A structural and functional analysis of the anti-
tumour antibody 14F7: providing new insight into the NeuGc GM3 binding mechanism
Thesis for the degree of Philosophiae Doctor By
Hedda Johannesen
Department of Biosciences Department of Chemistry
The Faculty of Mathematics and Natural Sciences University of Oslo
Acknowledgements
First and foremost, I would like to thank my main supervisor Ute Krengel. Thank you for creating a project based on my interests, for always having my back, supporting and guiding me. You are an amazing person, a brilliant supervisor, and I have really appreciated working with you. A huge thanks to my co-supervisor Hans-Petter Hersleth for giving me the opportunity to take a PhD in the first place. Thank you for all the good conversations, the teaching experiences, the trips and conferences we have had, and for always being such a kind- and full-hearted person. Another huge thanks to my co-supervisor Geir Åge Løset for pointing me in the right direction, your knowledge is incredible, and I am very thankful for all your inputs and our discussions. I would also like to direct a thanks to K. Kristoffer Andersson for making it possible to take this PhD in the first place.
I am grateful that I got the chance to work in the laboratory of Derek Logan for six months in Lund, Sweden. Thank you for supervising me and including me in your work. Moreover, thanks to Ünal Coskun for letting me work in your lab in Dresden, and an e massive thanks to Michal Grzybek for using so much of your time teaching and guiding me, discussing and being so genuinely kind, patient and brilliant, I really appreciated all your help. Thanks to Gregor Anderluh for allowing me to do the protein-liposome SPR spectroscopy in your lab in Ljubljana and thanks to Aleksandra Šakanović for helping with the experiments, as well as all the fun. I would also like to thank Nina F. J. Edin and Joe A. Sandvik for allowing me to do the eukaryote cell work and flow cytometry in your lab and supervising me in the process.
A huge thanks to Marta, Marie, Ingvild and Inger. I still miss sharing office with you guys, all the girl talks and good conversations both scientifically and socially. Thank you Ingvild for helping me with a change of scenery. An extra thanks to Marta for always being there, ready for helping me out on the lab or just hanging outs. For all great time at the conferences and the fun yet to come.
way, you are amazing. To Kaare for helping me re-familiarize with the project, for all the great discussions and being an impeccable writer. Thank you, Helen, you were our lab-mom, I miss you, your unfathomable kindness and your organisational skills. A special thanks to Julie for always being so kind, fun, genuine and a good friend (+super smart). Thanks to Henrik for being my fellow PhD in the group, it’s always fun plotting new ways of making fun of the Danish (sorry/not-sorry). Thanks to Dani for being both a good friend and support in the end of my master, to Daniel and Øyvind for making the lab work during our master more social and fun. Thanks to Kira for good conversations and of course the much-appreciated spell check in the last few days before delivery (so, if you find an error it’s not my fault ok?). Thanks to Joel, Tamjid, Clara, Abelone, Natalia, Dipankar, Steffi, Betty, Helene, Josefine, Victoria and Nico, making work a good place to be, for contributing with a positive social and work environment.
Thanks to my master students Kristoffer and Giacomo for giving me someone to discuss the project on a day-to-day basis, keeping me company, question my routines and help sharpening my brain.
In addition, I would like to thank the bachelor gang: Victoria, Ellen, Lena, Ida and Marte. Our study sessions and social community throughout our education really helped me to thrive in academia. I also want to give a special thanks to the social community at IBV, Lisa, Erna, Bie, Madeleine, Bettina, Ignacio, Mads, Jedrek, Melanie, Ida, Ada, Niels, Helga and Anette, making every work-event a social highlight.
A massive thanks to Øystein Foss for introducing me to science communication during my bachelor. You really made an impact on my life, letting me play around and have fun with chemistry, making things explode, burn and react and more importantly helping me start communicating science out to a broader audience. It has given me so much energy and joy. I would also like to direct a thanks to Line A. H. Valbø and Morten Berntsen at the Chemistry administration for going out of their way to help me setup home office when Sars-Cov-2 restrictions appeared. And thank Anna V.B. Mazzarella and Andreas Beesoo-Amundsen at IBV for helping me through the submission process.
I am indebted to my friends and family, who has constantly supported me throughout my ups and downs. A special thanks to Rohit for taken me in under your wing when I lived in Lund, Sweden. You and Francesco were amazing when I arrived and did not know anyone. You shared your social network and always supporting me in my work. An extra thanks to Jennifer for also becoming my good friend in Lund. You really stepped in when Francesco passed away,
thank you so much for the good conversations and the lovely girl time. I also want to give an extra thanks to Fredrik for your relentless encouragement and good conversations, and to my mom and dad for always supporting me.
I would like to direct a distinct thanks to Sars-Cov-2 for giving me an extra month dedicated to writing, and thanks for screwing up the last month of intensive lab work. Thank you for probably saving me the expenses for a PhD-party, and for not giving me a reason to buy an expensive PhD-defence dress. Thanks…
It has been a challenging journey with a lot of unexpected turns, that has tested my patience, adaptability, courage and so much more. But it has also been an experience that has shaped and developed me as a person, with immense support from all of you. I just want to say thank you and mention how lucky I am to know you all.
Finally, I owe Jonas a debt of gratitude for his everlasting patience, for always being understanding when I am frustrated or sad. You have been amazing the past few weeks, dealing with my stress, anxiety and helping me out whenever you could, you are incredible! And importantly, thanks for all the comfort, fun, joy, and love you have brought into my life.
June, 2020
Abstract
The focus of this thesis has been to elucidate the molecular basis for the specificity of an antibody termed 14F7 that exhibits potential for immunotherapy in cancer. When a healthy cell transforms into a cancerous cell, it exhibits altered gene expression and proliferation. Some of these changes can be utilised to distinguish a cancerous from a normal cell, and thus be exploited in cancer immunotherapy for diagnostic and therapeutic purposes. One such change, is the expression of the NeuGc GM3 ganglioside that is composed of a trisaccharide with a hydrophobic tail. Gangliosides are important components in a cell membrane, helping the cells communicate with each other. NeuGc GM3 is highly similar to the most common ganglioside in humans, namely the NeuAc GM3, differing only by a terminal hydroxyl-group at NeuGc instead of a hydrogen in NeuAc (Figure 1A). NeuGc GM3 is located at the cell surface of several different tumours, but does not exist in healthy human cells. The ganglioside is recognised by a monoclonal IgG antibody named 14F7. To be able to exploit the favourable properties of 14F7, it would be beneficial to know precisely how the NeuGc GM3 ganglioside and the 14F7 antibody interact at both the molecular and atomic level.
To investigate how they interact at the atomic level it is desirable to obtain a protein structure.
To this end, we generated four recombinant 14F7 variants in single chain variable fragment (scFv) format, expressed the proteins using an established periplasmic expression system and crystallised one of the scFvs. Recently, we soaked the trisaccharide part of NeuGc GM3 into the scFv crystals, revealing the binding mode of the ganglioside, where 14F7 utilises water molecules to create an extensive hydrogen-network to recognise the special hydroxyl-group of carbohydrate (Figure 1BC).
In follow-up studies, we investigated the interaction between 14F7 and NeuGc GM3 when the ganglioside was embedded in a lipid bilayer. Here, we observed that there was a concentration threshold for 14F7 binding. However, this threshold could be reduced if the similar ganglioside NeuAc GM3 was present together with NeuGc in the membrane. Binding was also abrogated if the membrane curvature was high. Together, this indicates that the 14F7 binding is likely
In this work we have designed, produced, characterised and crystallised 14F7 in scFv format.
Then, we have used liposomes to investigate 14F7-NeuGc GM3 binding dynamics. Together, this provides new insight into the functionality of 14F7 and highlights its potential within immunotherapy.
