63; N, 2.25. Found: C, 80.84; H, 6.67; N, 2.42.
H--D-MeAsp-OH, ((3S)--D-methylaspartic acid). The amino acid erythro-16 (200 mg, 1.6 mmol) was suspended in methanol (2.84 mL) and 10% Pd/C catalyst (82 mg, wet, 53.7% water) was added. The reaction mixture was hydrogenated overnight at a slightly positive hydrogen pressure and filtered through a syringe filter. The filtrate was diluted 1:1 with aqueous HCl (0.1
M) and filtered twice through a syringe filter. The volatiles were removed and the residue was coevapotated with EtOAc (1x) and toluene (2x). The resulting solid was treated with TFA (0.7 mL) according to GP4. The crude product was purified by HPLC (1% B isocratic over 10 min, tR = 4.0 min) to give the formic acid salt of H--D-MeAsp-OH (Armstrong et al. 2007) as a
D white amorphous solid (40 mg, 85%). The optical rotation ([] 23.5 -35.91 (c 1.07, 5M HCl))
D was in agreement with published values for H--D-MeAsp-OH ([] 22 -31.0 (c 2.00, 5M HCl))
D (Armstrong et al. 2007) and enantiomeric (3R)--l-methylaspartic acid ([] 21 +34.3 (c 2.05,
3 1 5M HCl)) (Sakaguchi et al. 2004). A coupling constant JCH-CH of 9.0 Hz determined by H
NMR (400 MHz, D2O, 300 K, pD > 14) additionally verifies the erythro configuration of H--
1 D-MeAsp-OH (Bochenska and Biernat 1972). H NMR (400 MHz, D2O, 300 K, pD > 14) 2.58
(d, J = 9.0 Hz, 1H, H-), 1.75 (dq, J = 8.9, 7.2 Hz, 1H, H-), 0.44 (d, J = 7.1 Hz, 3H, -CH3).
N-(tert-Butoxycarbonyl)-O-prop-2-yn-1-yl-L-tyrosine (Boc-Tyr(Prg)-OH) (18) (dos Anjos et al. 2007). 2-tert-Butoxycarbonylamino-3-[4-(prop-2-ynyloxy)phenyl]-propionic acid propargyl ester (Deiters et al. 2003) (3.633 g, 10.17 mmol) was dissolved in 1M KOH solution
67 4. Manuscript II (Synthesis of MC-LF and derivates) in methanol (20.34 mL, 20.34 mmol) and sonicated for 2h at 35 °C. The methanol was removed under reduced pressure. The residue was dissolved in H2O (30 mL) and the aqueous phase was washed with Et2O (2x30 mL). The aqueous phase was acidified with KHSO4 (1 M) to pH = 2 and extracted with EtOAc (2x30 mL). The EtOAc extracts were combined, washed with brine, dried over MgSO4, and the volatiles were removed under reduced pressure. Product 18 was obtained as a white amorphous solid (3.1 g, 95%) and used without further purification. The analytical data for 18 were in agreement with the published ones (dos Anjos et al. 2007). 1H
NMR (400 MHz, CDCl3) 7.12 (d, J = 8.5 Hz, 2H, Ar), 6.92 (d, J = 8.6 Hz, 2H, Ar), 4.92 (d, J
= 7.0 Hz, 1H, NH), 4.92 (d, J = 2.4 Hz, 2H, -OCH2), 4.56 (b, 1H, H-), 3.14 (dd, J = 13.9, 5.2
Hz, 1H, H-), 3.04 (dd, J = 13.8, 5.5 Hz, 1H, H-), 2.51 (t, J = 2.3 Hz, 1H, -C≡CH), 1.42 (s,
9H, t-Bu).
Fmoc--D-Glu(Ot-Bu)-N-MeSecPh-D-Ala-Leu--D-MeAsp(Ot-Bu)-Phe-OH (21a).
Tetrapeptide 4 (61 mg, 63 mol) and dipeptide 5a (50 mg, 72 mol) were deprotected according to GP2 and GP3, respectively. The products 19 and 20a were dissolved in DMF (1 mL) and coupled according to GP6 using HATU (28 mg, 74 mol) and collidine (25 L, 188 mol). The crude product was purified by FC (CH2Cl2/i-PrOH 99:1 to 95:5) to give 21a (49 mg, 60%) as a
1 white amorphous solid: Rf = 0.65 (CH2Cl2/i-PrOH 95:5); H NMR (600 MHz, DMSO-d6, 360
K) 8.22 (d, J = 8.0 Hz, 1H, NH), 7.94 (d, J = 7.3 Hz, 2H, Ar), 7.86 (d, J = 7.5 Hz, 2H, Ar),
7.69 – 7.66 (m, 4H, 3x H-Ar, NH), 7.61 (s, 1H, NH), 7.60 – 7.52 (m, 3H, 2x H-Ar, NH), 7.49
(d, J = 6.9 Hz, 2H, Ar), 7.40 (t, J = 7.4 Hz, 2H, Ar), 7.31 (t, J = 7.4 Hz, 2H, Ar), 7.29 – 7.22 (m,
7H, Ar), 7.21 – 7.17 (m, 1H, Ar), 5.45 (d, J = 16.4 Hz, 1H, CH2 Pac), 5.42 (d, J = 16.4 Hz, 1H,
CH2 Pac), 5.08 – 5.00 (b, 1H, H- NMeSecPh), 4.72 (td, J = 8.8, 5.0 Hz, 1H, H- Phe), 4.36 –
4.26 (m, 4H, H- MeAsp, H- Ala, -OCH2-CH- Fmoc), 4.25 – 4.19 (m, 2H, H- Leu, -OCH2-
CH- Fmoc), 4.03 – 3.98 (m, 1H, H- Glu), 3.52 – 3.45 (m, 1H, H- Phe), 3.25 (dd, J = 14.1,
5.0 Hz, 1H, H- Phe), 3.17 – 3.11 (m, 1H, H- NMeSecPh), 2.99 (dd, J = 14.2, 9.4 Hz, 1H, H-
Phe), 2.96 – 2.93 (m, 1H, H- MeAsp), 2.87 (bs, 3H, -NCH3), 2.42 – 2.34 (m, 2H, 2x H-
68 4. Manuscript II (Synthesis of MC-LF and derivates)
Glu), 2.06 – 1.99 (m, 1H, H- Glu), 1.93 – 1.83 (m, 1H, H- Glu), 1.63 – 1.55 (m, H- Leu),
1.53 – 1.48 (m, 2H, 2x H- Leu), 1.41 (s, 9H, t-Bu), 1.34 (s, 9H, t-Bu), 1.22 (d, J = 6.8 Hz, 3H,
CH3 Ala), 0.88 (d, J = 7.1 Hz, 3H, CH3 MeAsp), 0.85 (d, J = 6.6 Hz, 3H, 3xH- Leu), 0.80 (d, J
+ = 6.5 Hz, 3H, 3xH- Leu); HRMS (ESI-TOF) m/z: [M+H] Calcd for C69H85N6O14Se 1301.52835; Found 1301.52994.
Fmoc--D-Glu(Ot-Bu)-N-MeSecPh-D-Ala-Leu--D-MeAsp(Ot-Bu)-Phe-d5-OH (21b).
Tetrapeptide 4 (150 mg, 164.5 mol) and dipeptide 5b (110 mg, 158.1 mol) were deprotected according to GP2 and GP3, respectively. The crude peptides 19 and 20b were dissolved in DMF
(1 mL) and coupled according to GP6 using HATU (90 mg, 237.2 mol) and collidine (160 L,
1207 mol). The crude product was purified by FC (CH2Cl2/i-PrOH 99:1 to 95:5) to give 21b
1 (180 mg, 87%) as a white amorphous solid: Rf = 0.65 (CH2Cl2/i-PrOH 95:5); The H NMR spectrum (400 MHz) recorded at 300 K showed two sets of signals (ratio approx. 5:1) and peak broadening due to the occurrence of two rotamers of the N-methylated amide bond; due to incomplete H/D-exchange some remaining NH-protons are visible. 1H NMR (400 MHz,
CD3OD, 300 K) 8.68 (d, J = 8.8 Hz, 1H, NH), 7.97 – 7.87 (m, 4H, 2x H-Ar, 2x NH), 7.78 (d,
J = 7.6 Hz, 2H, Ar), 7.71 (d, J = 7.7 Hz, 1H, NH), 7.62 (t, J = 6.9 Hz, 3H, Ar), 7.51 – 7.42 (m,
4H, Ar), 7.37 (t, J = 7.4 Hz, 1H, Ar), 7.31 – 7.19 (m, 5H, Ar), 7.04 (d, J = 8.6 Hz, 1H, NH) 5.49
– 5.46 (m, 2H, CH2 Pac), 4.88 (dd, J = 9.7, 4.1 Hz, 1H, H- Phe-d5), 4.44 – 4.38 (m, 1H, H-
Ala), 4.37 – 4.21 (m, 5H, H- Leu, -OCH2-CH- Fmoc, H- Ala, -OCH2-CH- Fmoc), 4.20 – 4.08
(m, 2H, H- NMeSecPh, H- Glu), 3.66 (dd, J = 13.3, 5.0 Hz, 1H, H- NMeSecPh), 3.47 (dd,
J = 13.3, 9.7 Hz, 1H, H- NMeSecPh), 3.40 (dd, J = 14.1, 4.3 Hz, 1H, H- Phe-d5), 3.20 (s,
3H, -NCH3, minor rotamer), 3.11 (s, 3H, -NCH3, major rotamer), 3.05 – 2.94 (m, 2H, H- Phe- d5, H- MeAsp), 2.54 – 2.44 (m, 1H, H- Glu), 2.25 – 2.11 (m, 2H, H- Glu, H- Glu), 2.00 –
1.89 (m, 1H, H- Leu), 1.87 – 1.71 (m, 2H, H- Leu, H- Glu), 1.61 – 1.48 (m, 1H, H- Leu),
1.45 (s, 9H, t-Bu), 1.38 (s, 9H, t-Bu), 1.19 (d, J = 7.3 Hz, 3H, CH3 Ala), 0.97 (d, J = 6.6 Hz, 3H,
69 4. Manuscript II (Synthesis of MC-LF and derivates)
+ 3xH- Leu), 0.86 (m, 6H, 3xH- Leu, CH3 MeAsp); HRMS (ESI-Orbitrap) m/z: [M+H] Calcd for C69H80D5N6O14Se 1306.55973; Found 1306. 56074.
Fmoc--D-Glu(Ot-Bu)-N-MeSecPh-D-Ala-Leu--D-MeAsp(Ot-Bu)-Tyr(Prg)-OH (21c).
Tetrapeptide 4 (130 mg, 134 mol) and dipeptide 5c (109 mg, 146 mol) were deprotected according to GP2 and GP3, respectively. The products 19 and 20c were dissolved in DMF (1 mL) and coupled according to GP6 using HATU (56 mg, 146 mol) and collidine (53 L, 403
mol). The crude product was purified by FC (CH2Cl2/i-PrOH 99:1 to 95:5) to give 21c (116
1 mg, 76%) as a white amorphous solid: Rf = 0.63 (CH2Cl2/i-PrOH 95:5); H NMR (600 MHz,
DMSO-d6, 360 K) 8.19 (d, J = 7.3 Hz, 1H, NH), 7.94 (d, J = 7.2 Hz, 2H, Ar), 7.86 (d, J = 7.5
Hz, 2H, Ar), 7.71 – 7.65 (m, 3H, 3x H-Ar, NH), 7.60 (d, J = 8.6 Hz, 2H, 2x NH), 7.55 (t, J = 7.7
Hz, 2H, Ar), 7.49 (d, J = 8.1 Hz, 2H, Ar), 7.40 (t, J = 7.4 Hz, 2H, Ar), 7.31 (t, J = 7.4 Hz, 2H,
Ar), 7.29 – 7.22 (m, 4H, 3x H-Ar, NH), 7.18 (d, J = 8.6 Hz, 2H, Ar), 6.89 (d, J = 8.6 Hz, 2H,
Ar), 5.46 – 5.41 (m, 2H, CH2 Pac), 5.08 – 5.00 (m, 1H, H- NMeSecPh), 4.72 (d, J = 2.3 Hz,
2H, -OCH2- Tyr(Prg)), 4.66 (td, J = 8.6, 5.3 Hz, 1H, H- Tyr(Prg)), 4.36 – 4.26 (m, 4H, H-
MeAsp, H- Ala, -OCH2-CH- Fmoc), 4.25 – 4.19 (m, 2H, H- Leu, -OCH2-CH- Fmoc), 4.01
(dd, J = 13.4, 8.4 Hz, 1H, H- Glu), 3.52 – 3.46 (m, 1H, H- NMeSecPh), 3.32 (t, J = 2.3 Hz,
1H, -C≡CH), 3.18 (dd, J = 14.3, 5.1 Hz, 1H, H- Tyr(Prg)), 3.16 – 3.12 (m, 1H, H-
NMeSecPh), 2.99 – 2.91 (m, 2H, H- Tyr(Prg), H- MeAsp), 2.87 (bs, 3H, -NCH3), 2.44 –
2.35 (m, 2H, 2x H- Glu), 2.06 – 1.99 (m, 1H, H- Glu), 1.93 – 1.85 (m, 1H, H- Glu), 1.63 –
1.56 (m, 1H, H- Leu), 1.53 – 1.48 (m, 2H, 2x H- Leu), 1.41 (s, 9H, t-Bu), 1.34 (s, 9H, t-Bu),
1.23 (d, J = 6.9 Hz, 3H, CH3 Ala), 0.91 (d, J = 7.2 Hz, 3H, CH3 MeAsp), 0.85 (d, J = 6.6 Hz,
3H, 3xH- Leu), 0.81 (d, J = 6.5 Hz, 3H, 3xH- Leu); HRMS (ESI-TOF) m/z: [M+H]+ Calcd for C72H87N6O15Se 1355.53891; Found 1355.54122.
Boc-Adda--D-Glu(Ot-Bu)-N-MeSecPh-D-Ala-Leu--D-MeAsp(Ot-Bu)-Phe-OH (22a). The hexapeptide 21a (100 mg, 77.0 mol) was N-terminally deprotected according to GP3. The deprotected peptide was coupled with Boc-Adda-OH 3 (Pearson et al. 2000) (31 mg, 70.0 mol)
70 4. Manuscript II (Synthesis of MC-LF and derivates) according to GP6 using HATU (29 mg, 70.0 mol) and collidine (28 L, 210 mol) in DMF (1 mL). The crude product was purified by FC (95:5 CH2Cl2/i-PrOH, Rf = 0.55) to give phenacyl protected heptapeptide 2a as a white amorphous solid. The phenacyl group of 2a (50 mg, 33.5
mol) was removed according to GP2 and the crude product was purified by semi-preparative
RP-HPLC (gradient: 75-100% B in 20 min, tR = 19.9 min) to give 22a as a white solid (30 mg,
1 31% from 21a). H NMR (600 MHz, DMSO-d6, 360 K) 7.91 (d, J = 4.7 Hz, 1H, NH), 7.82 (d,
J = 7.0 Hz, 1H, NH), 7.77 (d, J = 6.9 Hz, 1H, NH), 7.68 (b, 1H, NH), 7.50 – 7.49 (m, 2H, Ar),
7.30 – 7.21 (m, 7H, Ar), 7.21 – 7.14 (m, 6H, Ar), 6.27 (s, 1H, NH), 6.09 (d, J = 15.7 Hz, 1H, H-
5 Adda), 5.48 (dd, J = 15.7, 6.6 Hz, 1H, H-4 Adda), 5.38 (d, J = 9.6 Hz, 1H, H-7 Adda), 5.12 –
5.05 (m, 1H, H- NMeSecPh), 4.50 – 4.44 (m, 1H, H- Phe), 4.32 – 4.28 (m, 1H, H- Ala),
4.26 – 4.20 (m, H- Leu, H- MeAsp), 4.15 – 4.12 (b, 1H, H- Glu), 4.09 (dd, J = 15.1, 6.9
Hz, 1H, H-3 Adda), 3.51 (d, J = 12.1 Hz, 1H, H- NMeSecPh), 3.27 – 3.23 (m, 1H, H-9 Adda),
3.18 (s, 3H, -OCH3), 3.10 (b, 2H, H- Phe, H- NMeSecPh), 2.94 – 2.91 (m, 1H, H- MeAsp),
2.88 (dd, J = 13.9, 8.8 Hz, 1H, H- Phe), 2.83 (s, 3H, -NCH3), 2.74 (dd, J = 14.0, 4.8 Hz, 1H,
H-10 Adda), 2.66 (dd, J = 14.0, 7.3 Hz, 1H, H-10 Adda), 2.61 – 2.54 (m, 2H, H-2 Adda, H-8
Adda), 2.38 – 2.29 (m, 2H, 2xH- Glu), 2.03 – 1.97 (m, 1H, H- Glu), 1.87 -1.80 (m, 1H, H-
Glu), 1.62 – 1.58 (m, 1H, H- Leu), 1.56 (s, 3H, 3xH-6’ Adda), 1.52 – 1.48 (m, 2H, H- Leu),
1.40 (s, 9H, t-Bu), 1.38 (s, 9H, t-Bu), 1.36 (s, 9H, t-Bu), 1.23 (d, J = 7.1 Hz, 3H, CH3 Ala), 1.05
(d, J = 7.0 Hz, 3H, 3xH-2’ Adda), 0.97 (d, J = 6.8 Hz, 3H, 3xH-8’ Adda), 0.90 (d, J = 7.1 Hz,
3H, CH3 MeAsp), 0.86 (d, J = 6.6 Hz, 3H, 3xH- Leu), 0.82 (d, J = 6.5 Hz, 3H, 3xH- Leu);
+ HRMS (ESI-TOF) m/z: [M+H] Calcd for C71H104N7O15Se 1374.67624; Found 1374.67841.
Boc-Adda--D-Glu(Ot-Bu)-N-MeSecPh-D-Ala-Leu--D-MeAsp(Ot-Bu)-Phe-d5-OH (22b).
The hexapeptide 21b (90 mg, 69.0 mol) was N-terminally deprotected according to GP3. The deprotected peptide was coupled with Boc-Adda-OH 3 (Pearson et al. 2000) (25 mg, 58.0 mol) according to GP6 using HATU (33 mg, 87.0 mol) and collidine (20 L, 145 mol) in DMF (2 mL). The crude product and purified by FC (95:5 CH2Cl2/i-PrOH, Rf = 0.55) to give phenacyl
71 4. Manuscript II (Synthesis of MC-LF and derivates) protected heptapeptide 2b as a white solid. The phenacyl group of 2b (63 mg, 42.1 mol) was removed according to GP2 and the crude product was purified by semi-preparative RP-HPLC
(gradient: 75-100% B in 20 min, tR = 18.5 min) to give 22b as a white amorphous solid (23 mg,
25% from 21b). The 1H NMR spectrum (600 MHz) recorded at 300 K showed two sets of signals
(ratio approx. 2:1) and peak broadening due to the occurrence of two rotamers of the N-
1 methylated amide bond. H NMR (600 MHz, CD3OD, 300 K) 7.56 – 7.52 (m, 2H, Ar), 7.33 –
7.27 (m, 3H, Ar), 7.26 – 7.23 (m, 2H, Ar), 7.20 – 7.14 (m, 3H, Ar), 6.22 (d, J = 15.5 Hz, 1H, H-
5 Adda), 5.59 – 5.51 (m, 1H, H-4 Adda), 5.39 (d, J = 9.7 Hz, 1H, H-7 Adda), 4.62 – 4.55 (m,
1H, H- Phe-d5), 4.36 – 4.29 (m, 5H, H-Glu, H-Leu, H-Ala, H-Mdha), 4.25 – 4.19 (m,
1H, H-3 Adda major rotamer), 4.06 – 4.03 (m, 1H, H-3 Adda major rotamer), 3.67 (d, J = 12.7
Hz, 1H, 1H, H- NMeSecPh), 3.41 – 3.35 (m, 1H, H- NMeSecPh), 3.22 (s, 1H, 3H, -OCH3),
3.27 – 3.17 (m, 2H, H- Phe-d5, H-9 Adda), 3.11 – 2.99 (m, 1H, H- MeAsp), 3.05 (s, 3H, -
NCH3), 2.93 (dd, J = 13.8, 9.2 Hz, 1H, H- Phe-d5), 2.80 (dd, J = 13.9, 3.7 Hz, 1H, H-10 Adda),
2.66 (dd, J = 13.9, 7.4 Hz, 2H, H-10 Adda, H-2 Adda), 2.59 (dq, J = 16.7, 6.7 Hz, 1H, H-8
Adda), 2.45 – 2.39 (m, 1H, H- Glu), 2.36 – 2.25 (m, 2H, H- Glu, H- Glu major rotamer),
2.06 – 1.98 (m, 1H, H- Glu minor rotamer) 1.95 – 1.87 (m, 1H, H- Leu), 1.78 – 1.68 (m,
2H, H- Leu, H- Glu), 1.62 (s, 3H, 3xH-6’ Adda), 1.58 – 1.51 (m, 1H, H- Leu), 1.46 (s, 9H, t-Bu), 1.42 (s, 9H, t-Bu), 1.41 (s, 9H, t-Bu), 1.25 (d, J = 7.4 Hz, 3H, CH3 Ala), 1.16 (s, 3H, 3xH-
2’ Adda), 1.02 (d, J = 6.7 Hz, 3H, 3xH-8’ Adda), 0.96 (d, J = 6.6 Hz, 3H, CH3 MeAsp), 0.91 (d,
J = 7.1 Hz, 3H, 3xH- Leu), 0.88 (d, J = 6.5 Hz, 3H, 3xH- Leu); HRMS (ESI-Orbitrap) m/z:
+ [M+H] Calcd for C71H99D5N7O15Se 1379.70640; Found 1379. 70814.
Boc-Adda--D-Glu(Ot-Bu)-N-MeSecPh-D-Ala-Leu--D-MeAsp(Ot-Bu)-Tyr(Prg)-OH
(22c). The hexapeptide 21c (110 mg, 81.2 mol) was N-terminally deprotected according to
GP3. The deprotected peptide was coupled with Boc-Adda-OH 3 (Pearson et al. 2000) (33 mg,
73.8 mol) according to GP6 using HATU (31 mg, 81.2 mol) and collidine (30 L, 226 mol) in DMF (1 mL). The crude product was purified by FC (95:5 CH2Cl2/i-PrOH) to give phenacyl
72 4. Manuscript II (Synthesis of MC-LF and derivates) protected heptapeptide 2c as a white solid (85 mg). The phenacyl group of 2c (50 mg, 33.5 mol) was removed according to GP2 and the crude product was purified by semi-preparative RP-
HPLC (gradient: 75-100% B in 20 min, tR = 18.4 min) to give 22c as a white amorphous solid
(34 mg, 54% from 21c). The 1H NMR spectrum (600 MHz) recorded at 300 K showed two sets of signals (ratio approx. 2:1) and peak broadening due to the occurrence of two rotamers of the
1 N-methylated amide bond. H NMR (600 MHz, CD3OD, 300 K) 7.55 (d, J = 6.2 Hz, 2H, Ar),
7.32 – 7.27 (m, 3H, Ar), 7.25 (t, J = 7.5 Hz, 2H, Ar), 7.20 – 7.16 (m, 3H, Ar), 7.12 (d, J = 7.7
Hz, 2H, Ar), 6.88 (d, J = 8.5 Hz, 2H, Ar), 6.22 (d, J = 15.8 Hz, 1H, H-5 Adda), 5.59 – 5.51 (m,
1H, H-4 Adda), 5.39 (d, J = 9.8 Hz, 1H, H-7 Adda), 4.67 (d, J = 2.2 Hz, 2H, -OCH2-), 4.59 –
4.53 (m, 1H, H- Tyr(Prg)), 4.40 – 4.28 (m, 5H, H- Leu, H- Glu, H- Ala, H- MeAsp, H-
NMeSecPh), 4.23 (b, 1H, H-3 Adda), 3.69 – 3.63 (m, 1H, H- NMeSecPh major rotamer),
3.57 – 3.51 (m, 1H, H- NMeSecPh minor rotamer), 3.40 – 3.35 (m, 1H, H- NMeSecPh major rotamer) 3.22 (s, 3H, -OCH3), 3.23 – 3.18 (m, 1H, H-9 Adda), 3.16 – 3.11 (m, 1H, H-
Tyr(Prg)), 3.04 (b, 3H, -NCH3, H- MeAsp), 2.91 (s, 1H, -C≡CH), 2.87 (dd, J = 13.9, 9.4 Hz,
1H, H- Tyr(Prg)), 2.80 (dd, J = 13.8, 3.7 Hz, 1H, H-10 Adda), 2.68 – 2.63 (m, 2H, H-2 Adda,
H-10 Adda), 2.63 – 2.56 (m, 1H, H-8 Adda), 2.44 – 2.37 (m, 1H, H- Glu), 2.34 – 2.25 (m, 2H,
H- Glu, H- Glu major rotamer), 2.04 – 1.98 (m, 1H, H- Glu minor rotamer) 1.95 – 1.88 (m,
1H, H- Leu), 1.78 – 1.68 (m, 2H, H- Leu H- Glu), 1.62 (s, 3H, 3xH-6’ Adda), 1.61 – 1.51
(m, 1H, H- Leu), 1.46 (s, 9H, t-Bu), 1.42 (s, 18H, 2xt-Bu), 1.25 (d, J = 7.4 Hz, 3H, CH3 Ala),
1.19 – 1.12 (b, 3H, 3xH-2’ Adda), 1.02 (d, J = 6.7 Hz, 3H, 3xH-8’ Adda), 0.96 (d, J = 6.6 Hz,
13 3H, 3xH- Leu), 0.93 (d, J = 7.2 Hz, 3H, CH3 MeAsp), 0.88 (d, J = 6.5 Hz, 3H, 3xH- Leu); C
NMR (151 MHz, CD3OD, 300 K): = 177.1 (C=O), 176.3 (C=O), 175.8 (C=O), 172.4 (C=O),
175.2 (C=O), 172.6 (C=O), 171.8 (C=O), 170.4 (C=O), 158.1 (-C-O-CH2-), 140.6 (C, Ar),
137.2 (C-7 Adda), 137.0 (C-5 Adda), 134.0 (C-6 Adda), 133.9 (2C, Ar), 131.3 (2C, Ar), 131.1
(C, Ar), 130.5 (2C, Ar), 130.4 (2C, Ar), 129.2 (2C, Ar), 128.3 (C, Ar), 127.1 (C-4 Adda), 127.1
(2C, Ar), 116.0 (2C, Ar), 88.4 (C-9 Adda), 83.2 (-C(CH3)3), 82.9 (-C(CH3)3), 80.5 (-C(CH3)3),
73 4. Manuscript II (Synthesis of MC-LF and derivates)
79.9 (-C≡CH), 76.8 (-C≡CH), 64.0 (C-), 58.8 (-OCH3), 56.7 (C-), 56.6 (-OCH2-), 55.9 (C-),
54.9 (C-), 53.8 (C-), 53.5 (C-), 50.7 (C- Ala), 45.3 (C-2 Adda), 41.9 (-NCH3), 41.2 (C-
Leu), 39.1 (C-10 Adda), 37.9 (C- Tyr(Prg)), 37.8 (C-8 Adda), 37.7 (C- MeAsp), 30.8 (C-
Glu), 28.8 (3C, -C(CH3)3), 28.4 (3C, -C(CH3)3), 28.4 (C- Glu), 28.2 (3C, -C(CH3)3), 27.0 (C-
NMeSecPh), 25.8 (C- Leu), 23.9 (C- Leu), 21.3 (C- Leu), 17.5 (CH3 Ala), 16.6 (C-8’
Adda), 16.3 (CH3 MeAsp), 15.7 (C-2’ Adda), 13.2 (C-6’ Adda); HRMS (ESI-Orbitrap) m/z:
+ [M+H] Calcd for C74H106N7O16Se 1428.68558; Found 1428. 68698.
4 7 [Phe-d5 , NMeSecPh ]-Microcystin-LF (23b). The heptapeptide 22b (21.7 mg, 15.7 mol) was macrocyclized using pentafluorophenol (5.7 mg, 30.7 mol) and DCC (4 mg, 19.2 mol) according to GP7. The crude product was purified by semi-preparative RP-HPLC (gradient: 50-
75% B in 25 min, tR = 20.5 min) and macrocyclic peptide 23b was obtained as a white amorphous
1 solid (5 mg, 28%). H NMR (600 MHz, CD3OD, 300 K) 8.04 – 8.02 (m, 1H, NH MeAsp), 8.00 (d, J = 6.4 Hz, 1H, NH Leu), 7.51 (d, J = 7.7 Hz, 2H, Ar), 7.35 – 7.29 (m, 3H, Ar), 7.27 –
7.24 (m, 2H, Ar), 7.22 – 7.14 (m, 4H, 3xH-Ar, NH Ala), 6.32 (d, J = 15.5 Hz, 1H, H-5 Adda),
5.49 (d, J = 9.7 Hz, 1H, H-7 Adda), 5.43 (dd, J = 15.4, 9.0 Hz, H-4 Adda), 4.65 – 4.59 (m, 1H,
H-3 Adda), 4.48 (dd, J = 12.0, 3.6 Hz, 1H, H- Phe-d5), 4.46 – 4.39 (m, 2H, H- MeAsp, H-
Ala), 4.19 – 4.12 (m, 2H, H- Glu, H- Leu), 4.04 (dd, J = 9.7, 4.9 Hz, 1H, H- NMeSecPh),
3.73 – 3.66 (m, 2H, 2xH- NMeSecPh), 3.47 – 3.41 (m, 1H, H- Phe-d5), 3.31 – 3.27 (m, 1H,
H-9 Adda) 3.26 (s, 3H,-OCH3), 3.19 (s, -NCH3), 2.93 (dd, J = 7.0, 3.0 Hz, 1H, H- MeAsp),
2.84 (dd, J = 14.0, 4.6 Hz, 1H, 1H, H-10 Adda), 2.81 – 2.76 (m, 1H, H-2 Adda), 2.70 (dd, J =
13.9, 7.2 Hz, 1H, H-10 Adda), 2.66 – 2.59 (m, 1H, H-8 Adda), 2.55 (dd, J = 14.2, 11.9 Hz, 1H,
H- Phe-d5), 2.20 – 2.13 (m, 1H, H- Glu), 1.96 – 1.89 (m, 1H, H- Leu), 1.84 – 1.75 (m, 2H,
H- Glu, H- Leu), 1.64 (s, 3H, 3xH-6’ Adda), 1.58 – 1.47 (m, 3H, H- Glu, H- Glu, H-
Leu), 1.08 (d, J = 6.8 Hz, 3H, 3xH-2’ Adda), 1.04 (d, J = 6.7 Hz, 3H, 3xH-8’ Adda), 0.97 (d, J
= 6.6 Hz, 3H, 3xH- Leu), 0.93 (d, J = 7.4 Hz, 3H, CH3 Ala), 0.88 (d, J = 6.6 Hz, 3H, 3xH-
74 4. Manuscript II (Synthesis of MC-LF and derivates)
+ Leu), 0.72 (d, J = 7.1 Hz, 3H, CH3 MeAsp); HRMS (ESI-TOF) m/z: [M+H] Calcd for
C58H73D5N7O12Se 1149.5182; Found 1149.5175.
[NMeSecPh7]-Microcystin-LY(Prg) (23c). The heptapeptide 22c (29 mg, 20.3 mol) was macrocyclized using pentafluorophenol (7.3 mg, 39.6 mol) and DCC (5.1 mg, 24.8 mol) according to GP7. The crude product was purified by semi-preparative RP-HPLC (gradient: 50-
80% B in 30 min, tR = 20.6 min) and the macrocyclic peptide 23c was obtained as a white
1 amorphous solid (8 mg, 33%). H NMR (600 MHz, CD3OD, 300 K) 8.83 (d, J = 9.6 Hz, 1H,
NH Tyr(Prg)), 8.19 (d, J = 8.9 Hz, 1H, NH MeAsp), 8.00 (d, J = 6.5 Hz, 1H, NH Glu), 7.52 (dd,
J = 8.0, 1.4 Hz, 2H, Ar), 7.36 – 7.28 (m, 3H, Ar), 7.26 (t, J = 7.5 Hz, 2H, Ar), 7.22 – 7.16 (m,
4H, 3xH-Ar, NH Adda), 7.12 (d, J = 8.4 Hz, 1H, NH Ala), 7.09 (d, J = 8.6 Hz, 2H, Ar), 6.84 (d,
J = 8.8 Hz, 2H, Ar), 6.34 (d, J = 15.5 Hz, 1H, H-5 Adda), 5.50 (d, J = 9.9 Hz, 1H, H-7 Adda),
5.39 (dd, J = 15.5, 8.7 Hz, 1H, H-4 Adda), 4.70 – 4.65 (m, 1H, H-3 Adda), 4.64 (d, J = 2.4 Hz,
2H, -OCH2-), 4.51 (dd, J = 8.9, 3.0 Hz, 1H, H- MeAsp), 4.49 – 4.41 (m, 2H, H- Ala, H-
Tyr(Prg)), 4.25 (dd, J = 9.2, 5.6 Hz, 1H, H- Glu), 4.15 (ddd, J = 11.2, 6.5, 3.9 Hz, 1H, H-
Leu), 4.06 (dd, J = 9.8, 4.8 Hz, 1H, H- NMeSecPh), 3.75 – 3.65 (m, 2H, 2xH- NMeSecPh),
3.41 (dd, J = 14.1, 3.3 Hz, 1H, H- Tyr(Prg)), 3.30 – 3.26 (m, 1H, H-9 Adda), 3.26 (s, 3H,-
OCH3), 3.21 (s, 3H, -NCH3), 2.91 (t, J = 2.4 Hz, 1H, -C≡CH), 2.87 (qd, J = 7.2, 3.0 Hz, 1H, H-
MeAsp), 2.84 (dd, J = 14.0, 4.8 Hz, 1H, H-10 Adda), 2.70 (dd, J = 14.0, 7.3 Hz, 1H, H-10
Adda), 2.66 – 2.59 (m, 2H, H-2 Adda, H-8 Adda), 2.49 (dd, J = 14.1, 12.0 Hz, 1H, H-
Tyr(Prg)), 2.18 – 2.12 (m, 1H, H- Glu), 1.90 – 1.84 (m, 1H, H- Glu), 1.84 – 1.74 (m, 2H, H-
Leu, H- Leu), 1.64 (s, 3H, 3xH-6’ Adda), 1.59 – 1.54 (m, 1H, H- Glu), 1.52 – 1.43 (m, 2H,
H-Glu, H- Leu), 1.09 (d, J = 6.9 Hz, 3H, 3xH-2’ Adda), 1.04 (d, J = 6.7 Hz, 3H, 3xH-8’
Adda), 0.97 (d, J = 4.7 Hz, 3H, 3xH- Leu), 0.96 (d, J = 5.5 Hz, 3H, CH3 Ala), 0.88 (d, J = 6.5
13 Hz, 3H, 3xH- Leu), 0.77 (d, J = 7.2 Hz, 3H, CH3 MeAsp); C NMR (151 MHz, CD3OD, 300
K): = 178.3 (C=O), 176.41 (C=O), 176.38 (C=O), 175.80 (C=O), 175.77 (C=O), 175.3 (C=O),
174.9 (C=O), 171.64 (C=O), 171.54 (C=O), 158.0 (-C-O-CH2-), 140.5 (C Ar), 139.4 (C-5 Adda),
75 4. Manuscript II (Synthesis of MC-LF and derivates)
137.7 (C-7 Adda), 133.6 (C-6 Adda), 133.4 (2xC Ar), 131.7 (C Ar), 131.1 (2xC Ar Tyr(Prg)),
130.9 (C Ar), 130.5 (2xC Ar), 129.2 (2xC Ar), 128.5 (C Ar), 127.1 (2xC Ar), 125.8 (C-4 Adda),
116.0 (2xC Ar Tyr(Prg)), 88.4 (C-9 Adda), 79.8 (-C≡CH), 76.8 (-C≡CH), 67.5 (C-
NMeSecPh), 58.6 (-OCH3), 56.5 (-OCH2-), 56.0 (C-3 Adda), 55.31, 55.28, 55.19 (m, 3C, C-
Leu, C- MeAsp, C- Tyr(Prg)), 53.2 (C- Glu), 49.7 (C- Ala), 46.0 (C-2 Adda), 40.9 (-
NCH3), 40.7 (C- Leu), 40.4 (C- MeAsp), 39.0 (C-10 Adda), 37.7 (C-8 Adda), 37.3 (C-
Tyr(Prg)), 32.8 (C- Glu), 29.4 (C- Glu), 26.4 (C- NMeSecPh), 25.8 (C- Leu), 23.6 (CH3
Leu), 21.2 (CH3 Leu), 17.3 (CH3 Ala), 16.5 (C-8’ Adda), 16.4 (C-2’ Adda), 15.2 (CH3 MeAsp),
+ 12.9 (C-6’ Adda); HRMS (ESI-Orbitrap) m/z: [M+H] Calcd for C61H80N7O13Se 1198.49738;
Found 1198. 49957.
