The Role of BAG6 in Protein Quality Control

The role of BAG6 in protein quality control

A thesis submitted to The University of Manchester for the degree of Doctor of Philosophy in the Faculty of Biology, Medicine and Health

2017

Yee Hui Koay

School of Biological Sciences

List of contents Page List of tables 5 List of figures 6 List of abbreviations 9 Abstract 13 Declaration 14 Copyright statement 14 Acknowledgement 15 Chapter 1: Introduction 1.1 Protein folding and quality control 16 1.2 Degradation of misfolded proteins 21 1.3 Protein biosynthesis and quality control at the endoplasmic reticulum 24 1.4 BAG6 31 1.4.1 BAG6 structure 32 1.4.2 BAG6 in tail-anchored protein targeting 34 1.4.3 BAG6 in degradation of mislocalised proteins 35 1.4.4 BAG6 triages targeting and degradation 39 1.4.5 BAG6 in endoplasmic reticulum-associated degradation 43 1.5 Aims and objectives of study 46 Chapter 2: Materials and methods 2.1 DH5α competent cells preparation 47 2.2 Plasmid DNA preparation 47 2.3 Plasmid construction 49 2.3.1 Myc-BirA-pcDNA5/FRT/TO 49 2.3.2 BAG6-myc-BirA-pcDNA5/FRT/TO 51 2.3.3 BAG6(∆N)-myc-BirA-pcDNA5/FRT/TO 51

2.3.4 HA3-XBP1u G519C(∆HR2)-pcDNA3.1(+) 52 2.4 Cell culture 52

2.5 Stable inducible cell line generation and induction 53 2.6 siRNA transfection 53 2.7 Transient transfection for immunoblotting and immunofluorescence microscopy 54 2.8 Treatment with proteasome and lysosomal protease inhibitors 55 2.9 Endo Hf treatment 55 2.10 Cycloheximide chase 55 2.11 SDS-PAGE and immunoblotting 56 2.12 Immunofluorescence microscopy 58 2.13 Cell cracking 59 2.14 Co-immunoprecipitation 60 2.15 Denaturing immunoprecipitation 60 2.16 BioID 61 2.17 Bioinformatics 62 2.18 Data and statistical analysis 62 Chapter 3: Role of BAG6 and UBR4 in ERAD 3.1 Introduction 64 3.2 BAG6 contributes to degradation of Op91 66 3.3 BAG6 and UBR4 are involved in ERAD of opsin-degron 71 3.4 UBR4 does not associate stably with BAG6 or opsin-degron 73 3.5 Depletion of UBR4 increases opsin-degron ubiquitination 75 3.6 UBR4 knockdown increases association of BAG6 with 80 opsin-degron and RNF126 3.7 Discussion 82 Chapter 4: Novel role of BAG6 in the UPR 4.1 Introduction 87 4.2 BAG6 interacts with XBP1u 90 4.3 BAG6 is required for efficient XBP1u degradation 95 4.4 BAG6 interacts with the C-terminal hydrophobic region of full length XBP1u in order to promote degradation 99 4.5 BAG6 depletion impairs XBP1u ubiquitination 103

4.6 Discussion 105 Chapter 5: The use of BioID to identify potential BAG6 substrates and/or interacting factors 5.1 Introduction 111 5.2 Generation and characterisation of BAG6-myc-BirA construct 113 5.3 Optimisation of BioID protocol 117 5.4 The use of transient expression system for efficient biotinylation 120 5.5 Mass spectrometry data analysis 123 5.5.1 Comparison of technical replicates 123 5.5.2 Pre-enrichment analysis of myc-BirA vs BAG6-myc-BirA 125 5.5.3 Comparison of biological replicates 126 5.5.4 Identification of potential BAG6 substrates 128 5.5.5 Identification of potential BAG6 interacting factors 135 5.6 Discussion 138 Chapter 6: Conclusions and future work 6.1 Conclusions 146 6.2 Future work 148 Chapter 7: References 150 Chapter 8: Supplementary data 165

Final word count: 41748

List of tables Page Table 2.1 Plasmid constructs used in this study 48 Table 2.2 Standard reaction setup and PCR cycling condition for site-directed mutagenesis using KOD Hot Start DNA Polymerase 50

Table 2.3 PCR cycling condition for subcloning 51 Table 2.4 siRNA sequences used in this study 54 Table 2.5 Primary and secondary antibodies used in immunoblotting 57 Table 2.6 Primary and secondary antibodies used in immunofluorescence microscopy 59

Table 5.1 Mean LFQ intensity of proteins found in both biological replicates and having at least two times higher intensity with BAG6-myc-BirA sample than myc-BirA sample 131 Table 5.2 Hydrophobicity prediction using Kyte and Doolittle scale with window size of 20 and threshold of 0 136

Table 5.3 Potential BAG6-associated protein quality control factors with gene ontology related to ubiquitination pathway 139

List of figures

Page Figure 1.1 Illustration on a protein folding energy landscape 18 Figure 1.2 The Hsp70 and TRiC/CCT chaperone systems 20 Figure 1.3 The ubiquitination cascade and chain formation 22 Figure 1.4 The three major protein translocation pathways: co- or post-translational translocation dependent on Sec61 and Sec61-independent post-translational translocation 25 Figure 1.5 The mammalian endoplasmic reticulum-associated degradation 28 Figure 1.6 The mammalian unfolded protein response signalling pathways 30 Figure 1.7 Schematic representation of human BAG6 protein domains and its putative binding partners 33 Figure 1.8 The yeast and mammalian guided entry of tail-anchor pathway components 36 Figure 1.9 Model illustrating BAG6 role in protein quality control in the cytosol 37 Figure 1.10 Model illustrating BAG6 role in endoplasmic reticulum pre-emptive quality control 40 Figure 1.11 Current model illustrating BAG6 triages protein targeting and degradation 41 Figure 1.12 Model illustrating BAG6 role in endoplasmic reticulum- associated degradation 44 Figure 3.1 UBR family proteins and the mammalian N-end rule pathway 65 Figure 3.2 BAG6 overexpression stabilises Op91 and causes nuclear re-localisation of Op91 67 Figure 3.3 Degradation of glycosylated form of Op91 is delayed with BAG6 knockdown 70 Figure 3.4 Degradation of opsin-degron is delayed with BAG6 and UBR4 knockdown 72 Figure 3.5 BAG6 and opsin-degron do not co-immunoprecipitate UBR4 74 Figure 3.6 UBR4 knockdown does not significantly alter BAG6 steady state level 76

Figure 3.7 UBR4 knockdown increases total cellular ubiquitination and BAG6-associated polyubiquitinated species 78 Figure 3.8 UBR4 knockdown increases opsin-degron ubiquitination 79 Figure 3.9 UBR4 knockdown increases opsin-degron-BAG6 and BAG6- RNF126 interactions 81 Figure 3.10 Proposed model on UBR4 function in ERAD in correlation with BAG6 86 Figure 4.1(A) Schematic representation of human XBP1 protein domains 88 Figure 4.1(B) Model illustrating the splicing event of XBP1 mRNA under ER stress 88 Figure 4.2 BAG6 re-localises and stabilises XBP1u 91 Figure 4.3 BAG6 co-purifies XBP1u with proteasome inhibition 94 Figure 4.4 Proteasome inhibition delays XBP1u degradation 96 Figure 4.5 BAG6 knockdown delays XBP1u degradation 98 Figure 4.6 UBR4 and RNF126 knockdown have no effect on XBP1u degradation 99 Figure 4.7 BAG6 cannot interact with XBP1 variant lacking HR2. XBP1 HR2 mutant is more stable and does not require BAG6 for degradation 101 Figure 4.8 BAG6 knockdown decreases XBP1u ubiquitination 104 Figure 4.9 Proposed model on BAG6-mediated XBP1u degradation 108 Figure 5.1 Model for application of BioID method 112 Figure 5.2 BAG6-myc-BirA re-localises and stabilises Op91 114 Figure 5.3 RIPA buffer extracts more proteins and fewer proteins are being washed away 118 Figure 5.4 Transient overexpression system is used for BioID experiment 121 Figure 5.5 Box plots of LFQ intensity and MS/MS count with respect to the three experiment conditions and technical replicates for BAG6-myc-BirA and BAG6-myc-BirA+Bz samples 124 Figure 5.6 Venn diagrams, histograms and scatter plots of proteins identified from LFQ intensity and MS/MS count of myc-BirA and BAG6-myc-BirA samples 127

Figure 5.7(A) Box plot of non-normalised and median-normalised LFQ intensities with respect to the four experiment conditions and two biological replicates except for Bag6(∆N)-myc-BirA with no biological replicate 129 Figure 5.7(B) Venn diagram of proteins identified from myc-BirA sample of two biological replicates 129 Figure 5.7(C) Venn diagram of proteins identified from BAG6-myc-BirA sample of two biological replicates 129 Figure 5.7(D) Dendrogram showing how the experiments cluster 129 Figure 5.8 Flow chart for the identification of potential BAG6 substrates 130 Figure 5.9 Venn diagram of proteins identified from BAG6-myc-BirA and BAG6(∆N)-myc-BirA samples 137

List of abbreviations

ADP adenosine diphosphate AMP adenosine monophosphate AP-MS affinity purification coupled to mass spectrometry ATE1 arginyl-tRNA--protein transferase 1 ATF4 cyclic AMP-dependent transcription factor ATF-4 ATF6 cyclic AMP-dependent transcription factor ATF-6 ATP adenosine triphosphate BAG6 BCL2-associated athanogene 6 BAT3 HLA-B associated transcript 3 BiP immunoglobulin heavy chain-binding protein BirA biotin ligase CAML calcium-modulating cyclophilin ligand CHIP E3 ubiquitin-protein ligase CHIP DMEM Dulbecco’s Modified Eagle’s Medium EDEM1 ER degradation-enhancing alpha-mannosidase-like protein 1 eIF2α eukaryotic translation initiation factor 2-alpha Endo Hf endoglycosidase Hf ER endoplasmic reticulum ERAD endoplasmic reticulum-associated degradation ER pQC endoplasmic reticulum pre-emptive quality control GET guided entry of tail-anchor gp78 E3 ubiquitin-protein ligase AMFR HECT homologous to the E6-AP carboxyl terminus Hrd1 E3 ubiquitin-protein ligase synoviolin Hsc70 heat shock cognate 71 kDa protein Hsp70 heat shock 70 kDa protein Hsp40 heat shock 40 kDa protein

IPOD insoluble protein deposit IRE1 serine/threonine-protein kinase/endoribonuclease IRE1 JUNQ juxta nuclear quality control compartment KCMF1 E3 ubiquitin-protein ligase KCMF1 LC-MS/MS liquid chromatography-tandem mass spectrometry LC3 microtubule-associated proteins 1A/1B light chain 3 LFQ label-free quantification MARCH6 E3 ubiquitin-protein ligase MARCH6 NLS nuclear localisation signal NO nitric oxide Npl4 nuclear protein localization protein 4 homolog NTAN1 protein N-terminal asparagine amidohydrolase NTAQ1 protein N-terminal glutamine amidohydrolase OpD opsin degron Op91 N-terminal 91 amino acids of opsin OS9 protein OS-9 PCR polymerase chain reaction PERK PRKR-like endoplasmic reticulum kinase PHD plant homeodomain Pi inorganic phosphate PINK1 serine/threonine-protein kinase PINK1 PPi inorganic pyrophosphate p97 15S Mg(2+)-ATPase p97 subunit RAD6A ubiquitin-conjugating enzyme E2 A RAD6B ubiquitin-conjugating enzyme E2 B RING really interesting new gene RNF5 E3 ubiquitin-protein ligase RNF5 RNF126 E3 ubiquitin-protein ligase RNF126 RNF139 E3 ubiquitin-protein ligase RNF139

RNF170 E3 ubiquitin-protein ligase RNF170 Rpn10 26S proteasome regulatory subunit RPN10 Rpn11 26S proteasome regulatory subunit RPN11 Rpn13 proteasome regulatory particle non-ATPase 13 RtcB tRNA-splicing ligase RtcB homolog SDS-PAGE sodium dodecyl sulfate-polyacrylamide gel electrophoresis Sec61 protein transport protein Sec61 Sec61β protein transport protein Sec61 subunit beta SGTA small glutamine-rich tetratricopeptide repeat-containing protein alpha Sgt2 small glutamine-rich tetratricopeptide repeat-containing protein 2 SOD1 superoxide dismutase 1 SPP signal peptide peptidase SRP signal recognition particle SRPR signal recognition particle receptor TA tail-anchored TCRα T-cell receptor alpha chain TMEM129 E3 ubiquitin-protein ligase TM129 TRC35 transmembrane domain recognition complex 35 kDa subunit TRC40 transmembrane domain recognition complex 40 kDa ATPase subunit TRiC/CCT TCP-1 Ring Complex/ chaperonin containing TCP-1 UBA ubiquitin-associated UBL ubiquitin-like Ubl4A ubiquitin-like protein 4A UBR4 E3 ubiquitin-protein ligase UBR4 UBXD8 UBX domain-containing protein 8 Ufd1 ubiquitin recognition factor in ER-associated degradation protein 1 UPR unfolded protein response WRB tail-anchored protein insertion receptor XBP1 X-box-binding protein 1

XTP3B XTP3-transactivated gene B protein

Abstract Institution: The University of Manchester Name: Yee Hui Koay Degree: Doctor of Philosophy Thesis Title: The role of BAG6 in protein quality control Date: 2017

Proteins must be folded into their correct three-dimensional structure for functions. Failure of proteins to achieve their native conformation is harmful to cells because these proteins are prone to aggregate and engage in non-native interactions. Cells possess extensive protein quality control systems to deal with misfolded proteins, either by refolding the proteins or degrading terminally misfolded proteins. BAG6 has been implicated in cellular protein quality control systems, specifically by recognising stretches of hydrophobic amino acids, recruiting E3 ubiquitin ligase(s) to promote substrate ubiquitination and preventing substrate aggregation. Such stretches of hydrophobic amino acids are typically found in transmembrane domains and endoplasmic reticulum targeting sequences, and BAG6 has been shown to play a role in the quality control and degradation of these proteins. In addition, cytoplasmic proteins may also contain hydrophobic sequences but these would normally be buried within the native structure, and only become exposed if the proteins misfold. However, the precise role of BAG6 in these pathways and the wider contribution of BAG6 to the handling of other aggregation-prone proteins in the cytosol are yet to be studied. Hence, the aim of this project was to determine BAG6 substrate specificity and to gain a better understanding of BAG6 role by identifying endogenous BAG6 substrates and co-factors through targeted and unbiased approaches. UBR4 has been identified as a protein that interacts with a BAG6 substrate (Sec61β) and interacting factor (SGTA) in two independent experiments. UBR4 was found to play a role in endoplasmic reticulum-associated degradation of a BAG6 substrate (opsin-degron), possibly by affecting BAG6-substrate and BAG6-E3 ligase interactions. The unspliced form of XBP1 was predicted to be a BAG6 substrate since XBP1 is a cytoplasmic protein with hydrophobic domain likely to be exposed in the cytoplasm for BAG6 recognition. BAG6 was shown to interact with XBP1 through the hydrophobic domain and affect XBP1 turnover through ubiquitination. Unspliced XBP1 acts as a negative regulator of the spliced XBP1, which is an important transcription factor in the unfolded protein response. BioID was performed as an unbiased approach for BAG6 substrates and co-factors identification. Biotin ligase fused to BAG6 biotinylated proteins coming into close proximity to BAG6, and biotinylated proteins were isolated and analysed with mass spectrometry. Based on bioinformatic analysis of the mass spectrometry data, hydrophobicity of BAG6 ‘substrates’ varied greatly. Biochemical analysis is needed in the future to validate these BAG6-substrate interactions. Together, these results further our understanding of BAG6 role in the protein quality control, especially in the unfolded protein response which was reported for the first time.

Declaration

No portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning.

Copyright statement i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes. ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made. iii. The ownership of certain Copyright, patents, designs, trademarks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions. iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://documents.manchester.ac.uk/display.aspx?DocID=24420), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.library.manchester.ac.uk/about/regulations/) and in The University’s policy on Presentation of Theses

Acknowledgement

I have been enjoying my Ph.D. study for the past 4 years and this is only made possible with the people I have known throughout the project. Firstly, I would like to thank my supervisor, Dr. Lisa Swanton for her guidance, tolerance and support as such this thesis would not have been possible without her. Special thanks to my co-supervisors, Professor Stephen High, Professor Simon Hubbard, Professor Stuart Pickering-Brown, and my advisor, Dr. Andrew Gilmore for their generous help in many topics related to my research work. My sincere appreciation goes to my lovely colleagues (Carolina Uggenti, Loic Gazquez, Kit Briant, Eleanor French and Ahmed Boukerrou) including all members of the Pool and Ford labs for providing me a lively, carefree and joyful working environment. I would like to express my gratitude to all members of the High, Allan, Woodman and Lowe labs for sharing plasmids, antibodies and equipment. I would also like to acknowledge the staffs of the Faculty of Biology, Medicine and Health for their continuous assistance in all aspects when I was conducting my project. I genuinely thank MyBrainSc Scholarship (Ministry of Education Malaysia) and President’s Doctoral Scholar Award (University of Manchester) for keeping me out of financial problems while pursuing my dream in the UK. Lastly, I am grateful to my SEA friends, my family, and my husband for their endless support.

1 Introduction

Proteins are macromolecules that perform a diverse range of structural and biochemical roles including enzymatic reaction, ligand binding and cell signalling. Such diverse functionalities largely depend upon their three-dimensional structures. Folding proteins may face several kinetic barriers before reaching their native states and failure to attain the correct conformations results in the production of misfolded proteins. Misfolded proteins are prone to aggregate and therefore pose a threat to cell viability. Cells possess several quality control mechanisms to refold or degrade misfolded proteins before they harm the cells. Failure of protein quality control pathways can lead to harmful accumulation of aggregation-prone proteins, as occurs in many human diseases including neurodegenerative diseases. The protein BCL2- associated athanogene 6 (BAG6) has been implicated in the quality control system of cells, specifically by recognising stretches of hydrophobic amino acids.

1.1 Protein folding and quality control

Proteins need to be folded into their native three-dimensional structures for function. The three-dimensional structure is ultimately determined by the amino acid sequence (Baker, 2000; Schaeffer and Daggett, 2011; Kim et al., 2013; Guzman and Gruebele, 2014). The primary amino acid sequence contains all the information required for a protein to achieve its correct conformation, through hydrogen bonding, Van der Waals forces, electrostatic forces, hydrophobic interactions and protein backbone angles. Even minor changes in the primary amino acid sequence can cause a protein to have total structural change (Guo et al., 2004; Khan and Vihinen, 2007; Alexander et al., 2009). In most cases, a polypeptide chain will first attain secondary structure (alpha helix or beta sheet) before folding into its tertiary globular structure with increased stability (Schaeffer and Daggett, 2011; Dill and MacCallum, 2012). The protein folding process can be viewed by the statistical thermodynamics of polymer (Dill and MacCallum, 2012; Kim et al., 2013; Englander and Mayne, 2014). Each protein has a distinct conformational space, which is the shape of its respective energy landscape. Conformational entropy studies using foldable polymer models have shown that there are more higher-energy unfolded species undergoing random folding into fewer lower-energy intermediates, meaning that the energy landscape is 16 funnel-shaped (Dill and MacCallum, 2012; Kim et al., 2013; Englander and Mayne, 2014) (Figure 1.1). Proteins undergoing folding may be trapped by several kinetic barriers before reaching their native state; trapped partially folded or misfolded intermediates may be of lower energy state than the native proteins (Figure 1.1), and this increases the time for the folding process to be completed (Kim et al., 2013). The structure of a protein is also affected by protein-protein and protein-solvent interactions (Prabhu and Sharp, 2006; Hong, 2014). Folding of proteins comprising hydrophobic amino acids is largely driven by the hydrophobic effect, where hydrophobic side chains are buried in the protein structure forming a hydrophobic core while polar side chains are exposed on the surface of the protein (Kyte, 2003; Sarkar and Kellogg, 2010; Kim et al., 2013; Lupas and Alva, 2017). Minimising exposure of hydrophobic side chains to water molecules in the cytoplasm is the principle behind the folding process and thus stabilises a protein structure. Molecular chaperones are especially important in successful folding of many proteins and function by recognising non-native protein conformations especially exposed hydrophobic stretches in folding intermediates or misfolded proteins to prevent irreversible aggregation reactions while allowing sufficient time for proteins to achieve their native conformations (Kim et al., 2013; Saibil, 2013; Brandvold and Morimoto, 2015; Bose and Chakrabarti, 2017). Hence, molecular chaperones restrict the conformational space and direct proteins down the funnel (Figure 1.1).

De novo protein folding may begin co-translationally at the ribosome (Kramer et al., 2009; Pechmann et al., 2013; Javed et al., 2017). The ribosome exit tunnel plays active roles in nascent chain folding: it allows limited folding of nascent chains into secondary or simple tertiary conformations (Kramer et al., 2009; Pechmann et al., 2013; Javed et al., 2017) and recruits chaperones and targeting factors (Kramer et al., 2009; Pechmann et al., 2013). In the cytosol, the heat shock 70 kDa protein (Hsp70) family and heat shock 40 kDa protein (Hsp40) co-factor act as holdases that maintain nascent polypeptides in their unfolded states. Holdase activity is an adenosine triphosphate (ATP)-independent process but de-novo folding or refolding of misfolded proteins requires ATP for substrates binding and release by Hsp70 (Brehme and Voisine, 2016). Hsp70 is usually present in ATP-bound state with very low intrinsic ATPase activity (Brandvold and Morimoto, 2015; Brehme and Voisine, 2016). When Hsp40 recruits unfolded polypeptide to Hsp70, Hsp40

Figure 1.1: Illustration on a protein folding energy landscape. Unfolded polypeptide folds into lower-energy intermediate before attaining native state to perform cellular functions. Chaperones are needed to assist in directing the energetically favourable non-native conformation to the native state and prevent non- native conformation from converting into the misfolded state. (Adapted from Kim et al., 2013)

18 facilitates ATP hydrolysis in Hsp70 resulting in an increased Hsp70-substrate affinity by closing Hsp70 substrate binding domain. Engagement of a nucleotide exchange factor to the Hsp70-substrate complex causes an exchange of adenosine diphosphate (ADP) for ATP, opening the substrate binding domain and releasing fully or partially folded polypeptide. Released polypeptide undergoes spontaneous folding until it is captured by Hsp40 for another Hsp70 cycle unless the final conformation is achieved (Kim et al., 2013; Shiber and Ravid, 2014; Brandvold and Morimoto, 2015; Brehme and Voisine, 2016) (Figure 1.2A). TCP-1 Ring Complex/ chaperonin containing TCP-1 (TRiC/CCT) is important for co- and post-translational folding of certain proteins that cannot be folded with other chaperones (Spiess et al., 2004; Roh et al., 2015). TRiC/CCT serves as a large and undisturbed folding barrel where unfolded polypeptide is sequestered into the central cavity away from the crowded cytoplasm for productive folding (Spiess et al., 2004; Kim et al., 2013; Saibil, 2013; Brandvold and Morimoto, 2015). Apical protrusions act as a lid to open or close the barrel during ATPase cycle where substrate encapsulation is critical for folding (Spiess et al., 2004; Kim et al., 2013; Saibil, 2013; Brandvold and Morimoto, 2015) (Figure 1.2B). Small heat shock proteins are ATP-independent chaperones that loosen aggregates and hold them for subsequent solubilisation and refolding by previously described chaperones Hsp70/Hsp40 and TRiC/CCT (Kim et al., 2013; Mogk and Bukau, 2017).

Newly synthesised proteins may fail to attain the correct native state due to intrinsic (underlying genetic mutations, transcriptional or translational errors, ageing) or extrinsic (exposure to environmental stressors such as reactive oxygen species) factors which disrupt the folding process and result in the production of misfolded species (Amm et al., 2014; Shiber and Ravid, 2014). Existing native proteins may also be susceptible to misfolding, particularly those that exhibit conformational flexibility (Kim et al., 2013; Amm et al., 2014; Koldewey et al., 2017). Misfolding occurs when folding intermediates become trapped at a lower-energy point than the native state in the folding landscape and hence fail to attain the native conformations (Figure 1.1). Abnormal protein conformations are potentially harmful to cells because they may disrupt native interactions and create novel non-native interactions, both of which can perturb cell function (Kim et al., 2013; Koldewey et al., 2017). Therefore, cells possess protein quality control systems in order to identify and

Figure 1.2: The Hsp70 and TRiC/CCT chaperone systems. (A) Heat shock 40 kDa protein (Hsp40) recruits unfolded polypeptide to adenosine triphosphate (ATP)- bound heat shock 70 kDa protein (Hsp70) and facilitates ATP hydrolysis to close Hsp70 substrate binding domain. Nucleotide exchange factor (NEF) then exchanges adenosine diphosphate (ADP) for ATP, opening the substrate binding domain and releasing native protein for functions or folding intermediate for further rounds of folding. Pi, inorganic phosphate. (Adapted from Brehme and Voisine, 2016) (B) Unfolded polypeptide and ATP are recruited to TCP-1 Ring Complex (TRiC). ATP hydrolysis causes lid closure for folding to occur until native conformation is achieved. (Adapted from Spiess et al., 2004) 20 degrade the misfolded or non-native proteins that are continuously generated in living cells. These quality control pathways recognise exposed non-native features of abnormally folded proteins, such as hydrophobic stretches that would normally be buried inside the native structure (Kriegenburg et al., 2012; Shiber and Ravid, 2014). Molecular chaperones play key roles not only in protein folding but also quality control. Once non-native proteins are recognised, there is often a triage decision whether to refold the proteins or to degrade them through ubiquitin-proteasome system/autophagy-lysosome system, both cases involving molecular chaperones (Kriegenburg et al., 2012; Stoecklin and Bukau, 2013; Amm et al., 2014; Shiber and Ravid, 2014). Although the triage decision making is still not clearly understood, cells are predicted to prioritise refolding over degradation (Shiber and Ravid, 2014). However, a considerable proportion of newly synthesised proteins are thought to be degraded soon after synthesis, implying that they fail to attain correct conformations usually due to translational errors or post-translational processes that inhibit folding (Schubert et al., 2000; Kleiger and Mayor, 2014).

1.2 Degradation of misfolded proteins

A large proportion of misfolded proteins are degraded by the ubiquitin- proteasome system. A cascade of enzyme-catalysed reactions are required to covalently attach ubiquitin to target proteins (Husnjak and Dikic, 2012; Amm et al., 2014; Kleiger and Mayor, 2014; Ciechanover, 2015). This cascade is initiated when an E1 enzyme activates ubiquitin in the presence of ATP and transfers the activated ubiquitin to an E2 enzyme, which forms a complex with an E3 ubiquitin ligase. The human genome encodes two E1 enzymes, more than 30 E2 enzymes and approximately 600 known E3 ligases (Kawabe and Brose, 2011; Husnjak and Dikic, 2012; Kleiger and Mayor, 2014). Substrate specificity relies on E3 recognising a target protein and bridging the substrate and E2 enzyme for ubiquitin transfer to the substrate (Figure 1.3A). Ubiquitin modification usually occurs on lysine residues. The ubiquitination cycle may be repeated until a polyubiquitin chain is formed (Husnjak and Dikic, 2012; Kleiger and Mayor, 2014; Ciechanover, 2015). Ubiquitin has 7 internal lysine residues which can be conjugated with additional ubiquitin molecules, generating polyubiquitin chain with different types of linkage (Husnjak

Figure 1.3: The ubiquitination cascade and chain formation. (A) Ubiquitin (Ub) is activated by E1 ubiquitin-activating enzyme in the presence of adenosine triphosphate (ATP) and activated ubiquitin is transferred to E2 ubiquitin-conjugating enzyme. E3 ubiquitin ligase recognises and engages misfolded substrate which then forms a complex with E2 enzyme to facilitate ubiquitin transfer to the substrate. AMP, adenosine monophosphate; PPi, inorganic pyrophosphate. (B) Ubiquitination can happen only once (monoubiquitination) or repeatedly (multiple monoubiquitination or homotypic polyubiquitination) on the same misfolded substrate (yellow oval). Ubiquitin modification occurs on lysine (Lys) residues and ubiquitin has seven internal lysine residues, generating polyubiquitin chain with different types of linkage. (Adapted from Husnjak and Dikic, 2012)

22 and Dikic, 2012; Kleiger and Mayor, 2014) (Figure 1.3B). Not all polyubiquitin chain linkages act as a signal for proteasomal degradation; some polyubiquitin- modified proteins act as signalling molecules for diverse cellular functions (Figure 1.3B). Lys48-linked chain is a canonical signal for proteasomal degradation but more evidence is needed to confirm the role of Lys11- and Lys29-linked chains in proteasomal degradation (Akutsu et al., 2016; Swatek and Komander, 2016). Polyubiquitinated proteins destined for degradation are escorted/delivered to the proteasome by a series of ubiquitin-binding protein. One of the ubiquitin-binding domain families, the ubiquitin-associated (UBA) domain interacts with the ubiquitin- like (UBL) domain of itself or of other shuttle proteins then the UBL domain binds the proteasome (Hicke et al., 2005; Husnjak and Dikic, 2012). The 19S proteasome has two intrinsic ubiquitin receptors, 26S proteasome regulatory subunit RPN10 (Rpn10) and proteasome regulatory particle non-ATPase 13 (Rpn13), which bind polyubiquitinated substrate proteins or UBA-UBL-substrate protein complexes (Husnjak and Dikic, 2012; Kleiger and Mayor, 2014). 26S proteasome regulatory subunit RPN11 (Rpn11), also at the 19S proteasome, is a deubiquitinating enzyme which cleaves off the entire polyubiquitin chain so that the unfolded substrate can be translocated into the catalytic core of the 20S proteasome (Kleiger and Mayor, 2014; Yao, 2015). Polyubiquitin chain binding and cleavage are tightly regulated to make sure the de-ubiquitinated substrate does not return to the cytoplasm. Therefore, a key criterion for successful degradation is that the substrate needs to achieve a minimum polyubiquitin chain length where proteasome binding and ubiquitin chain cleavage can happen simultaneously (Kleiger and Mayor, 2014). Degradation of misfolded proteins often requires chaperones. Chaperones are thought to discriminate substrates and triage them for folding or degradation of substrates by interacting with different co-chaperones (Fernández-Fernández et al., 2017). It is important to maintain substrate solubility because non-native proteins tend to aggregate in the absence of chaperones and aggregated proteins cannot be degraded by the proteasome despite being ubiquitinated (Shiber and Ravid, 2014).

Aggregated proteins that are resistant to proteasomal degradation may evoke a different quality control mechanism, autophagy. Autophagy is a cellular mechanism for the degradation of long-lived proteins, old/damaged organelles and aggregated proteins by the lysosome (Kaushik and Cuervo, 2012; Murrow and

Debnath, 2013). Autophagy is divided into 3 main pathways: macroautophagy, microautophagy and chaperone-mediated autophagy (Kaushik and Cuervo, 2012; Murrow and Debnath, 2013), the former of which contributes to degradation of aggregated proteins. In cases where aggregated proteins are ubiquitinated, aggregates recruit the ubiquitin-binding protein p62 which in turn binds microtubule-associated proteins 1A/1B light chain 3 (LC3) and initiates formation of a double-membrane autophagosome around the aggregates (Kaushik and Cuervo, 2012; Murrow and Debnath, 2013). Autophagosomes fuse with lysosomes, delivering the aggregates to the interior of the lysosome for degradation (Kaushik and Cuervo, 2012; Murrow and Debnath, 2013). The ubiquitin-proteasome system and autophagy are not mutually exclusive, and considerable overlap exists between the two pathways (Ding et al., 2007; Korolchuk et al., 2010; Lilienbaum, 2013). Some misfolded proteins are degraded by both pathways (Ding et al., 2007; Korolchuk et al., 2010; Lilienbaum, 2013). In addition, impaired activity of the ubiquitin-proteasome system induces autophagy, whilst inhibition of autophagy leads to proteasome substrates accumulation (Ding et al., 2007; Korolchuk et al., 2010; Lilienbaum, 2013).

1.3 Protein biosynthesis and quality control at the endoplasmic reticulum

Up to one-third of the eukaryotic proteome has to be translocated across the endoplasmic reticulum (ER) membrane (Vembar and Brodsky, 2008; Hanulová and Weiss, 2012). Besides ER localised proteins, plasma membrane proteins and proteins secreted from cells or residing in the secretory pathway are first translocated into the ER. There are several distinct pathways for targeting nascent polypeptides to the ER: classical protein transport protein Sec61 (Sec61)-dependent co-translational translocation, Sec61-dependent post-translational translocation and Sec61- independent post-translational translocation (Figure 1.4). During co-translational translocation, a hydrophobic signal sequence at the N-terminus of secretory proteins or the first transmembrane domain of the translating polypeptide is recognised by signal recognition particle (SRP) as it emerges from the ribosome. SRP binding transiently slows down translation, and delivers the ribosome-nascent chain complex to the ER membrane by interacting with the SRP receptor on the ER. The hydrophobic sequence is then transferred from SRP to the Sec61 translocon complex

Figure 1.4: The three major protein translocation pathways: co- or post- translational translocation dependent on Sec61 and Sec61-independent post- translational translocation. If signal recognition particle (SRP) recognises a signal sequence, translational halts and co-translational translocation takes place. Transmembrane domain recognition complex 40 kDa ATPase subunit (TRC40) captures tail-anchored protein at the C-terminus transmembrane domain (TMD) and assists insertion through protein transport protein Sec61 (Sec61)-independent post- translational translocation. Sec61-dependent post-translational insertion does not involve both SRP and TRC40 but requires heat shock 70 kDa protein (Hsp70) and heat shock 40 kDa protein (Hsp40) chaperone system to hold hydrophobic domain (HD) and maintain nascent polypeptide in unfolded state. ER, endoplasmic reticulum; Sec61, protein transport protein Sec61; SRPR, signal recognition particle receptor; CAML, calcium-modulating cyclophilin ligand; WRB, tail-anchored protein insertion receptor WRB; ATP, adenosine triphosphate. (Adapted from Ast and Schuldiner, 2011)

25 which forms a channel in the ER membrane. Protein synthesis resumes and the nascent polypeptide is directly translocated into the ER (Nyathi et al., 2013; Saraogi and Shan, 2013). The hydrophobic signal sequence or transmembrane domain is integrated into the ER bilayer through a lateral gate in the Sec61 channel (Rapoport et al., 2017). This strategy ensures that hydrophobic sequences are not exposed to the cytoplasm where they could aggregate. In yeast, and potentially in mammals, some nascent polypeptides fail to effectively engage SRP due to their mild hydrophobicity or structural features of the signal sequence, and are therefore released into the cytoplasm upon completion of translation, and are translocated using the post- translational translocation pathway. Since the polypeptides are fully translated, chaperones such as Hsp70 and Hsp40 prevent aggregation and maintain the polypeptides in a loosely folded form to be translocated by the Sec61 complex (Ast and Schuldiner, 2013; Johnson et al., 2013). Some proteins are translocated post- translationally - after completion of translation and release of the protein from the ribosome. This is best studied for so-called tail-anchored (TA) proteins which lack an N-terminal signal sequence but possess a single transmembrane domain at their extreme C-terminus. As a result, the transmembrane domain only emerges from the ribosome once the whole polypeptide has been synthesised and released from the ribosome, which precludes co-translational SRP binding (Hegde and Keenan, 2011; Casson et al., 2016). Different chaperone-mediated TA protein insertion pathways have been described involving SRP, heat shock cognate 71 kDa protein (Hsc70)/Hsp40 or transmembrane domain recognition complex 40 kDa ATPase subunit (TRC40) (Hegde and Keenan, 2011). The client specificity of each pathway is believed to depend largely on chaperones engaged by TA proteins based on the hydrophobicity of the transmembrane domain sequence (Hegde and Keenan, 2011). However, in each pathway, binding to chaperones or other proteins occurs to prevent aggregation, particularly of the hydrophobic domain, and maintain the polypeptides in a conformation competent for translocation. Despite these multiple pathways, targeting of nascent polypeptides to the ER is not perfectly efficient. Proteins may fail to insert into or translocate across the ER membrane due to the inherent properties of the signal sequence, malfunction of SRP or other chaperones in the targeting pathways (Levine et al., 2005). As a result, nascent polypeptides possessing hydrophobic signal sequences or transmembrane domains are mislocalised to the cytoplasm (Levine et al., 2005). Indeed, for some proteins, up to 20 % fail to be

26 properly translocated (Rane et al., 2004). Conditions of ER stress also reduce the efficiency of translocation (Kang et al., 2006; Orsi et al., 2006; Rane et al., 2008). Mislocalised proteins are usually implicated for their pathological roles in the cytosol as they contain a shared feature of hydrophobicity in their targeting domains; therefore these proteins need to be promptly degraded by the ubiquitin-proteasome system to prevent aggregation and other non-specific interactions in the cytosol.