Figure 1: A. Graphic representation of antibody B. View of the 14F7 NeuGc GM3 binding site (scFv crystal structure PDB: 6S2I). The trisaccharide of NeuGc GM3 is highlighted with a dashed box.
C. Schematic of GM3 structure, where R can be either a hydrogen (NeuAc GM3) or a hydroxyl-group (NeuGc GM3). The protein image in PyMOL, the ganglioside structure was created with ChemDraw, and the overall layout with biorender.com
Sammendrag
Arbeidet i denne avhandlingen forsøker å belyse spesifisiteten til et antistoff som har en potensiell fremtid innen kreftterapi. Når en frisk celle får uønskede mutasjoner vil den som oftest dø, men noen ganger kan den i stedet danne en kreftcelle. Kreftceller er kjent for å ha ukontrollert celledeling, samt uregelmessig genregulering og protein produksjon. Slike forandringer gjør det mulig å separere friske og syke celler, og dette kan man utnytte i kreft- immunterapi og kreft-diagnostikk. Et kreftspesifikt molekyl er NeuGc GM3 gangliosidet, som består av et trisakkarid med en hydrofob lipid-hale (Figur 1C). Gangliosider er viktige komponenter i cellemembranen som blant annet hjelper celler å kommunisere. NeuGc GM3 er molekylært svært lik en av de vanligste gangliosidene i kroppen, NeuAc GM3. Det eneste som skiller dem er at NeuGc har en hydroksyl-gruppe der NeuAc har et hydrogen atom. NeuGc GM3 er blitt identifisert på celleoverflaten til et bredt utvalg av kreftsvulster, men finnes ikke på overflaten til friske celler. Gangliosidet NeuGc GM3 kan bli gjenkjent av det monoklonale IgG antistoffet 14F7 (Figur 1A). For å bruke 14F7 i immunterapi er det gunstig å vite så mye som mulig om hvordan 14F7 og NeuGc GM3 interagerer på molekylært og overordnet nivå.
En teknikk for å få informasjon på molekylært nivå er å se på krystallstrukturen. I den forbindelse designet vi fire varianter av rekombinant 14F7 i scFv-format (fra engelsk: single- chain variable-fragment). Vi produserte proteinene ved å bruke en etablert protokoll for periplasmisk Escherichia coli-uttrykk, for deretter å rense og krystallisere 14F7 scFv. Vi har også nylig klart å få et innblikk i hvordan 14F7 binder til NeuGc GM3 ved å tilsette trisakkarid- delen av NeuGc GM3 til scFv krystaller for deretter å løse strukturen av komplekset (Figur 1B).
Strukturen viste at 14F7 klarer å skille mellom NeuGc og NeuAc ved å benytte seg av et hydrogenbindingsnettverk mellom proteinet, karbohydratet og vannmolekyler.
For å utforske hvordan 14F7 og NeuGc GM3 interagerer på et mer overordnet nivå, har vi designet et oppsett hvor NeuGc GM3 er en del av en modellmembran. Her observerte vi at 14F7 måtte overkomme en NeuGc GM3 konsentrasjonsavhengig terskel for å binde til gangliosidet.
Til vår store overraskelse, kunne denne terskelen senkes ved å tilsette det svært like NeuAc
I dette arbeidet har vi designet, produsert, karakterisert og krystallisert 14F7 i scFv format, for deretter å undersøke hvordan 14F7 interagerer med NeuGc GM3 m i en modellmembran.
Tilsammen gir dette arbeidet ny innsikt for hvordan 14F7 gjenkjenner og interagerer med NeuGc GM3, samt bidrar til å kvalifisere 14F7 for fremtidig bruk innen immunterapi.
Figur 1. A. Skjematisk fremstilling av et antistoff. B. 14F7 scFv (PDB: 6S2I) hvor trisakkaridet til NeuGc GM3 er markert i en stiplet boks. C. GM3 gangliosidet hvor R kan enten være et hydrogen (NeuAc GM3), eller en hydroksylgruppe (NeuGc GM3. Protein bildet i PyMOL, gangliosidstrukturen er laget i ChemDraw og Figuren er satt sammen med biorender.com
TABLE OF CONTENTS
ACKNOWLEDGEMENTS _________________________________________________________________ I ABSTRACT _____________________________________________________________________________ V SAMMENDRAG ________________________________________________________________________ VII ABBREVIATIONS _______________________________________________________________________ XI LIST OF PAPERS _______________________________________________________________________ XV 1 INTRODUCTION ____________________________________________________________________ 1
1.1 EXPOSING THE SWEET SIDE OF THE CELL _____________________________________________________ 1 1.2 MEMBRANE ORGANISATION ____________________________________________________________ 2 1.3 GLYCOLIPIDS ______________________________________________________________________ 6 1.4 THE ORIGIN OF NEUGC ______________________________________________________________ 17 1.5 THE ADAPTIVE IMMUNE SYSTEM ________________________________________________________ 24 1.6 IMMUNOTHERAPY __________________________________________________________________ 39 1.7 MONOCLONAL ANTIBODY 14F7 ________________________________________________________ 50 1.8 WHERE THE FIELD IS HEADING __________________________________________________________ 61 2 AIMS ______________________________________________________________________________ 63 3 SUMMARY OF PAPERS I-II _________________________________________________________ 65 4 DISCUSSION AND FUTURE PERSPECTIVE ___________________________________________ 67
4.1 SCFV CONSTRUCT DESIGN _____________________________________________________________ 67 4.2 ANTIBODY VERSUS SCFV AFFINITY ________________________________________________________ 68 4.3 14F7 SPECIFICITY __________________________________________________________________ 69 4.4 MEMBRANE INTERACTIONS ____________________________________________________________ 70 4.5 14F7 MAY INDUCE ONCOSIS-LIKE CELL DEATH _______________________________________________ 72 4.6 14F7 POTENTIAL IN CANCER IMMUNOTHERAPIES _____________________________________________ 74 5 METHODS _________________________________________________________________________ 77 5.1 CLONING, EXPRESSION AND PURIFICATION __________________________________________________ 77 5.2 X-RAY CRYSTALLOGRAPHY _____________________________________________________________ 81
Abbreviations
14F7hT Humanised monoclonal 14F7 IgG1 antibody Ab1 Original antibody
Ab2 Anti-idiotypic antibody Ab3 Anti-anti-idiotypic antibody ACT Adoptive T-cell therapy ADC Antibody-drug-conjugate
ADCC Antibody-dependent cellular cytotoxicity ATP Adenosine triphosphate
APC Antigen-presenting cell BCR B-cell receptor
CAR-T Chimeric antigen T-cell receptor C1q Complement component 1q C1 Construct 1, 14F7 scFv CD1 Cluster of differentiation 1
CDC Complement-dependent cytotoxicity CDR Complementarity-determining region CH Constant heavy domain
CMAHP CMAH pseudogene
DAMPS Damage-associated molecular patters DLS Dynamic light scattering
E. coli Escherichia coli
EGF Epidermal growth factor
EGFR Epidermal growth factor receptor
EIA Electrochemiluminescence immunoassay ELISA Enzyme-linked immunosorbent assay Fab Fragment, antigen binding
Fc Fragment, crystallisable FcRn Neonatal Fc-receptor
FDA USA Food and Drug administration Fv Fragment, variable
GalCer Galactosylceramid Gb3 Globotriasylceramide
GF Growth factor
GFR Growth factor receptor GlcCer Glucosylceramide
GM1 Monosialotetrahexosylganglioside GM3 Monosialodihexosylganglioside GSL Glycosphingolipid
GUV Giant unilamellar vesicle
HAMA Human anti-murine antibody
HD Hanganutziu-Deicher
HRP Horse radish peroxidase
Ig Immunoglobulin
IHC Immunohistochemistry
IL Interleukin
iNK Invariant natural killer cell
IPTG Isopropyl 𝛽-d-1-thiogalactopyranoside LacCer Lactosylceramide
Ld Liquid disordered state Lo Liquid ordered state Lnm Lymph node metastasis LUV Large unilamellar vesicle mAb Monoclonal antibody
MHC Major histocompatibility complex MLV Multilamellar vesicle
MSD-ECL Meso scale discovery -electrochemiluminescence
MP Macrophage
Neu3 Neuraminidase 3
NeuAc N-acetyl neuraminic acid
NeuGc GM3 Neu5Gcα2-3Galβ1-4Glcβ monosialodihexosylganglioside NK Natural killer cell
NSCLC Non-small cell lung carcinoma
OMPC Outer membrane-protein complex of Neisseria meningitides Ori Origin of replication
PDB Protein Data Bank
Pns Primary nervous system of paediatric patients scFv Single-chain variable fragment
SDHB Succinate dehydrogenase complex subunit B SO Solid ordered state
SPR Surface plasmon resonance TCR T-cell receptor
TFH T-follicular helper-cell TLC Thin-layer chromatography Tm Melting temperature
VH Variable heavy domain VL Variable light domain
VSSP Very-small-size proteoliposomes
VT Verotoxin
WT Wild type
List of papers
This work is based on the following papers I-II, which will be referred to in the text by their Roman numerals:
I. Crystal structure of an L chain optimised 14F7 anti-ganglioside Fv suggests a unique tumour-specificity through an unusual H-chain CDR3 architecture Kaare Bjerregaard-Andersen#, Hedda Johannesen#, Noha Abdel-Rahman, Julie E.
