Thesis
Reference
HP1 protein Chp2 selectively recruits nucleosome remodeler through non-canonical interaction
LEOPOLD, Karoline
Abstract
The establishment and maintenance of heterochromatic regions within the genome is an important factor for chromosome stability and con- trolled gene expression. It is dependent on many protein factors that are highly conserved from yeast to human. Among them are HP1 pro- teins that recognize heterochromatin-specific methylation marks and are involved in the recruitment of effector proteins. In my thesis, I will focus on the fission yeast HP1 protein Chp2 and its interaction with the Snf2/HDAC- containing remodeling complex SHREC, a homolog of NuRD complexes in higher organisms. Previously, Chp2 was found to be functionally and biochemically associated with the chromatin remodeler Mit1, a subunit of the SHREC complex. However, details of this interaction were unknown. I was able to solve the structure of the Chp2-Mit1 complex with high resolution and could show that an extensive interface between the two proteins provides high-affinity binding. The data reveals an unusual mode of HP1-client interaction and provides an example how specificity between different HP1 proteins canbe achieved. My work adds to the current [...]
LEOPOLD, Karoline. HP1 protein Chp2 selectively recruits nucleosome remodeler through non-canonical interaction. Thèse de doctorat : Univ. Genève, 2018, no. Sc. 5206
DOI : 10.13097/archive-ouverte/unige:104618 URN : urn:nbn:ch:unige-1046181
Available at:
http://archive-ouverte.unige.ch/unige:104618
Thomas Schalch
HP1 protein Chp2 selectively recruits nucleosome remodeler through
non-canonical interaction
T hèse
présentée á la Faculté des sciences de l’Université de Genève pour obtenir le grade de Docteur ès sciences, mention biologie
par
Karoline L eopold
de
Everswinkel (A llemagne )
Thèse No 5206
Genève
Université de Genève 2018
Thomas Schalch
HP1 protein Chp2 selectively recruits nucleosome remodeler through
non-canonical interaction
T hèse
présentée á la Faculté des sciences de l’Université de Genève pour obtenir le grade de Docteur ès sciences, mention biologie
par
Karoline L eopold
de
Everswinkel (A llemagne )
Thèse No 5206
Genève
Université de Genève 2018
1 a b s t r a c t 3
2 r é s u m é 4
3 a c k n o w l e d g m e n t s 5
4 i n t r o d u c t i o n 7
Chromatin structure and function . . . 7
DNA packaging in eukaryotes . . . 7
Chromatin organisation inS. pombe . . . 8
Heterochromatic states and the histone code . . . 10
Histone deacetylases . . . 10
Histone methyltransferases . . . 11
HP1proteins . . . 11
Chromatin remodeler . . . 13
NuRD complexes . . . 15
Heterochromatic gene silencing in S. pombe . . . 16
Transcriptional and postranscriptional gene silencing . 16 The RITS complex connects TGS and PTGS . . . 17
The Snf2/HDAC-containing repressor complex SHREC 17 S. pombeHP1homologs Chp2and Swi6 . . . 21
Aim of the thesis . . . 23
5 r e s u lt s 25 Chp2 interacts with the N-terminus of Mit1to repress transcription . . . 25
Crystal structure of the Chp2-Mit1complex reveals cryp- tic domain . . . 26
Chp2and Mit1engage in a high-affinity interaction . 31 The extended Chp2–Mit1interface is required for high affinity . . . 33
Swi6and Chp2differ substantially in their role for SHREC recruitment . . . 35
Disrupting the Chp2–Mit1interaction impairs silencing 36 Mit1ATPase activity is stimulated by nucleosomes but not free DNA . . . 37
The interface sequence is not well conserved . . . 38
The observed binding mode is not a universal mecha- nism for HP1binding . . . 39
High-affinity binding is mediated by the extended in-
teraction interface . . . 42
Swi6and Chp2exhibit different binding specificities . 43 Competitive binding assays can reveal preferred inter- action partners . . . 44
In vivo experiments demonstrate impact of disturbing Chp2-Mit1binding . . . 44
GBP-GFP system as a tool for studying interactions . . 45
Information about Mit1interaction partners allows dis- section of Mit1function . . . 46
Conclusion . . . 47
7 m at e r i a l s a n d m e t h o d s 49 Small-scale protein expression in Sf9insect cells . . . . 49
Pulldown experiment from insect cells . . . 49
Large-scale protein expression in insect cells . . . 49
Protein purification from insect cells for activity assay 50 Protein expression in E. coli . . . 50
Protein purification from E. coli . . . 51
Chd1purification . . . 51
Western blotting . . . 52
Limited proteolysis and mass-spectrometry analysis . 52 Crystallisation and structure processing . . . 52
Fluorescence polarization . . . 53
Isothermal titration calorimetry . . . 53
Sequence alignment . . . 53
ATPase assay . . . 53
Yeast two-hybrid screening . . . 54
Generation ofS. pombestrains . . . 54
Silencing assaysS. pombe . . . 55
Small-scale protein extraction fromS. pombe . . . 55
S. pombemedia . . . 56
Table1: Crystallographic table . . . 57
Table2: ITC data . . . 59
Table3:S. pombestrains used in this study . . . 60
Table4: Plasmids used in this study . . . 61
8 a b b r e v i at i o n s 63
9 b i b l i o g r a p h y 67
The establishment and maintenance of heterochromatic regions within the genome is an important factor for chromosome stability and con- trolled gene expression. It is dependent on many protein factors that are highly conserved from yeast to human. Among them are HP1pro- teins that recognize heterochromatin-specific methylation marks and are involved in the recruitment of effector proteins. While humans have three HP1 isoforms, in the model organism fission yeast two HP1 proteins are known and were reported to have different sets of interaction partners. In my thesis, I will focus on the fission yeast HP1 protein Chp2and its interaction with the Snf2/HDAC- containing re- modeling complex SHREC, a homolog of NuRD complexes in higher organisms. Previously, Chp2 was found to be functionally and bio- chemically associated with the chromatin remodeler Mit1, a subunit of the SHREC complex. However, details of this interaction were un- known. My goal was to study the molecular basis of HP1-dependent SHREC localisation and understand the different roles of the HP1pro- teins in this context. I was able to solve the structure of the Chp2-Mit1 complex with high resolution and could show that an extensive inter- face between the two proteins provides high-affinity binding. Using anin vivosystem, I found that this interaction is required for recruit- ment of Mit1 to heterochromatin and silencing of gene expression.
The data reveals an unusual mode of HP1-client interaction and pro- vides an example how specificity between different HP1proteins can be achieved. Taken together, my work adds to the current knowledge about the highly conserved class of HP1 proteins and deepens our understanding of the molecular basis of their function.
L’établissement et le maintien des régions d’hétérochromatine du gé- nome est un facteur important pour la stabilité des chromosomes et le contrôle de l’expression génétique. Ces processus dépendent de nom- breux facteurs protéiques qui sont conservés au cours de l’évolution, notamment entre la levure et l’homme. Parmi ces facteurs, les pro- téines HP1 reconnaissent des méthylations spécifiques aux regions d’hétérochromatine et sont impliquées dans le recrutement de pro- téines effectrices. Contrairement aux humains qui possèdent trois iso- formes d’HP1, on n’en connaît que deux chez l’organisme modèle de la levure à fission, chacune ayant un ensemble de partenaires d’inter- action différent. Dans ma thèse, je vais me concentrer sur la protéine HP1 de la levure à fission Chp2, et son interaction avec le complexe de remodelage associé à une histone désacétylase SHREC, qui est un homologue des complexes NuRD des organismes supérieurs. Des études précédentes ont établi que Chp2 interagissait fonctionnelle- ment et biochimiquement avec le facteur de remodelage de la chro- matine Mit1, une sous-unité du complexe SHREC. Cependant, les dé- tails de cette interaction restaient inconnus. Mon but était d’étudier les bases moléculaires de la localisation HP1-dépendante de SHREC, et de comprendre les différents rôles des protéines HP1dans ce con- texte. J’ai pu résoudre la structure du complexe Chp2-Mit1en haute résolution, et démontrer qu’une interface étendue entre les deux pro- téines permettait une liaison de forte affinité. En utilisant un système in vivo, j’ai découvert que cette interaction est nécessaire pour le re- crutement de Mit1 au niveau de l’hétérochromatine et l’extinction des gènes. Les données révèlent un mode d’interaction inhabituel en- tre une protèine HP1 et son partenaire, et illustrent comment peut être obtenue une spécificité entre différentes protéines HP1. En con- clusion, mon travail enrichit les connaissances actuelles sur la famille très conservée des protéines HP1, et approfondit notre compréhen- sion des bases moléculaires de leur fonction.