4.4 Results and discussion
4.4.1 Retrosynthetic Analysis.
For the synthesis of 1a-c we followed a fragment-based strategy using tert-butyl esters as protecting groups for -D-MeAsp3 and -D-Glu6 in order to suppress aspartimide formation
(Lauer et al. 1995). In contrast to the previous approach (Humphrey et al. 1996), in which N- methylphosphonylsarcosine was incorporated and subsequently converted into Mdha7 by a
Horner-Wadsworth-Emmons reaction, we incorporated N-methylphenylselenocysteine
(NMeSecPh) as Mdha precursor that was transformed into Mdha in the final step of the synthesis.
For the macrocyclization, we decided to use the peptide bond between residues 4 and 5 which is also the cyclization site during biosynthesis of MC (Tillett et al. 2000). In addition, this disconnection has proven reliable in the former synthetic approach (Humphrey et al. 1996).
76 4. Manuscript II (Synthesis of MC-LF and derivates)
Figure 4-4: Retrosynthetic analysis of MC derivatives 1a-c The linear precursors 2a-c were synthesized in a convergent manner from fragments 3, 4 and 5a-c. Several syntheses of Boc-Adda-OH 3 have been reported (Bauer and Armstrong 1999;
Beatty et al. 1992; Clave et al. 2010; Cundy et al. 1999; Deiters and Schultz 2005; Panek and
Hu 1997; Pearson et al. 2000; Samy et al. 1999; Valentekovich and Schreiber 1995). We followed the route developed by Pearson et al. that delivers 3 in 13 steps with the best overall yield (Pearson et al. 2000). Fmoc groups served as N-terminal protection of fragments 4 and 5a- c. The presence of Fmoc, tert-butyl ester and the phenylselenocysteine moiety in fragments 4 and 5a-c required a C-terminal protection which can be cleaved in presence of these three groups and withstands the conditions of Fmoc deprotection. In our approach we used the phenacyl (Pac) protecting group which is removed under mild reductive conditions (Kokinaki et al. 2005).
Tetrapeptide 4 and dipeptides 5a-c were obtained from NMeSecPh derivative 6 (Riggs
Costerison 2002), -D-MeAsp derivative 7 and the building blocks 8a (Kokinaki et al. 2005),
8b, and 8c.
77 4. Manuscript II (Synthesis of MC-LF and derivates)
Figure 4-5: Synthesis of tetrapeptide fragment 4 HATU represents 1-[bis(dimethylamino)methylene]-1H-1,2,3- triazolo[4,5b]pyridinium 3-oxide hexafluorophosphate.
4.4.2 Synthesis of Tetrapeptide 4.
We started our synthesis with the preparation of tetrapeptide fragment 4 (Figure 4-5).
Boc-NMeSecPh-OH 6 was synthesized starting from Boc-N-methyl-L-serine 9 using the strategy developed by van der 0 and co-workers for the synthesis of Boc-SecPh-OH (Okeley et al. 2000) and coupled with dipeptide 11 to give 12 (Riggs Costerison 2002). Tripeptide 12 was then deprotected with TFA and coupled with Fmoc-D-Glu-Ot-Bu to give fragment 4.
78 4. Manuscript II (Synthesis of MC-LF and derivates)
Figure 4-6: Synthesis of Fmoc-D-MeAsp-Ot-Bu 7
4.4.3 Synthesis of Dipeptides 5a-c.
In position 3 of the vast majority of the MC erythro--methyl D-aspartic acid (D-MeAsp) is found. Up to date several synthetic routes towards derivatives of this amino acid have been published (Armstrong et al. 2007; Bauer and Armstrong 1999; Humphrey et al. 1996; Samy et al. 1999; Valentekovich and Schreiber 1995). However, all these published approaches feature methyl or ethyl ester protecting groups, which are crucial for the generation of the stereocenter at the -carbon atom. Since the literature-known building blocks cannot be easily transformed into 7, we synthesized Fmoc-D-MeAsp-Ot-Bu 7 starting from D-aspartic acid 13 as shown in
Figure 4-6. The amino group of 13 was benzylated by reductive amination and the -carboxy group was regioselectively esterified under acidic conditions to give benzyl ester 14. The free - carboxy group of 14 was tert-butylated and subsequently the secondary amine was protected with the phenylfluorenyl (PhFl) group to give 15. This sterically hindered group suppresses deprotonation at the -carbon and allows enolate formation by selective deprotonation at the - position of 15 in the subsequent step. Thus, treatment of 15 with lithium bis(trimethylsilyl)amide
(LHMDS) followed by addition of methyl iodide led to exclusive methylation of the -position to give erythro-16 and threo-16 in a ratio of 5:1. The stereoselectivity of this step can be explained by preferred formation of the (Z)-lithium enolate which adopts a hydrogen-in-plane
79 4. Manuscript II (Synthesis of MC-LF and derivates) conformation that is attacked opposite to the bulky nitrogen protecting groups (Humphrey et al.
1994). The isomers erythro-16 and threo-16 were readily separated by column chromatography and the configuration of erythro-16 was verified by complete deprotection and subsequent NMR analysis (Bochenska and Biernat 1972) (see Supporting Information) as well as optical rotation measurement (Armstrong et al. 2007). Hydrogenolysis of erythro-16 followed by Fmoc protection of the free amino group gave building block 7.
Figure 4-7: A) Synthesis of alkyne labelled building block 8c. B) Synthesis of dipeptide fragments 5a-cb bFor definitions of a-c, see Figure 4-4.
The phenacyl protected propargyl tyrosine building block Boc-Tyr(Prg)-OPac 8c (Prg = propargyl) was synthesized in three steps starting from Boc-Tyr-OH 17 (Figure 4-7A). Double propargylation of 17 followed by saponification of the propargyl ester gave 18 that was alkylated with phenacyl bromide yielding 8c in a high yield. 8c as well as literature known 8a (Kokinaki et al. 2005) and analogously prepared, isotopically labelled 8b were N-terminally deprotected and coupled with D-MeAsp derivative 7 to give dipeptides 5a-c (Figure 4-7B).
80 4. Manuscript II (Synthesis of MC-LF and derivates)
Figure 4-8: Synthesis of linear heptapeptides 22a-c by fragment couplingsb bFor definitions of a-c, see Figure 4-4
4.4.4 Synthesis of Linear Heptapeptides
With all fragments in hand, heptapeptides 2a-c were assembled (Figure 4-8). The C- terminal phenacyl ester of tetrapeptide 4 was reductively cleaved to give 19. Fmoc deprotection of dipeptides 5a-c was performed with diluted (5%) piperidine in DMF to avoid possible cleavage of the Pac ester giving 20a-c. The subsequent fragment coupling of 19 and a slight
81 4. Manuscript II (Synthesis of MC-LF and derivates) excess of 20a-c was performed with HATU/collidine leading to 21a-c. Under these conditions isomerization at the C-terminal Leu residue of fragment 19 is minimized as it was also reported by others (Carpino and El-Faham 1994; Humphrey et al. 1996). The hexapeptides 21a-c were
N-terminally deprotected and Boc-Adda-OH 3 was coupled to give heptapeptides 2a-c. Removal of the Pac group led to macrocyclization precursors 22a-c. We found that it was essential to purify compounds 22a-c by RP-HPLC to prevent by-product formation during the subsequent macrocyclization step.
Figure 4-9: Deprotection and macrocyclization
4.4.5 Macrocyclization and Final Steps
The C-termini of 22a-c were activated as pentafluorophenyl (Pfp) ester and the t-butyl esters along with the Boc group were removed by TFA treatment (Figure 4-9). Macrocyclization was induced under basic conditions applying a two-phase system of chloroform and phosphate buffer (pH = 9.5) to give cyclopeptides 23a-c which were purified by RP-HPLC. According to
HPLC approximately 20% (in case of 23a and 23b) and 10% (in case of 23c), respectively, of an isomeric product, possibly due to epimerization at C-terminal amino acid during
82 4. Manuscript II (Synthesis of MC-LF and derivates) macrocyclization, were observed. Subsequent selenoxide elimination under mild oxidative conditions gave the desired MC derivatives 1a-c. All final products were fully characterized by one- and two-dimensional NMR spectroscopy and HRMS. The analytical data of synthetic MC-
LF (1a) were identical to those of a commercial MC-LF sample confirming the structure of the synthetic material. In addition, co-injection of both a mixture of natural and synthetic 1a as well as a mixture of 1a and 1b led to single peaks in the LC-MS chromatograms (Supporting
Information).
Figure 4-10: PPP1 inhibition assay with natural MC-LF and synthetic compounds 1a, 1c, and 23c. (Values from three independent experiments).
Inhibition of Protein Phosphatase-1. In order to confirm that the biological activities of synthetic and natural MC-LF are identical, we determined the potency of both compounds to inhibit the hydrolysis of p-nitrophenyl phosphate catalysed by protein phosphatase-1 (PPP1,
Figure 4-10). As expected, the IC50 values obtained with this robust assay are very similar within the accuracy of this assay (synthetic 1a: IC50 = 870 pM, natural 1a: IC50 = 1.2 nM) further verifying the authenticity of the synthetic material. We also evaluated the inhibitory potency of the propargylated derivative 1c and its precursor 23c lacking the Michael acceptor Mdha. The similar IC50 value of 1c (IC50 = 1.7 nM) shows that the modification at position 4 has only a
83 4. Manuscript II (Synthesis of MC-LF and derivates) minor influence on phosphatase binding and inhibition as expected from the crystal structure of
PPP1 in complex with MC-LR (Goldberg et al. 1995). Phenylselenocystein derivative 23c, lacking the capability to covalently bind to PP1, has a similar inhibitory potency (IC50 = 2.1 nM).
Our results show that neither the modification at position 4 nor the one at position 7 of the scaffold disrupt PP1 inhibition. The fact that the absence of the Michael acceptor in compound
23c does not result in reduced inhibition potency is in line with previous findings that have shown that covalent binding of the toxin to PPP1 is a slow process and not required for phosphatase inhibition which is rather achieved by the initial fast non-covalent interaction. However, it is remarkable that the large side chain of NMeSecPh is tolerated during this binding process, thereby dramatically expanding the application domain of derivatized MC.
4.5 Conclusion
In summary, we developed a novel strategy for the synthesis of MC. Application of the
Fmoc/t-Bu protecting group strategy enabled the isomerization-free synthesis of MC-LF as well as a deuterated and an alkyne-labeled derivative thereof. The incorporation of a phenylselenocysteine moiety as precursor for the Mdha residue gave access to a novel potent and reversible phosphatase inhibitor that cannot undergo covalent binding to the protein. The synthesized MC derivatives can be broadly applied for improved MC detection and quantification and pave the way for future approaches to understanding the biological roles of
MC in various organisms.
84 4. Manuscript II (Synthesis of MC-LF and derivates)
4.6 Supplementary material
Due to spacial limitations the supplementary matrial (49 pages) to this article will not be shown here. But it can be accessed through the online repository of the Journal of Organic
Chemistry (https://pubs.acs.org/doi/suppl/10.1021/acs.joc.7b00175).
4.7 Acknowledgements
We thank Maximilian Häfner for his help during the synthesis of 2b and 21b and Prof.
A. Richard Chamberlin (UC Irvine, USA) for providing several synthetic intermediates for the synthesis of Boc-Adda-OH 3. S. A. and D. R. D. were supported by a grant from the Arthur and
Aenne Feindt Foundation, the Marsden Fund of the Royal Society of New Zealand (12-UOW-
087), and the Marie Curie International Research Staff Exchange Scheme Fellowship (PIRSES-
GA-2011-295223).
85 5. Manuscript III (UPLC-MS/MS detection of microcystins)
5. Manuscript III (UPLC-MS/MS detection of microcystins)
Simultaneous detection of 14 microcystin congeners from tissue
samples using UPLC- ESI-MS/MS and two different deuterated
synthetic microcystins as internal standards
Stefan Altaner1, Jonathan Puddick2, Valerie Fessard3, Daniel Feurstein1,4, Ivan
Zemskov5, Valentin Wittmann5 and Daniel R. Dietrich1*
1Human and Environmental Toxicology, University of Konstanz, Konstanz, Germany 2Cawthron Institute, Nelson, New Zealand 3Toxicology of Contaminants Unit, French Agency for Food, Environmental and Occupational Health and Safety, ANSES, 35300 Fougères, France 4Dr. Feurstein Medical Hemp GmbH, HANAFSAN, Hauptstr. 19A, 6840 Götzis, Austria; [email protected] 5Organic and Bioorganic Chemistry, University of Konstanz, Konstanz, Germany
Published in: Toxins, 2019, 11 (7) doi: 10.3390/toxins11070388
5.1 Abstract
Cyanobacterial microcystins (MCs), potent serine/threonine-phosphatase inhibitors, pose an increasing threat to humans. Current detection methods are optimised for water matrices with only a few MC congeners simultaneously detected. However, as MC congeners are known to differ in their toxicity, methods are needed that simultaneously quantify the congeners present, thus allowing for summary hazard and risk assessment. Moreover, detection of MCs should be expanded to complex matrices, e.g., blood and tissue samples, to verify in situ MC
86 5. Manuscript III (UPLC-MS/MS detection of microcystins) concentrations, thus providing for improved exposure assessment and hazard interpretation. To achieve this, we applied two synthetic deuterated MC standards and optimised the tissue extraction protocol for the simultaneous detection of 14 MC congeners in a single ultra performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) run. This procedure was validated using plasma and liver homogenates of mice (male and female) spiked with deuterated MC standards. For proof of concept, tissue and plasma samples from mice i.p. injected with MC-LR and MC-LF were analysed. While MC-LF was detected in all tissue samples of both sexes, detection of MC-LR was restricted to liver samples of male mice, suggesting different toxicokinetics in males, e.g., transport, conjugation or protein binding. Thus, deconjugation/-proteinisation steps should be employed to improve detection of bound MC.
5.2 Introduction
In recent years, reports of cyanobacterial blooms have increased (Huisman et al. 2018b).
Water bodies experiencing cyanobacterial bloom formation often additionally contain toxins, like microcystins (MCs), which can pose a serious health threat to humans. This was demonstrated in Caruaru in 1996, when 76 patients died because of the use of contaminated water for hemodialysis (Azevedo et al. 2002). MCs are cyclic heptapeptides, comprised of rare and unique l- and d-amino acids and two variable l-amino acid positions (X and Z, Figure 5-1), which are also used for nomenclature, e.g., MC-LR has l-leucine in the X position and l-arginine in the Z position. Along with the variable amino acids, methylations and demethylations at other positions have led to more than 200 congeners being identified to date (Spoof and Catherine
2017).
87 5. Manuscript III (UPLC-MS/MS detection of microcystins)
Figure 5-1: General structure of microcystins where X and Z are variable l-amino acid positions. β-d-MeAsp is erythro-β-d-methylaspartate; Mdha, N-methyldehydroalanine, Adda, (2S,3S,8S,9S,4E,6E)-3-amino-9-methoxy-2,6,8-trimethyl-10-phenyl-4,6-decadienoic acid.
Differential transport of MC congeners by organic anion transporting polypeptides
(OATPs) has been shown to result in different cell and tissue loads (Fischer et al. 2010; Fischer et al. 2005). Additionally, export of MC is still under research, although recent data suggest export of MC by MRP2 in an MC congener-dependent fashion as observed earlier for OATPs
(Kaur and Dietrich 2018; Kaur et al. 2019b). As the liver displays high expression of OATPs and MC exposure mainly occurs via the oral route, the liver is generally accepted to be the main target organ. In the target cells, MC is either freely available (unbound) or covalently bound to serine/threonine-protein phosphatases (PPPs), conjugated to glutathione (Kondo et al. 1992) or other cysteine-containing polypeptides and proteins. The irreversible inhibition of PPPs by MCs, and consequently the downstream protein-hyperphosphorylation, is currently assumed to be the predominant mechanism underlying MC cytotoxicity. Most members of the PPP-family are known to be affected by MCs (Swingle et al. 2007), albeit with different susceptibilities (Garibo et al. 2014a). Indeed, structural features of MC congeners appear to have a major impact on the
“tightness” of MC binding to the catalytic subunit of PPP (Fontanillo and Köhn 2018; Mattila et al. 2000b), thus resulting in the observed differing PPP inhibition capacities. Thus, information as to whether a specific MC congener was involved in a given toxicity observed, and at what tissue concentration adverse effects manifest, is crucial for a better understanding of MC
88 5. Manuscript III (UPLC-MS/MS detection of microcystins) congener differences in exerting overt toxicities and consequently for better delineating risk for human health.
MC detection is generally carried out via ELISA or chromatographic methods coupled to mass-spectrometric analysis (LC-MS) (Hawkins et al. 2005; Sanseverino et al. 2017). ELISA methods may differ in the antibodies used, and thus their capability to detect the majority of MC congeners present in a given sample. Indeed, while antibodies raised against the arginine epitope in MC-LR primarily recognise arginine containing MC congeners (Gilroy et al. 2000), antibodies raised against Adda potentially recognise 85% of the known congeners (Niedermeyer
2014). Similarly, the MS-based MMPB method which detects the oxidation product MMPB (2- methyl-3-methoxy-4-phenylbutyric acid) of Adda (Neffling et al. 2010a; Sano et al. 1992), therefore showing similar coverage (86%). As with the ELISA, only the sum of all MC and not specific MC congeners can be quantified. An advantage of the MS-based MMPB method, however, is that protein-bound MC can be detected as both bound-MC and free MC are oxidised
(Neffling et al. 2010a), albeit whether bound-MC and free MC oxidation occurs with the same efficiency is currently under debate (Greer et al. 2017).
UPLC-MS/MS-based methods provide the possibility of simultaneous detection of different congeners in various sample types (Diez-Quijada et al. 2018a; Greer et al. 2016; Mallet
2017; Manubolu et al. 2018; Turner et al. 2017; Zervou et al. 2017). However, quantitation from complex matrices like blood or tissue homogenate is difficult due to matrix residues which may influence the signal (suppression or enhancement), thus leading to under- or over-estimation of the true amount of MC present. This could be alleviated with internal standards (ISs) that behave similarly to the MC congeners of interest analysed. Unfortunately, appropriate internal standards are still missing. Quantitation from various samples using ISs has been attempted previously
(Mallet 2017; Roegner et al. 2014a; Smith and Boyer 2009). However, in these approaches, either nodularin or thiolised MC-LR/-RR at the Mdha moiety was employed. Obviously all three
ISs used in the latter approaches will differ in their behaviour during sample extraction, elution
89 5. Manuscript III (UPLC-MS/MS detection of microcystins) and analysis. In contrast, incorporation of stable isotopes is the best option for the generation of an IS. The latter ensures that during laboratory handling and analysis, a stable isotope-labelled compound has near-identical behaviour and is subject to the same matrix effects as the actual analyte in question.
Detection of analytes from tissue is often complicated due to the need for tissue extraction, and thus tissue specific matrix effects. Generally, extraction methods are optimised for a given sample type and for a single or a small number of MC congeners, primarily representing rather hydrophilic congeners, e.g., MC-RR, -LR and -YR. However, in view of the fact that several different MC congeners co-occur in a given cyanobacterial bloom (Puddick et al. 2014; Turner et al. 2018a), including the highly toxic MC-LF (Feurstein et al. 2011), the present work aimed to establish a quantitation method for a wider range of congeners in complex matrices allowing the analysis of congener specific organ distribution. Thus, deuterated internal standards (D7-MC-LR and D5-MC-LF) were de novo synthesised, and present protocols for MC sample extraction were optimised and integrated into a UPLC-MS/MS-based analytical procedure. For proof of concept, tissue and plasma samples from MC-LR and -LF exposed mice were analysed.
5.3 Results
5.3.1 Method establishment and optimization
MC-spiked serum samples were extracted using a previously published procedure for detection via ELISA (Heussner et al. 2014a) and subsequently analysed using a previously established UPLC-MS/MS method optimised for MC in cyanobacterial culture extracts (Altaner et al. 2017). The extraction method (Figure 5-2) consisted of three times protein precipitation
(PP), followed by three times liquid-liquid partitioning (LLP) with n-hexane and subsequent solid-phase extraction (SPE). In the process, signal enhancement was observed for all MC
90 5. Manuscript III (UPLC-MS/MS detection of microcystins) congeners tested (data not shown), presumably due to carry-over of the previous sample as signal enhancement progressively increased with every additional injection. Therefore, a slower gradient with a prolonged washout phase (see Section 5.4) was used during UPLC, leading to stable detection of the MC congeners tested (comparable areas under the curve for multiple subsequent analyses of the same MC congener).
Figure 5-2: Scheme of the extraction highlighting different time points for spiking with MC congeners.
To determine whether and during which extraction steps potential loss of MC congeners would occur, MC spiking was introduced at different points of the extraction procedure (Figure
5-2) and the respective losses, i.e., recoveries of spiked MC, determined at each step using a backward approach, i.e., starting at the injection stage (Figure 5-2). Indeed, loss of signal was not expected at the injection stage into the UPLC-MS/MS (Over-Spike). Thus, the contribution of each phase of the extraction procedure to the loss of analyte was assessed by starting with the
SPE columns and then working backwards to the PP.
Samples were spiked (Middle-Spike) with MC congeners prior to application to the SPE columns and the recoveries compared for three different SPE columns (Table 5-1 and Figure
S5-1): Phenomenex StrataX (3 cc, 200 mg sorbent), Waters Oasis HLB (6cc, 200 mg sorbent) and Waters Oasis PRiME HLB (6cc, 200 mg sorbent). Although in a preliminary experiment,
91 5. Manuscript III (UPLC-MS/MS detection of microcystins)
Supelco Hybrid-SPE Phospholipid and Phenomenex C18-E columns were also tested, these did not provide reliable data (data not shown), and therefore were not considered any further. The
StrataX column presented overall recoveries ranging between 63.4 and 112.7%, with the majority of congeners showing a recovery of 70%.
Table 5-1: Congener dependent recovery during SPE and LLP steps of the extraction. Solid phase extraction (SPE) Liquid-liquid partitioning (LLP) StrataX HLB PRiME HLB % in MeOH % in n-Hexane Mean SD Mean SD Mean SD Mean SD Mean SD MC-RR 63.4 2.3 90.4 6.0 100.9 5.8 89.0 9.9 0.6 0.0 MC-YR 112.7 2.2 81.2 5.6 80.6 9.8 96.4 4.6 1.0 0.1 MC-LR 79.5 2.5 95.1 5.3 97.8 7.4 98.5 4.8 0.5 0.0 MC-FR 66.8 12.4 97.6 6.9 107.2 7.2 97.8 7.0 0.6 0.0 MC-WR 74.9 2.2 98.5 6.6 114.7 7.9 97.8 8.5 0.5 0.0 MC-RA 87.7 5.5 104.8 6.6 98.0 4.3 97.3 4.2 0.6 0.1 MC-Raba 105.9 0.0 108.7 10.9 102.3 7.4 100.0 6.2 0.0 0.0 MC-LA 76.0 2.1 72.6 7.1 57.9 5.6 93.6 2.8 0.8 0.0 MC-FA 67.9 3.3 81.6 8.5 62.2 3.7 93.1 4.7 1.0 0.1 MC-WA 75.0 1.4 87.3 4.6 70.2 5.6 95.8 1.8 3.1 1.0 MC-LAba 69.2 5.8 86.5 7.1 68.1 4.0 n.d. n.d. n.d. n.d. MC-FAba 68.4 4.6 86.9 8.5 78.3 2.5 n.d. n.d. n.d. n.d. MC-WAba 71.9 3.0 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. MC-LF n.d. n.d. n.d. n.d. 77.4 4.7 n.d. n.d. n.d. n.d.
D7-MC-LR n.d. n.d. 97.8 6.6 92.6 5.9 n.d. n.d. n.d. n.d.
D5-MC-LF n.d. n.d. 94.8 6.7 79.1 3.9 n.d. n.d. n.d. n.d.
For the determination of the loss of the SPE procedure, MC congeners in MeOH solution were spiked to the sample after protein precipitation and liquid-liquid partitioning, but before the SPE procedure (Middle-Spike). Liquid-liquid partitioning was performed with MC-spiked MeOH, which was topped with hexane. n.d.: not determined in that particular experiment. N = 3. SD = standard deviation.
The PRiME HLB column showed acceptable recoveries (80.6–114.7%) for congeners containing arginine residues (Table 5-1 and Figure S5-1), while a pronounced loss of non-
92 5. Manuscript III (UPLC-MS/MS detection of microcystins) arginated congeners was observed (recoveries: 57.9–78.3%). In contrast, the HLB column presented better recovery for all MC congeners tested (72.6–108.7%), with recoveries for the majority of congeners ranging between 85 and 100%. A two-way ANOVA analysis of the latter data suggested that the recoveries determined were not significantly different from a control recovery of 100 ± 10% obtained with MC congeners in MeOH solution (Over-Spike) applied to the UPLC-MS/MS without sample matrix and prior SPE column separation. Consequently, the
HLB column was employed for all subsequent MC congener recovery and quantitation experiments.
Liquid-Liquid Partitioning (LLP) experiments, carried out according to a previously published methodology (Heussner et al. 2014a) to remove hydrophobic matrix elements, were conducted to determine MC congener loss during LLP. For this, an MC congener mix was spiked into methanol (MeOH), which was subsequently overlaid with n-hexane, as in the extraction procedure (Figure 5-2). After 30 min, the n-hexane phase was separated from the MeOH phase; both fractions were dried and re-dissolved in MeOH, thus allowing for the determination of the distribution of MC congeners in the n-hexane and MeOH phases. Nearly 100% of each MC congener was found in the MeOH phase (Table 5-1, Figure S5-2), thus indicating that no, or at best, negligible, loss would occur during LLP due to the n-hexane clean-up.
Finally, the loss during protein precipitation with MeOH or acetonitrile (ACN) was investigated. Protein precipitates, spiked with an MC congener mix (Full-Spike), were extracted once (1×) or twice (2×) with the respective solvent (MeOH or ACN). The comparison of MeOH and ACN did not show a significant difference (2-way-ANOVA) in recovery (Table 5-2 and
Figure S5-3). Neither was there a significant increase (2-way-ANOVA) in recovery when protein precipitates were extracted twice. Hence, all subsequent experiments were carried out with a single MeOH extraction of the protein precipitates. As neither the SPE nor the LLP resulted in a demonstrable loss of MC congeners (Table 5-1), the poor recovery observed when the complete extraction procedure was followed through (Table 5-2), i.e., including the initial protein
93 5. Manuscript III (UPLC-MS/MS detection of microcystins) precipitation step suggested that the protein precipitation itself was responsible for the high loss
(poor recovery) of MC congeners observed.
Table 5-2: Total recovery of the extraction using different protein precipitation (PP) procedures. Methanol Acetonitrile 1× 2× 1× 2× Mean SEM Mean SEM Mean SEM Mean SEM MC-RR 43.2 2.1 52.3 5.0 44.6 2.6 47.1 2.8 MC-YR 35.1 2.2 47.4 7.1 43.8 4.2 47.2 3.0 MC-LR 43.8 2.3 49.4 5.2 45.9 2.8 46.7 2.6 MC-FR 44.3 2.3 56.3 6.0 50.8 2.9 52.1 3.5 MC-WR 41.0 3.0 54.0 5.6 45.6 3.3 48.0 3.2 MC-RA 50.9 2.2 62.0 7.5 54.8 2.5 55.7 2.7 MC-Raba 52.9 3.7 62.5 7.3 54.5 2.3 59.9 2.4 MC-LA 49.2 2.4 35.2 3.5 54.2 2.7 47.7 3.6 MC-FA 47.5 2.3 41.8 4.4 55.2 2.4 50.7 1.8 MC-WA 45.7 2.6 49.0 3.8 48.7 2.5 50.2 3.7 MC-LAba 49.7 2.8 47.0 6.0 57.7 3.0 56.6 3.3 MC-FAba 50.9 2.2 52.9 6.0 59.4 2.5 64.7 3.6 MC-WAba 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 MC-LF 59.2 5.3 61.1 5.8 69.0 6.4 64.3 6.2
D7-MC-LR 48.0 2.2 51.9 4.1 51.0 2.6 51.0 2.9
D5-MC-LF 61.0 2.5 67.4 6.0 70.0 3.1 71.6 3.3
n.d.: not determined in that particular experiment. N = 2 (MeOH) and 3 (ACN). SEM = standard error of the mean.
5.3.2 Use of internal standard
D5-MC-LF and D7-MC-LR presented with similar total recovery as their respective non- deuterated analogues (t-test, p = 0.31 MC-LR vs. D7-MC-LR and p = 0.79 MC-LF vs. D5-MC-
LF) (Table 5-2). Thus, use of these synthetic congeners could serve as an IS with which the recovery of individual MC congeners could be corrected. Hence, human serum was spiked with the MC congener mix, extracted with the established procedure (see above) and analysed using
94 5. Manuscript III (UPLC-MS/MS detection of microcystins)
UPLC-MS/MS (Figure 5-3). Quantification of MC congeners without the IS for correction
(Figure 5-3A) yielded the same result as obtained before, i.e., low recovery, thus demonstrating the robustness of the extraction and analytical procedure. When D5-MC-LF was used to quantify all other MC congeners (Figure 5-3B), recovery for non-arginine MCs was improved from around 40% to 80%; however, most arginine-containing congeners were overestimated. For example, when 10 µg/mL of D5-MC-LF were spiked to the sample, the resulting LC-MS/MS peak value was set as 100%, i.e., as if 100% recovery was achieved, and thus a correction factor for extraction loss was applied. The latter meaning that for the MC congeners to be determined the actual loss experienced during the extraction procedure was assumed to be identical to the loss observed for the IS. Accordingly, the actual values of the MC congeners to be determined were then multiplied with the correction factor for extraction loss as determined with the IS. In contrast, when D7-MC-LR (Figure 5-3C) was employed as IS, the recovery of arginine- containing congeners increased to 80–110%, whereas non-arginine-containing congeners were limited to a recovery of approximately 60%. The best results, i.e., recovery of most MC congeners of 100 ± 20%, were achieved when using a combination of D5-MC-LF and D7-MC-
LR as IS (Figure 5-3D), whereby D5-MC-LF was employed for the correction of the recoveries of non-arginine-containing and D7-MC-LR for arginine-containing MC congeners, respectively.
95 5. Manuscript III (UPLC-MS/MS detection of microcystins)
Figure 5-3: Quantification of MC in spiked human serum using D5-MC-LF and D7- MC-LR as internal standards (IS). The same dataset was analysed using different IS for the quantification of the other MC
congeners. A: no internal standard specified, B: D5-MC-LF as IS, C: D7-MC-LR as IS,
D: D7-MC-LR as IS for all arginine-containing congeners and D5-MC-LF for non- arginine-containing congeners. n = 3, Two-way ANOVA with Bonferroni Post-test to test difference from 100±20%. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.
5.3.3 Method validation
The established extraction procedure and UPLC-MS/MS detection method using deuterated MC ISs, developed for human serum (see above), was further tested by spiking murine serum and liver homogenate with the MC congener mix (Table 5-3). MC recoveries from mouse serum were similar to those observed for human serum regarding total recovery (Figure
3D vs. Table 3). In contrast, the recoveries from the more complex liver homogenate matrix provided mixed results. While, generally, arginine-containing MC congeners were recovered with an equally good recovery as those observed in murine serum, the recovery for some of the
96 5. Manuscript III (UPLC-MS/MS detection of microcystins) non-arginine-containing congeners, with the exception of MC-LF, ranged between ⅓ and ½ of the recoveries found for the respective MC congener in murine serum (Table 5-3).
Table 5-3: Validation parameters of the established methods. Mouse recovery (% of expected Linear range LOD (ng/ml) LOQ (ng/ml) result) (ng/ml) Congener Mouse Serum Mouse liver Mouse serum Mouse serum Mouse serum MC-RR 74.2% 78.7% 0.5 2 2 - 500 MC-YR 88.5% 82.9% 0.5 2 2 - 500 MC-LR 113.1% 81.2% 0.1 2 2 - 500 MC-FR 124.3% 86.3% - - - MC-WR 108.6% 94.5% - - - MC-RA 111.2% 120.9% - - - MC-RAba 106.4% 109.6% - - - MC-LA 86.1% 64.5% - - - MC-FA 100.0% 43.4% - - - MC-WA 96.8% 34.1% - - - MC-LAba 94.3% 42.2% - - - MC-FAba 97.4% 54.4% - - - MC-WAba 89.8% < LOQ - - - MC-LF 83.5% 90.6% (0.5) (2) (2 - 500)
D7-MC-LR 100.0% 100.0% - - -
D5-MC-LF 100.0% 100.0% - -
MC-RR, MC-YR and MC-LR were used as external standards. MC-LF was also part of the external standard mix, but was not used to quantify any of the other MC congeners as it was no certified reference material but served as reference for more
hydrophobic congeners. D7-MC-LR and D7-MC-LR were used as IS in every sample quantified. Recovery was calculated using the external standards in MeOH. Therefore, the recovery as calculated here shows the influence of the matrix. MC congener mix spiked mouse plasma and liver homogenates from untreated male animals were used for this experiment. Linear regressions for the linearity determination all showed R2 > 0.99.
MC-RR, MC-YR and MC-LR were used as external standards, as recommended by the
European Medicines Agency (EMA) (EMA 2011), to determine the LOD, LOQ and linear range
97 5. Manuscript III (UPLC-MS/MS detection of microcystins)
for the UPLC-MS/MS detection method used. LOD was defined as the minimal amount of the
analyte which provided for a discernible peak, while LOQ was defined as the minimal amount
of the analyte which produced a peak at least three times higher than the background (signal
above noise) and has an accuracy <±20%. The LOD for each was slightly lower, i.e., 0.5 ng/mL
for MC-RR and -YR and 0.1 ng/mL for MC-LR. The upper limit of the linear range might even
be higher, but no higher concentrations were tested. The linear range was defined to lie between
the LOQ and the highest standard concentration used in the experiment. Linear range for the
three MC congeners was between 2 and 500 ng/mL, all with R2 > 0.99.