Heggelund, Helene H. Mykland, Fana Abraha, Paula A. Bousquet, Lene S. Høydahl, Daniel Burschowsky, Gertrudis Rojas, Stefan Oscarson, Geir Åge Løset, Ute Krengel.
Scientific reports (2018) DOI: 10.1038/s41598-018-28918-5
II. Recognition of the N-glycolyl GM3 ganglioside by the anti-tumor antibody 14F7:
Crystal structure and membrane interactions
Hedda Johannesen#, Kaare Bjerregaard-Andersen#, Fana Abraha, Aleksandra Šakanović, Daniel Großer, Ünal Coskun, Gregor Anderluh, Stefan Oscarson, Ernesto Moreno, Michal Grzybek, Ute Krengel.
Manuscript
#Authors contributed equally
1 Introduction
“Nothing in biology makes sense except in the light of evolution”
(Theodosius Dobzhansky, 1973)
1.1 Exposing the sweet side of the cell
While cancer likely emerged in the same era as multicellular organisms (Aktipis et al., 2015), our first historical evidence for cancer dates back to the time of the dinosaurs (Rothschild et al., 1999). The disease we recognise as cancer today was initially documented by the Egyptians in 3000 B.C. (Breasted, 1930). Nevertheless, it took almost another 5000 years before a British surgeon, in 1775, noted that cancerous growth was common in the scrotum of chimney sweepers, and thus established the first cause of cancer (Brown and Thornton, 1957). Today, it is well known that some of the hallmarks of cancer include uncontrolled proliferation, genomic dysregulation, cellular immortality and metabolic dysregulation (Hanahan and Weinberg, 2011). If a cancer cell generates a mass of abnormal cells, it can be defined as a tumour.
Traditionally, the most common treatments are surgery, radiation- and chemotherapy, but there is no guarantee for relapse-free survival. Modern cancer treatment still includes these techniques but has also incorporated more sophisticated strategies such as pathway-inhibition and immunotherapy. These are all based on specifically targeting the cancer, through drug injections or activation of the patients’ immune system, intending to shield the healthy cells from damage. The challenges for immunotherapy are many, and one of them is detection of robust cancer-specific targets with minimal side effects. Even though cancer cells might display neo- and xenoantigens (molecules that arise from genetic mutations, or exogenous sources, respectively), they can adopt self-defence mechanisms shielding them from recognition. One such mechanism includes ganglioside shedding, where the cancer cells are thought to insert cancer-related gangliosides into immune cells: down-regulating the immune response by using
Since the sialic acid NeuGc is not naturally produced in humans, it is a cancer-associated antigen (Irie et al., 1998;Chou et al., 2002;Suzuki, 2006), making it an attractive target for passive and active immunotherapy, as well as for diagnostic purposes. An antibody raised to recognise the ganglioside is 14F7, the first reported antibody binding a carbohydrate epitope that belongs to the IgG family (Carr et al., 2000). 14F7 is a monoclonal antibody (mAb) that has been studied in phase I/II clinical trials targeting breast cancers and its binding capacity and specificity makes it an appealing candidate for immunotherapy (Oliva et al., 2006). Still, the binding mechanism and mode of interaction has remained unknown until now.
1.2 Membrane organisation
“When the earth came alive it began constructing its own membrane, for the general purpose of editing the sun”
(Thomas Lewis, 1974)
A cell is defined as the smallest functional unit of an organism, enclosed by a membrane bilayer, also known as the plasma membrane, consisting of lipids, proteins and other membrane-related molecules (Figure 1). The membrane functions not only as a vessel for its content, but also as a gatekeeper: regulating what enters and exits, sensing the surrounding environment and exchanging stimuli that affect both its interior and the outside world. Consequently, the membrane has a complex composition, dependent not only on the organism itself, but also on the cell subtype. The membrane bilayer of human cells consists of three main classes of lipids:
sterols, phospholipids and glycolipids. The term ‘lipid’ derives from the Greek word ‘lipos’ for fat. Lipids serve several functions, including energy storage, as messengers in signal transduction and molecular recognition processes and as components of plasma membranes.
The earliest experiment on the function of the plasma membrane was the microscope study of the membrane shape of erythrocytes (red blood cells) by W. Hewson in 1773. When Hewson added water to the cells they changed shape from flat to spherical, while they simply dissolved if too much water was added, providing the first evidence cellular osmosis (Kleinzeller, 1996).
Another important discovery was done by C. H. Schultz in 1836, who stained erythrocytes with iodine and managed to estimate the membrane thickness of the cell to approximately 220 Å
(Schultz, 1836), not too far from the currently accepted value of plasma membrane thickness of 50~100 Å. A few years later T. Schwann published his cell model where the word membrane was initially introduced (Schwann et al., 1847). From those early observations, a broad research field has emerged around the study of the cell membranes, which ultimately led to a ‘fluid mosaic’ model in the early 1970’s, (Singer and Nicolson, 1972). Among other things, the model explains how transmembrane proteins with an amphipathic structure can reside in a hydrophobic bilayer, and it posits that some lipids might specifically interact with each other, but are mostly fluently dispersed (Singer and Nicolson, 1972).
Figure 1. Schematic illustration of a membrane. The membrane consists of a lipid bilayer (1) with cholesterol (2) embedded into the membrane providing a stabilizing effect. Transmembrane (3) proteins and protein channels (4) have several different functions, such as ionic and molecular transmembrane transport and signal transduction. A variety of different peripheral proteins (5) can be associated with the membrane. At the cytosolic side of the membrane, the cytoskeleton (6) forms a network throughout the cells. The extracellular space is covered with different glycans, which can be either attached to proteins (7) or lipid-bound (8), as is the case for the ganglioside NeuAc GM3 (9). The Figure was created with biorender.com.
1.2.1.1 Membrane asymmetry
In the 70’s it became clear that erythrocyte membranes exhibit an asymmetric content distribution between their inner and outer lipid face. The first clue was the observation of an asymmetric arrangement of proteins between the inner and outer leaflet of the erythrocytes’ cell membrane (Bretscher, 1972;Verkleij et al., 1973). This was also the case for other cell types (Devaux, 1991;Zachowski, 1993), and it is now established as a standard property of cell membranes.