— The last bison,Switzerland
3 A C K N O W L E D G M E N T S
First of all, I would like to thank Thomas Schalch for accepting me in his lab, for never running out of project ideas and for giving me lots of useful advice. I would also like to thank the members of my TAC committee, David Shore and Michael Hothorn, for good ideas and constructive criticism during our meetings.
Another big thank you goes out to past and present members of the Schalch lab, especially Yvan Pfister for his excellent technical support, Babatunde Ekundayo for being the most cheerful person around and helping me a lot with both lab and mood issues and Christiane Brug- ger, for providing tea, chocolate, alcohol and knowledge. From day one, she showed me where to find things and how to run experi- ments and without her patience and helpfulness I would not have gotten the structure presented in this thesis.
Outside of the institute, I own a significant share of making it through the PhD to my climbing and skiing buddies in Geneva, most of all Hannes. Thank you for never dropping me, for doing all the scary slabs and for keeping me sane during these last few months.
I am very grateful for my friends outside of Geneva, especially Paul and Steffi. You would listen to my rants (both the big and the very little ones) and share my joy and frustration. I hope to always have you in my life wherever I go.
And last but not least, I want to thank my parents and my brother.
Your unconditional love was a strong support during the years of my PhD and I know that you have my back whatever I do.
c h r o m at i n s t r u c t u r e a n d f u n c t i o n
DNA packaging in eukaryotes
In the nucleus of eukaryotic cells, DNA is organized together with specialized proteins to achieve the necessary compaction. This macro- molecular DNA-protein complex is called chromatin (from the greek word for color, χρώμαchroma), a term first used in the 19th century by Walther Flemming for the substance inside the nucleus he could visualize with a staining procedure and observe thanks to advances in microscope technology [Flemming,1880]. Flemming was able to ac- curately describe the processes regarding nuclear reorganization pre- ceding cell division and established important groundwork for chro- matin related studies (Figure1). Soon after and based on Flemming’s work, the term chromosome was coined for the individual chromo- some particles that would undergo duplication and separation dur- ing the cell cycle [Waldeyer, 1888]. In the duplicated state, chromo- somes consist of two sister chromatids joined at the centromere, a specialized region of chromatin that acts as a docking platform for different proteins. In the early 20th century, the chromosome theory of inheritance was established based on the independent works of Walter Sutton and Theodor Boveri [Sutton,1902, Boveri,1904, Sutton, 1903] and correctly identified chromosomes as the factors responsible for inheritance.
While it was known since the studies of Richard Altmann and Al- brecht Kossel that chromatin consisted of DNA and protein [Altmann, 1894, Kossel,1884], nobody knew which substance was the carrier of genetic information. It was only in 1944that with the help of an ele- gant experiment, Avery et al. could show that desoxyribonucleic acid was the biologically active agent that determined the specific charac- teristics of pneumococcal cells [Avery et al.,1944]. Finally, their model of the DNA double helix enabled James Watson and Francis Crick to propose a copying mechanism of the molecule, a prerequisite for in- heritance [Watson and Crick,1953].
Today we know that chromatin consists of repeating units, the nucleo- somes. For each nucleosome,145-147bp of DNA are wrapped around the histone octamer, made from four different histone proteins in two copies each: H2A, H2B, H3 and H4. These repeating units are con-
Figure1: Chromosomes are reorganised during the cell cycle. Drawings of different stages of mitosis by Walther Flemming, published in his book “Zellsubstanz, Kern und Zelltheilung” (1882).
nected by stretches of nucleosome-free linker DNA and assembled into higher order structures stabilized by the linker histone H1[Korn- berg,1974, Kornberg,1977, Oudet et al.,1975, Olins and Olins,1974].
Non-histone proteins like chromatin remodelers and histone-modi- fying enzymes establish organisation of chromatin into distinct do- mains. This domain organization was first seen by Emil Heitz, who coined the terms heterochromatin for regions that remained condensed during interphase and euchromatin for regions that became less com- pact [Heitz, 1928]. In general, euchromatin is associated with highly transcribed regions while heterochromatin is in a less accessible and less transcribed state [Huisinga et al.,2006].
Chromatin organisation in S. pombe
The fission yeastSchizosaccharomyces pombe (S. pombe) is an excellent model organism for chromatin related studies due to its chromatin structure and conserved pathways for heterochromatin establishment and maintenance. The fission yeast genome contains 13.8 Mb orga- nized into three chromosomes (Figure2) [Kohli et al., 1977, Robinow, 1977].
While inSaccharomyces cerevisiae(S. cerevisiae) a short ~125bp consen- sus sequence acts as a point centromere, the centromeres of fission yeast are significantly larger and contain repetitive DNA sequences similar to satellite repeats in human centromeres. Both human and fis- sion yeast centromeres are therefore considered regional centromeres
Figure2: Chromosome architecture inS. pombe.Heterochromatic regions are high- lighted, while euchromatic regions are colored light grey (adapted from [Mizuguchi et al.,2015]).
Figure3: Comparison of centromere architecture in budding yeast, fission yeast and humans(adapted from [Verdaasdonk and Bloom,2011])
(Figure 3). A central asymmetric core region (cnt) is embedded into innermost inverted repeats (imr) and further surrounded by 20-100 kb of outer repetitive domains (otr) that consist of highly repetitive sequences dg and dh in chromosome specific orientations and repeat numbers [Nakaseko et al.,1987,Fishel et al.,1988,Hahnenberger et al., 1991,Clarke et al.,1986]. The central region (cnt+imr) is characterized by binding of histone variant CENP-A (Cnp1 in S. pombe) and fur- ther proteins responsible for kinetochore assembly [Takahashi et al., 2000]. In fission yeast, pericentromeres, telomeres, the mating type locus found on chromosome II and the ribosomal repeats (rDNA) on chromosome III are kept in a silent heterochromatic state (Figure2).
Heterochromatic states and the histone code
While the original definition of heterochromatin was based on its condensation throughout the cell cycle, the term was later further di- vided into facultative and constitutive heterochromatin to account for differences in underlying DNA sequence and silencing status [Brown, 1966]. Repetitive, usually gene-less DNA sequences such as centro- meres or telomeres that are kept in a transcriptionally silent state throughout the cell cycle are known as constitutive heterochromatin.
Facultative heterochromatin on the other hand contains genes that are transcribed or silenced according to the transcriptional program of a cell. Heterochromatin is defined by specific posttranslational modi- fications of the amino termini of histones. These modifications not only alter histone-histone or histone-DNA contacts, but also serve as binding platforms for other proteins. Different combinations of mod- ifications lead to different readouts of the same genetic information, like gene expression or gene silencing. This so-called “histone code“
and therefore the chromatin structure it defines can be stably inher- ited over many cell cycles and permits the establishment of different epigenetic states [Jenuwein and Allis,2001,Strahl and Allis,2000]. The most common histone modifications include acetylation and methy- lation of lysine residues in histone H3and H4.
Increased acetylation is a hallmark of transcribed regions and carried out by histone acetyltransferases (HAT) [Allfrey, 1966, Brownell et al., 1996]. HATs mainly target ε-amino groups of specific lysine residues and loss of regulation has been connected to different human can- cers [Marmorstein and Roth, 2001]. Heterochromatin on the other hand usually exhibits hypo-acetylated histone tails. Histone deacety- lation is catalyzed by histone deacetylases (HDAC) that are necessary for the formation and maintenance of heterochromatin [Yamada et al., 2005].