Secretory and membrane proteins translocated to the ER undergo post- translational modifications, folding and assembly prior to delivery to their final destination via vesicular transport (McCaffrey and Braakman, 2016; Vincenz- Donnelly and Hipp, 2017). As in the cytosol, folding is error-prone. Terminally misfolded proteins are recognised by the ER quality control system and targeted for degradation (Vincenz-Donnelly and Hipp, 2017). Degradation of misfolded ER proteins is more challenging than cytosolic proteins because there is a lack of degradation enzymes in the ER lumen and therefore, misfolded ER proteins are usually retrotranslocated back into the cytosol for degradation by the cytosolic proteasome. This pathway is termed ER-associated degradation (ERAD) and involves several overlapping stages (Christianson and Ye, 2014; Qi et al., 2017) (Figure 1.5). Factors involved in the initial recognition of misfolded proteins for ERAD include immunoglobulin heavy chain-binding protein (BiP), ER degradation- enhancing alpha-mannosidase-like protein 1 (EDEM1), protein OS-9 (OS9) and XTP3-transactivated gene B protein (XTP3B) (Christianson and Ye, 2014; Qi et al., 2017). Exactly how ERAD substrates are moved through the ER membrane is not known; it is proposed to involve a protein channel composed of multi-spanning proteins such as the transmembrane E3 ubiquitin ligases, the Sec61 complex, and/or other factors such as Derlin proteins (Christianson and Ye, 2014; Qi et al., 2017). In mammals, E3 ubiquitin-protein ligase synoviolin (Hrd1) and E3 ubiquitin-protein ligase AMFR (gp78) are two well characterised E3 ubiquitin ligases involved in ERAD. The active sites of both Hrd1 and gp78 are located at the cytoplasmic side of the ER membrane, positioned to ubiquitinate misfolded ER proteins upon retrotranslocation or dislocation (Christianson and Ye, 2014; Qi et al., 2017). The 15S Mg(2+)-ATPase p97 subunit (p97) can bind to both of these E3 ubiquitin ligases, and at the same time interacts with ubiquitinated ERAD substrates via its cofactors ubiquitin recognition factor in ER-associated degradation protein 1 (Ufd1) and

Figure 1.5: The mammalian endoplasmic reticulum-associated degradation. Misfolded ER protein is recognised by immunoglobulin heavy chain-binding protein (BiP), ER degradation-enhancing alpha-mannosidase-like protein 1 (EDEM1), protein OS-9 (OS9) and XTP3-transactivated gene B protein (XTP3B) and sent to the retrotranslocation site comprises of Derlin proteins, transmembrane E3 ubiquitin ligases and protein transport protein Sec61 (Sec61) complex. Retrotranslocated protein is ubiquitinated for 15S Mg(2+)-ATPase p97 subunit (p97) recognition and p97 extracts the endoplasmic reticulum-associated degradation substrate from the endoplasmic reticulum (ER) membrane. Hydrophobic domain is protected by shuttling factor en route to the proteasome for degradation. Proteasome contains stacked rings of heptameric proteolytic subunits (blue) and hexameric regulatory subunits (purple). Hrd1, E3 ubiquitin-protein ligase synoviolin; gp78, E3 ubiquitin- protein ligase AMFR; Ub, ubiquitin; (Adapted from Christianson and Ye, 2014)

28 nuclear protein localization protein 4 homolog (Npl4) (Christianson and Ye, 2014). p97 then uses the energy from ATP hydrolysis to extract ubiquitinated ERAD substrates from the ER membrane (Christianson and Ye, 2014). Protein may be fed directly into the proteasome as it is extracted from the ER, but for some ERAD substrates, retrotranslocation and dislocation can be uncoupled from degradation, meaning that the ERAD substrate may be extracted into the cytosol before being transferred into the proteasome (Brodsky, 2012; Ruggiano et al., 2014). Hence, membrane proteins undergoing ERAD may require special handling by ubiquitin- binding proteins en route to the proteasome. The majority of misfolded proteins are cleared from the ER by ERAD (Qi et al., 2017), but aggregates in the ER are sequestered into an ER-associated compartment and subsequently degraded by autophagy (Buchberger, 2014; Houck et al., 2014).

Even with the help of molecular chaperones, the ER folding capacity may be exceeded under conditions such as viral infection, glucose starvation or hypoxia, leading to build up of unfolded or misfolded proteins, a situation known as ER stress (Ryoo, 2016; Huang et al., 2017). The unfolded protein response (UPR) is a homeostatic mechanism to alleviate ER stress by reducing global translation, increasing expression of molecular chaperones and ERAD components, also lipid biosynthetic enzymes and trafficking factors (Frakes and Dillin, 2017; Lindholm et al., 2017). Mammalian cells possess three transmembrane ER stress sensors, the luminal domains of which detect the presence of misfolded proteins leading to activation of their cytoplasmic domains which mediate the UPR signalling pathways (Figure 1.6). PRKR-like endoplasmic reticulum kinase (PERK) is an ER transmembrane kinase that phosphorylates the eukaryotic translation initiation factor 2-alpha (eIF2α) resulting in translational attenuation to minimise newly synthesised proteins loaded into the ER (Harding et al., 1999). Phosphorylation of eIF2α also increases expression of certain proteins including the transcription factor cyclic AMP-dependent transcription factor ATF-4 (ATF4) which is responsible for amino acid metabolism to overcome low amino acid availability during ER stress (Harding et al., 2000). The second branch of the UPR pathway is initiated by cyclic AMP- dependent transcription factor ATF-6 alpha (ATF6-alpha) (Yoshida et al., 1998). Upon ER stress, ATF6-alpha is transported to the Golgi apparatus where it undergoes proteolytic cleavage by the site-1 and site-2 serine proteases. The cleaved N-terminus

A B C

Figure 1.6: The mammalian unfolded protein response signalling pathways. (A) PRKR-like endoplasmic reticulum kinase (PERK) is activated during endoplasmic reticulum (ER) stress by homo-oligomerisation and auto-phosphorylation (P). Activated PERK phosphorylates eukaryotic translation initiation factor 2-alpha (eIF2α) to reduce global translation but induce expression of cyclic AMP-dependent transcription factor ATF-4 (ATF4) important in amino acid metabolism. (B) Under conditions of ER stress, cyclic AMP-dependent transcription factor ATF-6 (ATF6) is transported to the Golgi apparatus and cleaved by the site-1 (SP1) and site-2 (SP2) serine proteases. ATF6-N terminus is an active transcription factor for chaperones and endoplasmic reticulum-associated degradation (ERAD) components. (C) Serine/threonine-protein kinase/endoribonuclease IRE1 (IRE1) homo-oligomerises and auto-phosphorylates upon ER stress, activating its endoribonuclease activity. X- box-binding protein 1 (XBP1) mRNA is cleaved by activated IRE1 to remove an intron and cleaved XBP1 mRNAs are ligated by tRNA-splicing ligase RtcB homolog (RtcB) for the production of XBP1s, a potent transcription factor for chaperones and ERAD components. HR2, hydrophobic domain 2. (Adapted from Hetz et al., 2013)

30 represents the active form enters the nucleus to act as a potent transcriptional activator of chaperone and ERAD component genes (Haze et al., 1999; Ye et al., 2000). The most conserved UPR pathway consists of serine/threonine-protein kinase/endoribonuclease IRE1 alpha (IRE1α), X-box-binding protein 1 (XBP1) and tRNA-splicing ligase RtcB homolog (RtcB). Upon detection of unfolded proteins, IRE1α homo-oligomerises and auto-phosphorylates, resulting in activation of its endoribonuclease. The IRE1 RNase then cleaves the XBP1 pre-mRNA as part of an unconventional splicing reaction, generating a spliced XBP mRNA which encodes for a potent transcription factor. IRE1 is located at the ER membrane, means that the unspliced XBP1 mRNA must be recruited to the ER membrane in order to undergo splicing and generate the active transcription factor. Due to the presence of a hydrophobic domain in the unspliced XBP1 protein and a specific peptide motif at the C-terminus that causes stalling of translation, signal recognition particle binds to the hydrophobic region and directs the XBP1u protein together with the ribosome and the mRNA to the ER, bringing the ribosome-nascent chain complex in contact with the activated IRE1α. Cleaved XBP1 mRNAs are ligated by RtcB for the production of spliced XBP1 protein (Yoshida et al., 2001; Yanagitani et al., 2009; Baltz et al., 2012; Lu et al., 2014). The main goal of the UPR is to restore homeostasis but prolonged or severe UPR induces cell death.

1.4 BAG6

Shielding of exposed hydrophobic domains and maintaining solubility of terminally misfolded substrates to prevent aggregation en route to the proteasome is a challenge for the protein quality control system. Even though heat shock proteins are able to bind and maintain the solubility of misfolded proteins, heat shock proteins do not interact directly with the proteasome, and therefore other proteasome- associating factors are needed to bridge the gap. BAG6 has a number of roles suggested, the best characterised of these involves the ability of BAG6 to bind and shield hydrophobic sequences in other proteins. BAG6 has also been shown to play a role in the ubiquitin-proteasome system. BAG6 has an ubiquitin-like domain that interacts with the ubiquitination machinery to promote substrate ubiquitination and degradation. Therefore, BAG6 is an important factor in the protein quality control

31 system and proteins having exposed hydrophobic stretches are potential substrates of BAG6 in the ubiquitin-proteasome system.

1.4.1 BAG6 structure

BAG6 is also known by the name HLA-B associated transcript 3 (BAT3) or scythe. The BAG6 gene is located at the short arm of human chromosome 6 within the major histocompatibility complex class III region (Wang and Liew, 1994; Ozaki et al., 1999). BAG6 contains a number of signature features: an ubiquitin-like domain at the N-terminus, proline-rich regions, a zinc finger-like domain, a nuclear localisation signal and a BAG domain at the C-terminus (Banerji et al., 1990; Wang and Liew, 1994; Lanning and Lafuse, 1997; Wu et al., 2004) (Figure 1.7). The ubiquitin-like domain shares structural similarity with ubiquitin, and is found in a wide range of proteins including those that function to shuttle ubiquitinated proteins to the proteasome, E3 ubiquitin ligases and deubiquitinating enzymes (Faesen et al., 2012; Husnjak and Dikic, 2012). The BAG domain is an evolutionarily conserved domain that defines the BAG-family of proteins (BAG1 to BAG6) (Behl, 2016). The BAG domain acts as nucleotide exchange factor for Hsp70, stimulating exchange of ADP for ATP, thus promoting substrate release from Hsp70 (Behl, 2016). However, the BAG6-BAG domain is shorter (56 residues compared to 76–112 residues in other BAG domains), and more than half of the residues important for heat shock protein interaction are not conserved. Furthermore, the crystal structure of the BAG6-BAG domain shows that these Hsp70 binding residues have different orientations to canonical BAG domain (Mock et al., 2015; Kuwabara et al., 2015). Even though BAG6 still binds Hsp70 through its BAG domain, the interaction is less efficient and BAG6 in fact inhibits Hsp70-mediated protein refolding (Thress et al., 2001; Wang et al., 2011; Kuwabara et al., 2015). Thus, BAG6-BAG domain is a non-canonical BAG domain. BAG6 was shown to initiate Hsp70 activation upon heat shock by stabilising Hsp70; however, accumulation of Hsp70 eventually leads to degradation of BAG6 (Corduan et al., 2009). Downregulation of BAG6 is crucial to facilitate protein refolding by Hsp70 (Corduan et al., 2009; Wang et al., 2011). The E3 ubiquitin ligase E3 ubiquitin-protein ligase CHIP (CHIP) was shown to bind BAG6 to prevent binding and degrading Hsp70 (Corduan et al., 2009). Once BAG6 is

Figure 1.7: Schematic representation of human BAG6 protein domains and its putative binding partners. A stable ternary complex is formed from BCL2- associated athanogene 6 (BAG6) (core component), ubiquitin-like protein 4A (Ubl4A) and transmembrane domain recognition complex 35 kDa subunit (TRC35). The direct interaction between Ubl4A and BAG6 is at the BAG domain while TRC35 binds to the nuclear localisation signal (NLS). Ubl4A can also bind to small glutamine-rich tetratricopeptide repeat-containing protein alpha (SGTA) when SGTA engages BAG6 at the ubiquitin-like (UBL) domain. The numbers denote respective amino acid positions. PR, proline-rich; ZF, zinc finger-like; Hsp70, heat shock 70 kDa protein. (Adapted from Lee and Ye, 2013)

33 degraded due to Hsp70 accumulation, CHIP is released to counteract on Hsp70 during cell recovery.

Despite the presence of a nuclear localisation signal (NLS), BAG6 is localised to the cytoplasm as well as the nucleus (Kwak et al., 2008). The reason for this dual localisation is that BAG6 is engaged in a stable ternary complex with two other proteins, ubiquitin-like protein 4A (Ubl4A) and transmembrane domain recognition complex 35 kDa subunit (TRC35), which masks its NLS (Mariappan et al., 2010). Within the complex, Ubl4A binds to BAG6 at the BAG domain while TRC35 interacts through the C-terminus, masking the NLS (Kawahara et al., 2013) (Figure 1.7). The BAG6-BAG domain is necessary and sufficient to bind Ubl4A, and furthermore Ubl4A cannot bind to the BAG domain of all the other BAG-family members, demonstrating the specificity of the interaction (Kuwabara et al., 2015). The BAG6-BAG domain does not acquire a stable secondary structure unless it tightly binds to Ubl4A into a stoichiometric complex (Kuwabara et al., 2015; Mock et al., 2015). Therefore, it is speculated that the primary role of the non-canonical BAG6-BAG domain is to promote hetero-dimerisation with Ubl4A (Kuwabara et al., 2015; Mock et al., 2015).

1.4.2 BAG6 in tail-anchored protein targeting

Membrane protein synthesis, targeting and insertion have to be tightly regulated as transmembrane domains are highly hydrophobic and readily aggregate in the aqueous environment of the cytosol. Tail-anchored proteins represent 3 to 5 % of total membrane proteins and they are not normally targeted co-translationally because their transmembrane domains are located at the C-terminus (Hegde and Keenan, 2011; Casson et al., 2016). Translation termination occurs before the transmembrane domain exits the ribosome exit tunnel, hence newly synthesised TA protein is released into the cytoplasm without engaging SRP, and TA protein needs other cytosolic factors for shielding the hydrophobic domain and targeting to the ER membrane (Hegde and Keenan, 2011). TRC40 interacts with protein transport protein Sec61 subunit beta (Sec61β) through the transmembrane domain and promotes membrane targeting of Sec61β (Stefanovic and Hegde, 2007). Later, components of the yeast guided entry of tail-anchor (GET) pathway for TA protein

34 membrane insertion are illustrated, including small glutamine-rich tetratricopeptide repeat-containing protein 2 (Sgt2) and Get1 to Get5, where Get3 is the yeast homolog of TRC40 (Schuldiner et al., 2008; Chang et al., 2010) (Figure 1.8). Sgt2 recruits TA protein and by interacting with Get4 and Get5, handovers the TA protein to Get3 which brings the TA protein to Get1 and Get2 that act as receptors at the ER membrane (Schuldiner et al., 2008; Chang et al., 2010). BAG6 is identified as a novel binding partner of Sec61β and BAG6 assists in TA protein insertion with TRC35 (yeast Get4) and Ubl4A (yeast Get5) (Leznicki et al., 2010; Mariappan et al., 2010; Hessa et al., 2011) (Figure 1.8). Even though BAG6 does not have a yeast homolog, BAG6 fits into the GET pathway through interaction between components of the BAG6 complex and to small glutamine-rich tetratricopeptide repeat-containing protein alpha (SGTA) (Winnefeld et al., 2006; Leznicki et al., 2010). This targeting pathway ensures shielding of the C-terminal transmembrane domain throughout the transport from the ribosome to the BAG6 complex, then the TRC40 until finally reaching the ER membrane.

1.4.3 BAG6 in degradation of mislocalised proteins

If targeting fails, membrane proteins become mislocalised to the cytoplasm and their hydrophobic domains are prone to aggregate and engage in inappropriate interactions in the aqueous environment. This is potentially harmful to cells and therefore factors that recognise, shield and degrade mislocalised proteins are important in maintaining cellular homeostasis (Kang et al., 2006; Rane et al., 2008). A number of studies over the past 7 years have implicated BAG6 in handling of mislocalised proteins, by recognising and shielding exposed hydrophobic domains in mislocalised proteins, and recruiting E3 ubiquitin ligase to promote substrate ubiquitination. A model mislocalised protein with two transmembrane domains was found to require BAG6 for efficient ubiquitination, and E3 ubiquitin-protein ligase RNF126 (RNF126) was found to be the key BAG6-dependent E3 ubiquitin ligase involved (Rodrigo-Brenni et al., 2014). RNF126 binds to the BAG6 ubiquitin-like domain and ubiquitinates BAG6 substrate on lysine residues adjacent to the hydrophobic domain (Rodrigo-Brenni et al., 2014) (Figure 1.9). BAG6 is also involved in the proteasome-mediated degradation of misfolded proteins and defective

Figure 1.8: The yeast and mammalian guided entry of tail-anchor pathway components. In yeast, small glutamine-rich tetratricopeptide repeat-containing protein 2 (Sgt2) transfers tail-anchored protein to guided entry of tail-anchor 3 (Get3) with the help of Get4 and Get5, the tail-anchored protein is then inserted into the endoplasmic reticulum (ER) membrane through Get1 and Get2. In mammalian cells, all components are conserved except BCL2-associated athanogene 6 (BAG6) that is only found in higher organism. BAG6 binds to transmembrane domain recognition complex 35 kDa subunit (TRC35/Get4) and ubiquitin-like protein 4A (Ubl4A/Get5) and facilitates tail-anchored protein insertion into the ER membrane. TMD, transmembrane domain; SGTA, small glutamine-rich tetratricopeptide repeat- containing protein alpha; CAML, calcium-modulating cyclophilin ligand; WRB, tail- anchored protein insertion receptor WRB. (Adapted from Hegde and Keenan, 2011)

Figure 1.9: Model illustrating BAG6 role in protein quality control in the cytosol. Mislocalised proteins, misfolded proteins and defective ribosomal products having exposed hydrophobic regions are sequestered by BCL-2 associated athanogene 6 (BAG6). BAG6 then recruits E3 ubiquitin ligase to promote substrate ubiquitination (Ub) for proteolysis degradation at the proteasome. Proteasome contains stacked rings of heptameric proteolytic subunits (blue) and hexameric regulatory subunits (purple). (Adapted from Kawahara et al., 2013)

37 ribosomal products that arise in the cytosol (Minami et al., 2010). Defective ribosomal products include polypeptides generated by premature termination and those that fail to fold or assemble with partner subunits during or soon after synthesis. BAG6 does not discriminate its substrates based on sequence or misfolding per se but overall hydrophobicity (Hessa et al., 2011; Tanaka et al., 2016). Prion protein truncated at both the N- and C-terminus is poorly ubiquitinated despite misfolded but prion proteins containing signal sequence from preprolactin and neuropeptide are efficiently ubiquitinated due to the unprocessed hydrophobic signals (Hessa et al., 2011). BAG6-CL1 (a model misfolded protein) interaction was also maintained in CL1 mutant having the same hydrophobic amino acids arranging in a different way (Tanaka et al., 2016). Therefore, TA proteins or other membrane proteins are not substrates of BAG6 for subsequent ubiquitination unless they are released into the cytosol inappropriately during their delivery to the ER with unprocessed or non- inserted long linear hydrophobic stretches, which differentiate them from cytosolic proteins. A study has suggested that even modest reduction of hydrophobicity impairs BAG6 recognition and binding (Mariappan et al., 2010). This is in contrast to classical molecular chaperones such as Hsp70 which binds substrates with a region of moderate hydrophobicity present in most cytosolic proteins (Hessa et al., 2011). BAG6 physically associates with the proteasome and the BAG6 N-terminus binds polyubiquitinated substrates which are mostly cycloheximide-sensitive newly synthesised polypeptides (Minami et al., 2010). Taken together, BAG6 couples synthesis and degradation by handing newly synthesised polyubiquitinated polypeptides directly to the proteasome for degradation.

Conditions of ER stress caused by accumulation of misfolded proteins in the ER have been shown to reduce the efficiency of co-translational protein translocation into the ER (Kang et al., 2006). This is thought to occur due to sequestration of the chaperone BiP, which assists polypeptide translocation into the ER lumen, by unfolded proteins in the ER lumen (Kang et al., 2006). As a result of the reduced availability of BiP, proteins targeted to the ER via SRP fail to be efficiently translocated across the ER membrane. Instead, the translated polypeptides are ubiquitinated and rerouted to the proteasome for degradation without entering the ER. This protein quality control pathway is termed ER pre-emptive quality control (ER pQC) (Kang et al., 2006; Kadowaki et al., 2015). A role for BAG6 in pQC was identified by experiments showing that siRNA mediated knockdown stabilised the cleaved form of a soluble secretory protein transthyretin (ERAD substrate) and also the uncleaved form (pQC substrate) (Kadowaki et al., 2015). The Derlin family of proteins which interact with SRP and SRP receptor are proposed to function upstream of BAG6 in this pathway, receiving pQC substrates from SRP and relaying them to p97/BAG6 (Kadowaki et al., 2015) (Figure 1.10).

1.4.4 BAG6 triages targeting and degradation

Mislocalised or misfolded proteins need to be promptly degraded as they tend to aggregate among themselves or form non-specific interactions with important factors in the cytosol, both of which is deleterious to cells. However, premature degradation of newly synthesised proteins causes resource wastage and proteins may not reach designated location for their cellular functions. Therefore, targeting and degradation have to be carefully regulated and coordinated. Being involved in both targeting and degradation, understanding the mechanism behind the triage decision made by BAG6 towards its substrates is of great interest. BAG6 triage mechanism was studied in-vitro using a model tail-anchored protein together with important factors in the BAG6-dependent targeting and BAG6-dependent degradation pathways including BAG6, SGTA, TRC40 and RNF126 (Shao et al., 2017). In the model proposed by this study, SGTA acts upstream of BAG6 and TRC40 to bind TA proteins released from the ribosome, and at this point the TA protein’s fate is undecided (Chang et al., 2010; Shao et al., 2017). TA protein transfer from SGTA to TRC40 only requires the minimal BAG6 complex (C-terminus) where SGTA interacts with Ubl4A and TRC35 binds to TRC40 (Shao et al., 2017), agreeing with earlier report in yeast (Mock et al., 2015). On the other hand, N-terminus of BAG6 interacts with RNF126 for TA protein ubiquitination (Rodrigo-Brenni et al., 2014; Shao et al., 2017). These two observations lead to a predicted targeting module involving SGTA, cBAG6 and TRC40 and a degradation module comprises of SGTA, nBAG6 and RNF126 (Shao et al., 2017) (Figure 1.11). SGTA binds to either BAG6 (nBAG6) or Ubl4A (cBAG6) at a given time but not simultaneously (Leznicki et al., 2013), agreeing with the triage model. Even though BAG6 promotes membrane integration of Sec61β through the TRC40 pathway, BAG6 and TRC40 interact with

Figure 1.10: Model illustrating BAG6 role in endoplasmic reticulum pre- emptive quality control. Under conditions of endoplasmic reticulum (ER) stress, translocation is attenuated. Derlin protein interacts with signal recognition particle (SRP) and SRP receptor (SRPR) to capture nascent polypeptide to be handed over to 15S Mg(2+)-ATPase p97 subunit (p97) and BCL-2 associated athanogene 6 (BAG6) for degradation through the proteasome. Proteasome contains stacked rings of heptameric proteolytic subunits (blue) and hexameric regulatory subunits (purple). Ub, ubiquitin. (Adapted from Kadowaki et al., 2015)

Figure 1.11: Current model illustrating BAG6 triages protein targeting and degradation. Small glutamine-rich tetratricopeptide repeat-containing protein alpha (SGTA) associates with tail-anchored protein and either transfers the tail-anchored protein to BCL-2 associated athanogene 6 (BAG6) C-terminus targeting module or BAG6 N-terminus degradation module. Tail-anchored protein is readily associated and dissociated from SGTA, followed by BAG6 N-terminus then BAG6 C-terminus. Transfer of tail-anchored protein to the targeting module is not competed by BAG6 N-terminus while BAG6 N-terminus captures tail-anchored protein dissociated from SGTA much slower than SGTA, indicating that tail-anchored protein will have more targeting attempts before being degraded. Proteasome contains stacked rings of heptameric proteolytic subunits (blue) and hexameric regulatory subunits (purple). TMD, transmembrane domain; Ubl4A, ubiquitin-like protein 4A; TRC35, transmembrane domain recognition complex 35 kDa subunit; TRC40, transmembrane domain recognition complex 40 kDa ATPase subunit; ER, endoplasmic reticulum; CAML, calcium-modulating cyclophilin ligand; WRB, tail- anchored protein insertion receptor; RNF126, E3 ubiquitin-protein ligase RNF126; Ub, ubiquitin.

41 distinct populations of Sec61β (Leznicki et al., 2010), again supported BAG6 triage model proposed. When TRC40 is present together with the degradation module, only low level of TA protein ubiquitination is observed (Shao et al., 2017), opposed to an earlier work showing TRC40 has little effect on ubiquitination status of Sec61ß (Leznicki and High, 2012). Competitive captured assay reveals strongest preference of TA proteins for SGTA followed by nBAG6 then lastly TRC40 (Shao et al., 2017). However, TA proteins also most readily dissociate from SGTA and least from TRC40 (Shao et al., 2017). The rate of association and dissociation between the different factors has been observed as the key to the triage reaction (Shao et al., 2017). TA protein transfer from SGTA to TRC40 is rapid and private (not competed by nBAG6), while nBAG6 captures TA protein dissociated from SGTA much slower, indicating that TA protein may be recaptured by SGTA for more targeting attempts before it is committed to degradation (Shao et al., 2017) (Figure 1.11). Modified TA protein with an extra transmembrane domain was shown to interact with BAG6 more readily than TRC40 (Hessa et al., 2011), and this observation is still unable to explain with the current model.

BAG6-dependent ubiquitination is reversible through the action of SGTA (Leznicki and High, 2012; Wunderley et al., 2014). SGTA does not act passively by masking BAG6 recognition site on substrate but actively promotes deubiquitination of substrates that have undergone ubiquitination; this deubiquitination activity requires SGTA interaction with BAG6 (Leznicki and High, 2012; Wunderley et al., 2014). Therefore, the fate of mislocalised or misfolded proteins is predicted to depend upon competition between factors promoting ubiquitination (E2 and E3 enzymes) and factors promoting deubiquitination (SGTA) for the ubiquitin-like domain of BAG6 (Leznicki et al., 2013). This prediction is in agreement with a structural study showing that RNF126 (E3 enzyme) and SGTA have overlapping binding sites at the N-terminus of BAG6 and that the affinity of RNF126 binding to BAG6 is stronger than that of SGTA (Krysztofinska et al., 2016). RNF126 can also bind with lower affinity to the ubiquitin-like domain of Ubl4A in the BAG6 complex but the affinity of RNF126 to Ubl4A is rather low compared to its binding to BAG6, as opposed to the case for SGTA (Krysztofinska et al., 2016). In addition, SGTA was discovered to bind Rpn13, the ubiquitin receptor of the proteasome regulatory subunit (Leznicki et al., 2015) while Scythe (Xenopus BAG6) was shown to bind

Xenopus Rpn10 (Kikukawa et al., 2005). Since a polyubiquitin chain can bind both Rpn13 and Rpn10 simultaneously (Bhattacharyya et al., 2014), Rpn13-bound SGTA and Rpn10-bound BAG6 may compete with each other and control substrate access to the proteasome and hence determine the fate of the substrate (Leznicki et al., 2013).

1.4.5 BAG6 in endoplasmic reticulum-associated degradation

BAG6 was shown to interact with ERAD components Derlin-2, gp78 and UBX domain-containing protein 8 (UBXD8) (Claessen and Ploegh, 2011; Wang et al., 2011; Xu et al., 2013). BAG6 was also found to bind to the model ERAD substrate T-cell receptor alpha chain (TCRα) and TCRα degradation was delayed in cells depleted of BAG6, suggesting that BAG6 has a role in ERAD (Claessen and Ploegh, 2011; Wang et al., 2011). Similar to the effect seen with BAG6 knockdown, stabilisation of TCRα was observed when Ubl4A and TRC35 are knockdown, showing that the whole BAG6 complex is needed to facilitate ERAD (Wang et al., 2011). TRC35 keeps BAG6 in the cytosol by masking BAG6 nuclear localisation signal but the role of Ubl4A is still unclear (Wang et al., 2011). The BAG6 proline- rich region promotes the formation of BAG6 oligomers, allowing the ubiquitin-like domains to form multivalent interactions simultaneously with two ERAD components, the E3 ligase gp78 and UBXD8, so that BAG6 is recruited to the ER membrane where it is positioned to interact with ERAD substrates as they are extracted from the membrane by the action of p97 (Wang et al., 2011; Xu et al., 2013). The BAG6 oligomer is predicted to contain disordered segments which are used to bind a wide range of hydrophobic substrates (Xu et al., 2013). BAG6 is thought to act downstream of p97 (Wang et al., 2011) and act as a ‘holdase’ to minimise aggregation and maintain the substrate in unfolded state accessible for proteasome action (Xu et al., 2013) (Figure 1.12). With another model ERAD substrate, the multi-spanning membrane protein opsin-degron, delayed degradation was observed not only under BAG6 knockdown but also under BAG6 overexpression (Payapilly and High, 2014). Exogenously expressed BAG6 perturbs the stoichiometry of endogenous BAG6 complex and by this means disrupts BAG6 function (Payapilly and High, 2014).

Figure 1.12: Model illustrating BAG6 role in endoplasmic reticulum-associated degradation. Misfolded endoplasmic reticulum (ER) protein is ubiquitinated (Ub) by E3 ubiquitin-protein ligase AMFR (gp78) and retrotranslocated with the help of 15S Mg(2+)-ATPase p97 subunit (p97) into the cytosol. BCL-2 associated athanogene 6 (BAG6) forms a homo-oligomer so that BAG6 N-terminus can interact with gp78 and UBX domain-containing protein 8 (UBXD8) simultaneously. BAG6 then escorts the endoplasmic reticulum-associated degradation substrate to proteasome for degradation. Proteasome contains stacked rings of heptameric proteolytic subunits (blue) and hexameric regulatory subunits (purple). (Adapted from Lee and Ye, 2013)

In summary, many studies provide evidence that BAG6 plays a role in a range of protein quality control pathways, using its ability to identify and bind proteins in the cytosol that expose stretches of hydrophobic residues, including mislocalised proteins, misfolded proteins, defective ribosomal products, ER- associated degradation and ER pre-emptive quality control substrates. It is interesting to note that there are some differences in the way BAG6 functions to promote degradation of these different proteins. For example, the sequence in which BAG6 and E3 ubiquitin ligases function appears to be different in cytosolic and ER quality control. In the cytosolic quality control system, BAG6 binds to misfolded proteins and helps to ensure efficient ubiquitination by recruiting E3 ligases whilst BAG6 appears to act downstream of the E3 enzymes in ERAD.

1.5 Aims and objectives of study

Failure of protein quality control pathways can lead to harmful accumulation of aggregation-prone proteins, as occurs in many human diseases. Defining the range of quality control systems that cells use to handle such aggregation-prone proteins is vital for understanding the molecular basis of these diseases and the development of new treatments. Even though BAG6 is now firmly implicated in handling of different types of hydrophobic domains in the cytosol, the co-factors involved in BAG6- mediated pathways, such as E3 ubiquitin ligases and deubiquitinases, are not well defined. Furthermore, because studies so far have relied heavily on individual model substrates, endogenous or physiological substrates of BAG6 have not been defined, and we therefore have little understanding of the range of cellular substrates of BAG6. Hence, overarching aims of this project were to identify endogenous BAG6 substrates and co-factors. Two approaches were taken to address these aims:

1. Targeted approach a. To examine the role of UBR4 (co-factor) in BAG6-mediated ER- associated degradation (Chapter 3) b. To test whether XBP1 is a novel endogenous BAG6 substrate (Chapter 4) 2. Unbiased approach a. To identify endogenous BAG6-interacting proteins with BioID (Chapter 5) b. To predict determinant of BAG6 substrate specificity with bioinformatics analysis (Chapter 5)

2 Materials and methods

2.1 DH5α competent cells preparation

DH5α from glycerol stock was streaked onto LB agar (1 % w/v NaCl, 1 % w/v tryptone, 0.5 % w/v yeast extract, 1 % w/v agar) and incubated inverted at 37 °C overnight. A single colony was inoculated into 5 mL LB broth (1 % w/v NaCl, 1 % w/v tryptone, 0.5 % w/v yeast extract) and incubated at 37 °C overnight in an incubator shaker. Overnight bacteria culture was then inoculated into 100 mL SOB broth (2 % w/v tryptone, 0.5 % w/v yeast extract, 10 mM NaCl, 2.5 mM KCl, 10 mM

MgCl2, 10 mM MgSO4) and incubated at 37 °C with vigorous shaking until the optical density at the wavelength of 595 nm (O.D.595) reaches 0.4. Bacterial cells were transferred to pre-chilled 50 mL centrifuge tube and centrifuged at 4000 rpm for 10 minutes at 4 °C. Bacterial cells suspended in TFB1 buffer (10 mM RbCl, 50 mM MnCl2, 30 mM KOAc, 10 mM CaCl2, 15 % v/v glycerol) were incubated on ice for 30 minutes, centrifuged again at 4000 rpm for 10 minutes at 4 °C and re- suspended in TFB2 buffer (10 mM MOPS pH 7, 10 mM RbCl, 75 mM CaCl2, 15 % v/v glycerol). Competent cells were aliquoted into pre-chilled microfuge tubes and snap-frozen in liquid nitrogen before being stored at - 80 °C.

2.2 Plasmid DNA preparation

Plasmid constructs (Table 2.1) and DH5α competent cells were mixed and incubated on ice for 30 minutes then heat shocked at 42 °C in a water bath for 45 seconds before placing back on ice to recover for 5 minutes. LB broth (1 % w/v NaCl, 1 % w/v tryptone, 0.5 % w/v yeast extract) was added to transformation mixture and incubated at 37 °C for 30 minutes in an incubator shaker. Transformation mixture was spread on LB agar (1 % w/v NaCl, 1 % w/v tryptone, 0.5 % w/v yeast extract, 1 % w/v agar) containing either ampicillin (100 µg/mL) or kanamycin (50 µg/mL), both from Melford, based on resistance gene present in the plasmid transformed. LB agar plate was incubated inverted at 37 °C overnight. The next day, a single colony was inoculated into LB Broth with 100 µg/mL ampicillin or 50 µg/mL kanamycin and incubated at 37 °C with constant shaking (225 rpm) overnight. Plasmid purification

47 was done using QIAprep® Spin Miniprep Kit (Qiagen) according to manufacturer’s instruction. Plasmid DNA concentration and 260/280 nm ratio were measured using NanoDrop 1000 Spectrophotometer (Thermo Scientific).

Table 2.1: Plasmid constructs used in this study.

Plasmid construct Source BAG6-V5- Aishwarya Payapilly, pcDNA5/FRT/TO University of Manchester BAG6(∆N)-V5- Aishwarya Payapilly, pcDNA5/FRT/TO University of Manchester BAG6mNLS-V5- Aishwarya Payapilly, pcDNA5/FRT/TO University of Manchester Myc-BioID- Martin Lowe, University of Manchester pcDNA3.1 Myc-BirA- Constructed in this study pcDNA5/FRT/TO BAG6-myc-BirA- Constructed in this study pcDNA5/FRT/TO BAG6(∆N)-myc-BirA- Constructed in this study pcDNA5/FRT/TO Op91- Aishwarya Payapilly, pcDNA5/FRT/TO University of Manchester HA-Ubiquitin Lisa Swanton, University of Manchester HA3-XBP1u G519C- Martin Pool, University of Manchester pcDNA3.1(+) HA3-XBP1u Constructed in this study G519C(∆HR2)- pcDNA3.1(+) HA3- XBP1u Martin Pool, University of Manchester G519C(W256A)- pcDNA3.1(+) F3H8- XBP1u G519C-HA- Kenji Kohno, pcDNA3.1(+) Nara Institute of Science and Technology, Japan

2.3 Plasmid construction

2.3.1 Myc-BirA-pcDNA5/FRT/TO

Myc-BirA was subcloned from pcDNA3.1 (Table 2.1) to pcDNA5/FRT/TO using KOD Hot Start DNA Polymerase (Novagen) according to manufacturer’s instruction with slight modification. PCR mixture was prepared as shown in Table 2.2A, with forward primer 5’-CCGCTCGAGGAACAAAAACTCATC-3’ and reverse primer 5’-ATTGGGCCCTCACTTCTCTGCGCTTCT-3’. PCR cycling condition used was shown in Table 2.2B and PCR reaction was run in the thermo cycler Techne. PCR product was incubated with 1 µL DpnI (New England BioLabs) at 37 °C for 8 hours to digest all template DNAs. A small portion of the DpnI-treated PCR product was added with Sample Loading Buffer (Bioline), loaded onto 0.8 % agarose gel with SafeView Nucleic Acid Stain (NBS Biologicals Ltd) and agarose gel electrophoresis was performed at 100 V for 30 minutes in TAE buffer (40 mM Tris, 20 mM acetic acid, 1 mM EDTA pH 8) to check the PCR reaction. DpnI-treated PCR product was first purified by QIAquick® PCR Purification Kit (Qiagen) and then both vector and insert were restriction digested with compatible restriction enzyme pair in the presence of CutSmart™ Buffer (New England BioLabs) at 37 °C for 8 hours followed by 65 °C for 20 minutes. Restriction digested products were run in 0.8 % agarose gel at 100 v for 45 minutes and bands of interest were carefully excised from the agarose gel into Ultrafree®-MC-HV centrifugal filters (Merck). Tubes were centrifuged at maximum speed for 5 minutes and flow-through was purified once again by QIAquick® PCR Purification Kit. Vector was dephosphorylated with Shrimp Alkaline Phosphatase (New England BioLabs) at 37 °C for 4 hours followed by deactivation at 65 °C for 30 minutes. A molar ratio of 1:3 vector to insert were ligated with Quick Ligase (New England BioLabs) for 10 minutes at room temperature before transformation (Section 2.2). Purified plasmid was sent to Source BioScience for Sanger Sequencing. Positive clone was stored as glycerol stock at - 80 °C for long term usage.

Table 2.2: (A) Standard reaction setup and (B) PCR cycling condition for site- directed mutagenesis using KOD Hot Start DNA Polymerase.

A KOD buffer (10X) 5 MgSO4 (25 mM) 3 dNTPs (2 mM each) 5 PCR grade water 32 Sense (5’) primer (20 µM) 0.75 Antisense (3’) primer (20 µM) 0.75 Template DNA (10 ng/µL) 2.5 KOD Hot Start DNA Polymerase (1 U/µL) 1 Total reaction volume 50

B Step Temperature (°C) Time (m:s) Heated lid 105 Polymerase activation 95 2:00 Denaturation 95 0:20 Number of cycle Annealing Lowest primer Tm 0:30 30 Extension 70 2:00 Hold 4

To add a start codon to myc-BirA-pcDNA5/FRT/TO, site-directed mutagenesis was performed using KOD Hot Start DNA Polymerase (Novagen®) PCR mixture was prepared as shown in Table 2.2A on ice, with forward primer 5’- TGGCGGCCGCTCGAGATGGAACAAAAACTCATC-3’ and reverse primer 5’- GATGAGTTTTTGTTCCATCTCGAGCGGCCGCCA-3’. PCR cycling condition used was shown in Table 2.3 and PCR reaction was run in the thermo cycler Techne. PCR product was incubated with 1 µL DpnI (New England BioLabs) at 37 °C for 8 hours to digest all template DNAs. A small portion of the DpnI-treated PCR product was added with Sample Loading Buffer (Bioline), loaded onto 0.8 % agarose gel with SafeView Nucleic Acid Stain (NBS Biologicals Ltd) and agarose gel electrophoresis was performed at 100 V for 30 minutes in TAE buffer (40 mM Tris, 20 mM acetic acid, 1 mM EDTA pH 8) to check the PCR reaction. Remaining DpnI-treated PCR product was transformed as in Section 2.2 and purified plasmid was sent to Source BioScience for sequencing.