Around the same time it was discovered that epithelial cells had a different glycosphingolipid (GSL) content within the same membrane leaflet, defining a distinct lateral asymmetry (Figure 2). On the outer apical side there were much more GSLs compared to the outer basolateral side of the cell (Koichi et al., 1974;Brasitus and Schachter, 1980;Chapelle and Gilles-Baillien, 1983). The apical and basolateral side are separated by tight junctions, forming a barrier that prevents lipids in the outer leaflet to move between the two regions, while the lipids in the inner leaflet can diffuse freely (Dragsten et al., 1981;Dragsten et al., 1982;Spiegel et al., 1985;van Meer and Simons, 1986).
Figure 2. A simplified illustration of two epithelial cells. The different membrane sections are indicated, as well as the tight junctions that effectively separate the outer leaflet of the apical membrane from the rest, while gap junctions allow ions and small molecule-transport between the cells. The Figure was created with biorender.com.
1.2.1.2 Membrane sorting
The conundrum of how it was possible to establish and maintain different molecular compositions in the outer apical and basolateral membrane of epithelial cells was solved in the mid 80’s. It was discovered that newly synthesised glycoproteins located in the late trans-Golgi network were sorted into vesicles that were selectively transported to the apical or the basolateral side of the epithelial cells (Simons and Fuller, 1985;Griffiths and Simons, 1986;Pfeffer and Rothman, 1987;Hughson et al., 1988). Van Meer et al. (1987) used a fluorescent ceramide analogue to investigate how sphingolipids were sorted within epithelial cells. They discovered that GSLs, too, were sorted to the outer membrane leaflet of either the apical or basolateral side (van Meer et al., 1987). This was further elucidated by Klemm et al.
(2009), investigating the segregation of sphingolipids and sterols in yeast using immunoisolation combined with lipidomics. They confirmed that the trans-Golgi network was able to sort lipids, and that differences in lipid composition within the vesicles likely played a vital role in directing the sorting machinery (Klemm et al., 2009). Thus, it became clear that the vesicle components could direct trafficking to different regions of the membrane, explaining the lateral membrane separation observed in the case of epithelial cells.
1.2.1.3 Membrane domains
The lateral asymmetry observed between the apical and basolateral area of the epithelial cell membrane is an extreme form of membrane heterogeneity. However, a more confined asymmetrical formation is likely present within the same area of the lateral membrane, as local, transient clusters in the sub micrometre scale (<200 nm) (van Meer et al., 1987;Brown and Rose, 1992;Ahmed et al., 1997;Friedrichson and Kurzchalia, 1998;Varma and Mayor, 1998;Pralle et al., 2000). Such clusters could play a role in signal transduction pathways, where specific membrane molecules must assemble into functional domains to exert their function. A specific class of lipid clusters, named lipid rafts, are hypothesised to be transient, but relatively ordered membrane domains (Sezgin et al., 2017), enriched with cholesterol, (glyco)sphingolipids and raft-associated proteins (Brown and Rose, 1992;Schroeder et al., 1994;Hanada et al., 1995). Since the size of such clusters is too small for detection by optical microscopy techniques (lower limit of 250 nm resolution), they are mainly investigated
2009;Levental et al., 2009;Sáenz et al., 2012). Although structurally still elusive, biologically, it appears that rafts can provide a means for cell membranes to form dynamic platforms within the bilayer, functioning in membrane trafficking, signal transduction and cell polarization (Lingwood and Simons, Science 2010).
1.3 Glycolipids
“GSL masking [is] a potential dynamic ‘cloaking device’ in cellular physiology”
(Mahfoud et al., 2010)
Sphingolipids
Sphingolipids are a large group of molecules consisting of a hydrophobic tail that can be connected to a hydrophilic moiety. The hydrophilic head group can influence a wide range of functional traits; this includes cell-cell recognition, signalling events, growth, apoptosis, adhesion, migration and function as pathogen receptors (Bremer et al., 1986;Hakomori and Igarashi, 1995;Imberty and Varrot, 2008;Klokk et al., 2016). Sphingolipids were discovered in the 1870s from brain extracts, and due to the lack of knowledge about their function or role at the time, they were named after the mythical sphinx (Thudichum, 1884). They are predominantly found in the plasma membrane of cells, in the outer leaflet (Singer, 1974), and their structure consist of a long chain amino alcohol that is linked to a fatty acid via an amide bond, creating the ceramide (Figure 3A). The length of the two chains affect the molecular function and how the sphingolipid behaves in the membrane leaflet (Löfgren and Pascher, 1977;Kannagi et al., 1983). The resulting amphipathic molecule can consist of a hydrophobic tail and a hydrophilic head group protruding out of the membrane. This can lead to membrane stabilisation through lateral packing of the hydrophobic tails dominated by van der Waal forces and lateral hydrogen bonds if the head group is polar (Pascher, 1976). In tissues that are exposed to high physical stress, like the intestines, the sphingolipids generally contain head groups with extended hydrogen bonding capacity to give the membrane further stability (Singer, 1974). The most common sphingolipid is sphingomyelin, where the headgroup consists of phosphocholine.
If the head group consist of carbohydrates, a glycosphingolipid (GSL) is formed (Figure 3B).
Figure 3. Illustration of A. Sphingolipid consisting of a chemical group (R) connected to a ceramide that are composed of a sphingosine connected to a fatty acid chain. B. Glycosphingolipid, in this case globotriasylceramide (Gb3), where the carbohydrate head group is connected to an aglycone moiety, the ceramide. Galactose is represented with a yellow circle. The structures were created with ChemDraw, the overall layout with biorender.com.
Glycosphingolipids
While mammalian phospholipids usually contain fatty acid chains of 12-24 carbons, GSLs display longer fatty acid chains that are usually between 14-30 carbons or more (Schnaar, 2015).
This is in accordance with the finding that GSLs preferentially colocalise in the cholesterol- enriched Lo phase of model membranes, which generally has a thicker bilayer (Simons and Vaz, 2004;García-Sáez et al., 2007). The aglycone moiety of GSLs, which is the part connected to the glycan, consists of a ceramide for GSLs, contributes to additional variation for each GSL by its hydroxylation pattern, chain length and chain saturation (Figure 3B). GSLs are known to accumulate on the apical surface of epithelial cells (Koichi et al., 1974;Brasitus and Schachter, 1980;Chapelle and Gilles-Baillien, 1983), covering the cell surface like a sugar-lawn (Vicker
can also serve as receptors for several microorganisms (Karlsson, 1989), two examples being the glycosphingolipid monosialotetrahexosylganglioside (GM1), which is recognised by the cholera toxin, and globotriasylceramide (Gb3), which is the receptor of the Shiga-toxin and the Shiga-like toxin family of verotoxins (VT).
Crypticity
The term ‘glycolipid crypticity’ describes the inaccessibility for ligand interactions of the GSL molecules localised at the cell surface (Figure 4) (Hakomori, 1981;Hakomori, 1986;Lampio et al., 1986;Curatolo, 1987). The first evidence of glycolipid crypticity was reported by Koscielak et al. (1968), who produced anti-globoside antibodies from rabbits and investigated the expression of globosides on erythrocytes. Although there are large quantities of globoside on the erythrocyte cell surface, Koscielak et al. (1968) detected unexpectedly low binding by anti- globoside antibodies. However, when treating the erythrocytes with trypsin, the antibody reactivity was significantly enhanced, demonstrating that the globosides could be exposed by trypsin treatment. Further investigations showed that this was not only true for globoside, but also for other GSLs, and that there were differences in crypticity for these GSLs when comparing foetal and adult erythrocytes (Hakomori, 1969). This is the first reported finding that pools of GSLs, although present in the plasma membrane, were in fact inaccessible to GSL ligands, resulting in a completely new perspective on lipid organisation in the membrane.