Methylation is carried out by histone-methyltransferases (HMT) that are either arginine- or lysine-specific. While trimethylation of lysine 4 on histone3(H3K4me3) serves as an activating mark for transcrip- tion, trimethylation of lysine residue 9 on histone H3 (H3K9me3) is known to transcriptionally silence regions [Strahl et al.,1999, Lachner et al.,2001].
Histone deacetylases
Histone deacetylases remove the activating acetyl marks placed on histone tails by HATs, an enzymatic activity that was discovered long before the enzyme responsible for it was identified [Inoue and Fu-
jimoto, 1969, Taunton et al., 1996]. Based on sequence similarities, HDACs can be classified into four different classes I-IV [Gregoretti et al., 2004]. Class I HDACs are defined by their sequence similarity to the yeast Rpd3 protein and share a similar domain organisation, while class II HDACs are similar to the yeast protein Hda1and have further functional domains in addition to the HDAC domain [Seto and Yoshida, 2014]. Class IV has only one member that shares se- quence homology with both class I and class II HDACs [Gao et al., 2002]. Class III HDACs are called sirtuins and are non-classical hi- stone deacetylases, as they use NAD+ to catalyze the deacetylation reaction, while class I, II and IV are all zinc dependent [Imai et al., 2000, Finnin et al., 1999]. Histone deacetylases were found to be part of diverse gene regulation complexes with discrete biological func- tions [Wade et al., 1998, Hassig et al.,1997, You et al.,2001].
Histone methyltransferases
Lysine-specific histone methyltransferases can be divided in two fam- ilies: SET-domain (Su(var)3-9, Enhancer of Zeste, Trithorax) and non- SET-domain containing HMT. SET-domain containing HMT catalyze the methylation reaction using SAM as a methyl-group donor [Trievel et al., 2002]. These proteins are responsible for the heterochromatic H3K9 methyl mark (Su(var)3-9 in Drosophila, Suv39h1 in humans, Clr4 in fission yeast) and the conservation of the SET-domain be- tween organisms hints at the importance of its function [Rea et al., 2000, Ivanova et al.,1998, Aagaard et al.,1999]. Methylation of H3K9 has been established as a hallmark for silencing and serves as a bind- ing site for proteins containing chromodomains [Nakayama et al., 2001, Bannister et al.,2001].
HP1proteins
The eponymous founding member of the HP1(heterochromatin pro- tein1) family of proteins was the first protein in Drosophila found to interact specifically with heterochromatic regions and to be a strong suppressor of position effect variegation, that is cell-specific silenc- ing of euchromatic regions when placed adjacent to heterochromatin [James and Elgin, 1986, James et al., 1989, Eissenberg et al., 1990].
HP1 proteins are highly conserved between organisms and share their overall structure of a chromodomain (chromatin organization modifier, CD) that is connected by a linker region to a C-terminal chromo shadow domain (CSD), making them part of the chromod-
omain superfamily [Paro and Hogness, 1991, Assland and Stewart, 1995]. In humans, three HP1 protein family members are known, namely HP1α, HP1βand HP1γ(or CBX5, CBX1and CBX3), with iso- form specific localization and interaction partners [Lomberk et al., 2006, Nielsen et al.,2002a, Vassallo and Tanese,2002, Hayakawa et al., 2003]. Especially HP1γhas been linked to transcriptional elongation rather than to silencing [Vakoc et al., 2005]. All three isoforms bind to H3K9me3 and have repeatedly been reported to interact with the transcriptional repressor TIF1α, subunits of the chromatin assembly factor CAF1, the ATP-dependent chromatin remodelers SMCA4 and CHD3/CHD4and the zinc finger proteins POGZ and ADNP among others [Vermeulen et al.,2010,Rosnoblet et al.,2011,Hoffmeister et al., 2017]. However, some proteins like the ADNP-related ADNP2 zinc finger protein were reported to bind only HP1β and HP1γ but not HP1α, while zinc-finger protein Znf828 exclusively interacted with HP1α[Vermeulen et al.,2010, Rosnoblet et al.,2011].
Chromodomains are ~50 amino acid protein folds consisting of a three-stranded antiparallel beta-sheet packed against a C-terminal alpha helix and were found to be a shared property of many nu- clear proteins involved with chromatin [Ball et al.,1997, Koonin et al., 1995]. Chromo shadow domains were identified as variants of the chromodomain and are found exclusively in proteins also containing N-terminal chromodomains [Assland and Stewart,1995].
In HP1 proteins, the chromodomain specifically binds methylated H3K9 with affinities in the low micromolar range [Bannister et al., 2001, Lachner et al., 2001]. This precise binding is commissioned by an induced fit, where three conserved aromatic residues form an aro- matic cage upon binding of the methylated H3K9 peptide [Nielsen et al., 2002b, Jacobs and Khorasanizadeh, 2002]. In addition, affinity to H3K9-methylated nucleosomes seems to be modulated by phos- phorylation of N-terminal residues [Hiragami-Hamada et al., 2011, Nishibuchi et al.,2014].
The chromo shadow domain on the other hand is not involved with the recognition of methylated H3K9, but is responsible for dimer- ization and interaction with other proteins. These interactions were reported to be mediated by a consensus peptide pentamer found within many interaction partners, with proline at position 1, valine at position3and hydrophobic residues like leucine found at position 5[Smothers and Henikoff,2000, Murzina et al.,1999]. Structural stud- ies revealed that this motif is bound between the C-terminal tails of the chromo shadow domains across the symmetry axis of HP1dimers, forming a parallel beta sheet with one monomer tail and an antiparal- lel beta sheet with the other. In addition, key residues are accommo-
dated in hydrophobic pockets formed by the dimer interface [Thiru et al., 2004, Cowieson et al., 2000]. This mode of peptide binding is rather unusual since in most cases like the very common PDZ inter- action domains, peptides bind in a groove formed by residues of only one molecule instead of two [Nourry et al.,2003].
The hinge region in HP1 proteins has been reported to be involved with RNA-binding via positively-charged residues [Muchardt et al., 2002, Keller et al.,2012, Keller et al.,2013]. However, while inS. pombe Swi6RNA-binding seems to compete with H3K9me binding, in mam- malian HP1α both RNA- and methyl mark binding were reported to be necessary for localization to heterochromatin [Muchardt et al., 2002,Keller et al.,2012,Keller et al.,2013]. Furthermore, studies found the hinge region sufficient for binding to the linker histone H1[Nielsen et al.,2001, Hale et al.,2006]. The H3K9me recognition and the ability to dimerize and interact with other proteins make HP1a crucial adap- tor between heterochromatic sites and effector proteins and lead to a model for heterochromatin spreading, where HP1proteins would rec- ognize methylation marks and recruit histone methyltransferases to these sites, resulting in methylation of further H3K9residues [Bannis- ter et al., 2001, Lachner et al.,2001, Nakayama et al.,2001].
Chromatin remodeler
Among the most important effector proteins of chromatin organi- sation that ultimately define the structure of DNA packaging are ATP-dependent remodeling complexes, so called chromatin remod- elers. These ATPases use energy from ATP hydrolysis to locally alter histone-DNA contacts and move or exchange histone proteins. They belong to the superfamily of SF2 helicases and the family of Snf2 related enzymes, named after the S. cerevisiae Snf2 helicase-like pro- tein [Laurent et al.,1992]. This relationship is based on a shared Snf2- type ATPase domain consisting of two tandem RecA folds (DExx and HELICc, separated by a linker) that contain seven helicase-related motifs in a conserved organization [Eisen et al., 1995, Flaus et al., 2006]. Apart from this common feature however, they differ in their subunit organization, action and specificity. The different composi- tion of remodelers allows for classification into four major subfam- ilies: SWI/SNF, ISWI, CHD and INO80 [Flaus et al., 2006, Clapier and Cairns, 2009]. Each of these is defined by sequence-homology of unique domains within the ATPase domain and by surrounding subunits, making them specialized for different tasks and contexts (Figure4).