Table 5-4: Intra-day and inter-day precision. Day 1 Day 2 Both days RSD RSD Mean (ng/ml) SD (ng/ml) Mean (ng/ml) SD (ng/ml) RSD (%) (%) (%) MC-RR 193.0 10.0 5.2 208.3 44.7 21.5 5.4 MC-YR 31.1 3.4 10.8 32.7 6.5 19.9 3.5 MC-LR 502.7 22.8 4.5 682.9 44.0 6.4 21.5 MC-FR 271.9 14.5 5.3 383.0 43.5 11.4 24.0 MC-WR 81.6 16.1 19.7 127.3 16.9 13.3 31.0 MC-RA 73.5 9.4 12.8 75.8 13.5 17.8 2.2 MC-RAba 21.1 1.0 4.5 22.6 2.7 12.1 5.0 MC-LA 112.6 17.9 15.9 102.9 8.6 8.3 6.4 MC-FA 145.5 2.4 1.7 119.6 6.0 5.0 13.8 MC-WA 46.3 3.1 6.7 38.0 5.5 14.6 14.0 MC-LAba 89.0 10.2 11.5 87.4 6.5 7.4 1.3 MC-FAba 49.7 8.1 16.2 39.1 3.1 8.0 16.9 MC-WAba < LOQ n.d. n.d. < LOQ n.d. n.d. n.d. MC-LF < LOQ n.d. n.d. < LOQ n.d. n.d. n.d. D7-LR 10.0 0.0 0.0 10.0 0.0 0.0 0.0 D5-LF 10.0 0.0 0.0 10.0 0.0 0.0 0.0
Samples (n=3) have been measured on day 1. RSD is a measure for intraday precision. Identical samples were also measured on day 2 with slightly different RSD. The last column shows the RSD when mean of day 1 and day 2 are used.
98 5. Manuscript III (UPLC-MS/MS detection of microcystins)
To determine the precision of the UPLC-MS/MS detection method, intra-day and inter- day precision was determined. Identical samples were measured on two different days. Relative standard deviations (RSD) between 1.7 and 19.7% were observed for intra-day precision of day
1, although slightly higher similar RSDs (between 5.0 and 21.5%) were found on day 2. No obvious relationship between RSD and MC congener structure could be observed, i.e., precision for arginine-containing congeners and non-arginine-containing congeners was similar (Table 4).
When means of the individual days were compared (inter-day precision), RSD ranged from 1.3% to 31.0%, again with no difference in variability among arginine- and non-arginine containing congeners (Table 5-4).
5.3.4 MC levels in exposed mice
Liver and plasma samples of male and female Balb/c mice that were treated i.p. with 100
µg/kg bw MC-LR or MC-LF before sacrifice (24 h post-injection), were analysed using the extraction procedure and UPLC-MS/MS analytical method previously developed. While MC-
LF was quantifiable in plasma and liver tissue for both male and female mice (Figure 5-4), MC-
LR was detected only in the male liver tissue but not in female livers or the plasma samples of both sexes. It is to be noted that for both MC-LF and MC-LR, toxin levels in plasma and liver appeared to be lower in females than in males. The low number of animals available for this analysis prevented further investigation, i.e., corroboration of this observation. No MCs were observed in animals injected with 20 µg/kg bw of the toxins every second day for two weeks
(data not shown).
99 5. Manuscript III (UPLC-MS/MS detection of microcystins)
Figure 5-4: Analysis of MC levels in plasma and livers of exposed mice. Mice were injected with 100 µg/kg bw MC-LR or MC-LF. Values < LOD are shown as 0. Results are mean ± SD of two animals.
5.4 Discussion
The method described here demonstrated the simultaneous detection of 14 MC congeners in biological samples (serum, plasma, liver homogenate) using synthetic deuterated MC internal standards. The complete method presented acceptable analyte recovery, low LOD and LOQ, and limited intra- and inter-day variation (Table 3 and Table 4). In principle, this method is also amenable to the simultaneous detection and quantification of >14 MC congeners, albeit this would come at the expense of decreased sensitivity for individual MC congeners, as per time unit more parent masses would need to be analysed. The latter could be partly alleviated by narrowing the specific windows of analysis, as described in the Methods section and shown in
Table 5.
The quantification method presented here using two synthetic deuterated MCs as internal standards is, to the best of our knowledge, the first time this has been carried out for MCs.
Previous studies have either employed MC-LR with an attached thiol group at the Mdha residue as an internal standard for the HPLC-MS analysis of MC-LR, -RR and -LA (Smith and Boyer
2009), or similarly thiolised MC-LR and -RR in conjunction with MALDI-MS (Roegner et al.
100 5. Manuscript III (UPLC-MS/MS detection of microcystins)
2014a). The problem with the latter approach is that thiolised internal standards do not necessarily have the same behaviour during extraction and analysis as the parent compounds.
Moreover, these ISs are based on arginine-containing MC congeners, thus introducing a bias when measuring non-arginated MC congeners. In contrast, the ISs used here (D5-MC-LF and
D7-MC-LR) represent two major classes of MC, i.e., arginated and non-arginated congeners, and more importantly do not incorporate structural changes of the IS that would change behaviour during extraction and analysis. Other studies have suggested that heavy atoms are the best choice for labelling, because deuterium may change the analytical properties of the standards, e.g., retention time and hydrophobicity, possibly leading to altered recovery (Stokvis et al. 2005;
Wang et al. 2007). However, this was not observed during the present study with recovery
(Supplementary Figures S1 and S3) and retention times being indistinguishable from those observed for the respective naturally occurring (non-deuterated) MC congeners.
Although the new procedure for quantifying MC congeners in biological tissue samples is promising, it must be noted that the analysis is still restricted to free and unconjugated MC congeners, as it uses the parent ion mass of each congener analysed. It has been suggested that free MC only comprises a minor portion of the total MC load after exposure (Azevedo et al.
2002; Carmichael et al. 2001), so one would assume that the actual MC load is severely underestimated (Greer et al. 2017). Indeed, MC-LR was nearly undetectable in liver and plasma of mice injected i.p., with 100 µg/kg bw of MC-LR (Figure 4). While the lack of MC-LR detection in plasma samples could be interpreted to be the result of rapid uptake via the mOatp1b2 (Feurstein et al. 2009; Lu et al. 2008), lack of MC-LR detection in the liver samples could suggest either covalent protein binding or rapid excretion, and thus elimination from the liver. Indeed, as MC-LF was detectable in the liver samples and the covalent interaction of MC-
LF and MC-LR are assumed to be comparable, it is currently undiscernible whether the differences of liver tissue levels stem from differences in kinetics (uptake and excretion of MC congeners), dynamics (covalent binding) or indeed difference in extraction of free and protein
101 5. Manuscript III (UPLC-MS/MS detection of microcystins) bound MC. Recently, methods have been developed that are capable of cleaving the thiol bond formed during conjugation and thus release free MC using basic conditions (Miles 2017; Miles et al. 2016a; Zemskov et al. 2016). Moreover, the regioselective cleavage of the thiol bond in combination with the application of our IS to future in vivo or in vitro experiments would allow to quantify the amount of free, conjugated and protein-bound single MC congeners or MC congener mixtures.
ESI-MS/MS methods are known to be influenced by matrix components left in the sample subsequent to extraction (matrix-effects) (Trufelli et al. 2011). Although ISs were employed, the matrix still influenced the signal of the individual MC congeners. It is interesting to note that the liver matrix appeared to influence the recovery of the non-arginated MC congeners MC-LA, MC-FA, MC-WA, MC-LAba, MC-FAba, and MC-WAba (34.1–64.5% recovery) more profoundly than other non-arginated (MC-LF, 90.6%) or arginated MC congeners (78.7–120.9% recovery) (Table 3).
Ideally, method validation would be carried out with human and murine serum and plasma, albeit no marked difference was observed between the application of the method on human or mouse serum (Figure 3D vs. Table 3). Nevertheless, previous studies suggested stronger binding of MC congeners to human than to murine or other mammalian albumin (Cheng et al. 2006; Robinson et al. 1991), thus the recoveries obtained for 14 MC congeners in human serum (Figure 3) with those observed in murine plasma (Figure 4, Table 3) would suggest that the differences in albumin binding had an impact. In consequence, method validation would have to be carried out for every sample type prior to broad data analysis.
As method validation was carried out in murine plasma, more trust is placed in the results of the plasma analyses of the mouse experiments in vivo than in those in the liver homogenates.
Irrespective of the latter, MC could not be detected in plasma or liver tissue of mice which were injected with 20 µg/kg bw MC-LR or MC-LF every second day for 2 weeks, while mice exposed to 100 µg/kg bw MC-LR and LF showed detectable MC levels both in plasma and liver tissue
102 5. Manuscript III (UPLC-MS/MS detection of microcystins)
(Figure 4). Studies in humans (Chen et al. 2009; Zheng et al. 2017) observed serum levels of around 0.2–0.6 ng/mL in humans consuming contaminated food items and drinking water. Here, we could detect MCs in mouse plasma at around 10-15 ng/mL after injection of a single dose
100 µg/kg bw. This shows, that the detection limit is biologically relevant and that the range in which the presented method is linear fits to actually observed levels. MCs are known to be efficiently conjugated to glutathione (Pflugmacher et al. 1998a) and are subsequently rapidly excreted (Li et al. 2018; Wang et al. 2008), most likely via the bile into the small intestine. Thus, it may not be surprising that application of small concentrations of MCs every two days for 2 weeks resulted in non-detectable free MC in the plasma and liver of exposed mice. If a regioselective cleavage of the thiol bond had been applied to the plasma and liver homogenates prior to the MC extraction procedure, protein-bound MC would most likely have been detected.
Indeed, a recent study in pigs showed that oral consumption via drinking water over several weeks at low MC-LR doses (2 µg/kg bw) did not lead to any detectable free microcystins in plasma and livers, similar to our study. However protein-bound MC-LR could be detected in the livers at around 20 ng/mg using Lemieux oxidation (Greer et al. 2018). Thus, the implementation of regioselective cleavage to assess the fraction of bound MCs is essential, especially as it was found that around 85% of administered MC-LR amounts are found in a bound state in fish tissue
(Greer et al. 2017).
The fact that MC-LF was readily detectable in plasma and liver homogenates of male and female mice, whereas this was not the case for MC-LR, could suggest that either the murine
Oatp1b2 expressed in the liver is more proficient in taking up MC-LR than MC-LF or that murine hepatocytes conjugate and excrete MC-LR faster than MC-LF. Differences in binding affinity of MC-LF and-LR for the murine organic anion polypeptide transporter mOatp1b2 have been observed (Feurstein et al. 2009; Lu et al. 2008). Interestingly, females tended to show lower detectable amounts of MC-LF in both matrices than males, an observation that could stem from a sex-dependent expression levels of Oatp1b2 which is hormonally regulated, as observed for
103 5. Manuscript III (UPLC-MS/MS detection of microcystins) other murine Oatps (Cheng et al. 2006; Fu et al. 2019). However, the latter findings need to be taken with caution, as only two replicates were used. Nevertheless, similar quantities of MC-LR, based on wet weight, were found in livers of the mice in our study when compared to the livers of rats used in the study of Wang et al. (2008).
In summary, we established a procedure capable of simultaneously quantifying 14 MC congeners in plasma/serum and liver tissue samples. For the first time, de novo synthesised deuterated MC internal standards were used for quantification and resulted in a highly improved recovery of MC congeners. The method was validated using a matrix of plasma and was applied to the detection of administered MC in mouse plasma and liver samples. In the future, the regioselective cleavage of the thiol bond in combination with the application of our deuterated
ISs should allow for the quantification of free, conjugated (e.g., glutathione) and protein-bound single MC congeners or MC congener mixtures.
5.5 Material and methods
5.5.1 Materials
Human serum (off the clot) was obtained from Biochrom (Berlin, Germany) or from the
New Zealand Blood Service. Water was deionized water with 18.2 MΩ (Milli-Q). Formic acid
(Merck, Darmstadt, Germany), acetonitrile (Carl Roth, Karlsruhe, Germany) and solvents were of MS grade. Analytical standards for MC-RR, MC-LR and MC-YR were from DHI LAB products (Hørsholm, Denmark). SPE columns were either from Waters (Oasis HLB and PRiME
HLB, Eschborn, Germany) or from Phenomenex (StrataX, Auckland, New Zealand). They were used with a vacuum manifold by Macherey-Nagel (Düren, Germany).
For UPLC-MS/MS analytics, an internal standard was used which comprised synthetic, deuterated MC-LR and MC-LF. The synthesis of D5-MC-LF (containing phenylalanine with a deuterated phenyl group (five D-atoms) in position 4) was published previously (Zemskov et al.
104 5. Manuscript III (UPLC-MS/MS detection of microcystins)
2017). D7-MC-LR was synthesized in analogy to the published synthesis of D5-MC-LF employing leucine deuterated in the side chain (seven D-atoms) at position 2. These two standards were either used for spiking mouse plasma and liver samples (see below), or supplementing (100 ng/ml each) a mixture of MCs (MC-mix) extracted from Microcystis
CAWBG11 cells as published previously (Altaner et al. 2017; Puddick et al. 2014). The extract contained multiple microcystin congeners, among them MC-RR, -YR, -LR, -FR, - WR, -RA, -
RAba, -LA, -FA, -WA, -LAba, -FAba, -WAba and was additionally supplemented with each
100 ng/ml MC-LF. The so produced stock solution (MC congener mix) was used for spiking into samples at a 1:10 ratio to establish the extraction and detection method. MC concentrations ranged between 94 ng/mL (MC-WAba) and 2768 ng/mL (MC-LR) in the stock solution.
5.5.2 Sample generation for the establishment and validation
All liquid samples were handled in glass LC-vials to reduce loss due to adsorption, as previously recommended (Altaner et al. 2017; Heussner et al. 2014a).
Serum/plasma samples for method establishment (0.5 ml) were spiked with MC congener mix stock at a 1:10 ratio (MC mix:serum, vol/vol). A 1:10 dilution of the MC mix in methanol, serving as control for every experiment, was analysed without further treatment (Recovery control). Spiking was performed at different steps during the procedure (Figure 5-2); before extraction (Full-Spike), after extraction but before UPLC-MS/MS analysis (Over-Spike) or directly before the SPE step (Middle-Spike). These samples were used for generating the data in
Table 5-1, Table 5-2 and Figure 5-3.
Additional establishment and validation experiments were performed with plasma and livers of non-treated Balb/c mice obtained from the animal facility of the University of Konstanz.
Serum/plasma (0.25 ml) was used as is, while liver samples were homogenized prior to spiking.
For homogenization, livers were thawed on ice, pieces taken, weighed, and 200% ice-cold RIPA buffer was added (e.g. 125 mg liver + 250 µl RIPA). The sample was then homogenized using
105 5. Manuscript III (UPLC-MS/MS detection of microcystins) an electric drill with attached pestle while on ice. The final liver homogenate (250 µl) was used for spiking and subsequent toxin extraction, resulting in 83.3 mg of tissue used for one sample.
Spiking was carried out with the MC-congener mix at a ratio of 1:10 (vol/vol). These samples were used during the method validation (Table 5-3).
5.5.3 Extraction method for MC from blood and liver tissue samples
Samples (250 µl liver homogenates or plasma/serum) were transferred to glass reaction tubes and subjected to protein precipitation by adding three volumes 100% MeOH. Precipitates were centrifuged for 40 min at 4°C with 3023 rcf. Subsequently, supernatants were transferred to a clean glass tubes. Removal of lipophilic compounds from the supernatant was conducted using n-hexane (4 ml). Samples were vortexed, phases were allowed to separate for 30 min before the n-hexane was removed and discarded. The sample was diluted with MilliQ (52 volumes) to reduce the organic solvent content below 10% (v/v) in the final sample. Afterwards, samples were subjected to solid-phase extraction (SPE) clean-up using Waters Oasis HLB (6 cc,
200 mg sorbent). Columns were connected to a vacuum manifold, activated with 100% MeOH
(5 ml) and equilibrated with 10% MeOH (5 ml). Then sample was loaded, washed with MilliQ
(7 ml) and 20% MeOH (7 ml) before elution with 80% MeOH (5 ml). All SPE steps were performed at a maximum vacuum of 20 kPa. SPE eluents were dried with a SpeedVac before reconstitution in 0.25 or 0.5 ml pure MeOH and stored at -20°C until analysis. Samples of the animal study had to be concentrated further. In the latter case, samples were again dried using the SpeedVac and reconstituted in 12.5 µl pure MeOH.
5.5.4 UPLC-MS/MS analysis
Analyses were performed on a Waters Acquity H-class liquid chromatograph equipped with an Acquity BEH C18 column (1.6 µm, 2.1 × 50 mm) with a corresponding guard column kept at 40°C, coupled to a Waters XEVO TQ-S mass spectrometer. The used solvents were 10%
106 5. Manuscript III (UPLC-MS/MS detection of microcystins)
ACN (solvent A) and 90% ACN (solvent B) both containing 100 mM CH2O2 and 6 mM NH3 with a total flow of 0.4 ml/min. The used gradient started with 25% B, held for 30s at 25% B before increasing to 45% B over additional 30s. Over the following 180 s amount of solvent B was increased to 60% before being raised to 99% over 12 s where it was held for additional 30s.
Reequilibration back to 25% B was done over 78 s where it was held for another 60 s before the next sample injection. Injection volume was 5 µL. Compounds were ionized using a capillary voltage of 3 kV and a nebulizer pressure of 7.0 bar. Dissolution was achieved using a nitrogen flow of 1000 L/h at 500 °C. Analysis of the congeners was divided into five analysis windows, to reduce the number of parallel analysed congeners in order to maximise the time spent scanning for each individual compound. Analysis parameters for all MC congeners are shown in Table
5-5.
MC-RR, MC-YR and MC-LR were used as external standards for each analytical run.
Standards were used at three final concentration levels: 2, 10 and 100 ng/ml and employed for establishing a linear regression of the signal response with the injected amount. In the first experiments, these standards were used in MeOH, during method validation standards were used in blank extracted matrix. The external standard MC congener mix also contained MC-LF, which was not from a source classified as reference material. Therefore, it was not used to quantify any
MC levels, but only as point of comparison to the others standards during the validation. D7-
MC-LR as IS for all arginine-containing congeners and D5-MC-LF for non-arginine-containing congeners were either spiked into samples individually before extraction (animal experiment) or were part of the MC-mix used for establishment and validation.
107 5. Manuscript III (UPLC-MS/MS detection of microcystins)
Table 5-5: MS Parameters Analysis Parent Cone Collision Daughter External Internal Congener window mass Dwell time (s) voltage Energy mass (m/z) standard standard (min) (m/z) (V) (V)
MC-RR 0.3 – 1.6 519.7 135.1 0.027 40 27 MC-RR D7-MC-LR
MC-YR 0.3 – 1.6 1045.6 135.1 0.027 40 70 MC-YR D7-MC-LR
MC-LR 0.3 – 1.6 995.6 135.1 0.027 40 65 MC-LR D7-MC-LR
MC-FR 0.3 – 1.6 1029.6 135.1 0.027 40 65 MC-LR D7-MC-LR
MC-WR 0.3 – 1.6 1086.6 135.1 0.027 40 65 MC-LR D7-MC-LR
MC-RA 1.3 – 2.5 953.6 135.1 0.024 40 65 MC-LR D7-MC-LR
MC-Raba 1.3 – 2.5 967.6 135.1 0.024 40 65 MC-LR D7-MC-LR
MC-LA 1.3 – 2.5 910.6 135.1 0.024 40 65 MC-LR D7-MC-LF
MC-FA 1.3 – 2.5 944.6 135.1 0.024 40 65 MC-LR D7-MC-LF
MC-WA 1.3 – 2.5 983.6 135.1 0.024 40 65 MC-LR D5-MC-LF
MC-LAba 1.7 – 3.5 924.6 135.1 0.024 40 65 MC-LR D5-MC-LF
MC-FAba 1.7 – 3.5 958.6 135.1 0.024 40 65 MC-LR D5-MC-LF
MC-WAba 1.7 – 3.5 997.6 135.1 0.024 40 65 MC-LR D5-MC-LF
MC-LF 1.7 – 3.5 986.6 135.1 0.024 40 65 MC-LR D5-MC-LF
D7-MC-LR 1.0 – 1.8 1002.7 135.1 0.024 40 65 MC-LR -
D5-MC-LF 2.3 – 3.5 991.6 135.1 0.024 40 65 MC-LR -
The columns for internal and external standard denominate the congeners, which are used for the quantification of the respective conger during the UPLC-MS/MS analysis. Dwell times differ as the individual analytical windows do not contain the same numbers of congeners.
5.5.5 Animal samples
Balb/c mice, obtained from Janvier (Le Genest-Saint-Isle, France), were injected i.p. with
either MC-LR, MC-LF (20 µg/kg bw) or vehicle (water) every second day for 14 days, before
sacrifice of the animals and organ harvest. Additional mice were injected once with 100 µg/kg
bw MC-LF or MC-LR. Mice were sacrificed 24h after last injection. Blood was taken in
heparinized tubes, centrifuged and the remaining plasma was snap-frozen in liquid nitrogen.
Liver tissue was also snap-frozen in liquid nitrogen. Samples were stored at -80°C until further
preparation. Mice were housed in plastic cages, in an air-conditioned room (20–23°C) under a
108 5. Manuscript III (UPLC-MS/MS detection of microcystins)
12/12 h light/dark cycle with free access to rodent pellets and tap water. The animals were acclimatised at least 7 days prior to experimentation. Animal handling, exposure and organ removal was performed in the laboratory of ANSES (Agence nationale de sécurité sanitaire de l’alimentation, de l’environnement et du travail) Fougères in 2008 according to approved protocols by our Institute’s ethical committee on animal experimentation.
For spiking experiments, plasma samples were used as is, while livers were homogenized in analogy to the validation samples (see above). Every sample was spiked with 10 - 50 ng/ml
D7-MC-LR and D5-MC-LF, as IS, and samples stored at -20°C until extraction.
5.5.6 Data analyses and statistics
LC-MS/MS data was analysed using TargetLynx 4.1 integrated into MassLynx 4.1
(Waters). Data handling was performed with Microsoft excel, while GraphPad Prism 5 was used for statistical analysis and data visualization.
109 5. Manuscript III (UPLC-MS/MS detection of microcystins)
5.6 Supplementary material
Figure S5-1: Column comparison using a middle-spike in serum. Blank serum was extracted with the PP and LLP, then spiked with QC mix before SPE. Interfering residue seems lowest on HLB columns, higher with StrataX and Prime HLB. n.d.: not determined in that specific experiment. Here, human plasma was used. Two-way ANOVA with Bonferroni post-test was used to calculate significant difference from “artificial” 100 % control. **, p < 0.01; ***, p < 0.001; ****, p < 0.0001.
120 Methanol fraction Hexane fraction 100 **
80
60
40
20 Toxin (%Toxin control) of *** *** *** *** *** *** *** *** *** *** 0
MC-RRMC-YRMC-LRMC-FRMC-WRMC-RA MC-LAMC-FAMC-WA MC-RAba Figure S5-2: Distribution of microcystins in methanol and hexane fractions. Microcystins were spiked into methanol, hexane phase added, mixed, phases allowed to separate and, both phases dried. Microcystins were then reconstituted in MeOH and measured. Two-way ANOVA with Bonferroni post-test was used to calculate significant difference from “artificial” 100 % control. **, p < 0.01; ***, p < 0.001
110 5. Manuscript III (UPLC-MS/MS detection of microcystins)
Figure S5-3: Use of acetonitrile and methanol for protein precipitation. Serum was spiked before extraction. The protein precipitation step was either performed with methanol or acetonitrile for one or two times, SPE was performed using the Oasis HLB columns. No difference in recovery was observed, therefore the easiest option was taken for further experiments: single extraction with methanol. Human serum (off the clot was used).
5.7 Acknowledgements
Thanks to Jutta Fastner for the opportunity to gather very first MS/MS experience (S.A.) at the labs of the Umweltbundesamt (Berlin, Germany). Also, thanks to Michael Boundy
(Cawthron Institute) for helpful insights into MS-troubleshooting. We gratefully acknowledge
Tabea Zubel and Aswin Mangerich (Molecular Toxicology, University of Konstanz) for the possibility to use the UPLC-MS/MS instrument during most experiments.
111 6. Manuscript IV (In-silico prediction of microcystin toxicity)
6. Manuscript IV (In-silico prediction of microcystin toxicity)
Machine learning prediction of cyanobacterial toxin (microcystin)
toxicodynamics in humans
1* 2* 1 3 3 Stefan Altaner , Sabrina Jaeger , Regina Fotler , Ivan Zemskov , Valentin Wittman ,
Falk Schreiber2,4 and Daniel R. Dietrich1
1Human and Environmental Toxicology, University of Konstanz, Konstanz, Germany 2Life Science Informatics, University of Konstanz, Konstanz, Germany 3Organic and Bioorganic Chemistry, University of Konstanz, Konstanz, Germany 4Faculty of IT, Monash University, Melbourne, Australia *equal contribution
Published in: ALTEX, 2019, (online ahead of print) doi: 10.14573/altex.1904031
6.1 Abstract
Microcystins (MC) represent a family of cyclic peptides with approx. 250 congeners presumed harmful to human health due to their ability to inhibit ser/thr-proteinphosphatases
(PPP), albeit all hazard and risk assessments (RA) are based on data of one MC-congener (MC-
LR) only. MC congener structural diversity is a challenge for the risk assessment of these toxins, especially as several different PPPs have to be included in the RA. Consequently, the inhibition of PPP1, PPP2A and PPP5 was determined with 18 structurally different MC and demonstrated
MC congener dependent inhibition activity and a lower susceptibility of PPP5 to inhibition than
PPP1 and PPP2A. The latter data were employed to train a machine learning algorithm that should allow prediction of PPP inhibition (toxicity) based on MCs 2D chemical structure. IC50
112 6. Manuscript IV (In-silico prediction of microcystin toxicity) values were classified in toxicity classes and three machine learning models were used to predict the toxicity class, resulting in 80-90% correct predictions.
6.2 Introduction
Harmful (toxic) cyanobacterial blooms have become an important concern with regard to drinking water quality and safety. Some prominent examples of the latter are a bloom that affected almost 1000 km of the Barwon-Darling River, New South Wales, Australia, in
November and December 1991 (Bowling and Baker 1996), the deaths of renal dialysis patients in 1996 in Caruaru, Brazil (Azevedo et al. 2002), or the most recent closing of the drinking water supply for the inhabitants of Toledo, Ohio, USA, resulting from recurrent Microcystis aeruginosa blooms in Lake Erie (Berry et al. 2017). Of concern is the impression that cyanobacterial blooms in surface waters appear to be increasing with climate change (Huisman et al. 2018b). Of importance in conjunction with toxic cyanobacterial blooms is that several different toxins and congeners of a given toxin (e.g. microcystins (MC)) can co-occur in a given bloom (Dietrich and Hoeger 2005), toxin concentrations per cyanobacterial cell can change > ten-fold within a short time span of a bloom event (Wood et al. 2011), and that increased temperature may provide a growth advantage for toxin producing species (Kleinteich et al.
2012).
Microcystins, produced by several cyanobacteria species e.g. Microcystis spp.,
Dolichospermum spp. or Planktothrix spp. in water bodies worldwide (Preece et al. 2017), appear to be one of the toxins most frequently associated with drinking water, food supplement and/or food contamination and have resulted in human health morbidities and mortalities.
Structurally, MC are cyclic heptapeptides consisting of common L-amino acids, but also uncommon and unique amino acids. Their general structure is cyclo([D-Ala1]-[L-X2]-[β-D-
MeAsp3]-[L-Z4]-[Adda5]- -D-Glu6]-[Mdha7]). X and Z stand for variable L-amino acids, while β-D-MeAsp is erythro-β-D-methylaspartate, ADDA is (2S,3S,8S,9S,4E,6E)-3-amino-9-
113 6. Manuscript IV (In-silico prediction of microcystin toxicity) methoxy-2,6,8-trimethyl-10-phenyl-4,6-decadienoic acid and Mdha is N-methyldehydroalanine.
The variable positions, along with various (de)methylation sites (Figure 6-1 and Table S6-1), provide for currently 248 known MC congeners (Spoof and Catherine 2017), albeit new MC congeners are continuously discovered. However, contrary to recent stipulations (Huisman et al.
2018b), the toxicity is known but for a very few of the 248 MC congeners.
Figure 6-1: Consensus structure of microcystins and the synthetic variations produced for this study. The dashed lines represent the single amino acids of the heptapeptide structure. Further details can be found in Table S6-1. Amba = (2S,3S)-3-amino-2-methylbutanoic acid, Anda = (2S,3S,4E,6E)-3-amino-2-methylnona-4,6-dienoic acid, Dhb = (E)-2-amino-2- butenoic acid, MSecPh = N-methyl-Se-phenyl-L-selenocysteine, Prg = propargyl
Indeed, the World Health Organization (WHO) provisional guideline value of 1 µg/L for the risk assessment of MC in drinking water (World Health Organization 2017), is based entirely on the toxicological data of MC-LR and the assumption that MC-LR is the most toxic of the MC congeners known. This WHO guideline value heavily relies on a 90-day toxicity study in mice
114 6. Manuscript IV (In-silico prediction of microcystin toxicity)
(Fawell et al. 1999). Toxicity can be split into two critical components: toxicokinetics (cellular uptake, distribution, metabolism and elimination) and toxicodynamics (the interaction with cellular molecules resulting in an observed adverse outcome) (Dellafiora et al. 2018; EFSA PPR
Panel et al. 2018). Cellular uptake of MC is primarily governed by organic anion polypeptide
(OATP) and thus by the type and level of OATP expression in a given cell, as well as by the affinity and capacity of the respective OATP for transporting the different MC congeners
(Fischer et al. 2010). Cellular export of MC (conjugates) is still under debate, as involved exporters have so far not been unambiguously determined. However, the comparison of rodents
(mouse and rat) with humans demonstrated that rodents are poor surrogates for humans specifically with regard to type of OATP expressed in the various tissues and the affinity and capacity of expressed OATPs for specific MC congener transport (Feurstein et al. 2011). The fact that humans demonstrated major differences in OATP expression and thus susceptibility to
MC (Fischer et al. 2010) only compounded the fact that current risk assessment premises, based on surrogate species and one single MC congener e.g. the WHO guideline value, could severely underestimate the potential toxicities of MC due to their congener specific kinetics.
Similarly, questions arose with regard to the toxicodynamics of MC congeners. Indeed,
MC are very potent inhibitors of the catalytic subunits of ser/thr-protein phosphatases (PPP), albeit MC congeners differ with regard to their PPP inhibition cabability (Hastie et al. 2005;
Hoeger et al. 2007; Mackintosh et al. 1990). The family of PPPs in humans has seven members
(PPP1, PPP2A, PPP2B (Calcineurin), PPP4, PPP5, PPP6 and PPP7), whose catalytic subunits are structurally similar (Shi 2009), display protein sequence homology of up to 65% (checked with Clustal Omega), and have defined substrate specificities and therefore various functions
(Pereira et al. 2011b). Most of the PPPs are expressed ubiquitously albeit at different levels in the various organs, while PPP7 is specific to retina and brain (Cohen 2004). Inhibition of PPPs by MC occur prima facie via reversible followed by covalent binding of MC to the catalytic subunit of the respective PPP (Mackintosh et al. 1990).
115 6. Manuscript IV (In-silico prediction of microcystin toxicity)
Dysregulated phospho-protein homeostasis, subsequent to PPP inhibition, thus hyperphosphorylation of numerous phosphate-regulated enzymes and the deregulation of fundamental cellular processes, e.g. disruption of the cytoskeleton, represents the toxicodynamic process. As PPP differ in their susceptibility to inhibition by MC (Hoeger et al. 2007) and congeners differ in their capacity to inhibit specific PPP (Hoeger et al. 2007), the observed toxicity as manifested in the respective organs (liver, brain, kidney) are not only the result of
MC toxicokinetics but also of toxicodynamics (Fischer et al. 2010). However, to date research on MC toxicodynamics primarily focused on the interaction of MC congeners with PPP1 and
PPP2A, whereby only a select few MC congeners (predominantly MC-LR, -RR, -LA, and -LF) were tested (Garibo et al. 2014a; Hoeger et al. 2007).
In view of the ever increasing number of MC congeners identified (Spoof and Catherine
2017), yet lacking availability to synthetize these in sufficient purity and amount for in vitro, nor in vivo, testing, an in silico approach using toxicodynamic data could provide for a first step towards a better “toxicity assessment with relevance for humans” of uncharacterized MCs.
However, as there are yet insufficiently robust testing systems to address the toxicokinetic component of toxicity, only the toxicodynamic component, i.e. PPP inhibition, was addressed in the work presented. Indeed, although a number of in vitro models have been put forward that allow studying the uptake of MC congeners (Feurstein et al. 2009; Feurstein et al. 2010;
Feurstein et al. 2011; Fischer et al. 2010; Monks et al. 2007), studying the efflux from cells is much more difficult as intracellular MC would kill the cells before a proper efflux model could be established. Thus, despite recent advances in studying MC efflux using in vitro membrane vesicle approaches (Kaur et al. 2019a), complete kinetic models encompassing influx and efflux kinetics of MCs have yet been impossible to establish. Accordingly, the aim of this study was to develop a comprehensive dataset of toxicodynamics i.e. the PPP inhibitory capacities of a limited number of MC congeners. These in vitro data were then used as a comparative basis driving an in silico approach using machine learning (ML). Thus, the PPP inhibition capacity (toxicity) of
116 6. Manuscript IV (In-silico prediction of microcystin toxicity)
18 structurally diverse MC congeners was determined using ser/thr-PPP (PPP1, PPP2A and
PPP5) in a colorimetric protein phosphatase inhibition assay. For the latter, a number of synthetic
MC derivatives were generated according to previously published procedures (Fontanillo et al.
2016; Zemskov et al. 2017). Among these was a variant with modified stereochemistry at the
Adda5 residue (i.e., the enantiomer of Adda was used) ([enantio-Adda5]-MC-LF) as well as variants with simplified residues at the Adda5 position ([Anda5]-MC-LY(Prg) and [Amba5]-
MC-LY(Prg)). The modified amino acids in these synthetic MCs in positions other than X2 and
Z4 are indicated by adding the modification in square brackets before the name of the MC-XZ derivative. MC-LY(Prg) denotes variants with L-leucine in position X2 and O-propargylated L- tyrosine in position Z4 (Figure 6-1 and Table S6-1). Results were classified into three categories
(toxic, less toxic, non-toxic) and the toxicity predicted based on chemical structure via the ML approach described below.
Machine learning is widely used in the field of bioinformatics to predict bioactivity or molecular properties (e.g. solubility) of compounds, protein folding, etc. Despite the latter advances, it is still difficult to employ ML in pharmacology or toxicology, as datasets are often smaller and more heterogeneous compared to datasets from other domains (Wu et al. 2018).
Indeed, although ML has been employed for the prediction of cyanobacterial blooms based on satellite data (Chang et al. 2014) or the production of toxins based on environmental factors
(Taranu et al. 2017), ML has so far not been used to predict the toxicity of MC congeners. To encode molecules or proteins for ML, a fixed size numerical vector is needed (Wu et al. 2018).
One approach to encode molecules and proteins is Mol2vec (Jaeger et al. 2018) and ProtVec
(Asgari and Mofrad 2015), respectively, which are inspired by natural language processing. Both approaches are based on the Word2vec approach, which is generating vector representation of words to capture semantic meaning (Mikolov et al. 2013). The vectors are obtained by training a deep neural network based on a database of different text (so-called corpus) and results in a dense, high-dimensional representation of words. This procedure is a pre-training, resulting in a
117 6. Manuscript IV (In-silico prediction of microcystin toxicity) look-up table of words and vectors, which can be extracted later for various applications (e.g.
ML). To encode molecules with the Mol2vec approach a large corpus of a database or collection of molecular structures has to be generated (Figure 6-2A). The Morgan fingerprint is calculated for each molecule, but instead of hashing identifiers of substructures in a bit vector, identifiers
(or words) are extracted and ordered to a sentence to represent a molecule. By this procedure, a molecular lookup table of molecular substructures and vectors is generated, which is able to capture chemical relationship between substructures. To represent a new molecule, a Morgan fingerprint is generated, identifiers are extracted, and looked up in the pre-trained model. Then, all substructure vectors are summed up to represent one molecule (Figure 6-3), which results in a fixed size numerical representation of the molecule (Jaeger et al. 2018).