Koscielak et al. (1968) were also the first to formulate an explanation for GSL crypticity, suggesting that it may be induced by protein crowding. This was further investigated by Peters et al. (1983) by examining the potential of macromolecules to induce GSL crypticity by shielding the glycolipids from neuraminidase recognition. For this purpose, they used albumin and different dextrans polymers covalently linked to an oleic acid for membrane incorporation.
Their results indicated that physical shielding by non-specific macromolecules, such as albumin and dextrans, was unlikely to be the source of the glycolipids’ crypticity, and suspected that it may actually be driven by specific interactions, as well as glycolipid dynamics, and membrane lipid composition (Peters et al., 1983).
Figure 4. A schematic representation covering some of the aspects of GSL crypticity. Glycolipids, such as Gb3 (1) and GM1s (2) located in the outer leaflet of the cell membrane are exposed for ligand interactions as in the case of verotoxin (3) and cholera-toxin (4). Aglycone variations such as ceramide chain length, saturation and hydroxylation and GSL-GSL interactions can modulate GSL binding accessibility. Externals factors, including cholesterol (5) influence crypticity by affecting membrane packing and its ability to tilt the GSL. Other factors that affects crypticity includes fatty acid chain length of the surrounding lipids, whether the lipids are saturated (6) or unsaturated (7), as well as the presence of interaction partners, including proteins (8), ions (9) and the cytoskeleton (10). N-acetyl neuraminic acid is depicted as a purple diamond; galactose – yellow circle; yellow square – N-acetyl galactosamine;
glucose – blue circle. The Figure was prepared with biorender.com.
1.3.3.1 Aglycone diversity and its effect on crypticity
Systematic investigation of aglycone diversity was executed by Stewart and Boggs (1993).
They explored carbohydrate GSL exposure using one of the simplest GSLs, galactosylceramide (GalCer), in model membranes and used GalCer oxidation by galactose oxidase to detect GalCer exposure. By increasing the fatty acid chain length from 16 to 26 carbons, they
of the ceramide was hydroxylated. Essentially, this indicated that the aglycones, fatty acid length and hydroxylation state affect GSL crypticity.
Another GSL type that has been thoroughly investigated regarding its crypticity is the neutral GSL globoside Gb3, also named CD77 or Pk blood group antigen (Johannes and Römer, 2010).
Gb3 is recognised by several proteins, including the Shiga-like verotoxins (Figure 4, Nr2) VT1 and VT2c of Escherichia coli (Bergan et al., 2012). Although both toxins are Gb3-specific, they bind Gb3 with different affinity (Head et al., 1991). Kiarash et al. (1994) compared the binding of two verotoxins, the VT1 and VT2c, against 14 different species of Gb3 in model membranes and revealed that both toxins favoured unsaturated Gb3 with a fatty acid exceeding 14 carbons, but they exhibited preferences for different Gb3 species. Whereas Gb3 with 22 carbons fatty acid chain was the best receptor for VT1, VT2c preferred a Gb3 species with a shorter fatty acid, containing 18 carbons (Kiarash et al., 1994). Molecular modelling suggested that the two verotoxins might use different binding sites, explaining the variation in binding for the two Gb3
variants (Nyholm et al., 1995). This is a good example of how GSL binding to different interaction partners, although similar, can be regulated through ceramide modifications, resulting in specific induced crypticity towards different ligands.
1.3.3.2 Lipid environment
The lipid environment also affects GSL crypticity, including Gb3 ligand availability. VT1 and VT2c binding to Gb3 has been shown to be inversely related to the surrounding phospholipid chain length (Arab and Lingwood, 1996). Thus, the exposure of different Gb3 epitopes can be regulated by the surrounding lipid membrane, rather than by the conformation of the carbohydrate portion. This has been extensively investigated by Schubert et al. (2020) who compared Gb3 binding to two lectins: LecA from Pseudomonas aeruginosa and the StxB B- subunit from Shigella dysenteriae. While both lectins recognise Gb3, they bind to different Gb3- containing membrane domains on the apical side of epithelial cells. In addition, StxB had the ability to recognise Gb3 on the primary cilium (Schubert et al., 2020). By adjusting the lipid composition, Schubert et al. (2020) managed to reconstitute these effects, demonstrating that StxB prefer a more ordered Gb3 conformation surrounded by a higher amount of saturated lipids. Based on these results, they hypothesised that the two proteins recognise different Gb3
carbohydrate orientations (Schubert et al., 2020) and further validated the use of cell membrane investigations combined with model membranes as a powerful tools to investigate lipid organisation and function.
Additionally, the lipid bilayer might not only change the GSL exposure, it can potentially also modulate it, as demonstrated for GalCer in membrane models (Stewart and Boggs, 1993). While detecting galactose oxidase enzyme activity on liposomes embedded with GalCer with and without sphingomyelin, it became clear that sphingomyelin had the ability to protect GalCer from enzyme recognition (Stewart and Boggs, 1993). Intriguingly, sphingomyelin was also able to modulate the enzyme recognition based on different GalCer species. In the absence of sphingomyelin, GalCer processing by galactose oxidase decreased with increased chain length of the fatty acid portion of the ceramide. When sphingomyelin was included as a membrane component, the enzyme activity remained constant regardless of the GalCer fatty acid chain length (Stewart and Boggs, 1993).
1.3.3.3 Cholesterol masking
While testing for verotoxin recognition of Gb3 on frozen erythrocyte sections, Lingwood et al.
(2011) observed that association of verotoxin with Gb3 decreased when cholesterol was extracted from the tissue. Further investigations of this phenomenon in model membranes containing either the glycolipid GM1 or Gb3, demonstrated that detection by their respective ligands (cholera or verotoxin, respectively) was clearly reduced in the presence of cholesterol, confirming its direct role in GSLs crypticity. Their molecular dynamic simulations of membrane-embedded GM1, showed similar findings to those reported by Yahi et al. (2010), i.e. that cholesterol can potentially induce a tilted GSL conformation of the glycan moiety, thereby explaining how cholesterol can contribute in regulating receptor crypticity (Lingwood et al., 2011). Furthermore, the cholesterol-dependent GM1 masking was maximized when the lipid composition heterogeneity was minimised, demonstrating that the inhibition can also be related to the membrane plane (Lingwood et al., 2011).
It has also been suggested that cholesterol-masking might be a general mechanism to induce GSL crypticity. Mahfoud et al. (2010) used detergent resistant model membrane vesicles to test cholesterol masking of Gb3, GalCer, SGC or GM1. The lipid vesicles were separated by a density gradient into two fractions, a small (~5 %) lighter fraction, and a larger (~95 %) heavier fraction. Binding to the different GSLs was probed using their respective ligands, and their
recognition (Mahfoud et al., 2010). They repeated this experiment by creating detergent resistant membrane vesicles of Vero cells and observed the same trend where VT only recognised Gb3 in the smaller lighter fractions of vesicles, indicating the potential physical relevance of cholesterol as a masking agent (Mahfoud et al., 2010).
1.3.3.4 GSL-GSL interactions
Mahfoud et al. (2010) also investigated how GSL-GSL interaction affected the crypticity of Gb3. They observed that Gb3 had affinity for the two GSLs GalCer and glucosylceramide (GlcCer) using a TLC overlay technique. They therefore investigated how their mutual interaction affected verotoxin recognition of Gb3 in detergent resistant vesicles. These vesicles either contained Gb3 and cholesterol only, or together with either of the two GSLs. When one of the other GSLs were present, verotoxin binding was significantly improved. Consequently, these two GSLs had the ability to counter the cholesterol effect and unmask GSL. Since Mahfoud et al. (2010) observed specific affinity between Gb3 together with either GalCer or GlcCer, they hypothesised that the presence of another GSL could counter cholesterol masking by direct GSL-GSL interactions.