Table 2.3: PCR cycling condition for site-directed mutagenesis using KOD Hot Start DNA Polymerase.

Step Temperature (°C) Time (m:s) Heated lid 105 Polymerase activation 95 2:00 Denaturation 95 0:20 Number of cycle Annealing Lowest primer Tm 0:30 25 Extension 70 6:00 Hold 4

2.3.2 BAG6-myc-BirA-pcDNA5/FRT/TO

Subcloning for BAG6-myc-BirA-pcDNA5/FRT/TO was performed using KOD Hot Start DNA Polymerase (Novagen) according to manufacturer’s instruction with slight modification. PCR mixture was prepared as shown in Table 2.2A, with forward primer 5’-CCGCTTAAGATGGAGCCTAATGATAG-3’ reverse primer 5’- GGACTCGAGAGGATCATCAGCAAAG-3’ and BAG6-V5-pcDNA5/FRT/TO as template DNA. PCR cycling condition used was shown in Table 2.2B and PCR reaction was run in the thermo cycler Techne. Downstream processing was performed as described in Section 2.3.1.

2.3.3 BAG6(∆N)-myc-BirA-pcDNA5/FRT/TO

Subcloning for BAG6(∆N)-myc-BirA-pcDNA5/FRT/TO was performed using KOD Hot Start DNA Polymerase (Novagen) according to manufacturer’s instruction with slight modification. PCR mixture was prepared as shown in Table 2.2A, with forward primer 5’-TTACTTAAGATGCAACCGCAGCACAG-3’ reverse primer 5’-GGACTCGAGAGGATCATCAGCAAAG-3’ and BAG6-V5-pcDNA5/FRT/TO as template DNA. PCR cycling condition used was shown in Table 2.2B and PCR reaction was run in the thermo cycler Techne. Downstream processing was performed as described in Section 2.3.1.

2.3.4 HA3-XBP1u G519C(∆HR2)-pcDNA3.1(+)

For mutagenesis involving deletion of the hydrophobic region HR2 from

HA3-XBP1u-pcDNA3.1(+) (Table 2.1), KOD Hot Start DNA Polymerase (Novagen) was used according to manufacturer’s instruction with slight modification. PCR mixture was prepared as shown in Table 2.2A, with forward primer 5’- [Phos]ACAACTTGGACCCAGTCATGTTCTTCAAAT-3’ (primer sequence starts straight after the deleted region) and reverse primer 5’- [Phos]GTTCTGGAGGGGTGACAACTGGGCCTGCAC-3’ (primer sequence ends directly before the deleted region). PCR cycling condition used was shown in Table 2.2B and PCR reaction was run in the thermo cycler Techne. DpnI-treated PCR product was purified by QIAquick® PCR Purification Kit (Qiagen), ligated with Quick Ligase (New England BioLabs) for 10 minutes at room temperature before transformation (Section 2.2) and purified plasmid was sent to Source BioScience for sequencing.

2.4 Cell culture

All culture media were purchased from Sigma-Aldrich® unless specified otherwise. HeLa cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10 % v/v Fetal Bovine Serum (Biowest), 2 mM L- Glutamine solution and 0.1 mM MEM Non-essential Amino Acid solution. Cells were maintained at 37 °C with 5 % CO2. When cells reached 80 to 90 % confluency, cells were rinsed with Dulbecco’s Phosphate Buffered Saline, detached with Trypsin- EDTA solution and seeded at the density required for experiments or at 1:10 for maintenance. Cells were counted using Bright-Line® Hemacytometer (Reichert) after trypan blue staining. Cells were usually seeded into 12-well culture plate at 2.5 x 104 cells per cm2 and scaled-up linearly when necessary.

2.5 Stable inducible cell line generation and induction

Flp-In™ T-REx™ HeLa cell line (Stephen Taylor, University of Manchester) was seeded at 2.5 x 104 cells per cm2 into 6-well culture plate to reach 80 % confluency on the following day for transfection. For each well, 100 µL OPTI- MEM® I (Gibco) was mixed with 2700 ng pOG44, 300 ng DNA of interest in pcDNA™5/FRT/TO, 3 µL Plus™ reagent (Invitrogen) and incubated for 5 minutes before adding into a separate tube containing 100 µL OPTI-MEM® I (Gibco) with 7.5 µL Lipofectamine® LTX reagent (Invitrogen). Transfection mixture was mixed and incubated for 20 minutes then added drop-wise onto cells in 1 mL fresh complete medium. On the next day, cells were rinsed, trypsinised and transferred to 10 cm culture dish with selective media (complete media added with 200 µg/mL Hygromycin B from Invitrogen and 4 µg/mL Blasticidin S from InvivoGen). The medium was replaced every two days until cell death ceased and significant numbers of viable cells were observed (approximately 2 to 3 weeks). Cells were then split, maintained and seeded for experiment as described in Section 2.4. Different concentrations of tetracycline (Sigma) (up to 1000 ng/mL) were added to cells to induce expression of the protein of interest.

2.6 siRNA transfection

Double-stranded siRNA (Table 2.4) transfection was performed using INTERFERin® siRNA Transfection Reagent (Polyplus). HeLa cells were seeded at 1.25 x 104 cells per cm2 into 12-well culture plate to reach about 50 % confluency for siRNA transfection the following day. Transfection was performed following the manufacturer’s guidelines with slight modification. 0.6 µL siRNA stocks (20 µM) were added to 100 µL OPTI-MEM® I (Gibco), then 3 µL INTERFERin® was added, and the mixture was vortexed and incubated at room temperature for 10 minutes. 100 µL of the transfection mixture was added drop-wise onto cells in 0.5 mL fresh complete media. Cells were harvested for analysis after 24 to 72 hours based on specific experimental setup.

Table 2.4: siRNA sequences used in this study.

siRNA target siRNA sequence Non-targeting 5’- GUAUGAAGGUGGCCGUGAAUU-3’ 3’-UUCAUACUUCCACCGGCACUU -5’ (Dharmacon) BAG6 3’ UTR 5’- GCUCUAUGGCCCUUCCUCAUU-3’ 3’-UUCGAGAUACCGGGAAGGAGU -5’ (Dharmacon) BAG6 CDS 5’- CAGCUCCGGUCUGAUAUACAAUU-3’ 3’-UUGUCGAGGCCAGACUAUAUGUU -5’ (Aishwarya Payapilly, University of Manchester) UBR4 5’- CUACGAAGCUGCCGACAAAUU-3’ 3’-UUGAUGCUUCGACGGCUGUUU -5’ (Dharmacon) RNF126 5′- CCGGATTATATCTGTCCAAGAUU-3′ 3’-UUGGCCUAAUAUAGACAGGUUCU -5’ (Aishwarya Payapilly, University of Manchester)

2.7 Transient transfection for immunoblotting and immunofluorescence microscopy

HeLa cells were seeded at 2.5 x 104 cells per cm2 into 12-well culture plate to reach around 80 % confluency for plasmid DNA transfection the following day. Transfection was performed using Lipofectamine® LTX and Plus™ Reagent (Invitrogen) following the manufacturer’s guidelines with slight modification: 0.5 µg plasmid DNA (for co-transfection, 0.25 µg of each plasmid DNA was used) and 0.5 µL Plus™ Reagent were diluted in 100 µL OPTI-MEM® I (Gibco). In a separate tube, 1.25 µL Lipofectamine® LTX Reagent was diluted in 100 µL OPTI-MEM® I (Gibco). After 5 minutes incubation, transfection mixture from both tubes was mixed and incubated further for 20 minutes. 100 µL transfection mixture was added drop-wise onto cells in 0.5 mL fresh complete media and cells were analysed after 24 or 48 hours based on experimental setup.

For some of the immunofluorescence microscopy experiments, HeLa cells were seeded at 2.5 x 104 cells per cm2 into 12-well culture plate and plasmid DNA transfection was performed using jetPEI® DNA Transfection Reagent (Polyplus). jetPEI® (2 µL) and plasmid DNA (0.5 µg) were individually diluted in 150 mM NaCl (50 µL). Diluted jetPEI® was added to diluted DNA, mixed by vortexing and stood

54 for 25 minutes before adding 100 µL drop-wise onto cells in 0.5 mL fresh complete media.

2.8 Treatment with proteasome and lysosomal protease inhibitors

100 nM bortezomib (Selleckchem, prepared as 100 µM stock in DMSO) was used to inhibit proteasome and was added directly to fresh complete media 4 to 6 hours before cells were collected for analysis. In some experiments (as indicated), 10 nM bortezomib (Selleckchem, prepared as 10 µM stock in DMSO), and combination of 0.5 mM leupeptin (Merck, prepared as 50 mM stock in water) and 1 µg/mL Pepstatin A (Sigma, prepared as 1 mg/mL stock in water) used to inhibit lysosomal protein degradation were added to cells 18 hours prior to analysis by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting.

2.9 Endo Hf treatment

Proteins were denatured by heating cell lysate at 100 °C for 10 minutes in SDS-PAGE sample buffer (2 % w/v SDS, 5 % v/v glycerol and 0.01 % w/v bromophenol blue in 31.25 mM Tris pH 7.6 with freshly added 100 mM DTT). 30 µL cell lysate was then treated with 1 µL Endo Hf (New England BioLabs) and incubated at 37 °C for 3 hours. Samples were then analysed by SDS-PAGE and immunoblotting.

2.10 Cycloheximide chase

Cycloheximide chase assay is used to inhibit protein synthesis and measure the rate of degradation of different model proteins. Cells were grown in 12-well culture plate and transfected or induced to express the protein of interest. A final concentration of 100 µg/mL cycloheximide (VWR) was added to each well containing fresh complete media and cells were harvested in SDS-PAGE sample buffer (2 % w/v SDS, 5 % v/v glycerol and 0.01 % w/v bromophenol blue in 31.25 mM Tris pH 7.6 with freshly added 100 mM DTT) immediately (0 hour), or after

55 specific time intervals in the presence of cycloheximide as indicated in individual experiments. Cell lysates were analysed by immunoblotting using antibodies specific to the protein of interest and a loading control (tubulin or actin), allowing the amount of protein remaining at each time point to be calculated.

2.11 SDS-PAGE and immunoblotting

Cells were rinsed twice with PBS (137 mM NaCl, 3 mM KCl, 10 mM

Na2HPO4, 2 mM KH2PO4), and lysed directly in SDS-PAGE sample buffer (2 % w/v SDS, 5 % v/v glycerol and 0.01 % w/v bromophenol blue in 31.25 mM Tris pH 7.6) containing 100 mM DTT. Lysates were sonicated at 4 °C for 5 minutes with 30- second on/off intervals. Samples were denatured by heating at 70 °C for 10 minutes before separating on 8 to 16 % polyacrylamide gels. SDS-PAGE was performed under 30 mA current in SDS running buffer (0.08 % w/v SDS and 190 mM glycine in 25 mM Tris). Proteins were immediately transferred onto Odyssey® Nitrocellulose Membranes (LI-COR Biosciences) in transfer buffer (150 mM glycine, 20% v/v methanol in 20 mM Tris) by applying 300 mA current for 1 hour. After blocking membranes with 5 % w/v milk in TBS (150 mM NaCl in 20 mM Tris pH 7.4) for 20 minutes, membranes were incubated with primary antibody (Table 2.5) made up in 5 % milk overnight at 4 °C. Membranes were washed thrice in TBS before incubating with secondary antibody (Table 2.5) made up in 5 % milk at room temperature for 1 hour. Membranes were then scanned using Odyssey® SA Infrared Imaging System (LI-COR Biosciences) with 100 µm resolution in the 700 nm and 800 nm channels using the intensity range of 7 to 12, images analysed using Odyssey Sa (LI-COR Biosciences) and bands quantified with Image Studio Lite (LI-COR Biosciences). Membranes were incubated with other antibodies when necessary. Proteins that are difficult to detect would be analysed first. For UBR4 detection, lysates were not sonicated and samples were run in 4 to 16 % gradient polyacrylamide gels and transferred onto Odyssey® Nitrocellulose Membranes at 4 °C by applying constant 20 V for 16 hours. Membranes were blocked with Odyssey® Blocking Buffer PBS (LI-COR Biosciences) for 30 minutes before incubating with primary and secondary antibodies in Odyssey® Blocking Buffer PBS.

Table 2.5: Primary and secondary antibodies used in immunoblotting.

Primary antibody Species Dilution Source BAG6 Chicken 1:5000 Abcam ab37751 BAG6 Rabbit 1:1000 Quentin Roecbuck, University of Manchester BirA Mouse 1:1000 Martina Maric, Institute of Medical Biology, Singapore Flag Mouse 1:1000 Sigma F3165 HA Mouse 1:1000 Santa Cruz Biotechnology sc-7392 HA Rabbit 1:1000 Sigma H6908 Myc Mouse 1:1000 Sigma M5546 Opsin Mouse 1:1000 Quentin Roecbuck, University of Manchester RNF126 Rabbit 1:3000 Abcam ab87256 Tubulin Rabbit 1:5000 Abcam ab4074 UBR4 Rabbit 1:1000 Abcam ab86738 V5 Mouse 1:5000 Invitrogen R960-25 V5 Rabbit 1:5000 Abcam ab9116

Secondary antibody Dilution Source IRDye® 800CW Donkey anti- 1:10000 LI-COR Biosciences Chicken P/N 926-32218 IRDye® 800CW Donkey anti- 1:10000 LI-COR Biosciences Mouse P/N 926-32212 IRDye® 680RD Donkey anti- 1:10000 LI-COR Biosciences Mouse P/N 926-68072 IRDye® 800CW Donkey anti- 1:10000 LI-COR Biosciences Rabbit P/N 926-32213 IRDye® 680RD Donkey anti- 1:10000 LI-COR Biosciences Rabbit P/N 926-68073

Others Dilution Source IRDye® 800CW Streptavidins 1:10000 LI-COR Biosciences P/N 926-32230

2.12 Immunofluorescence microscopy

All chemicals were purchased from Sigma-Aldrich® and prepared in PBS unless specifies otherwise. Cells were fixed with 3 % v/v formaldehyde for 20 minutes at room temperature, rinsed twice with PBS, the unreacted aldehyde groups quenched twice by incubating with PBS containing a few drops of 1 M glycine (Fisher Scientific) pH 8 for 5 minutes each, then permeabilised with 0.1 % (v/v) Triton X-100 for 4 minutes before incubation with primary antibody (Table 2.6) for 30 minutes at room temperature. After three washes with PBS (137 mM NaCl, 3 mM

KCl, 10 mM Na2HPO4, 2 mM KH2PO4) (5 minutes each), coverslips were incubated with Alexa Fluor® 488/594 secondary antibody (Life Technologies) (Table 2.6) for 30 minutes and nucleus was stained with 0.25 µg/mL DAPI for 5 minutes. Lastly, coverslips were mounted onto glass slides with 7 µL ProLong® Gold antifade reagent (Molecular Probes) after a brief rinse in water. Samples were analysed with Olympus BX60 microscope (Olympus Corporation) and images were captured with CoolSNAP EZ camera (Photometrics®) using MetaMorph® software (Molecular Devices). Images were later processed with ImageJ Launcher.

Table 2.6: Primary and secondary antibodies used in immunofluorescence microscopy.

Primary antibody Species Dilution Source BAG6 Chicken 1:100 Abcam ab37751 BAG6 Rabbit 1:100 Quentin Roecbuck, University of Manchester HA Mouse 1:200 Santa Cruz Biotechnology sc-7392 HA Rabbit 1:200 Sigma H6908 Opsin Mouse 1:100 Quentin Roecbuck, University of Manchester V5 Mouse 1:200 Invitrogen R960-25 V5 Rabbit 1:200 Abcam ab9116

Secondary antibody Dilution Source Alexa Fluor® 488 goat anti- 1:500 Life Technologies chicken IgG (H+L) A-11039 Alexa Fluor® 488 goat anti- 1:500 Life Technologies mouse IgG (H+L) A-11001 Alexa Fluor® 594 goat anti- 1:200 Life Technologies mouse IgG (H+L) A-11005 Alexa Fluor® 488 goat anti- 1:500 Life Technologies rabbit IgG (H+L) A-11008 Alexa Fluor® 594 goat anti- 1:200 Life Technologies rabbit IgG (H+L) A-11012

2.13 Cell cracking

T-REx™ HeLa cells were seeded into 10 cm culture dish and induced expression with 1000 ng/mL tetracycline (Sigma) for 24 hours. Trypsinised cells were collected by centrifugation at 2000 rpm for 3 minutes, rinsed twice with PBS

(137 mM NaCl, 3 mM KCl, 10 mM Na2HPO4, 2 mM KH2PO4), swelled in hypotonic ™ buffer (10 mM KAc, 1 mM MgAc2, 1 mM PMSF, 1mM DTT and 1X Halt Protease Inhibitor Cocktail from Thermo Scientific in 10 mM HEPES pH7.4) for 15 minutes on ice and cracked for 30 times by forcing cells over 10 µm ball bearing fitted in pre- chilled Cell Homogenizer (Isobiotec). Cells were collected into pre-chilled microfuge tubes, centrifuged at 600 g for 10 minutes at 4 °C and supernatant adjusted to 50 mM HEPES (pH 7.4), 100 mM KAc and 2 mM MgAc2 prior to co- immunoprecipitation.

2.14 Co-immunoprecipitation

Cells were seeded into 6-well culture plate, transfected with siRNA and/or plasmid or induced expression with 1000 ng/mL tetracycline (Sigma) for 24 hours before 100 nM bortezomib (Selleckchem) was added to fresh complete media 4 to 6 hours where indicated. Cells were rinsed twice with PBS (137 mM NaCl, 3 mM KCl,

10 mM Na2HPO4, 2 mM KH2PO4) and lysed in digitonin immunoprecipitation buffer ™ (2 % w/v digitonin, 150 mM NaCl, 5 mM MgCl2, 1 mM PMSF and 1 mM Halt Protease Inhibitor Cocktail in 25 mM Tris-HCl pH7.4) or Triton X-100 immunoprecipitation buffer (1 % v/v Triton X-100, 150 mM NaCl, 1 mM PMSF and 1 mM Halt™ Protease Inhibitor Cocktail in 25 mM Tris-HCl pH7.4) with 1 hour gentle agitation at 4 °C. 10 % of cell lysates obtained by centrifugation at 6500 rpm for 10 minutes at 4 °C were kept as ‘Input’ and the remaining cell lysates were incubated with primary antibody and Protein A Resin (GenScript) (if IgG class antibodies from rabbit is used) or Protein G Resin (GenScript) (if IgG class antibodies from mouse is used) overnight at 4 °C with constant agitation. Resins were rinsed thrice with detergent-free immunoprecipitation buffer to remove non- specifically bound proteins. Proteins of interest were then eluted from the resins with heating in SDS-PAGE sample buffer (2 % w/v SDS, 5 % v/v glycerol and 0.01 % w/v bromophenol blue in 31.25 mM Tris pH 7.6 with freshly added 100 mM DTT) and analysed by SDS-PAGE and immunoblotting. For V5 pulldown, cell lysate was incubated with Anti-V5 Agarose (Sigma) with agitation for 2.5 hours then washed and processed as described earlier.

2.15 Denaturing immunoprecipitation

Cells were seeded into 6-well culture plate, transfected with siRNA and/or plasmid or induced expression with 1000 ng/mL tetracycline (Sigma) for 24 hours before 100 nM bortezomib (Selleckchem) was added to fresh complete media 4 to 6 hours. Cell culture media were replaced with 20 mM NEM (Sigma-Aldrich) for 10 minutes on ice, rinsed twice with PBS (137 mM NaCl, 3 mM KCl, 10 mM Na2HPO4,

2 mM KH2PO4) and lysed in RIPA buffer (150 mM NaCl, 0.5 % w/v sodium deoxycholate, 1 % v/v Triton X-100, 0.1 % v/v SDS, 1 mM PMSF and 1 mM Halt™ Protease Inhibitor Cocktail in 25 mM Tris-HCl pH7.4) with 1 hour gentle agitation at

4 °C. 10 % of cell lysates obtained by centrifugation at 6500 rpm for 10 minutes at 4 °C were kept as ‘Input’ and the remaining cell lysates were incubated with primary antibody and Protein A Resin (GenScript) (if IgG class antibodies from rabbit is used) or Protein G Resin (GenScript) (if IgG class antibodies from mouse is used) overnight at 4 °C with constant agitation. Resins were rinsed thrice with RIPA buffer to remove non-specifically bound proteins. Ubiquitinated proteins were then eluted from the resins with heating in SDS-PAGE sample buffer (2 % w/v SDS, 5 % v/v glycerol and 0.01 % w/v bromophenol blue in 31.25 mM Tris pH 7.6 with freshly added 100 mM DTT) and analysed by SDS-PAGE and immunoblotting.

2.16 BioID

HeLa cells were seeded into 10 cm culture dish, transfected the next day with BirA, BAG6-BirA or BAG6(∆N)-BirA constructs for 24 hours and incubated with 50 µM biotin (Sigma-Aldrich) in fresh complete media for another 24 hours. 100 nM bortezomib (Selleckchem) was added into one of the two culture dishes transfected with BAG6-BirA construct to look at the interactor profile under proteasome inhibition. After 24 hours, cell culture media was removed completely, rinsed twice with PBS (137 mM NaCl, 3 mM KCl, 10 mM Na2HPO4, 2 mM KH2PO4) and lysed in RIPA immunoprecipitation buffer (150 mM NaCl, 0.5 % w/v sodium deoxycholate, 1 % v/v Triton X-100, 0.1 % v/v SDS, 1 mM PMSF and 1 mM Halt™ Protease Inhibitor Cocktail in 25 mM Tris-HCl pH7.4) with 1 hour gentle agitation at 4 °C. 1 % of cell lysates obtained by centrifugation at 6500 rpm for 10 minutes at 4 °C were kept as ‘Input’ and the remaining cell lysates were incubated with pre- equilibrated Dynabeads™ MyOne™ Streptavidin C1 (Invitrogen) overnight at 4 °C with constant agitation. Dynabeads were rinsed thrice with RIPA buffer to remove non-specifically bound proteins then rinsed with 50 mM Tris-HCl for three times to remove all residual detergent that will interfere with mass spectrometry. 5 % of dynabeads suspended in 50 mM Tris-HCl pH 7.4 were collected to be analysed by immunoblotting. Dynabeads were reconstituted in a small volume of 50 mM ammonium bicarbonate, snap-frozen and stored at - 80 °C before samples were sent to Sanford Burnham Prebys Medical Discovery Institute for mass spectrometry. Two technical replicates and two biological replicates were performed unless specified

61 otherwise. Raw data were processed with MaxQuant by Sanford Burnham Prebys Medical Discovery Institute and processed data were received in a Microsoft Excel spreadsheet. Box plots, histograms, scatter plots and dendrogram were generated in R with the help of Dr. Craig Lawless. Venn diagrams were generated through http://www.interactivenn.net/. Enriched gene ontology related to ubiquitination pathway was analysed by http://metascape.org/gp/index.html#/main/step1 and presented in a table.

2.17 Bioinformatics

Proteins enriched with BAG6-myc-BirA from the BioID experiment were subjected to bioinformatic analysis. Hydrophobicity prediction Perl script was prepared by Prof. Simon Hubbard. Proteins not annotated to the cytoplasm were excluded for further analysis. Kyte-Doolittle hydrophobic scale was used with hydrophobic window set at 20 and hydrophobic threshold set at 0. Command prompt was used to run the Perl script and output file in .txt containing Uniprot entry, number of hydrophobic region, percentage of hydrophobic region, and the entire polypeptide marked ‘-’ for non-hydrophobic region and ‘H’ for hydrophobic region was generated.

2.18 Data and statistical analysis

Data collected from 3 independent experiments were plotted as mean ± standard error of mean (SEM) using GraphPad Prism 7 (GraphPad Software). Two- way ANOVA or two-tailed t-test were performed where appropriate, to determine the p-value. p-value of less than 0.05 (*p<0.05) was considered statistically significant. The half-life of protein was estimated with one phase decay curve in GraphPad Prism 7. Time was plotted on the X-axis and percentage of protein remained was plotted on the Y-axis. The model for the one phase decay is

where is the value of Y when the curve intercepts the Y-axis (set as 100). is the Y value when the curve approaches infinity along the X-axis (set as 0).

, and have the same units (in this case %). is the rate constant, expressed in reciprocal of the X-axis time units. If the time unit () is in minute, then is expressed as minute-1. Half-life was estimated through . .

3 Role of BAG6 and UBR4 in ERAD

3.1 Introduction

E3 ubiquitin-protein ligase UBR4 (UBR4) was identified interacting with Sec61β and SGTA in two proteomic experiments (unpublished data from High Lab). In order to identify cellular factors that play a role in ER targeting of tail-anchored proteins, purified recombinant Sec61β (a tail-anchored protein) variants were used in pulldown experiments to isolate proteins in reticulocyte lysate that bound to the exposed transmembrane domain of Sec61β. BAG6 and UBR4 were identified to interact with Sec61β through the Sec61β transmembrane domain. Similar experiment using purified SGTA also identified UBR4 as well as BAG6. Since UBR4 interacts with both Sec61β (BAG6 substrate) and SGTA (BAG6 interacting factor), it was of interest to determine if UBR4 interacts directly or indirectly with BAG6, and whether UBR4 contributes to BAG6 mediated protein quality control.

UBR4 is a member of the UBR family of mammalian E3 ubiquitin ligases that contain a UBR box motif. The UBR box motif is a 70 amino acids zinc-finger- like domain that is shared among the 7 members of the family, UBR1 to UBR7 (Figure 3.1A). The UBR motif is characterised by its ability to bind so-called N- degrons – proteins containing specific N-terminal amino acid residues that cause rapid degradation of themselves via the N-end rule (Tasaki et al., 2005; Dissmeyer et al., 2017) (Figure 3.1B). The N-end rule predicts the stability of a protein depending on the identity of its N-terminal amino acid residue, where certain N-terminal residues serve as degradation signal (Liu et al., 2016; Dissmeyer et al., 2017). The N-terminal amino acid residue can be naturally occurring or can be generated after proteolytic cleavage and post-translational modification of precursor to expose internally embedded residue (Kim et al., 2013). The N-end rule pathway comprises of N-degrons, N-recognins and the proteasome (Liu et al., 2016). UBR proteins play a central role in the N-end rule pathway by acting as N-recognins and catalysing ubiquitination of internal lysine residues of N-degrons, resulting in targeting for ubiquitin proteasome-dependent proteolysis (Sriram and Kwon, 2010; Kim et al., 2013). Unlike the other UBR family members, UBR4 does not in fact contain active ubiquitin ligase domain such as really interesting new gene (RING), homologous to 64

Figure 3.1: UBR family proteins and the mammalian N-end rule pathway. (A) Diagrammatic representation of the seven members of the UBR family comprising an UBR domain (yellow box) and an ubiquitin ligase domain (blue hexagon) in six of them except UBR4 with a cysteine-rich domain (CRD) of unknown function (green pentagon). RING, really interesting new gene; HECT, homologous to the E6- AP carboxyl terminus; PHD, plant homeodomain. (B) Tertiary, secondary and primary N-terminal destabilising residues are indicated by single-letter amino acid code and the remaining amino acid sequences by yellow ovals. NTAN1, protein N- terminal asparagine amidohydrolase; NTAQ1, protein N-terminal glutamine amidohydrolase; NO, nitric oxide; ATE1, arginyl-tRNA--protein transferase 1.

65 the E6-AP carboxyl terminus (HECT), F-box or plant homeodomain (PHD) (Panchenko, 2016; Zheng et al., 2016; Zheng and Shabek, 2017) (Figure 3.1A), and hence is not predicted to catalyse ubiquitination of substrate proteins. However, UBR4 does contain a functional UBR box which has been shown to bind N-degrons, and a cysteine-rich domain of unknown function (Tasaki et al., 2005) (Figure 3.1A). UBR4 is a very large protein of 570 kDa (Tasaki et al., 2005) and UBR4 has been proposed to function instead as an adaptor or scaffold for other E3 ligases (Rinschen et al., 2016). UBR4 stably associates with the proteasome, but whether this association is direct, through ubiquitinated substrates or auto-ubiquitination is not known (Besche et al., 2009). UBR4 is also associated with proteins destined for autophagic degradation by the lysosome (Kim et al., 2013). Therefore, UBR4 may be involved in selective proteolysis of short-lived proteins that are substrates of the N- end rule via the proteasome, and also bulk degradation through the lysosome. To this end, two model proteins that utilised BAG6 for degradation were used to examine the role of UBR4 in quality control of mislocalised protein and ERAD.

3.2 BAG6 contributes to degradation of Op91

A truncated version of the 7 transmembrane domain protein bovine opsin, termed Op91, has previously been used as a model mislocalised protein that is rapidly degraded via a proteasome-dependent pathway (Wunderley et al., 2014; Leznicki et al., 2015). Op91 consists of the N-terminal 91 amino acids of opsin, including the entire first transmembrane domain and part of the second transmembrane domain, with two glycosylation sites (Figure 3.2A). Due to its short length, translation of Op91 is completed before the first hydrophobic transmembrane domain fully emerges from the ribosome exit tunnel to engage signal recognition particle. Thus, the polypeptide is not efficiently translocated into the ER and becomes mislocalised in the cytosol (Wunderley et al., 2014; Leznicki et al., 2015). Several bands were detected in lysates of cells transiently transfected with plasmid encoding Op91 by blotting with anti-opsin antibody; a higher molecular weight species is thought to be the doubly glycosylated form and a lower molecular weight species representing the non-glycosylated form (Figure 3.2B, lane 1). This was confirmed by treatment of lysate with endoglycosidase Hf (Endo Hf) to cleave high

B C

Figure 3.2: BAG6 overexpression stabilises Op91 and causes nuclear re- localisation of Op91. (A) Diagrammatic representation of Op91 comprising two N- glycosylation sites (triangle) at amino acid 2 and 15, first and part of second transmembrane domains (blue box). (B) HeLa cells were transfected with plasmid encoding Op91 and a portion of cell lysate was treated with Endo Hf. gly, glycosylation. (C) HeLa cells transfected with plasmid encoding Op91 were treated with bortezomib (Bz), leupeptin+pepstatin (L+P) or co-transfected with plasmid encoding BAG6-V5 (oe). gly, glycosylation. (D) HeLa cells were transfected with plasmid encoding Op91 and further treated as indicated. Cells were fixed with formaldehyde and stained with opsin, BAG6 or V5 antibodies. White arrows show the cytosolic and re-localised nuclear Op91. This is a representative image from at least 2 independent experiments. Scale bar = 10 µm.

67 mannose and some hybrid types of N-linked glycoproteins, which resulted in the disappearance of the higher molecular weight band (around 17 kDa marker) and increased intensity of the smaller species (around 7 kDa marker) (Figure 3.2B, lane 2). Hence, these species were concluded to represent glycosylated and non- glycosylated forms, respectively. Since N-glycosylation occurs within the lumen of the ER, these results indicate that a considerable proportion of the exogenously expressed Op91 was in fact inserted into the ER and only a proportion was mislocalised.

Treatment of cells with the proteasome inhibitor, bortezomib increased levels of both the non-glycosylated Op91 and the doubly glycosylated form (Figure 3.2C, lane 2). Non-glycosylated Op91 level was not increased by lysosomal protease inhibitors (Figure 3.2C, lane 3), suggesting that mislocalised Op91 is a substrate of the proteasome and not the lysosome. Transient overexpression of BAG6 also increased the level of non-glycosylated mislocalised Op91 (Figure 3.2C, lane 4). This is consistent with the suggestion that overexpression of BAG6 has a dominant negative effect on its substrates (Payapilly and High, 2014). BAG6 functions with Ubl4A and TRC35 in a trimeric complex (Mariapan et al., 2010) and thus increasing BAG6 levels may disrupt the stoichiometry of the complex, delaying degradation of its substrate.

Immunofluorescence microscopy of cells transfected with plasmid encoding Op91 revealed only background staining with anti-opsin antibody (Figure 3.2D, panel 1), most likely due to rapid degradation of Op91 leading to very low steady state expression level (Figure 3.2B and 3.2C). Following treatment with bortezomib, Op91 accumulated in punctate structures (Figure 3.2D, panel 2), possibly due to aggregation of Op91 which contains exposed hydrophobic domains. Overexpressed BAG6 localised to the nucleus (Figure 3.2D, panel 3), consistent with previous findings (Leznicki et al., 2010; Leznicki et al., 2013; Payapilly and High, 2014; Wunderley et al., 2014). This is suggested to be due to exposure of the BAG6 NLS when BAG6 is expressed at higher level than TRC35 which would normally bind and mask BAG6 NLS (Wang et al., 2011). Notably, co-expression of BAG6 with Op91 caused re-localisation of Op91 to the nucleus (Figure 3.2D, panel 3). This suggests that BAG6 binds to Op91 in cells, and is in line with published data showing overexpressed BAG6 drives nuclear re-localisation of two GET pathway

68 substrates, Sec61β and Sed5 (Leznicki et al., 2010; Leznicki et al., 2013). Together these data are consistent with Op91 being a BAG6 substrate and thus Op91 represents a suitable model to examine the role of UBR4 in mislocalised protein degradation.

Detection of UBR4 was challenging due to its very large size, and several methods of cell lysis were tested to optimise detection. Using the most widely cited commercially available antibody, UBR4 could be detected as a single species above the 245 kDa marker when cells were lysed directly in SDS-PAGE sample buffer and analysed on 4-16 % polyacrylamide gel (Figure 3.3A). Sonication of lysates resulted a complete loss of the anti-UBR4 signal, but freeze-thaw cycles were well tolerated (Figure 3.3A). Having established conditions for UBR4 detection, the effect of UBR4 depletion on degradation of Op91 was examined with cycloheximide chase experiments. Cycloheximide is a protein synthesis inhibitor that affects the translocation step during protein translation and causes an elongation halt. Hence, cells were treated with UBR4 or scrambled siRNA, transiently transfected with plasmid encoding Op91 and then treated with cycloheximide to inhibit protein synthesis. The levels of Op91 remaining in cells were then determined by SDS- PAGE and immunoblotting to provide a readout of Op91 degradation. Following addition of cycloheximide, the intensity of both the doubly glycosylated and non- glycosylated forms of Op91 decreased with time, suggesting that both forms were degraded over this period (Figure 3.3B). In order to examine the rate of degradation, the Op91 signal intensity at each time point was quantified and expressed relative to the signal present at the start of the chase (t = 0). In cells treated with a non-targeting siRNA, the doubly glycosylated Op91 was more rapidly degraded than the non- glycosylated form, with 30 % of the doubly glycosylated and 50 % of the non- glycosylated form remaining after a 60-minute chase (Figure 3.3C and 3.3D). Degradation of the non-glycosylated Op91 was not altered in cells depleted of UBR4, despite efficient knockdown (Figure 3.3B, lanes 9-12 and 3.3D), suggesting that UBR4 is not required for degradation of this mislocalised protein. Surprisingly, BAG6 knockdown also failed to stabilise non-glycosylated Op91 (Figure 3.3B, lanes 5-8 and 3.3D), in contrast to previous work implicating BAG6 in degradation of mislocalised membrane proteins (Hessa et al., 2011). It is possible that alternative degradation pathways compensate for lack of BAG6 (Rodrigo-Brenni et al., 2014).

Op91-2gly C D Op91-0gly 120 120 * siNT 100 * 100 siBAG6 80 * 80 siUBR4 60 60

40 40

20 20 Remaining Op91-2gly (%) 0 Remaining Op91-0gly (%) 0 0204060 0 204060 CHX chase (min) CHX chase (min)

Figure 3.3: Degradation of glycosylated form of Op91 is delayed with BAG6 knockdown. (A) HeLa cell lysates were divided and proccessed as indicated by sonication and/or freeze-thaw before analysing with immunoblotting. (B) HeLa cells were treated with the indicated siRNAs and transiently transfected with plasmid encoding Op91. Cells were then treated with cycloheximide (CHX). Cells were harvested 0, 15, 30 and 60 minutes after cycloheximide treatment and lysates analysed by immunoblotting. siNT, non-targeting siRNA; gly; glycosylation; Kd, knockdown. (C, D) The intensity of glycosylated (C) and non-glycosylated (D) forms of Op91 signal were quantified using Image Studio Lite software then expressed relative to amount of Op91 at 0 hour (set to 100%). Data shown are the mean ± SEM of 3 independent experiments. *p<0.05 using Two-way ANOVA.

In contrast, BAG6 knockdown caused a significant stabilisation of the doubly glycosylated Op91 (Figure 3.3B, lanes 5-8 and 3.3C). This species of Op91 must have inserted into the ER membrane where glycosylation occurs, hence its degradation is likely to occur through the ERAD pathway. This suggests that BAG6 is required for efficient ERAD of Op91 in line with previous reports on the involvement of BAG6 in ERAD (Claessen and Ploegh, 2011; Wang et al., 2011; Payapilly and High, 2014). However, UBR4 knockdown had no effect on the doubly glycosylated Op91 (Figure 3.3B, lanes 9-12 and 3.3C). Large amount of sample was analysed with immunoblotting in order to detect Op91 and loading control was saturated at that point (Figure 3.3B, IB: Tubulin). Therefore, Op91 signal was not normalising to tubulin signal and this might account for some errors in the quantification.