Figure 6-2: Application of Word2vec to molecules and proteins. The procedure results in a lookup table, where 300-dimensional vectors can be extracted to have a numerical representation. A) For molecules, identifiers generated by Morgan fingerprint are considered as words, and their ordering as a sentence or molecule. B) For proteins, all possible 3-grams are considered as words, and by applying a sliding window over protein sequence, three sentences are generated to represent a protein. Modified from Jaeger et al. (2018).
118 6. Manuscript IV (In-silico prediction of microcystin toxicity)
To encode a protein with the ProtVec approach (Figure 6-2B), a large corpus of a database or collection of protein sequences has to be generated. Therefore, all possible n-grams of a protein sequence are generated by applying a sliding window over a protein sequence. This results in n-sentences to represent a protein. Those n-sentences are then used to generate the lookup table of protein n-grams. To represent a new protein, all n-grams are generated, looked up in the pre-trained model, and summed up to represent a new protein (Figure 6-3), which results in a fixed size numerical representation of the protein (Asgari and Mofrad 2015; Jaeger et al. 2018).
PPP inhibition data (IC50’s) gained from the in vitro assays for the different MC congeners were then classified into three classes of “toxicity” (Table 6-2, Table 6-1 and Figure
6-3) and ML models generated, using the encoded MC congeners and PPP vectors (Figure 6-2 and Figure 6-3). These were then trained with different features and classifiers to classify MC congeners into the toxicity classes, as shown in the ML flowchart (Figure 6-4). The latter approach thus allowed to predict the toxicity of MC congeners and to compare the predictions with the true findings from the in vitro PPP inhibition assays.
Figure 6-3: Workflow for feature generation for toxicity classification. After downloading a pre-trained model for molecular structures and pre-training a protein model, vectors are extracted for substructures of 18 Microcystin congeners and the 3-grams of the 3 PPP‘s. To represent a molecule or protein, the respective vectors are summed up. Then, either 300-dimensional molecule vector is extracted, or combined with protein vector, to build a 600-dimensional vector. As the data set is highly imbalanced for the different classes, synthetic minority oversampling technique (SMOTE) is applied, to have the same number of compounds for each class. Modified from Jaeger et al. (2018).
119 6. Manuscript IV (In-silico prediction of microcystin toxicity)
Figure 6-4: After feature generation and pre-processing of the data, respective target values (toxicity class) are combined with the feature vector and used for a machine learning classification. Two different validation methods were used. For both validation methods, three machine learning models were set up, and majority voting was used for final prediction, building a so-called consensus model. Afterwards, performance was evaluated with a confusion matrix and evaluation metrics.
6.3 Material and methods
6.3.1 Materials
Microcystins were obtained either from Enzo Life Sciences (MC-RR, -LR, -YR, -WR, -
LA, -LY, -LF, -LW, -HilR, -HtyR, [-D-Asp3]-MC-RR and [-D-Asp3]-MC-LR) or were synthesized (MC-LY(Prg), [enantio-Adda5]-MC-LF, [Anda5]-MC-LY(Prg), [Amba5]-MC-
LY(Prg), [MSecPh7]-MC-LY(Prg)) using previously published procedures (Fontanillo et al.
2016; Zemskov et al. 2017). [-D-Asp3, Dhb7]-MC-RR was a gift from Judith Blom
(University of Zürich, Switzerland).
Microcystins were dissolved in pure methanol to 100 µM. Actual concentrations were determined using UV spectroscopy at 238 nm (using the extinction coefficient of MC-LR of
39800 mol l-1 cm-1) and stocks were stored at -20°C until used for serial dilutions. Stocks of
120 6. Manuscript IV (In-silico prediction of microcystin toxicity)
[Anda5]-MC-LY(Prg), [Amba5]-MC-LY(Prg) were quantified by dissolving weighed amounts in an appropriate volume of methanol, as photometric quantification was not possible due to lack of the characteristic absorption peak at 238 nm (missing Adda chain).
rPPP1 (rabbit skeletal muscle) was obtained from New England Biolabs (P0754L, product discontinued). hPPP2A (human red blood cells) was from Promega (V6311, product discontinued). pET32a(+)-TrxA-6His-hPPP5 was generated by GenScript using human PPP5 sequence (NCBI Accession NP_006238.1) with GenScripts codon optimization for E.coli.
6.3.2 Expression of 6xHis-hPPP5 in E.coli
pET32a(+)-TrxA-6His-hPPP5 was transformed into chemical competent BL21-
CodonPlus(DE3)-RP E.coli cells (Agilent, 230255) via heat-shock. After selection on ampicillin/chloramphenicol-agar, a single colony was picked and cultivated in 2 ml LB- ampicillin/chloramphenicol (amp/cam) medium for 6h. Afterwards, the pre-culture was added to 500 ml of TB-amp/cam (Terrific Broth supplemented with ampicillin and chloramphenicol) medium and incubated over night at 37°C while shaking (220 rpm). The following day OD600 was measured to ensure growth in the exponential phase (OD600 = ~4). Controls were performed using heat-shock treated BL21-CodonPlus(DE3)-RP E. coli cells without plasmid.
6.3.3 Purification of 6xHis-hPPP5
Cultures of transformed BL21-CodonPlus(DE3)-RP E. coli cells were centrifuged at
5000×g for 10 min at 4°C. Pellets were washed with 10 ml STE buffer (10 mM Tris-HCl pH 8.0,
100 mM NaCl, 1 mM EDTA) prior to resuspension in 10 ml cold PBS buffer + 1% protease inhibitor cocktail (Sigma, P8849). Cell lysis was achieved using a Branson Sonifier 250 with five times 20 pulses. Samples were cooled on ice between each set of 20 pulses. Lysate was cleared from cell debris by centrifugation for 45 min at 18000×g at 4°C. The cleared lysate was incubated with 2.5 ml (50:50 slurry) equilibrated Ni-NTA-agarose beads (Biozym, 2631105) on
121 6. Manuscript IV (In-silico prediction of microcystin toxicity) an overhead-tumbler at 4°C over-night. After centrifugation at 4°C and 800×g for 5 min, the supernatant was discarded and beads were washed three times with 2.5 ml wash buffer (50 mM
NaH2PO4, 300 mM NaCl, 60 mM imidazole). Elution was performed by incubating the beads for 10 min at 4°C on an overhead-tumbler in 2 ml elution buffer (50 mM NaH2PO4, 300 mM
NaCl, 100 mM imidazole). Samples were taken from each step for subsequent SDS-PAGE analysis. To avoid high imidazole concentrations in the final sample, buffer exchange via dialysis against storage buffer (20 mM TrisHCl pH 8, 100 mM NaCl) was performed using 2 L volumes for about 18 h with two buffer exchanges. 10% glycerol was added to the final samples before aliquoting and liquid nitrogen snap-freezing. All steps were either performed on ice or within a cooling room (4°C). Samples were stored at -80°C until further use. In addition, one analogous purification was performed using the heat-shock treated BL21-CodonPlus(DE3)-RP E.coli without plasmid (“empty expression”) for control purposes.
6.3.4 SDS-PAGE analysis
SDS-PAGE analysis was performed using 10% SDS-gels in a Bio-Rad Mini-PROTEAN vertical electrophoresis system. 6 µl 6 × SDS-sample buffer was added to 30 µl sample and the samples loaded into the gel pockets. Samples were not corrected for protein amount, but for the volume of the original fraction in such a fashion that consistently 0.3% of the original fractions were used (filled up to 30 µl with MilliQ). Gels were run at 100 V for 60 min and stained using colloidal Coomassie solution at 4°C while shaking over-night. Images were taken using a desktop scanner.
6.3.5 Mass spectrometry
Mass-spectrometry was employed to confirm the identity of the TrxA-6×His-hPPP5 expressed. Samples were run on SDS-gels as described above and bands considered to contain
TrxA-6×xHis-hPPP5 were cut out, divided into ~1 mm² squares and submitted to the Proteomics
122 6. Manuscript IV (In-silico prediction of microcystin toxicity)
Core Facility of the University Konstanz. Proteomic analyses of trypsin digested fragments were carried out with an LTQ Orbitrap Discovery (Thermo Fisher Scientific, Bremen Germany) coupled to an Eksigent 2D-nano HPLC (Eksigent, USA). Data were analysed using Mascot software (Matrix Science).
6.3.6 Phosphatase activity assay
The phosphatase activity of the TrxA-6xHis-hPPP5 fraction (hPPP5) was assayed using a colorimetric phosphatase activity assay with para-nitrophenylphosphate (pNPP) as substrate and buffers according to Heresztyn and Nicholson (Heresztyn and Nicholson 2001). Purified hPPP5 was tested undiluted and in 1:1 serial dilutions (total 11 concentrations, highest dilution
1:1024) using the enzyme diluent buffer (52 mM Tris pH 7, 2 mM MnCl2, 0.5 mg/ml BSA, 1 mM DTT, 0.5 mM NaOAc, 123.5 µM EGTA) for dilutions. 20 µl of each dilution was pipetted into a well of a polystyrene flat bottom 96-well plate to which 200 µl of the substrate solution
(62.5 mM Tris pH 8.1, 26 mM MgCl2, 0.2 mM MnCl2, 0.5 mg/ml BSA, 2 mM DTT, 1 mM
NaOAc, 24 mM pNPP) was given. Testing was carried out in technical triplicates. Colour development at 37°C and 405 nm was measured every 10 minutes over a period of 4 h. Linear regressions of each dilution over time were plotted and slopes obtained. Slopes obtained were compared to corresponding slopes of rPPP1 (protein amount and specific activity). Accordingly, this allowed calculation of the volume of the hPPP5 fraction corresponding to three U of rPPP1
(hPPP5: 0.83 U/µl, hPPP1: 2.5 U/µl). U here is defined as the amount of enzyme needed to dephosphorylate 1 nmol of pNPP in 1 minute at 30°C.
6.3.7 Colorimetric protein phosphatase inhibition assay (cPPIA)
The assay employed is based on the previously published procedures (Fischer et al. 2010;
Heresztyn and Nicholson 2001). Serial dilutions of each MC congener were produced in MilliQ in LC-vials, whereby the most concentrated MC sample contained a maximum of 5% methanol.
123 6. Manuscript IV (In-silico prediction of microcystin toxicity)
20 µl of each dilution was pipetted into a polystyrene flat-bottom 96-well plate in triplicates.
Three U of each phosphatase were diluted in 2120 µl enzyme dilution buffer and 20 µl of the solution were added to each well (corresponding to about 0.07 U per well). Enzyme solution (20
µl) in addition to 20 µl of MilliQ (three replicates) served as 100% activity control (no inhibition) while 20 µl enzyme dilution buffer lacking enzyme (three replicates) as well as 20 µl MilliQ served as background control (no enzyme activity). The plate was incubated at 37°C for 5 min to ensure interaction of microcystins with the PPP. To each well 200 µl of substrate solution was pipetted and the plate was immediately read at 405 nm using an Infinite 200 Pro microplate reader (Tecan, Männedorf, Switzerland). The plate was then incubated at 37°C for 3h before being measured again at 405 nm. PPP activity was calculated by subtracting the start value (0h) from the end value (3h) and compared to 100% activity. IC50 were calculated using GraphPad
Prism 5.0 software via a 5-PL non-linear regression with anchorage points and constraints between 100% and 0%. Replication: n = at least 3 for PPP2A and at least 5 for PPP1 and PPP5, each in technical duplicates or triplicates. The analyses of [Amba5]-MC-LF and [Anda5]-MC-
LF had 5 biological replicates but no technical triplicates due to shortage of pure testing material.
6.3.8 Data analyses and statistics
Data analyses were carried out using Microsoft Excel Professional Plus 2013, while
Graphpad Prism 5 was used for statistics and data representation. Data pre-processing and machine learning were carried out using python version 3.6.6.
6.3.9 Machine learning (ML)
The PPP inhibition capabilities (toxicodynamics) of the different MC congeners, expressed as IC50’s (Table 6-1), were used to train a ML model (Figure 6-2, Figure 6-3 and
Figure 6-4). In order to allow for the adaption into the ML model, the two primary factors, i.e.
MC congeners (molecules) and PPP (proteins), had to be transformed to numerical vectors
124 6. Manuscript IV (In-silico prediction of microcystin toxicity)
(Figure 6-2). To transform molecules to a vector, the Mol2vec approach, as described in Jaeger et al. (2018) was used with a pre-trained model (Turk 2018). To transform proteins to a vector, the ProtVec approach (Asgari and Mofrad 2015) was used (Figure 6-2) and a model trained on
UniProt sequences as described in (Jaeger et al. 2018). Subsequently, the models were applied to encode MC congeners and the PPPs (UniProt ID: PPP1 (P62136), PPP2A (P67775) and PPP5
(P53041)) as vectors.
PPP inhibition data (IC50’s) of the different MC congeners (Table 6-1) were classified into three classes of “toxicity” (Table 6-2, Figure 6-3 and Figure 6-4). The original data set consisted of 47 data points of which 31 data points classified as toxic, seven classified as less- toxic and nine classified as non-toxic. This classification chosen was arbitrary, albeit the most toxic classification includes MC congeners with relevance to human intoxications and the WHO guideline (Azevedo et al. 2002; Berry et al. 2017; Dietrich and Hoeger 2005; WHO 1999). As this is a rather small data set for ML and sufficient samples are crucial for ML performance,
Synthetic Minority Oversampling Technique (SMOTE) was applied to mimic a balanced dataset and thereby increase prediction performance of minority classes (Chawla et al. 2002). Indeed,
SMOTE generates new, artificial data points for minority class by variation of the feature vector representing original data points. SMOTE implementation in imbalanced learn (version 0.3.3)
(Lemaître et al. 2017) was used with standard settings and a ratio of 1.0. This procedure resulted in 31 data points per class, resulting in 93 data points in total (Table 6-2).
Three different ML models were trained with different features and classifiers to classify
MC congeners into the toxicity classes. Two models were trained with a random forest (RF) classifier implemented in scikit-learn (version 0.8.0) (Pedregosa et al. 2011) and the XGBoost implementation (version 0.8.0) of Gradient Boosting Machines (GBM) classifier (Friedman
2001), respectively. As feature, vectorized structural data of the congeners (Mol2vec) was used.
Additionally, one model was trained with a random forest classifier with molecular and protein data as feature (Table S6-3, Figure 6-3 and Figure 6-4). To merge molecule and protein
125 6. Manuscript IV (In-silico prediction of microcystin toxicity) information into one vector, structural data of molecules was vectorized with Mol2vec and protein data was vectorized with ProtVec and then concatenated. Subsequently, hyper- parameters were tuned (Table S6-3) to derive the best model. The final model used majority voting of these three models (Mol2vec with RF, Mol2vec with GBM and Mol2vec + ProtVec with RF) to classify a compound (Figure 6-4).
For training and evaluation of an ML algorithm, the data set had to be split into a training and a test set. Those two sets had to be strictly separated, because when data points from the training set were used in the test set, the evaluation would always result in high performance, because the model already knew the data point. Here, two procedures were applied to split the data set: 1) using 80% of the data points (75) for training and 20% of the data points (18) for testing the performance and 2) by five-fold cross validation (Figure 6-4). Applying five-fold cross validation results in four-folds of the data set used for training and one-fold of the data set used for testing. This procedure is repeated, until every fold was once the test set, resulting in five ML models. This procedure has the advantage, that a standard deviation of performance between the models can be calculated to get a better estimation of model performance and robustness (Table S6-4). For both procedures data points are randomly assigned to the respective set or fold. For this reason, each procedure and ML was repeated 50 times, to test whether performance was robust and independent of the random assignment of data points for training.
To finally estimate performance of the ML model, different performance metrics were employed (precision, recall, F-score, see Table S6-4 and Figure S6-1). In addition, the confusion matrix (Figure 6-5) was checked, to estimate how many and which molecules were classified correctly or falsely. Performance metrics and confusion matrix were used as implemented in scikit-learn (Pedregosa et al. 2011).
126 6. Manuscript IV (In-silico prediction of microcystin toxicity) 6.4 Results
Full-length human PPP5 was bacterially expressed in BL21 Codon Plus E. coli with several attached tags: Thioredoxin A (TrxA) for solubility, 6-Histag for purification, S-Tag for a possible second purification and a thrombin-site (TrxA-6His-S-PPP5). PPP5 identity was confirmed using mass spectrometry after purification using Ni-NTA beads (Table S6-2). To ensure that the observed activity was due to PPP5, bacteria without a plasmid were grown, purified and tested. These purified proteins did not show PPP activity (Figure S6-2).
To develop a dataset of MC congener dependent “toxicity” (PPP inhibition activity) for the ML model, 18 different MC congeners were tested in three PPP (rPPP1, hPPP2A, hPPP5 expression #1). The 18 MC congeners spanned the known spectrum of hydrophobicity, had different molecular weights, and contained common as well as unusual modifications of the consensus structure (Figure 6-1 and Table S6-1). The in vitro PPP inhibition assays provided for well fitted (R2) concentration-inhibition response curves (Table 6-1, Figure S6-3, Figure S6-4 and Figure S6-5), whereby the derived IC50’s were subsequently used for the ML approach
(Figure 6-2, Figure 6-3 and Figure 6-4).
127 6. Manuscript IV (In-silico prediction of microcystin toxicity)
Table 6-1: IC50s of the tested MC congeners on rPPP1, hPPP2A and hPPP5. Congener rPPP1 hPPP2A hPPP5
CI95 CI95 IC50 CI95 2 2 2 IC50 (nM) R IC50 (nM) R R (nM) (nM) (nM) (nM) MC-RR 1.5 1.3 – 1.8 0.95 1.6 1.4 – 1.7 0.99 11.7 8.3 – 16.5 0.96 MC-LR 0.3 0.2 – 0.4 0.93 0.5 0.4 – 0.5 0.99 5.1 4.0 – 6.6 0.97 MC-YR 1.3 1.2 – 1.5 0.99 n.d. n.d. n.d. 5.1 4.3 – 6.1 0.99 MC-WR 1.2 1.0 – 1.5 0.94 1.0 0.8 – 1.1 0.97 5.6 4.2 – 7.6 0.97 MC-LA 1.9 1.4 – 2.7 0.86 0.7 0.5 – 0-9 0.93 6.1 4.3 – 8.7 0.96 MC-LY 0.8 0.7 – 0.9 0.99 n.d. n.d. n.d. 4.1 3.1 – 5.4 0.97 MC-LF 2.0 1.5 – 2.6 0.90 1.4 1.3 – 1.4 0.99 4.7 3.5 – 6.3 0.97 MC-LW 1.2 1.0 – 1.4 0.97 0.7 0.7 – 0.8 0.99 2.5 2.0 – 3.2 0.98 MC-HilR 0.6 0.5 – 0.8 0.99 n.d. n.d. n.d. 4.2 3.5 – 5.1 0.99 MC-HtyR 0.7 0.6 – 0.8 0.99 n.d. n.d. n.d. 4.7 3.6 – 6.0 0.96 [-D-Asp3]-MC-RR 45.0 39.3 – 51.6 0.99 n.d. n.d. n.d. 167.1 131.8 – 211.8 0.97 [-D-Asp3]-MC-LR 0.9 0.7 – 1.0 0.99 n.d. n.d. n.d. 10.2 8.3 – 12.5 0.99 [-D-Asp3, Dhb7]-MC-RR 62.0 51.7 – 74.3 0.96 84.3 80.7 - 87.8 0.99 877.1 692.6 – 1111 0.97 MC-LY(Prg) 1.7 1.3 – 2.2 0.95 0.4 0.2 – 0.3 0.99 1.7 1.2 – 2.6 0.95 [MSecPh7]-MC-LY(Prg) 1.9 1.6 – 2.4 0.97 0.9 0.7 – 1.1 0.94 18.2 10.7 – 31.1 0.91 [enantio-Adda5]-MC-LF ------[Amba5]-MC-LY(Prg) 449800 – 1991 - 520817 0.98 2135 0.99 54063 37431 - 78087 0.95 603048 2291 [Anda5]-MC-LY(Prg) 1724 1434 - 2072 0.98 n.d. n.d. n.d. 2420 1690 - 3467 0.96
IC50 are calculated with GraphPad Prism 5 after 5PL-nonlinear regression of at least three (hPPP2A) or five (rPPP1 and hPPP5) individual replicates using technical duplicates or triplicates. n.d. not determined (PPP2A not available any more, discontinued by manufacturer).
128 6. Manuscript IV (In-silico prediction of microcystin toxicity)
The comparison of IC50’s obtained with MC congeners in the three PPPs tested, demonstrated that of the ten MC congeners available for testing in PPP1 and 2A, five MC congeners had a comparable IC50 values, while five MC congeners were more selective towards
PPP2A, possibly suggesting a slightly higher sensitivity of PPP2A toward MC congeners (Table
6-1 and Figure S6-6). In contrast, several-fold higher concentrations of MC congeners were necessary to achieve 50% inhibition of PPP5 phosphatase activity. Notable exceptions to the latter were MC-LY(Prg) and [Amba5]-MC-LY(Prg), to which the PPP5 susceptibility to inhibition was comparable and but lower than observed for PPP1.
The importance of structural differences with regard to binding to the catalytic subunit of PPPs was dramatically demonstrated with the comparison of MC-LF and the de novo synthetized [enantio-Adda5]-MC-LF. While MC-LF inhibited all three PPPs tested, the corresponding [enantio-Adda5]-MC-LF had no PPP inhibitive activity at all (Table 6-1). In contrast, other structurally similar derivatives, i.e. MC-LY(Prg) and [MSecPh7]-MC-LY(Prg), both having a propargyloxy residue at position 4 (Figure 6-1) of the phenylalanine moiety, show only slightly reduced PPP inhibiting activity, if any, when compared to the parent MC-LF.
However, if the Adda-residue is shortened to [Amba5]-MC-LY(Prg) or [Anda5]-MC-LY(Prg)
(Figure 6-1 and Figure S6-1) a marked reduction in PPP inhibiting capacity is found (Table 6-1).
The latter observations suggest that structural changes of the amino acid Adda (enantiomeric configuration or shortened Adda-side chain) prohibited or reduced functional interaction with the catalytic subunits and thus inhibition of the PPPs. On the other hand, structural changes to the phenylalanine moiety at position 4 or the Mdha at position 7 had limited impact on PPP activity. Similarly, exchanging leucine in MC-LR and tyrosine in MC-YR for a homoisoleucine
(MC-HilR) and a homotyrosine (MC-HtyR) at position 2 had no significant effect on PPP inhibition capacity (Table 6-1). However, changing the methylation of β-D-MeAsp at position 3 of MC-RR to demethylated [β-D-Asp3]-MC-RR and [β-D-Asp3,Dhb7]-MC-RR, resulted in a decreased PPP inhibition capacity, thus suggesting that structural changes involving L-amino
129 6. Manuscript IV (In-silico prediction of microcystin toxicity) acid residues at position 3 could have an impact on the inhibition of PPPs. When comparing the impact of structural changes of the Adda side-chain at position 5 with changes of β-D-MeAsp at position 3, it appears that the former had a much more pronounced impact on the binding of MC congeners into the catalytic subunit and thus inhibition of PPPs.
The above IC50 values were classified in toxicity classes (Table 6-2) and used for an ML approach. As the distribution of data points was not similar among classes, oversampling was applied, resulting in 31 data points per class, adding up to a total of 93 (Table 6-2, Figure 6-3 and Figure 6-4). The resulting data was used for ML via two different approaches: using 80% of the data for training and 20% for testing of prediction (80/20) and by 5-fold cross validation
(CV) with 50-times repetition (see Figure 6-4, Table S6-4 and 6.3.9 for details).
Table 6-2: “Toxicity” classes assigned to MC congeners. Number data points Number data points IC50 (nM) Class Description (original) after oversampling ≤10 0 Toxic 31 31 > 10 ≤ 1000 1 Less-toxic 7 31 > 1000 2 Non-toxic 9 31
Total 47 93
Classification is based on their PPP inhibitive capabilities (expressed as IC50) for each of the three PPPs tested.
Subsequently, all data points were used and majority voting of three ML models was employed to predict the toxicity class of every data point (Table S6-3). Both approaches of splitting data in training and test set performed well for toxicity class prediction with a precision above 0.8 and a recall and F-score mostly above 0.8 (Table S6-4). Since cross-validation is more suitable for small data sets (Beleites et al. 2013), we primarily focused on CV results, although
80/20 results are provided in Table S6-4 as well. CV predictions were compared with the true classification according to classified IC50 values (Figure 6-5A).
130 6. Manuscript IV (In-silico prediction of microcystin toxicity)
Figure 6-5: Confusion matrix of microcystin toxicity prediction using 5-fold cross validation. Either the whole dataset was used for training (A), or the stereoisomer [enantio- Adda5]-MC-LF) was dropped for the training of the algorithm (B). The identity of all wrong classifications are given in the individual squares.
25 of the 31 toxicity data points were predicted correctly, while two were wrongly predicted as less toxic and four as non-toxic. Interestingly, the misclassified MC congeners primarily involved MC-LF variants. For example, the stereoisomer of MC-LF ([enantio-Adda5]-
MC-LF) was classified as toxic in PPP2A, despite that it is not toxic. The latter resulted most likely from the fact that the training vector generation was not trained for chirality of the molecules in question. Indeed, the vector generated for [enantio-Adda5]-MC-LF and MC-LF would be identical, albeit the values entered into the ML algorithm would read “non-toxic” and
“toxic” and thus result in wrong classifications by ML. To test this, the same approach was chosen, yet while omitting the [enantio-Adda5]-MC-LF-variant from the training and test sets of the ML algorithm (Figure 6-5B). In consequence, both natural MC-LF and propargylated MC-
LF variants were classified correctly. Moreover, [β-D-Asp3]-MC-LR (in PPP1) was now moved from originally being wrongly predicted as non-toxic, to be now predicted as less toxic, despite its true affiliation in the “toxic” class. Similarly, the prediction of [β-D-Asp3]-MC-LR (in PPP5) moved from non-toxic to toxic, despite its true affiliation in the less-toxic class. Finally, MC-RR
131 6. Manuscript IV (In-silico prediction of microcystin toxicity)
(in PPP5) moved from the correct prediction of “less toxic” to the wrong prediction of “toxic”.
Overall, the ML model trained without the MC-LF stereoisomer performed better, producing fewer false negative predictions.
6.5 Discussion
Ser/thr-PPP are an evolutionary very old family of enzymes, with PPP1 and PPP2A being two of the most slowly evolving enzymes (Cohen 2004). PPP1 usually works together with a regulatory subunit, while the PPP2A-holoenzyme additionally comprises a scaffold subunit (Shi
2009). In this regard PPP2A is similar to PPP4 and PPP6, whose holoenzymes are also comprised of three subunits, i.e. scaffold, regulatory and catalytic (Brewis et al. 1993; Stefansson et al.
2008). Moreover, PPP4 and PPP6 are more closely related to PPP2A than to any other member of the PPP family (Shi 2009). Thus, it can be assumed that PPP4 and PPP6 should behave similarly to MC inhibition as PPP2A. PPP5 is the most distant from the other members of the
PPP family (Andreeva and Kutuzov 2001). PPP5 is expressed as a single peptide combining the catalytic domain with a TPR-domain, which interacts with the peptides c-terminus to act as auto- inhibitory domain (Yang et al. 2005).
The catalytic subunits of PPP1, PPP2A and PPP5, representing three members of distinct subfamilies of the PPP family, were used for inhibition assays with various MC congeners. In view of the absence of pure human PPP1 available for in vitro testing, PPP1 from rabbit skeletal muscle was employed as it displays 100% protein sequence identity to human PPP1 (hPPP1; analysed with Clustal Omega and Geneious (Biomatters)) and is therefore considered to be equal in structure and enzymatic performance to hPPP1. In contrast, hPPP2A and hPPP5 have only
43% and 37% sequence homology with hPPP1, while hPPP5 and hPPP2A are 40% homologous
(analysed with Clustal Omega). Despite sequence differences, the 3D-structures of the respective catalytic subunits align quite well (see Figure S6-7), suggestive of similar size and structure
132 6. Manuscript IV (In-silico prediction of microcystin toxicity) restrictions for MC interaction with the hydrophobic groove close to the catalytically active center of the respective PPPs.
Indeed, despite that PPPs have been reported to be inhibited by MC earlier (Fischer et al.
2010; Garibo et al. 2014a; Hastie et al. 2005; Mackintosh et al. 1990), the data presented here are unique as they compare the MC-mediated inhibition of the catalytic subunits of three different PPPs in parallel with an hitherto unprecedented number of MC congeners, including synthetically derived structural variants. The data demonstrated that MC-RR, -LR, -YR, and [β-
D-Asp3, Dhb7]-MC-RR presented with similar IC50 for PPP1 and PPP2A (Table 6-1), while
PPP2A was ≥2-fold more sensitive to the more hydrophobic MC-LW, -LA, -LF and the synthetic
MC-LF derivatives (MC-LY(Prg), [MSecPh7]-MC-LY(Prg) and [Amba5]-MC-LY(Prg) (Figure
6-1). Although ([Amba5]-MC-LY(Prg) displayed an approx. 243-fold higher toxicity towards
PPP2A than to PPP1, this was not the case for the other two OPrg-containing congeners
([MSecPh7]-MC-LY(Prg) and [Anda5]-MC-LY(Prg)). Indeed, it has previously been described that MC variants with reduced Adda5-sidechains show a tendency to bind more effectively to
PPP2A than to PPP1(Fontanillo et al. 2016).
In principle the MC congener’s inhibition capacity of PPP5 followed the same trend as observed for PPP1 and PPP2A, albeit being 4-200-fold lower. With the exception of [enantio-
Adda5]-MC-LF, showing absence of binding to all three PPPs tested (Table 6-1), earlier assumptions regarding size and structure restrictions for MC interaction with the respective PPPs could not be corroborated. Indeed, MC congeners apparently do not bind as tightly to the catalytic subunit of PPP5 as to the catalytic subunits of PPP1 and 2A. Exception to the latter, surprisingly, were the synthetic MC-LY(Prg) and [Anda5]-MC-LY(Prg), sharing similar inhibition capabilities in PPP1 and PPP5 (Table 6-1). Asp3 variants of MC-LR appeared to be of comparable toxicity as MC-LR, while in contrast β-D-Asp3 variants of MC-RR were all dramatically less toxic than MC-RR across all PPPs tested. Although MC-LR is considered to be the most toxic of all congeners (World Health Organization 2017), this appears to apply only
133 6. Manuscript IV (In-silico prediction of microcystin toxicity) to PPP1 when considering toxicodynamic data. Indeed, MC-LW, MC-LF, MC-LY(Prg) and
[MSecPh7]-MC-LY(Prg) presented with comparable inhibiting capabilities as MC-LR in
PPP2A. Moreover, in PPP5 MC-WR, -YR, -LY, -LA, -HilR, -HtyR were of comparable toxicity while MC-LF and MC-LY(Prg) were more toxic than MC-LR. The latter suggests that using the toxicity equivalent factors concept (TEF), i.e. all MC congeners equaling in toxicity to MC-LR, would under- and overestimate the potential toxicodynamic capacity present in a given cyanobacterial bloom. Moreover, the fact that MC congeners have been demonstrated to present with significant differences with regard to OATP transport (Feurstein et al. 2011; Fischer et al.
2005), whereby MC-LR and-RR are transported less efficiently than e.g. MC-LA,-LW, or -LF, compounds the problems mentioned with using the TEF as originally proposed by Dietrich and
Hoeger (Dietrich and Hoeger 2005). Indeed, in a realistic setting employing a guidance value of
1 µg MC-LRequivalent /L for drinking water (Falconer and Humpage 2005) and using summary detection methods e.g. ELISA (Fischer et al. 2001) without concurrent LC-MS/MS confirmation of MC congeners present (Puddick et al. 2014), could severely under- or overestimate the toxicity of a MC congener mixture in a given water sample contaminated by a toxic cyanobacterial bloom. Indeed, there are several reports of multiple co-occurring MC congeners in a given cyanobacterial bloom (Falconer et al. 1994; Kleinteich et al. 2018; Puddick et al.
2014), thus demonstrating the reality of having to deal with mixture exposures of different toxicities in a human hazard and risk assessment scenario. The question then needs to be raised as to how one could deal with the uncertainties of having more than 248 putative MC congeners
(Spoof and Catherine 2017) on one hand, yet absence of relevant toxicity data for the majority of these MC congeners, on the other hand, that would allow for appropriate hazard and risk assessment. The latter discrepancy is exacerbated by the fact that for the most of the 248 putative
MC congeners there is no purified material available to actually test the MC using in vitro toxicokinetic and -dynamic assays and thus to provide for a minimal dataset that could be of toxicological relevance for humans.
134 6. Manuscript IV (In-silico prediction of microcystin toxicity)
One approach, albeit yet limited to the toxicodynamic component of the apical toxicity, is the employed ML approach for roughly categorizing MC congeners into groups of “toxic, slightly toxic, and non-toxic” MC and thus to make preliminary predictions. Although the available data for training of the ML approach was restricted to 18 MC congeners, the number of samples was artificially increased with the oversampling technique SMOTE. This approach introduced uncertainty in the data, which might have caused overlapping data points of different classes and therefore the wrong classification of compounds. Therefore, more MC should be tested, including rare MC variants (e.g. doubly-demethylated congeners) to expand the existing data set for better model performance and essentially more fine-scaled predictions (i.e. more toxicity classes). Inclusion of other PPP inhibitors, e.g. the structurally related nodularins
(Rinehart et al. 1988) and structurally unrelated anabaenopeptins (Spoof et al. 2015) should ensure a better and more sensitive predictive performance. However, irrespective of the potential uncertainties experienced, both models used (Figure 6-5 and Table S6-4) provided for 80-90% correct predictions of toxicity class. More importantly however, modulation of the training set allowed for improved prediction (Figure 6-5), whereby most of the few wrongly predicted MC were found in a higher toxicity class, thereby overestimating the true toxicity. As overestimation of toxicity would be more precautionary with regard to potential hazard and risk, this caveat of the ML approach was considered acceptable.
Obviously categorizing MC congeners into toxicodynamic classes will not resolve the problem of having to assess the potential hazard of mixtures of different MC congeners present in a given surface water. Additionally, the use of calculated TEF, as attempted by Garibo et al.
(2014a) with six different MC congeners, or the TEF calculated from the PPP inhibition data in this study (Table 6-3), will not alleviate the problem of lacking toxicokinetic data. Although there have been past efforts to characterize the uptake into human cells via OATPs (Fischer et al. 2010; Fischer et al. 2005; Monks et al. 2007) and current efforts are under way to characterize the efflux of MC congeners from human epithelial cells using human exporter expressing insect
135 6. Manuscript IV (In-silico prediction of microcystin toxicity) membrane vesicle systems (Kaur et al. 2019a), we are still far from actually being able to calculate individual MC congener toxicokinetics or even considering integrating toxicokinetic- and dynamic data into toxicologically based kinetic and -dynamic modelling.
Table 6-3: MC congener toxicity equivalency factors (TEFs).
PPP1 PPP2A PPP5
IC50 (nM) TEF IC50 (nM) TEF IC50 (nM) TEF MC-RR 1.5 0.20 1.6 0.31 11.7 0.44 MC-LR 0.3 1.00 0.5 1.00 5.1 1.00 MC-YR 1.3 0.22 n.d. n.d. 5.1 1.00 MC-WR 1.2 0.24 1.0 0.50 5.6 0.91 MC-LA 1.9 0.15 0.7 0.71 6.1 0.84 MC-LY 0.8 0.36 n.d. n.d. 4.1 1.24 MC-LF 2.0 0.15 1.4 0.36 4.7 1.09 MC-LW 1.2 0.24 0.7 0.66 2.5 2.02 MC-HilR 0.6 0.49 n.d. n.d. 4.2 1.20 MC-HtyR 0.7 0.45 n.d. n.d. 4.7 1.09 [-D-Asp3]-MC-RR 45.0 0.01 n.d. n.d. 167.1 0.03 [-D-Asp3]-MC-LR 0.9 0.34 n.d. n.d. 10.2 0.50 [-D-Asp3, Dhb7]-MC-RR 62.0 0.01 84.3 0.01 877.1 0.01 MC-LY(Prg) 1.7 0.17 0.4 1.76 1.7 2.95 [MSecPh7]-MC-LY(Prg) 1.9 0.15 0.9 0.54 18.2 0.28 [enantio-Adda5]-MC-LF ------[Amba5]-MC-LY(Prg) 520817 0.000 2135 0.000 54063 0.00 [Anda5]-MC-LY(Prg) 1724 0.000 n.d. n.d. 2420 0.00
TEFs were determined by defining the IC50 of MC-LR as 1 and the others in relation to
MC-LR based on their IC50. n.d. not determined in that particular experiment.