Figure4: Chromatin remodeler families.Chromatin remodelers can be divided into different families according to their domain architecture (adapted from [Clapier and Cairns,2009])
s w i/s n f s u b f a m i ly The SWI/SNF subfamily is defined by a helicase-SANT associated (HSA) domain involved in protein inter- action, a post-HSA domain and a C-terminal bromodomain, known for binding to acetylated histones [Szerlong et al.,2008, Haynes et al., 1992]. Members of this family have been reported to reorganize chro- matin by nucleosome sliding or eviction and form large, highly con- served remodeling complexes [Mohrmann and Peter Verrijzer, 2005, Tang et al.,2010].
i s w i s u b f a m i ly The defining attributes of the ISWI subfamily are the SANT and SLIDE domains towards the C-terminus of ATPases.
These domains are crucial for histone tail recognition and binding and link it to the ATPase activity [Boyer et al., 2004]. An important feature of ISWI remodelers is their ability to regularly space nucleo- somes on DNA [Gelbart et al.,2001].
c h d s u b f a m i ly The CHD subfamily harbors two tandem chro- modomains N-terminal of the ATPase domain that have been shown to interact with methylated H3 tails in humans and Drosophila, but not in yeast [Woodage et al.,1997,Flanagan et al.,2005,Flanagan et al., 2007]. Here, the chromodomains are reported to regulate the ATPase activity [Stockdale et al., 2006, Hauk et al., 2010]. The structure of yeast Chd1 bound to the nucleosome revealed that in the free state, the chromodomains keep the ATPase in an open conformation. Upon binding of the nucleosome, they interact with nucleosomal DNA, thereby releasing and activating the ATPase domain [Farnung et al., 2017]. CHD remodelers can slide nucleosomes away from DNA ends to center them on a piece of DNA, affect the assembly of nucleosomes and are involved with the positioning of histone variants [Stockdale et al., 2006, Lusser et al., 2005, Konev et al., 2007]. Members of the CHD subfamily can act as transcription activators like Chd1 as part
of the SAGA complex in yeast, or as transcription repressors like Mi- 2as part of NuRD complexes in higher eukaryotes [Pray-Grant et al., 2005, Tong et al.,1998]. To account for differences between CHD sub- family members, they can be further classified into three subfami- lies according to additional domains: the Chd1-Chd2 subfamily, the Chd3-Chd4 subfamily and the Chd5-Chd9 subfamily [Marfella and Imbalzano, 2007]. The Chd1-Chd2 subfamily is characterized by an additional DNA-binding domain located in the C-terminus of the protein [Stokes and Perry, 1995]. The second subfamily, also called the Mi-2 subfamily, lacks this C-terminal DNA-binding domain, but harbors an N-terminal PHD (plant homeo domain) instead, a zinc- finger like motif that is involved with histone tail binding [Shi et al., 2006,Peña et al.,2006,Musselman et al.,2009]. The third subfamily has been defined by the presence of additional domains in the C-terminal region.
i n o 8 0 s u b f a m i ly Finally, the INO80 subfamily is characterized by a split ATPase domain and has important roles in transcriptional activation and DNA damage repair [Bakshi et al., 2004, Shen et al., 2000].
NuRD complexes
NuRD (nucleosome remodeling and histone deacetylation) complexes were identified in various organisms from plants to mammals and were unique in that they connected histone deacetylation and ATP- dependent chromatin remodeling in one complex [Tong et al., 1998, Wade et al., 1998, Xue et al., 1998, Zhang et al., 1998]. Mi-2α and Mi-2β remodeler (also called CHD3 and CHD4) confer the ATPase activity to NuRD and at least in mammals are mutually exclusive, while HDAC1and HDAC2are responsible for deacetylation of lysine residues [Hoffmeister et al., 2017, Taunton et al., 1996]. Further non- enzymatic subunits include MTA1-3, MBD2/MBD3, RBBP4/7 and GATAD2A and GATAD2B.
MTA1, MTA2and MTA3 are metastasis-associated proteins and con- tain several functional domains, among them the BAH-domain (bromo adjacent homology) involved in protein interactions, the ELM do- main involved in complex dimerization and the SANT domain that together with the ELM domain recruits the histone deacetylase [Mil- lard et al.,2013]. Despite their sequence homology, the MTA proteins form distinct NuRD complexes [Fujita et al.,2003]. MBD2and MBD3 are methyl-CpG binding domain proteins, however only MBD2 is able to bind methylated DNA while MBD3 uses its MBD domain
to interact with MTA2 and HDAC1 [Hendrich and Bird, 1998, Saito and Ishikawa, 2002]. MBD2 and MBD3 are assembled into NuRD complexes with different specificities and were found to be mutu- ally exclusive [Le Guezennec et al.,2006]. RBBP4and7(also RBAp46 and RBAp48) are highly homologous WD40 retinoblastoma-binding (RB) proteins and are involved with histone binding [Murzina et al., 2008]. GATAD2A and GATAD2B (also known as p66αand p66β) have been shown to bind to several NuRD subunits and are responsible for structural integrity [Brackertz,2006]. Overall, MTA and MBD pro- teins seem to define target specificity of NuRD, while RB and p66 proteins are general structure factors of protein complexes [Lai and Wade,2011].
The structure of the core NuRD complex revealed that MTA1 is a dimer and that one MTA1 molecule recruits two RBBP4/7, leading to four RBBP4/7proteins in the complex [Millard et al.,2016]. While the core components are highly conserved, the exact stoichiometry and accessory factors are dynamic and allow for a variety of func- tions [Torchy et al.,2015].
Interestingly, NuRD complexes were found to locate to sites of DNA damage and directly interact there with HP1proteins via their remod- eler subunits [Hoffmeister et al., 2017, Rosnoblet et al.,2011]. NuRD components, mainly MTA proteins, are reported to be involved in various cancers and their upregulation is linked to poor prognosis, making NuRD an interesting drug target that necessitates the devel- opment of complex specific drugs since many subunits are shared with other regulatory complexes [Lai and Wade,2011, Millard et al., 2017].
h e t e r o c h r o m at i c g e n e s i l e n c i n g i n s. p o m b e Transcriptional and postranscriptional gene silencing
Heterochromatic silencing of genes can occur both on the transcrip- tional and on the post-transcriptional level. Transcriptional gene si- lencing (TGS) refers to the repression of transcription through epi- genetic mechanisms such as DNA methylation, specific nucleosome placement or the aforementioned repressive histone marks. All these mechanisms converge on the regulation of RNA polymerase II (RNAP II) that decides whether a gene can be transcribed [Shimada et al., 2009,Kouzarides,2002]. Post-transcriptional gene silencing (PTGS) on the other hand depends on RNA interference (RNAi) and other RNA modifying processes and acts via effector complexes that incorpo-
rate small single-stranded RNA molecules [Gregory et al.,2005,Mont- gomery et al.,1998].
The RITS complex connects TGS and PTGS
InS. pombe, the RNA-induced initiation of transcriptional gene silenc- ing complex (RITS) targets the mating-type region and centromeres and leads to chromatin modifications that initiate and spread hete- rochromatin [Verdel et al.,2004]. The RNA-directed RNA polymerase complex (RDRC) contains the RNA-directed RNA polymerase sub- unit Rdp1 and uses transcripts stemming from inverted repeat re- gions like the centromere to synthesize double stranded RNA (dsRNA) [Motamedi et al., 2004, Sugiyama et al., 2005]. These dsRNA can be used by the ribonuclease III-like enzyme Dicer for the production of small interfering RNA (siRNA) that are subsequently loaded as single stranded siRNA onto the Argonaute protein Ago1 in the context of the Argonaute siRNA chaperone (ARC) [Buker et al., 2007]. siRNA- loaded Ago1 can then recruit the RITS complex to complementary regions. Its interaction partners in the RITS complex are Tas3 and Chp1, a chromodomain containing protein that can bind to methy- lated H3K9 and thereby stabilize localisation of the complex to het- erochromatic regions (Figure 5) [Verdel et al., 2004, Partridge et al., 2002,Schalch et al.,2009a]. Furthermore, siRNA-loaded Ago1contain- ing RITS can directly interact with RDRC, leading to the formation of a self-enforcing feedback loop for the production of siRNA [Mo- tamedi et al.,2004]. This interaction also requires the histone methyl- transferase Clr4that is responsible for H3K9methylation which leads to formation and spreading of heterochromatin. Clr4is the only his- tone methyltransferase in S. pombeand is part of the Clr4 containing complex (or CLr4-Rik1-Cul4, CLRC) that in turn interacts with RITS, providing a direct link between siRNA mediated silencing and H3K9 methylation based transcriptional silencing (Figure 5) [Bayne et al., 2010, Horn et al.,2005, Hong et al.,2005, Nakayama et al.,2001]. Clr4- mediated methylation of H3K9 provides a binding platform for the fission yeast HP1 proteins Swi6and Chp2that remain bound to the heterochromatic mating type locus throughout the cell cycle and act as a mark for the silenced state [Nakayama et al.,2000].