3.3 BAG6 and UBR4 are involved in ERAD of opsin-degron

In order to examine whether UBR4 contributes to degradation of a defined ERAD substrate, a variant of bovine opsin containing a dibasic degron motif (L46R and G51K) in the first transmembrane domain was used (Figure 3.4A). This protein is retained in the ER and degraded via ERAD (Ray-Sinha et al., 2009), and BAG6 has been shown to play a role in its degradation (Payapilly and High, 2014). To determine the potential role of UBR4 in this process, HeLa cells stably expressing opsin-degron under an inducible promoter were transfected with UBR4 or non- targeting siRNA, then tetracycline was added to induce expression of opsin-degron prior to carrying out cycloheximide chase assays. A single band of around 32 kDa was observed in cell lysate upon immunoblotting with anti-opsin antibody (Figure 3.4B), consistent with the expected size of glycosylated opsin-degron (Payapilly and High, 2014). Opsin-degron is more stable than Op91, and so a 4-hour cycloheximide chase was performed. In cells treated with non-targeting siRNA, more than 90 % of opsin-degron was degraded after 4 hours of cycloheximide treatment (Figure 3.4B, lanes 1-4 and 3.4C), giving an estimated half-life of 1.2 hours based on exponential one-phase decay (Figure 3.4D). BAG6 knockdown significantly delayed degradation of opsin-degron (Figure 3.4B, lanes 5-8 and 3.4C). The half-life of opsin-degron was increased to approximately 2 hours in BAG6 depleted cells (Figure 3.4D),

A B

100 siNT C siBAG6 * siUBR4

50 * Remaining OpD (%) OpD Remaining

0 01234 CHX chase (h)

D siNT siBAG6 siUBR4 One phase decay Y0 (%) = 100 = 100 = 100

Plateau (%) = 0 = 0 = 0 -1 K (h ) 0.56 0.35 0.30 Half Life (h) 1.2 2.0 2.3

Figure 3.4: Degradation of opsin-degron is delayed with BAG6 and UBR4 knockdown. (A) Diagrammatic representation of opsin-degron comprising two N- glycosylation sites (blue triangle) at amino acid 2 and 15, seven transmembrane domains (blue box) with a dibasic degron motif (red cross) in the first transmembrane domain. ER, endoplasmic reticulum. (B) HeLa cells were treated with the indicated siRNAs and induced to express opsin-degron with tetracycline. Cells were then treated with cycloheximide (CHX). Cells were harvested 0, 1, 2 and 4 hours after cycloheximide treatment and lysates analysed by immunoblotting. siNT, non-tageting siRNA; Kd, knockdown. (C) The intensity of the opsin-degron signal was quantified using Image Studio Lite software and normalised to tubulin, then expressed relative to the amount of opsin-degron at 0 hour (set to 100%). Data shown are the mean ± SEM of 3 independent experiments. *p<0.05 using two-way ANOVA. (D) An estimate of opsin-degron half-life under different treatments with one phase decay curve.

72 degron half-life of 1.1 hours and 2 hours for non-targeting and BAG6-targeting siRNAs treated cells, respectively. Strikingly, an even greater delay in the degradation of opsin-degron was observed after depletion of UBR4 (Figure 3.4B, lanes 9-10 and 3.4C), which increased the half-life to 2.3 hours (Figure 3.4D). The great stabilisation effect on opsin-degron with UBR4 knockdown provided evidence that UBR4 is involved in ERAD of this membrane protein.

3.4 UBR4 does not associate stably with BAG6 or opsin-degron

Depletion of BAG6 and UBR4 were shown to delay opsin-degron degradation. An involvement of UBR4 in ERAD has not previously been reported. Hence, it was of interest to determine how UBR4 might carry out this role. To this end, the ability of UBR4 to interact with BAG6 or opsin-degron was examined.

In order to facilitate these experiments, HeLa cells stably expressing BAG6 with a C-terminal V5 tag (BAG6-V5) under a tetracycline inducible promoter were generated using Flp-In™ TREx™ system (Ward et al., 2011). Protein-protein interactions can be sensitive to detergent, and therefore cell extracts were prepared with different detergents. HeLa cells expressing BAG6-V5 were lysed in digitonin or Triton X-100 containing buffer, then cell lysates were prepared and BAG6-V5 was immunoprecipitated using anti-V5 agarose. Non-induced cells were used as a control to detect non-specific interactions between proteins and agarose beads. Immunoprecipitated proteins were analysed by SDS-PAGE and immunoblotting. BAG6 was efficiently immunoprecipitated from cell lysates prepared from both detergents (Figure 3.5A, lanes 6 and 8). RNF126 is the main reported BAG6- interacting E3 ligase (Rodrigo-Brenni et al., 2014), so it was used as a positive control for the co-immunoprecipitation. A strong RNF126 signal was observed in the anti-V5 immunoprecipitates from both digitonin and Triton X-100 lysates of cells induced to express BAG6-V5 (Figure 3.5A, lanes 6 and 8), but was not detected in control non-induced cells (Figure 3.5A, lanes 5 and 7). Although UBR4 was readily detected in the cell lysate (Figure 3.5A, lanes 1-4), none was observed in the V5- immunoprecipitates under any of the different conditions used for the co- immunoprecipitation (Figure 3.5A, lanes 5-8). Thus, even under mild lysis conditions that maintained BAG6-RNF126 interactions, UBR4 was not detected in

Figure 3.5: BAG6 and opsin-degron do not co-immunoprecipitate UBR4. (A) HeLa cells were induced to express BAG6-V5 with tetracycline (Tet) or left non- induced as indicated and lysed by digitonin or Triton X-100 buffer. Cell lysate was incubated with V5-agarose and immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). At least two independent experiments were analysed. (B) HeLa cells were induced to express opsin-degron with tetracycline (Tet) then treated with bortezomib (Bz) for 4 hours as indicated before lysis in digitonin buffer. Cell lysate was immunoprecipitated with anti-opsin antibody and immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). At least two independent experiments were analysed.

74 association with BAG6. These results suggest that UBR4 and BAG6 do not interact, or that any interaction is very labile.

In order to examine whether UBR4 interacts with the ERAD substrate, opsin- degron expressing HeLa cells were lysed in digitonin buffer and cell lysates immunoprecipitated with anti-opsin antibody. Previous work showed that proteasome inhibition stabilised the interaction between opsin-degron and BAG6 (Payapilly and High, 2014), and hence cells were treated with or without proteasome inhibitor. Opsin-degron immunoprecipitation was efficient (Figure 3.5B, lanes 7 and 8), and the presence of BAG6 was evident in opsin-degron-immunoprecipitates from cells treated with proteasome inhibitor (Figure 3.5B, lane 8). UBR4 was detected in the cell lysate (Figure 3.5B, ‘Input’ lanes 1-4), but could not be co- immunoprecipitated with opsin-degron (Figure 3.5B, lanes 5-8), even following treatment of cells with bortezomib. These data suggest that UBR4 does not stably associate with either BAG6 or the ERAD substrate opsin-degron.

Next, the possibility that UBR4 depletion affects opsin-degron ERAD indirectly by altering BAG6 levels was tested. As shown in Figure 3.6A and 3.6B, UBR4 knockdown by 83 % reduced endogenous BAG6 levels slightly, to 76 % of control cells. Conversely, BAG6 knockdown by 71 % increased UBR4 levels a bit. The reduction in BAG6 levels was not statistically significant, and it seems unlikely that UBR4 knockdown inhibits opsin-degron degradation by affecting levels of BAG6.

3.5 Depletion of UBR4 increases opsin-degron ubiquitination

BAG6 has been shown to bind a wide range of substrates and recruit E3 ligases through its ubiquitin-like domain in order to promote substrate ubiquitination for proteasomal degradation (Hessa et al., 2011; Rodrigo-Brenni et al., 2014). This raised the possibility that UBR4 might impact on ubiquitination of BAG6 substrates. Thus, the effect of UBR4 knockdown on total levels of protein ubiquitination and polyubiquitin conjugates associated with BAG6 was tested. Non-induced opsin- degron cells were transfected with non-targeting siRNA or siRNA targeting UBR4 for 48 hours and then co-transfected with plasmids encoding BAG6-V5 and HA-

B 1.5 siNT siBAG6 1.0 siUBR4

0.5

0.0

Protein level relative to siNT (au) siNT to relative level Protein BAG6 UBR4

Figure 3.6: UBR4 knockdown does not significantly alter BAG6 steady state level. (A) HeLa cells were treated with the indicated siRNAs for 72 hours and lysates analysed by immunoblotting. siNT, non-targeting siRNA; Kd, knockdown. (B) The intensity of BAG6 and UBR4 signals were quantified using Image Studio Lite software and normalised to tubulin, then expressed relative to their respective signals in the non-targeting knockdown sample (siNT). Data shown are the mean + SEM of 3 independent experiments. p>0.05 using two-tailed t-test.

Ubiquitin for 24 hours. Cells were treated with bortezomib to prevent rapid turnover of polyubiquitinated opsin-degron then lysed in digitonin buffer before performing non-denaturing immunoprecipitation with anti-V5 agarose. Immunoprecipitated proteins were analysed by SDS-PAGE and immunoblotting with anti-HA antibody to detect ubiquitin. BAG6-V5 was efficiently immunoprecpiated by the V5-agarose (Figure 3.7A, lanes 4 and 5) but did not bind non-specifically to untagged agarose beads (Figure 3.7A, lanes 3). Blotting with anti-HA antibody revealed a ‘smear’ of high molecular weight bands in the cell lysates; these are typical of polyubiquitinated proteins (Minami et al., 2010; Hessa et al., 2011; Leznicki et al., 2012). Interestingly, cells depleted of UBR4 had noticeably higher levels of cellular polyubiquitinated proteins as judged by these high species (Figure 3.7A, lane 2 vs 1 and 3.7B), suggesting that loss of UBR4 has an impact on protein ubiquitination globally. A considerable amount of these high molecular weight anti-HA reactive species were co-precipitated with BAG6-V5 (Figure 3.7A, lanes 4 and 5) but not with the control agarose beads (Figure 3.7A, lane 3). These are likely to represent ubiquitinated proteins that are associated with BAG6 such as BAG6 substrates. Interestingly, a greater amount of anti-HA reactive polyubiquitinated proteins were co-precipitated with BAG6 (Figure 3.7A, lane 5 vs 4 and 3.7B) when UBR4 was knocked down. This suggests that loss of UBR4 increases global protein ubiquitination and promotes BAG6 recruitment of polyubiquitinated substrates.

Next, it was of interest to determine UBR4 role on ubiquitination of a specific substrate. Therefore, the effect of UBR4 depletion on ubiquitination of opsin-degron was examined. Opsin-degron cells were treated with UBR4 or BAG6 siRNA for 48 hours, transfected with plasmid encoding HA-tagged ubiquitin and induced to express opsin-degron for a further 24 hours. Cells were treated with proteasome inhibitor to inhibit degradation of polyubiquitinated opsin-degron, and lysed with buffer containing SDS to disrupt non-covalent interactions prior to immunoprecipitation of opsin-degron. As observed above, UBR4 depletion resulted in an increase in total levels of polyubiquitin conjugates detected in cell lyates by blotting with anti-HA antibody (Figure 3.8A, lane 2), when compared to BAG6 siRNA and control siRNA (Figure 3.8A, lanes 1 and 3). Control immunoprecipitation with mouse IgG showed no non-specific binding of HA- ubiquitin to the beads (Figure 3.8A, lane 4). Immunoprecipitation with anti-opsin

B 8 10 07 siNT siUBR4 6 10 07

4 10 07

2 10 07 HA-Ubiquitin signal

0 Total BAG6-associated

Figure 3.7: UBR4 knockdown increases total cellular ubiquitination and BAG6- associated polyubiquitinated species. (A) Non-induced opsin-degron cells were transfected with siRNAs for 48 hours then co-transfected with plasmids encoding BAG6-V5 and HA-ubiquitin (Ub) for 24 hours. Cells were then treated with bortezomib for 4 hours before lysis in digitonin buffer. Cell lysate was incubated with untagged-agarose (Control) or V5-agarose and immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). siNT, non-tageting siRNA. (B) The intensities of the HA-ubiquitin signal in the input (Total) and immunoprecipitated fraction (BAG6- associated) were quantified using Image Studio Lite software. Data shown are the mean and individual values of two independent experiments.

B 6 10 07 siNT siUBR4 4 10 07 siBAG6

2 10 07 HA-Ubiquitinsignal

0 Total Opsin-degron

Figure 3.8: UBR4 knockdown increases opsin-degron ubiquitination. (A) Cells were transfected with siRNAs for 48 hours then transfected with plasmid encoding HA-ubiquitin (Ub) and induced to express opsin-degron with tetracycline for 24 hours. Cells were then treated with bortezomib for 4 hours before lysis in SDS containing buffer. Cell lysate was incubated with mouse IgG (Control) or anti-opsin antibody and immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). siNT, non-tageting siRNA. (B) The intensities of the HA-ubiquitin signal in the input (Total) and immunoprecipitated fraction (Opsin-degron) were quantified using Image Studio Lite software. Data shown are the mean and individual values of two independent experiments.

79 antibody isolated a range of high molecular weight proteins detected by anti-HA antibody (Figure 3.8A, lanes 5-7). Since immunoprecipitation was done under denaturing conditions, these high molecular weight proteins are likely to represent HA-ubiquitinated opsin-degron. Cells expressing opsin-degron showed a low level of ubiquitination after non-targeting and BAG6 knockdown (Figure 3.8A, lanes 5 and 7). It is surprising that BAG6 knockdown did not affect opsin-degron ubiquitination as has been found previously (Payapilly and High, 2014). The relatively low efficiency of BAG6 knockdown (Figure 3.8A, lane 3 vs 1) may have contributed to apparent lack of effect of BAG6 depletion on opsin-degron ubiquitination. Strikingly however, a much greater amount of polyubiquitinated opsin-degron was isolated from cells depleted of UBR4 (Figure 3.8A, lane 6), suggesting that UBR4 depletion increased opsin-degron ubiquitination. Together these results suggest that UBR4 knockdown somehow increases levels of ubiquitinated opsin-degron, and also total cellular polyubiquitin-conjugated proteins. This could conceivably be due to increased ubiquitination, reduced de-ubiquitination, and/or reduced proteasomal degradation. At least one more biological replicate has to be performed in order to confirm this finding statistically.

3.6 UBR4 knockdown increases association of BAG6 with opsin-degron and RNF126

Having found that UBR4 knockdown increased ubiquitination of opsin- degron, the possibility that UBR4 somehow altered the interaction between BAG6 and opsin-degron was tested. HeLa cells stably expressing opsin-degron were transfected with non-targeting or UBR4 siRNA for 48 hours, then tetracycline was added to induce expression of opsin-degron for 24 hours. Transfected cells were treated with proteasome inhibitor for 4 hours before lysis in digitonin buffer. Opsin- degron was immunoprecipiated and co-precipitation of BAG6 was measured under the same conditions as illustrated in Section 3.4. Opsin-degron was efficiently isolated using anti-opsin antibody (Figure 3.9A, lanes 5 and 6) but not the control mouse IgG (Figure 3.9A, lane 4), confirming the specificity of the immunoprecipitation. BAG6 was co-immunoprecipitated with opsin-degron from lysates of cells treated with non-targeting or UBR4 siRNA (Figure 3.9A, lanes 5 and

A C

B D

0.03

0.02

0.01

0.00

R4 siNT iUB s

Figure 3.9: UBR4 knockdown increases opsin-degron-BAG6 and BAG6- RNF126 interactions. (A) Cells were transfected with siRNAs for 48 hours then induced to express opsin-degron with tetracycline for 24 hours. Cells were then treated with bortezomib for 4 hours before lysis in digitonin buffer. Cell lysate was incubated with mouse IgG (Control) or anti-opsin antibody and immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). siNT, non-tageting siRNA. (B) Cells were transfected with siRNAs for 48 hours then induced to express BAG6-V5 with tetracycline for 24 hours before lysis in digitonin buffer. Cell lysate was incubated with untagged-agarose (Control) or V5-agarose and immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). siNT, non-tageting siRNA. (C, D) The intensity of BAG6 and RNF126 signals were quantified using Image Studio Lite software and normalised to opsin and V5, respectively. Data shows the mean and individual values of two independent experiments.

6), but was not detected in the control IgG immunoprecipitates (Figure 3.9A, lane 4). Quantification of the BAG6 and opsin-degron signals in the immunoprecipitated material revealed that the relative amount of BAG6 that was co-immunoprecipitated with opsin-degron increased by approximately 25 % when UBR4 was knocked-down (Figure 3.9C). This observation is interesting as it suggests that lack of UBR4 may enhance or stabilise the interaction between BAG6 and the substrate opsin-degron.

Next, the impact of UBR4 depletion on the interaction between BAG6 and the E3 ligase RNF126 was examined. Cells were transfected with non-targeting or UBR4 siRNA for 48 hours then induced to express BAG6-V5 for 24 hours. Cells were then lysed in digitonin buffer and co-immunoprecipitation was performed with anti-V5 agarose as illustrated in Section 3.4. BAG6 was immunoprecipitated by V5- agarose (Figure 3.9B, lane 5 and 6) but not by untagged-agarose (Figure 3.9B, lane 4). RNF126 was co-purified specifically by BAG6 (Figure 3.9B, lanes 5 and 6), RNF126 was not detected in control immunoprecipitates where BAG6 was absent (Figure 3.9B, lane 4). UBR4 siRNA increased the amount of RNF126 co- immunoprecipitated with BAG6-V5 (Figure 3.9B, lane 6 vs 5, IB:RNF126), the amount of BAG6-V5 was also increased (Figure 3.9B, lane 6 vs 5, IB:V5); but when this was accounted for by normalising amount of immunoprecipitated RNF126 to BAG6, an increase was still observed (Figure 3.9D). Taken together, these results suggest that UBR4 might affect opsin-degron ubiquitination by increasing BAG6 interactions with opsin-degron and RNF126. At least one more biological replicate has to be performed in order to confirm these findings statistically.

3.7 Discussion

UBR4 was studied here as it had been identified in two independent mass spectrometry experiments aiming to discover factors involved in tail-anchored protein biogenesis (unpublished data from High Lab). In these experiments, recombinant Sec61β, a variant lacking the transmembrane domain (Sec61β∆TMD), and another variant in which the lysine residues were substituted with arginine (Sec61β∆K), were used to detect interactions dependent on hydrophobicity and ubiquitination, respectively. In addition to BAG6 and its binding partners Ubl4A and TRC35, UBR4 was shown to interact with Sec61β in a transmembrane domain- and

82 ubiquitination-dependent manner. The role of BAG6 in the quality control of hydrophobic proteins through the ubiquitin-proteasome system has been well documented (Leznicki et al., 2010; Minami et al., 2010; Hessa et al., 2011; Tanaka et al., 2016), and these results raised the possibility that UBR4 might coordinate with BAG6 in this role. Supporting this idea, UBR4 was also recovered in association with SGTA, which functions together with BAG6, facilitating capture of hydrophobic substrate proteins (Shao et al., 2017) and antagonizing BAG6 mediated ubiquitination (Leznicki and High, 2012; Xu et al., 2012). Since Sec61β and SGTA are a substrate and an interacting factor of BAG6, respectively, it was speculated that UBR4 might be involved in some common pathways with BAG6.

Initial studies with a model mislocalised protein did not identify a role for UBR4 (Figure 3.3B, 3.3C and 3.3D). However when an established BAG6- interacting model ERAD substrate opsin-degron was examined, UBR4 depletion was found to delay degradation of opsin-degron (Figure 3.4B, 3.4C and 3.4D) in a similar fashion with BAG6 depletion, judged by the amount of opsin-degron remained after two hours of cycloheximide treatment. To compare the half-life of opsin-degron with data published in Payapilly and High, 2014, one-phase decay curve was generated and similar half-life was estimated. This observation is interesting as the role of UBR4 is not well understood and UBR4 has not previously been implicated in ERAD. Degradation of a mitochondrial single-pass membrane protein serine/threonine-protein kinase PINK1 (PINK1) has been shown to involve UBR4 (Yamano and Youle, 2015). However, PINK1 is not an ER localised protein and therefore not an ERAD substrate. PINK1 recruits the E3 ligase Parkin and causes mitophagy when mitochondria are damaged. Under homoeostatic conditions, PINK1 is cleaved, retrotranslocated into the cytosol and degraded through the N-end rule pathway due to the destabilising N-terminus generated by the cleavage (Yamano and Youle, 2015). Hence, a role for UBR4 in degradation of PINK1 may be via the classical N-end pathway. Opsin-degron, on the other hand, is not expected to possess an N-degron. Asparagine (N) is the second amino acid in opsin after methionine, and since opsin undergoes N-terminal acetylation (Ovchinnikov, 1982), it does not have an exposed destabilising residue to be recognised as N-degron. Nevertheless, UBR family proteins have reported roles in N-end rule independent degradation of unfolded or misfolded cytosolic proteins (Nillegoda et al., 2010; Sultana et al., 2012).

Furthermore, in yeast, a role of Ubr1 in ERAD has been illustrated using two polytopic transmembrane proteins (Stolz et al., 2013). The ability of UBR4 to bind the ER (Shim et al., 2008) and the proteasome (Besche et al., 2009) has further supported the role of UBR4 in ERAD. In the future, it will be important to determine whether the effect of UBR4 is specific for opsin-degron or the effect is also observed for other ERAD substrates.

UBR4 lacks an active E3 ligase domain and therefore cannot directly mediate ubiquitination as proposed for other UBR proteins (Panchenko, 2016; Zheng et al., 2016; Zheng and Shabek, 2017). Alternatively, UBR4 may function as a scaffold for interaction of cellular components (Rinschen et al., 2016). However, a stable interaction was not detected between UBR4 and either BAG6 or opsin-degron (Figure 3.5A and 3.5B). Furthermore, BAG6 was associated with opsin-degron with or without UBR4 (Figure 3.9A), suggesting that UBR4 is not required for BAG6 to bind substrates. Affinity purification coupled mass spectrometry of ubiquitin- conjugating enzyme E2 A (RAD6A) and ubiquitin-conjugating enzyme E2 B (RAD6B) (E2 ubiquitin-conjugating enzymes) identified the E3 ubiquitin-protein ligase KCMF1 (KCMF1) and UBR4 as binding partners (Hong et al., 2015). KCMF1 was reported to sit in the centre of this E2-E3 complex by interacting RAD6A/B with its N-terminus and interacting UBR4 with its C-terminus (Hong et al., 2015). A similar mechanism involving interaction with an active E3 ligase could potentially underlie the function of UBR4 in ERAD. However, UBR4 depletion was in fact found to cause increased ubiquitination of opsin-degron (Figure 3.8A and 3.8B). In relation to ubiquitination of opsin-degron, it was found that UBR4 knockdown increased or stabilised the association between BAG6 and opsin-degron (Figure 3.9A and 3.9C), and between BAG6 and RNF126 (Figure 3.9B and 3.9D). Although the E3 ligase responsible for opsin-degron ubiquitination has not been identified, RNF126 is the BAG6-associated E3 ligase shown to play a role in mislocalised protein degradation (Rodrigo-Brenni et al., 2014). Hence, it is possible that UBR4 knockdown increases ubiquitination of opsin-degron by enhancing opsin-degron- BAG6-RNF126 association. Other main ERAD E3 ligases such as HRD1, E3 ubiquitin-protein ligase MARCH6 (MARCH6), E3 ubiquitin-protein ligase RNF5 (RNF5), E3 ubiquitin-protein ligase RNF139 (RNF139), E3 ubiquitin-protein ligase RNF170 (RNF170) or E3 ubiquitin-protein ligase TM129 (TMEM129) (Smith et al.,

2011; van den Boomen and Lehner, 2015) may well contribute to opsin-degron degradation, in particular gp78 which interacts with BAG6 (Xu et al., 2013; Zhang et al., 2015). It will be interesting to determine if UBR4 increases association of opsin- degron and/or BAG6 with these. Indeed, the global increase in levels of cellular polyubiquitin conjugates observed with UBR4 depletion (Figure 3.7A and 3.7B) suggested that UBR4 plays a more general role in regulating ubiquitin homeostasis. This could be related to enhanced ubiquitination, reduced de-ubiquitination and/or reduced degradation of ubiquitinated proteins.

A model illustrating how UBR4 fits into BAG6-mediated ERAD is proposed: UBR4 is a very large protein that might hinder BAG6 interaction with other proteins; when UBR4 is depleted, BAG6 has increased interactions with its substrate and E3 ligase, thus enhancing ubiquitination of the substrate (Figure 3.10). BAG6 was thought to recruit different E3 ligases when engaging different substrates and so UBR4 might have a role in determining which E3 ligase is recruited to BAG6 but it is also possible that UBR4 only regulates a subset of BAG6-E3 ligase interaction. Hyper-ubiquitination caused by depletion of UBR4 has precluded opsin-degron degradation and the nature of the hyper-ubiquitin chain formed might contribute to the delayed in degradation. Another possible explanation would be that some deubiquitinating enzymes required for removal of ubiquitin prior to degradation are not recruited or active when UBR4 is depleted. It will be interesting to investigate if hyper-ubiquitinated opsin-degron is integrated into the ER membrane or is extracted into the cytosol. ERAD substrates undergo two rounds of ubiquitination and deubiquitination (Liu and Ye, 2012; Olzmann et al., 2013; Wang et al., 2013). First round of ubiquitination occurs prior to retrotranslocation to allow recognition by p97, then substrate is deubiquitinated to be threaded through p97 central cavity and extracted from the ER. Substrate is then ubiquitinated again to allow proteasome targeting and deubiquitinated for degradation. Conceivably, hyper-ubiquitination could prevent effective deubiquitination and block extraction by p97 or degradation by proteasome. One of the limitations/caveats of this work is that some experiments were only done twice, thus could not assess statistic. Also, the lack of effect of BAG6 on opsin-degron ubiquitination has to be looked into. Future work will aim at assessing the substrate specificity of UBR4 in ERAD and understanding how UBR4 causes hyper-ubiquitination that is restricted for degradation.

Figure 3.10: Proposed model on UBR4 function in ERAD in correlation with BAG6. BCL-2 associated athanogene 6 (BAG6) captures opsin-degron then recruits a yet to identify E3 ubiquitin ligase to ubiquitinate opsin-degron for degradation at the proteasome. Proteasome contains stacked rings of heptameric proteolytic subunits (blue) and hexameric regulatory subunits (purple). In the absence of E3 ubiquitin-protein ligase UBR4 (UBR4), BAG6 - substrate (opsin-degron) - E3 ligase complex increases interactions among themselves and causes hyper-ubiquitination of opsin-degron. However, increased polyubiquitination has slowed down opsin-degron degradation through an unknown mechanism. ER, endoplasmic reticulum; UBXD8, UBX domain-containing protein 8; gp78, E3 ubiquitin-protein ligase AMFR; Ub, ubiquitin; p97, 15S Mg(2+)-ATPase p97 subunit; DUB, deubiquitinating enzyme.

4 Novel role of BAG6 in the UPR

4.1 Introduction

BAG6 has been implicated in a wide range of functions, but recent work has focused on its role in cytosolic protein quality control. BAG6 has the ability to bind exposed hydrophobic stretches and this is likely to underlie its capability contributing to the quality control of different types of unfolded proteins in the cytosol (Hessa et al., 2011; Tanaka et al., 2016). In these pathways, BAG6 is suggested to have a ‘holdase’ activity to minimise aggregation of misfolded proteins containing hydrophobic sequences and also to facilitate ubiquitination and thus proteasomal degradation. Identification of endogenous BAG6 substrates has become an important goal to better understand role of BAG6 in cytosolic protein quality control. A targeted approach was used, whereby the literature was searched for cytoplasmic proteins that contain stretches of hydrophobic residues, since this is known to be a feature recognised by BAG6 (Hessa et al., 2011; Tanaka et al., 2016). The unspliced form of the transcription factor XBP1 was identified as a strong candidate as it is known to possess a hydrophobic region at its C-terminus (Yanagitani et al., 2009; 2011).

XBP1 is a key component of the first discovered and most conserved UPR pathway, also the only UPR pathway presents in Saccharomyces cerevisiae (Brodsky and Skach, 2011; Wu et al., 2014). IRE1 is an ER transmembrane endoribonuclease which acts as a proximal sensor for ER stress. IRE1 is activated by homo- oligomerisation and auto-phosphorylation upon detection of unfolded proteins in the ER lumen. Activated IRE1 cleaves membrane-associated XBP1 mRNA to remove the 26 bases intron and the cleaved mRNAs are ligated by the tRNA-splicing ligase, RtcB (Yoshida et al., 2001; Yanagitani et al., 2009; Baltz et al., 2012; Lu et al., 2014). This cleavage resulted in a frame-shift displacement of the C-terminal region of XBP1 mRNA, encoding a protein product of 376 amino acids instead of 261 amino acids for the unspliced protein (Figure 4.1A). Under non-stressed condition, unspliced XBP1 protein is expressed at low levels and gets rapidly degraded via the proteasome (Yoshida et al., 2001; 2006). When splicing occurs, spliced XBP1 is translated with an activation domain at the C-terminus for its activity as a 87

Figure 4.1: (A) Schematic representation of human XBP1 protein domains. Unspliced and spliced X-box-binding protein 1 (XBP1) proteins share common N- terminus containing the first hydrophobic region (HR1) and a basic-region leucine zipper motif for homo- and hetero-oligomerisation. The C-terminus of both proteins is distinct; with unspliced XBP1 containing the second hydrophobic region (HR2) and the spliced protein comprises an activation domain. K, lysine residues important for ubiquitination. (B) Model illustrating the splicing event of XBP1 mRNA under ER stress. During endoplasmic reticulum (ER) stress, translation of the XBP1 nascent chain polypeptide stalls after the HR2 and the nascent chain binds the ER membrane through the HR2 and brings with it the ribosome-nascent chain complex for mRNA splicing by the activated serine/threonine-protein kinase/endoribonuclease IRE1 (IRE1). P, phosphorylation. (Adapted from Yanagitani et al., 2009)

88 transcription factor (Yoshida et al., 2001; 2006) (Figure 4.1A).

The UPR is an essential homeostatic pathway and it is implicated in many diseases (Walter and Ron, 2011), hence understanding how it is regulated is really important. The RNase IRE1 responsible for generating the active/spliced XBP1 is located on the ER membrane; therefore mechanism is required to bring the unspliced XBP1 mRNA to the ER membrane for it to undergo cleavage. This is thought to be achieved by the hydrophobic region of the unspliced XBP1 protein which interacts with the ER membrane (Yanagitani et al., 2009; 2011). There are two reported hydrophobic regions in unspliced XBP1 (HR1 and HR2) but only HR1 is shared among the unspliced and spliced XBP1 (Yanagitani et al., 2009) (Figure 4.1A). Yanagitani et al. (2009, 2011) have proposed a model whereby translation of XBP1 mRNA is paused after the hydrophobic HR2 region has emerged from the ribosome exit tunnel, and the whole mRNA-ribosome-nascent chain complex is targeted to the ER by the HR2, bringing the XBP1 mRNA into contact with the IRE1 RNase (Figure 4.1B). Current evidence is conflicting, with one study reporting that unspliced XBP1 protein is integrated into the ER membrane and subsequently degraded via an ERAD-like pathway (Chen et al., 2014), whilst others conclude that unspliced XBP1 protein is delivered to the ER by signal recognition particle but is then rejected by the translocon resulting in mislocalisation to the cytosol and proteasomal degradation (Plumb et al., 2015; Kanda et al., 2016).

In addition to targeting XBP1 mRNA to the ER, the unspliced XBP1 protein plays an important role in regulating levels of the active transcription factor, spliced XBP1. During recovery from ER stress, unspliced XBP1 protein forms a complex with spliced XBP1 protein through the basic-region leucine zipper motif and promotes its degradation at the proteasome to terminate UPR (Yoshida et al., 2006). Therefore, unspliced XBP1 protein is a negative regulator of the active XBP1 transcription factor, making it important to understand the mechanisms that regulate unspliced XBP1 protein levels. BAG6 is an attractive candidate for mediating protein quality control, being associated with the ribosome (Mariappan et al., 2010) and the ER (Claessen and Ploegh, 2011; Wang et al., 2011; Xu et al., 2013) and having the capacity to bind exposed hydrophobic sequences (Hessa et al., 2011; Tanaka et al., 2016).

4.2 BAG6 interacts with XBP1u

In order to test whether BAG6 interacts with unspliced XBP1 protein (XBP1u), the ability of overexpressed BAG6 to re-localise substrates to the nucleus was used as an initial readout for BAG6 interaction. Endogenous BAG6 localises to the nucleus and the cytoplasm (Kwak et al., 2008; Kämper et al., 2012). However, overexpressed BAG6 localises to the nucleus due to endogenous levels of TRC35 failing to mask the nuclear localisation signal of BAG6 (Wang et al., 2011), and has been shown to cause re-localisation of substrate proteins to the nucleus (Leznicki et al., 2010; Leznicki et al., 2013; Figure 3.2D). An N-terminally tagged recombinant

XBP1, HA3-XBP1∆u; hereafter named HA-XBP1u, with a silent nucleotide substitution to prevent splicing of XBP1 mRNA by IRE1 was used (Yanagitani et al., 2011). HeLa cells were co-transfected with plasmids encoding HA-XBP1u and V5- tagged BAG6 variants. Overexpressed BAG6-V5 showed clear nuclear localisation as expected (Figure 4.2A, panel 1). However, HA-XBP1u was also predominantly detected in the nucleus regardless of BAG6 (Figure 4.2A, panel 1), meaning that the nuclear re-localisation assay could not be used as a readout for interaction. As an alternative, BAG6 with a mutation in the nuclear localisation signal which localises to the cytoplasm (Payapilly and High, 2014) was used to test whether this variant of BAG6 could re-localise HA-XBP1u from the nucleus to the cytoplasm. BAG6∆NLS- V5 was seen in the cytoplasm (Figure 4.2A, panel 2) and was able to re-localise HA- XBP1u to the cytoplasm (Figure 4.2A, panel 2, white arrow), indicating a possible interaction between BAG6 and XBP1u in cells.

In order to confirm this potential interaction biochemically, HeLa cells stably expressing BAG6-V5 under an inducible promoter were transiently transfected with plasmid encoding HA-XBP1u, and lysates subjected to co-immunoprecipitation with anti-V5 agarose. Since BAG6-substrate interaction could be sensitive to detergents, cell lysates were prepared with Triton X-100 and digitonin. In addition, cells were treated with proteasome inhibitor bortezomib prior to co-immunoprecipitation to stabilise HA-XBP1u and maximise potential interaction as shown for opsin-degron (Figure 3.5B). BAG6 pulldown was efficient and specific; no BAG6 was left in the unbound fraction (Figure 4.2B, lanes 6 and 12) and BAG6 was only detected in immunoprecipitates from induced cells (Figure 4.2B, lanes 2, 4, 8 and 10). Comparable amounts of BAG6 were immunoprecipitated from Triton X-100 and

Figure 4.2: BAG6 re-localises and stabilises XBP1u. (A) HeLa cells were co- transfected with plasmids encoding HA-XBP1u and BAG6 variants, and after 24 hours fixed with formaldehyde and stained with HA and V5 antibodies followed by Alexa Fluor®-conjugated secondary antibodies. White arrows show the re-localised HA-XBP1u. This is a representative image from at least 2 independent experiments. Scale bar = 10 µm. (B) HeLa cells were transfected with plasmid encoding HA- XBP1u and induced to express Bag6-V5 with tertracycline (Tet) or left non-induced as indicated. Cells were then treated with bortezomib for 4 hours before lysis in Triton X-100 (Tx-100) or digitonin buffer. Cell lysate was incubated with V5- agarose and immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). The heavy chain (HC) and light chain (LC) of the immunoprecipitation antibody are indicated. At least two independent experiments were analysed. (C) HeLa cells were transfected with plasmid encoding HA-XBP1u and then treated with bortezomib for 4 hours before lysis in digitonin buffer. Cell lysate was split into two equal portions and incubated with either a control antibody (anti-V5) or anti-BAG6. Immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). The heavy chain (HC) and light chain (LC) of the immunoprecipitation antibody are indicated. At least two independent experiments were analysed. (D) HeLa cells were co- transfected with plasmids encoding FLAG-XBP1u and empty vector/BAG6-V5, and lysates analysed by SDS-PAGE and immunoblotting with anti-FLAG, anti-V5 and anti-tubulin followed by IRDye®-conjugated secondary antibodies. At least two independent experiments were analysed.

C D

92 digitonin lysates (Figure 4.2B, lanes 4 and 10). Transiently expressed HA-XBP1u was detected as a band between 32 and 46 kDa by blotting with anti-HA antibody. Blotting immunoprecipitated material revealed that HA-XBP1u specifically co- precipitated with BAG6-V5, as indicated by the band observed in anti-V5 immunoprecipitates from cells induced to express BAG6-V5 (Figure 4.2B, lanes 4 and 10) but not from non-induced cells. (Figure 4.2B, lanes 3 and 9). Digitonin buffer appeared to preserve the interaction between BAG6 and XBP1 better as more HA-XBP1u was co-immunoprecipitated (Figure 4.2B, lane 10 vs 4), and was therefore used in all future experiments.

In order to examine the interaction of XBP1 with endogenous BAG6, cells were transiently transfected with another XBP1u construct, FLAG8His3-XBP1∆u- HA (hereafter named FLAG-XBP1u), and lysates immunoprecipitated with antibody specific for BAG6 (N-terminus) or a control IgG. Although FLAG-XBP1u was detected in both control and BAG6 immunoprecipitates, the intensity of the FLAG- XBP1u band in BAG6 immunoprecipitate was more intense (Figure 4.2C, lane 3 vs 2), consistent with an interaction between XBP1u and BAG6. Together, these results provide further evidence that BAG6 does indeed interact with XBP1u.

Overexpressed BAG6 stabilised its substrate due to disruption of the BAG6 complex stoichiometry leading to reduced BAG6 function (Figure 3.2C). HeLa cells were co-transfected with empty vector or plasmids encoding BAG6-V5 and FLAG- XBP1u, and it was clearly shown that exogenous BAG6 increased the level of FLAG-XBP1u detected (Figure 4.2D, lane 2 vs 1), further suggesting a novel interaction between BAG6 and XBP1u.

In order to determine whether proteasome inhibition is required to preserve the BAG6-XBP1u interaction, cells co-expressing BAG6-V5 and HA-XBP1u were left untreated or treated with bortezomib for 4 hours prior to co-immunoprecipitation. Increased HA-XBP1u steady state level in cells treated with bortezomib (Figure 4.3, lanes 1 and 2 vs 3) was consistent with proteasomal degradation of XBP1u, but no obvious effect of proteasome inhibition on levels of BAG6 (Figure 4.3, lanes 2 and 3). As seen previously, HA-XBP1u was detected in anti-V5 immunoprecipitates from cells induced to express BAG6 but not from non-induced cells when cells were treated with bortezomib (Figure 4.3, lanes 4 and 5). However, if cells were not

Figure 4.3: BAG6 co-purifies XBP1u with proteasome inhibition. HeLa cells stably expressing BAG6-V5 under an inducible promoter were transfected with plasmid encoding HA-XBP1u and treated with tetracycline (Tet) and/or bortezomib (Bz) for 4 hours as indicated before lysis in digitonin buffer. Cell lysate was incubated with V5-agarose and immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). At least two independent experiments were analysed.