136 6. Manuscript IV (In-silico prediction of microcystin toxicity) 6.6 Supplementary Material
Figure S6-1: Calculation of prediction measurements. Precision is defined as the ability of a model to retrieve relevant results, recall as the ability of a model to retrieve relevant results and the F- Score as harmonic mean of precision and recall.
3,5
3
2,5 PPP5 1:8 2 PPP5 1:128 1,5 empty 1:8 1 empty 1:128
0,5 Absorbance @ 405 nm (AU) 0 0 20 40 60 80 100 120 140 -0,5 Time (min)
Figure S6-2: Comparison of the hPPP5 expression with the empty expression. Different dilutions of the two respective fractions were analysed towards their activity to dephosphorylate pNPP.
137 6. Manuscript IV (In-silico prediction of microcystin toxicity)
Figure S6-3: Inhibition curves for PPP1. Only upper error bars (SD) are shown for reasons of clarity.
Figure S6-4: Inhibition curves for PPP2A. Only upper error bars (SD) are shown for reasons of clarity.
138 6. Manuscript IV (In-silico prediction of microcystin toxicity)
Figure S6-5: Inhibition curves for PPP5. Only upper error bars (SD) are shown for reasons of clarity.
139 6. Manuscript IV (In-silico prediction of microcystin toxicity)
Figure S6-6: Comparison IC50s on the three tested phosphatases. Biological replicates were plotted individually using GraphPad Prism 5 and individual
IC50 are noted. Black bar shows the median of the IC50.
Figure S6-7: Overlay of the structures of the used PPP. PyMOL was used using the PDB entries for PPP1 (4MOV, green), PPP2A (2IE4, teal) and PPP5 (4ZX2, magenta).
140 6. Manuscript IV (In-silico prediction of microcystin toxicity)
Table S6-1: Congener-dependent modifications. Congener X (2) Z (4) Position 3 Position 7 Position 5 MC-RR Arginine Arginine -D-MeAsp Mdha Adda MC-LR Leucine Arginine -D-MeAsp Mdha Adda MC-YR Tyrosine Arginine -D-MeAsp Mdha Adda MC-WR Tryptophane Arginine -D-MeAsp Mdha Adda MC-LA Leucine Alanine -D-MeAsp Mdha Adda MC-LY Leucine Tyrosine -D-MeAsp Mdha Adda MC-LF Leucine Phenylalanine -D-MeAsp Mdha Adda MC-LW Leucine Tryptophane -D-MeAsp Mdha Adda MC-HilR Homoisoleucine Arginine -D-MeAsp Mdha Adda MC-HtyR Homotyrosine Arginine -D-MeAsp Mdha Adda [-D-Asp3]-MC-RR Arginine Arginine -D-Asp Mdha Adda [-D-Asp3]-MC-LR Leucine Arginine -D-Asp Mdha Adda [-D-Asp3, Dhb7]-MC-RR Arginine Arginine -D-Asp Dhb Adda MC-LY(Prg) Leucine Tyrosine(Prg) -D-MeAsp Mdha Adda [enantio-Adda5]-MC-LF Leucine Phenylalanine -D-MeAsp Mdha enantio-Adda [MSecPh7]-MC-LY(Prg) Leucine Tyrosine(Prg) -D-MeAsp MSecPh Adda [Amba5]-MC-LY(Prg) Leucine Tyrosine(Prg) -D-MeAsp Mdha Amba [Anda5]-MC-LY(Prg) Leucine Tyrosine(Prg) -D-MeAsp Mdha Anda
Hil (homoisoleucine) and Hty (homotyrosine) are variants of isoleucine and tyrosine respectively, which have an additional CH2 compared to the parent amino acids. For structures of amino acids see Figure 1.
141 6. Manuscript IV (In-silico prediction of microcystin toxicity)
Table S6-2: Results of the mass spectrometric analysis. # Unique MW Accession Description Score Coverage # Proteins # Peptides # PSMs # AAs calc. pI Peptides [kDa] P53041 Serine/threonine-protein phosphatase 5 OS=Homo sapiens GN=PPP5C PE=1 SV=1 - 18779.21 69.74 10 35 35 626 499 56.8 6.28 [PPP5_HUMAN] C6EG32 Thioredoxin OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_4259 PE=4 SV=1 - 5021.36 66.97 1 6 6 128 109 11.8 4.88 [C6EG32_ECOBD] C6EG74 Glutamine--fructose-6-phosphate aminotransferase [isomerizing] OS=Escherichia coli 2972.68 67.16 1 33 33 89 609 66.9 5.87 (strain B / BL21-DE3) GN=glmS PE=3 SV=1 - [C6EG74_ECOBD] C6ECY2 Chaperonin GroEL OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3888 PE=3 2072.26 72.26 1 29 29 68 548 57.3 4.94 SV=1 - [C6ECY2_ECOBD] C6EB40 Chaperone protein DnaK OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3605 1381.54 52.98 1 23 26 43 638 69.1 4.97 PE=3 SV=1 - [C6EB40_ECOBD] C6EE39 Translation elongation factor Tu OS=Escherichia coli (strain B / BL21-DE3) 1286.33 59.39 2 16 18 45 394 43.3 5.45 GN=ECBD_4053 PE=3 SV=1 - [C6EE39_ECOBD] C6ECZ5 Lysine--tRNA ligase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3901 PE=3 1226.75 45.35 1 13 20 44 505 57.8 5.24 SV=1 - [C6ECZ5_ECOBD] C6EKZ1 Chaperone protein HtpG OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3183 1221.43 57.05 1 24 27 42 624 71.4 5.21 PE=3 SV=1 - [C6EKZ1_ECOBD] C6E9U5 Bifunctional polymyxin resistance protein ArnA OS=Escherichia coli (strain B / BL21- 1112.67 48.18 1 21 21 37 660 74.2 6.87 DE3) GN=ECBD_1404 PE=3 SV=1 - [C6E9U5_ECOBD] C6EAU2 Pyruvate dehydrogenase complex dihydrolipoamide acetyltransferase OS=Escherichia coli 1019.86 40.16 1 16 16 30 630 66.1 5.17 (strain B / BL21-DE3) GN=ECBD_3504 PE=4 SV=1 - [C6EAU2_ECOBD] C6EI71 Ribosomal protein S1 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_2684 PE=4 975.04 45.42 1 21 21 31 557 61.1 4.98 SV=1 - [C6EI71_ECOBD] C6EJZ8 ATP-dependent chaperone ClpB OS=Escherichia coli (strain B / BL21-DE3) 841.83 31.39 1 19 19 25 857 95.5 5.52 GN=ECBD_1092 PE=4 SV=1 - [C6EJZ8_ECOBD] C6ECU7 Superoxide dismutase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_1987 PE=4 815.45 80.31 1 9 9 26 193 21.3 5.95 SV=1 - [C6ECU7_ECOBD] C6EL63 Cysteine synthase A OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_1267 PE=4 812.89 56.04 1 12 12 17 323 34.5 6.06 SV=1 - [C6EL63_ECOBD] C6EJ93 Enolase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0950 PE=3 SV=1 - 672.19 46.30 1 12 12 19 432 45.6 5.48 [C6EJ93_ECOBD]
142 6. Manuscript IV (In-silico prediction of microcystin toxicity)
C6EII3 Phosphoglycerate kinase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0812 575.89 42.12 1 11 12 23 387 41.1 5.22 PE=3 SV=1 - [C6EII3_ECOBD] C6ECN8 Threonine--tRNA ligase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_1926 537.73 18.22 1 11 11 18 642 74.0 6.19 PE=3 SV=1 - [C6ECN8_ECOBD] C6EIL8 Lysine--tRNA ligase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0847 PE=3 529.29 34.06 1 8 15 25 505 57.6 5.24 SV=1 - [C6EIL8_ECOBD] C6EAU1 Dihydrolipoamide dehydrogenase OS=Escherichia coli (strain B / BL21-DE3) 522.12 32.49 1 11 11 14 474 50.7 6.15 GN=ECBD_3503 PE=4 SV=1 - [C6EAU1_ECOBD] C6EJL4 Succinate dehydrogenase flavoprotein subunit OS=Escherichia coli (strain B / BL21-DE3) 484.14 23.30 1 10 10 13 588 64.4 6.27 GN=ECBD_2937 PE=3 SV=1 - [C6EJL4_ECOBD] C6EA90 Proline--tRNA ligase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3424 PE=3 446.70 25.70 1 10 10 12 572 63.6 5.19 SV=1 - [C6EA90_ECOBD] P00761 Trypsin OS=Sus scrofa PE=1 SV=1 - [TRYP_PIG] 442.41 25.11 1 4 4 14 231 24.4 7.18 C6ECS9 Pyruvate kinase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_1969 PE=4 SV=1 426.45 27.02 1 10 10 17 470 50.7 6.09 - [C6ECS9_ECOBD] C6EC50 Glyceraldehyde-3-phosphate dehydrogenase OS=Escherichia coli (strain B / BL21-DE3) 397.55 29.61 1 6 7 11 331 35.5 7.11 GN=ECBD_1865 PE=3 SV=1 - [C6EC50_ECOBD] C6EL29 Trigger factor OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3221 PE=3 SV=1 - 383.65 34.26 1 11 11 13 432 48.2 4.88 [C6EL29_ECOBD] C6EJ92 CTP synthase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0949 PE=3 SV=1 - 380.09 16.88 1 8 8 13 545 60.3 5.94 [C6EJ92_ECOBD] C6EFU4 Short-chain dehydrogenase/reductase SDR OS=Escherichia coli (strain B / BL21-DE3) 374.98 25.19 1 4 5 8 262 27.8 5.87 GN=ECBD_2329 PE=1 SV=1 - [C6EFU4_ECOBD] C6EA50 Aminoacyl-histidine dipeptidase OS=Escherichia coli (strain B / BL21-DE3) 348.95 25.98 1 8 9 12 485 52.9 5.52 GN=ECBD_3384 PE=4 SV=1 - [C6EA50_ECOBD] C6EBG1 Transaldolase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3610 PE=3 SV=1 - 334.15 38.49 2 9 9 12 317 35.2 5.21 [C6EBG1_ECOBD] C6EGV5 Isocitrate dehydrogenase, NADP-dependent OS=Escherichia coli (strain B / BL21-DE3) 322.82 22.36 1 6 6 11 416 45.7 5.40 GN=ECBD_2463 PE=4 SV=1 - [C6EGV5_ECOBD] C6EGZ6 Malate dehydrogenase, NAD-dependent OS=Escherichia coli (strain B / BL21-DE3) 315.07 23.08 1 5 5 7 312 32.3 5.77 GN=ECBD_0511 PE=3 SV=1 - [C6EGZ6_ECOBD] C6EL61 Phosphoenolpyruvate-protein phosphotransferase OS=Escherichia coli (strain B / BL21- 311.30 14.43 1 6 6 8 575 63.5 4.87 DE3) GN=ECBD_1265 PE=4 SV=1 - [C6EL61_ECOBD] C6EGG1 Ribosomal protein L6 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0446 PE=3 302.07 41.81 1 6 6 12 177 18.9 9.70 SV=1 - [C6EGG1_ECOBD] C6EBI7 Purine nucleoside phosphorylase OS=Escherichia coli (strain B / BL21-DE3) 296.62 37.24 1 6 6 10 239 25.9 5.66 GN=ECBD_3636 PE=3 SV=1 - [C6EBI7_ECOBD]
143 6. Manuscript IV (In-silico prediction of microcystin toxicity)
C6EIH4 Transketolase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0803 PE=4 SV=1 - 296.10 20.97 1 7 8 11 663 72.2 5.67 [C6EIH4_ECOBD] C6EB22 4-hydroxy-tetrahydrodipicolinate reductase OS=Escherichia coli (strain B / BL21-DE3) 252.77 17.95 1 3 3 6 273 28.7 5.76 GN=ECBD_3585 PE=3 SV=1 - [C6EB22_ECOBD] C6EI79 Formate acetyltransferase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_2692 245.29 11.97 1 7 7 8 760 85.3 6.01 PE=4 SV=1 - [C6EI79_ECOBD] C6EI52 Asparaginyl-tRNA synthetase OS=Escherichia coli (strain B / BL21-DE3) 240.13 23.82 1 7 8 10 466 52.5 5.31 GN=ECBD_2665 PE=3 SV=1 - [C6EI52_ECOBD] C6EFH1 Phosphoenolpyruvate carboxykinase (ATP) OS=Escherichia coli (strain B / BL21-DE3) 235.16 15.37 1 6 6 9 540 59.6 5.71 GN=ECBD_0342 PE=3 SV=1 - [C6EFH1_ECOBD] C6EG71 ATP synthase F1, beta subunit OS=Escherichia coli (strain B / BL21-DE3) 233.94 25.87 1 9 9 9 460 50.3 5.01 GN=ECBD_4300 PE=3 SV=1 - [C6EG71_ECOBD] C6E9W2 Glycerophosphoryl diester phosphodiesterase OS=Escherichia coli (strain B / BL21-DE3) 232.32 24.30 1 6 6 6 358 40.8 5.60 GN=ECBD_1421 PE=4 SV=1 - [C6E9W2_ECOBD] C6EII4 Fructose-bisphosphate aldolase, class II OS=Escherichia coli (strain B / BL21-DE3) 231.68 20.61 1 5 5 8 359 39.1 5.86 GN=ECBD_0813 PE=4 SV=1 - [C6EII4_ECOBD] C6EAB5 Translation elongation factor Ts OS=Escherichia coli (strain B / BL21-DE3) 222.18 24.73 1 6 6 10 283 30.4 5.29 GN=ECBD_3449 PE=3 SV=1 - [C6EAB5_ECOBD] C6ECY6 Aspartate ammonia-lyase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3892 206.81 17.78 1 6 7 10 478 52.3 5.29 PE=4 SV=1 - [C6ECY6_ECOBD] C6EE34 50S ribosomal protein L10 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_4048 200.67 31.52 1 4 4 4 165 17.7 8.98 PE=3 SV=1 - [C6EE34_ECOBD] C6ECA9 Alpha,alpha-phosphotrehalase OS=Escherichia coli (strain B / BL21-DE3) 195.51 11.62 1 5 5 6 551 63.8 5.87 GN=ECBD_3794 PE=4 SV=1 - [C6ECA9_ECOBD] C6EI89 Seryl-tRNA synthetase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_2702 192.63 10.47 1 4 4 5 430 48.4 5.50 PE=3 SV=1 - [C6EI89_ECOBD] C6EE35 Ribosomal protein L1 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_4049 PE=3 191.29 29.06 1 5 5 7 234 24.7 9.64 SV=1 - [C6EE35_ECOBD] C6EK89 Peroxiredoxin OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3047 PE=1 SV=1 - 190.58 32.09 1 4 4 7 187 20.7 5.17 [C6EK89_ECOBD] C6EEX2 Glycerol kinase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_4098 PE=3 SV=1 184.75 12.95 1 6 6 8 502 56.2 5.50 - [C6EEX2_ECOBD] C6ELH0 Riboflavin biosynthesis protein RibD OS=Escherichia coli (strain B / BL21-DE3) 182.02 16.08 1 4 4 7 367 40.2 7.83 GN=ECBD_3247 PE=4 SV=1 - [C6ELH0_ECOBD] C6EJP6 Phosphoglucomutase, alpha-D-glucose phosphate-specific OS=Escherichia coli (strain B / 170.33 13.92 1 5 5 6 546 58.4 5.71 BL21-DE3) GN=ECBD_2973 PE=4 SV=1 - [C6EJP6_ECOBD]
144 6. Manuscript IV (In-silico prediction of microcystin toxicity)
C6EI54 Aspartate transaminase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_2667 166.81 14.14 1 5 5 5 396 43.5 5.77 PE=4 SV=1 - [C6EI54_ECOBD] H6VRF8 Keratin 1 OS=Homo sapiens GN=KRT1 PE=3 SV=1 - [H6VRF8_HUMAN] 164.03 6.52 8 4 4 4 644 66.0 8.12 C6EE66 Catalase/peroxidase HPI OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_4081 160.52 11.16 1 5 6 6 726 80.0 5.31 PE=3 SV=1 - [C6EE66_ECOBD] C6EGH1 DNA-directed RNA polymerase, alpha subunit OS=Escherichia coli (strain B / BL21-DE3) 160.52 28.57 1 5 7 8 329 36.5 5.06 GN=ECBD_0456 PE=3 SV=1 - [C6EGH1_ECOBD] C6EK88 Alkyl hydroperoxide reductase, F subunit OS=Escherichia coli (strain B / BL21-DE3) 147.66 10.55 1 4 4 4 531 57.4 5.69 GN=ECBD_3046 PE=3 SV=1 - [C6EK88_ECOBD] C6EDZ9 Glucose-6-phosphate isomerase OS=Escherichia coli (strain B / BL21-DE3) 144.54 10.75 1 3 4 4 549 61.5 6.29 GN=ECBD_4012 PE=3 SV=1 - [C6EDZ9_ECOBD] C6EG37 Ketol-acid reductoisomerase OS=Escherichia coli (strain B / BL21-DE3) GN=ilvC PE=3 137.36 11.61 1 5 5 5 491 54.0 5.31 SV=1 - [C6EG37_ECOBD] C6EKK1 Inosine-5'-monophosphate dehydrogenase OS=Escherichia coli (strain B / BL21-DE3) 135.63 15.57 1 4 4 5 488 52.0 6.42 GN=ECBD_1178 PE=3 SV=1 - [C6EKK1_ECOBD] C6EGH0 Ribosomal protein S4 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0455 PE=3 135.16 20.87 1 4 4 5 206 23.5 10.05 SV=1 - [C6EGH0_ECOBD] C6EE78 Heat shock protein HslVU, ATPase subunit HslU OS=Escherichia coli (strain B / BL21- 134.17 9.71 1 3 4 6 443 49.6 5.35 DE3) GN=ECBD_4093 PE=3 SV=1 - [C6EE78_ECOBD] C6EG69 ATP synthase F1, alpha subunit OS=Escherichia coli (strain B / BL21-DE3) 131.99 11.31 1 4 4 6 513 55.2 6.13 GN=ECBD_4298 PE=3 SV=1 - [C6EG69_ECOBD] C6EH56 NusA antitermination factor OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0571 131.09 5.25 1 2 2 4 495 54.8 4.64 PE=3 SV=1 - [C6EH56_ECOBD] C6EGF2 Ribosomal protein S3 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0437 PE=3 124.61 30.47 1 3 4 4 233 26.0 10.27 SV=1 - [C6EGF2_ECOBD] C6EAJ7 Methionine--tRNA ligase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_1544 118.56 7.68 1 4 4 4 677 76.2 5.94 PE=3 SV=1 - [C6EAJ7_ECOBD] C6EJK5 Cytochrome bd ubiquinol oxidase subunit I OS=Escherichia coli (strain B / BL21-DE3) 116.12 8.24 1 3 3 3 522 58.2 6.81 GN=ECBD_2928 PE=4 SV=1 - [C6EJK5_ECOBD] C6EGQ9 Ribose-phosphate pyrophosphokinase OS=Escherichia coli (strain B / BL21-DE3) GN=prs 113.00 7.42 1 2 2 2 337 36.6 5.74 PE=3 SV=1 - [C6EGQ9_ECOBD] C6EGG9 Ribosomal protein S11 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0454 PE=3 112.01 18.60 1 1 2 3 129 13.8 11.33 SV=1 - [C6EGG9_ECOBD] C6EG40 Dihydroxy-acid dehydratase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_4268 111.31 6.01 1 2 3 4 616 65.5 6.01 PE=3 SV=1 - [C6EG40_ECOBD] C6EGM4 Extracellular solute-binding protein family 5 OS=Escherichia coli (strain B / BL21-DE3) 109.66 6.63 1 3 3 3 558 62.6 6.37 GN=ECBD_2379 PE=4 SV=1 - [C6EGM4_ECOBD]
145 6. Manuscript IV (In-silico prediction of microcystin toxicity)
C6EG30 Transcription termination factor Rho OS=Escherichia coli (strain B / BL21-DE3) 109.07 14.32 1 4 4 4 419 47.0 7.25 GN=ECBD_4257 PE=3 SV=1 - [C6EG30_ECOBD] C6EJI9 2,3-bisphosphoglycerate-dependent phosphoglycerate mutase OS=Escherichia coli (strain B 108.50 14.00 1 3 3 3 250 28.5 6.18 / BL21-DE3) GN=gpmA PE=3 SV=1 - [C6EJI9_ECOBD] C6EBA1 Molecular chaperone Hsp31 and glyoxalase 3 OS=Escherichia coli (strain B / BL21-DE3) 107.23 7.77 1 2 2 2 283 31.2 6.02 GN=hchA PE=2 SV=1 - [C6EBA1_ECOBD] C6EJQ4 Glutaminyl-tRNA synthetase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_2981 104.01 10.65 1 4 5 6 554 63.4 6.28 PE=3 SV=1 - [C6EJQ4_ECOBD] C6EGE9 Ribosomal protein L2 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0434 PE=3 100.49 7.69 1 2 2 2 273 29.8 10.93 SV=1 - [C6EGE9_ECOBD] C6ECC2 Inorganic diphosphatase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3807 95.13 30.68 1 4 5 6 176 19.7 5.17 PE=3 SV=1 - [C6ECC2_ECOBD] C6EGB5 Peptidylprolyl isomerase FKBP-type OS=Escherichia coli (strain B / BL21-DE3) 95.00 11.73 1 2 2 2 196 20.8 5.05 GN=ECBD_0400 PE=4 SV=1 - [C6EGB5_ECOBD] C6ECH2 Adenylosuccinate synthetase OS=Escherichia coli (strain B / BL21-DE3) GN=purA PE=3 94.31 10.88 1 3 4 5 432 47.3 5.49 SV=1 - [C6ECH2_ECOBD] C6EJL0 Succinyl-CoA synthetase, beta subunit OS=Escherichia coli (strain B / BL21-DE3) 92.27 6.19 1 2 2 2 388 41.4 5.52 GN=ECBD_2933 PE=3 SV=1 - [C6EJL0_ECOBD] C6EBX6 Glucose-6-phosphate 1-dehydrogenase OS=Escherichia coli (strain B / BL21-DE3) 90.65 9.57 1 4 4 4 491 55.7 5.76 GN=ECBD_1787 PE=3 SV=1 - [C6EBX6_ECOBD] C6EJL1 Dihydrolipoyllysine-residue succinyltransferase component of 2-oxoglutarate dehydrogenase complex OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_2934 90.47 6.67 1 2 2 2 405 44.0 5.81 PE=3 SV=1 - [C6EJL1_ECOBD] C6EAL9 Alcohol dehydrogenase GroES domain protein OS=Escherichia coli (strain B / BL21-DE3) 86.89 6.07 1 2 2 2 346 37.4 6.38 GN=ECBD_1566 PE=4 SV=1 - [C6EAL9_ECOBD] C6EKL0 Polyphosphate kinase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_1187 PE=3 79.87 4.36 1 2 2 2 688 80.4 8.92 SV=1 - [C6EKL0_ECOBD] C6EE36 Ribosomal protein L11 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_4050 75.65 16.20 1 2 2 3 142 14.9 9.63 PE=3 SV=1 - [C6EE36_ECOBD] C6EE00 Aspartate kinase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_4013 PE=4 SV=1 72.24 4.90 1 2 2 2 449 48.5 5.11 - [C6EE00_ECOBD] C6EBW1 Aspartyl-tRNA synthetase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_1772 72.19 3.90 1 2 2 3 590 65.9 5.69 PE=4 SV=1 - [C6EBW1_ECOBD] C6EI04 Glutathionylspermidine amidase OS=Escherichia coli (strain B / BL21-DE3) 71.67 4.36 1 2 2 2 619 70.5 5.26 GN=ECBD_0750 PE=4 SV=1 - [C6EI04_ECOBD] C6EKK2 GMP synthase [glutamine-hydrolyzing] OS=Escherichia coli (strain B / BL21-DE3) 63.84 7.24 1 2 3 4 525 58.6 5.39 GN=ECBD_1179 PE=3 SV=1 - [C6EKK2_ECOBD]
146 6. Manuscript IV (In-silico prediction of microcystin toxicity)
C6EFE8 Glycerol-3-phosphate dehydrogenase OS=Escherichia coli (strain B / BL21-DE3) 62.14 4.59 1 2 2 2 501 56.7 7.44 GN=ECBD_0316 PE=3 SV=1 - [C6EFE8_ECOBD] C6EKF6 Glycine hydroxymethyltransferase OS=Escherichia coli (strain B / BL21-DE3) 60.73 7.43 1 2 2 2 417 45.3 6.48 GN=ECBD_1133 PE=3 SV=1 - [C6EKF6_ECOBD] C6EBX3 Pyruvate kinase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_1784 PE=4 SV=1 60.20 7.50 1 1 3 3 480 51.3 6.68 - [C6EBX3_ECOBD] C6EGH2 Ribosomal protein L17 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0457 57.55 18.11 1 2 2 2 127 14.4 11.05 PE=3 SV=1 - [C6EGH2_ECOBD] C6EJL8 Citrate synthase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_2941 PE=3 SV=1 57.31 5.15 1 1 2 2 427 48.0 6.68 - [C6EJL8_ECOBD] C6EHZ8 Cytochrome-c3 hydrogenase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0744 57.27 6.53 1 2 2 2 567 62.5 6.28 PE=4 SV=1 - [C6EHZ8_ECOBD] C6EAB9 2,3,4,5-tetrahydropyridine-2,6-dicarboxylate N-succinyltransferase OS=Escherichia coli 56.81 8.03 1 2 2 2 274 29.9 5.74 (strain B / BL21-DE3) GN=ECBD_3453 PE=3 SV=1 - [C6EAB9_ECOBD] C6EHB9 3-oxoacyl-(Acyl-carrier-protein) synthase 2 OS=Escherichia coli (strain B / BL21-DE3) 55.41 13.56 1 1 3 3 413 43.0 6.09 GN=ECBD_2506 PE=4 SV=1 - [C6EHB9_ECOBD] C6EKZ0 Adenylate kinase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3182 PE=3 54.54 10.28 1 1 2 2 214 23.6 5.76 SV=1 - [C6EKZ0_ECOBD] C6EB56 6-phosphogluconate dehydrogenase, decarboxylating OS=Escherichia coli (strain B / BL21- 50.74 5.98 1 3 3 3 468 51.5 5.07 DE3) GN=ECBD_1630 PE=3 SV=1 - [C6EB56_ECOBD] C6EH63 DEAD/DEAH box helicase domain protein OS=Escherichia coli (strain B / BL21-DE3) 46.89 3.82 1 2 2 2 629 70.5 8.72 GN=ECBD_0578 PE=3 SV=1 - [C6EH63_ECOBD] C6EGF8 Ribosomal protein L5 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0443 PE=3 44.64 26.26 1 2 3 3 179 20.3 9.48 SV=1 - [C6EGF8_ECOBD] P02533 Keratin, type I cytoskeletal 14 OS=Homo sapiens GN=KRT14 PE=1 SV=4 - 44.54 3.81 2 1 2 3 472 51.5 5.16 [K1C14_HUMAN] C6EGE5 Ribosomal protein S10 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_0430 PE=3 39.15 20.39 1 2 2 2 103 11.7 9.69 SV=1 - [C6EGE5_ECOBD] C6ECB1 Anaerobic ribonucleoside-triphosphate reductase OS=Escherichia coli (strain B / BL21- 36.70 2.11 1 1 2 2 712 80.0 6.83 DE3) GN=ECBD_3796 PE=4 SV=1 - [C6ECB1_ECOBD] C6EAB6 Ribosomal protein S2 OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_3450 PE=3 31.42 7.88 1 2 2 3 241 26.7 7.14 SV=1 - [C6EAB6_ECOBD] C6EBJ7 Peptide chain release factor 3 OS=Escherichia coli (strain B / BL21-DE3) 31.18 4.73 1 1 2 2 529 59.5 5.97 GN=ECBD_3646 PE=3 SV=1 - [C6EBJ7_ECOBD] C6EKI4 PepB aminopeptidase OS=Escherichia coli (strain B / BL21-DE3) GN=ECBD_1161 PE=3 30.83 4.45 1 1 2 2 427 46.2 5.77 SV=1 - [C6EKI4_ECOBD] First two hits were PPP5 and thioredoxin, which are both part of the expressed fusion protein. Other proteins were found in lower quantity.
147 6. Manuscript IV (In-silico prediction of microcystin toxicity)
Table S6-3: Summary of settings of different machine learning models trained. Feature ML algorithm Tuned hyper-parameters Number of estimators Maximum tree depth Mol2vec + ProtVec RF 50 5 Mol2vec RF 50 5 Mol2vec GBM 100 3 The following abbreviations were used: ML: Machine Learning, RF: Random Forests, GBM: Gradient Boosting Machines
Table S6-4: Evaluation of machine learning performance (mean ± standard deviation) on different classes. Class precision recall F-score CV Combined 0 (non-toxic) 0.81 ± 0.15 0.80 ± 0.16 0.79 ± 0.10 Model 1 (less toxic) 0.87 ± 0.11 0.87 ± 0.15 0.86 ± 0.10 2 (toxic) 0.92 ± 0.87 ± 0.17 0.88 ± 0.10 0.10 RF 0 (non-toxic) 0.80 ± 0.14 0.81 ± 0.16 0.79 ± 0.11 (Mol2vec+ 1 (less toxic) 0.89 ± 0.11 0.90 ± 0.13 0.88 ± 0.09 ProtVec) 2 (toxic) 0.91 ± 0.11 0.83 ± 0.16 0.86 ± 0.11 RF 0 (non-toxic) 0.80 ± 0.15 0.80 ± 0.16 0.78 ± 0.10 (Mol2vec) 1 (less toxic) 0.87 ± 0.11 0.85 ± 0.15 0.85 ± 0.10 2 (toxic) 0.92 ± 0.10 0.87 ± 0.17 0.88 ± 0.11 XGB 0 (non-toxic) 0.81 ± 0.14 0.78 ± 0.16 0.78 ± 0.11 (Mol2vec) 1 (less toxic) 0.86 ± 0.11 0.87 ± 0.14 0.85 ± 0.10 2 (toxic) 0.92 ± 0.10 0.87 ± 0.15 0.88 ± 0.10 80 /20 Combined 0 (non-toxic) 0.80 ± 0.15 0.79 ± 0.17 0.77 ± 0.11 Model 1 (less toxic) 0.82 ± 0.16 0.86 ± 0.14 0.83 ± 0.11 2 (toxic) 0.91 ± 0.11 0.88 ± 0.15 0.88 ± 0.11 RF 0 (non-toxic) 0.80 ± 0.17 0.80 ± 0.16 0.78 ± 0.12 (Mol2vec+ 1 (less toxic) 0.84 ± 0.16 0.90 ± 0.15 0.85 ± 0.12 ProtVec) 2 (toxic) 0.89 ± 0.12 0.83 ± 0.16 0.85 ± 0.12 RF 0 (non-toxic) 0.79 ± 0.16 0.79 ± 0.18 0.77 ± 0.11 (Mol2vec) 1 (less toxic) 0.83 ± 0.16 0.85 ± 0.14 0.83 ± 0.10 2 (toxic) 0.91 ± 0.11 0.88 ± 0.15 0.88 ± 0.11 XGB 0 (non-toxic) 0.81 ± 0.16 0.78 ± 0.17 0.78 ± 0.11 (Mol2vec) 1 (less toxic) 0.82 ± 0.16 0.87 ± 0.13 0.83 ± 0.10 2 (toxic) 0.89 ± 0.12 0.88 ± 0.14 0.88 ± 0.11
148 6. Manuscript IV (In-silico prediction of microcystin toxicity)
6.7 Acknowledgements
We gratefully acknowledge the Arthur-und-Aenne-Feindt foundation (Hamburg, Germany), the Konstanz Research School Chemical Biology (KoRS-CB) and CHARM (BW- Wassernetzwerk, Baden-Württemberg, Germany) for financial support.
149 7. General discussion
7. General discussion
7.1 Recommendations for the handling of microcystin-containing samples
The presented work tries to give recommendations for laboratory handling of microcystins and how to avoid inaccuracies before and during analytical processes. In the past, a few studies highlighted, that MCs can be adsorbed on surfaces of common laboratory plastics e.g. polypropylene or polystyrene. In general, those plastics provide a hydrophobic surface to which rather hydrophobic congeners like MC-LF, MC-LA or MC-LW tend to adsorb (Heussner et al. 2014a). Other studies demonstrate, that adsorption is dependent on the percentage of methanol or acetonitrile (Hyenstrand et al. 2001a; Hyenstrand et al. 2001b). Interestingly, acetonitrile displays an optimum at around 50%. Lower and higher concentrations lead to adsorption. In contrast, methanol decreases adsorption linearly up to around 60% and then does not further increase or decrease afterwards (Hyenstrand et al. 2001a). Furthermore, differences in adsorption have been observed after storage in differently pure aqueous solutions. Adsorption of MC-LR, MC-LA and MC-LF to polypropylene was highest in pure, deionized water, although lesser for MC-LR than for the other two. When MCs were dissolved in drinking water adsorption was decreased for all of the three tested MC, although MC-LF still presented with strong adsorption. In contrast, all three MCs displayed greatly reduced adsorption (recovery up to 100%) when dissolved in surface water from an east American reservoir (Kamp et al. 2016). An explanation here could be that deionized water obviously contains only a minimum of additional substances, compared to drinking water and especially surface water. Present substances might on the one hand lead to different ionization of MCs, thus decreasing hydrophobic interactions with the plastic surface, or on the other hand might saturate possible binding sites thereby preventing MC adsorption. Nevertheless, during laboratory handling, pure solvents are favourable, as downstream analytical methods might display interference with present impurities. Thus, it was recommended to store and handle MC solutions in glass containers in methanol and dilution should only take place immediately before analysis if necessary (Heussner et al. 2014a). In addition to the adsorption to plastic ware, Rogers et al. (2015) observed MC adsorption to glass-fibre material (GF/C filters) in methanolic extracts, which could be mostly alleviated by adding low amounts of formic acid to the methanol. Interestingly, the work presented here, showed somewhat different results (see section 3 or Altaner et al. (2017)). Here,
150 7. General discussion
MCs dissolved in acidic aqueous and low-percentage methanol solutions generally adsorbed more readily to glass and plastic surfaces during pipetting than when solved in non-acidified solvents. However, an increase in the methanol content to above 40% could abolish the adsorption to glass, as well as plastic surfaces, irrespective of acidification. Thus, to avoid loss of MCs due to adsorption and hence inaccuracies during following analysis, several recommendations for the handling of microcystin-containing solutions are given: 1.) Optimally, use at least 40% methanol as a solvent during any step whenever possible. If lower methanol concentrations have to be used, e.g. during ELISA analysis, 2.) Use dilutions of methanol containing samples but perform any storage in methanol containing solvents. Dilution should be performed immediately before subsequent analysis. However, sometimes dilution of the sample does not work, e.g. when using samples with low MC content. In such a case, 3.) Avoid acidification of the sample. 4.) Avoid polypropylene, whenever possible. The use of glass vials (e.g. LC vials) for storage purposes instead of Eppendorf-reaction tubes is recommended. 5.) However, also the handling in methanol might be problematic, as it has a low partial pressure, thus it evaporates more easily thereby changing concentrations of the dissolved content. Hence, handled containers should be cooled properly, e.g. by working on ice and should be closed immediately after aspiration to avoid evaporation.