The Snf2/HDAC-containing repressor complex SHREC
The Snf2/HDAC-containing repressor complex (SHREC) was found to act onS. pombeheterochromatin in parallel to the siRNA- involving
Figure5: The silencing machinery inS. pombe.Transcriptional and posttranscrip- tional gene silencing mechanisms converge on regulating RNA polymerase II activ- ity.
RITS machinery by using transcriptional gene silencing mechanisms (Figure 5) [Yamada et al., 2005, Sugiyama et al., 2007]. SHREC is a multienzyme effector complex containing four core proteins, two of them having enzymatic functions, namely histone deacetylation and chromatin remodeling. This makes it a homolog of NuRD complexes found in higher organisms.
c l r 1 The clr1(cryptic loci regulator) gene was first identified as a locus where mutations led to de-repression of the mating type locus and aberrant meiosis of cells [Thon and Klar, 1992]. The clr1 gene encodes for the1238 amino acid C2H2-type zinc finger protein Clr1, whose deletion or mutation leads to a global upregulation of gene expression [Hansen et al.,2005]. In the SHREC complex, Clr1 serves as a long flexible scaffolding protein onto which the other subunits are assembled [Motamedi et al., 2008a]. Expression levels of Clr1are relatively low with only ~300molecules per cell [Carpy et al., 2014, Marguerat et al.,2012].
c l r 2 Like Clr1, Clr2 was initially identified as a protein whose mutation led to de-repression of the silent mating type locus [Thon et al.,1994, Ekwall and Ruusala, 1994]. Clr2is a 62kDa protein with no sequence homology to known domains and localizes not only to
regions of heterochromatin associated with SHREC, but can also be found at the core region of centromeres [Sugiyama et al.,2007]. Upon deletion of Clr2, histones were found to be hyperacetylated, while overexpression had no influence on silencing [Bjerling et al., 2004].
Structural studies revealed binding of Clr2to the C-terminal region of Clr1and showed that the protein can be divided into three domains, one of them with high structural similarity to MBD domains (methyl- CpG binding domain). This domain was found to bind nucleic acids and its functionality is required for SHREC-mediated silencing [Job et al.,2016].
c l r 3 Clr3 was first identified in the same screens for silencing defects as Clr1 [Thon et al., 1994, Ekwall and Ruusala, 1994]. It is a 687 aa class II histone deacetylase and responsible for removing acetyl marks from lysine14on histone H3, a potential prerequisite for H3K9methylation by Clr4[Nakayama et al.,2001,Bjerling et al.,2002].
While it mostly colocalizes with other SHREC components, it can also be found alone at certain sites, possibly indicating functions beyond SHREC-mediated silencing [Sugiyama et al.,2007]. The localization of Clr3to the mating type locus is dependent on Chp2and Swi6among other factors, and leads to nucleation and subsequent spreading of heterochromatin across the mat2/3region. Notably, while direct inter- action between Clr3and Swi6has been reported, only a small fraction of Clr3was bound by Swi6, indicating a rather transient and dynamic interaction [Sugiyama et al., 2007, Yamada et al., 2005]. Localization to the telomeric loci on the other hand is mediated by Ccq1and Taz1 through direct interaction between the former and Clr3 [Sugiyama et al.,2007]. In addition to the catalytic HDAC domain, Clr3has a C- terminal Arb2domain [Job et al.,2016, Buker et al.,2007]. The crystal structure of the Clr1-Clr3 complex revealed that Clr3acts as a dimer with an extensive dimerization interface encompassing both the Arb2 domain and the HDAC domain. Both domains were also found to be required for silencing and for the interaction with Clr1, which is hap- pening via the zinc fingers of Clr1towards the C-terminus [Job et al., 2016]. Clr3is a lot more abundant than Clr1, with about3000proteins per cell during vegetative growth phase [Carpy et al.,2014,Marguerat et al.,2012].
m i t 1 Mit1(Mi2-like protein interacting with Clr three1) was iden- tified in pulldown experiments with Clr3 and named for its high similarity with Mi-2/CHD3 proteins [Sugiyama et al., 2007]. Mit1 is an 1418 aa CHD type chromatin remodeller, containing the Snf2- type helicase domain and a PHD finger domain commonly found
within Mi-2proteins of the CHD subfamily [Marfella and Imbalzano, 2007]. While initial studies reported the absence of a chromodomain, a loosely defined chromodomain could be identified later that is nec- essary for Mit1 silencing function, as is the PHD domain [Creamer et al., 2014]. Like other Mi-2 type remodelers, Mit1 moves nucleo- somes to the center of DNA pieces in vitro[Creamer et al., 2014]. In the context of SHREC, Mit1 binds towards the N-terminus of Clr1 with its C-terminus, forming a tightly intertwined complex [Job et al., 2016]. In concordance with other SHREC components, Mit1locates to all major heterochromatic sites but is strongly reduced at those sites in strains lacking Swi6[Sugiyama et al.,2007].
c h p 2 In addition to the four core proteins, the380aa HP1protein Chp2 was found to associate with SHREC [Motamedi et al., 2008a, Creamer et al.,2014]. Like all HP1proteins, Chp2is organized into a chromodomain, a hinge region and a C-terminal chromo shadow do- main and has extensive sequence similarity with Swi6. Loss of Chp2 leads to de-repression of the mating-type region and affects the ex- pression levels of a similar set of genes as Mit1 [Thon and Verhein- Hansen,2000, Job et al.,2016].
s h r e c o r g a n i s at i o n a n d f u n c t i o n With the chromatin re- modeler Mit1bound towards the N-terminus of Clr1and the HDAC Clr3 and Clr2 bound to the C-terminus of Clr1, the overall architec- ture of SHREC can be divided into two distinct modules, each har- boring one of SHREC’s two enzymatic functions. These modules are connected by Clr1, acting as a long flexible linker. This structural or- ganisation is reflected in transcriptional regulation. It was found that for yeast lacking individual SHREC components, gene expression pro- files for clr1Δ, clr2Δand clr3Δsignificantly overlapped, while mit1Δ and chp2Δ behaved in a different way but similar to one another.
These observations suggest differential recruitment mechanisms for the two functional modules of SHREC (Figure6) [Job et al.,2016, Mo- tamedi et al., 2008a]. The finding that Clr2 harbors an MBD-like do- main and is capable of binding nucleic acids raised the possibility of recruitment of the HDAC module chromatin via Clr2, while the association of Mit1 with Chp2made recruitment via the histone tail binding abilities of the latter likely [Job et al., 2016, Creamer et al., 2014]. However, interaction between Clr3 and Swi6 and dependence of SHREC on Swi6for proper localization also suggests involvement of Swi6[Yamada et al.,2005, Sugiyama et al.,2007].
Apart from the enzymatic functions, SHREC shares further similari- ties with NuRD. SHREC subunit Clr1 shares some sequence homol-
Figure6: SHREC is organized into two distinct modules.The remodeler module encompasses Mit1and Chp2, while the HDAC module is made up from Clr2and Clr3. Clr1acts as a linker between both parts. SHREC subunits are represented by their structures where available.
ogy with the NuRD subunit GATAD2B. Furthermore, NuRD com- plexes contain MBD proteins that are involved in recruitment of the complex to DNA, while Clr2contains an MBD domain capable of nu- cleic acid binding [Lai and Wade, 2011]. However, NuRD assembly differs between organisms and cell types, while SHREC is a lot better defined.