94 treated with bortezomib, virtually no HA-XBP1u was co-immunoprecipitated, despite BAG6-V5 being efficiently immunoprecipitated (Figure 4.3, lane 6). This loss of HA-XBP1u in BAG6-V5 immunoprecipitates might be due in part to the lower levels of HA-XBP1u in the absence of proteasome inhibitor. However, quantification revealed that whilst HA-XBP1u levels in lysate were 65 % lower in absence of proteasome inhibitor, HA-XBP1u was virtually undetectable in the BAG6-V5 immunoprecipitates, suggesting that the interaction was increased under conditions of proteasome inhibition. Hence, it may be inferred that BAG6 interacts with XBP1 that is en route to degradation.

4.3 BAG6 is required for efficient XBP1u degradation

In order to determine whether BAG6 contributes to turnover of XBP1u, conditions for monitoring degradation of the unspliced protein were established. HA- XBP1u was transiently expressed in HeLa cells which were then either treated with bortezomib or left untreated. Treatment with bortezomib caused an increase in the steady state level of HA-XBP1u (Figure 4.4A, lane 2 vs 1), consistent with proteasomal degradation of the unspliced XBP1 protein (Yoshida et al., 2001; 2006).

Cycloheximide chase assays were carried out to measure the rate of HA- XBP1u degradation. Four-hour treatment with bortezomib caused a dramatic increase in levels of HA-XBP1u (Figure 4.4A). To minimise differences in the starting amount of HA-XBP1u prior to the cycloheximide chase, bortezomib was added to transfected cells for only an hour before cycloheximide addition, and included throughout the chase. There was no obvious difference in levels of HA- XBP1u at time 0 in untreated or bortezomib-treated cells (Figure 4.4B, lanes 1 and 4). In the absence of proteasome inhibitor, approximately 65 % of HA-XBP1u was lost after two hours (Figure 4.4B, lanes 1-3 and 4.4C); with an estimated half-life of 1.2 hours based on exponential one-phase decay (Figure 4.4D). Treatment of cells with proteasome inhibitor caused a significant delay in HA-XBP1u degradation (Figure 4.4B, lanes 4-6 and 4.4C) with half-life increased to 2.5 hours (Figure 4.4D). These results are consistent with previous reports of endogenous XBP1u being rapidly degraded by the proteasome (Yoshida et al., 2001; 2006) and overexpressed XBP1u having a half-life of approximately 1 hour (Yoshida et al., 2001).

A B

C 100 Untreated Bortezomib *

50 Remaining XBP1u (%) XBP1u Remaining 0 0.0 0.5 1.0 1.5 2.0 CHX chase (h)

D Untreated Bortezomib One phase decay Y0 (%) = 100 = 100 Plateau (%) = 0 = 0 K (h-1) 0.58 0.28 Half Life (h) 1.2 2.5

Figure 4.4: Proteasome inhibition delays XBP1u degradation. (A) HeLa cells transfected with plasmid encoding HA-XBP1u were treated with bortezomib (Bz) as indicated for 4 hours and lysates analysed by SDS-PAGE and immunoblotting with anti-HA and anti-tubulin followed by IRDye®-conjugated secondary antibodies. (B) HeLa cells transfected with plasmid encoding HA-XBP1u were treated with bortezomib (Bz) as indicated for 1 hour. Cells were then treated with cycloheximide (CHX). Cells were harvested 0, 1 and 2 hours after cycloheximide treatment and lysates analysed by immunoblotting. (C) The intensity of the HA (XBP1u) signal was quantified using Image Studio Lite software and normalised to tubulin, then expressed relative to the amount of HA-XBP1u at 0 hour (set to 100%). Data shown are the mean ± SEM of 4 independent experiments. *p<0.05 using two-way ANOVA. (D) An estimate of HA-XBP1u half-life under different treatments with one phase decay curve.

In order to examine the impact of depleting BAG6 on HA-XBP1u turnover, cycloheximide chase assays were performed on cells transfected with non-targeting siRNA or siRNA targeting the 3’ untranslated region of BAG6. In cells treated with non-targeting siRNA, HA-XBP1u was rapidly lost following addition of cycloheximide (Figure 4.5A, lanes 1-3 and 4.5C), consistent with rapid degradation seen previously. BAG6 siRNA efficiently depleted BAG6 (Figure 4.5A, lane 4 vs 1). Notably, a significant delay in HA-XBP1u degradation was observed when BAG6 was depleted (Figure 4.5A, lanes 4-6, 4.5C and 4.5E). A second independent BAG6 siRNA targeting the coding region was used to deplete BAG6. This also caused a significant delay in HA-XBP1u degradation (Figure 4.5B, 4.5D and 4.5F). These results suggest that BAG6 contributes to the degradation of XBP1u. However, degradation of HA-XBP1u was not affected in cells depleted of UBR4 and RNF126 (Figure 4.6), suggesting that BAG6 involvement in XBP1 turnover is independent of UBR4 and RNF126 functions.

4.4 BAG6 interacts with the C-terminal hydrophobic region of full length XBP1u in order to promote degradation

The C-terminal hydrophobic region of XBP1u (HR2; residues 186 to 208) (Figure 4.1A) is recognised by signal recognition particle (Plumb et al., 2015; Kanda et al., 2016) and mediates ER targeting of both XBP1 protein and mRNA (Yanagitani et al., 2009; 2011) (Figure 4.1B). To test whether BAG6 interacts with XBP1u by recognising this hydrophobic region, a variant of HA-XBP1u lacking HR2 was generated (HA-XBP1u-∆HR2) and its association with endogenous BAG6 examined. HeLa cells were transfected with full length or HA-XBP1u-∆HR2 constructs, treated with proteasome inhibitor, lysed in digitonin-containing buffer and endogenous BAG6 was immunoprecipitated using anti-BAG6 antibody. As seen previously, full length HA-XBP1u co-immunoprecipitated with endogenous BAG6 but not with the control antibody (Figure 4.7A, lane 5 vs 4). In contrast, HA-XBP1u- ∆HR2 did not co-immunoprecipitate with endogenous BAG6 above the control level (Figure 4.7A, lane 7 vs lane 6), despite being expressed at comparable levels to the full length HA-XBP1u (Figure 4.7A, lanes 1 and 2) and similar BAG6 immunoprecipitation efficiency (Figure 4.7A, lanes 5 and 7). These results provide

A B

C D

100 100 siNT * siNT * siBAG6 siBAG6/2 * *

50 50 Remaining XBP1u (%) XBP1u Remaining Remaining XBP1u (%) XBP1u Remaining 0 0 012 012 CHX chase (H) CHX chase (H) E F

siNT siBAG6 siNT siBAG6/2 One phase decay One phase decay Y0 (%) = 100 = 100 Y0 (%) = 100 = 100 Plateau (%) = 0 = 0 Plateau (%) = 0 = 0 -1 K (h ) 0.49 0.22 K (h-1) 0.40 0.18 Half Life 1.4 3.1 Half Life 1.7 3.9 (h) (h)

Figure 4.5: BAG6 knockdown delays XBP1u degradation. (A, B) HeLa cells were transfected with the indicated siRNAs and plasmid encoding HA-XBP1u. Cells were then treated with cycloheximide (CHX). Cells were harvested 0, 1 and 2 hours after cycloheximide treatment and lysates analysed by immunoblotting. siNT, non- targeting siRNA; Kd, knockdown. (C, D) The intensity of the HA (XBP1u) signal was quantified using Image Studio Lite software and normalised to tubulin, then expressed relative to the amount of HA-XBP1u at 0 hour (set to 100%). Data shown are the mean ± SEM of 3 independent experiments. *p<0.05 using two-way ANOVA. (E, F) An estimate of HA-XBP1u half-life under different treatments with one phase decay curve.

B 100 siNT siUBR4 siRNF126

50 Remaining XBP1u (%) XBP1u Remaining 0 012 CHX chase (H)

C siNT siUBR4 siRNF126 One phase decay Y0 (%) = 100 = 100 = 100 Plateau (%) = 0 = 0 = 0 K (h-1) 0.51 0.62 0.52 Half Life (h) 1.4 1.1 1.3

Figure 4.6: UBR4 and RNF126 knockdown have no effect on XBP1u degradation. (A) HeLa cells were transfected with the indicated siRNAs and plasmid encoding HA-XBP1u. Cells were then treated with cycloheximide (CHX). Cells were harvested 0, 1 and 2 hours after cycloheximide treatment and lysates analysed by immunoblotting. siNT, non-tageting siRNA; Kd, knockdown. (B) The intensity of the HA (XBP1u) signal was quantified using Image Studio Lite software and normalised to tubulin, then expressed relative to the amount of HA-XBP1u at 0 hour (set to 100%). Data shown are the mean ± SEM of 3 independent experiments. p>0.05 using two-way ANOVA. (C) An estimate of HA-XBP1u half-life under different treatments with one phase decay curve.

99 evidence that BAG6 interacts with XBP1u via the hydrophobic HR2. Since XBP1u contains another hydrophobic HR1 region, XBP1u variant lacking the HR1 region could be tested to assess the specificity of BAG6 interaction with hydrophobic region.

Since the hydrophobic region HR2 of XBP1u was required for BAG6 binding, and BAG6 was required for optimal degradation of XBP1u, the next question was whether XBP1u lacking HR2 was more stable than the full length protein. To address this, the degradation of HA-XBP1u-∆HR2 and HA-XBP1u was examined using cycloheximide chase assays. These revealed a higher steady state level of HA- XBP1u-∆HR2 at the start of the chase (Figure 4.7B, lane 4 vs 1), and less HA- XBP1u-∆HR2 was lost over time compared to HA-XBP1u (Figure 4.7B and 4.7C). These results show that the HR2 region of XBP1u promotes degradation of XBP1u, consistent with published work showing that the C-terminal region of XBP1u is responsible for its rapid degradation (Yoshida et al., 2001; 2006). However, the data did not reach statistical significant, hence more experimental repeats should be performed to test the hypothesis.

To further test this model, the effect of BAG6 siRNA on degradation of HA- XBP1u-∆HR2 was examined. In direct contrast to what was seen with full length XBP1u, depletion of BAG6 had no impact on the rate at which HA-XBP1u-∆HR2 was degraded following addition of cycloheximide (Figure 4.7D and 4.7E). Thus the ability of BAG6 to promote degradation of XBP1u is dependent on the presence of the C-terminal hydrophobic region of XBP1u. This is an important result as it shows that the inhibitory effect of BAG6 siRNA on XBP1u degradation is not due to a more general effect of BAG6 depletion on cellular quality control processes. Together, these data suggest a model whereby BAG6 recognises XBP1u by binding to the exposed HR2 and subsequently promotes its proteasomal degradation.

A peptide sequence locating after the HR2 region is responsible for stalling the translation of XBP1u and allowing targeting to the ER membrane. Stalling during translation of XBP1 is required to allow sufficient time for recognition of nascent XBP1 by signal recognition particle and subsequent targeting to the ER (Yanagitani et al., 2011; Kanda et al., 2016). Since BAG6 is also proposed to associate with the ribosome (Mariappan et al., 2010) and recognise proteins with hydrophobic sequences that fail to insert into the ER (Hessa et al., 2011), it was of interest to test

100

Figure 4.7: BAG6 cannot interact with XBP1 variant lacking HR2. XBP1 HR2 mutant is more stable and does not require BAG6 for degradation. (A) HeLa cells were transfected with HA-XBP1u variants and then treated with bortezomib for 4 hours before lysis in digitonin buffer. Cell lysate was incubated with either a control antibody (-) or anti-BAG6 antibody (+). Immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). The heavy chain (HC) and light chain (LC) of the immunoprecipitation antibody are indicated. At least two independent experiments were analysed. (B) HeLa cells were transfected with plasmids encoding HA-XBP1u or HA-XBP1u-∆HR2. Cells were then treated with cycloheximide (CHX). Cells were harvested 0, 1 and 2 hours hours after cycloheximide treatment and lysates analysed by immunoblotting. (C, E) The intensity of the HA (XBP1u) signal was quantified using Image Studio Lite software and normalised to tubulin, then expressed relative to amount of respective protein at 0 hour (set to 100%). Data shows the mean ± SEM of 3 independent experiments. p>0.05 using two-way ANOVA. (D) As in (B), except that cells were transfected with the indicated siRNAs and plasmid encoding HA-XBP1u-∆HR2. siNT, non-targeting siRNA; Kd, knockdown. (F) HeLa cells were co-transfected with plasmids encoding BAG6-V5 and empty vector(-)/FLAG-XBP1u-HA(+) and then treated with bortezomib for 4 hours before lysis in digitonin buffer. Cell lysate was incubated with anti-HA. Immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). The heavy chain (HC) and light chain (LC) of the immunoprecipitation antibody are indicated. At least two independent experiments were analysed.

101

B C

100 XBP1u deltaHR2

50 Remaining XBP1 (%) XBP1 Remaining 0 012 CHX chase (H)

D E

120 siNT 100 siBAG6 80

RemainingdeltaHR2 (%) 0 012 CHX chase (H)

102 whether stalling would have any impact on binding of BAG6 to XBP1. A tryptophan (W) residue at position 256 in XBP1 is a critical part of the stalling sequence and substitution with alanine (W256A) inhibits stalling and ER targeting of XBP1u (Yanagitani et al., 2011; Kanda et al., 2016). Thus, plasmids encoding HA-XBP1u- W256A was transfected into HeLa cells, treated with proteasome inhibitor, lysed in digitonin buffer and endogenous BAG6 immunoprecipitated. BAG6 co-purified HA- XBP1u- W256A as well as wild type HA-XBP1u (Figure 4.7A, lane 9), indicating that ribosomal stalling is not required for BAG6-XBP1u interaction.

BAG6 could potentially interact co-translationally with the nascent XBP1u at the ribosome-nascent chain complex, with the full length protein released into the cytosol, or both. To address this, FLAG-XBP1u-HA which has a HA-tag at the C- terminus was used to ensure that anti-HA immunoprecipitation would only isolate the full length XBP1u. Here, HeLa cells were co-transfected with plasmid encoding BAG6-V5 and empty vector or plasmid encoding FLAG-XBP1u-HA. Cells were treated with proteasome inhibitor, lysed in digitonin buffer and immunoprecipitated with anti-HA antibody. BAG6 was co-immunoprecipitated with full length FLAG- XBP1u-HA (Figure 4.7F, lane 4) but was not seen in control immunoprecipitates (Figure 4.7F, lane 3). Together, these results suggest that BAG6 interacts with full length XBP1u via the C-terminal hydrophobic region of XBP1u, and does not require translational stalling of XBP1u on the ribosome.

4.5 BAG6 depletion impairs XBP1u ubiquitination

In order to better understand how loss of BAG6 caused a delay in XBP1 degradation, the impact of BAG6 siRNA on XBP1u polyubiquitination was examined. HeLa cells were treated with BAG6 or non-targeting siRNA, then co- transfected with plasmids encoding FLAG-XBP1u and HA-ubiquitin to facilitate detection of ubiquitinated XBP1u. Cells were treated with proteasome inhibitor in order to prevent degradation of polyubiquitinated XBP1u, prior to lysis under denaturing conditions and immunoprecipitation with anti-FLAG antibody or mouse IgG. Looking at the Input, it appeared that BAG6 knockdown caused a slight reduction of total polyubiquitinated proteins (Figure 4.8A, lane 3), in agreement with a previous report (Payapilly and High, 2014). Control immunoprecipitation

103

B 1.0

0.5

0.0 6

IP ubiquitin relativeto FLAG (au) T iN G s A iB s

Figure 4.8: BAG6 knockdown decreases XBP1u ubiquitination. (A) HeLa cells were transfected with siRNAs for 48 hours then co-transfected with plasmids encoding FLAG-XBP1u and HA-ubiquitin for 24 hours. Cells were then treated with bortezomib for 4 hours before lysis in SDS containing buffer. Cell lysate was incubated with mouse IgG (Control) or anti-FLAG antibody and immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting together with 10 % of the cell lysate (Input). The heavy chain (HC) and light chain (LC) of the immunoprecipitation antibody are indicated. siNT, non-targeting siRNA; Ub, ubiquitin. (B) The intensity of the HA (ubiquitin) signal was quantified using Image Studio Lite software and normalised to FLAG then expressed relative to siNT (set to 1). Data shows the mean ± SEM of 3 independent experiments. p>0.05 using two-tailed t-test.

104 isolated only very low levels of HA, showing that HA-ubiquitin did not associate non-specifically with the beads (Figure 4.8A, lane 4). In contrast, FLAG-XBP1u immunoprecipitates contained a range of high molecular weight anti-HA reactive species, most likely representing ubiquitinated XBP1u, given the denaturing conditions used for the immunoprecipitation (Figure 4.8A, lane 5). Less HA- ubiquitin was detected in the FLAG-XBP1u immunoprecipitates from BAG6 siRNA treated cells (Figure 4.8A, lane 6). Even though there was less FLAG-XBP1u immunoprecipitated from these cells, after normalising HA-ubiquitin signals to FLAG-XBP1u signals, nearly 40 % less HA-ubiquitin was detected in the FLAG immunoprecipitates from BAG6 siRNA treated cells (Figure 4.8B). These results might suggest that loss of BAG6 reduces XBP1u ubiquitination, and this might underlie the effect of BAG6 depletion to inhibit XBP1u degradation. However, calculated p-value was 0.12, indicating that 12 % of the same observation might occur by chance, hence more experimental repeats should be performed before coming to a conclusion.

4.6 Discussion

Mammalian cells have evolved to possess at least 3 UPR signalling pathways with proximal sensors named PERK, ATF6 and IRE1, with XBP1 as one of the key components of the IRE1 pathway. Under conditions of ER stress, ribosome-stalled XBP1 recruits its ribosome-nascent chain complex to the ER membrane for splicing of its mRNA by activated IRE1α (Yoshida et al., 2001; Yanagitani et al., 2009; Yanagitani et al., 2011). Spliced mRNAs are ligated and the new mRNA has a reading frame shift that encodes for a larger spliced XBP1 (XBP1s) (Yoshida et al., 2001). The replaced C-terminus of XBP1s contains an activation domain for XBP1s function in downstream gene expression of chaperones and ERAD components to alleviate ER stress (Yoshida et al., 2001). Under non-stressed conditions, XBP1u is expressed at a low level and is rapidly degraded (Yoshida et al., 2001). However, increased expression of XBP1u is observed during recovery from ER stress (Yoshida et al., 2006). XBP1u forms a hetero-oligomer with the transcription factor XBP1s and as a result of the nuclear exclusion signal and degradation signal at the C- terminal end of XBP1u, the complex is exported from the nucleus and degraded by

105 the proteasome (Yoshida et al., 2006). Thus, XBP1u acts as a negative regulator for XBP1s, helping to ensure UPR signalling is attenuated when ER homeostasis is restored. It is therefore important to understand how XBP1u level is regulated in cells since prolonged XBP1 signalling may have detrimental effect on recovery from stress. In this study, BAG6 was shown to interact with XBP1u through the hydrophobic HR2 region and promote XBP1u degradation potentially by regulating XBP1u ubiquitination.

Both exogenous and endogenous BAG6 could co-immunoprecipitate XBP1u (Figure 4.2B and 4.2C) and the interaction was stabilised with proteasome inhibition (Figure 4.3). BAG6 interaction with its substrate opsin-degron was also reported to require proteasome inhibition for detection (Payapilly and High, 2014; Chapter 3), consistent with XBP1u being a proteasomal substrate (Yoshida et al., 2001; 2006). BAG6 did not interact with an XBP1u variant lacking the hydrophobic HR2 region (Figure 4.7A). The XBP1u HR2 region is important in targeting to the ER membrane to allow cleavage of XBP1u mRNA by IRE1 upon UPR induction (Yanagitani et al., 2009). However, this region is not recognised as a transmembrane domain by the Sec61 translocon and is thus not integrated into the ER membrane (Plumb et al., 2015; Kanda et al., 2016). Exposure of this hydrophobic region could be problematic for cells due to inappropriate interactions (Kim et al., 2013; Koldewey et al., 2017). Degradation of XBP1u was delayed when BAG6 was depleted (Figure 4.5) whereas the HR2 mutant that did not interact with BAG6 was much more stable (Figure 4.7B and 4.7C). The most recent membrane protein triage model proposed that SGTA captures substrates through rapid on/off rate binding, then substrates get rapid ‘private’ transfer to TRC40 leading to targeting, with transfer to BAG6 for ubiquitination being dependent on dissociation from SGTA and recapture by BAG6 (Shao et al., 2017). Therefore, one possibility would be that HR2, targeted to the ER by signal recognition particle but then released by the translocon, is initially captured by SGTA but is not successfully transferred to TRC40 due to its properties and thus ends up being captured by BAG6. This possibility could be tested by depleting SGTA and looking at effect on BAG6 interaction/degradation of XBP1u. Alternatively, action of BAG6 to promote degradation of XBP1u may involve BAG6 acting separately from SGTA to recognise HR2. HR2 rejected by the translocon, which was reported to happen on more than 90 % of ER targeting XBP1u (Plumb et

106 al., 2015), could be transferred from signal recognition particle to BAG6. Under conditions of ER stress, the translocon rejects proteins targeted to the ER, and these proteins are handed over to Derlin proteins and then onto p97/BAG6 (Kadowaki et al., 2015). BAG6 could sit around the ER ready to mop up proteins with hydrophobic domains, as suggested for its role in ERAD (Claessen and Ploegh, 2011; Wang et al., 2011; Payapilly and High, 2014; Xu et al., 2013). Proposed model on BAG6 function in unfolded protein response was illustrated in Figure 4.9.

It was possible to detect ubiquitinated XBP1u and BAG6 knockdown reduced XBP1u ubiquitination (Figure 4.8). The importance of ubiquitination for XBP1u degradation was illustrated by the lack of degradation when two or three lysine residues were substituted for arginine residues (Lee et al., 2003). One of the lysine residues studied locates before the HR2 region while the other two locate after the HR2 (Lee et al., 2003) (Figure 4.1A). The E3 ligase(s) recruited by BAG6 favourably ubiquitinates lysine residues adjacent to regions of hydrophobicity in BAG6-interacting substrate (Rodrigo-Brenni et al., 2014), which would be the 3 lysine residues discussed above. These observations are consistent with the model proposed here that BAG6 might be a key player in XBP1u ubiquitination. The two lysine residues after the HR2 region locate within the XBP1u C-terminus thought to be the degradation domain, leading to a prediction whereby XBP1u residues 209-261 is important for controlling stability (Yoshida et al., 2006) because the region contains two BAG6-dependent ubiquitination sites. However, an HR2 deleted version of XBP1u that could not interact with BAG6 was still degraded (Figure 4.7B and 4.7C), albeit at a slower rate. This variant could be degraded by BAG6- independent route either proteasomal or non-proteasomal. Navon et al. (2010) suggested that XBP1u was also degraded through an ubiquitin-independent pathway by showing in-vitro data on XBP1u degradation with purified 20S proteasome in the absence of the ubiquitination machinery. XBP1u interacted with the proteasome directly, possibly by opening the proteasome gate and diffused into the catalytic core (Navon et al., 2010). The ability of XBP1u to be degraded by ubiquitination- dependent and -independent pathways has been used to explain the very rapid degradation of XBP1u.

BAG6 interacted with full length XBP1u, and there was no requirement for translational stalling (Figure 4.7A and 4.7F). Hence, it seems likely that BAG6 does

107

Figure 4.9: Proposed model on BAG6-mediated XBP1u degradation. Translating X-box-binding protein 1 (XBP1) is targeted by signal recognition particle (SRP) to the endoplasmic reticulum (ER) membrane but less than 10 % is inserted into the ER membrane (1) with the rest being rejected by the protein transport protein Sec61 and mislocalised into the cytosol (2). Cytosolic XBP1 might be captured by small glutamine-rich tetratricopeptide repeat-containing protein alpha (SGTA) and handed over to BAG6 or captured by BAG6 independent of SGTA for ubiquitination and subsequent degradation. Derlin proteins might capture SRP-engaged XBP1 to be handed over to 15S Mg(2+)-ATPase p97 subunit (p97) and BAG6 for degradation through the proteasome (3). Proteasome contains stacked rings of heptameric proteolytic subunits (blue) and hexameric regulatory subunits (purple). SRPR, signal recognition particle receptor; IRE1, serine/threonine-protein kinase/endoribonuclease IRE1; Ub, ubiquitin. (Adapted from Plumb et al., 2015)

108 not solely ubiquitinate XBP1u at the ribosome. There is some discrepancy in the literature regarding XBP1u localisation. Specifically, whether XBP1u is a soluble protein that becomes a peripheral membrane protein under conditions of ER stress (Yanagitani et al., 2009) or XBP1u is a single-pass transmembrane protein (Chen et al., 2014) is still unclear. XBP1u was also reported to localise to the nucleus (Kanda et al., 2016), as observed in this study (Figure 4.2A). BAG6 plays role in both cytoplasmic quality control (Minami et al., 2010; Hessa et al., 2011; Rodrigo-Brenni et al., 2014; Tanaka et al., 2016) and ERAD (Claessen and Ploegh, 2011; Wang et al., 2011; Payapilly and High, 2014; Xu et al., 2013), thus could contribute to degradation of XBP1u anchored in the ER membrane or released into the cytosol, as per models suggested above. If XBP1u is a cytoplasmic protein, degradation of XBP1u by BAG6 would resemble degradation of mislocalised proteins. However, siRNA mediated knockdown of the E3 ligase RNF126 which is known to ubiquitinate BAG6-associated model mislocalised proteins (Rodrigo-Brenni et al., 2014) had no effect on XBP1u degradation (Figure 4.6), indicating that BAG6 utilises other E3 ligases for XBP1u ubiquitination. If XBP1u is integrated into the ER membrane then it could be degraded through ERAD. Indeed, membrane integrated XBP1u was proposed to be subjected to ERAD involving cleavage by signal peptide peptidase (SPP) (Chen et al., 2014). SPP cleaved XBP1u within an unrecognized type II transmembrane domain and the peptide fragments were released for proteasomal degradation independent of p97 (Chen et al., 2014). Although XBP1u was stabilised by two SPP inhibitors, the cleaved species were almost undetectable (Chen et al., 2014), suggesting that XBP1u cleavage by SPP was minimal. Other ERAD E3 ligase especially gp78 which interacts with BAG6 (Xu et al., 2013; Zhang et al., 2015) might be involved in XBP1u ubiquitination and could be tested in the future.

XBP1u forms a complex with XBP1s and promotes XBP1s degradation at the proteasome (Yoshida et al., 2006). Therefore, increasing the abundance of XBP1u by reducing its degradation upon BAG6 depletion would be predicted to inhibit XBP1s signalling in the UPR. Indeed, a variant of XBP1u resistant to SPP cleavage have reduced BiP-reporter activity, a measurement of UPR induction (Chen et al., 2014). However, two other XBP1u mutants with increased stability were shown to promote transcription of XBP1s-specific and non-specific UPR target genes (Tirosh et al.,

109

2006). Both XBP1u mutants used were non-spliceable: the first mutant had a premature stop codon after the intron while the second mutant had two prolines important for secondary structure formation substituted (Tirosh et al., 2006). Hence, the increased expression of UPR genes observed with these two mutants might not represent the actual situation as the ability of these mutants to engage XBP1s was not tested. XBP1u also interacted with activated ATF6, another transcription factor of the UPR, and promoted proteasome-mediated degradation of ATF6, in a similar fashion to XBP1s (Yoshida et al., 2009). Upon ER stress, ATF6-alpha is transported to the Golgi apparatus to undergo proteolytic cleavage and the cleaved N-terminus then enters the nucleus to act as a potent transcriptional activator of chaperones and ERAD component genes (Haze et al., 1999; Ye et al., 2000). To test whether BAG6 regulates XBP1s functionally, qPCR on XBP1s-specific and non-specific UPR target genes would be performed on cells depleted of BAG6.

110

5 The use of BioID to identify potential BAG6 substrates and/or interacting factors

5.1 Introduction

Although BAG6 has been reported to recognise its substrate through regions of hydrophobic amino acids, BAG6 does not interact with all mislocalised proteins or aggregation-prone proteins in general (unpublished data from Swanton Lab). The minimum length of hydrophobic domain needed for BAG6 binding has not been defined and remains a key unanswered question. Mariappan et al. (2010) mutated a known BAG6 substrate Sec61β which contains a hydrophobic transmembrane domain of 21 amino acids, and showed that substitution of 3 of the hydrophobic amino acids to charged arginine residues effectively prevent the interaction with BAG6. Hessa et al. (2011) showed that fluorescent GFP proteins possessing hydrophobic signal sequences of 21 to 40 amino acids became substrates of BAG6- dependent ubiquitination in vitro, but it was unclear whether these proteins would become substrates for BAG6 mediated quality control in vivo. Furthermore, most studies of BAG6 function have utilised individual model substrates and little is known about its endogenous substrates. To better understand BAG6 substrate specificity determinant, an unbiased proteomic approach termed BioID was used to identify endogenous protein interactions in living cells based on proximity labelling (Roux et al., 2012) (Figure 5.1). This is a proximity labelling approach in which the protein of interest, here BAG6, is fused to a minimal biotin ligase (BirA) domain which catalyses covalent attachment of biotin moiety to primary amines, mainly the side chain of lysines, of proteins located in close proximity within ~ 10 nm (Kim et al., 2014) to BAG6. Biotin-labelled proteins can then be isolated on streptavidin- conjugated beads and analysed with mass spectrometry in order to identify BAG6 substrates. Many BAG6 substrates are likely to be targeted for degradation and thus interactions might be labile. Hence, biotinylation was also performed under conditions where degradation is blocked, with proteasome inhibitor treatment, which would stabilise some rapidly turnover proteins.

111

Figure 5.1: Model for application of BioID method. Biotin ligase (BirAWT) activates biotin (yellow-filled circle) and transfers reactive biotin (red-filled circle) to substrates with an acceptor sequence. Promiscuous form of biotin ligase (BirA) transfers reactive biotin to any substrates within the radius of 10 nm. BAG6, BCL2- associated athanogene 6. (Adapted from Kim and Roux, 2016)

112

Another important question worth addressing would be how BAG6 promotes ubiquitination of its various substrates and which other factors are involved? Overexpression of BAG6 was shown to promote ERAD substrate ubiquitination while BAG6 knockdown inhibits ubiquitination (Payapilly and High, 2014). Immunodepletion of BAG6 was shown to inhibit mislocalised protein ubiquitination in vitro and mislocalised PrP is stabilised by BAG6 siRNA in vivo (Hessa et al., 2011; Rodrigo-Brenni et al., 2014). The BAG6 UBL domain is generally thought to be required for BAG6 substrate ubiquitination (Minami et al., 2010; Hessa et al., 2011), possibly by recruiting E3 ubiquitin ligases to the substrates. Two E3 ligases, RNF126 and gp78 have been shown to bind to BAG6 and may play a role in ubiquitinating mislocalised proteins and ERAD substrates, respectively. Whether BAG6 directs different ligases towards different clients or if there is overlap in their function is still not known. RNF126 does not seem to promote ubiquitination of XBP1u (Chapter 4), suggesting the formal assumption might be true. N-terminally truncated BAG6 with the ubiquitin-like domain deleted was used in addition to full length BAG6 to evaluate domain needed for substrate binding and ubiquitination.

5.2 Generation and characterisation of BAG6-myc-BirA construct

The sequence encoding myc-BirA (R118G, a highly promiscuous form that demonstrates undiscriminating biotinylation activity) (Choi-Rhee et al., 2004) was sub-cloned into pcDNA5/FRT/TO with a multiple cloning site at the 5’ end. The BAG6 coding sequence was amplified with PCR to have suitable restriction site at both ends for insertion into pcDNA5/FRT/TO at the 5’ side of the myc-BirA coding sequence to produce BAG6-myc-BirA. A BAG6 N-terminally deletion variant was also generated with the same method to produce BAG6∆N-myc-BirA to study the importance of BAG6 ubiquitin-like domain containing N-terminus on substrate binding and ubiquitination. Myc-BirA was used as a control to determine background biotinylation. Plasmids encoding myc-BirA, BAG6-myc-BirA and BAG6∆N-myc-BirA were transfected into HeLa cells and expression of the constructs analysed with immunoblotting using anti-myc antibody. In Figure 5.2A, myc-BirA appeared as a single band between 32 and 46 kDa (lane 1). BAG6-myc- BirA appeared between 190 to 245 kDa (Figure 5.2A, IB: Myc, lane 2), which is

113

Figure 5.2: BAG6-myc-BirA re-localises and stabilises Op91. (A) HeLa cells were transfected with expression plasmids as indicated, and lysates analysed by SDS-PAGE and immunoblotting with anti-myc, anti-BAG6 and anti-tubulin followed by IRDye®-conjugated secondary antibodies. Bands of interest were shown with black arrows. exo, exogenous; endo, endogenous. (B) HeLa cells were co- transfected with plasmids encoding BAG6-myc-BirA and Op91, and after 24 hours fixed with formaldehyde and stained with BAG6 and opsin antibodies followed by Alexa Fluor®-conjugated secondary antibodies. This is a representative image from at least 2 independent experiments. Scale bar = 10 µm. (C) HeLa cells were co- transfected with plasmids encoding Op91 and myc-BirA/BAG6-myc-BirA/BAG6- V5, and lysates analysed by SDS-PAGE and immunoblotting with indicated primary antibodies followed by IRDye®-conjugated secondary antibodies. Bands of interest were shown with black arrows. (D) HeLa cells stably expressing BAG6-myc-BirA under an inducible promoter were induced with tetracycline (Tet) as indicated. Cells were lysed in digitonin buffer and cell lysate was incubated with anti-myc and immunoprecipitated proteins (IP) were eluted with SDS sample buffer and analysed by immunoblotting.

114

C D

115 bigger than endogenous BAG6 (Figure 5.2A, IB: BAG6, lane 2) showing that BAG6 was properly fused with BirA. BAG6∆N-myc-BirA was detected between 135 to 190 kDa and it was not detected with anti-BAG6 antibody targeting the N-terminus (Figure 5.2A, lane 3). Myc-BirA was expressed several-fold higher than BAG6- myc-BirA and BAG6∆N-myc-BirA, possibly due to myc-BirA having a smaller size. The smaller bands observed with anti-myc antibody for all the 3 constructs might represent degradation fragments.

BAG6-myc-BirA should behave similarly to endogenous BAG6 for accurate detection of endogenous interactions. To test this, two methods were used to examine the function of BAG6-myc-BirA: nuclear re-localisation and stabilisation of substrate with BAG6 overexpression (Chapter 3 and Chapter 4). HeLa cells were co- transfected with plasmids encoding BAG6-myc-BirA and Op91 (Chapter 3) and cells were fixed and stained to look at localisation of the two proteins. Anti-BAG6 antibody stained both endogenous and exogenous BAG6 and therefore, cytoplasmic staining of BAG6 was observed but much higher intensity of BAG6 was observed in the nucleus, representing the overexpressed BAG6-myc-BirA (Figure 5.2B, IF: BAG6). Op91 is unstable and gets degraded quickly (Chapter 3) but in cells where BAG6-myc-BirA was co-expressed, Op91 was stabilised and was concentrated in the nucleus (Figure 5.2B, IF: Opsin). HeLa cells co-transfected with plasmids encoding Op91 plus myc-BirA, BAG6-myc-BirA or BAG6-V5 were used to look at Op91 level with immunoblotting. Both glycosylated and non-glycosylated forms of Op91 were only faintly detected in cells co-transfected with plasmid encoding myc-BirA, but non-glycosylated Op91 was greatly stabilised when BAG6-myc-BirA was co- expressed, similar to BAG6-V5 (Figure 5.2C). Furthermore, BAG6 was shown to be able to co-immunoprecipitate RNF126 (Chapter 3) and here again, BAG6-myc- BirA clearly co-immunoprecipitated RNF126 under the same lysis conditions (Figure 5.2D, lane 4). All these data provide evidence that BAG6-myc-BirA retains the normal function of BAG6 and therefore BAG6-myc-BirA is suitable to be used in following BioID experiments.

116

5.3 Optimisation of BioID protocol

In order to identify BAG6 interacting partners using BioID, a method for generating and isolating biotinylated proteins was required. Therefore, the initial goal was to establish conditions for sample preparation that would be used for mass spectrometry analysis. Stable inducible cells for each construct were generated using Flp-In™ TREx™ system where tetracycline was then used to induce expression of the protein of interest. To determine the extent and specificity of biotinylation by the BAG6-myc-BirA, cells were treated under different combinations of induction and biotinylation. Cells were lysed in RIPA buffer containing 0.1 % SDS and biotinylated proteins were isolated with NeutrAvidin agarose, eluted with SDS- PAGE sample buffer, analysed on polyacrylamide gel and detected by infrared dye- conjugated streptavidin. A low level of biotinylation was observed in lysates of cells that were induced to express BAG6-myc-BirA in the presence of biotin, but not in non-induced cells or cultures in the absence of biotin (Figure 5.3A, lanes 1-3). The most prominent biotinylated band had a similar size to BAG6-myc-BirA, suggesting that BAG6-myc-BirA was itself biotinylated (Figure 5.3A, lane 3). When analysed materials bound to NeutrAvidin agarose, two non-specific bands were detected with the infrared dye-conjugated streptavidin in the absence of BAG6-myc-BirA (Figure 5.3A, lane 7) and biotin (Figure 5.3A, lane 8). In contrast, a range of biotinylated proteins were detected in the NeutrAvidin-pulldown from cells that had been induced to express BAG6-myc-BirA in the presence of biotin (Figure 5.3A, lane 9). These data show that BAG6-myc-BirA efficiently biotinylated proteins and biotinylated proteins were successfully isolated. In order to check if BAG6-myc-BirA was able to biotinylate known BAG6-interacting factor, materials bound to NeutrAvidin agarose were blotted with anti-RNF126 antibody. However, RNF126 was not detected in pulldown using NeutrAvidin agarose (Figure 5.3A, lane 9), suggesting that RNF126 was not biotinylated by BAG6-myc-BirA. This is surprising because BAG6-myc- BirA could co-immunoprecipitate RNF126 (Figure 5.2D) and RNF126 has 4 internal lysine residues for biotinylation. Thus, it is important to bear in mind the downside of these BioID experiments, since some interacting proteins may not be biotinylated (Roux, 2013; Li et al., 2017).