7.2 Optimization of microcystin extraction from tissue
Since the first report of toxicity of cyanobacteria in the 19th century (Francis 1878) a lot of research effort has been invested in methods that are able to detect MCs in various matrices such as lake water, various tissue of various species and in the producing cyanobacteria themselves. Hence, many different extraction methods have been proposed which were then subsequently coupled to varying detection methods like ELISA or chromatographic methods. Here, obviously the complexity of the extraction is determined by the complexity of the matrix from which analytes should be separated. Blood and tissue homogenates are very complex matrices. They contain both a lot of protein and lipids, stemming mostly from membranes (around 10% dry weight of cells) (Cooper and Hausman 2000). Therefore, these components
151 7. General discussion may be detected non-specifically by accident and hence may generate a background during the detection of specific analytes, i.e. microcystins. One of the first used methods comprised just a dilution step of blood or organ homogenate followed by a simple solid phase extraction step, using commercially available C18 columns and ultimately an ELISA method for detection (Lin and Chu 1994). Tencalla and Dietrich (1997) added a centrifugation step after homogenization of liver tissue to remove cell debris, before MC quantification. Another group successfully used a simple extraction method consisting only of freeze-drying tissue material, dissolving it again in methanol while being sonicated in a water bath before filtering through 0.45 µm filters (Sipiä et al. 2001). For the extraction of MC from human samples which stemmed from the patients of the Caruaru incident, a three step extraction was proposed: Precipitation of proteins with methanol and centrifugation, liquid-liquid partitioning of the methanol extract with n-hexane followed by a solid phase extraction using specialized C18 columns. With this method the researchers were able to employ ELISA and chromatographic methods on serum samples (Hilborn et al. 2005; Yuan et al. 2006). This method later on was used for further ELISA experiment and some improvements were made to the protocol to minimize analyte loss during the extraction. These improvements consisted of using glass containers where possible and keeping the solvent amount high until directly before analysis (Heussner et al. 2014a; Heussner et al. 2014b). Laboratories have used also combinations and modifications (less/additional steps) of the described methods, hence there is no universally applicable method proposed so far. Researchers of other fields (pesticides, mycotoxins, (veterinary) drugs) have similar problems in regard to many different used extraction procedures and analyte recovery. Thus, Mol et al. (2008) proposed a generic protocol for extraction of those compounds from various food matrices. In their publication, they propose a method, which has been adopted, although slightly modified, by researchers in the microcystin field for their extraction from fish tissue. To homogenized tissue samples 75% acetonitrile was given, incubated, centrifuged and the supernatant subjected to liquid-liquid partitioning with hexane before analysis, without an additional SPE step (Geis-Asteggiante et al. 2011). The extraction procedure, that was used in the present work (described in the third manuscript, section 5) was generally based on the approach of Heussner et al. (2014b), but was shortened and the above mentioned recommendations (see section 7.1) were incorporated where possible. Consequently, acidification of the sample for solid phase extraction (SPE) by solvation of the dried sample in 5% acetic acid was replaced with simply diluting the sample to contain below 10% methanol just before SPE. On the one hand, this avoids adsorption of MC
152 7. General discussion to the used glass container due to acidification, but on the other hand also minimizes the time needed, as previously samples were dried overnight. In order to further speed up the process, both, the protein precipitation step as well as the liquid-liquid-partitioning step with hexane were each only performed once without repetition as done previously. Hydrophobic congeners (e.g. MC-LF and -LW) might interact with cellular membranes, as they have been shown to interact with lipid monolayers to a higher degree than hydrophilic MC-LR (Vesterkvist and Meriluoto 2003). This could mean, that those might be discarded when only the aqueous intracellular fraction or the aqueous fraction of blood from the liquid-liquid partitioning step are analysed. However, it has been demonstrated that microcystins (hydrophobic and hydrophilic) do not go over to the hexane phase, in a liquid-liquid-partitioning situation and thus analyte loss due to hydrophobic interactions is unlikely. Nevertheless, some material might be lost due to pelleting the membrane fraction, as adsorption to membrane bound proteins might occur. The SPE procedure finally was similar to the previously used one by Heussner et al. (2014a). Although different SPE columns were tested, still Oasis HLB columns are recommended. As discussed earlier, samples were not loaded in acidic solvents but in low percentage methanol, thus column equilibration was also performed with 10% methanol. The final method comprised two washing steps: One with MilliQ and one with 20% methanol in order to achieve a cleaner extract. Sample elution was performed with at least 10 volumes of 80% methanol as previously, thus still requiring a final concentration step using a SpeedVac. Generally, the proposed method still takes one day of laboratory time and a final concentration step, which optimally is performed overnight (time needed in SpeedVac: ~ 8h) and sample reconstitution on the second day. Nevertheless, laboratory time until analysis is effectively halved and analysis can be performed on the second day. Thus, also storage time is efficiently decreased resulting in fewer adsorption events but also decreases turnaround time until a sample is analysed, but keeps the advantage of three independent extractions resulting in a cleaner sample. Immunoaffinity columns equipped with antibodies against MCs, could be an alternative to the classically used SPE method and have been successfully employed in the clean-up of MCs from environmental samples (Aranda-Rodriguez et al. 2003; Kondo et al. 2005; Kondo et al. 2002; Samdal et al. 2018). They might provide cleaner samples, due to specific interactions with MCs, but on the other hand bear the risk of missing congeners, which do not react with the used antibody. This could be alleviated, when antibodies are used, which react with many different MC for example when they are raised against structural elements of the MC scaffold
153 7. General discussion present in most congeners, such as the Adda chain. Such an antibody was produced by Fischer et al. (2001) for the use in ELISA methods but until now not used for immunoaffinity purification of MCs. Although the extraction here is considered a good improvement over previous methods, it still has a major drawback that is actually common to many methods used for MCs: Only free MCs are extracted but due to their ability to covalently bind proteins via thiol conjugation, e.g. to PPPs, glutathione or probably also albumin, a major portion of MCs might be co-precipitated with the proteins during the addition of high-percentage methanol at the start of the extraction and hence are not available for detection. Ideally, a further step should be included that aids in the deconjugation of the thiol bonds formed between protein and MCs. Different methods have been proven to be able to deconjugate the formed thiol bond using either oxidation (Miles 2017) or alkaline treatment (Miles et al. 2016b; Zemskov et al. 2016). These should be incorporated as a first step during extraction of biological material. Miles (2017) proposed to include the oxidation method into extraction as it might lead to less analytical artefacts and suggested to additionally include a proteolytic step to decrease the size of peptides thus increasing the yield. This seems to be a feasible approach, but may increase the turnaround time of the complete procedure again, as proteolysis (e.g. tryptic digest) and deconjugation could probably not be performed on the same day as the extraction due to the prolonged incubation times needed. Nevertheless, it would provide for a much better picture of the MC content in a given sample.
7.3 UPLC-MS/MS analysis
Along with the extraction method, an UPLC-MS/MS based detection and quantification method was presented in the work at hand. Chromatographic methods have been used for quite some time in the detection of MCs and also recently fast and reliable UPLC-MS/MS methods have been published (Greer et al. 2016; Manubolu et al. 2018; Turner et al. 2017), nevertheless the present method displays some advantages over these. UPLC methods bear the possibility of analysing several analytes at the same time, due to being very sensitive. Thus the aim was, to establish detection and quantitation for many MC congeners in one analytical run. A similar approach was taken by Greer et al. (2016) who included six congeners (MC-LR, -YR, -RR, - LA, -LY and -LF) in their method and then later on by Turner et al. (2017) who added MC- LW, -WR, -HilR, -HtyR as well as [Asp3]MC-LR and [Dha7]MC-LR, thus being able to detect and quantify twelve congeners. In here, we report the possibility to detect and quantify 14 individual congeners (MC-RR, -YR, -LR, -FR, -WR, -RA, -Raba, -LA, -FA, -WA, -LAba, - FAba, MC-WAba and -LF). When these two works are combined in one method using the MS-
154 7. General discussion parameters published, they would be able to detect 18 congeners. The downside of detecting many congeners at once, is the loss of analytical sensitivity for each individual analyte, because per time unit fewer fragments of each molecule pass the detector, i.e. dwell time for each analyte is decreased. To reduce this problem, analytical windows were used in a way that not every congener is monitored over the whole period of the analytical run, but only around its expected retention time, thereby slightly increasing the limit of detection for all congeners. Additional smaller windows could be added to increase individual dwell time for each conger analysed at the risk of missing congeners when their actual retention time is deviated from the expected retention time. Puddick et al. (2016b) established UPLC-MS/MS detection of even more congeners and indeed the presented work is an adaptation of their method. At the beginning of method development here, all congeners searched for in their method were also part of the method at hand, but some were neglected due to very low amounts present in the used extract for establishment and thus non-reliable results during validation and establishment. Theoretically, it would be easy to include them again thereby increasing simultaneously detected congeners. The same is true for the MC-related toxin nodularin. The major advantage of the present method however is the use of internal standards, which are structurally identical to two naturally occurring MC congeners from the hydrophilic
(D7-MC-LR) as well as the rather hydrophobic (D5-MC-LF) spectrum. Structural identity was achieved by full synthesis of MC-LR and MC-LF both containing deuterium instead of hydrogen in some parts. In D5-MC-LF, five deuterium atoms were incorporated at the aromatic ring structure of the phenylalanine while D7-MC-LR was modified with seven deuterium atoms at the leucine residue in position 2. Although other studies used internal standards (Mallet 2017; Roegner et al. 2014b; Smith and Boyer 2009), none used fully synthesized and thus structurally identical ones. The advantage of structurally identical standards is the same behaviour during handling and analytical procedures, i.e. they are experiencing the same effects like adsorption, ionisation or signal enhancement/suppression. It is discussed that deuterium may alter hydrophobicity and hence retention times, and therefore rather other stable isotopes like 13C or 15N should be used (Stokvis et al. 2005; Wang et al. 2007), nevertheless this could not be observed here. As the internal standard is spiked into samples before extraction at a known concentration which can be specified during the analytical run, it serves as a control for loss during the whole procedure. The underlying assumption here being, that the internal standard behaves like the actual analytes (see above). This is somewhat skewed when much more analyte than internal standard is present, because small losses result in a bigger percentile difference when small or high amounts are compared, e.g. 1 ng loss of 10 ng represents 10%, while 1 ng
155 7. General discussion of 100 ng is only 1%. Nevertheless, the used internal standards improved analytical precision and reproducibility. Neffling et al. (2010b) described MC congener dependent signal suppression in serum. Arginine containing congeners showed less suppression while hydrophobic congeners (MC- LY, -LW and -LF) were suppressed to a degree they could not even be detected anymore. The authors discuss that stronger matrix effects (i.e. more noise) were observed around the elution times of these congeners. They also point out that inefficient ionization of those congeners may lead to lower recovery. However, the loss due to signal suppression/enhancement can be effectively corrected when internal standards are employed. It has been suggested, that for mass spectrometric analyses of hydrophobic MC congeners, negative ionization method is preferable, as it provides better detection, i.e. higher sensitivity (Ferranti et al. 2009; Gambaro et al. 2012). Nevertheless, here all congeners where measured in positive ion mode, as it is not possible to switch ionization mode during one run. Kohoutek et al. (2010) found that a LC-MS method, only relying on the m/z ratio of parent microcystins (RR, -LR and -YR), is more prone to false positives in complex matrices, as similar eluting peaks with similar masses to the MC are mistakenly quantified as MC. It was found, that using the LC-MS/MS method, relying on two fragments (parent and daughter ion) of each compound, shows a better signal-to-noise ratio. Thus the simpler LC-MS method overestimated the concentration about 2-fold. Furthermore, using multiple fragments for one + congeners, including [M+NH4 ] when appropriate solvents are used, is encouraged by some researchers to increase specificity (Draper et al. 2009). Due to the already high number of simultaneous detection events, our method uses only one daughter fragment produced from every parent ion, to increase dwell time for each fragment which should lead to a more stable detection (see above).
7.4 Serine-/Threonine-Proteinphosphatase (PPP) inhibition
Manuscript IV (see section 6) shifts from the detection and quantitation of MCs towards their toxicity i.e. their inhibition capacity of ser/thr-PPPs. The family of PPPs is evolutionary very old and PPP1 and PPP2A are evolving very slowly (Cohen 2004). They can be found in many branches of the evolutionary tree, e.g. PPP1 in mammalians are 80% similar in sequence to PPP1 from yeast and 60% similar to plants (Bastan et al. 2014) and can presumably even be found in prokaryotes (Macek et al. 2008). This underlines the important role of PPPs for cellular functionality and explains that interference with phosphorylation of homeostasis leads to cellular stress.
156 7. General discussion
Here, a comprehensive dataset was compiled that integrates data for the inhibition of three PPPs with 18 structurally diverse MC congeners. The three chosen PPPs (PPP1, PPP2A and PPP5) stem from three distinct groups among the PPP family, as PPP1 functions as a heterodimer, PPP2A as a heterotrimer and PPP5 as a single peptide. Here, only the catalytic subunits of the PPPs were used for inhibition tests with the different MC. The catalytic subunits are expressed ubiquitously throughout the human body (Pereira et al. 2011a). The interaction of the catalytic subunit with regulatory subunits is responsible for the substrate specificity. Regulation of PPPs by their respective subunits is very complex due to various reasons. Firstly, often many regulatory subunits are known which are expressed differentially in tissues thus allowing for differences in dephosphorylation in the respective organs, e.g. 13 regulatory subunits of PP2A are known, which can be found in various tissues (Seshacharyulu et al. 2013). Secondly, the expression of the regulatory subunits themselves may be regulated by external and internal factors. And finally, the holoenzymes themselves can be post-translationally modified by phosphorylations and/or methylations thereby altering activity, e.g. methylation of the catalytic subunit of PPP2A leads to increased binding to the regulatory subunits (Cho and Xu 2007). Nevertheless, all of the catalytic subunits used here were efficiently inhibited by various MCs, thus demonstrating the possibility of inhibition of the holoenzyme, especially as it has been shown that the holoenzymes do not change significantly upon inhibition of their catalytic subunits, but their interaction networks changes (Yadav et al. 2017). The study at hand used naturally occurring MCs, but also synthetic variants that display specific structural changes and thus allow for investigation of important structural elements regarding PPP inhibition. The obtained data corroborate the importance of the Adda-residue. A missing or reduced Adda-chain greatly decreases toxicity towards all tested PPPs, while stereoisomers of Adda do not show any toxicity. Interestingly, a congener’s inability for covalent inhibition of the PPPs did not reduce inhibition, underpinning that covalent inhibition is not necessary for MC toxicity. The work also highlights, that although similar inhibition patterns among the PPPs were observed, inhibition by individual congeners towards the individual PPP may differ. The used PPPs were either commercially available (PPP1/PPP2A), or heterologously expressed (PPP5). Interestingly, some sources state that bacterially expressed PPP5 is essentially inactive (Bastan et al. 2014; Chen and Cohen 1997; Chen et al. 1994). Activity could be stimulated in these studies with either partial proteolysis or addition of polyunsaturated fatty acids (Chen and Cohen 1997). Nevertheless, here active human PPP5 was expressed in bacteria without the need of activation. One possible explanation might be the addition of various tags
157 7. General discussion to the N-terminus, which might hinder the inhibition of the active site through the TPR domain. Therefore, it was tested whether using thrombin to cleave off the tags, would render PPP5 inactive. However, even after cleavage PPP5 remained active (data not shown). Cleavage left a small peptide still attached to the N-terminus of the protein (GSGMKETAAAKFERQHMDSPDLGTDDDDK, 29 amino acids), though it was assumed that this does not interfere with the binding of the TPR domain to the active site, as the N-terminus seems to be not important for this interaction (Yang et al. 2005). Several low-abundance impurities of the PPP5 fraction were observed using mass spectrometry and SDS-PAGE, which most likely stem from unspecific interactions during the Ni-NTA purification. However, only negligible phosphatase activity was observed in an empty expression (Figure S6-2). The reason for the discrepancy regarding the active of bacterially expressed PPP5 between this and other studies therefore still remains unknown. Fischer et al. (2010) compared MC-LR, -RR, -LF, and -LA on PPP1 and PPP2A, using the same assay and enzyme sources as the study at hand and indeed obtained similar IC50 for the four congeners and two PPPs. Additionally, a previous study found similar differences between MC-RR and [Asp3,Dhb7]MC-RR on PPP1 (about 30-fold) and PPP2A (about 54- fold), while [Asp3]MC-RR displayed a similar IC50 like MC-RR (Hoeger et al. 2007), which was not the case in the present study. The apparent difference between the two studies cannot be explained, especially as IC50 were produced using the same analytical method, calculations and programs. In a review, Zhang et al. (2013) summarized IC50 of MC-LR for several PPPs including PPP1, PPP2A and PPP5. There MC-LR has a similar IC50 on all three PPP (1 nM), although it is not quite clear, where the values for PPP5 stem from. The 1 nM value fits to the data from this work regarding PPP1 and PPP2A, but is around five-fold lower than the observed value for PPP5 herein (Zhang et al. 2013). Another study finds an IC50 for MC-LR on PPP5 between 1 and 15 nM, which fits to the present results (Yang et al. 2005). An additional study also remarks that PPP5 is likely to be less sensitive towards MC-LR than PPP1 and PPP2A, a fact also obvious in the study at hand (Chen et al. 1994). Garibo et al. (2014b) compared MC- RR, -LR, -YR, -LY, -LW and -LF regarding PPP1 and PPP2A, using the same PPP1, but a different PPP2A. Interestingly, the authors found values for MC-LR and MC-RR about three to ten-fold higher than in the present study, while the IC50 of the other four MC were around 40 to 300-fold higher in their study (Garibo et al. 2014b). It is not obvious why the data differs so massively, but might be explained (at least for MC-LY, -LW and -LF) with adsorption to laboratory-ware during the assay (see section 7.1). It has been published previously, that rather hydrophobic MC adsorb to commonly used plastic lab ware when handled in low percentages
158 7. General discussion of methanol (Altaner et al. 2017; Heussner et al. 2014a). This might lead to apparently increased
IC50 values in the work of Garibo et al. (2014b). In the present study, it was tried to handle MC in as high methanol as possible and using LC-vials instead of 1.5 ml plastic reaction tubes (“Eppendorf-cups”) when dilutions of the MC stocks had to be made. Nevertheless, the reason for the difference in IC50 regarding MC-YR (factor ~57) remains unknown. Interestingly,
Ikehara et al. (2009), who investigated the inhibition of PPP2A by several MCs, found IC50 about 6 – 332 fold different to the ones reported here, but most around 6 to 8 fold different. The
[Asp3,Dhb7]-variant of MC-RR showed the highest discrepancy of about 332 fold to the IC50 presented here. The difference again is not trivial to explain, especially as a similar assays was used. PPP2A was expressed in insect cells in the work of Ikehara et al. (2009), but purified from human blood cells in the present study. It has been discussed, whether the expression system of PPP might play a role in the observed activity (Liang et al. 2018), thus differences could stem from the different expression. Interestingly, there seemed to be a selection of PPP2A over the other two tested PPPs when congeners with shortened Adda residues are employed. This was already previously observed with similar MC variants (Fontanillo and Köhn 2018; Fontanillo et al. 2016). As PPP4 and PPP6 are closest related to PPP2A (Shi 2009), it would be interesting, whether those also display the observed selectivity. Unfortunately, although work is in progress to express PPP4 and PPP6 in sufficient amounts for similar toxicity testing, this effort is not yet fully achieved, as expression seems to be no straight forward (unpublished work Eva Riehle and Jahn Nitschke, Master’s theses). Dietrich and Hoeger (2005) proposed to use TEFs to describe the toxicity of other MC congeners in relation to MC-LR. Garibo et al. (2014b) started to use this concept in their study using six MCs. Consequently, Table 6-3 lists the TEFs generated from the present dataset for the individual phosphatases. Obviously, the TEF depends largely on which PPP is used for the assessment, e.g. the TEF for MC-YR, differs >2-fold between PPP1 and PPP2 and >3-fold between PPP1 and PPP5. Therefore, it is proposed to use the highest observed TEF to provide for a more conservative risk assessment. Most likely, TEFs are changing when toxicokinetic data is also incorporated, as differential transport of congeners has been observed (Fischer et al. 2010). In that regard, Ikehara et al. (2009) noted that the observed order of the toxicity of MCs determined with inhibition assays does not correlate with the toxicity observed with human normal hepatocytes, thus underlining, that differential transport of MCs via OATPs (and potential cellular exporters) plays a pivotal role.
159 7. General discussion
7.5 Computational chemistry and machine learning
There are different possibilities for machine learning algorithms that can be used to predict activity of a compound towards an enzyme. Typically, these are used during drug design, to either find molecules which may be active against a target, or to investigate off-target effects of a drug candidate (Krusemark 2012). For machine learning involving chemical structures, these have to be transferred into features for description and subsequent comparison. In the present work, Mol2vec was used to generate vectors which describe the chemical structure of a molecule. Mol2vec is an adaptation of Word2vec, an algorithm developed for the description of words in a sentence. In Mol2vec, a molecule is represented as sentence, made up by individual substructures (words) (Jaeger et al. 2018). Classical molecular descriptors, e.g. the Morgan fingerprint (MF), have the disadvantage, that information about the compound may be lost during the generation of the fingerprint. Additionally, transferring the MF into a vector results in a very sparse vector, as they are generated via hashing the information of the presence (1) or absence (0) of set of known substructures into a vector. As a certain chemical structure only has a limited number of substructures, most positions in the hashed vectors are represented by 0. This is alleviated by using Mol2vec as here, the information about substructures are not represented by their presence or absence, but all present substructures are extracted as ‘words’ of a ‘sentence’ that represents a molecule. The result here being a vector that represents the whole molecule (‘sentence’) made up of vectors of the substructures (‘words’). This was considered to be advantageous over the MF, as the resulting vector only contains meaningful chemical information for a particular compound, e.g. a microcystin congener. On the other hand, a drawback of the Mol2vec machine learning approach is, that it does not incorporate information about stereochemistry into the generated molecule vector. Thus, as described earlier (see section 6.4), the congeners displaying enantiomeric Adda-components are not recognized as different from the ‘standard’ counterparts (e.g. [enantio-Adda5]-MC-LF vs. MC-LF). This is due to the fact that the vectors for [enantio-Adda5]-MC-LF and MC-LF do not differ, but the associated toxicity differs, which in turn lead to erroneous learning by the algorithm. Nevertheless, Mol2vec was chosen as an appropriate tool for the representation of MC, as it allows for a representation which can be easily combined with other features, such as the chemical structure of the target PPP via ProtVec. Machine learning itself relies on the input data, in the present case, chemical information of MCs from Mol2vec and corresponding data (here, IC50 values against PPP). Machine learning algorithms then try to find patterns in that data that describe the data on the basis of
160 7. General discussion the structural features of the MCs. In the present work, random forest (RF) and gradient boosting machine (GBM) algorithms were used. Both are based on decision trees, but use both more than one tree. RF generates many different trees and uses a majority voting to finally classify a compound (Breiman 2001), while GBM tries to describe the data with different trees, which end up at the same conclusion, i.e. classification (Chen and Guestrin 2016). Furthermore, so-called deep neural networks could be used for the machine learning, but were not employed here, as they tend to overfit data and additionally, the decision for a certain classification cannot be extracted from the algorithm, and are thus not recommended for the use with Mol2vec (Jaeger et al. 2018). Furthermore, SMOTE had to be used to generate additional artificial data points, in order to have enough data for the machine learning task. Thus, complementary data from actual existing MC congeners would be favourable. In order to achieve this, the present machine learning algorithm could be transformed into an online platform, where other researchers could add their data regarding (additional) MC congeners and PPPs, thereby creating a larger data base. Optimally, this platform works iteratively, adding the newly supplied data into the learning set - after being reviewed for quality - thus increasing predictive power.
7.6 Risk assessment and determination of sample toxicity
As described earlier (see section 1.5.1) the current risk assessment for MCs by the WHO suffers some major drawbacks, i.e. it is solely based on MC-LR and the used data is around 20 years old. Thus it was described as ‘provisional’ already when being published in 1999. In the meanwhile, many congeners other than MC-LR have been identified and shown to be at least as toxic as MC-LR or even more effective. Furthermore, MC-LR often is not the most prevalent congener as shown by Birbeck et al. (2019) where MC-LA is found more frequently in blooms and MC-RR shows the highest amounts in individual samples of Michigan water bodies. Similarly, Diez-Quijada et al. (2018b) reviewed, that most cases of animal intoxication did not occur by MC-LR but by MC-LF, MC-LA and MC-YR. Thus, the risk assessment needs to be revised. Currently, the WHO is working on an updated version of said risk assessment, but a publication date is not yet available and no information about the revisions which are applied are published by now. The approaches in the present work may help to improve future assessments in different ways. The toxicodynamic portion of MC effects can be described much more representative as data is available for different congeners. Additionally, the data shows that the various target enzymes (ser/thr-PPPs) are affected differentially. Although not all human PPPs are described
161 7. General discussion here, the used ones represent three distinct groups (see section 7.4). Combining the data, TEFs were established, that describe the toxicity of a given MC congener in relation to MC-LR thus giving the opportunity to consider the differential toxicity of structural MC variants. Of course, the TEFs varies when different PPPs are considered. Nevertheless, for a conservative risk assessment the highest TEF should be used to account for data uncertainties and population variation, e.g. MC-LF displays a TEF of 0.24 on PPP1, 0.66 on PP2A and 2.02 on PPP5, thus 2.02 should be used, as it describes MC-LF to be more toxic than MC-LR, thereby allowing for a prudent assessment. Additionally, in the present work, laboratory handling and procedures for the detection of MCs have been worked over and improved. This allows for more accurate detection of individual MC congeners and thus provides the basis for meaningful use of the TEF concept. In order to properly classify a certain bloom or food source as a risk, researchers and authorities should use a combined approach: A combination of ELISA methods, with UPLC- MS/MS and subsequently use the power of existing databases and possibly even prediction tools. In essence, ELISA methods should be used to quickly screen through samples to determine whether microcystins are present. If so, this sample should be further analysed with UPLC-MS/MS techniques that allow for congener identification. Of course, if appropriate funding and resources are given the initial ELISA step could be omitted and samples could be directly analysed for congener presence. Once it is known which congeners are present in the sample, they can be looked up in a data base listing experimental IC50 values or other toxicological data, which would be necessary to evaluate the risk of a given sample. Using previously established TEFs, the overall expected toxicity of a certain sample can be determined subsequently. Furthermore, in the case of a missing TEF, e.g. due to unavailability of pure material, machine learning can be employed to predict the toxicity of a certain congener based on its chemical structure, thus closing the knowledge gap when more than 200 congeners (Spoof and Catherine 2017) are considered which cannot be tested in the laboratory individually. Thus, employing the presented tools, general risk assessment, as well as, toxicity determination of a given sample, can be improved. Nevertheless, it should not be forgotten, that the toxicodynamic part of the MC toxicity is not complete without the toxicokinetic portion, i.e. the transport of MCs into target cells by OATPs, the cellular export presumably via multi-drug resistance related proteins (e.g. MDR or MRP) and the expression of those transporters in the target cells. Here again, congener- dependent differences are known to occur regarding uptake by OATPs (Feurstein et al. 2009; Fischer et al. 2010; Fischer et al. 2005). Furthermore, preliminary data show that export of MCs
162 7. General discussion by MRP2 also is congener dependent (Kaur and Dietrich 2018). Thus, further research in that direction is crucial for the full description of the toxicity of MCs in general and for each congener individually.
7.7 An example for the necessity of congener dependent observations
Birbeck et al. (2019) analysed MC levels in Michigan lakes with a LC-MS/MS method, which was able to detect multiple MC congeners. They found several positive samples, two of which are presented here in Table 7-1 and Table 7-2. When the total measured amounts are compared with the total amounts after correction using the respective established TEFs with the data for PPP5, which always was the highest TEF, it is obvious that the picture is different for both samples. The Hudson lake sample (Table 7-1) displays lower toxicity after correction using the TEF, while the sample of Houghton lake (Table 7-2) is more toxic than anticipated due to a high amount of MC-LA. Thus, assuming that all congeners are equal to MC-LR regarding their toxicity, or only working with MC-LR analysis may significantly underestimate the actual toxicity of a given sample.
Table 7-1: Application of the TEF concept on a bloom sample from Hudson lake. measured TEF TEF TEF Equivalent using TEF of PPP5
µg/L PPP1 PPP2A PPP5 calculated µg/L
(Asp3)MC-RR 80.02 0.005 0.006 0.006 0.48 MC- RR 1830.26 0.196 0.309 0.435 796.16 MC-YR 860.62 0.236 0.495 0.909 782.30 MC-HtyR n.d. 0.445 n.d. 1.091 0.00 MC-LR 1080.87 1 1 1 1080.87 (Asp3)MC-LR 70.89 0.34 n.d. 0.5 35.45 MC-HilR 44.57 0.486 n.d. 1.202 53.57 MC-WR 62.45 0.218 n.d. 1 62.45 MC-LA 582.95 0.145 0.356 1.087 633.67 MC-LY 23.59 0.362 n.d. 1.238 29.20 MC-LF 12.34 0.243 0.659 2.016 24.88 total 4648.56 3499.03 % of measured 75.27 Data from (Birbeck et al. 2019), sample from Hudson lake, Michigan, USA in August 2016.
163 7. General discussion
Table 7-2: Application of the TEF concept on a bloom sample from Houghton lake. measured TEF TEF TEF Equivalent using TEF of PPP5
µg/L PPP1 PPP2A PPP5 calculated µg/L
(Asp3)MC-RR 0.005 0.006 0.006 0.00 MC-RR 13.05 0.196 0.309 0.435 5.68 MC-YR 0.236 0.495 0.909 0.00 MC-HtyR 0.445 n.d. 1.091 0.00 MC-LR 224.91 1 1 1 224.91 (Asp3)MC-LR 0.34 n.d. 0.5 0.00 MC-HilR 0.486 n.d. 1.202 0.00 MC-WR 0.218 n.d. 1 0.00 MC-LA 1728.62 0.145 0.356 1.087 1879.01 MC-LY 6.26 0.362 n.d. 1.238 7.75 MC-LF 0.243 0.659 2.016 0.00 total 1972.84 2117.35 % of measured 107.32 Data from (Birbeck et al. 2019), sample from Houghton lake, Michigan, USA in August 2016.
Lake Houghton is a popular fishing area, where Crappie (Pomoxis spp.) is a common game species. Crappie seems to be a species that is affected by MCs when living in areas with frequent blooms (Schmidt et al. 2013), and as it is a popular pan fish, the possibility of human exposure is given, although the Crappie samples in that particular study did not exceed safe levels. On the other hand, the levels of MC found in the Michigan lakes by Birbeck et al. (2019) are much higher (20 – 40 fold) than the ones in the study of Schmidt et al. (2013), thus it can be assumed, that also MC levels in fish were higher. Unfortunately, no data about MC contamination of fish species living in the Michigan lakes are available. Nevertheless, the presented data underlines the necessity of congener dependent risk assessment of lake and especially tissue samples.
164 8. Conclusions and outlook
8. Conclusions and outlook
This work deals with different aspects which should be considered when dealing with the health risks of cyanobacterial MCs to humans. Thus, laboratory handling of MCs was investigated to provide for more reproducible results. Additionally, extraction of MCs from complex matrices such as serum/plasma or tissue was examined, optimized and coupled with UPLC-MS/MS method for the simultaneous detection of many different MC congeners. The method furthermore uses internal standards to control for inaccuracies and loss during the extraction and analytical process. Complementary to this work, congener dependent inhibition of major MC targets (PPP1 and PPP2A) along with a lesser researched target (PPP5) was investigated. The generated data subsequently was used to train a machine learning algorithm that correlates structural information of individual congeners to their ability to inhibit PPP1, PPP2A and PPP5, thus creating a possibility for MC congener dependent toxicity prediction. Taken together, the present work tries to improve aspects of the risk assessment for MCs by shifting the focus away from the oversimplified examination of only MC-LR and towards congener dependent observation. Nevertheless, to fully achieve this goal further work is needed, especially regarding toxicokinetics of individual congeners and problems in conjunction with the analysis of protein bound MCs. Additionally, the established prediction tool would benefit from additional data, thus more congeners should be tested e.g. other [Asp3]-variants. A way to achieve this would be an online platform running the classification and prediction algorithm, to allow users worldwide to contribute their data thus increasing precision of the predictions. Although PPP2A is closely related to PPP4 and PPP6, those should also be integrated into the prediction model. Furthermore, detection of MC congeners could and should be improved further. The major drawback of the presented UPLC-MS/MS method is, that the protein-bound fraction, which is assumed to be as large as 80% of the total MC content in biological samples, is not available for detection. Hence, additional extraction steps might be a worthwhile improvement, e.g. including base-catalysed or oxidative deconjugation of thiol bonds, thereby releasing free MCs, which subsequently could be picked up by the MS/MS analysis. Using LC-MS/MS, the challenge will always be to select which congeners are looked for, as even with the current instruments it is not realistically feasible to analyse for all the known congeners and so far non-described congeners in routine analysis. Techniques like SWATH (sequential window acquisition of all theoretical fragment ion spectra) are capable of this task, but compatible instruments are not used in routine analysis (Arnhard et al. 2015; Roemmelt et al. 2014). Therefore, quantitated amounts will always be underestimated.
165 8. Conclusions and outlook
Nevertheless, the work presented here represents a step towards a congener-dependent risk assessment of samples contaminated with MCs. These can either be water samples, bloom material, tissue samples or BGAS, as mainly extraction procedures need to be adjusted for the different matrices, while the UPLC-MS/MS technique generally is applicable for all samples.
166 9. Record of contribution
9. Record of contribution
Manuscript I Adsorption of Ten Microcystin Congeners to Common Laboratory-Ware Is Solvent and Surface Dependent Stefan Altaner, Jonathan Puddick, Susanna A. Wood and Daniel R. Dietrich
Experiments were planned by S. Altaner and J. Puddick with discussion by and suggestions of S. Wood. Experiments, data analysis and graphical representation of data was done by S. Altaner. S. Altaner wrote the first draft of the manuscript which was discussed with all other authors. All authors added to the writing of the manuscript.
Manuscript II Total Synthesis of Microcystin-LF and Derivatives Thereof Ivan Zemskov, Stefan Altaner, Daniel R. Dietrich and Valentin Wittmann
Study design was performed by I. Zemskov, D. Dietrich and V. Wittmann. Synthesis was performed by I. Zemskov. S. Altaner planned and performed the inhibition assays with the help of I. Zemskov. Main draft of the manuscript was written by I. Zemskov (synthesis part) and a minor part by S. Altaner (inhibition assays). All authors corrected, amended and complemented the manuscript.
Manuscript III Simultaneous detection of 14 microcystin congeners from tissue samples using UPLC- ESI-MS/MS and two different deuterated synthetic microcystins as internal standards Stefan Altaner, Jonathan Puddick, Valerie Fessard, Daniel Feurstein, Ivan Zemskov, Valentin Wittmann and Daniel R. Dietrich
S. Altaner designed the study with help of J. Puddick. Method establishment and experiments were performed by S. Altaner. Synthetic microcystins were synthesized by I. Zemskov under the supervision of V. Wittmann (PhD Thesis I. Zemskov). D. Feurstein and V. Fessard
167 9. Record of contribution performed the work on the mice. Manuscript was written by S. Altaner with amendments and corrections by all authors. Manuscript IV Machine learning prediction of cyanobacterial toxin (microcystin) toxicodynamics in humans Stefan Altaner*, Sabrina Jaeger*, Regina Fotler, Ivan Zemskov, Valentin Wittmann, Falk Schreiber and Daniel R. Dietrich *equal contribution
Study was designed by S. Altaner, D. Dietrich, S. Jaeger and F. Schreiber. Experiments for PPP and PPP2a were performed by S. Altaner, while PPP5 experiments were performed by R. Fotler under the supervision of S. Altaner (Bachelor’s thesis R. Fotler). Synthetic microcystin variants were produced by I. Zemskov under the supervision of V. Wittmann (PhD Thesis I. Zemskov). Machine learning and prediction models were performed by S. Jaeger. Manuscript was written by S. Altaner and S. Jaeger (machine learning part). All authors corrected, amended and complemented the manuscript.