S. pombe HP1homologs Chp2and Swi6
In fission yeast, two HP1 homologs are known, the aforementioned Swi6 and Chp2. They act downstream of H3K9 methylation and are both necessary for fully repressive heterochromatin, with the loss of Swi6 affecting silencing more, while a Δchp2Δswi6 double mutant proved to have the strongest effect [Fischer et al., 2009, Sadaie et al.,
2008]. Not only are the expression levels between the HP1 proteins very different, with more than19,000molecules of Swi6estimated to be present in one cell in contrast to only240copies of Chp2, but these levels were also shown to be crucial for proper function. Overexpres- sion of either protein led to silencing defects. While overexpression of Chp2 seems to dilute Swi6 from its binding sites, overexpression of Swi6 caused growth defects through affecting chromosome segre- gation [Sadaie et al., 2008].
Furthermore, experiments with chimeric proteins indicated that the chromo shadow domains are specific for each protein and cannot sub- stitute each other. Since the CSD have been reported to be involved with protein interactions, this pointed to different binding partners.
Indeed, pulldown experiments have identified specific sets of inter- acting proteins [Fischer et al.,2009, Motamedi et al.,2008a].
Swi6 was found to interact with many nuclear factors (Figure 5).
These involve the essential protein Sap1 with a role in mating type switching, components of the Clr6HDAC complex and the telomere- associated protein Ccq1 [Arcangioli et al., 1994, Nicolas et al., 2007, Sugiyama et al., 2007]. Also Atf1/Pcr1 could be identified [Jia et al., 2004]. Interestingly, Atf1/Pcr1was found to be crucial for recruiting of Clr3 to the mat locus, while Ccq1 is responsible for SHREC re- cruitment to the telomeres [Yamada et al.,2005,Sugiyama et al.,2007].
Furthermore, subunits of chromatin remodeling complexes and the CHD type remodeler Hrp1, that is involved with CENP-A loading, were identified as Swi6 binding partners [Walfridsson et al., 2005].
Also interaction with Mis4, a protein important for the assembly of the cohesin complex, was observed and further confirmed in co- immunoprecipitation experiments [Furuya et al.,1998]. This supported earlier studies implicating Swi6 with cohesin recruitment to hete- rochromatin [Nonaka et al.,2002]. Finally, the JmjC protein Epe1and the SHREC subunit Mit1 could be identified among the set of pro- teins interacting with Swi6[Zofall and Grewal,2006, Sugiyama et al., 2007, Fischer et al.,2009].
Chp2 on the other hand was found to exclusively purify with sub- units of the SHREC complex, namely Clr1 and Mit1, indicating spe- cific interaction between Chp2and SHREC. However, mutating Clr3 showed more silencing defects than mutating Chp2, indicating an ad- ditional Chp2-independent role of SHREC [Fischer et al., 2009]. In addition, other studies have reported interaction between both Chp2 and Swi6and Shugoshin (Sgo1), a highly conserved kinetochore pro- tein, which is recruited to centromeric heterochromatin [Yamagishi et al., 2008, Kitajima et al., 2004]. While Sgo1-Swi6 binding was re- ported to be in the low micromolar range, binding to Chp2 was
~6times weaker [Isaac et al.,2017].
Another distinguishing feature between Chp2 and Swi6seems to be their oligomerization behaviour. Both proteins exist as a dimer in so- lution but only Swi6 has been reported to tetramerize at higher con- centrations via CD-CD interactions [Canzio et al.,2011]. Furthermore, it was proposed that the CD of Swi6possesses a loop similar to part of the H3tail that occupies the H3K9me2/3binding site. It is released upon nucleosome binding to free the CD that can then engage in in- teractions with neighbouring nucleosomes. This auto-inhibitory loop is absent in Chp2 which might add to differences between the HP1 proteins [Canzio et al., 2013, Isaac et al., 2017]. In addition to these functions, only Swi6 has been reported to directly interact with the RNAi machinery and bind to heterochromatic RNA, while this could not be observed for Chp2 [Motamedi et al., 2008a]. Binding of RNA to HP1 proteins has been reported to happen via basic residues in hinge region [Maison and Almouzni,2004, Muchardt et al.,2002].
For Swi6, also the N-terminus and the CD seem to be involved with RNA binding in a way that makes it mutually exclusive with the recognition of methylated H3K9. According to this model, RNA bind- ing would lead to eviction of Swi6from heterochromatin [Keller et al., 2012]. In addition, the ability to bind RNA seems to restrict Swi6 to heterochromatic regions and the RNA-dependent eviction of Swi6 from heterochromatin inhibits heterochromatin spreading [Keller et al., 2013].
a i m o f t h e t h e s i s
The main goal of this thesis is to understand how the remodeler unit of the Snf2/HDAC complex SHREC is recruited to heterochromatin.
While involvement of the two HP1proteins Swi6and Chp2has been reported and Chp2 can be seen as a component of the SHREC com- plex, reports differ with regards to the importance of either protein.
Furthermore, no data exists showing clearly how any of those interac- tions are defined and what they depend on. As different sets of inter- actors have also been reported for HP1isoforms in higher organisms, we want to expand the current knowledge of how these specializa- tions are ensured. Using interaction studies, we want to resolve the minimal components required for interaction between SHREC and HP1 proteins and subsequently employ X-ray crystallization to un- derstand the molecular details of this interaction. This information should give us valuable insight into the basis of the different roles of Chp2 and Swi6that have been reported before. In addition, we want to study the importance of our in vitroresults in vivo, usingS. pombe
strains carrying mutations in the implicated proteins. We want to un- derstand the role the observed interactions have, both for SHREC function and for distinction between Swi6 and Chp2 and we will be employing silencing assays as a read-out.
Chp2interacts with the N-terminus of Mit1to repress transcription
Previous experiments have suggested that the HP1protein Chp2 in- teracts with the SHREC complex through the N-terminus of Clr1 or through the chromatin remodeler Mit1 [Motamedi et al., 2008b].
Since this interaction interface is a strong candidate for mediating Chp2-specific function in heterochromatin we decided to investigate its molecular nature. Sequence analysis of Clr1revealed two potential PxVxL motifs within the N-terminus of the protein (Figure7A). How- ever, co-expression of Chp2with the N-terminal half of Clr1(1-500) in insect cells did not yield interacting partners (Figure 7B). The same was true for the second HP1protein inS. pombe, Swi6(Figure7B).
Figure7: No interaction was seen between Clr1and either Swi6HP1or Chp2HP1.(A) Domain architecture of Clr1and the two HP1proteins Chp2and Swi6. (B) Pulldown experiments from insect cells and western blot analysis reveal no Chp2 or Swi6 bound to Clr1.
Therefore, we focussed on Mit1and its individual domains (Figure8A).
Co-expression of Mit1 and Chp2 in insect cells produced a complex that could be recovered in pulldown experiments (Figure 8B). Inter- estingly, the N-terminus of Mit1, which is not predicted to harbor any domains, proved to be both required and sufficient for mediat- ing Chp2interaction in the heterologous insect cell system. Mutation of two predicted PxVxL motifs in this region (N2F, Figure 8B) did not abolish the interaction. In order to map the domains responsi- ble for this interaction we subjected Mit1 and Chp2 and their do- mains to a yeast-two-hybrid assay (Figure 8C). The yeast-two-hybrid results show that the Mit1N-terminus and the Chp2chromo shadow domain are each necessary and sufficient for interaction. Under the highly stringent selection of the quadruple drop-out medium Chp2
Figure 8: The Mit1N-terminus binds Chp2HP1. (A) Domain organisation of the Mit1 chromatin remodeler and the HP1 homolog Chp2. (B) StrepII pull-down of Mit1fragments with Clr1and Chp2shows that the ternary complex forms in a het- erologous expression system. The interaction with Chp2depends on the N-terminal domain of Mit1. (C) Yeast-2-hybrid interaction study shows that the chromo shadow domain is required for Mit1-Chp2interaction. (D) Comparative growth assays for various Mit1constructs show that the Mit1N-terminus is required for ura4reporter silencing at pericentromeric regions.
full length protein is required to sustain growth, which raises the possibility that Mit1 interacts with both CSD and CD of Chp2. The observed interaction suggests that the N-terminus of Mit1 mediates the recruitment of SHREC by Chp2 for heterochromatin silencing.