To compare the profile of proteins biotinylated by each of the different constructs, HeLa cells stably expressing myc-BirA, BAG6-myc-BirA or BAG6∆N-

117

A B

Figure 5.3: RIPA buffer extracts more proteins and fewer proteins are being washed away. (A) HeLa cells stably expressing BAG6-myc-BirA under an inducible promoter were induced with tetracycline (Tet) or left non-induced as indicated. Cells were then incubated with biotin for 24 hours before lysis in RIPA buffer. Cell lysate was incubated with NeutrAvidin agarose and pulled-down proteins (PD) were eluted with SDS sample buffer and analysed by immunoblotting with IRDye®-conjugated Streptavidin and anti-RNF126 followed by IRDye®-conjugated secondary antibodies. (B) Myc-BirA/BAG6-myc-BirA/BAG6∆N-myc-BirA were induced and then incubated with biotin for 24 hours before lysis in BioID buffer. Cell lysate was incubated with NeutrAvidin agarose and pulled-down proteins (PD) were eluted with SDS sample buffer and analysed by immunoblotting. (C) HeLa cells stably expressing BAG6-myc-BirA were induced with tetracycline (Tet) and incubated with biotin for 24 hours before lysis in RIPA or BioID buffer. Cell lysate (I) was incubated with dynabeads and pulled-down proteins (B) were eluted with SDS sample buffer and analysed by immunoblotting. Equal amount was used for comparison. L, lysate prior centrifugation; UB, unbound; W, wash.

118 myc-BirA were induced in the presence of biotin and lysed in buffer containing 0.2 % SDS. Biotinylated proteins were isolated with NeutrAvidin agarose, eluted with SDS-PAGE sample buffer, analysed on polyacrylamide gel and detected by infrared dye-conjugated streptavidin. Lysates of cells expressing myc-BirA contained only the non-specific bands, suggesting that myc-BirA did not extensively biotinylate protein non-specifically (Figure 5.3B, lane 4). Biotinylated proteins were detected in the NeutrAvidin-pulldown from cells induced to express BAG6-myc-BirA (Figure 5.3B, lane 5) and less biotinylated proteins were detected from cells expressing BAG6∆N-myc-BirA (Figure 5.3B, lane 6), agrees with previous understanding in the function of BAG6 N-terminus engaging factors such as E3 ubiquitin ligases.

In a trial experiment, HeLa cells stably expressing myc-BirA, BAG6-myc- BirA and BAG6∆N-myc-BirA under an inducible promoter were induced with different concentrations of tetracycline to achieve comparable expression among the constructs. Cells were lysed in buffer containing 0.2 % SDS and biotinylated proteins were isolated with Streptavidin-coupled Dynabeads, eluted with SDS-PAGE sample buffer and run slightly into a polyacrylamide gel for mass spectrometry analysis. Streptavidin-coupled Dynabeads were used instead of NeutrAvidin agarose as advised by the mass spectrometry facility. However, only a small amount of proteins with very low MS/MS count were detected. The BioID protocol calls for a 2 % SDS washing step of the Streptavidin-coupled Dynabeads to reduce non-specific protein binding (Kim et al., 2016). To make sure that biotinylated proteins were not being washed off the beads under high SDS conditions, washing with RIPA buffer containing 0.1% SDS was also tested. Cells were efficiently lysed in both methods (Figure 5.3C, lanes 1 and 8), and the centrifugation step used to clarify lysates did not cause loss in biotinylated proteins (Figure 5.3C, lanes 2 and 9). A significantly higher amount of biotinylated proteins were isolated with RIPA method when compared to BioID method (Figure 5.3C, lane 3 vs 10). Furthermore, the 2 % SDS washing step caused some biotinylated proteins to dissociate from the beads, resulting in the detection of biotinylated proteins in the wash buffer (Figure 5.3C, lane 12). In contrast, the RIPA buffer did not show any protein loss upon all washing steps (Figure 5.3C, lanes 5-7). Even though low SDS condition might cause increased non-specific interactions, analysing proteins enriched in the different samples when compared to the myc-BirA control would help in discriminating

119 specific against non-specific interactions. Based on these observations, samples were prepared following the RIPA method. Stable expression system sometimes only overexpresses a protein few folds above the endogenous level which might not be high enough for efficient biotinylation to occur. Hence, a transient expression system was tested in the hope of getting higher expression level leading to more efficient biotinylation to increase the protein hits from mass spectrometry.

5.4 The use of transient expression system for efficient biotinylation

Different amounts of Streptavidin-coupled Dynabeads were used to optimise the BioID experiment due to the change from a lower expression stable system to a higher expression transient system. According to the BioID protocol, 5 % (v/v) of Streptavidin-coupled Dynabeads were used (Roux et al., 2012). A higher (8.5 %) and a lower (3 %) amount of dynabeads were tested. Analysis of materials bound to dynabeads demonstrated that reducing the amount of dynabeads resulted in less biotinylated proteins being isolated (Figure 5.4A, lanes 2-4). Looking at the unbound fraction, it appeared that the 5 % dynabeads recommended was not sufficient for pulling down all biotinylated proteins generated in the transient overexpression system (Figure 5.4A, lane 6). However, increasing dynabeads to 8.5 % resulted in at least 90 % of the biotinylated proteins being pulled-down from the lysates (Figure 5.4A, lane 5). The use of RIPA buffer meant that no biotinylated proteins were lost from the dynabeads during washes (Figure 5.4A, lanes 8-10).

In order to unbiasedly identify BAG6 substrates and interacting factors, HeLa cells were transiently transfected with plasmids encoding myc-BirA, BAG6-myc- BirA or BAG6∆N-myc-BirA and biotin was added to the culture media for 24 hours. One out of two BAG6-myc-BirA transfected dishes of cells was treated with bortezomib for 6 hours. Cells were lysed in RIPA buffer containing 0.1 % SDS and incubated with 8.5 % Streptavidin-coupled Dynabeads then washed with 0.1 % SDS thrice. A small portion of each sample was kept for immunoblotting and the remaining samples were washed 3 times with 50 mM Tris.Cl pH 7.4 and dynabeads were stored in minimum volume of 50 mM ammonium bicarbonate for 1D LC- MS/MS. Western blotting the lysates and bound materials showed that all constructs were expressed at comparable levels (Figure 5.4B, IB: BirA, lanes 1-4), and that the

120

Figure 5.4: Transient overexpression system is used for BioID experiment. (A) HeLa cells were transfected with plasmid encoding BAG6-myc-BirA and then incubated with biotin for 24 hours before lysis in RIPA buffer. Cell lysate (I) was incubated with different amount of dynabeads and pulled-down proteins (B) were eluted with SDS sample buffer and analysed by immunoblotting. 10% of I, UB and W were used. UB, unbound; W, wash. (B) HeLa cells were transfected as indicated and then incubated with biotin for 24 hours with one sample treated with bortezomib (Bz) for 6 hours before lysis in RIPA buffer. Cell lysate was incubated with dynabeads and a smal amount of pulled-down proteins (PD) were eluted with SDS sample buffer and analysed by immunoblotting. The remaining PD was sent for mass spectrometry. Bands of interest were shown with black arrows. UB, unbound.

121

122 total cellular biotinylation for BAG6-myc-BirA and BAG6∆N-myc-BirA were equal (Figure 5.4B, lanes 2-4) with myc-BirA having minimal biotinylation (Figure 5.4B, lane 1). With dynabeads pulldown, myc-BirA isolated the lowest amount of biotinylated proteins followed by BAG6∆N-myc-BirA (Figure 5.4B, lanes 5 and 8). However, there was no significant increase in biotinylated proteins pulled-down when bortezomib was added (Figure 5.4B, lane 7 vs 6).

5.5 Mass spectrometry data analysis

Having optimised the generation and isolation of biotinylated proteins, the next set of experiments aimed to unbiasedly identify potential BAG6 substrates and interacting factors.

5.5.1 Comparison of technical replicates

Samples from 3 experimental conditions were sent for liquid chromatography-tandem mass spectrometry (LC-MS/MS): myc-BirA, BAG6-myc- BirA and BAG6-myc-BirA with proteasome inhibitor bortezomib. Two technical replicates were performed for BAG6-myc-BirA and BAG6-myc-BirA with bortezomib, but more samples were used for myc-BirA in the first run and the second run had the same amount of samples as BAG6-myc-BirA and BAG6-myc-BirA with bortezomib. Myc-BirA first run and second run were slightly different because a ‘loading evaluation’ was performed. The raw data from mass spectrometry was processed with MaxQuant and processed data was received in a Microsoft Excel spreadsheet. A box plot of label-free quantification (LFQ) intensity with respect to the different experimental conditions and technical replicates were plotted (Figure 5.5A). LFQ intensity is used to determine the relative amount of proteins in different samples without isotopic labelling. Peptide-ion intensity is extracted from MS1 and the peptide peak area or peak height is used as a quantitative measurement of the peptide concentration, while MS2 is used for identification of the peptide. Extracted peptide-ion intensity is then matched and compared across samples with a prerequisite where most proteins do not change abundance under different experimental conditions. LFQ intensity for each protein is calculated as the best

123

Figure 5.5: Box plots of (A) label-free quantification (LFQ) intensity and (B) MS/MS count with respect to the three experiment conditions and technical replicates for BAG6-myc-BirA and BAG6-myc-BirA with bortezomib (Bz) samples.

124 estimation after all the pair-wise peptide comparisons (Cox et al., 2014). The line in the box represents the median/mid-point of the dataset for the respective conditions/replicates. 75 % of the data falls below the upper quartile while 25 % of the data falls below the lower quartile. The upper and lower whiskers indicate the maximum and minimum values of the dataset excluding outliers, with outliers showing in black-filled circles. Highly similar plots were observed for all 3 experimental conditions and two technical replicates (Figure 5.5A), showing that the LFQ intensity could represent normalised data. Even though more samples were loaded into the first myc-BirA run, a lower intensity plot was observed after normalisation. Other data obtained with equal amount of samples loaded fell within the same range with BAG6-myc-BirA and bortezomib showing slightly lower intensity; meaning that all datasets were similarly distributed. To check which parameter would be better for downstream enrichment analysis, MS/MS count was plotted (Figure 5.5B). MS/MS spectral is generated when intact proteins or smaller peptides are subjected to ionization and subsequent fragmentation. MS/MS spectral is used not only for protein identification but also quantification. MS/MS count denotes the total number of spectral assigned to the same protein as a measure of its abundance (Nahnsen et al., 2013). As expected, MS/MS count for the first myc-BirA run had higher median than the second run. BAG6-myc-BirA and BAG6-myc-BirA with bortezomib and their technical replicates were very much similar but the myc- BirA run had the lowest median even though no difference was observed for the upper and lower quartiles.

5.5.2 Pre-enrichment analysis of myc-BirA vs BAG6-myc-BirA

Protein identification and quantification by mass spectrometry is a sensitive method but it is challenging at the same time, especially with samples prepared from affinity purification. Proteins can bind non-specifically to agarose beads used for affinity purification (Mellacheruvu et al., 2013), and in the case of BioID, endogenous biotinylation is not uncommon (Li et al., 2017). False positive or false negative identification can be reduced with proper control(s) included in all experiments, to maximise hit of genuine interactions. Myc-BirA was used as a negative control in this study to determine if BAG6-myc-BirA could isolate a distinct

125 population of proteins from myc-BirA, proteins that interact specifically with BAG6. Venn diagrams were constructed from pre-enrichment MaxQuant data using either LFQ intensity (Figure 5.6A) or MS/MS count (Figure 5.6B). A total of 1210 proteins were identified based on LFQ intensity and 396 more proteins were identified based on MS/MS count. MS/MS count gave higher number of identified proteins because proteins having only one spectrum were included, but this group of proteins would be excluded for downstream analysis due to low confidence on the occurrence. The majority of proteins identified (788 proteins and 1156 proteins with LFQ intensity and MS/MS count, respectively) were common between myc-BirA and BAG6-myc-BirA. Myc-BirA had 82 and 130 more unique proteins identified with LFQ intensity and MS/MS count, respectively, compared to BAG6-myc-BirA. This might be due to the fact that BAG6-myc-BirA is large compared to myc-BirA, and the C-terminal biotin-ligase in BAG6-myc-BirA might be less accessible for biotinylating its interacting proteins. However, BAG6-myc-BirA had slightly more unique proteins identified using LFQ intensity (170 proteins) than MS/MS count (160 proteins). Histograms of LFQ intensity (Figure 5.6C) and MS/MS count (Figure 5.6D) of proteins detected in myc-BirA and BAG6-myc-BirA showed largely overlapped area. Density, also called frequency density, is the number of proteins falls into each LFQ intensity or MS/MS count unit. Proteins of interest would fall in ‘BAG6-myc-BirA only’ area towards the higher LFQ intensity or MS/MS count and LFQ intensity gave bigger non-overlapped area. When myc-BirA vs BAG6-myc-BirA were plotted into scatter plots using LFQ intensity (Figure 5.6E) and MS/MS count (Figure 5.6F), the correlation line was skewed towards myc-BirA, indicating that more proteins had higher intensities or counts in the negative control, with LFQ intensity showing better correlation. Adding together, LFQ intensity was thought to be a better parameter for downstream analysis.

5.5.3 Comparison of biological replicates

Two biological replicates were analysed for myc-BirA, BAG6-myc-BirA and BAG6-myc-BirA with bortezomib samples. Sample prepared from BAG6 N- terminus deleted variant with myc-BirA ligated at the C-terminus (BAG6(∆N)-myc- BirA) was also included for analysis. The mean LFQ intensity of the two technical

126

A B

C D

E F

Figure 5.6: Venn diagrams of proteins identified from (A) label-free quantification (LFQ) intensity and (B) MS/MS count of myc-BirA and BAG6-myc-BirA samples. Histograms of (C) LFQ intensity and (D) MS/MS count of proteins detected in myc- BirA and BAG6-myc-BirA samples. Scatter plots of myc-BirA and BAG6-myc-BirA samples using (E) LFQ intensity and (F) MS/MS count.

127 replicates in the first biological replicate was compared with the second biological replicate (Figure 5.7A). Log-10 scale was used to respond to skewness towards large values. Slightly higher LFQ intensity plots were observed for all the different experimental conditions in the second biological replicate. However, after median- normalisation, both biological replicates closely resembled each other, indicating that the datasets were similarly distributed. Proteins identified with myc-BirA and BAG6- myc-BirA samples were plotted into two Venn diagrams, comparing the two biological replicates (Figure 5.7B and 5.7C). It was obvious that the second biological replicate identified more proteins than the first replicate with little overlap between them. For myc-BirA, 1040 proteins were identified in the first replicate while 1776 proteins were identified in the second replicate, with only 429 proteins similarly identified in both replicates (Figure 5.7B). The same trend was observed for BAG6-myc-BirA, first replicate gave 958 proteins hit while second replicate gave 1902 protein hits, with 397 proteins overlapped between them (Figure 5.7C). It is worth mentioning that more proteins were isolated with BAG6-myc-BirA in replicate two in relation to myc-BirA, in contrast to the first replicate. The two biological replicates were sent for analysis at a different time and this might account for the difference in the two biological replicates observed. A dendrogram was constructed based on LFQ intensity clustered in the Spearman distance (Figure 5.7D). There was a clear separation of biological replicates over the different experimental conditions. Interestingly, myc-BirA was well separated from the other experiments in within each biological replicate cluster.

5.5.4 Identification of potential BAG6 substrates

As the two biological replicates were not well correlated, only proteins identified in both of the replicates were analysed, with the rest of the protein hits being excluded from further analysis. The complete analysis flow is illustrated in Figure 5.8. BAG6-myc-BirA identified 958 proteins in the first replicate and 1902 proteins in the second replicate, 397 proteins were found in both replicates. Enrichment analysis was performed on these 397 common protein hits, and only proteins that were uniquely detected with BAG6-myc-BirA or had LFQ intensity at least two times (arbitrary) higher than the myc-BirA control (Table 5.1) were used

128

B C

Figure 5.7: (A) Box plot of non-normalised and median-normalised label-free quantification (LFQ) intensities with respect to the four experiment conditions and two biological replicates except for Bag6(∆N)-myc-BirA sample with no biological replicate. Venn diagrams of proteins identified from (B) myc-BirA and (C) BAG6- myc-BirA samples of two biological replicates. (D) Dendrogram showing how the experiments cluster.

129

Figure 5.8: Flow chart for the identification of potential BAG6 substrates.

130

Table 5.1: Mean label-free quantification (LFQ) intensity and fold change of proteins found in both biological replicates, were unique with BAG6-myc-BirA or had at least two times higher intensity with BAG6-myc-BirA sample than myc-BirA sample. Fold change shows the mean of two biological replicates or “,” is used to distinguish the first and second replicates.

LFQ intensity Fold change Gene name Protein name (mean) Large proline-rich protein Unique BAG6 5.45 X 1012 BAG6 UBL4A Ubiquitin-like protein 4A 4.63 X 108 Unique Golgi to ER traffic protein 4 1.91 X 106 Unique, 0 GET4 homolog first replicate Barrier-to-autointegration Unique, 2.0 BANF1 4.01 X 108 factor Transcription intermediary 8.0 TRIM28 3.85 X 108 factor 1-beta X-ray repair cross- 2.5 XRCC6 2.80 X 108 complementing protein 6 SF3A1 Splicing factor 3A subunit 1 2.28 X 108 2.5 SNW domain-containing 4.0 SNW1 2.13 X 108 protein 1 Lysine-specific demethylase 10.2 KDM3B 1.81 X 108 3B CYR61 Protein CYR61 1.50 X 108 2.0, unique UDP-N-acetylglucosamine- 2.9 peptide N- OGT 1.15 X 108 acetylglucosaminyltransferase 110 kDa subunit FAM50A Protein FAM50A 9.26 X 107 15.2 Nuclear ubiquitous casein and Unique, 17.8 NUCKS1 cyclin-dependent kinase 8.00 X 107 substrate 1 RPS21 40S ribosomal protein S21 7.87 X 107 Unique, 2.1 Transcription initiation factor Unique, 2.2 GTF2A2 7.33 X 107 IIA subunit 2 Proteasome activator complex 8.8 PSME3 6.54 X 107 subunit 3 Serine/threonine-protein 3.5 PPP4R2 phosphatase 4 regulatory 5.58 X 107 subunit 2 G patch domain-containing Unique, 11.3 GPATCH8 5.45 X 107 protein 8 Ubiquitin-like modifier- Unique, 3.1 UBA1 5.37 X 107 activating enzyme 1 DGCR14 Protein DGCR14 4.83 X 107 Unique, 4.4 Polyglutamine-binding protein Unique PQBP1 4.65 X 107 1 Heterogeneous nuclear 2.0, unique HNRNPA2B1 4.22 X 107 ribonucleoproteins A2/B1

131

Uncharacterized protein Unique, 4.9 KIAA1143 4.21 X 107 KIAA1143 Ribonuclease P protein subunit Unique, 3.2 RPP30 4.15 X 107 p30 General transcription factor IIE Unique, 13.9 GTF2E1 4.06 X 107 subunit 1 NEDD8-conjugating enzyme Unique, 6.5 UBE2M 3.81 X 107 Ubc12 HSPA4 Heat shock 70 kDa protein 4 3.58 X 107 Unique Zinc finger and BTB domain- Unique, 10.9 ZBTB21 3.57 X 107 containing protein 21 WW domain-binding protein Unique WBP11 3.45 X 107 11 Protein strawberry notch Unique, 3.1 SBNO1 3.12 X 107 homolog 1 ATP-dependent RNA helicase Unique DHX8 3.12 X 107 DHX8 QRICH1 Glutamine-rich protein 1 2.66 X 107 Unique MKI67 FHA domain- Unique, 2.2 NIFK interacting nucleolar 2.51 X 107 phosphoprotein PRPF38A Pre-mRNA-splicing factor 38A 2.34 X 107 2.2, unique TMOD3 Tropomodulin-3 2.20 X 107 Unique, 3.4 X-ray repair cross- 2.1, unique XRCC5 2.13 X 107 complementing protein 5 FOXK1 Forkhead box protein K1 2.06 X 107 Unique, 2.8 Histone deacetylase complex Unique, 3.6 SAP18 1.84 X 107 subunit SAP18 SAP domain-containing Unique, 2.4 SARNP 1.84 X 107 ribonucleoprotein KIAA0020 Pumilio homolog 3 1.76 X 107 Unique CDK-activating kinase Unique MNAT1 1.72 X 107 assembly factor MAT1 Histone-lysine N- Unique WHSC1L1 1.59 X 107 methyltransferase NSD3 Paired amphipathic helix Unique, 2.4 SIN3A 1.51 X 107 protein Sin3a PRIM2 DNA primase large subunit 1.26 X 107 Unique Na(+)/H(+) exchange Unique SLC9A3R1 1.25 X 107 regulatory cofactor NHE-RF1 RCC2 Protein RCC2 1.14 X 107 Unique G patch domain-containing Unique GPATCH11 1.01 X 107 protein 11 CDKN2AIP CDKN2A-interacting protein 1.01 X 107 Unique Insulin-like growth factor 2 Unique IGF2BP1 9.97 X 106 mRNA-binding protein 1 Polyadenylate-binding protein Unique PABPN1 9.91 X 106 2 MLLT10 Protein AF-10 9.88 X 106 Unique

132

NELFA Negative elongation factor A 9.04 X 106 Unique Ribonuclease P protein subunit Unique RPP38 8.18 X 106 p38 CREB-regulated transcription Unique CRTC3 7.43 X 106 coactivator 3 SNCG Gamma-synuclein 6.67 X 106 Unique, 2.4 FAM122B Protein FAM122B 6.61 X 106 Unique Coiled-coil domain-containing Unique CCDC94 6.25 X 106 protein 94 MBNL1 Muscleblind-like protein 1 4.62 X 106 Unique MLLT6 Protein AF-17 4.46 X 106 Unique Periodic tryptophan protein 1 Unique PWP1 4.23 X 106 homolog Pre-mRNA-splicing factor Unique DHX38 ATP-dependent RNA helicase 3.99 X 106 PRP16 Chromatin assembly factor 1 Unique CHAF1A 3.61 X 106 subunit A FAM98B Protein FAM98B 3.35 X 106 Unique Arginine/serine-rich coiled-coil Unique RSRC2 3.05 X 106 protein 2 Paired amphipathic helix Unique SIN3B 2.98 X 106 protein Sin3b TSC22 domain family protein Unique TSC22D1 2.84 X 106 1 DNA replication licensing Unique MCM6 1.35 X 106 factor MCM6 DNA-directed RNA Unique POLR2B 1.25 X 106 polymerase II subunit RPB2 Mediator of RNA polymerase Unique MED21 1.08 X 106 II transcription subunit 21

133 for functional analysis (Schweingruber et al., 2016; Schopp et al., 2017). BAG6 is engaged in a stable ternary complex with two other proteins, Ubl4A and TRC35 (Mariappan et al., 2010), and Ubl4A was identified as one of the 70 enriched proteins, validating the approach. However, TRC35 came out in the first replicate but not the second, due to the difference in the replicates discussed in Section 5.5.3. Functional enrichment of identified proteins can be found in the Supplementary data 8.1. In one of the experimental conditions, bortezomib was added to inhibit proteasomal degradation. As shown earlier in Chapter 3 and Chapter 4, BAG6- substrate interactions were greatly stabilised in the presence of proteasome inhibition. Out of the 70 BAG6-myc-BirA enriched proteins, 28 of them were further enriched at least twice with proteasome inhibition.

BAG6 has been shown to interact with proteins with hydrophobic amino acids (Hessa et al., 2011; Tanaka et al., 2016), but exactly how hydrophobic or how long the stretch of hydrophobic amino acids is needed for interaction has not been defined. To have a better understanding of BAG6 substrate specificity, the 70 BAG6- myc-BirA enriched proteins were subjected to bioinformatic analysis to assess the hydrophobicity pattern in each of the proteins. As BAG6 is involved in cytosolic protein quality control (Leznicki et al., 2010; Minami et al., 2010; Hessa et al., 2011; Tanaka et al., 2016), proteins not annotated to the cytoplasm were excluded from analysis. However, proteins that have not been annotated were not excluded, to prevent losing potential substrates. 23 cytosolic annotated proteins and 10 non- annotated proteins were analysed with Perl script. Specific analysis parameters were adjusted, namely the hydrophobic scale, the hydrophobic window and the hydrophobic threshold. In this study, Kyte-Doolittle hydrophobic scale (Kyte and Doolittle, 1982) was used with hydrophobic window set at 20 and hydrophobic threshold set at 0. Each amino acid has been assigned a hydrophobic index depending on how easily the side chain of the amino acid transfers between the water and vapour phases (Kyte and Doolittle, 1982). The hydrophobic window specifies the number of amino acids to be considered as a segment for hydrophobicity calculation. For example, if the hydrophobic window is set at 20, the first value is the average of the hydrophobic index for amino acids 1 to 20, and the second value is the average of the hydrophobic index for amino acids 2 to 21, and so on. Hydrophobic window of 20 was used because it is the length of amino acid chain estimated to span

134 the lipid bilayer, which would be sufficient for BAG6 interaction (Leznicki et al., 2010; Mariappan et al., 2010; Hessa et al., 2011). The hydrophobic threshold was set to 0, so all positive values were considered hydrophobic. Command prompt was used to run the Perl script and output file in .txt contained Uniprot entry, number of hydrophobic region, percentage of hydrophobic region, and the entire polypeptide marked ‘-’ for non-hydrophobic region and ‘H’ for hydrophobic region (Supplementary data 8.2). Maximum length of continuous hydrophobic region and whether the proteins were enriched with bortezomib were illustrated in Table 5.2. Two known BAG6-interacting proteins, rhodopsin (Chapter 3) and XBP1 (Chapter 4) were included in the analysis as positive controls. The minimum percentage of hydrophobic region was observed for rhodopsin (11 %) with the maximum of 86 % found in KIAA1143, suggesting that all proteins could potentially interact with BAG6 if hydrophobicity was the sole determinant. 31 out of 35 proteins had more than 20 continuous hydrophobic regions, agreeing with study showing BAG6 interaction with long linear hydrophobic stretches (Mariappan et al., 2010). 13 out of 33 proteins were enriched when the proteasome was inhibited. Future work will aim at validating the interactions of these 13 candidate proteins with BAG6 biochemically.

5.5.5 Identification of potential BAG6 interacting factors

BAG6 ubiquitin-like domain is located at the N-terminus and it is important for RNF126 (E3 ubiquitin ligase) and SGTA (co-chaperone) interactions (Winnefeld et al., 2006; Leznicki et al., 2010; Rodrigo-Brenni et al., 2014). BAG6(∆N)-myc- BirA was generated and used in conjunction with BAG6-myc-BirA for the identification of protein quality control factors interacting with BAG6 at the N- terminus, which would be selectively biotinylated by BAG6-myc-BirA and not BAG6(∆N)-myc-BirA. 620 proteins were isolated with both constructs, 94 proteins were isolated only with the N-terminus deleted variant and 255 proteins were isolated only with BAG6-myc-BirA (Figure 5.9). The 255 proteins interacted with BAG6-myc-BirA and enriched at least twice over myc-BirA were imported into Metascape for further analysis. ‘Ubiquitin’ was used as a specific enriched gene ontology term to search through the list of the 255 proteins and 27 of them were

135

Table 5.2: Hydrophobicity prediction using Kyte and Doolittle scale with window size of 20 and threshold of 0. N.D, not determined in this experiment.

Maximum length Percentage of continuous Enriched with Gene name hydrophobic hydrophobic bortezomib? region region BANF1 39 15 Yes CCDC94 77 137 No CRTC3 59 89 No DCAF7 27 15 Yes FAM122B 56 40 Yes FAM98B 38 25 No GPATCH11 71 106 Yes GPATCH8 74 452 No HNRNPA2B1 81 98 No HSPA4 47 78 No IGF2BP1 43 34 No KIAA1143 86 78 No MBNL1 31 31 Yes OGT 27 21 No PABPN1 28 38 Yes PFKP 68 71 Yes PPP4R2 29 23 No PRIM2 77 115 No PSME3 51 37 No QRICH1 43 36 Yes RCC2 43 63 Yes RPS21 14 2 Yes RSRC2 84 263 No SAP18 65 24 No SBNO1 42 109 Yes SLC9A3R1 63 65 No SNCG 43 30 No TMOD3 34 41 No TSC22D1 57 61 Yes UBA1 36 28 No UBE2M 53 39 No UBL4A 41 21 No WBP11 71 111 Yes RHO 11 11 N.D XBP1 52 69 N.D

136

Figure 5.9: Venn diagram of proteins identified from BAG6-myc-BirA and BAG6(∆N)-myc-BirA samples.

137 associated with ubiquitin pathway in some way (Table 5.3). Ideally, two more biological replicates should be done so that statistical analysis can be performed on detected interactors.

5.6 Discussion

Proteins exist in complicated interconnected networks and these protein- protein interactions are crucial in all cellular processes. In order to determine the function of a particular protein, identifying other proteins in the same network will offer an informative insight. Affinity purification coupled to mass spectrometry (AP- MS) has been widely used for studying protein-protein interactions. However, cells have to be broken open to extract cellular contents for affinity purification and cell lysis disrupts weak/transient interactions (Mehta and Trinkle-Mulcahy, 2016) which are important in constructing the protein interactome. Thus, only high-affinity/stable in-vitro interactions are detected. To overcome the downside of AP-MS, proximity- dependent labelling approaches such as BioID have been developed and advanced (Roux et al., 2012). BioID utilises Escherichia coli biotin ligase used for biotinylating acetyl-CoA carboxylase (Chapman-Smith and Cronan, 1999). Wild- type biotin ligase generates reactive biotinoyl-AMP from biotin and ATP, and remains bound to biotinoyl-AMP until it recognises acetyl-CoA carboxylase and transfers the reactive biotin to acetyl-CoA carboxylase. When wild-type biotin ligase is used in proximity labelling, it can only biotinylate substrates with a recognition/acceptor sequence (Li et al., 2017). A mutation R118G was introduced into the active site of wild-type biotin ligase to generate a promiscuous form of biotin ligase (BirA) with reduced affinity to biotinoyl-AMP (Choi-Rhee et al., 2004) so that the reactive biotinoyl-AMP can be transferred more efficiently to any substrate proteins within the radius of 10 nm (Kim et al., 2014). Biotinylation is a type of covalent modification where biotin is conjugated onto primary amines such as those found in the lysine side chain (Roux et al., 2012), hence transient interactions in-vivo can be detected, making BioID an effective approach in identifying novel interactions. Since biotin is added to the cell media for 24 hours to allow maximal biotinylation over the course of BioID, interactions of low abundance proteins will have increased chance to be detected (Roux, 2013).

138

Table 5.3: Potential BAG6-associated protein quality control factors with gene ontology related to ubiquitination pathway.

Gene Name Biological process Molecular function ATRIP GO:0070530 K63-linked polyubiquitin binding GO:0031593 polyubiquitin binding GO:0043130 ubiquitin binding CBFA2T3 GO:0032436 positive regulation of proteasomal ubiquitin-dependent protein catabolic process CHFR GO:0032436 GO:0061630 positive regulation of proteasomal ubiquitin protein ligase activity ubiquitin-dependent protein GO:0061659 catabolic process ubiquitin-like protein ligase GO:0032434 activity regulation of proteasomal GO:0004842 ubiquitin-dependent protein ubiquitin-protein transferase catabolic process activity CHP1 GO:0031397 negative regulation of protein ubiquitination DNAJB9 GO:0030433 ubiquitin-dependent ERAD pathway FBXL18 GO:0031146 GO:0004842 SCF-dependent proteasomal ubiquitin-protein transferase ubiquitin-dependent protein activity catabolic process GO:0019787 GO:0043161 ubiquitin-like protein transferase proteasome-mediated ubiquitin- activity dependent protein catabolic process GO:0000209 protein polyubiquitination FBXL2 GO:0031146 SCF-dependent proteasomal ubiquitin-dependent protein catabolic process GO:0006513 protein monoubiquitination FZD8 GO:0031625 ubiquitin protein ligase binding HIST1H2BD GO:0016567 protein ubiquitination HIST1H2BH GO:0016567 protein ubiquitination

139

HIST1H2BJ GO:0016567 protein ubiquitination HIST1H2BL GO:0016567 protein ubiquitination LNX1 GO:0042787 GO:0004842 protein ubiquitination involved in ubiquitin-protein transferase ubiquitin-dependent protein activity catabolic process GO:0006511 ubiquitin-dependent protein catabolic process MUL1 GO:0031625 ubiquitin protein ligase binding GO:0044389 ubiquitin-like protein ligase binding NEUROD2 GO:0016567 protein ubiquitination PIAS1 GO:0032436 GO:0031625 positive regulation of proteasomal ubiquitin protein ligase binding ubiquitin-dependent protein catabolic process PSMA2 GO:0051437 positive regulation of ubiquitin- protein ligase activity involved in regulation of mitotic cell cycle transition GO:0051436 negative regulation of ubiquitin- protein ligase activity involved in mitotic cell cycle GO:1904667 negative regulation of ubiquitin protein ligase activity PSMD12 GO:0051437 positive regulation of ubiquitin- protein ligase activity involved in regulation of mitotic cell cycle transition GO:0051436 negative regulation of ubiquitin- protein ligase activity involved in mitotic cell cycle GO:1904667 negative regulation of ubiquitin protein ligase activity RIPK2 GO:0031398 positive regulation of protein ubiquitination RNF133 GO:0051865

140

protein autoubiquitination GO:0016567 protein ubiquitination SENP1 GO:0032435 GO:0070137 ubiquitin-like negative regulation of proteasomal protein-specific endopeptidase ubiquitin-dependent protein activity catabolic process SNX9 GO:0031625 ubiquitin protein ligase binding UBA7 GO:0006511 GO:0004839 ubiquitin-dependent protein ubiquitin activating enzyme catabolic process activity GO:0004842 ubiquitin-protein transferase activity UBXN1 GO:1903094 GO:0071796 negative regulation of protein K6-linked polyubiquitin binding K48-linked deubiquitination GO:0036435 GO:1903093 K48-linked polyubiquitin binding regulation of protein K48-linked GO:0031593 polyubiquitin deubiquitination binding GO:2000157 negative regulation of ubiquitin- specific protease activity UFC1 GO:0019787 ubiquitin-like protein transferase activity USP49 GO:0035616 GO:0004843 histone H2B conserved C-terminal thiol-dependent ubiquitin-specific lysine deubiquitination protease activity GO:0016578 GO:0036459 histone deubiquitination thiol-dependent ubiquitinyl GO:0016579 hydrolase activity protein deubiquitination VPS36 GO:0043328 GO:0043130 protein targeting to vacuole ubiquitin binding involved in ubiquitin-dependent protein catabolic process via the multivesicular body sorting pathway GO:0043162 ubiquitin-dependent protein catabolic process via the multivesicular body sorting pathway

141

BAG6 has been shown to play roles in the cytosolic protein quality control by recognising hydrophobic substrate, recruiting E3 ubiquitin ligase and preventing substrate aggregation for degradation of ubiquitinated substrate at the proteasome (Leznicki et al., 2010; Minami et al., 2010; Hessa et al., 2011; Tanaka et al., 2016). By far, most of the work done to elucidate BAG6 functions in the protein quality control system relies on model substrate proteins. Even though BAG6 was proposed to interact with proteins containing hydrophobic region (Hessa et al., 2011; Tanaka et al., 2016), BAG6 interaction with model substrate proteins of varying hydrophobicity was unpredictable (unpublished data from Swanton Lab), leading to difficulty in identifying suitable model substrate proteins for BAG6 study. BioID was performed in the hope of identifying BAG6 physiological substrates so that BAG6 substrate specificity and functions can be studied more closely. Fusion of BirA (35 kDa) to BAG6 or any protein of interest may affect not only the size but folding, function and also localisation of the protein (Kim and Roux, 2016). Therefore, BAG6-myc-BirA was first validated for its localisation and functionality before performing the BioID experiment. Overexpressed BAG6 is localised to the nucleus due to the disruption of its complex with Ubl4A and TRC35 where TRC35 keeps BAG6 in the cytoplasm by masking BAG6 nuclear localisation signal (Wang et al., 2011). Overexpressed BAG6-myc-BirA was shown to concentrate in the nucleus (Figure 5.2B), indicating that BirA fused at the C-terminus of BAG6 has not changed BAG6 localisation. Functionality wise, BAG6-myc-BirA stabilised BAG6 substrate Op91 (also due to disruption of the BAG6 complex) (Figure 5.2C) and co- immunoprecipitated RNF126 (Figure 5.2D) as efficient as BAG6, showing that BAG6-myc-BirA has also not compromised functionally.

Myc-BirA, BAG6-myc-BirA and BAG6-myc-BirA with proteasome inhibition samples were sent for LC-MS/MS for the identification of BAG6 endogenous substrates. Technical replicates were performed and data was shown to be distributed similarly between different samples (Figure 5.5). When comparing LFQ intensity to MS/MS count for use in downstream analysis, LFQ intensity was chosen as it was shown to represent normalised data across samples (Figure 5.5A). LFQ intensity also gave bigger region of interest with histogram (Figure 5.6C) and better correlation to myc-BirA with scatter plot (Figure 5.6E). The second biological replicate had higher LFQ intensity (Figure 5.7A) and higher protein hits (Figure

142

5.7B and 5.7C) for all samples analysed. The dendrogram clearly shown that both biological replicates were distinct from each other, with myc-BirA separated from BAG6-myc-BirA in both replicates (Figure 5.7D). BAG6-myc-BirA with and without proteasome inhibition was closely clustered, indicating that proteasome inhibition was not critical in the majority of BAG6-substrate interactions (Figure 5.7D). Due to the differences in the two biological replicates, only proteins found in both replicates unique to BAG6-myc-BirA or at least twice enriched with BAG6- myc-BirA over myc-BirA were analysed. BAG6 engages in a stable trimeric complex with Ubl4A and TRC35 (Mariapan et al., 2010) and Ubl4A was identified to interact uniquely with BAG6-myc-BirA in both of the BioID experiments (Table 5.1). However, TRC35 was only detected in the first but not the second biological replicate (Table 5.1). Besides TRC35, a number of potential interacting proteins were unique in one experiment but not the other (Table 5.1), which could be the weakness of the BioID approach or the sample preparation in these experiments. Cytosolic annotated proteins and non-annotated proteins were subjected to hydrophobicity prediction using prepared Perl Script but the percentage hydrophobic region and the maximum length of continuous hydrophobic region varied greatly between potential substrates (Table 5.2). This might indicate that hydrophobicity is not being as important as previously thought for BAG6 interaction as BAG6 interacting rhodopsin (Chapter 3) was relatively non-hydrophobic (Table 5.2). However, no conclusion could be made before interactors were validated biochemically. Besides identifying endogenous substrates, other factors working together with BAG6 in the protein quality control pathways were also of interest, especially E3 ubiquitin ligases. Ubiquitination is a transient process that is difficult to be captured with traditional biochemical methods. Also, E3 ligase-substrate affinity is usually low and ligase is capable of interacting different substrates, making detection of the interactions challenging. BioID has been used to detect interactions in the ubiquitination pathway (Kim and Roux, 2016; Li et al., 2017) and here, 27 BAG6-associating factors were successfully identified by having ‘ubiquitin’ term in their gene ontologies (Table 5.3). Again, biochemical analysis has to be performed before any further conclusion can be made.