168 10. Record of achievements
10. Record of achievements
Journal articles Altaner S, Puddick J, Wood SA, Dietrich DR (2017) Adsorption of Ten Microcystin Congeners to Common Laboratory-Ware Is Solvent and Surface Dependent. Toxins 9(4) doi:10.3390/toxins9040129
Zemskov I, Altaner S, Dietrich DR, Wittmann V (2017) Total Synthesis of Microcystin-LF and Derivatives Thereof. Journal of Organic Chemistry 82(7):3680-3691 doi:10.1021/acs.joc.7b00175
Altaner S, Puddick J, Fessard V, Feurstein D, Zemskov I, Wittmann V and Dietrich DR; Simultaneous detection of 14 microcystin congeners from tissue samples using UPLC- ESI-MS/MS and two different deuterated synthetic microcystins as internal standards; Toxins 11(7) doi:10.3390/toxins11070388
Altaner S*, Jaeger S*, Fotler R, Zemskov I, Wittmann V, Schreiber F, Dietrich DR; Machine learning prediction of cyanobacterial toxin (microcystin) toxicodynamics in humans;ALTEX [ePUB ahead of print], doi: 10.14573/altex.1904031
*equal contribution
Poster presentations Altaner S*, Nitschke J, Giménez-Papiol G, Zemskov I, Wittmann V, Fontanillo M, Köhn, M, and Dietrich DR; Inhibition of various human phosphatases by different microcystin congeners; presented at the 54th annual Meeting of the Society of Toxicology (2015), San Diego (USA) (peer-reviewed abstract, published in The Toxicologist supplemented to Toxicological Sciences, 2015, Vol. 144, #187).
Altaner S, Puddick J, Wood SA and Dietrich DR*; Adsorption of microcystins to common lab ware depends on the solvent and surface; presented at the 56th annual Meeting of the Society of Toxicology (2017), Baltimore (USA) (peer-reviewed abstract, published in The Toxicologist supplemented to Toxicological Sciences, 2017, Vol. 156, #1889).
169 10. Record of achievements
Kaur G*, Altaner S, Zemskov I, Wittmann V, and Dietrich DR; Effect of modification of microcystin structure on cellular uptake and cytotoxicity in OATP transfected HEK cells; presented at the 56th annual Meeting of the Society of Toxicology (2017), Baltimore (USA) (peer-reviewed abstract, published in The Toxicologist supplemented to Toxicological Sciences, 2017, Vol. 156, #3154).
Altaner S*, Zemskov I, Wittmann V, Puddick J and Dietrich DR; An UPLC-ESI- MS/MS method for the simultaneous detection of at least 17 microcystin congeners from complex matrices; presented at the 57th annual Meeting of the Society of Toxicology (2018), San Antonio (USA) (peer-reviewed abstract, published in The Toxicologist supplemented to Toxicological Sciences, 2018, Vol. 162, #2424).
Altaner S*, Zemskov I, Wittmann V, Puddick J and Dietrich DR; An UPLC-ESI- MS/MS method for the simultaneous detection of at least 17 microcystin congeners from complex matrices; presented at the poster session of the Graduate School Biological Sciences of the University Konstanz (2018), updated version of the above poster.
* presenting author
Talks Microcystin – Toxicity and Case studies, 2h-lecture as part of the ‘Advanced Course Human-and Environmental Toxicology’, annually 2014 – 2018.
Scholarship and funding Full scholarship by the ‘Arthur-und-Aenne-Feindt-Stiftung’ (Hamburg, Germany) that included a stipend for S. Altaner, consumables and travel costs. Grant for the scholarship was written by S. Altaner and D. Dietrich in 2014 with subsequent extension.
S. Altaner was a member of the ‘Graduate School Biological Sciences’ (GBS) of the University of Konstanz during the whole time of the PhD work.
170 11. Bibliography
11. Bibliography
Aggen JB, Humphrey JM, Gauss C-M, Huang H-B, Nairn AC, Chamberlin AR (1999) The design, synthesis, and biological evaluation of analogues of the serine-threonine protein phosphatase 1 and 2A selective inhibitor microcystin LA: rational modifications imparting PP1 selectivity. Bioorganic and Medicinal Chemistry 7(3):543-564 Ahlborg UG, Hanberg A, Kenne K (1992) Risk assessment of polychlorinated biphenyls (PCBs). Nordic Council of Ministers Altaner S, Puddick J, Wood SA, Dietrich DR (2017) Adsorption of Ten Microcystin Congeners to Common Laboratory-Ware Is Solvent and Surface Dependent. Toxins (Basel) 9(4) Amé MV, Wunderlin DAJW, Air,, Pollution S (2005) Effects of Iron, Ammonium and Temperature on Microcystin Content by a Natural Concentrated Microcystis Aeruginosa Population. 168(1):235-248 Andreeva AV, Kutuzov MA (2001) PPP Family of Protein Ser/Thr Phosphatases: Two Distinct Branches? Molecular Biology and Evolution 18(3):448-452 Aranda-Rodriguez R, Kubwabo C, Benoit FM (2003) Extraction of 15 microcystins and nodularin using immunoaffinity columns. Toxicon 42(6):587-99 Armstrong A, Geldart SP, Jenner CR, Scutt JN (2007) Organocatalytic Synthesis of β- Alkylaspartates via β-Lactone Ring Opening. The Journal of Organic Chemistry 72(21):8091-8094 Arnhard K, Gottschall A, Pitterl F, Oberacher H (2015) Applying 'Sequential Windowed Acquisition of All Theoretical Fragment Ion Mass Spectra' (SWATH) for systematic toxicological analysis with liquid chromatography-high-resolution tandem mass spectrometry. Analytical and Bioanalytical Chemistry 407(2):405-14 Asgari E, Mofrad MR (2015) Continuous Distributed Representation of Biological Sequences for Deep Proteomics and Genomics. PLoS One 10(11):e0141287 Azevedo SMFO, Carmichael WW, Jochimsen EM, Rinehart KL, Lau S, Shaw GR, Eaglesham GK (2002) Human intoxication by microcystins during renal dialysis treatment in Caruaru—Brazil. Toxicology 181–182(0):441-446 Backer LC, Carmichael W, Kirkpatrick B, Williams C, Irvin M, Zhou Y, Johnson TB, Nierenberg K, Hill VR, Kieszak SM, Cheng Y-S (2008) Recreational Exposure to Low Concentrations of Microcystins During an Algal Bloom in a Small Lake. Marine Drugs 6(2):389-406 Backer LC, Landsberg JH, Miller M, Keel K, Taylor TK (2013) Canine cyanotoxin poisonings in the United States (1920s-2012): review of suspected and confirmed cases from three data sources. Toxins 5(9):1597-1628 Barford D, Das AK, Egloff M (1998) The Structure and Mechanism of Protein Phosphatases: Insights into Catalysis and Regulation. Annual Review of Biophysics and Biomolecular Structure 27(1):133-164 Bastan R, Eskandari N, Sabzghabaee AM, Manian M (2014) Serine/Threonine phosphatases: classification, roles and pharmacological regulation. Int J Immunopathol Pharmacol 27(4):473-84 Bauer SM, Armstrong RW (1999) Total Synthesis of Motuporin (Nodularin-V). Journal of the American Chemical Society 121(27):6355-6366 Bautista AC, Moore CE, Lin Y, Cline MG, Benitah N, Puschner B (2015) Hepatopathy following consumption of a commercially available blue-green algae dietary supplement in a dog. BMC Veterinary Research 11:136 Beatty MF, Jennings-White C, Avery MA (1992) Stereocontrolled synthesis of (2S,3S,8S,9S,4E,6E)-3-amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,6-dienoic
171 11. Bibliography
acid (Adda), the amino acid characteristic of microcystins and nodularin. Journal of the Chemical Society, Perkin Transactions 1(13):1637-1641 Behrens SH, Grier DG (2001) The charge of glass and silica surfaces. The Journal of Chemical Physics 115(14):6716-6721 Beleites C, Neugebauer U, Bocklitz T, Krafft C, Popp J (2013) Sample size planning for classification models. Analytica Chimica Acta 760:25-33 Berry MA, Davis TW, Cory RM, Duhaime MB, Johengen TH, Kling GW, Marino JA, Den Uyl PA, Gossiaux D, Dick GJ, Denef VJ (2017) Cyanobacterial harmful algal blooms are a biological disturbance to Western Lake Erie bacterial communities. Environ Microbiol 19(3):1149-1162 Besant PG, Attwood PV (2005) Mammalian histidine kinases. Biochimica et Biophysica Acta 1754(1-2):281-90 Bieczynski F, De Anna JS, Pirez M, Brena BM, Villanueva SSM, Luquet CM (2014) Cellular transport of microcystin-LR in rainbow trout (Oncorhynchus mykiss) across the intestinal wall: Possible involvement of multidrug resistance-associated proteins. Aquatic Toxicology 154:97-106 Birbeck JA, Westrick JA, O'Neill GM, Spies B, Szlag DC (2019) Comparative Analysis of Microcystin Prevalence in Michigan Lakes by Online Concentration LC/MS/MS and ELISA. Toxins 11(1) Bochenska M, Biernat J (1972) Aminodicarboxylic acids. Part V. Stereochemistry of - Methylaspartic acids. Rocz Chem 46:1429-1433 Böttcher T, Pitscheider M, Sieber SA (2010) Natural Products and Their Biological Targets: Proteomic and Metabolomic Labeling Strategies. Angew Chem Int Ed 49(15):2680- 2698 Bouaicha N, Maatouk I (2004) Microcystin-LR and nodularin induce intracellular glutathione alteration, reactive oxygen species production and lipid peroxidation in primary cultured rat hepatocytes. Toxicology Letters 148(1-2):53-63 Bowling L, Baker PD (1996) Major cyanobacterial bloom in the Barwon-Darling River, Australia, in 1991, and underlying limnological conditions, vol 47, Brautigan DL (2013) Protein Ser/Thr phosphatases - the ugly ducklings of cell signalling. FEBS Journal 280(2):324-45 Breiman L (2001) Random Forests. J Machine Learning 45(1):5-32 Brewis ND, Street AJ, Prescott AR, Cohen PT (1993) PPX, a novel protein serine/threonine phosphatase localized to centrosomes. The EMBO Journal 12(3):987-996 Bullerjahn GS, McKay RM, Davis TW, Baker DB, Boyer GL, D'Anglada LV, Doucette GJ, Ho JC, Irwin EG, Kling CL, Kudela RM, Kurmayer R, Michalak AM, Ortiz JD, Otten TG, Paerl HW, Qin B, Sohngen BL, Stumpf RP, Visser PM, Wilhelm SW (2016) Global solutions to regional problems: Collecting global expertise to address the problem of harmful cyanobacterial blooms. A Lake Erie case study. Harmful Algae 54:223-238 Buratti FM, Manganelli M, Vichi S, Stefanelli M, Scardala S, Testai E, Funari E (2017) Cyanotoxins: producing organisms, occurrence, toxicity, mechanism of action and human health toxicological risk evaluation. Archives of Toxicology 91(3):1049-1130 Campos A, Vasconcelos V (2010) Molecular mechanisms of microcystin toxicity in animal cells. International Journal of Molecular Sciences 11(1):268-87 Carmichael WW (1997) The Cyanotoxins. In: Callow JA (ed) Advances in Botanical Research. vol 27. Academic Press, p 211-256 Carmichael WW, Azevedo S, An JS, Molica R, Jochimsen EM, Lau S, Rinehart KL, Shaw GR, Eaglesham GK (2001) Human fatalities from cyanobacteria: chemical and biological evidence for cyanotoxins. Environmental Health Perspectives 109(7):663 Carmichael WW, Biggs DF, Gorham PR (1975) Toxicology and pharmacological action of anabaena flos-aquae toxin. Science 187(4176):542-4
172 11. Bibliography
Carmichael WW, Biggs DF, Peterson MA (1979) Pharmacology of anatoxin-a, produced by the freshwater cyanophyte Anabaena flos-aquae NRC-44-1. Toxicon 17(3):229-36 Carpino LA, El-Faham A (1994) Effect of Tertiary Bases on O-Benzotriazolyluronium Salt- Induced Peptide Segment Coupling. The Journal of Organic Chemistry 59(4):695-698 Chang N, Vannah B, Yang YJ (2014) Comparative Sensor Fusion Between Hyperspectral and Multispectral Satellite Sensors for Monitoring Microcystin Distribution in Lake Erie. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 7(6):2426-2442 Chawla NV, Bowyer KW, Hall LO, Kegelmeyer WP (2002) SMOTE: synthetic minority over- sampling technique. Journal of artificial intelligence research 16:321-357 Chen J, Xie P, Li L, Xu J (2009) First identification of the hepatotoxic microcystins in the serum of a chronically exposed human population together with indication of hepatocellular damage. Toxicological Sciences 108(1):81-9 Chen L, Xie P (2016) Mechanisms of microcystin-induced cytotoxicity and apoptosis. Mini Reviews in Medicinal Chemistry Chen MX, Cohen PTW (1997) Activation of protein phosphatase 5 by limited proteolysis or the binding of polyunsaturated fatty acids to the TPR domain. FEBS Letters 400(1):136- 140 Chen MX, McPartlin AE, Brown L, Chen YH, Barker HM, Cohen PT (1994) A novel human protein serine/threonine phosphatase, which possesses four tetratricopeptide repeat motifs and localizes to the nucleus. The EMBO Journal 13(18):4278-4290 Chen T, Guestrin C (2016) XGBoost: A Scalable Tree Boosting System. Paper presented at the Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, California, USA, Cheng X, Maher JM, Lu H, Klaassen CD (2006) Endocrine regulation of gender-divergent mouse organic anion transporting polypeptide (Oatp) expression. Molecular Pharmacology Cho US, Xu W (2007) Crystal structure of a protein phosphatase 2A heterotrimeric holoenzyme. Nature 445(7123):53-57 Chorus I, Bartram J (1999) Toxic cyanobacteria in water: A guide to their public health consequences, monitoring and management, London Clave G, Ronco C, Boutal H, Kreich N, Volland H, Franck X, Romieu A, Renard PY (2010) Facile and rapid access to linear and truncated microcystin analogues for the implementation of immunoassays. Organic & Biomolecular Chemistry 8(3):676-90 Codd GA, Bell SG, Kaya K, Ward CJ, Beattie KA, Metcalf JS (1999a) Cyanobacterial toxins, exposure routes and human health. European Journal of Phycology 34(04):405-415 Codd GA, Metcalf JS, Beattie KA (1999b) Retention of Microcystis aeruginosa and microcystin by salad lettuce (Lactuca sativa) after spray irrigation with water containing cyanobacteria. Toxicon 37(8):1181-1185 Cohen PTW (2004) Overview of protein serine/threonine phosphatases. In: Ariño Jn, Alexander DR (eds) Protein Phosphatases. Springer Berlin Heidelberg, Berlin, Heidelberg, p 1-20 Colyer CL, Kinkade CS, Viskari PJ, Landers JP (2005) Analysis of cyanobacterial pigments and proteins by electrophoretic and chromatographic methods. Analytical and Bioanalytical Chemistry 382(3):559-569 Cooper GM, Hausman RE (2000) The cell. Sinauer Associates Sunderland Craig M, Luu HA, McCready TL, Holmes CFB, Williams D, Andersen RJ (1996) Molecular mechanisms underlying the interaction of motuporin and microcystins with type-1 and type-2A protein phosphatases. Biochemistry and Cell Biology 74(4):569-578 Cronin MTD, Dietrich D (2017) Computational toxicology to support predictions of ADMET properties and as a promising tool for 21st Century risk assessment. Computational Toxicology 1:1-2
173 11. Bibliography
Cundy DJ, Donohue AC, McCarthy TD (1999) An asymmetric synthesis of ADDA and ADDA- glycine dipeptide using the β-lactam synthon method. Journal of the Chemical Society, Perkin Transactions 1(5):559-568 Davis TW, Berry DL, Boyer GL, Gobler CJ (2009) The effects of temperature and nutrients on the growth and dynamics of toxic and non-toxic strains of Microcystis during cyanobacteria blooms. Harmful Algae 8(5):715-725 Dawson JF, Holmes CF (1999) Molecular mechanisms underlying inhibition of protein phosphatases by marine toxins. Frontiers in Bioscience 4:D646-58 de Maagd PG-J, Hendriks AJ, Seinen W, Sijm DT (1999) pH-dependent hydrophobicity of the cyanobacteria toxin microcystin-LR. Water Research 33(3):677-680 Deiters A, Cropp TA, Mukherji M, Chin JW, Anderson JC, Schultz PG (2003) Adding Amino Acids with Novel Reactivity to the Genetic Code of Saccharomyces Cerevisiae. Journal of the American Chemical Society 125(39):11782-11783 Deiters A, Schultz PG (2005) In vivo incorporation of an alkyne into proteins in Escherichia coli. Bioorganic & Medicinal Chemistry Letters 15(5):1521-1524 Dietrich D, Hoeger S (2005) Guidance values for microcystins in water and cyanobacterial supplement products (blue-green algal supplements): a reasonable or misguided approach? Toxicology and Applied Pharmacology 203(3):273-289 Diez-Quijada L, Guzman-Guillen R, Prieto Ortega AI, Llana-Ruiz-Cabello M, Campos A, Vasconcelos V, Jos A, Camean AM (2018a) New Method for Simultaneous Determination of Microcystins and Cylindrospermopsin in Vegetable Matrices by SPE- UPLC-MS/MS. Toxins 10(10) Diez-Quijada L, Prieto AI, Guzman-Guillen R, Jos A, Camean AM (2018b) Occurrence and toxicity of microcystin congeners other than MC-LR and MC-RR: A review. Food and Chemical Toxicology 125:106-132 dos Anjos JV, Sinou D, de Melo SJ, Srivastava RM (2007) Synthesis of Triazole-Linked Manno- and Glucopyranosyl Amino Acids. Synthesis 2007(17):2647-2652 Dunn PJ, Haener R, Rapoport H (1990) Stereoselective synthesis of 2,3-diamino acids. 2,3- Diamino-4-phenylbutanoic acid. The Journal of Organic Chemistry 55(17):5017-5025 Egloff MP, Cohen PT, Reinemer P, Barford D (1995) Crystal structure of the catalytic subunit of human protein phosphatase 1 and its complex with tungstate. Journal of Molecular Biology 254(5):942-59 EMA (2011) Guideline on bioanalytical method validation. In: Committee for Medicinal Products for Human Use (CHMP) (ed). European Medicines Agency, London Falconer IR, Burch MD, Steffensen DA, Choice M, Coverdale OR (1994) Toxicity of the blue‐ green alga (cyanobacterium) Microcystis aeruginosa in drinking water to growing pigs, as an animal model for human injury and risk assessment. Environmental Toxicology and Water Quality 9(2):131-139 Falconer IR, Humpage AR (2005) Health risk assessment of cyanobacterial (blue-green algal) toxins in drinking water. International Journal of Environmental Research and Public Health 2(1):43-50 Falconer IR, Yeung DS (1992) Cytoskeletal changes in hepatocytes induced by Microcystis toxins and their relation to hyperphosphorylation of cell proteins. Chemico-Biological Interactions 81(1-2):181-96 Fawell JK, Mitchell RE, Everett DJ, Hill RE (1999) The toxicity of cyanobacterial toxins in the mouse: I microcystin-LR. Human & Experimental Toxicology 18(3):162-7 Ferranti P, Fabbrocino S, Nasi A, Caira S, Bruno M, Serpe L, Gallo P (2009) Liquid chromatography coupled to quadruple time-of-flight tandem mass spectrometry for microcystin analysis in freshwaters: method performances and characterisation of a novel variant of microcystin-RR. Rapid Communications in Mass Spectrometry 23(9):1328-36
174 11. Bibliography
Feurstein D, Holst K, Fischer A, Dietrich DR (2009) Oatp-associated uptake and toxicity of microcystins in primary murine whole brain cells. Toxicology and Applied Pharmacology 234(2):247-255 Feurstein D, Kleinteich J, Heussner AH, Stemmer K, Dietrich DR (2010) Investigation of Microcystin Congener–Dependent Uptake into Primary Murine Neurons. Environmental Health Perspectives 118(10):1370 Feurstein D, Stemmer K, Kleinteich J, Speicher T, Dietrich DR (2011) Microcystin Congener– and Concentration-Dependent Induction of Murine Neuron Apoptosis and Neurite Degeneration. Toxicological Sciences 124(2):424-431 Fewer David P, Wahlsten M, Österholm J, Jokela J, Rouhiainen L, Kaasalainen U, Rikkinen J, Sivonen K (2013) The Genetic Basis for O-Acetylation of the Microcystin Toxin in Cyanobacteria. Chemistry & Biology 20(7):861-869 Fischer A, Hoeger SJ, Stemmer K, Feurstein DJ, Knobeloch D, Nussler A, Dietrich DR (2010) The role of organic anion transporting polypeptides (OATPs/SLCOs) in the toxicity of different microcystin congeners in vitro: A comparison of primary human hepatocytes and OATP-transfected HEK293 cells. Toxicology and Applied Pharmacology 245(1):9- 20 Fischer WJ, Altheimer S, Cattori V, Meier PJ, Dietrich DR, Hagenbuch B (2005) Organic anion transporting polypeptides expressed in liver and brain mediate uptake of microcystin. Toxicology and Applied Pharmacology 203(3):257-263 Fischer WJ, Garthwaite I, Miles CO, Ross KM, Aggen JB, Chamberlin AR, Towers NR, Dietrich DR (2001) Congener-Independent Immunoassay for Microcystins and Nodularins. Environmental Science & Technology 35(24):4849-4856 Fladmark KE, Brustugun OT, Hovland R, Boe R, Gjertsen BT, Zhivotovsky B, Doskeland SO (1999) Ultrarapid caspase-3 dependent apoptosis induction by serine/threonine phosphatase inhibitors. Cell Death and Differentiation 6(11):1099-108 Fladmark KE, Brustugun OT, Mellgren G, Krakstad C, Boe R, Vintermyr OK, Schulman H, Doskeland SO (2002) Ca2+/calmodulin-dependent protein kinase II is required for microcystin-induced apoptosis. Journal of Biological Chemistry 277(4):2804-11 Fontanillo M, Köhn M (2018) Microcystins: Synthesis and structure–activity relationship studies toward PP1 and PP2A. Bioorganic & Medicinal Chemistry 26(6):1118-1126 Fontanillo M, Zemskov I, Häfner M, Uhrig U, Salvi F, Simon B, Wittmann V, Köhn M (2016) Synthesis of Highly Selective Submicromolar Microcystin-Based Inhibitors of Protein Phosphatase (PP)2A over PP1. Angewandte Chemie (English Edition) 128(45):14191- 14195 Francis G (1878) Poisonous Australian Lake. Nature 18(444) Friedman JH (2001) Greedy function approximation: A gradient boosting machine. Ann Statist 29(5):1189-1232 Fu Q, Chen M, Anderson JT, Sun X, Hu S, Sparreboom A, Baker SD (2019) Interaction Between Sex and Organic Anion-Transporting Polypeptide 1b2 on the Pharmacokinetics of Regorafenib and Its Metabolites Regorafenib-N-Oxide and Regorafenib-Glucuronide in Mice. Clinical and Translational Science Fujiki H, Suganuma M (1993) Tumor promotion by inhibitors of protein phosphatases 1 and 2A: The okadaic acid class of compounds. Advances in Cancer Research 61:143-94 Gambaro A, Barbaro E, Zangrando R, Barbante C (2012) Simultaneous quantification of microcystins and nodularin in aerosol samples using high-performance liquid chromatography/negative electrospray ionization tandem mass spectrometry. Rapid Communications in Mass Spectrometry 26(12):1497-1506 Garibo D, Flores C, Ceto X, Prieto-Simon B, Del Valle M, Caixach J, Diogene J, Campas M (2014a) Inhibition equivalency factors for microcystin variants in recombinant and
175 11. Bibliography
wild-type protein phosphatase 1 and 2A assays. Environmental Science and Pollution Research International Garibo D, Flores C, Ceto X, Prieto-Simon B, Del Valle M, Caixach J, Diogene J, Campas M (2014b) Inhibition equivalency factors for microcystin variants in recombinant and wild-type protein phosphatase 1 and 2A assays. Environmental Science and Pollution Research International:10652-60 Geis-Asteggiante L, Lehotay SJ, Fortis LL, Paoli G, Wijey C, Heinzen H (2011) Development and validation of a rapid method for microcystins in fish and comparing LC-MS/MS results with ELISA. Analytical and Bioanalytical Chemistry 401(8):2617-30 Gilroy DJ, Kauffman KW, Hall RA, Huang X, Chu FS (2000) Assessing potential health risks from microcystin toxins in blue-green algae dietary supplements. Environmental Health Perspectives 108(5):435-9 Goldberg J, Huang H-b, Kwon Y-g, Greengard P, Nairn AC, Kuriyan J (1995) Three- dimensional structure of the catalytic subunit of protein serine/threonine phosphatase- 1. Nature 376(6543):745-753 GraphPad Grubb's Test Quick Calculator. In. https://www.graphpad.com/quickcalcs/Grubbs1.cfm Accessed 31.03.2017 2017 Greer B, Maul R, Campbell K, Elliott CT (2017) Detection of freshwater cyanotoxins and measurement of masked microcystins in tilapia from Southeast Asian aquaculture farms. Analytical and Bioanalytical Chemistry 409(16):4057-4069 Greer B, McNamee SE, Boots B, Cimarelli L, Guillebault D, Helmi K, Marcheggiani S, Panaiotov S, Breitenbach U, Akcaalan R, Medlin LK, Kittler K, Elliott CT, Campbell K (2016) A validated UPLC-MS/MS method for the surveillance of ten aquatic biotoxins in European brackish and freshwater systems. Harmful Algae 55:31-40 Greer B, Meneely JP, Elliott CT (2018) Uptake and accumulation of Microcystin-LR based on exposure through drinking water: An animal model assessing the human health risk. Scientific Reports 8(1):4913 Griffiths DJ, Saker ML (2003) The Palm Island mystery disease 20 years on: a review of research on the cyanotoxin cylindrospermopsin. Environmental Toxicology 18(2):78- 93 Hagenbuch B, Meier PJ (2004) Organic anion transporting polypeptides of the OATP/SLC21 family: phylogenetic classification as OATP/SLCO superfamily, new nomenclature and molecular/functional properties. Pflügers Archiv 447(5):653-665 Harke MJ, Steffen MM, Gobler CJ, Otten TG, Wilhelm SW, Wood SA, Paerl HW (2016) A review of the global ecology, genomics, and biogeography of the toxic cyanobacterium, Microcystis spp. Harmful Algae 54:4-20 Hastie CJ, Borthwick EB, Morrison LF, Codd GA, Cohen PT (2005) Inhibition of several protein phosphatases by a non-covalently interacting microcystin and a novel cyanobacterial peptide, nostocyclin. Biochimica et Biophysica Acta 1726(2):187-93 Hawkins PR, Novic S, Cox P, Neilan BA, Burns BP, Shaw G, Wickramasinghe W, Peerapornpisal Y, Ruangyuttikarn W, Itayama T (2005) A review of analytical methods for assessing the public health risk from microcystin in the aquatic environment. J WATER SUPPLY RES T 54(8):509-518 Heresztyn T, Nicholson BC (2001) Determination of cyanobacterial hepatotoxins directly in water using a protein phosphatase inhibition assay. Water Research 35(13):3049-3056 Heussner AH, Altaner S, Kamp L, Rubio F, Dietrich DR (2014a) Pitfalls in microcystin extraction and recovery from human blood serum. Chemico-Biological Interactions 223(0):87-94 Heussner AH, Mazija L, Fastner J, Dietrich DR (2012) Toxin content and cytotoxicity of algal dietary supplements. Toxicology and Applied Pharmacology 265(2):263-271
176 11. Bibliography
Heussner AH, Winter I, Altaner S, Kamp L, Rubio F, Dietrich DR (2014b) Comparison of two ELISA-based methods for the detection of microcystins in blood serum. Chemico- Biological Interactions 223:10-17 Hiemstra T, De Wit JCM, Van Riemsdijk WH (1989) Multisite proton adsorption modeling at the solid/solution interface of (hydr)oxides: A new approach. Journal of Colloid and Interface Science 133(1):105-117 Hilborn ED, Carmichael WW, Yuan M, Azevedo SM (2005) A simple colorimetric method to detect biological evidence of human exposure to microcystins. Toxicon 46(2):218-21 Hoeger SJ, Schmid D, Blom JF, Ernst B, Dietrich DR (2007) Analytical and functional characterization of microcystins [Asp3]MC-RR and [Asp3,Dhb7]MC-RR: consequences for risk assessment? Environmental Science & Technology 41(7):2609- 16 Holland A, Kinnear S (2013) Interpreting the possible ecological role(s) of cyanotoxins: compounds for competitive advantage and/or physiological aide? Marine Drugs 11(7):2239-2258 Hooser SB, Beasley VR, Basgall EJ, Carmichael WW, Haschek WM (1990) Microcystin-LR- induced ultrastructural changes in rats. Veterinary Pathology 27(1):9-15 Huisman J, Codd GA, Paerl HW, Ibelings BW, Verspagen JMH, Visser PM (2018a) Cyanobacterial blooms. Nature Reviews: Microbiology 16(8):471-483 Huisman J, Codd GA, Paerl HW, Ibelings BW, Verspagen JMH, Visser PM (2018b) Cyanobacterial blooms. Nature Reviews Microbiology 16(8):471-483 Humphrey JM, Aggen JB, Chamberlin AR (1996) Total synthesis of the serine-threonine phosphatase inhibitor microcystin-LA. Journal of the American Chemical Society 118(47):11759-11770 Humphrey JM, Bridges RJ, Hart JA, Chamberlin AR (1994) 2,3-Pyrrolidinedicarboxylates as Neurotransmitter Conformer Mimics: Enantioselective Synthesis via Chelation- Controlled Enolate Alkylation. The Journal of Organic Chemistry 59(9):2467-2472 Hyenstrand P, Metcalf JS, Beattie KA, Codd GA (2001a) Effects of adsorption to plastics and solvent conditions in the analysis of the cyanobacterial toxin microcystin-LR by high performance liquid chromatography. Water Research 35(14):3508-11 Hyenstrand P, Metcalf JS, Beattie KA, Codd GA (2001b) Losses of the cyanobacterial toxin microcystin-LR from aqueous solution by adsorption during laboratory manipulations. Toxicon 39(4):589-594 IARC (2011) Ingested Nitrate and Nitrite, and Cyanobacterial Peptide Toxins, vol 94. International Agency for Research on Cancer, Lyon Ibelings BW, Backer LC, Kardinaal WE, Chorus I (2015) Current approaches to cyanotoxin risk assessment and risk management around the globe. Harmful Algae 49:63-74 Ibelings BW, Chorus I (2007) Accumulation of cyanobacterial toxins in freshwater "seafood" and its consequences for public health: a review. Environmental Pollution 150(1):177- 92 Ikehara T, Imamura S, Sano T, Nakashima J, Kuniyoshi K, Oshiro N, Yoshimoto M, Yasumoto T (2009) The effect of structural variation in 21 microcystins on their inhibition of PP2A and the effect of replacing cys269 with glycine. Toxicon 54(4):539-544 Ito E, Kondo F, Terao K, Harada K-I (1997) Neoplastic nodular formation in mouse liver induced by repeated intraperitoneal injections of microcystin-LR. Toxicon 35(9):1453- 1457 Jaeger S, Fulle S, Turk S (2018) Mol2vec: Unsupervised Machine Learning Approach with Chemical Intuition. J Chem Inf Model 58(1):27-35 James KJ, Crowley J, Duphard J, Lehane M, Furey A (2007) Anatoxin-a and Analogues: Discovery, Distribution, and Toxicology Phycotoxins: Chemistry and Biochemistry.