In order to test the functional relevance of the Mit1 N-terminus we deleted residues 1-200 of Mit1 at the endogenous genomic locus in S. pombeand subjected this mutant to comparative growth assays in a otr1R::ura4 background that permits to monitor heterochromatin integrity in the pericentromeric region of chromosome I (Figure 8D).
Loss of the Mit1 N-terminus clearly leads to loss of silencing of the ura4marker gene, thereby demonstrating the importance of this Mit1 region in heterochromatin formation.
Crystal structure of the Chp2-Mit1complex reveals cryptic domain
Based on the functional and biochemical evidence that Chp2-depen- dent recruitment of SHREC is mediated by the N-terminus of Mit1, we decided to determine the atomic structure of the interaction inter- face. We therefore coexpressed the N-terminus of Mit1(Mit1(1-200)) with Chp2in bacteria. However, size-exclusion chromatography (SEC) revealed that proteins were aggregating and no well-folded complex could be obtained. Changing the Mit1 construct to also include the
PHD domain (Mit1(1-299)) did improve protein behaviour slightly, but only after removal of the Chp2chromodomain and linker region from the construct (Chp2CSD) could we purify good amounts of pro- tein complex eluting in a single peak (Figure 9A). In order to deter- mine the minimal structural unit that harbors the interaction we had to resort to limited proteolysis on the bacterially expressed construct consisting of the Chp2CSD and Mit1residues1-299(Figure 9B). The limited proteolysis with thermolysin revealed a stable Mit1fragment that corresponds to residues1-81 as determined by mass spectrome- try (Figure9C). Based on this information, we coexpressed Mit1(1-81) with Chp2 CSD in bacteria and were able to purify a stable well- behaved complex (Figure9D).
Figure9: Limited proteolysis reveals minimal complex.(A) The Mit1(1-299)-Chp2 CSD complex elutes as a single peak from SEC and peak fractions were used for limited proteolysis. (B) Limited proteolysis experiment with increasing amounts of protease yielded two stable fragments that withstood even highest protease concen- trations. (C) MALDI analysis (upper panel) after sample reduction reveals two main ions (marked in purple), one of them corresponding well to Chp2CSD. MS/MS frag- mentation (lower panel) of the second fragment matches composition of Mit1(1-81) (Mit1fragments marked in purple). (D) The minimal interaction complex as identi- fied in the limited proteolysis experiment followed by mass spectrometry analysis elutes as a single peak from SEC and was set up for crystallization.
This minimal Chp2-Mit1complex crystallized readily as small irregu- larly shaped crystals (Figure 10A). Initial optimization that included lowering both pH and salt concentration yielded bigger and very regular crystals that diffracted well (Figure 10B). Small cube-shaped crystals could be obtained by adding 1,6-hexanediol to the original crystallization condition (Figure 10C). Manually setting up this con- ditions as 1 μl hanging drops resulted in single large cube-shaped crystals that diffracted to high resolution (Figure10D).
Figure10: The minimal Chp2-Mit1complex crystallizes readily.(A) Small irregular crystals in the initial hit. (B) Example of optimization condition leading to different crystal shapes. (C) Crystallization condition used to grow crystals for data collection.
(D) Crystal fills out the entire0.4mm loop, top and side view.
The collected dataset allowed us to determine the structure of the minimal Chp2-Mit1 complex by molecular replacement, using the chromo shadow domain of Swi6 as a model (PDB ID 1E0B), at a resolution of 1.6 Å (Table1). Electron density is well defined for the entire CSD of Chp2, and we observe an equally well defined density for residues8-81of Mit1, which allowed for the building of an atomic model of the Mit1 N-terminus. The structure shows that one CSD dimer of Chp2binds to one Mit1molecule (Figure11A), resulting in a stoichiometry of2:1. While the homodimer of Chp2CSD is symmet- ric, binding of Mit1 across the symmetry axis breaks this symmetry and yields an asymmetric complex in which the two copies of Chp2 contribute unequally to Mit1 binding. Intriguingly, the structure fur- ther reveals a small domain in Mit1 that packs against one of the CSD domains. Easily recognizable, this domain features a chromo- domain fold as demonstrated by the close resemblance to the Chp1 chromodomain (Figure 11B) [Schalch et al., 2009b]. The cryptic Mit1 chromodomain-like (CDL) domain however lacks an aromatic cage and the binding groove for a methyl-lysine histone peptide. Instead, this binding groove of the domain is occupied by a CSD of Chp2 (Figure11A).
Further upstream of the CDL we observe an extended polypeptide that follows the surface of the CSD and is anchored in the canonical CSD peptide-binding groove formed by the dimerization of the Chp2 CSDs. The structure thereby defines the equivalent of the PxVxL mo-
Figure11: Crystal structure of Chp2HP1-Mit1complex reveals extensive interface.
(A) Overview of the Chp2-Mit1crystal structure. Eye symbols with letters indicate viewing angles for corresponding details panels in Fig.12A-D. (B) Superposition of the Mit1CDL domain with the chromodomain of Chp1bound to an H3K9trimethyl peptide (PDBID:3G7L, rmsd=1.38Å).
tif in the Chp2-Mit1 interaction as residues9-13of Mit1, correspond- ing to the sequence CkIvV (Figure 12A). This motif deviates signifi- cantly from the motif expected for an HP1-client interaction, which is defined as [PVL]-x-[VL]-x-[VL]. However, the mode of interaction is identical to reported HP1-PxVxL structures, with the ends of the C-termini of the Chp2 CSDs forming a deep groove between them, which accommodates the Mit1peptide to form a three-stranded mixed β-sheet. Mit1Cys9, Ile11and Val13point into the bottom of the bind- ing groove, which is lined by two symmetric sets of residues Tyr372, Tyr373, His376, Ile377, Phe379, Ile377from both Chp2protomers. The bottom of the binding grove is thus mostly hydrophobic and provides three pockets for harboring Cys9, Ile11 and Val13. Ile11 points down into the groove along the two-fold symmetry axis of the Chp2 CSD dimer, and therefore occupies the center of the two-fold symmetric
binding grove. Ile11 is a large hydrophobic side chain, and Chp2ac- commodates this extra bulk by rotating the Tyr373 side chain so it packs the aromatic face of the tyrosine side against the methyl groups of Ile11. The outer pocket where the side chains of Cys9and Val13are found is formed by Tyr372and His376from one protomer and Phe379 and Ile377from the other protomer. This pocket is relatively spacious and easily accommodates the cysteine and valine residue, but with- out visibly strong complementarity or formation of specific hydrogen bonds. Thus, the structure suggests that the Mit1-Chp2 equivalent of the HP1-PxVxL interaction is based on a CkIvV motif that binds in a well-defined manner to a hydrophobic groove due to the hydro- gen bonds established by β-sheet formation. Based on the structure, we predict that the groove accommodates small to medium size hy- drophobic side chains, but without high complementarity or hydro- gen bonding necessary for highly specific readout.
Residues 15-31C-terminal of the CkIvV motif in Mit1closely follow the surface of one of the Chp2 molecules in the CDS dimer. We call this region the linker region between the CkIvV motif and the CDL.
The linker contains a short310helix formed by a hydrophobic stretch (residues 18-22). The three leucines on this helix all fit into comple- mentary pockets in the Chp2surface (Figure12B).
After this hydrophobic interaction follows a series of polar and acidic residues, which form a hydrogen-bonding network with Chp2residues Lys316, His356and Asp357(Figure 12C). Lys316forms a salt bridge with Mit1Asp26, while His356engages the side chain of Mit1Glu29 in a bifurcated hydrogen bond. Asp357 establishes one hydrogen bond with Mit1Tyr25and one with Thr30. Due to these hydrophobic and polar interactions the entire linker region of Mit1is tightly asso- ciated with the Chp2 surface and we predict it to play a significant role in mediating the interaction between Chp2and Mit1.