1.3.3.5 Glycolipid diversity and its complexity
There are several other factors that can regulate glycolipid crypticity. This includes the membrane plane (Strömberg et al., 1991;Nyholm and Pascher, 1993a;b), divalent ions (Menikh et al., 1997), the involvement of the cytoskeleton (Mahfoud et al., 2010) and GSL-protein interactions (Russo et al., 2016). As already mentioned, GSL-GSL interactions can attenuate cholesterol-induced crypticity, but the presence of carbohydrates might also have the opposite effect as demonstrated by Lampio et al. (1986). Lampio et al. (1986) used galactose oxidase to investigate the enzyme’s ability to recognise several different glycolipids on erythrocytes, including asialo-GM3. This GSL remained undetectable until the cells were treated with neuraminidase. The neuraminidase induced removal of the sialic acids yielded a 4-8 fold increase in asialo-GM3 recognition, which suggested that carbohydrate crowding on the cell surface could affect GSL crypticity (Lampio et al., 1986).
GSL crypticity is a vast and constantly expanding field of study, and it is part of an intricate pattern of GSL regulation; however, other aspects influencing crypticity are outside the scope of this thesis and will not be discussed further.
1.3.3.6 Physiological role of crypticity
The physiological significance of crypticity could be to prevent undesired interaction, such as binding of bacterial toxins, as well as a means to regulate ligand interactions. GSL crypticity could potentially be used to avoid GSL-related autoimmune disorders (Quarles et al., 1986).
For example, healthy human serum contains naturally occurring anti- Gb3 antibodies (Huflejt et al., 2009). Masking of Gb3 through crypticity might prevent the autologous recognition of healthy cell, thus preventing autoimmune reactions. Further insights into the functional role of crypticity are provided by the finding that GSLs with different ceramide composition can be found at different developmental stages (O'Brien et al., 1964;Svennerholm, 1968;Shimomljra and Kishimoto, 1983;Palestini et al., 1990) and tissues (Ogawa-Goto et al., 1990). The different membrane composition modulates the binding pattern exposed on the cell surface, likely defining distinct functional roles for different cell types and regulating their interaction, providing tissue-specific diversity and adaptability throughout the organisms life-cycle.
Gangliosides
Gangliosides are generally characterised by a ceramide moiety decorated by a glycan head- group containing one or more sialic acids: nine-carbon carbohydrates with a carboxylic acid group. Gangliosides were discovered in nerve-tissue in the 1940s by Ernst Klenk, who named them after the German word for neurons ‘Ganglionzellen’ (Klenk, 1942). The ceramide portion is made in the endoplasmic reticulum before the carbohydrate moiety is added (Hanada et al., 2003). The first carbohydrate added to the ceramide for all gangliosides (except GM4) is a glucose followed by a galactose, creating lactosylceramide (LacCer). Thereafter the structure is further modulated by a range of glycosyltransferases, sialyltransferases and sialidases resulting in hundreds of different structures (Schwarz and Futerman, 1996;Kolter et al., 2002;Merrill, 2002;Futerman and Hannun, 2004;Yu et al., 2007;Shevchenko and Simons, 2010;Yu et al., 2011). The tight regulation of ganglioside synthesis generates tissue- and cell- specific ganglioside patterns that are readily incorporated into the plasma membrane (Maccioni, 2007). One of the most common gangliosides in humans, the monosialodihexosylganglioside (GM3), is also one of the simplest (Kiguchi et al., 1990). The enzyme sialyltransferase I adds
hydroxylase (cmah), which replaces the C5-hydrogen with a hydroxyl group, resulting in N- glycolyl GM3 (NeuGc GM3, Figure 5B) (Schlenzka et al., 1996).
Figure 5. Schematic image of monosialodihexosylganglioside (GM3) consisting of a ceramide moiety embedded in the lipid membrane and decorated by a glycan head consisting of a glucose, a galactose a sialic acid residue. A. If the sialic acid has a terminal hydrogen (in bold) it is a NeuAc sialic acid, generating NeuAc GM3. B. If the sialic acid has a hydroxyl group (in bold) it is a NeuGc GM3. N-acetyl neuraminic acid is represented as a purple diamond; N-glycolyl neuraminic acid – light blue diamond;
galactose – yellow circle; glucose – blue circle. The molecular structures were prepared in ChemDraw and modified with biorender.com.
1.3.4.1 Ganglioside functions
Gangliosides are ubiquitous components of cell membranes. As such, they are involved in several physiological processes including apoptosis, differentiation, cell-adhesion, migration and proliferation, by modulating cell signalling between cell-to-cell and cell-to-matrix interactions (Kojima and Hakomori, 1991;Zurita et al., 2001;Proia, 2003;Daniotti et al., 2006;Todeschini and Hakomori, 2008;Lopez and Schnaar, 2009;Julien et al., 2013). Another important ganglioside function worth mentioning is its function as a receptor for pathogen toxins, such as for the cholera toxin, which recognises the GM1 ganglioside (Heggelund et al., 2015). Cells can also actively shed gangliosides from the membrane. These circulating gangliosides can then be taken up by other cells where they are randomly inserted into the cell
membrane. This happens for normal healthy cells, but ganglioside shedding is especially upregulated during cancer progression in tumours such as glioblastomas, renal carcinoma, melanoma and astrocytoma (Kloppel et al., 1977;Portoukalian et al., 1978;Olshefski and Ladisch, 1996;Lauc and Heffer-Lauc, 2006;Lardone et al., 2014).
NeuAc GM3 can form clusters as part of detergent-resistant membrane-domains or by associating with the caveolae (Kabayama et al., 2007). NeuAc GM3-containing clusters were reported to be sensitive to cholesterol depletion (Fujita et al., 2007), and were abrogated when the cytoskeleton were perturbed (Fujita et al., 2009). Such clusters were also shown to be associated with a variety of transductor molecules, including c-src in mice (Iwabuchi et al., 1998;Prinetti et al., 1999). NeuAc GM3’s exact role in these mechanisms is unknown, but NeuAc GM3 clusters seems to function within cell mobility, growth and signalling events (Merritt et al., 1984;Kojima and Hakomori, 1989;1991;Iwabuchi et al., 1998;Prinetti et al., 1999;Slevin et al., 1999;Visco et al., 2000;Fujita et al., 2007;Todeschini and Hakomori, 2008;Fujita et al., 2009;Hakomori, 2010;Prinetti et al., 2011). The different functions of NeuGc compared to NeuAc clusters, however, has yet to be elucidated.
1.3.4.2 Gangliosides’ role in cancer
The first indication that GSLs composition might change during oncogenesis was discovered in the late 1960’s (Hakomori and Murakami, 1968). Today it is well known that cancer cells exhibit aberrant ganglioside expression patterns (Hakomori, 1996;Birklé et al., 2003;Daniotti et al., 2016). This can easily occur during oncogenesis due to blockage of GSL synthesis that can lead to accumulation of ganglioside structural intermediates (Hakomori, 1981). If the cancer cells express anomalous glycosyltransferases that are usually expressed in other tissues this will also contribute to aberrant ganglioside patters (Hakomori, 1981). Tumours can utilise the gangliosides regulatory effect to initiate in immunoediting. Thus, the immune system and the cancer cells are interacting in a cat and mouse-like relationship where the immune system tries to eradicate the tumour, while the tumour adapts and evolves to avoid detection (Dunn et al., 2002;Dunn et al., 2004). Tumours defend themselves against the immune system by creating an immunosuppressive environment (Boligan et al., 2015). This includes ganglioside shedding,
neuroblastoma tumour cells (Ladisch and Wu, 1985;Ladisch et al., 1994b). The GD2 species containing a shorter fatty acid chain are selectively released (Li and Ladisch, 1991). This is likely related to function, as synthetic GD2 with shorter fatty acids have been shown to more efficiently suppress lymphocyte activity in vitro (Ladisch et al., 1995). The underlying mechanism might be that shorter fatty acids makes the carbohydrate epitope more available, since the GD2 ceramide could be replaced with a synthetic lipid of similar length (Ladisch et al., 1994a).