Due to the dynamic nature of protein-protein interactions, distinguishing real interactions from random interactions has been challenging. Since reactive biotinoyl-

143

AMP is released from BirA into the cellular environment, biotinoyl-AMP will biotinylate any protein in close proximity; hence a positive hit may not at all indicate a direct physical interaction (Kim and Roux, 2016). Random interactions are especially common in highly expressed proteins because they are likely to present around reactive biotinoyl-AMP. On the other hand, proteins with low intensity or count from mass spectrometry analysis do not necessarily indicate false positive interactions. In many cases, these interactions are potentially more relevant than interactions of high abundance proteins (Kim and Roux, 2016). Proper control(s) has to be included in all BioID experiments to discriminate real interactions from random interactions. BirA without fusion of a protein of interest is commonly used as a negative control (Kim and Roux, 2016) where proteins enriched less than two times over the control are excluded for downstream analysis. BirA only control was the only available control but not the ideal control for this study, mainly due to the large size difference between BAG6-myc-BirA and myc-BirA, which had a direct effect on biotinylation efficiency. BAG6-myc-BirA variant with mutation(s) at BAG6 substrate binding site that completely loses ability to bind substrates would be the ideal control, but the substrate binding site has yet to be defined. Care is to be taken when interpreting negative data because some known interacting partners were not identified with BioID (Roux, 2013; Li et al., 2017). Even though BAG6-myc-BirA was able to co-immunoprecipitate RNF126, pulldown of biotinylated proteins from lysates prepared from cells transfected with plasmid encoding BAG6-myc-BirA in the presence of biotin could not detect RNF126 (Figure 5.3A). RNF126 has lysine residues but the lysine residues might not be present at the surface and hence are not accessible for biotinylation. Although biotin is generally non-toxic, biotinylation may impair certain protein functions (Kim and Roux, 2016). An obvious example would be protein ubiquitination, as biotinylated proteins would have their lysine residues occupied and thus ubiquitination event was inhibited (Roux, 2013).

Putting together, analysis of the mass spectrometry data from two biological replicates has generated diverse protein hits (Figure 5.7). The two replicates were subjected to mass spectrometry at different times and this might account for the variability in the data obtained. To increase reliability of data, biological replicates should be analysed in the same mass spectrometry run with samples randomised. In order to perform statistical analysis on the data obtained, more biological replicates

144 are needed. Most importantly, potential novel interactions should also be validated biochemically to be designated as genuine interactions.

145

6 Conclusions and future work

6.1 Conclusions

In order to better understand the precise role of BAG6 in the protein quality control system, it is crucial to look at factors interacting with BAG6 in the different pathways. In Chapter 3, the role of UBR4 in BAG6-mediated quality control pathways was investigated. UBR4 was co-purified with Sec61β (BAG6 substrate) and SGTA (BAG6-interacting factor) in two independent experiments, leading to speculation that UBR4 might be involved in some common pathways with BAG6. UBR4 is a member of the family of E3 ubiquitin ligase but it does not have a typical E3 ligase domain. UBR4 knockdown had no effect on a model mislocalised protein (Op91) but degradation of an ERAD substrate (opsin-degron) was delayed. BAG6 knockdown has also caused delay in the degradation of opsin-degron. Op91 and opsin-degron are model substrate proteins that have been used in studies aiming to elucidate BAG6 functions in mislocalised protein degradation and ERAD, respectively (Ray-Sinha et al., 2009; Payapilly and High, 2014; Wunderley et al., 2014; Leznicki et al., 2015). However, it was surprising that instead of reducing opsin-degron ubiquitination, depletion of UBR4 actually caused massive increase in not only opsin-degron ubiquitination but total cellular ubiquitination. Even though direct interactions were not detected between UBR4 and BAG6 and also opsin- degron, knocking down UBR4 has improved BAG6-opsin-degron interaction and also BAG6-RNF126 (E3 ligase) interaction. The working model is that UBR4 hinders BAG6-substrate-E3 ligase complex coming together, and by this mean delays substrate degradation.

BAG6 has been shown to interact with model proteins having exposed stretches of hydrophobic amino acids and that the interactions were reduced or lost when hydrophobicity was reduced. In Chapter 4, BAG6 was shown to interact with a novel substrate, the unspliced XBP1 that is a key component of the unfolded protein response. Unspliced XBP1 has a hydrophobic region to engage the ER membrane to undergo splicing under conditions of ER stress. Under normal conditions, unspliced XBP1 is localised to the cytoplasm and being degraded rapidly. This ‘miclocalised’ species of XBP1 has hydrophobic domain exposed in the 146 cytoplasm, hence representing a good substrate for BAG6. BAG6 interacted with unspliced XBP1 by targeting the hydrophobic domain. BAG6 depletion reduced unspliced XBP1 ubiquitintaion, leading to a delay in XBP1 degradation. This is the first time BAG6 was shown to play a role in the unfolded protein response. This is an important finding because BAG6 affects unspliced XBP1 turnover which in turn controls the level of spliced XBP1, which is one of the most important transcription factors responsible for the unfolded protein response.

It is obvious that BAG6 interacts with hydrophobic amino acids but the minimum length of hydrophobicity needed for interaction is not defined. In Chapter 5, BioID was performed to unbiasedly biotinylate proteins that appeared in close proximity to BAG6. BioID is a powerful approach utilising covalent biotin modification on substrate proteins for the detection of transient interactions or interactions involving low abundance proteins, which is challenging when using other affinity-purification methods. However, proximity labelling might be random and non-specific, especially for highly expressed proteins. Self-biotinylation or biotinylation-resistant of a certain proteins have to be taken into account when analysing the BioID data. Even though biotinylated proteins do not indicate genuine substrates until confirmed by more direct biochemical analysis, proper controls in the assay and careful analysis of the mass spectrometry data will provide an insight of the hydrophobicity needed for a substrate to interact with BAG6. As mentioned above, identifying quality control factors in BAG6-mediating pathways is also important to unfold BAG6 role in more details. BAG6 recruits E3 ubiquitin ligase through the N-terminus and therefore N-terminus deleted variant was used as a negative control in the BioID experiment aiming to identify interacting factors. Proteins interacted with the full length BAG6 and not the N-terminus deleted variant, with ‘ubiquitin’ annotated in their gene ontology, would be of interest for future study.

To conclude, BAG6 interacts with different substrates and hydrophobicity is thought to be one of the important criteria in BAG6 substrate. Analysis of the BioID data and future validation of the interactors may provide important information about BAG6 substrate specificity determinant so that BAG6 role in the different pathways can be further elucidated. It may seem that BAG6 has different roles on distinct substrates, possibly by engaging different factors, such that UBR4 was involved in

147 opsin-degron degradation but has no role in unspliced XBP1 degradation, in which both are BAG6 substrates.

6.2 Future work

It would be of interest to validate potential BAG6 interacting factors in the ubiquitination pathway obtained from the BioID experiment with co- immunoprecipitation. Involvement of those positive candidates in the ubiquitination and degradation of opsin-degron and unspliced XBP1 would be examined to check if BAG6 engages different factors towards different substrates. In the case of opsin- degron, BAG6 interaction with the identified factor would be evaluated under conditions of UBR4 knockdown to test the proposed model where depletion of UBR4 causes increased opsin-degron ubiquitination by enhancing BAG6-factor interaction. The occurrence of hyper-ubiquitination on other BAG6 substrates and the nature of the hyper-ubiquitination chain would also be studied. The interaction between BAG6 and its potential substrates from the BioID experiment would be measured by the ability of BAG6 to induce nuclear localisation, stabilisation and ubiquitination of these proteins. Positive candidates would then be subjected to a series of bioinformatic analysis for the determination of common hydrophobicity features in the amino acid sequence and also structure. Then, it would be of interest to determine if BAG6 can target neurodegenerative-linked proteins having ‘enough’ hydrophobic regions, prevent them from aggregation and promote their timely degradation. Future work would also examine whether adding a hydrophobic domain converts proteins to being BAG6 substrates and determine whether the position of hydrophobic domain matters. For these experiments, increasing lengths of leucine/isoleucine residues would be added to non-BAG6 substrates, to determine whether this converts these misfolded proteins towards BAG6. It would be of great interest to determine whether directing toxic aggregation-prone proteins towards BAG6 mediated pathways alters their aggregation and/or toxicity. Weisberg et al. (2012) has reported that directing superoxide dismutase 1 (SOD1) G93A from the juxta nuclear quality control compartment (JUNQ) to the insoluble protein deposit (IPOD) increased cell viability.

148

The domain(s) needed for BAG6 to interact with different substrates is not well defined. Previous studies have suggested the poorly defined central proline-rich region is required for substrate binding (Leznicki et al., 2013; Xu et al., 2013; Payapilly and High, 2014). On the other hand, the N-terminal region was said to be involved in binding naturally occurring dislocated ERAD substrates (Claessen et al., 2014). To address this question, BAG6 variants having mutation or deletion in important domain would be constructed, and the role of each domain on association with different substrates (mislocalised proteins and ERAD substrates) determined through nuclear re-localisation assay and co-immunoprecipitation. Together, such studies would provide a more complete picture of how BAG6 plays its role in the cytosolic protein quality control.

149

7 References

Akutsu, M., Dikic, I. and Bremm, A. (2016) Ubiquitin chain diversity at a glance. J Cell Sci, 129(5), 875-80.

Alexander, P.A., He, Y., Chen, Y., Orban, J. and Bryan, P.N. (2009) A minimal sequence code for switching protein structure and function. Proc Natl Acad Sci U S A, 106(50), 21149-54.

Amm, I., Sommer, T. and Wolf, D.H. (2014) Protein quality control and elimination of protein waste: the role of the ubiquitin-proteasome system. Biochim Biophys Acta, 1843(1), 182-96.

Ast, T. and Schuldiner, M. (2013) All roads lead to Rome (but some may be harder to travel): SRP-independent translocation into the endoplasmic reticulum. Crit Rev Biochem Mol Biol, 48(3), 273-88.

Back, S.H., Lee, K., Vink, E. and Kaufman, R.J. (2006) Cytoplasmic IRE1alpha- mediated XBP1 mRNA splicing in the absence of nuclear processing and endoplasmic reticulum stress. J Biol Chem, 281(27), 18691-706.

Baker, D. (2000) A surprising simplicity to protein folding. Nature, 405(6782), 39- 42.

Baltz, A.G., Munschauer, M., Schwanhäusser, B., Vasile, A., Murakawa, Y., Schueler, M., Youngs, N., Penfold-Brown, D., Drew, K., Milek, M., Wyler, E., Bonneau, R., Selbach, M., Dieterich, C. and Landthaler, M. (2012) The mRNA-bound proteome and its global occupancy profile on protein-coding transcripts. Mol Cell, 46(5), 674-90.

Banerji, J., Sands, J., Strominger, J.L. and Spies, T. (1990) A gene pair from the human major histocompatibility complex encodes large proline-rich proteins with multiple repeated motifs and a single ubiquitin-like domain. Proc Natl Acad Sci U S A, 87(6), 2374-8.

Behl, C. (2016) Breaking BAG: The Co-Chaperone BAG3 in Health and Disease. Trends Pharmacol Sci, 37(8), 672-88.

Besche, H.C., Haas, W., Gygi, S.P. and Goldberg, A.L. (2009) Isolation of mammalian 26S proteasomes and p97/VCP complexes using the ubiquitin- like domain from HHR23B reveals novel proteasome-associated proteins. Biochemistry, 48(11), 2538-49.

Bhattacharyya, S., Yu, H., Mim, C. and Matouschek, A. (2014) Regulated protein turnover: snapshots of the proteasome in action. Nat Rev Mol Cell Biol, 15(2), 122-33.

150

Bose, D. and Chakrabarti, A. (2017) Substrate specificity in the context of molecular chaperones. IUBMB Life, doi: 10.1002/iub.

Brandvold, K.R. and Morimoto, R. (2015) The Chemical Biology of Molecular Chaperones--Implications for Modulation of Proteostasis. J Mol Biol, 427(18), 2931-47.

Brehme, M. and Voisine, C. (2016) Model systems of protein-misfolding diseases reveal chaperone modifiers of proteotoxicity. Dis Model Mech, 9(8), 823-38.

Brodsky, J.L. and Skach, W.R. (2011) Protein folding and quality control in the endoplasmic reticulum: Recent lessons from yeast and mammalian cell systems. Curr Opin Cell Biol, 23(4), 464-75.

Brodsky, J.L. (2012) Cleaning up: ER-associated degradation to the rescue. Cell, 151(6), 1163-7.

Buchberger, A. (2014) ERQC autophagy: Yet another way to die. Mol Cell, 54(1), 3- 4.

Casson, J., McKenna, M. and High, S. (2016) On the road to nowhere: cross-talk between post-translational protein targeting and cytosolic quality control. Biochem Soc Trans, 44(3), 796-801.

Chang, Y.W., Chuang, Y.C., Ho, Y.C., Cheng, M.Y., Sun, Y.J., Hsiao, C.D. and Wang, C. (2010) Crystal structure of Get4-Get5 complex and its interactions with Sgt2, Get3, and Ydj1. J Biol Chem, 285(13), 9962-70.

Chapman, R.E. and Walter, P. (1997) Translational attenuation mediated by an mRNA intron. Curr Biol, 7(11), 850-9.

Chapman-Smith, A. and Cronan, J.E. Jr. (1999) Molecular biology of biotin attachment to proteins. J Nutr, 129(2S Suppl), 477S-484S.

Chen, C.Y., Malchus, N.S., Hehn, B., Stelzer, W., Avci, D., Langosch, D. and Lemberg, M.K. (2014) Signal peptide peptidase functions in ERAD to cleave the unfolded protein response regulator XBP1u. EMBO J, 33(21), 2492-506.

Choi-Rhee, E., Schulman, H. and Cronan, J.E. (2004) Promiscuous protein biotinylation by Escherichia coli biotin protein ligase. Protein Sci, 13(11), 3043-50.

Christianson, J.C. and Ye, Y. (2014) Cleaning up in the endoplasmic reticulum: ubiquitin in charge. Nat Struct Mol Biol, 21(4), 325-35.

Ciechanover, A. (2015) The unravelling of the ubiquitin system. Nat Rev Mol Cell Biol, 16(5), 322-4.

151

Claessen, J.H. and Ploegh, H.L. (2011) BAT3 guides misfolded glycoproteins out of the endoplasmic reticulum. PLoS One, 6(12), e28542.

Corduan, A., Lecomte, S., Martin, C., Michel, D. and Desmots, F. (2009) Sequential interplay between BAG6 and HSP70 upon heat shock. Cell Mol Life Sci, 66(11-12), 1998-2004.

Cox, J., Hein, M.Y., Luber, C.A., Paron, I., Nagaraj, N. and Mann, M. (2014) Accurate proteome-wide label-free quantification by delayed normalization and maximal peptide ratio extraction, termed MaxLFQ. Mol Cell Proteomics, 13(9), 2513-26.

Cox, J.S. and Walter, P. (1996) A novel mechanism for regulating activity of a transcription factor that controls the unfolded protein response. Cell, 87(3), 391-404.

Dill, K.A. and MacCallum, J.L. (2012) The protein-folding problem, 50 years on. Science, 338(6110), 1042-6.

Ding, W.X., Ni, H.M., Gao, W., Yoshimori, T., Stolz, D.B., Ron, D. and Yin, X.M. (2007) Linking of autophagy to ubiquitin-proteasome system is important for the regulation of endoplasmic reticulum stress and cell viability. Am J Pathol, 171(2), 513-24.

Dissmeyer, N., Rivas, S. and Graciet, E. (2017) Life and death of proteins after protease cleavage: protein degradation by the N-end rule pathway. New Phytol, doi: 10.1111/nph.

Englander, S.W. and Mayne, L. (2014) The nature of protein folding pathways. Proc Natl Acad Sci U S A, 111(45), 15873-80.

Faesen, A.C., Luna-Vargas, M.P. and Sixma, T.K. (2012) The role of UBL domains in ubiquitin-specific proteases. Biochem Soc Trans, 40(3), 539-45.

Fernández-Fernández, M.R., Gragera, M., Ochoa-Ibarrola, L., Quintana-Gallardo, L. and Valpuesta, J.M. (2017) Hsp70 - a master regulator in protein degradation. FEBS Lett, doi: 10.1002/1873-3468.12751.

Frakes, A.E. and Dillin, A. (2017) The UPR^ER: Sensor and Coordinator of Organismal Homeostasis. Mol Cell, 66(6), 761-771.

Guo, H.H., Choe, J. and Loeb, L.A. (2004) Protein tolerance to random amino acid change. Proc Natl Acad Sci U S A, 101(25), 9205-10.

Guzman, I. and Gruebele, M. (2014) Protein folding dynamics in the cell. J Phys Chem B, 118(29), 8459-70.

Hanulová, M. and Weiss, M. (2012) Protein sorting and membrane-mediated interactions. Biophys Rev, 4(2), 117-124. 152

Harding, H.P., Zhang, Y. and Ron, D. (1999) Protein translation and folding are coupled by an endoplasmic-reticulum-resident kinase. Nature, 397(6716), 271-4.

Harding, H.P., Novoa, I., Zhang, Y., Zeng, H., Wek, R., Schapira, M. and Ron, D. (2000) Regulated translation initiation controls stress-induced gene expression in mammalian cells. Mol Cell, 6(5), 1099-108.

Haze, K., Yoshida, H., Yanagi, H., Yura, T. and Mori, K. (1999) Mammalian transcription factor ATF6 is synthesized as a transmembrane protein and activated by proteolysis in response to endoplasmic reticulum stress. Mol Biol Cell, 10(11), 3787-99.

Hegde, R.S. and Keenan, R.J. (2011) Tail-anchored membrane protein insertion into the endoplasmic reticulum. Nat Rev Mol Cell Biol, 12(12), 787-98.

Hessa, T., Sharma, A., Mariappan, M., Eshleman, H.D., Gutierrez, E. and Hegde, R.S. (2011) Protein targeting and degradation are coupled for elimination of mislocalized proteins. Nature, 475(7356), 394-7.

Hetz, C., Chevet, E. and Harding, H.P. (2013) Targeting the unfolded protein response in disease. Nat Rev Drug Discov, 12(9), 703-19.

Hicke, L., Schubert, H.L. and Hill, C.P. (2005) Ubiquitin-binding domains. Nat Rev Mol Cell Biol, 6(8), 610-21.

Hong, H. (2014) Toward understanding driving forces in membrane protein folding. Arch Biochem Biophys, 564, 297-313.

Hong, J.H., Kaustov, L., Coyaud, E., Srikumar, T., Wan, J., Arrowsmith, C. and Raught, B. (2015) KCMF1 (potassium channel modulatory factor 1) Links RAD6 to UBR4 (ubiquitin N-recognin domain-containing E3 ligase 4) and lysosome-mediated degradation. Mol Cell Proteomics, 14(3), 674-85.

Houck, S.A., Ren, H.Y., Madden, V.J., Bonner, J.N., Conlin, M.P., Janovick, J.A., Conn, P.M. and Cyr, D.M. (2014) Quality control autophagy degrades soluble ERAD-resistant conformers of the misfolded membrane protein GnRHR. Mol Cell, 54(1), 166-79.

Huang, N., Yu, Y. and Qiao, J. (2017) Dual role for the unfolded protein response in the ovary: adaption and apoptosis. Protein Cell, 8(1), 14-24.

Husnjak, K. and Dikic, I. (2012) Ubiquitin-binding proteins: decoders of ubiquitin- mediated cellular functions. Annu Rev Biochem, 81, 291-322.

Javed, A., Christodoulou, J., Cabrita, L.D. and Orlova, E.V. (2017) The ribosome and its role in protein folding: looking through a magnifying glass. Acta Crystallogr D Struct Biol, 73(Pt 6), 509-521.

153

Johnson, N., Powis, K. and High, S. (2013) Post-translational translocation into the endoplasmic reticulum. Biochim Biophys Acta, 1833(11), 2403-9.

Kadowaki, H., Nagai, A., Maruyama, T., Takami, Y., Satrimafitrah, P., Kato, H., Honda, A., Hatta, T., Natsume, T., Sato, T., Kai, H., Ichijo, H. and Nishitoh, H. (2015) Pre-emptive Quality Control Protects the ER from Protein Overload via the Proximity of ERAD Components and SRP. Cell Rep, 13(5), 944-56.

Kämper, N., Kessler, J., Temme, S., Wegscheid, C., Winkler, J. and Koch, N. (2012) A novel BAT3 sequence generated by alternative RNA splicing of exon 11B displays cell type-specific expression and impacts on subcellular localization. PLoS One, 7(4), e35972.

Kanda, S., Yanagitani, K., Yokota, Y., Esaki, Y. and Kohno, K. (2016) Autonomous translational pausing is required for XBP1u mRNA recruitment to the ER via the SRP pathway. Proc Natl Acad Sci U S A, 113(40), E5886-E5895.

Kang, S.W., Rane, N.S., Kim, S.J., Garrison, J.L., Taunton, J. and Hegde, R.S. (2006) Substrate-specific translocational attenuation during ER stress defines a pre- emptive quality control pathway. Cell, 127(5), 999-1013.

Kaushik, S. and Cuervo, A.M. (2012) Chaperones in autophagy. Pharmacol Res, 66(6), 484-93.

Kawabe, H. and Brose, N. (2011) The role of ubiquitylation in nerve cell development. Nat Rev Neurosci, 12(5), 251-68.

Kawahara, H., Minami, R. and Yokota, N. (2013) BAG6/BAT3: emerging roles in quality control for nascent polypeptides. J Biochem, 153(2), 147-60.

Kawahara, T., Yanagi, H., Yura, T. and Mori, K. (1997) Endoplasmic reticulum stress-induced mRNA splicing permits synthesis of transcription factor Hac1p/Ern4p that activates the unfolded protein response. Mol Biol Cell, 8(10), 1845-62.

Kawahara, T., Yanagi, H., Yura, T. and Mori, K. (1998) Unconventional splicing of HAC1/ERN4 mRNA required for the unfolded protein response. Sequence- specific and non-sequential cleavage of the splice sites. J Biol Chem, 273(3), 1802-7.

Khan, S. and Vihinen, M. (2007) Spectrum of disease-causing mutations in protein secondary structures. BMC Struct Biol, 7, 56.

Kikukawa, Y., Minami, R., Shimada, M., Kobayashi, M., Tanaka, K., Yokosawa, H. and Kawahara, H. (2005) Unique proteasome subunit Xrpn10c is a specific receptor for the antiapoptotic ubiquitin-like protein Scythe. FEBS J, 272(24), 6373-86.

154

Kim, D.I., Birendra, K.C., Zhu, W., Motamedchaboki, K., Doye, V. and Roux, K.J. (2014) Probing nuclear pore complex architecture with proximity-dependent biotinylation. Proc Natl Acad Sci U S A, 111(24), E2453-61.

Kim, D.I., Jensen, S.C. and Roux, K.J. (2016) Identifying Protein-Protein Associations at the Nuclear Envelope with BioID. Methods Mol Biol, 1411, 133-46.

Kim, D.I. and Roux, K.J. (2016) Filling the Void: Proximity-Based Labeling of Proteins in Living Cells. Trends Cell Biol, 26(11), 804-817.

Kim, Y.E., Hipp, M.S., Bracher, A., Hayer-Hartl, M. and Hartl, F.U. (2013) Molecular chaperone functions in protein folding and proteostasis. Annu Rev Biochem, 82, 323-55.

Kleiger, G. and Mayor, T. (2014) Perilous journey: a tour of the ubiquitin- proteasome system. Trends Cell Biol, 24(6), 352-9.

Koldewey, P., Horowitz, S., Bardwell, J.C.A. (2017) Chaperone-client interactions: Non-specificity engenders multifunctionality. J Biol Chem, 292(29), 12010- 12017.

Korolchuk, V.I., Menzies, F.M. and Rubinsztein, D.C. (2010) Mechanisms of cross- talk between the ubiquitin-proteasome and autophagy-lysosome systems. FEBS Lett, 584(7), 1393-8.

Kramer, G., Boehringer, D., Ban, N. and Bukau, B. (2009) The ribosome as a platform for co-translational processing, folding and targeting of newly synthesized proteins. Nat Struct Mol Biol, 16(6), 589-97.

Kriegenburg, F., Ellgaard, L., Hartmann-Petersen, R. (2012) Molecular chaperones in targeting misfolded proteins for ubiquitin-dependent degradation. FEBS J, 279(4), 532-42.

Krysztofinska, E.M., Martínez-Lumbreras, S., Thapaliya, A., Evans, N.J., High, S. and Isaacson, R.L. (2016) Structural and functional insights into the E3 ligase, RNF126. Sci Rep, 6:26433.

Kuwabara, N., Minami, R., Yokota, N., Matsumoto, H., Senda, T., Kawahara, H. and Kato, R. (2015) Structure of a BAG6 (Bcl-2-associated athanogene 6)-Ubl4a (ubiquitin-like protein 4a) complex reveals a novel binding interface that functions in tail-anchored protein biogenesis. J Biol Chem, 290(15), 9387-98.

Kwak, J.H., Kim, S.I., Kim, J.K. and Choi, M.E. (2008) BAT3 interacts with transforming growth factor-beta (TGF-beta) receptors and enhances TGF- beta1-induced type I collagen expression in mesangial cells. J Biol Chem, 283(28), 19816-25.

155

Kyte, J. and Doolittle, R.F. (1982) A simple method for displaying the hydropathic character of a protein. J Mol Biol, 157(1), 105-32.

Kyte, J. (2003) The basis of the hydrophobic effect. Biophys Chem, 100(1-3), 193- 203.

Lanning, D. and Lafuse, W.P. (1997) Localization of the casein kinase II beta- subunit gene within the mouse H-2 complex class III region and comparison of expression with Bat genes. Mamm Genome, 8(7), 519-21.

Lee, A.H., Iwakoshi, N.N., Anderson, K.C. and Glimcher, L.H. (2003) Proteasome inhibitors disrupt the unfolded protein response in myeloma cells. Proc Natl Acad Sci U S A, 100(17), 9946-51.

Lee, J.G. and Ye, Y. (2013) Bag6/Bat3/Scythe: a novel chaperone activity with diverse regulatory functions in protein biogenesis and degradation. Bioessays, 35(4), 377-85.

Levine, C.G., Mitra, D., Sharma, A., Smith, C.L. and Hegde, R.S. (2005) The efficiency of protein compartmentalization into the secretory pathway. Mol Biol Cell, 16(1), 279-91.

Leznicki, P., Clancy, A., Schwappach, B. and High, S. (2010) Bat3 promotes the membrane integration of tail-anchored proteins. J Cell Sci, 123(Pt 13), 2170- 8.

Leznicki, P. and High, S. (2012) SGTA antagonizes BAG6-mediated protein triage. Proc Natl Acad Sci U S A, 109(47), 19214-9.

Leznicki, P., Roebuck, Q. P., Wunderley, L., Clancy, A., Krysztofinska, E. M., Isaacson, R. L., Warwicker, J., Schwappach, B. and High, S. (2013) The association of BAG6 with SGTA and tail-anchored proteins. PLoS One, 8(3), e59590.

Leznicki, P., Korac-Prlic, J., Kliza, K., Husnjak, K., Nyathi, Y., Dikic, I. and High, S. (2015) Binding of SGTA to Rpn13 selectively modulates protein quality control. J Cell Sci, 128(17), 3187-96.

Li, P., Li, J., Wang, L. and Di, L.J. 2017 Proximity labeling of interacting proteins: Application of BioID as a discovery tool. Proteomics, doi: 10.1002/pmic.201700002.

Lilienbaum, A. (2013) Relationship between the proteasomal system and autophagy. Int J Biochem Mol Biol, 4(1), 1-26.

Lindholm, D., Korhonen, L., Eriksson, O. and Kõks, S. (2017) Recent Insights into the Role of Unfolded Protein Response in ER Stress in Health and Disease. Front Cell Dev Biol, 5, 48.

156

Liu, Y. and Ye, Y. (2012) Roles of p97-associated deubiquitinases in protein quality control at the endoplasmic reticulum. Curr Protein Pept Sci, 13(5), 436-46.

Liu, Y., Liu, C., Dong, W. and Li, W. (2016) Physiological functions and clinical implications of the N-end rule pathway. Front Med, 10(3), 258-70.

Lu, Y., Liang, F.X. and Wang, X. (2014) A synthetic biology approach identifies the mammalian UPR RNA ligase RtcB. Mol Cell, 55(5), 758-70.

Lupas, A.N. and Alva, V. (2017) Ribosomal proteins as documents of the transition from unstructured (poly)peptides to folded proteins. J Struct Biol, 198(2), 74- 81.

Mariappan, M., Li, X., Stefanovic, S., Sharma, A., Mateja, A., Keenan, R.J. and Hegde, R.S. (2010) A ribosome-associating factor chaperones tail-anchored membrane proteins. Nature, 466(7310), 1120-4.

McCaffrey, K. and Braakman, I. (2016) Protein quality control at the endoplasmic reticulum. Essays Biochem, 60(2), 227-235.

Mehta, V. and Trinkle-Mulcahy, L. (2016) Recent advances in large-scale protein interactome mapping. F1000Res, 5, pii: F1000 Faculty Rev-782.

Mellacheruvu, D., Wright, Z., Couzens, A.L., Lambert, J.P., St-Denis, N.A., Li, T., Miteva, Y.V., Hauri, S., Sardiu, M.E., Low, T.Y., Halim, V.A., Bagshaw, R.D., Hubner, N.C., Al-Hakim, A., Bouchard, A., Faubert, D., Fermin, D., Dunham, W.H., Goudreault, M., Lin, Z.Y., Badillo, B.G., Pawson, T., Durocher, D., Coulombe, B., Aebersold, R., Superti-Furga, G., Colinge, J., Heck, A.J., Choi, H., Gstaiger, M., Mohammed, S., Cristea, I.M., Bennett, K.L., Washburn, M.P., Raught, B., Ewing, R.M., Gingras, A.C. and Nesvizhskii, A.I. (2013) The CRAPome: a contaminant repository for affinity purification-mass spectrometry data. Nat Methods, 10(8), 730-6.

Minami, R., Hayakawa, A., Kagawa, H., Yanagi, Y., Yokosawa, H., Kawahara, H. (2010) BAG-6 is essential for selective elimination of defective proteasomal substrates. J Cell Biol, 190(4), 637-50.

Mock, J.Y., Chartron, J.W., Zaslaver, M., Xu, Y., Ye, Y. and Clemons, W.M.Jr. (2015) Bag6 complex contains a minimal tail-anchor-targeting module and a mock BAG domain. Proc Natl Acad Sci U S A, 112(1), 106-11.

Mogk, A. and Bukau, B. (2017) Role of sHsps in organizing cytosolic protein aggregation and disaggregation. Cell Stress Chaperones, doi: 10.1007/s12192-017-0762-4.

Mulligan, V.K. and Chakrabartty, A. (2013) Protein misfolding in the late-onset neurodegenerative diseases: common themes and the unique case of amyotrophic lateral sclerosis. Proteins, 81(8), 1285-303.

157

Murrow, L. and Debnath, J. (2013) Autophagy as a stress-response and quality- control mechanism: implications for cell injury and human disease. Annu Rev Pathol, 8, 105-37.

Nahnsen, S., Bielow, C., Reinert, K. and Kohlbacher, O. (2013) Tools for label-free peptide quantification. Mol Cell Proteomics, 12(3), 549-56.

Navon, A., Gatushkin, A., Zelcbuch, L., Shteingart, S., Farago, M., Hadar, R. and Tirosh, B. (2010) Direct proteasome binding and subsequent degradation of unspliced XBP-1 prevent its intracellular aggregation. FEBS Lett, 584(1), 67- 73.

Nillegoda, N.B., Theodoraki, M.A., Mandal, A.K., Mayo, K.J., Ren, H.Y., Sultana, R., Wu, K., Johnson, J., Cyr, D.M. and Caplan, A.J. (2010) Ubr1 and Ubr2 function in a quality control pathway for degradation of unfolded cytosolic proteins. Mol Biol Cell, 21(13), 2102-16.

Nyathi, Y., Wilkinson, B.M. and Pool, M.R. (2013) Co-translational targeting and translocation of proteins to the endoplasmic reticulum. Biochim Biophys Acta, 1833(11), 2392-402.

Olzmann, J.A., Kopito, R.R. and Christianson, J.C. (2013) The mammalian endoplasmic reticulum-associated degradation system. Cold Spring Harb Perspect Biol, 5(9), pii: a013185.

Orsi, A., Fioriti, L., Chiesa, R. and Sitia, R. (2006) Conditions of endoplasmic reticulum stress favor the accumulation of cytosolic prion protein. J Biol Chem, 281(41), 30431-8.

Ovchinnikov, Y.A. (1982) Rhodopsin and bacteriorhodopsin: structure-function relationships. FEBS Lett, 148(2), 179-91.

Ozaki, T., Hanaoka, E., Naka, M., Nakagawara, A. and Sakiyama, S. (1999) Cloning and characterization of rat BAT3 cDNA. DNA Cell Biol, 18(6), 503-12.

Panchenko, M.V. (2016) Structure, function and regulation of jade family PHD finger 1 (JADE1). Gene, 589(1), 1-11.

Payapilly, A. and High, S. (2014) BAG6 regulates the quality control of a polytopic ERAD substrate. J Cell Sci, 127(Pt 13):2898-909.

Pechmann, S., Willmund, F. and Frydman, J. (2013) The ribosome as a hub for protein quality control. Mol Cell, 49(3), 411-21.

Plumb, R., Zhang, Z.R., Appathurai, S. and Mariappan, M. (2015) A functional link between the co-translational protein translocation pathway and the UPR. Elife, 4.

158

Powers, E.T., Morimoto, R.I., Dillin, A., Kelly, J.W. and Balch, W.E. (2009) Biological and chemical approaches to diseases of proteostasis deficiency. Annu Rev Biochem, 78, 959-91.

Prabhu, N. and Sharp, K. (2006) Protein-solvent interactions. Chem Rev, 106(5), 1616-23.

Qi, L., Tsai, B. and Arvan, P. New Insights into the Physiological Role of Endoplasmic Reticulum-Associated Degradation. (2017) Trends Cell Biol, 27(6), 430-440.

Rane, N.S., Yonkovich, J.L. and Hegde, R.S. (2004) Protection from cytosolic prion protein toxicity by modulation of protein translocation. EMBO J, 23(23), 4550-9.

Rane, N.S., Kang, S.W., Chakrabarti, O., Feigenbaum, L. and Hegde, R.S. (2008) Reduced translocation of nascent prion protein during ER stress contributes to neurodegeneration. Dev Cell, 15(3), 359-70.

Rapoport, T.A., Li, L. and Park, E. (2017) Structural and Mechanistic Insights into Protein Translocation. Annu Rev Cell Dev Biol, doi: 10.1146/annurev-cellbio- 100616-060439.

Ray-Sinha, A., Cross, B.C., Mironov, A., Wiertz, E. and High, S. (2014) Endoplasmic reticulum-associated degradation of a degron-containing polytopic membrane protein. Mol Membr Biol, 26(8), 448-64.

Rinschen, M.M., Bharill, P., Wu, X., Kohli, P., Reinert, M.J., Kretz, O., Saez, I.7., Schermer, B., Höhne, M., Bartram, M.P., Aravamudhan, S., Brooks, B.R., Vilchez, D., Huber, T.B., Müller, R.U., Krüger, M. and Benzing, T. (2016) The ubiquitin ligase Ubr4 controls stability of podocin/MEC-2 supercomplexes. Hum Mol Genet, 25(7), 1328-44.

Rodrigo-Brenni, M.C., Gutierrez, E. and Hegde, R.S. (2014) Cytosolic quality control of mislocalized proteins requires RNF126 recruitment to Bag6. Mol Cell, 55(2), 227-37.

Roh, S.H., Kasembeli, M., Bakthavatsalam, D., Chiu, W. and Tweardy, D.J. (2015) Contribution of the Type II Chaperonin, TRiC/CCT, to Oncogenesis. Int J Mol Sci, 16(11), 26706-20.

Roux, K.J., Kim, D.I., Raida, M. and Burke, B. (2012) A promiscuous biotin ligase fusion protein identifies proximal and interacting proteins in mammalian cells. J Cell Biol, 196(6), 801-10.

Roux, K.J. (2013) Marked by association: techniques for proximity-dependent labeling of proteins in eukaryotic cells. Cell Mol Life Sci, 70(19), 3657-64.

159

Rüegsegger, U., Leber, J.H. and Walter, P. (2001) Block of HAC1 mRNA translation by long-range base pairing is released by cytoplasmic splicing upon induction of the unfolded protein response. Cell, 107(1), 103-14.

Ruggiano, A., Foresti, O. and Carvalho, P. (2014) Quality control: ER-associated degradation: protein quality control and beyond. J Cell Biol, 204(6), 869-79.

Ryoo, H.D. (2016) Long and short (timeframe) of endoplasmic reticulum stress- induced cell death. FEBS J, 283(20), 3718-3722.

Saibil, H. (2013) Chaperone machines for protein folding, unfolding and disaggregation. Nat Rev Mol Cell Biol, 14(10), 630-42.

Saraogi, I. and Shan, S.O. (2011) Molecular mechanism of co-translational protein targeting by the signal recognition particle. Traffic, 12(5), 535-42.

Sarkar, A. and Kellogg, G.E. (2010) Hydrophobicity--shake flasks, protein folding and drug discovery. Curr Top Med Chem, 10(1), 67-83.