177 11. Bibliography
Kamp L, Church JL, Carpino J, Faltin-Mara E, Rubio F (2016) The effects of water sample treatment, preparation, and storage prior to cyanotoxin analysis for cylindrospermopsin, microcystin and saxitoxin. Chemico-Biological Interactions 246:45-51 Kaur G, Dietrich DR The role of human MRP2 in cellular efflux of microcystin congeners (Poster). In: 57th Annual Meeting of the Society of Toxicology, San Antonio, USA, 11th - 15th March 2018 2018. Kaur G, Zemskov I, Wittmann V, Dietrich DR Toxicokinetics and dynamics of microcystin congeners are structure-dependent (Poster). In: 58th Annual Meeting of the Society of Toxicology, Baltimore, USA, 10th - 14th March 2018 2019. Kellmann R, Mihali TK, Neilan BA (2008) Identification of a saxitoxin biosynthesis gene with a history of frequent horizontal gene transfers. Journal of Molecular Evolution 67(5):526-38 Klein AR, Baldwin DS, Silvester E (2013) Proton and iron binding by the cyanobacterial toxin microcystin-LR. Environmental Science & Technology 47(10):5178-84 Kleinteich J, Puddick J, Wood SA, Hildebrand F, Laughinghouse HI, Pearce DA, Dietrich DR, Wilmotte A (2018) Toxic Cyanobacteria in Svalbard: Chemical Diversity of Microcystins Detected Using a Liquid Chromatography Mass Spectrometry Precursor Ion Screening Method. Toxins 10(4) Kleinteich J, Wood SA, Küpper FC, Camacho A, Quesada A, Frickey T, Dietrich DR (2012) Temperature-related changes in polar cyanobacterial mat diversity and toxin production. Nature Climate Change 2:356 Kohoutek J, Adamovsky O, Oravec M, Simek Z, Palikova M, Kopp R, Blaha L (2010) LC-MS analyses of microcystins in fish tissues overestimate toxin levels-critical comparison with LC-MS/MS. Analytical and Bioanalytical Chemistry 398(3):1231-7 Kokinaki S, Leondiadis L, Ferderigos N (2005) A Novel and Efficient Method for Cleavage of Phenacylesters by Magnesium Reduction with Acetic Acid. Organic Letters 7(9):1723- 1724 Kondo F, Harada Ki, Ueno Y (2005) Method for Quantifying Microcystins in Fish Using Immunoaffinity Purification. Journal of Liquid Chromatography & Related Technologies 28(19):3025-3033 Kondo F, Ikai Y, Oka H, Okumura M, Ishikawa N, Harada K, Matsuura K, Murata H, Suzuki M (1992) Formation, characterization, and toxicity of the glutathione and cysteine conjugates of toxic heptapeptide microcystins. Chemical Research in Toxicology 5(5):591-6 Kondo F, Ito Y, Oka H, Yamada S, Tsuji K, Imokawa M, Niimi Y, Harada K, Ueno Y, Miyazaki Y (2002) Determination of microcystins in lake water using reusable immunoaffinity column. Toxicon 40(7):893-9 Konst H, McKercher PD, Gorham PR, Robertson A, Howell J (1965) Symptoms and pathology produced by toxic Microcystis aeruginosa NRC-1 in laboratory and domestic animals. Canadian Journal of Comparative Medicine and Veterinary Science 29(9):221-8 Krusemark CJ (2012) Drug Design: Structure- and Ligand-Based Approaches edited by Kenneth M. Merz, Dagmar Ringe, and Charles H. Reynolds. The Quarterly Review of Biology 87(2):176-176 Kulms D, Dussmann H, Poppelmann B, Stander S, Schwarz A, Schwarz T (2002) Apoptosis induced by disruption of the actin cytoskeleton is mediated via activation of CD95 (Fas/APO-1). Cell Death and Differentiation 9(6):598-608 Lacap-Bugler DC, Lee KK, Archer S, Gillman LN, Lau MCY, Leuzinger S, Lee CK, Maki T, McKay CP, Perrott JK, de los Rios-Murillo A, Warren-Rhodes KA, Hopkins DW, Pointing SB (2017) Global Diversity of Desert Hypolithic Cyanobacteria. Frontiers in Microbiology 8:867
178 11. Bibliography
Lambert TW, Boland MP, Holmes CFB, Hrudey SE (1994) Quantitation of the Microcystin Hepatotoxins in Water at Environmentally Relevant Concentrations with the Protein Phosphatase Bioassay. Environmental Science & Technology 28(4):753-755 Lauer JL, Fields CG, Fields GB (1995) Sequence dependence of aspartimide formation during 9-fluorenylmethoxycarbonyl solid-phase peptide synthesis. Letters in Peptide Science 1(4):197-205 Lawton LA, Edwards C, Codd GA (1994) Extraction and high-performance liquid chromatographic method for the determination of microcystins in raw and treated waters. Analyst 119(7):1525-30 Li W, He J, Chen J, Xie P (2018) Excretion pattern and dynamics of glutathione detoxification of microcystins in Sprague Dawley rat. Chemosphere 191:357-364 Li Y, Chen JA, Zhao Q, Pu C, Qiu Z, Zhang R, Weiqun S (2011) A cross-sectional investigation of chronic exposure to microcystin in relationship to childhood liver damage in the Three Gorges Reservoir Region, China. Environmental Health Perspectives 119(10):1483-1488 Liang CT, Lin YS, Huang YC, Huang HL, Yang JQ, Wu TH, Chang CF, Huang SJ, Huang HB, Lin TH (2018) Characterization of the interactions between inhibitor-1 and recombinant PP1 by NMR spectroscopy. Scientific Reports 8(1):50 Liang G, Xie P, Chen J, Yu T (2011) Comparative studies on the pH dependence of DOW of microcystin-RR and -LR using LC-MS. Scientific World Journal 11:20-6 Lin J-R, Chu FS (1994) Kinetics of Distribution of Microcystin LR in Serum and Liver Cytosol of Mice: An Immunochemical Analysis. Journal of Agricultural and Food Chemistry 42(4):1035-1040 Lu H, Choudhuri S, Ogura K, Csanaky IL, Lei X, Cheng X, Song PZ, Klaassen CD (2008) Characterization of organic anion transporting polypeptide 1b2-null mice: essential role in hepatic uptake/toxicity of phalloidin and microcystin-LR. Toxicological Sciences 103(1):35-45 Lyon-Colbert A, Su S, Cude C (2018) A Systematic Literature Review for Evidence of Aphanizomenon flos-aquae Toxigenicity in Recreational Waters and Toxicity of Dietary Supplements: 2000(-)2017. Toxins 10(7) Macek B, Gnad F, Soufi B, Kumar C, Olsen JV, Mijakovic I, Mann M (2008) Phosphoproteome Analysis of E. coli Reveals Evolutionary Conservation of Bacterial Ser/Thr/Tyr Phosphorylation. Molecular & Cellular Proteomics 7(2):299-307 Mackintosh C, Beattie KA, Klumpp S, Cohen P, Codd GA (1990) Cyanobacterial microcystin- LR is a potent and specific inhibitor of protein phosphatase-1 and phosphatase-2a from both mammals and higher-plants. FEBS Letters 264(2):187-192 Mallet C (2017) Analysis of Microcystins RR, LY, and YR in Bottled, Tap, and Surface Water Using ACQUITY UPLC Systems with 2D-LC Technology. In: Waters. http://www.waters.com/waters/library.htm?cid=511436&lid=134826892&locale=de_ DE Accessed 09.11.2018 Manubolu M, Lee J, Riedl KM, Kua ZX, Collart LP, Ludsin SA (2018) Optimization of extraction methods for quantification of microcystin-LR and microcystin-RR in fish, vegetable, and soil matrices using UPLC-MS/MS. Harmful Algae 76:47-57 Martin WF, Garg S, Zimorski V (2015) Endosymbiotic theories for eukaryote origin. Philosophical transactions of the Royal Society of London Series B, Biological sciences 370(1678):20140330-20140330 Mattila K, Annila A, Rantala TT (2000a) Metals ions mediate the binding of cyanobacterial toxins to human protein phosphatase 1 - A computational study. Acta Universitatis Ouluensis A A 351
179 11. Bibliography
Mattila K, Annila A, Rantala TT (2000b) Metals ions mediate the binding of cyanobacterial toxins to human protein phosphatase 1 - A computational study, vol A 351. University of Oulu, Oulu McLellan NL, Manderville RA (2017) Toxic mechanisms of microcystins in mammals. Toxicology research 6(4):391-405 Mehrotra AP, Webster KL, Gani D (1997) Design and preparation of serine–threonine protein phosphatase inhibitors based upon the nodularin and microcystin toxin structures: Part 1. Evaluation of key inhibitory features and synthesis of a rationally stripped-down nodularin macrocycle. Journal of the Chemical Society, Perkin Transactions 1(17):2495-2512 Meier-Abt F, Hammann-Hanni A, Stieger B, Ballatori N, Boyer JL (2007) The organic anion transport polypeptide 1d1 (Oatp1d1) mediates hepatocellular uptake of phalloidin and microcystin into skate liver. Toxicology and Applied Pharmacology 218(3):274-9 Michalak AM, Anderson EJ, Beletsky D, Boland S, Bosch NS, Bridgeman TB, Chaffin JD, Cho K, Confesor R, Daloğlu I, DePinto JV, Evans MA, Fahnenstiel GL, He L, Ho JC, Jenkins L, Johengen TH, Kuo KC, LaPorte E, Liu X, McWilliams MR, Moore MR, Posselt DJ, Richards RP, Scavia D, Steiner AL, Verhamme E, Wright DM, Zagorski MA (2013) Record-setting algal bloom in Lake Erie caused by agricultural and meteorological trends consistent with expected future conditions. Proceedings of the National Academy of Sciences 110(16):6448-6452 Mikalsen B, Boison G, Skulberg OM, Fastner J, Davies W, Gabrielsen TM, Rudi K, Jakobsen KS (2003) Natural Variation in the Microcystin Synthetase Operon mcyABC and Impact on Microcystin Production in Microcystis Strains. Journal of Bacteriology 185(9):2774-2785 Miles CO (2017) Rapid and Convenient Oxidative Release of Thiol-Conjugated Forms of Microcystins for Chemical Analysis. Chemical Research in Toxicology 30(8):1599- 1608 Miles CO, Sandvik M, Nonga HE, Ballot A, Wilkins AL, Rise F, Jaabaek JA, Loader JI (2016a) Conjugation of Microcystins with Thiols is Reversible: Base-Catalyzed Deconjugation for Chemical Analysis. Chemical Research in Toxicology:860-70 Miles CO, Sandvik M, Nonga HE, Ballot A, Wilkins AL, Rise F, Jaabaek JA, Loader JI (2016b) Conjugation of Microcystins with Thiols is Reversible: Base-Catalyzed Deconjugation for Chemical Analysis. Chemical Research in Toxicology Miller MA, Kudela RM, Mekebri A, Crane D, Oates SC, Tinker MT, Staedler M, Miller WA, Toy-Choutka S, Dominik C, Hardin D, Langlois G, Murray M, Ward K, Jessup DA (2010) Evidence for a novel marine harmful algal bloom: cyanotoxin (microcystin) transfer from land to sea otters. PloS One 5(9) Milutinović A, Zorc-Pleskovič R, Živin M, Vovk A, Serša I, Šuput D (2013) Magnetic Resonance Imaging for Rapid Screening for the Nephrotoxic and Hepatotoxic Effects of Microcystins. Marine Drugs 11(8):2785-2798 Mol HG, Plaza-Bolanos P, Zomer P, de Rijk TC, Stolker AA, Mulder PP (2008) Toward a generic extraction method for simultaneous determination of pesticides, mycotoxins, plant toxins, and veterinary drugs in feed and food matrixes. Analytical Chemistry 80(24):9450-9 Monks NR, Liu S, Xu Y, Yu H, Bendelow AS, Moscow JA (2007) Potent cytotoxicity of the phosphatase inhibitor microcystin LR and microcystin analogues in OATP1B1- and OATP1B3-expressing HeLa cells. Molecular Cancer Therapeutics 6(2):587-98 Neffling M-R, Lance E, Meriluoto J (2010a) Detection of free and covalently bound microcystins in animal tissues by liquid chromatography–tandem mass spectrometry. Environmental Pollution 158(3):948-952
180 11. Bibliography
Neffling M-R, Spoof L, Quilliam M, Meriluoto J (2010b) LC–ESI-Q-TOF-MS for faster and accurate determination of microcystins and nodularins in serum. Journal of Chromatography B 878(26):2433-2441 Niedermeyer T (2014) Microcystin congeners described in the literature, Version 5. In. 10.6084/m9.figshare.880756 Accessed January 2017 2017 Nishiwaki-Matsushima R, Ohta T, Nishiwaki S, Suganuma M, Kohyama K, Ishikawa T, Carmichael WW, Fujiki H (1992) Liver tumor promotion by the cyanobacterial cyclic peptide toxin microcystin-LR. Journal of Cancer Research and Clinical Oncology 118(6):420-4 Nishizawa T, Asayama M, Fujii K, Harada K, Shirai M (1999) Genetic analysis of the peptide synthetase genes for a cyclic heptapeptide microcystin in Microcystis spp. Journal of Biochemistry 126(3):520-9 OECD (2018) The Adverse Outcome Pathways development programme workplan. In. http://www.oecd.org/chemicalsafety/testing/projects-adverse-outcome-pathways.htm Accessed 28.06 2018 Ohtani I, Moore RE, Runnegar MTC (1992) Cylindrospermopsin: a potent hepatotoxin from the blue-green alga Cylindrospermopsis raciborskii. Journal of the American Chemical Society 114(20):7941-7942 Okeley NM, Zhu Y, van der Donk WA (2000) Facile Chemoselective Synthesis of Dehydroalanine-Containing Peptides. Organic Letters 2(23):3603-3606 Olsen JV, Blagoev B, Gnad F, Macek B, Kumar C, Mortensen P, Mann M (2006) Global, In Vivo, and Site-Specific Phosphorylation Dynamics in Signaling Networks. Cell 127(3):635-648 Omidi A, Esterhuizen-Londt M, Pflugmacher S (2018) Still challenging: the ecological function of the cyanobacterial toxin microcystin – What we know so far. Toxin Reviews 37(2):87-105 OpenStax (2013) Cancer and the cell cycle Biology. Rice University, Houston Paerl HW, Gardner WS, McCarthy MJ, Peierls BL, Wilhelm SW (2014) Algal blooms: noteworthy nitrogen. Science 346(6206):175 Paerl HW, Huisman J (2008) Blooms like it hot. Science 320(5872):57-8 Paerl HW, Otten TG (2013) Harmful cyanobacterial blooms: causes, consequences, and controls. Microbial Ecology 65(4):995-1010 Panek JS, Hu T (1997) Asymmetric Synthesis of (2S,3S,8S,9S)-N-Boc ADDA: Application of a Palladium(0)-Catalyzed Cross-Coupling Reaction of Trisubstituted Olefins. The Journal of Organic Chemistry 62(15):4914-4915 Pearson C, Rinehart KL, Sugano M, Costerison JR (2000) Enantiospecific Synthesis of N-Boc- Adda: A Linear Approach. Organic Letters 2(18):2901-2903 Pearson LA, Dittmann E, Mazmouz R, Ongley SE, D’Agostino PM, Neilan BA (2016) The genetics, biosynthesis and regulation of toxic specialized metabolites of cyanobacteria. Harmful Algae 54:98-111 Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, Prettenhofer P, Weiss R, Dubourg V, Vanderplas J, Passos A, Cournapeau D, Brucher M, Perrot M, Duchesnay E (2011) Scikit-learn: Machine Learning in Python. J Mach Learn Res 12:2825-2830 Pelley J (2016) Taming Toxic Algae Blooms. ACS Central Science 2(5):270-273 Pereira SR, Vasconcelos VM, Antunes A (2011) The phosphoprotein phosphatase family of Ser/Thr phosphatases as principal targets of naturally occurring toxins. 41(2):83-110 Pflugmacher S, Wiegand C, Oberemm A, Beattie KA, Krause E, Codd GA, Steinberg CEW (1998a) Identification of an enzymatically formed glutathione conjugate of the cyanobacterial hepatotoxin microcystin-LR: the first step of detoxication. Biochimica et Biophysica Acta (BBA) - General Subjects 1425(3):527-533
181 11. Bibliography
Pflugmacher S, Wiegand C, Oberemm A, Beattie KA, Krause E, Codd GA, Steinberg CEW (1998b) Identification of an enzymatically formed glutathione conjugate of the cyanobacterial hepatotoxin microcystin-LR: the first step of detoxication. BBA-Gen Subjects 1425(3):527-533 Pouria S, de Andrade A, Barbosa J, Cavalcanti RL, Barreto VT, Ward CJ, Preiser W, Poon GK, Neild GH, Codd GA (1998) Fatal microcystin intoxication in haemodialysis unit in Caruaru, Brazil. Lancet 352(9121):21-6 Preece EP, Hardy FJ, Moore BC, Bryan M (2017) A review of microcystin detections in Estuarine and Marine waters: Environmental implications and human health risk. Harmful Algae 61:31-45 Puddick J, Prinsep MR, Wood SA, Cary SC, Hamilton DP (2016a) Modulation of microcystin congener abundance following nitrogen depletion of a Microcystis batch culture. Aquatic Ecology:1-12 Puddick J, Prinsep MR, Wood SA, Cary SC, Hamilton DP, Holland PT (2015) Further characterization of glycine-containing microcystins from the McMurdo dry valleys of Antarctica. Toxins 7(2):493-515 Puddick J, Prinsep MR, Wood SA, Kaufononga SA, Cary SC, Hamilton DP (2014) High levels of structural diversity observed in microcystins from Microcystis CAWBG11 and characterization of six new microcystin congeners. Marine Drugs 12(11):5372-95 Puddick J, Wood SA, Hawes I, Hamilton DP (2016b) Fine-scale cryogenic sampling of planktonic microbial communities: Application to toxic cyanobacterial blooms. Limnology and Oceanography: Methods 14(9):600-609 Puschner B, Galey FD, Johnson B, Dickie CW, Vondy M, Francis T, Holstege DM (1998) Blue-green algae toxicosis in cattle. Journal of the American Veterinary Medical Association 213(11):1605-7, 1571 Qu M, Lefebvre DD, Wang Y, Qu Y, Zhu D, Ren W (2014) Algal blooms: proactive strategy. Science 346(6206):175-6 Rantala A, Fewer DP, Hisbergues M, Rouhiainen L, Vaitomaa J, Börner T, Sivonen K (2004) Phylogenetic evidence for the early evolution of microcystin synthesis. Proceedings of the National Academy of Sciences of the United States of America 101(2):568-573 Reisner M, Carmeli S, Werman M, Sukenik A (2004) The cyanobacterial toxin cylindrospermopsin inhibits pyrimidine nucleotide synthesis and alters cholesterol distribution in mice. Toxicological Sciences 82(2):620-7 Riggs Costerison J (2002) An Approach to the Total Synthesis of Protein Phosphatase 1 and 2A Inhibitor Microcystin-LR. PhD Thesis, University of Illinois: Urbana Rinehart KL, Harada K, Namikoshi M, Chen C, Harvis CA, Munro MHG, Blunt JW, Mulligan PE, Beasley VR, Dahlem AM, Carmichael WW (1988) Nodularin, microcystin, and the configuration of Adda. Journal of the American Chemical Society 110(25):8557-8558 Ritchie SM, Battey NH (1996) Phosphorylation and the Cytoskeleton. In: Verma DPS (ed) Signal Transduction in Plant Growth and Development. Springer Vienna, Vienna, p 113-139 Rivasseau C, Martins S, Hennion MC (1998) Determination of some physicochemical parameters of microcystins (cyanobacterial toxins) and trace level analysis in environmental samples using liquid chromatography. Journal of Chromatography A 799(1-2):155-69 Robinson NA, Pace JG, Matson CF, Miura GA, Lawrence WB (1991) Tissue distribution, excretion and hepatic biotransformation of microcystin-LR in mice. Journal of Pharmacology and Experimental Therapeutics 256(1):176-82 Roegner AF, Schirmer MP, Puschner B, Brena B, Gonzalez-Sapienza G (2014a) Rapid Quantitative Analysis of Microcystins in Raw Surface Waters with MALDI MS
182 11. Bibliography
Utilizing Easily Synthesized Internal Standards. Toxicon : official journal of the International Society on Toxinology 78:94-102 Roegner AF, Schirmer MP, Puschner B, Brena B, Gonzalez-Sapienza G (2014b) Rapid Quantitative Analysis of Microcystins in Raw Surface Waters with MALDI MS Utilizing Easily Synthesized Internal Standards. Toxicon 78:94-102 Roemmelt AT, Steuer AE, Poetzsch M, Kraemer T (2014) Liquid chromatography, in combination with a quadrupole time-of-flight instrument (LC QTOF), with sequential window acquisition of all theoretical fragment-ion spectra (SWATH) acquisition: systematic studies on its use for screenings in clinical and forensic toxicology and comparison with information-dependent acquisition (IDA). Analytical Chemistry 86(23):11742-9 Rogers S, Puddick J, Wood SA, Dietrich DR, Hamilton DP, Prinsep MR (2015) The effect of cyanobacterial biomass enrichment by centrifugation and GF/C filtration on subsequent microcystin measurement. Toxins 7(3):821-34 Romero M, Muro-Pastor AM, Otero A (2011) Quorum sensing N-acylhomoserine lactone signals affect nitrogen fixation in the cyanobacterium Anabaena sp. PCC7120. FEMS Microbiology Letters 315(2):101-8 Roy-Lachapelle A, Solliec M, Bouchard MF, Sauvé S (2017) Detection of Cyanotoxins in Algae Dietary Supplements. Toxins 9(3):76 Rzymski P, Poniedzialek B (2014) In search of environmental role of cylindrospermopsin: a review on global distribution and ecology of its producers. Water Research 66:320-337 Sakaguchi K, Yamamoto M, Kawamoto T, Yamada T, Shinada T, Shimamoto K, Ohfune Y (2004) Synthesis of optically active β-alkyl aspartate via [3,3] sigmatropic rearrangement of α-acyloxytrialkylsilane. Tetrahedron Letters 45(30):5869-5872 Samdal I, Van Nieuwenhove I, Lovberg K, Kilcoyne J, Wright E, Sandvik M, Uhlig S, Miles C Immunoaffinity columns for clean-up and concentration of algal toxins (Poster). In: 18th International Conference on Harmful Algae, Nantes, France, 2018. Samy R, Kim HY, Brady M, Toogood PL (1999) Total Synthesis of Motuporin and 5-[l-Ala]- Motuporin. The Journal of Organic Chemistry 64(8):2711-2728 Sano T, Nohara K, Shiraishi F, Kaya K (1992) A Method for Micro-Determination of Total Microcystin Content in Waterblooms of Cyanobacteria (Blue-Green Algae). International Journal of Environmental Analytical Chemistry 49(3):163-170 Sanseverino I, António DC, Loos R, Lettieri T (2017) Cyanotoxins: methods and approaches for their analysis and detection. In: Centre JR (ed). Luxembourg Schmidt JR, Shaskus M, Estenik JF, Oesch C, Khidekel R, Boyer GL (2013) Variations in the microcystin content of different fish species collected from a eutrophic lake. Toxins 5(5):992-1009 Schmidt JR, Wilhelm SW, Boyer GL (2014) The fate of microcystins in the environment and challenges for monitoring. Toxins 6(12):3354-87 Schwarzenberger A, Sadler T, Motameny S, Ben-Khalifa K, Frommolt P, Altmüller J, Konrad K, von Elert E (2014) Deciphering the genetic basis of microcystin tolerance. BMC Genomics 15(1):776-776 Seshacharyulu P, Pandey P, Datta K, Batra SK (2013) Phosphatase: PP2A structural importance, regulation and its aberrant expression in cancer. Cancer Letters 335(1):9- 18 Sharif DI, Gallon J, Smith CJ, Dudley E (2008) Quorum sensing in Cyanobacteria: N-octanoyl- homoserine lactone release and response, by the epilithic colonial cyanobacterium Gloeothece PCC6909. Isme j 2(12):1171-82 Shi Y (2009) Serine/Threonine Phosphatases: Mechanism through Structure. Cell 139(3):468- 484
183 11. Bibliography
Shoemaker JA, Tettenhorst DR, Delacruz A (2015) Determination of microcystins and nodularin in drinking water by solid phase extraction and liquid chromatography/tandem mass spectrometry (LC/MS/MS). In: Agency USEP (ed) Method 544. National Exposure Research Laboratory, Office of Research and Development, Cincinnati, Ohio 45268 Shreder KR, Liu Y, Nomanhboy T, Fuller SR, Wong MS, Gai WZ, Wu J, Leventhal PS, Lill JR, Corral S (2004) Design and Synthesis of AX7574: A Microcystin-Derived, Fluorescent Probe for Serine/Threonine Phosphatases. Bioconjugate Chemistry 15(4):790-798 Sipiä VO, Kankaanpää HT, Flinkman J, Lahti K, Meriluoto JA (2001) Time-dependent accumulation of cyanobacterial hepatotoxins in flounders (Platichthys flesus) and mussels (Mytilus edulis) from the northern Baltic Sea. Environmental Toxicology 16(4):330-6 Smith JL, Boyer GL (2009) Standardization of microcystin extraction from fish tissues: a novel internal standard as a surrogate for polar and non-polar variants. Toxicon 53(2):238-45 Smith JL, Schulz KL, Zimba PV, Boyer GL (2010) Possible mechanism for the foodweb transfer of covalently bound microcystins. Ecotoxicology and Environmental Safety 73(5):757-761 Snyder LR (1978) Classification of the solvent properties of common liquids. Journal of Chromatographic Science 16(6):223-234 Soares RM, Yuan M, Servaites JC, Delgado A, Magalhaes VF, Hilborn ED, Carmichael WW, Azevedo SM (2006) Sublethal exposure from microcystins to renal insufficiency patients in Rio de Janeiro, Brazil. Environmental Toxicology 21(2):95-103 Spoof L, Blaszczyk A, Meriluoto J, Ceglowska M, Mazur-Marzec H (2015) Structures and Activity of New Anabaenopeptins Produced by Baltic Sea Cyanobacteria. Marine Drugs 14(1):8 Spoof L, Catherine A (2017) Appendix 3: Tables of Microcystins and Nodularin. In: Meriluoto J, Spoof L, Codd G (eds) Handbook of Cyanobacterial Monitoring and Cyanotoxin Analysis. John Wiley & Sons Stefansson B, Ohama T, Daugherty AE, Brautigan DL (2008) Protein phosphatase 6 regulatory subunits composed of ankyrin repeat domains. Biochemistry 47(5):1442-51 Steiner K, Hagenbuch B, Dietrich DR (2014) Molecular cloning and functional characterization of a rainbow trout liver Oatp. Toxicology and Applied Pharmacology 280(3):534-42 Steiner K, Zimmermann L, Hagenbuch B, Dietrich D (2015) Zebrafish Oatp-mediated transport of microcystin congeners. Archives of Toxicology Stokvis E, Rosing H, Beijnen JH (2005) Stable isotopically labeled internal standards in quantitative bioanalysis using liquid chromatography/mass spectrometry: necessity or not? Rapid Communications in Mass Spectrometry 19(3):401-7 Sun KH, Silverman A (1945) Lewis acid-base theory applied to glass. Journal of the American Ceramic Society 28(1):8-11 Suria H, Chau LA, Negrou E, Kelvin DJ, Madrenas J (1999) Cytoskeletal disruption induces T cell apoptosis by a caspase-3 mediated mechanism. Life Sciences 65(25):2697-707 Swingle M, Ni L, Honkanen RE (2007) Small Molecule Inhibitors of Ser/thr Protein Phosphatases: Specificity, Use and Common Forms of Abuse. Methods in molecular biology (Clifton, NJ) 365:23-38 Takumi S, Komatsu M, Furukawa T, Ikeda R, Sumizawa T, Akenaga H, Maeda Y, Aoyama K, Arizono K, Ando S, Takeuchi T (2010) p53 Plays an important role in cell fate determination after exposure to microcystin-LR. Environmental Health Perspectives 118(9):1292-1298 Tappan E, Chamberlin AR (2008) Activation of protein phosphatase 1 by a small molecule designed to bind to the enzyme's regulatory site. Chemistry & Biology 15(2):167-74
184 11. Bibliography
Taranu ZE, Gregory-Eaves I, Steele RJ, Beaulieu M, Legendre P (2017) Predicting microcystin concentrations in lakes and reservoirs at a continental scale: A new framework for modelling an important health risk factor. 26(6):625-637 Taylor C, Quinn RJ, Alewood P (1996a) Synthesis of cyclic peptides modelled on the microcystin and nodularin rings. Bioorganic & Medicinal Chemistry Letters 6(17):2107-2112 Taylor C, Quinn RJ, Suganuma M, Fujiki H (1996b) Inhibition of protein phosphatase 2A by cyclic peptides modelled on the microcystin ring. Bioorganic & Medicinal Chemistry Letters 6(17):2113-2116 Tencalla F, Dietrich D (1997) Biochemical characterization of microcystin toxicity in rainbow trout (Oncorhynchus mykiss). Toxicon 35(4):583-95 Theiss WC, Carmichael WW, Wyman J, Bruner R (1988) Blood pressure and hepatocellular effects of the cyclic heptapeptide toxin produced by the freshwater cyanobacterium (blue-green alga) Microcystis aeruginosa strain PCC-7820. Toxicon 26(7):603-13 Tillett D, Dittmann E, Erhard M, von Dohren H, Borner T, Neilan BA (2000) Structural organization of microcystin biosynthesis in Microcystis aeruginosa PCC7806: an integrated peptide-polyketide synthetase system. Chemistry & Biology 7(10):753-64 ToledoBlade (2014) The Toledo Watercrisis Toledo Blade. Tomitani A, Knoll AH, Cavanaugh CM, Ohno T (2006) The evolutionary diversification of cyanobacteria: Molecular–phylogenetic and paleontological perspectives. Proceedings of the National Academy of Sciences 103(14):5442-5447 Tricker E, Arvand A, Kwan R, Chen GY, Gallagher E, Cheng G (2011) Apoptosis induced by cytoskeletal disruption requires distinct domains of MEKK1. PloS One 6(2):e17310 Trifiro G, Barbaro E, Gambaro A, Vita V, Clausi MT, Franchino C, Palumbo MP, Floridi F, De Pace R (2016) Quantitative determination by screening ELISA and HPLC-MS/MS of microcystins LR, LY, LA, YR, RR, LF, LW, and nodularin in the water of Occhito lake and crops. Analytical and Bioanalytical Chemistry Trufelli H, Palma P, Famiglini G, Cappiello A (2011) An overview of matrix effects in liquid chromatography–mass spectrometry. Mass Spectrometry Reviews 30(3):491-509 Turk S (2018) Mol2Vec - an unsupervised machine learning approach to learn vector representations of molecular substructures. In. https://github.com/samoturk/mol2vec Accessed 08.08 2018 Turner AD, Dhanji-Rapkova M, O'Neill A, Coates L, Lewis A, Lewis K (2018a) Analysis of Microcystins in Cyanobacterial Blooms from Freshwater Bodies in England. Toxins 10(1) Turner AD, Dhanji-Rapkova M, O'Neill A, Coates L, Lewis A, Lewis K (2018b) Analysis of Microcystins in Cyanobacterial Blooms from Freshwater Bodies in England. Toxins 10(1) Turner AD, Waack J, Lewis A, Edwards C, Lawton L (2017) Development and single- laboratory validation of a UHPLC-MS/MS method for quantitation of microcystins and nodularin in natural water, cyanobacteria, shellfish and algal supplement tablet powders. Journal of Chromatography B: Analytical Technologies in the Biomedical and Life Sciences 1074-1075:111-123 Turner PC, Gammie AJ, Hollinrake K, Codd GAJBBMJ (1990) Pneumonia associated with contact with cyanobacteria. 300(6737):1440 UN (2008) Trends in global water use by sector. In: Environment Programme (ed). Vital Water Graphics UN (2015) World Population Prospects: The 2015 Revision. In: Department of Economic and Social Affairs. Population Division (ed).
185 11. Bibliography
Valentekovich RJ, Schreiber SL (1995) Enantiospecific total synthesis of the protein phosphatase inhibitor motuporin. Journal of the American Chemical Society 117(35):9069-9070 van Apeldoorn ME, van Egmond HP, Speijers GJ, Bakker GJ (2007) Toxins of cyanobacteria. Molecular Nutrition & Food Research 51(1):7-60 Vesterkvist PS, Meriluoto JA (2003) Interaction between microcystins of different hydrophobicities and lipid monolayers. Toxicon 41(3):349-55 Vichi S, Lavorini P, Funari E, Scardala S, Testai E (2012) Contamination by Microcystis and microcystins of blue-green algae food supplements (BGAS) on the Italian market and possible risk for the exposed population. Food and Chemical Toxicology 50(12):4493- 9 Vincent WF (2002) Cyanobacterial Dominance in the Polar Regions. In: Whitton BA, Potts M (eds) The Ecology of Cyanobacteria: Their Diversity in Time and Space. Springer Netherlands, Dordrecht, p 321-340 Wang Q, Xie P, Chen J, Liang G (2008) Distribution of microcystins in various organs (heart, liver, intestine, gonad, brain, kidney and lung) of Wistar rat via intravenous injection. Toxicon 52(6):721-7 Wang S, Cyronak M, Yang E (2007) Does a stable isotopically labeled internal standard always correct analyte response? A matrix effect study on a LC/MS/MS method for the determination of carvedilol enantiomers in human plasma. Journal of Pharmaceutical and Biomedical Analysis 43(2):701-7 Webster KL, Maude AB, O'Donnell ME, Mehrotra AP, Gani D (2001) Design and preparation of serine–threonine protein phosphatase inhibitors based upon the nodularin and microcystin toxin structures. Part 3. Journal of the Chemical Society, Perkin Transactions 1(14):1673-1695 Whitton BA, Potts M (2007) The ecology of cyanobacteria: their diversity in time and space. Springer Science & Business Media WHO, UNESCO, UNEP (1999a) Chapter 4. Human health aspects. In: Chorus I, Bartram J (eds) Toxic Cyanobacteria in Water: A guide to their public health consequences monitoring and management. E & FN Spon, p 125-160 WHO, UNESCO, UNEP (1999b) Chapter 11. Fieldwork: site inspection and sampling. In: Chorus I, Bartram J (eds) Toxic Cyanobacteria in Water: A guide to their public health consequences monitoring and management. E & FN Spon, p 317-333 WHO, UNESCO, UNEP (1999c) Chapter 12. Determination of cyanobacteria in the laboratory. In: Chorus I, Bartram J (eds) Toxic Cyanobacteria in Water: A guide to their public health consequences monitoring and management. E & FN Spon, p 334-361 WHO, UNESCO, UNEP (1999d) Chapter 13. Laboratory analysis of cyanotoxins. In: Chorus I, Bartram J (eds) Toxic Cyanobacteria in Water: A guide to their public health consequences monitoring and management. E & FN Spon, p 362-400 Wilson EK (2014) Danger From Microcystins In Toledo Water Unclear Chemical and Engineering News. American Chemical Society Wood SA, Dietrich DR (2011) Quantitative assessment of aerosolized cyanobacterial toxins at two New Zealand lakes. Journal of Environmental Monitoring 13(6):1617-1624 Wood SA, Dietrich DR, Cary SC, Hamilton DP (2012) Increasing Microcystis cell density enhances microcystin synthesis: a mesocosm study. Inland Waters 2(1):17-22 Wood SA, Rueckert A, Hamilton DP, Cary SC, Dietrich DR (2011) Switching toxin production on and off: intermittent microcystin synthesis in a Microcystis bloom. Environmental Microbiology Reports 3(1):118-24 World Health Organization (2003a) Cyanobacterial toxins: Microcystin-LR in Drinking-water: Background document for development of WHO Guidelines for Drinking-water
186 11. Bibliography
Quality. In. http://www.who.int/water_sanitation_health/water- quality/guidelines/chemicals/microcystin/en/ Accessed 04.06.2018 World Health Organization (2003b) Guidelines for safe recreational water environments. World Health Organization, Malta World Health Organization (2017) Guidelines for Drinking-Water Quality, 4 edn. World Health Organization, Geneva Wu Z, Ramsundar B, Feinberg EN, Gomes J, Geniesse C, Pappu AS, Leswing K, Pande V (2018) MoleculeNet: a benchmark for molecular machine learning. Chem Sci 9(2):513- 530 Xie L, Rediske RR, Gillett ND, O’Keefe JP, Scull B, Xue Q (2016) The impact of environmental parameters on microcystin production in dialysis bag experiments. Scientific Reports 6:38722 Yadav L, Tamene F, Goos H, van Drogen A, Katainen R, Aebersold R, Gstaiger M, Varjosalo M (2017) Systematic Analysis of Human Protein Phosphatase Interactions and Dynamics. Cell Syst 4(4):430-444.e5 Yang J, Roe SM, Cliff MJ, Williams MA, Ladbury JE, Cohen PTW, Barford D (2005) Molecular basis for TPR domain-mediated regulation of protein phosphatase 5. The EMBO Journal 24(1):1-10 Yu S-Z (1995) Primary prevention of hepatocellular carcinoma. Journal of Gastroenterology and Hepatology 10(6):674-682 Yuan M, Carmichael WW, Hilborn ED (2006) Microcystin analysis in human sera and liver from human fatalities in Caruaru, Brazil 1996. Toxicon 48(6):627-640 Zegura B (2016) An Overview of the Mechanisms of Microcystin-LR Genotoxicity and Potential Carcinogenicity. Mini Reviews in Medicinal Chemistry 16(13):1042-62 Zemskov I, Altaner S, Dietrich DR, Wittmann V (2017) Total Synthesis of Microcystin-LF and Derivatives Thereof. Journal of Organic Chemistry 82(7):3680-3691 Zemskov I, Kropp HM, Wittmann V (2016) Regioselective Cleavage of Thioether Linkages in Microcystin Conjugates. Chemistry- A European Journal 22(31):10990-7 Zervou SK, Christophoridis C, Kaloudis T, Triantis TM, Hiskia A (2017) New SPE-LC- MS/MS method for simultaneous determination of multi-class cyanobacterial and algal toxins. Journal of Hazardous Materials 323(Pt A):56-66 Zhai C, Zhang P, Shen F, Zhou C, Liu C (2012) Does Microcystis aeruginosa have quorum sensing? FEMS Microbiology Letters 336(1):38-44 Zhang M, Yogesha SD, Mayfield JE, Gill GN, Zhang Y (2013) Viewing serine/threonine protein phosphatases through the eyes of drug designers. Febs j 280(19):4739-60 Zheng C, Zeng H, Lin H, Wang J, Feng X, Qiu Z, Chen JA, Luo J, Luo Y, Huang Y, Wang L, Liu W, Tan Y, Xu A, Yao Y, Shu W (2017) Serum microcystin levels positively linked with risk of hepatocellular carcinoma: A case-control study in southwest China. Hepatology 66(5):1519-1528 Zhou M, Tu WW, Xu J (2015) Mechanisms of microcystin-LR-induced cytoskeletal disruption in animal cells. Toxicon 101:92-100 Zhou Y, Xu X, Yu B, Yu G (2017) Characterization of in vitro effects of microcystin-LR on intestinal epithelial cells. 32(5):1539-1547 Zilliges Y, Kehr JC, Meissner S, Ishida K, Mikkat S, Hagemann M, Kaplan A, Borner T, Dittmann E (2011) The cyanobacterial hepatotoxin microcystin binds to proteins and increases the fitness of microcystis under oxidative stress conditions. PloS One 6(3):e17615
187