Ganglioside shedding can for example reduce necrosis (Ziegler-Heitbrock et al., 1992) and disrupt the development and maturation of immune cells (Peguet-Navarro et al., 2003).
Gangliosides can also impair fragment crystallisable (Fc) receptor expression on B-cells (Hoon et al., 1989;Kimata, 1995), they can directly downregulate T-cell proliferation by preventing the growth-factor interleukin (IL)-2 cytokines to bind to T-cells (Chu and Sharom, 1993), or reduce the major histocompatibility complex class II (MHCII) expression, preventing immune recognition (Caldwell et al., 2003;Shen et al., 2005). There are several other mechanisms by which gangliosides affect the immune system, helping the tumour escape from immune recognition. There is also accumulating evidence indicating that tumour ganglioside shedding is not only immune-repressive, but also associated with increased cell growth, and that ganglioside shedding can potentially accelerate proangiogenic (blood vessel formation) activity in vivo (Liu et al., 2014a).
As gangliosides accumulate in the outer leaflet of the membrane, they are valuable targets for cancer therapy. Therefore, knowledge of how gangliosides are regulated in cancer cells could help develop more specific drugs against tumours. To be able to understand how ganglioside regulation relates to cancer progression, it is important to investigate the ganglioside composition both within membrane domains, and the variations between different tissues (Zamfir, 2014;Groux-Degroote et al., 2015). In addition to upregulation of certain gangliosides in tumours, there is also an increased expression of neoantigens and xenoantigens, including expression of NeuGc GM3 that is normally not present in healthy human tissues (Zhang et al., 1997;Tangvoranuntakul et al., 2003;Nguyen et al., 2005). Such antigens are especially valuable as targets for passive and active immunotherapy (see Section 1.6 Immunotherapy).
1.4 The origin of NeuGc
“In physics, if there is something that we don’t quite understand, we call it ‘complex’.
Alternatively, we often call it ‘effective’. However, as a synonym, we could also call it ‘sugary’”
(Ilpo Vattulainen, 2020)
The absence and reappearance of NeuGc
While NeuGc is a common sialic acid in most deuterostomes, some species lack NeuGc expression. This includes among others the pinnipeds and musteloids (e. g. seals, weasels, badgers and otters) (Ng et al., 2014), a variety of birds (including chicken, duck and turkey), as well as zebrafish and platypus (Schauer et al., 2009), sperm whales (Peri et al., 2018), some species of dogs, bats (Yasue et al., 1978;Peri et al., 2018) and new world monkeys (Springer et al., 2014). Last, but not least, it is absent in humans (Varki, 2010;Dhar et al., 2019), due to a mutation in the CMAH gene encoding the CMAH enzyme, responsible for converting NeuAc to NeuGc. In humans, the gene lacks 95 base pairs in exon 3, giving rise to a frameshift and early termination, resulting in a CMAH pseudogene (CMAHP) (Irie et al., 1998;Chou et al., 2002;Suzuki, 2006). The human CMAHP only codes for 72 amino acids compared to the 590 amino acid protein found in chimpanzees (Muchmore et al., 1998;Diaz et al., 2009). It is hypothesised that CMAHP was quickly established within the human population around 2.5-3 million years ago, before the human brain was fully developed (Varki, 2001a;Chou et al., 2002).
The rapid fixation of CMAHP might have been due to a pandemic specifically targeting NeuGc, resulting in selection towards NeuGc-depleted individuals (Martin et al., 2005;Hayakawa et al., 2006), or due to sexual selection, where NeuGc-deficient females would create anti-NeuGc antibodies against NeuGc-containing sperm (Ghaderi et al., 2011). As a result of NeuGc deficiency, the human glycosylation pattern has adapted, and there are up to ten or more direct genetic changes in genes related to sialic acid production that can be traced directly to this event (Altheide et al., 2006;Varki and Varki, 2007;Varki, 2009).
GM3’s role in tumour development
1.4.2.1 NeuAc GM3 and cancer progression
As Vattulainen indicates in his quote, “the physical properties of carbohydrates are complex in its functions” (Vattulainen 2020). Although gangliosides have a cognate origin, their diversity is great and one ganglioside might have different roles depending on its concentration and localisation. A common way to investigate the effect of a certain ganglioside is exogenous addition of the ganglioside to cells for random incorporation (Bremer et al., 1986). However, depending on the cell type, the result might differ, such is indeed the case for NeuAc GM3. For endothelial cells, exogenous addition of NeuAc GM3 has been reported to feed the tumour nodules stimulating cell division (Slevin et al., 1999), while it was pro-apoptotic when added to neural progenitor cells and in human colonic carcinoma (Miller et al., 1998;Nakatsuji and Miller, 2001;Nojiri et al., 2002). Ganglioside research is therefore case-specific, making it difficult to postulate overall trends within tumour development. In general, the addition of NeuAc GM3 seems to induce apoptosis and inhibits growth of tumours related to the central nervous system (Visco et al., 2000;Noll et al., 2001). Exogenous addition of NeuAc GM3 might have a bright future for e.g. neuronal cancer treatment. Seyfried and Mukherjee (2010) report that brain tumours with high NeuAc GM3 levels grow slower compared to brain tumours with low NeuAc GM3 levels, and Case Western University holds a patent for the use of NeuAc GM3 for downregulating brain tumour growth (Miller et al., 1998).
1.4.2.2 The origin of NeuGc expression in human tissue
Since the CMAH gene is partly deleted, one hypothesis is that NeuGc is incorporated into tumour tissue from dietary sources (Figure 6A). Red meat, milk and other substances, such as caviar, contain high amounts of NeuGc glycoconjugates (Samraj et al., 2015).
Tangvoranuntakul et al. (2003) found traces of NeuGc in foetal tissue, in normal adult tissue (mainly in secretory epithelia and blood vessels), and a larger amount in breast carcinomas in humans. To investigate NeuGc incorporation they fed healthy humans with NeuGc. Most of the NeuGc was filtered out through the urine, but a small portion was shown to be incorporated into glycoproteins (Tangvoranuntakul et al., 2003). The small amount of NeuGc sialic acid incorporation in healthy tissue might be due to random uptake of sialic acids by the membrane sialic acid transporter sialin (SLC17A5), shuffling free sialic acids over the cell membrane, into the cytosol (Mancini et al., 1989;Verheijen et al., 1999).
Since small amounts of NeuGc have been identified in foetal tissues, it has given rise to the idea that NeuGc might be an oncofoetal gene (Hirabayashi et al., 1987;Tangvoranuntakul et al., 2003), but it is also possible that foetal NeuGc originates from the mother’s dietary uptake.
Oncogenes are genes that are associated with the causation of cancer. Since the CMAH gene in humans is partly deleted, the term ‘onco’ might not be an appropriate term for CMAHP (Varki, 2001b). Though, if the CMAHP gene is expressed, and the cells succeeds in salvaging the CMAH activity, combined with documented effect for NeuGc’s role in tumour progression, CMAHP could still be categorised as an oncogene or potentially an oncofoetal gene (Malykh et al., 2001;Bousquet et al., 2018). The human CMAHP gene terminates in the middle of a Rieske iron-sulphur cluster critical for the catalytic activity, rendering the gene product non-functional (Schlenzka et al., 1996). However, the CMAHP gene has a start codon downstream that could possibly ‘rescue’ protein expression, resulting in a CMAH-like enzyme that lacks a functional Rieske cluster domain (Malykh et al., 2001). The missing Rieske domain could potentially be replaced by other redox partners present in malignant cells. Bousquet at al. (2018) observed NeuGc-GM3-expression in HeLa cells (cervical cancer), when they were grown in NeuGc- deficient media (human serum). This was correlated with an upregulation of the mitochondrial succinate dehydrogenase complex subunit B (SDHB) that contains an iron-sulphur cluster.