The contacts between Chp2and the Mit1CDL domain extend the in- terface established by the hydrogen bonding network in the linker re- gion. Key residues in the Mit1CDL are Lys62and Tyr63(Figure12D).
Lys62 is involved in a hydrogen bond to the backbone oxygen of Asn339of Chp2and reciprocally the amide group of Asn339hydrogen- bonds to the backbone oxygen of Lys62. The same oxygen is also found in hydrogen-bonding distance with the Nε atom of Chp2’s Lys341. Mit1Tyr63is involved in a water-mediated hydrogen-bonding network that connects it to backbone atoms of Asn358, Ile359 and Leu342in Chp2. Furthermore, we find hydrophobic contacts of Mit1 Phe47, Val51and Ala53with Ile360and Ile359of Chp2(Figure12D).
Figure12: All parts of the interface contribute to binding between Chp2and Mit1.
Structures colored as in Fig.11. (A) Binding of the CkIvV motif to the groove formed by the Chp2CSD dimer. (B) Hydrophobic interactions of the linker domain of Mit1 with the surface of Chp2CSD. (C) Hydrogen bonding network between Mit1linker domain and Chp2CSD in the left of the panel, hydrophobic interactions between Mit1CDL and Chp2CSD on the right of the panel. (D) shows water-mediated con- tacts at the Mit1CDL-Chp2CSD interaction interface.
In summary, the Chp2-Mit1structure reveals an unexpectedly exten- sive interaction interface that buries 2521 Å2 of solvent-accessible surface area, corresponding to almost one quarter of the total sur- face. It consists of CkIvV motif, linker region and CDL domain, each of which provides a significant number of hydrogen bonds and hy- drophobic interactions. Side-chain specificity is observed less in the CkIvV motif and more in the linker and CDL part of the interaction interface. The structure proposes that all three parts are required for full binding affinity and for providing specific binding towards Chp2.
Chp2and Mit1engage in a high-affinity interaction
The structure of the Chp2–Mit1 interface suggests that the affinity of this interaction will be substantially different from the affinity for a typical HP1–client interaction. In order to determine the binding en- ergy of the Chp2–Mit1interaction we subjected the complex to anal- ysis by fluorescence polarization measurements. For this, we made use of the fact that Chp2 CSD contains only two cysteines, one of which is buried inside the domain, while the other is accessible to la- beling with thiol-reactive dyes. While we could measure an increase
of fluorescence polarization when incubating labeled Chp2CSD with increasing amounts of Mit1, we could not fit the data to calculate affin- ity. Interestingly, we noticed that fluorescence intensity increased as well as a function of Mit1concentration and we were able to calculate affinities in the single-digit micromolar range (Figure 13). However, affinity seemingly increased with longer incubation times.
Figure13: Fluorescence intensity measurements.Fluorescence of labeled Chp2in- creases as a function of Mit1concentration, but calculated affinities depend on the incubation time.
Since we could not explain the observed behavior and wanted a more robust way to determine binding energy, we resorted to isothermal calorimetry (ITC) measurements. Figure14A shows that combination of Mit1and Chp2resulted in an exothermic reaction with a dissocia- tion constant (Kd) of0.30 μM (Table 2). This Kd is on the lower end of the range of values reported previously for well-established HP1- client interactions, which varies from 0.2-10 μM [Isaac et al., 2017, Kang et al., 2011]. We monitored protein quality by size-exclusion chromatography and gel electrophoresis and found that the analyzed proteins were stable and not aggregated (Figure14B, C).
Figure14: ITC reveals nanomolar affinity between Chp2CSD and Mit1.(A) ITC experiment monitoring injection of Chp2 CSD into Mit1(1-81). (B) Size-exclusion chromatography and gel electrophoresis of Mit1(1-81) sample used for ITC. (C) Chp2 CSD sample.
The extended Chp2–Mit1interface is required for high affinity
In the crystal structure we observed that the binding surface consists of CkIvV motif, linker and CDL domain. To address the individual contributions of these Mit1fragments to binding energy, we subjected the interface to a deletion analysis and used ITC as a readout. Mutat- ing the motif or deleting the CDL both lead to a strong reduction in release of binding energy and resulted in a Kd of 5-12 μM, approxi- mately20-40-fold lower than the intact interface (Figure15A, Table2).
A construct with linker region and CDL left intact but missing the mo- tif altogether was designed and expressed but proved too unstable to purify. Since all Mit1 constructs still carried a StrepSUMO tag from purification, we also measured binding energy between tag alone and Chp2 CSD and could rule out any non-specific interaction contribut-
Figure15: The intact interface between Mit1and Chp2is needed for high affinity binding. (A) Injection of Chp2CSD into different Mit1constructs and tag alone as a negative control. The heat rate trace for injection of Chp2 CSD into Mit1(1- 31) is shown as an example. (B) SEC and gel electrophoresis for samples used for ITC, Mit1(1-81) I11R (C) ITC sample Mit1(1-31) shows some degradation that was accounted for in Kd calculations.
ing to the measured affinity between Mit1 and Chp2 (Figure 15A, Table2).
Interestingly, while we observe the stoichiometry in the crystal struc- ture to be 2 (the Chp2 CSD dimer binding to one molecule of Mit1) and this value to be reflected in the ITC measurements for Mit1(1-81) and Mit1(1-31), the stoichiometry changes to1 upon mutation of the motif (Table 2). We assume the mutation I11R of Mit1 to not fit the binding groove formed by the Chp2 CSD dimer while the interface between linker or CDL and Chp2is not disrupted. This would allow for two Mit1molecules to bind to the Chp2CSD dimer since the only part of Mit1contacting both copies of Chp2is the motif.
Again, sample quality was monitored by size-exclusion chromatog- raphy and gel electrophoresis (Figure 15B, C). For Mit1(1-31) we ob-
served a shoulder in the SEC peak that corresponded to a fraction of protein showing some degradation on the gel. We accounted for this degradation by adjusting the protein concentration used for calculat- ing affinity accordingly.
These experiments corroborate the observations made in the struc- ture and show that the extensive interface is required to provide a high-affinity interface for coupling Chp2and Mit1. Furthermore, this suggests that an intact interface is also required for Chp2 and Mit1 function in heterochromatic gene silencing.
Swi6and Chp2differ substantially in their role for SHREC recruitment Since Swi6was reported to be involved with SHREC recruitment and to directly interact with Mit1, we wanted to understand how this in- teraction compares to the high-affinity Mit1-Chp2interaction [Fischer et al., 2009]. To this end, we purified the chromo shadow domain of Swi6 and subjected it to ITC measurements. In addition, we tested binding to a known Swi6interaction partner, a PxVxL-motif contain- ing Sgo1 peptide [Isaac et al.,2017]. In addition to reproducing the reported affinity for the Sgo1-Swi6interaction within error, we found that Mit1binds Swi6with the same affinity within error. Furthermore, we could show that Sgo1similarly interacted with Chp2(Figure16A).
This points to an interaction network between theS. pombe HP1 pro- teins and their partners that seems to be more general and is based on PxVxL-like motifs being accommodated by the binding groove with- out strict sequence specificity. The interaction between Chp2and Mit1 on the other hand is ~50times tighter and depends on more than just the motif binding.
Size-exclusion chromatography and gel electrophoresis allowed us to verify that also the Sgo1 peptide and Swi6 CSD are stable and not aggregated (Figure 16B, C). However, we noticed that while stoi- chiometry of the Swi6-Mit1complex seemed to be2as predicted from PxVxL-based interactions, for both Swi6-Sgo and Chp2-Sgo1we mea- sured a binding ratio of1:1. Since proteins elute as a single peak from size-exclusion chromatography and show no degradation on gel elec- trophoresis, we assume them to be functional and therefore cannot explain why the binding ratio appears to be different than expected.
Further experiments will have to be conducted to rule out artefacts and understand possible alternative binding modes.