They hypothesised that the SDHB protein could potentially replenish the CMAHP deleted gene- region, creating a functional CMAH-like enzyme complex (Figure 6B) (Bousquet et al., 2018).
Although the hypothesis is intriguing, the need for oxygen in the hydroxylation of NeuAc to NeuGc during hypoxic growth would limit the enzyme efficiency, and would require the involvement of a more complex mechanism.
Yin et al. (2006) observed a similar correlation between NeuGc expression and hypoxic conditions as Bosquet et al. (2018), but suggest that NeuGc expression has a dietary origin by identifying a heavily upregulated protein, the sialic acid transporter sialin (SLC17A5) (Figure 6A). Yin et al. (2006) succeeded to obtain steady surface expression of NeuGc in hypoxic LS174T human colon adenocarcinoma cells, but the expression was abolished when growing the cells in sialic-acid-free media under the same hypoxic conditions for 4 weeks. The addition of NeuGc to the sialic-acid-free media led to increased NeuGc incorporation and a
hypoxic conditions. They observed NeuGc expression, but without any indications of de novo synthesis of NeuGc. Only when transfecting the murine cmah gene into SKOV3 cells (SKOV3 KI-cmah) did the cells express NeuGc continually (Dorvignit et al., 2019). Blanco et al. (2011c) observed a correlation with NeuGc GM3 amount and hypoxic squamous cell carcinoma samples during immunohistochemistry (IHC) analysis. The further into the centre the tumour sample was taken, the higher NeuGc-GM3-expression. One explanation for why it might be beneficial for hypoxic cancer cells to upregulate sialin could be because of the Pasteur effect.
Louis Pasteur discovered that yeast consumed more carbohydrate during anaerobic growth compared to aerobic (Tyndall and Pasteur, 1878;Porter, 1961). This is due to increased adenosine triphosphate (ATP) consumption, combined with decreased efficiency in ATP production during anaerobic (Nelson, 2005). Therefore, the cell needs more glucose substrate to produce ATP, and might solve this by increasing uptake of carbohydrates. The observation that NeuGc expression is upregulated during hypoxia might also explain NeuGc levels in foetal tissue, since hypoxia is also common in embryogenesis (Lee et al., 2001). Another hypothesis for NeuGc origin has been proposed by Malykh et al. (2001) suggesting that glycolyl-CoA as a donor in a re-N-acylation reaction might be the origin of NeuGc in human tumours (Figure 6C). Whatever the origin, the presence of NeuGc in human tumours makes NeuGc GM3 an interesting target for immunotherapy and diagnostic purposes (Higashi et al., 1985;Marquina et al., 1996;Malykh et al., 2001;Blanco et al., 2011c;Blanco et al., 2015b;Pilco-Janeta et al., 2019).
Figure 6. Illustration of the three main hypotheses for the origin of NeuGc in human tumours. A. Dietary uptake over the cell membrane, by for example the sialic acid transporter sialin, potentiated by hypoxia (Yin et al., 2006). B. De novo synthesis of NeuGc by rescuing the CMAHP gene through combining the N- and C-terminal parts of the CMAH enzyme with a Rieske cluster domain creating a CMAH-like protein that can convert NeuAc to (Malykh et al., 2001;Bousquet et al., 2018). C. De novo synthesis of NeuGc by utilising glycolyl-CoA (Malykh et al., 2001). The Figure was created with biorender.com.
1.4.2.3 NeuGc GM3 stimulates tumour proliferation
It is well established that NeuAc GM3 has a down-regulatory effect on cell growth by inhibiting growth factors (GF) in certain tumours (Wang et al., 2001). NeuAc does not directly interact with the GF itself (Bremer and Hakomori, 1982;Bremer et al., 1984), but rather inhibits the growth factor receptor (GFR) by blocking dimerization, thereby preventing GF-induced tyrosine autophosphorylation and downstream signalling pathways (Figure 7A) (Bremer et al., 1986;Chung et al., 2009;Coskun et al., 2011). Miljan et al. (2002) and Coskun et al. (2011) both demonstrated GFRs’ specific affinity for NeuAc GM3. Metastatic NeuAc GM3 production is downregulated in melanocytes, causing increased cell growth (Ravindranath et al., 1991). By
dimerization by carbohydrate-carbohydrate interactions due to binding of N-linked glycans on EGFR (Yoon et al., 2006;Kawashima et al., 2009). It is likely that one of the glycan anchors on EGFR is K649, since mutating this residue (K642G) disrupts all regulatory effects (Coskun et al., 2011).
Figure 7. EGFR regulation by GM3. A. Normally NeuAc GM3 downregulates EGFR by blocking dimerization (Bremer et al., 1986;Chung et al., 2009;Coskun et al., 2011). B. NeuGc GM3 can potentially occupy the same position as NeuAc GM3 but does not necessarily have the same downregulatory function, likely due to the hydroxyl group at the terminal sialic acid. The active EGFR complex is autophosphorylated at its cytosolic domain leading to further activation cascades. The Figure was created with biorender.com.
There are several techniques the tumour might use to downregulate NeuAc GM3 to increase cell proliferation. One such mechanism is to express the plasma-associated neuraminidase 3 (Neu3), an enzyme that removes sialic acids on the cell surface. Neu3 is commonly upregulated during tumour progression (Kakugawa et al., 2002;Ueno et al., 2006). By silencing Neu3,
Tringali et al. (2008) observed a 46 % increase in NeuAc-GM3-expression in human K562 leukaemia cells and the cells’ apoptotic resistance was diminished. Another technique would be to introduce a NeuAc GM3 competitor, as in the case of the cancer-associated ganglioside NeuGc GM3 (Figure 7B). Hayashi et al. (2013) found NeuGc-containing gangliosides in 85 of 92 non-small cell lung carcinoma (NSCLC) tumour samples, and reported that NeuGc GM3 stimulates cell division by interacting with EGFRs extracellular domain, outcompeting NeuAc GM3. As the only difference is a hydrogen on the NeuAc sialic acid, whereas NeuGc harbours a hydroxyl group, it is likely that this NeuAc-hydrogen is crucial for EGFR downregulation.
Furthermore, it is suggested that NeuGc GM3 might have a prognostic role by itself, due to its association with hypoxic conditions (Hayashi et al., 2013;Giussani et al., 2014;Pilco-Janeta et al., 2019) (see Section 1.4.2.4 for more details). Recent research has contributed to clarify how NeuGc GM3 and EGFR expression is correlated. Palomo et al. (2016;2018) found that EGFR and NeuGc GM3 were co-localised in 63 of 92 tumour samples independent of histological subtypes, and suggest that a combined treatment of NeuGc GM3 and EGFR might be an efficient strategy to increase the survival rate. Blanco et al. (2017) identified the co-expression of NeuGc and EGFR in 60/101 and 18/26 tumour samples in untreated primary tumours and metastases, respectively, in 22/31 samples of high-grade astrocytoma and in 4/6 samples of tumour recurrences.
1.4.2.4 NeuGc GM3 as a prognostic factor
Hypoxic tumours are known for their increased ability to form metastases and are therefore associated with poor prognosis (Rankin and Giaccia, 2016). The increased expression of NeuGc GM3 has been found to be upregulated during hypoxic conditions (Yin et al., 2006), marking NeuGc GM3 as a possible prognostic factor for poor survival. Blanco et al. (2013b;2015a) observed a correlation between aggressive biological behaviour and NeuGc GM3 in human neuroectodermal tumour samples, independent of tumour cell lineage. Observations by Lahera et al. (2014) provide further evidence for NeuGc GM3 having a prognostic role in human colon adenocarcinoma. They also detected increased NeuGc-GM3-expression in the metastatic murine cell line 3LL-D122 (Lewis lung carcinoma) compared to the primary tumours, indicating that NeuGc-GM3-expression was related to increased tumour