Saxena, S. and Caroni, P. (2011) Selective neuronal vulnerability in neurodegenerative diseases: from stressor thresholds to degeneration. Neuron, 71(1), 35-48.

Schaeffer, R.D. and Daggett, V. (2011) Protein folds and protein folding. Protein Eng Des Sel, 24(1-2), 11-9.

Schopp, I.M., Amaya Ramirez, C.C., Debeljak, J., Kreibich, E., Skribbe, M., Wild, K. and Béthune, J. (2017) Split-BioID a conditional proteomics approach to monitor the composition of spatiotemporally defined protein complexes. Nat Commun, 8, 15690.

Schubert, U., Antón, L.C., Gibbs, J., Norbury, C.C., Yewdell, J.W. and Bennink, J.R. (2000) Rapid degradation of a large fraction of newly synthesized proteins by proteasomes. Nature, 404(6779), 770-4.

Schuldiner, M., Metz, J., Schmid, V., Denic, V., Rakwalska, M., Schmitt, H.D., Schwappach, B. and Weissman, J.S. (2008) The GET complex mediates insertion of tail-anchored proteins into the ER membrane. Cell, 134(4), 634- 45.

Schweingruber, C., Soffientini, P., Ruepp, M.D., Bachi, A. and Mühlemann, O. (2016) Identification of Interactions in the NMD Complex Using Proximity- Dependent Biotinylation (BioID). PLoS One, 11(3), e0150239.

Shao, S., Rodrigo-Brenni, M.C., Kivlen, M.H. and Hegde, R.S. (2017) Mechanistic basis for a molecular triage reaction. Science, 355(6322), 298-302.

160

Shiber, A. and Ravid, T. (2014) Chaperoning proteins for destruction: diverse roles of Hsp70 chaperones and their co-chaperones in targeting misfolded proteins to the proteasome. Biomolecules, 4(3), 704-24.

Shim, S.Y., Wang, J., Asada, N., Neumayer, G., Tran, H.C., Ishiguro, K., Sanada, K., Nakatani, Y. and Nguyen, M.D. (2008) Protein 600 is a microtubule/endoplasmic reticulum-associated protein in CNS neurons. J Neurosci, 28(14), 3604-14.

Sidrauski, C., Cox, J.S. and Walter, P. (1996) tRNA ligase is required for regulated mRNA splicing in the unfolded protein response. Cell, 87(3), 405-13.

Smith, M.H., Ploegh, H.L. and Weissman, J.S. (2011) Road to ruin: targeting proteins for degradation in the endoplasmic reticulum. Science, 334(6059), 1086-90.

Spiess, C., Meyer, A.S., Reissmann, S. and Frydman, J. (2004) Mechanism of the eukaryotic chaperonin: protein folding in the chamber of secrets. Trends Cell Biol, 14(11), 598-604.

Sriram, S.M. and Kwon, Y.T. (2010) The molecular principles of N-end rule recognition. Nat Struct Mol Biol, 17(10), 1164-5.

Stefanovic, S. and Hegde, R.S. (2007) Identification of a targeting factor for posttranslational membrane protein insertion into the ER. Cell, 128(6), 1147- 59.

Stoecklin, G. and Bukau, B. (2013) Telling right from wrong in life - cellular quality control. Nat Rev Mol Cell Biol, 14(10), 613-5.

Stolz, A., Besser, S., Hottmann, H. and Wolf, D.H. (2013) Previously unknown role for the ubiquitin ligase Ubr1 in endoplasmic reticulum-associated protein degradation. Proc Natl Acad Sci U S A, 110(38), 15271-6.

Sultana, R., Theodoraki, M.A. and Caplan, A.J. (2012) UBR1 promotes protein kinase quality control and sensitizes cells to Hsp90 inhibition. Exp Cell Res, 318(1), 53-60.

Swatek, K.N. and Komander, D. (2016) Ubiquitin modifications. Cell Res, 26(4), 399-422.

Tanaka, H., Takahashi, T., Xie, Y., Minami, R., Yanagi, Y., Hayashishita, M., Suzuki, R., Yokota, N., Shimada, M., Mizushima, T., Kuwabara, N., Kato, R. and Kawahara, H. (2016) A conserved island of BAG6/Scythe is related to ubiquitin domains and participates in short hydrophobicity recognition. FEBS J, 283(4), 662-77.

161

Tasaki, T., Mulder, L.C., Iwamatsu, A., Lee, M.J., Davydov, I.V., Varshavsky, A., Muesing, M. and Kwon, Y.T. (2005) A family of mammalian E3 ubiquitin ligases that contain the UBR box motif and recognize N-degrons. Mol Cell Biol, 25(16), 7120-36.

Thress, K., Song, J., Morimoto, R.I. and Kornbluth, S. (2001) Reversible inhibition of Hsp70 chaperone function by Scythe and Reaper. EMBO J, 20(5), 1033-41.

Tirasophon, W., Welihinda, A.A. and Kaufman, R.J. (1998) A stress response pathway from the endoplasmic reticulum to the nucleus requires a novel bifunctional protein kinase/endoribonuclease (Ire1p) in mammalian cells. Genes Dev, 12(12), 1812-24. van den Boomen, D.J. and Lehner, P.J. (2015) Identifying the ERAD ubiquitin E3 ligases for viral and cellular targeting of MHC class I. Mol Immunol, 68(2 Pt A), 106-11.

Vembar, S.S. and Brodsky, J.L. (2008) One step at a time: endoplasmic reticulum- associated degradation. Nat Rev Mol Cell Biol, 9(12), 944-57.

Vincenz-Donnelly, L. and Hipp, M.S. (2017) The endoplasmic reticulum: A hub of protein quality control in health and disease. Free Radic Biol Med, 108, 383- 393.

Walter, P. and Ron, D. (2011) The unfolded protein response: from stress pathway to homeostatic regulation. Science, 334(6059), 1081-6.

Wang, Q., Liu, Y., Soetandyo, N., Baek, K., Hegde, R. and Ye, Y. (2011) A ubiquitin ligase-associated chaperone holdase maintains polypeptides in soluble states for proteasome degradation. Mol Cell, 42(6), 758-70.

Wang, R. and Liew, C.C. (1994) The human BAT3 ortholog in rodents is predominantly and developmentally expressed in testis. Mol Cell Biochem, 136(1), 49-57.

Wang, X., Yu, Y.Y., Myers, N. and Hansen, T.H. (2013) Decoupling the role of ubiquitination for the dislocation versus degradation of major histocompatibility complex (MHC) class I proteins during endoplasmic reticulum-associated degradation (ERAD). J Biol Chem, 288(32), 23295-306.

Wang, Y., Xing, P., Cui, W., Wang, W., Cui, Y., Ying, G., Wang, X. and Li, B. (2015) Acute Endoplasmic Reticulum Stress-Independent Unconventional Splicing of XBP1 mRNA in the Nucleus of Mammalian Cells. Int J Mol Sci, 16(6), 13302-21.

Ward, R.J., Alvarez-Curto, E. and Milligan, G. (2011) Using the Flp-In™ T-Rex™ system to regulate GPCR expression. Methods Mol Biol, 746, 21-37.

162

Weisberg, S.J., Lyakhovetsky, R., Werdiger, A.C., Gitler, A.D., Soen, Y. and Kaganovich, D. (2012) Compartmentalization of superoxide dismutase 1 (SOD1G93A) aggregates determines their toxicity. Proc Natl Acad Sci U S A., 109(39), 15811-6.

Winnefeld, M., Grewenig, A., Schnölzer, M., Spring, H., Knoch, T.A., Gan, E.C., Rommelaere, J. and Cziepluch, C. (2006) Human SGT interacts with Bag- 6/Bat-3/Scythe and cells with reduced levels of either protein display persistence of few misaligned chromosomes and mitotic arrest. Exp Cell Res, 312(13):2500-14.

Wu, H., Ng, B.S. and Thibault, G. (2014) Endoplasmic reticulum stress response in yeast and humans. Biosci Rep, 34(4), pii: e00118.

Wu, Y.H., Shih, S.F. and Lin, J.Y. (2004) Ricin triggers apoptotic morphological changes through caspase-3 cleavage of BAT3. J Biol Chem, 279(18), 19264- 75.

Wunderley, L., Leznicki, P., Payapilly, A. and High, S. (2014) SGTA regulates the cytosolic quality control of hydrophobic substrates. J Cell Sci, 127(Pt 21), 4728-39.

Xu, Y., Liu, Y., Lee, J. G. and Ye, Y. (2013) A ubiquitin-like domain recruits an oligomeric chaperone to a retrotranslocation complex in endoplasmic reticulum-associated degradation. J Biol Chem, 288(25), 18068-76.

Yamano, K. and Youle, R.J. (2013) PINK1 is degraded through the N-end rule pathway. Autophagy, 9(11), 1758-69.

Yanagitani, K., Imagawa, Y., Iwawaki, T., Hosoda, A., Saito, M., Kimata, Y. and Kohno, K. (2009) Cotranslational targeting of XBP1 protein to the membrane promotes cytoplasmic splicing of its own mRNA. Mol Cell, 34(2), 191-200.

Yanagitani, K., Kimata, Y., Kadokura, H. and Kohno, K. (2011) Translational pausing ensures membrane targeting and cytoplasmic splicing of XBP1u mRNA. Science, 331(6017), 586-9.

Yao, T. (2015) A timer to coordinate substrate processing by the 26S proteasome. Nat Struct Mol Biol, 22(9), 652-3.

Ye, J., Rawson, R.B., Komuro, R., Chen, X., Davé, U.P., Prywes, R., Brown, M.S. and Goldstein, J.L. (2000) ER stress induces cleavage of membrane-bound ATF6 by the same proteases that process SREBPs. Mol Cell, 6(6), 1355-64.

163

Yoshida, H., Haze, K., Yanagi, H., Yura, T. and Mori, K. (1998) Identification of the cis-acting endoplasmic reticulum stress response element responsible for transcriptional induction of mammalian glucose-regulated proteins. Involvement of basic leucine zipper transcription factors. J Biol Chem, 273(50), 33741-9.

Yoshida, H., Matsui, T., Yamamoto, A., Okada, T. and Mori, K. (2001) XBP1 mRNA is induced by ATF6 and spliced by IRE1 in response to ER stress to produce a highly active transcription factor. Cell, 107(7), 881-91.

Yoshida, H., Oku, M., Suzuki, M. and Mori, K. (2006) pXBP1(U) encoded in XBP1 pre-mRNA negatively regulates unfolded protein response activator pXBP1(S) in mammalian ER stress response. J Cell Biol, 172(4), 565-75.

Yoshida, H., Uemura, A. and Mori, K. (2009) pXBP1(U), a negative regulator of the unfolded protein response activator pXBP1(S), targets ATF6 but not ATF4 in proteasome-mediated degradation. Cell Struct Funct, 34(1), 1-10.

Zhang, T., Xu, Y., Liu, Y. and Ye, Y. (2015) gp78 functions downstream of Hrd1 to promote degradation of misfolded proteins of the endoplasmic reticulum. Mol Biol Cell, 26(24), 4438-50.

Zheng, N., Wang, Z. and Wei, W. (2016) Ubiquitination-mediated degradation of cell cycle-related proteins by F-box proteins. Int J Biochem Cell Biol, 73, 99- 110.

Zheng, N. and Shabek, N. (2017) Ubiquitin Ligases: Structure, Function, and Regulation. Annu Rev Biochem, 86, 129-157.

164

8 Supplementary data

Supplementary data 8.1 (A): Functional enrichment (biological process) of potential BAG6 substrates.

Pathway ID Pathway Observed Matching proteins in the network description gene (labels) count GO.0006397 mRNA 15 DGCR14,DHX38,DHX8,HNRNPA processing 2B1,MBNL1,MNAT1,PABPN1,PO LR2B,PPP4R2,PQBP1,PRPF38A,S AP18,SF3A1,SNW1,WBP11 GO.0010467 gene expression 38 CRTC3,DGCR14,DHX38,DHX8,F OXK1,GTF2A2,GTF2E1,HNRNPA 2B1,IGF2BP1,KDM3B,MED21,MK I67IP,MLLT10,MNAT1,PABPN1,P OLR2B,PPP4R2,PQBP1,PRPF38A, PSME3,PWP1,RPP30,RPP38,RPS2 1,SAP18,SARNP,SF3A1,SIN3A,SI N3B,SNW1,TRIM28,TSC22D1,WB P11,WHSC1L1,WHSC2,XRCC5,X RCC6,ZNF295 GO.0016070 RNA metabolic 36 CRTC3,DGCR14,DHX38,DHX8,F process OXK1,GTF2A2,GTF2E1,HNRNPA 2B1,KDM3B,MED21,MKI67IP,ML LT10,MNAT1,PABPN1,POLR2B,P PP4R2,PQBP1,PRPF38A,PWP1,RP P30,RPP38,RPS21,SAP18,SARNP, SF3A1,SIN3A,SIN3B,SNW1,TRIM 28,TSC22D1,WBP11,WHSC1L1,W HSC2,XRCC5,XRCC6,ZNF295 GO.0090304 nucleic acid 37 BANF1,CHAF1A,CRTC3,DGCR14 metabolic ,DHX38,DHX8,FOXK1,GTF2A2,G process TF2E1,HNRNPA2B1,KDM3B,ME D21,MKI67IP,MLLT10,MNAT1,P ABPN1,POLR2B,PPP4R2,PQBP1,P RPF38A,PWP1,RPP30,RPP38,RPS2 1,SAP18,SARNP,SF3A1,SIN3A,SI N3B,SNW1,TSC22D1,WBP11,WH SC1L1,WHSC2,XRCC5,XRCC6,Z NF295 GO.0016071 mRNA 16 DGCR14,DHX38,DHX8,HNRNPA metabolic 2B1,MBNL1,MNAT1,PABPN1,PO process LR2B,PPP4R2,PQBP1,PRPF38A,R PS21,SAP18,SF3A1,SNW1,WBP11 GO.0006396 RNA processing 17 DGCR14,DHX38,DHX8,HNRNPA 2B1,MNAT1,PABPN1,POLR2B,PP P4R2,PQBP1,PRPF38A,RPP30,RPP 165

38,RPS21,SAP18,SF3A1,SNW1,W BP11 GO.0008380 RNA splicing 13 DGCR14,DHX38,DHX8,HNRNPA 2B1,PABPN1,POLR2B,PPP4R2,PQ BP1,PRPF38A,SAP18,SF3A1,SNW 1,WBP11 GO.0006139 nucleobase- 38 BANF1,CHAF1A,CRTC3,DGCR14 containing ,DHX38,DHX8,FOXK1,GTF2A2,G compound TF2E1,HNRNPA2B1,KDM3B,ME metabolic D21,MKI67IP,MLLT10,MNAT1,P process ABPN1,PFKP,POLR2B,PPP4R2,PQ BP1,PRPF38A,PWP1,RPP30,RPP38 ,RPS21,SAP18,SARNP,SF3A1,SIN 3A,SIN3B,SNW1,TSC22D1,WBP1 1,WHSC1L1,WHSC2,XRCC5,XRC C6,ZNF295 GO.0034641 cellular nitrogen 39 BANF1,CHAF1A,CRTC3,DGCR14 compound ,DHX38,DHX8,FOXK1,GTF2A2,G metabolic TF2E1,HNRNPA2B1,KDM3B,ME process D21,MKI67IP,MLLT10,MNAT1,P ABPN1,PFKP,POLR2B,PPP4R2,PQ BP1,PRPF38A,PSME3,PWP1,RPP3 0,RPP38,RPS21,SAP18,SARNP,SF 3A1,SIN3A,SIN3B,SNW1,TSC22D 1,WBP11,WHSC1L1,WHSC2,XRC C5,XRCC6,ZNF295 GO.0000398 mRNA splicing, 10 DGCR14,DHX38,DHX8,HNRNPA via spliceosome 2B1,MBNL1,PABPN1,POLR2B,PQ BP1,SF3A1,SNW1 GO.0032774 RNA 27 CRTC3,DHX38,FOXK1,GTF2A2,G biosynthetic TF2E1,KDM3B,MED21,MKI67IP, process MLLT10,MNAT1,PABPN1,POLR2 B,PQBP1,PWP1,RPS21,SAP18,SA RNP,SIN3A,SIN3B,SNW1,TRIM28 ,TSC22D1,WHSC1L1,WHSC2,XR CC5,XRCC6,ZNF295 GO.0006351 transcription, 26 CRTC3,DHX38,FOXK1,GTF2A2,G DNA-templated TF2E1,KDM3B,MED21,MKI67IP, MLLT10,MNAT1,PABPN1,POLR2 B,PQBP1,PWP1,SAP18,SARNP,SI N3A,SIN3B,SNW1,TRIM28,TSC22 D1,WHSC1L1,WHSC2,XRCC5,XR CC6,ZNF295 GO.0044260 cellular 41 BANF1,CHAF1A,CRTC3,DCAF7, macromolecule DGCR14,DHX38,DHX8,FOXK1,G metabolic TF2A2,GTF2E1,HNRNPA2B1,KD process M3B,MKI67IP,MLLT10,MNAT1,O GT,PABPN1,POLR2B,PPP4R2,PQ BP1,PRPF38A,PSME3,PWP1,RPP3 0,RPP38,RPS21,SAP18,SARNP,SF

166

3A1,SIN3A,SIN3B,SNW1,TSC22D 1,UBA1,UBE2M,UBL4A,WBP11, WHSC2,XRCC5,XRCC6,ZNF295 GO.0016032 viral process 13 BANF1,CRTC3,GTF2A2,GTF2E1, MNAT1,PABPN1,POLR2B,PSME3 ,RPS21,SNW1,WHSC2,XRCC5,XR CC6 GO.0043170 macromolecule 42 BANF1,CHAF1A,CRTC3,DCAF7, metabolic DGCR14,DHX38,DHX8,FOXK1,G process TF2A2,GTF2E1,HNRNPA2B1,IGF 2BP1,KDM3B,MKI67IP,MLLT10, MNAT1,OGT,PABPN1,POLR2B,P PP4R2,PQBP1,PRPF38A,PSME3,P WP1,RPP30,RPP38,RPS21,SAP18, SARNP,SF3A1,SIN3A,SIN3B,SNW 1,TSC22D1,UBA1,UBE2M,UBL4A ,WBP11,WHSC2,XRCC5,XRCC6,Z NF295 GO.0034645 cellular 28 CHAF1A,CRTC3,DHX38,FOXK1, macromolecule GTF2A2,GTF2E1,KDM3B,MED21, biosynthetic MKI67IP,MLLT10,MNAT1,OGT,P process ABPN1,POLR2B,PQBP1,PWP1,SA P18,SARNP,SIN3A,SIN3B,SNW1, TRIM28,TSC22D1,WHSC1L1,WH SC2,XRCC5,XRCC6,ZNF295 GO.0051171 regulation of 29 CRTC3,CYR61,GTF2A2,GTF2E1, nitrogen HNRNPA2B1,IGF2BP1,KDM3B,K compound IAA0020,MED21,MLLT10,MLLT6 metabolic ,MNAT1,OGT,PABPN1,POLR2B,P process PP4R2,PQBP1,PSME3,SAP18,SAR NP,SBNO1,SIN3A,SIN3B,SNW1,T SC22D1,WHSC1L1,XRCC5,XRCC 6,ZNF295 GO.0075713 establishment of 3 BANF1,XRCC5,XRCC6 integrated proviral latency GO.0006368 transcription 5 GTF2A2,GTF2E1,MNAT1,POLR2 elongation from B,WHSC2 RNA polymerase II promoter GO.0031323 regulation of 34 CRTC3,CYR61,GTF2A2,GTF2E1, cellular HNRNPA2B1,IGF2BP1,KDM3B,K metabolic IAA0020,MED21,MKI67IP,MLLT1 process 0,MLLT6,MNAT1,OGT,PABPN1,P FKP,POLR2B,PPP4R2,PQBP1,PSM E3,SAP18,SARNP,SBNO1,SIN3A, SIN3B,SLC9A3R1,SNW1,TSC22D 1,UBL4A,WBP11,WHSC1L1,XRC C5,XRCC6,ZNF295

167

GO.0019042 viral latency 3 BANF1,XRCC5,XRCC6 GO.0044237 cellular 43 BANF1,CHAF1A,CRTC3,CYR61, metabolic DCAF7,DGCR14,DHX38,DHX8,F process OXK1,GTF2A2,GTF2E1,HNRNPA 2B1,KDM3B,MKI67IP,MLLT10,M NAT1,OGT,PABPN1,PFKP,POLR2 B,PPP4R2,PQBP1,PRPF38A,PSME 3,PWP1,RPP30,RPP38,RPS21,SAP 18,SARNP,SF3A1,SIN3A,SIN3B,S NW1,TSC22D1,UBA1,UBE2M,UB L4A,WBP11,WHSC2,XRCC5,XRC C6,ZNF295 GO.0006357 regulation of 17 CRTC3,CYR61,FOXK1,GTF2A2,H transcription NRNPA2B1,MED21,MLLT10,MN from RNA AT1,OGT,SAP18,SIN3A,SIN3B,SN polymerase II W1,TRIM28,WHSC2,XRCC6,ZNF promoter 295 GO.2000112 regulation of 26 CRTC3,CYR61,GTF2A2,GTF2E1, cellular HNRNPA2B1,IGF2BP1,KDM3B,K macromolecule IAA0020,MED21,MLLT10,MLLT6 biosynthetic ,MNAT1,OGT,PQBP1,SAP18,SAR process NP,SBNO1,SIN3A,SIN3B,SNW1,T SC22D1,WHSC1L1,WHSC2,XRCC 5,XRCC6,ZNF295 GO.0050684 regulation of 5 HNRNPA2B1,MBNL1,PABPN1,S mRNA AP18,SNW1 processing GO.0051252 regulation of 25 CRTC3,CYR61,GTF2A2,GTF2E1, RNA metabolic HNRNPA2B1,KDM3B,MED21,ML process LT10,MLLT6,MNAT1,OGT,PABP N1,POLR2B,PQBP1,SAP18,SARN P,SBNO1,SIN3A,SIN3B,SNW1,TS C22D1,WHSC1L1,XRCC5,XRCC6, ZNF295 GO.0010468 regulation of 27 CRTC3,CYR61,GTF2A2,GTF2E1, gene expression HNRNPA2B1,IGF2BP1,KDM3B,K IAA0020,MED21,MLLT10,MLLT6 ,MNAT1,OGT,PABPN1,POLR2B,P QBP1,SAP18,SARNP,SBNO1,SIN3 A,SIN3B,SNW1,TSC22D1,WHSC1 L1,XRCC5,XRCC6,ZNF295 GO.0010556 regulation of 26 CRTC3,CYR61,GTF2A2,GTF2E1, macromolecule HNRNPA2B1,IGF2BP1,KDM3B,K biosynthetic IAA0020,MED21,MLLT10,MLLT6 process ,MNAT1,OGT,POLR2B,PQBP1,SA P18,SARNP,SBNO1,SIN3A,SIN3B, SNW1,TSC22D1,WHSC1L1,XRCC 5,XRCC6,ZNF295 GO.0019219 regulation of 26 CRTC3,CYR61,GTF2A2,GTF2E1, nucleobase- HNRNPA2B1,KDM3B,MED21,ML

168

containing LT10,MLLT6,MNAT1,OGT,PABP compound N1,POLR2B,PPP4R2,PQBP1,SAP1 metabolic 8,SARNP,SBNO1,SIN3A,SIN3B,S process NW1,TSC22D1,WHSC1L1,XRCC5 ,XRCC6,ZNF295 GO.0060255 regulation of 32 CRTC3,CYR61,GTF2A2,GTF2E1, macromolecule HNRNPA2B1,IGF2BP1,KDM3B,K metabolic IAA0020,MED21,MKI67IP,MLLT1 process 0,MLLT6,MNAT1,OGT,PABPN1,P OLR2B,PPP4R2,PQBP1,PSME3,SA P18,SARNP,SBNO1,SIN3A,SIN3B, SLC9A3R1,SNW1,TSC22D1,UBL4 A,WHSC1L1,XRCC5,XRCC6,ZNF 295 GO.0071704 organic 43 BANF1,CHAF1A,CRTC3,DCAF7, substance DGCR14,DHX38,DHX8,FOXK1,G metabolic TF2A2,GTF2E1,HNRNPA2B1,IGF process 2BP1,KDM3B,MKI67IP,MLLT10, MNAT1,OGT,PABPN1,PFKP,POL R2B,PPP4R2,PQBP1,PRPF38A,PS ME3,PWP1,RPP30,RPP38,RPS21,S AP18,SARNP,SF3A1,SIN3A,SIN3 B,SNW1,TSC22D1,UBA1,UBE2M, UBL4A,WBP11,WHSC2,XRCC5,X RCC6,ZNF295 GO.0006355 regulation of 24 CRTC3,CYR61,GTF2A2,GTF2E1, transcription, HNRNPA2B1,KDM3B,MED21,ML DNA-templated LT10,MLLT6,MNAT1,OGT,PQBP 1,SAP18,SARNP,SBNO1,SIN3A,SI N3B,SNW1,TSC22D1,WHSC1L1, WHSC2,XRCC5,XRCC6,ZNF295 GO.2001141 regulation of 24 CRTC3,CYR61,GTF2A2,GTF2E1, RNA HNRNPA2B1,KDM3B,MED21,ML biosynthetic LT10,MLLT6,MNAT1,OGT,POLR process 2B,PQBP1,SAP18,SARNP,SBNO1, SIN3A,SIN3B,SNW1,TSC22D1,W HSC1L1,XRCC5,XRCC6,ZNF295 GO.0006366 transcription 11 DHX38,FOXK1,GTF2A2,GTF2E1, from RNA MNAT1,PABPN1,POLR2B,SNW1, polymerase II TRIM28,TSC22D1,WHSC2 promoter GO.0048024 regulation of 4 HNRNPA2B1,MBNL1,SAP18,SN mRNA splicing, W1 via spliceosome GO.0050434 positive 4 MNAT1,POLR2B,SNW1,WHSC2 regulation of viral transcription GO.0044238 primary 42 BANF1,CHAF1A,CRTC3,DCAF7, metabolic DGCR14,DHX38,DHX8,FOXK1,G

169

process TF2A2,GTF2E1,HNRNPA2B1,KD M3B,MKI67IP,MLLT10,MNAT1,O GT,PABPN1,PFKP,POLR2B,PPP4 R2,PQBP1,PRPF38A,PSME3,PWP1 ,RPP30,RPP38,RPS21,SAP18,SAR NP,SF3A1,SIN3A,SIN3B,SNW1,TS C22D1,UBA1,UBE2M,UBL4A,WB P11,WHSC2,XRCC5,XRCC6,ZNF2 95 GO.0031326 regulation of 26 CRTC3,CYR61,GTF2A2,GTF2E1, cellular HNRNPA2B1,IGF2BP1,KDM3B,K biosynthetic IAA0020,MED21,MLLT10,MLLT6 process ,MNAT1,OGT,POLR2B,PQBP1,SA P18,SARNP,SBNO1,SIN3A,SIN3B, SNW1,TSC22D1,WHSC1L1,XRCC 5,XRCC6,ZNF295 GO.0006974 cellular response 10 CDKN2AIP,CHAF1A,MNAT1,PO to DNA damage LR2B,PSME3,SNW1,TRIM28,UBA stimulus 1,XRCC5,XRCC6 GO.0080090 regulation of 31 CYR61,GTF2A2,GTF2E1,HNRNP primary A2B1,IGF2BP1,KDM3B,KIAA002 metabolic 0,MED21,MKI67IP,MLLT10,MLL process T6,MNAT1,OGT,PABPN1,POLR2 B,PPP4R2,PQBP1,PSME3,SAP18,S ARNP,SBNO1,SIN3A,SIN3B,SLC9 A3R1,SNW1,TSC22D1,UBL4A,W HSC1L1,XRCC5,XRCC6,ZNF295 GO.0008152 metabolic 44 BANF1,CHAF1A,CRTC3,CYR61, process DCAF7,DGCR14,DHX38,DHX8,F OXK1,GTF2A2,GTF2E1,HNRNPA 2B1,IGF2BP1,KDM3B,MKI67IP,M LLT10,MNAT1,OGT,PABPN1,PFK P,POLR2B,PPP4R2,PQBP1,PRPF3 8A,PSME3,PWP1,RPP30,RPP38,R PS21,SAP18,SARNP,SF3A1,SIN3A ,SIN3B,SNW1,TSC22D1,UBA1,UB E2M,UBL4A,WBP11,WHSC2,XRC C5,XRCC6,ZNF295 GO.1902680 positive 14 CRTC3,CYR61,FOXK1,GTF2A2,M regulation of ED21,MLLT10,MNAT1,OGT,POL RNA R2B,SIN3A,SNW1,TRIM28,WHSC biosynthetic 2,XRCC6 process GO.0009892 negative 19 CRTC3,FOXK1,HNRNPA2B1,IGF regulation of 2BP1,MKI67IP,MNAT1,OGT,PSM metabolic E3,RCC2,SAP18,SIN3A,SIN3B,SL process C9A3R1,SNW1,WHSC1L1,WHSC 2,XRCC5,XRCC6,ZNF295 GO.0050658 RNA transport 5 DHX38,HNRNPA2B1,IGF2BP1,PA BPN1,SARNP

170

GO.0043933 macromolecular 17 CHAF1A,CRTC3,GTF2A2,HSPA4, complex subunit KDM3B,MBNL1,MKI67IP,MNAT organization 1,OGT,RPS21,SAP18,SF3A1,SIN3 A,SIN3B,SLC9A3R1,TMOD3,TRI M28 GO.0006367 transcription 6 GTF2A2,GTF2E1,MNAT1,POLR2 initiation from B,SNW1,TRIM28 RNA polymerase II promoter GO.0006403 RNA 5 DHX38,HNRNPA2B1,IGF2BP1,PA localization BPN1,SARNP GO.0043484 regulation of 4 HNRNPA2B1,PQBP1,SAP18,SNW RNA splicing 1 GO.1903900 regulation of 5 MNAT1,POLR2B,SNW1,TRIM28, viral life cycle WHSC2

171

Supplementary data 8.1 (B): Functional enrichment (cellular component) of potential BAG6 substrates.

Pathway ID Pathway Observe Matching proteins in the network description d gene (labels) count GO.0044428 nuclear part 44 BANF1,CDKN2AIP,CHAF1A,CRT C3,DCAF7,DGCR14,DHX38,DHX 8,FAM50A,FOXK1,GTF2A2,GTF2 E1,HNRNPA2B1,KDM3B,KIAA00 20,MBNL1,MCM6,MED21,MKI67I P,MLLT10,MNAT1,OGT,POLR2B, PPP4R2,PQBP1,PRPF38A,PSME3, PWP1,QRICH1,RBM12,RCC2,RPP 30,RPP38,SAP18,SARNP,SF3A1,SI N3A,SIN3B,SNW1,TRIM28,WHSC 1L1,WHSC2,XRCC5,XRCC6 GO.0031981 nuclear lumen 42 BANF1,CDKN2AIP,CHAF1A,CRT C3,DCAF7,DHX38,DHX8,FAM50 A,FOXK1,GTF2A2,GTF2E1,HNRN PA2B1,KDM3B,KIAA0020,MBNL 1,MCM6,MED21,MKI67IP,MLLT1 0,MNAT1,OGT,PABPN1,POLR2B, PPP4R2,PQBP1,PSME3,PWP1,QRI CH1,RBM12,RCC2,RPP30,RPP38, SAP18,SARNP,SF3A1,SIN3A,SIN3 B,SNW1,WHSC1L1,WHSC2,XRC C5,XRCC6 GO.0005654 nucleoplasm 37 BANF1,CDKN2AIP,CRTC3,DCAF 7,DHX38,DHX8,FAM50A,FOXK1, GTF2A2,GTF2E1,HNRNPA2B1,K DM3B,MBNL1,MCM6,MED21,M KI67IP,MLLT10,MNAT1,OGT,PA BPN1,POLR2B,PPP4R2,PQBP1,PS ME3,QRICH1,RBM12,SAP18,SAR NP,SF3A1,SIN3A,SIN3B,SNW1,T RIM28,WHSC1L1,WHSC2,XRCC5 ,XRCC6 GO.0005634 nucleus 53 BANF1,CDKN2AIP,CHAF1A,CRT C3,DCAF7,DGCR14,DHX38,DHX 8,FAM50A,FAM98B,FOXK1,GTF2 A2,GTF2E1,HNRNPA2B1,IGF2BP 1,KDM3B,KIAA0020,MBNL1,MC M6,MED21,MKI67IP,MLLT10,ML LT6,MNAT1,NUCKS1,OGT,PFKP, POLR2B,PPP4R2,PQBP1,PRPF38A ,PSME3,PWP1,QRICH1,RBM12,R CC2,RPP30,RPP38,SAP18,SARNP, SF3A1,SIN3A,SIN3B,SNW1,TRIM 28,TSC22D1,UBA1,UBL4A,WHSC

172

1L1,WHSC2,XRCC5,XRCC6,ZNF2 95 GO.0030529 ribonucleoprotein 14 DGCR14,DHX38,DHX8,HNRNPA complex 2B1,IGF2BP1,MBNL1,PABPN1,PQ BP1,PRPF38A,RPP30,RPP38,RPS2 1,SF3A1,SNW1 GO.0043227 membrane- 56 BANF1,CDKN2AIP,CHAF1A,CRT bounded C3,DCAF7,DGCR14,DHX38,DHX organelle 8,FAM50A,FAM98B,FOXK1,GTF2 A2,GTF2E1,HNRNPA2B1,HSPA4,I GF2BP1,KDM3B,KIAA0020,MBN L1,MCM6,MED21,MKI67IP,MLLT 10,MLLT6,MNAT1,NUCKS1,OGT, PFKP,POLR2B,PPP4R2,PQBP1,PR PF38A,PSME3,PWP1,QRICH1,RB M12,RPP30,RPP38,SAP18,SARNP, SF3A1,SIN3A,SIN3B,SLC9A3R1,S NCG,SNW1,TRIM28,TSC22D1,UB A1,UBE2M,UBL4A,WHSC1L1,W HSC2,XRCC5,XRCC6,ZNF295 GO.0043231 intracellular 53 BANF1,CDKN2AIP,CHAF1A,CRT membrane- C3,DCAF7,DGCR14,DHX38,DHX bounded 8,FAM50A,FAM98B,FOXK1,GTF2 organelle A2,GTF2E1,HNRNPA2B1,IGF2BP 1,KDM3B,KIAA0020,MBNL1,MC M6,MED21,MKI67IP,MLLT10,ML LT6,MNAT1,NUCKS1,OGT,PFKP, POLR2B,PPP4R2,PQBP1,PRPF38A ,PSME3,PWP1,QRICH1,RBM12,R PP30,RPP38,SAP18,SARNP,SF3A1 ,SIN3A,SIN3B,SLC9A3R1,SNW1, TRIM28,TSC22D1,UBA1,UBL4A, WHSC1L1,WHSC2,XRCC5,XRCC 6,ZNF295 GO.0071013 catalytic step 2 6 DGCR14,DHX38,DHX8,HNRNPA spliceosome 2B1,SF3A1,SNW1 GO.0005681 spliceosomal 7 DGCR14,DHX38,DHX8,HNRNPA complex 2B1,PRPF38A,SF3A1,SNW1 GO.0044451 nucleoplasm part 12 GTF2A2,MED21,MNAT1,OGT,PO LR2B,PQBP1,SAP18,SARNP,SIN3 A,SIN3B,WBP11,WHSC2 GO.0044446 intracellular 43 BANF1,CDKN2AIP,CHAF1A,CRT organelle part C3,DCAF7,DGCR14,DHX38,DHX 8,FAM50A,FOXK1,GTF2A2,GTF2 E1,HNRNPA2B1,KDM3B,KIAA00 20,MBNL1,MCM6,MED21,MKI67I P,MLLT10,MNAT1,OGT,POLR2B, PPP4R2,PRPF38A,PSME3,PWP1,Q RICH1,RBM12,RPP30,RPP38,RPS 21,SAP18,SARNP,SF3A1,SIN3B,S

173

NCG,TMOD3,TRIM28,WHSC1L1, WHSC2,XRCC5,XRCC6 GO.0032991 macromolecular 31 CHAF1A,DCAF7,DGCR14,DHX38 complex ,DHX8,FAM98B,GTF2A2,HNRNP A2B1,IGF2BP1,MBNL1,MCM6,M NAT1,OGT,PABPN1,POLR2B,PPP 4R2,PQBP1,PRPF38A,RCC2,RPP3 0,RPP38,RPS21,SAP18,SARNP,SF 3A1,SIN3B,TRIM28,UBL4A,WHS C2,XRCC5,XRCC6 GO.0043564 Ku70:Ku80 2 XRCC5,XRCC6 complex GO.0043229 intracellular 51 BANF1,CDKN2AIP,CHAF1A,CRT organelle C3,DCAF7,DGCR14,DHX38,DHX 8,FAM50A,FAM98B,FOXK1,GTF2 A2,GTF2E1,HNRNPA2B1,IGF2BP 1,KDM3B,KIAA0020,MBNL1,MC M6,MED21,MKI67IP,MLLT10,ML LT6,MNAT1,NUCKS1,OGT,PFKP, POLR2B,PPP4R2,PRPF38A,PSME 3,PWP1,QRICH1,RBM12,RPP30,R PP38,RPS21,SAP18,SARNP,SF3A1 ,SNCG,TMOD3,TRIM28,TSC22D1 ,UBA1,UBL4A,WHSC1L1,WHSC2 ,XRCC5,XRCC6,ZNF295 GO.1902494 catalytic complex 12 DCAF7,GTF2A2,MED21,MNAT1, OGT,POLR2B,PPP4R2,RPP30,RPP 38,SAP18,SIN3A,SIN3B GO.0043226 organelle 53 BANF1,CDKN2AIP,CHAF1A,CRT C3,DCAF7,DGCR14,DHX38,DHX 8,FAM50A,FAM98B,FOXK1,GTF2 A2,GTF2E1,HNRNPA2B1,HSPA4,I GF2BP1,KDM3B,KIAA0020,MBN L1,MCM6,MED21,MKI67IP,MLLT 10,MLLT6,MNAT1,NUCKS1,OGT, PFKP,POLR2B,PPP4R2,PRPF38A, PSME3,PWP1,QRICH1,RBM12,RP P30,RPP38,RPS21,SAP18,SARNP, SF3A1,SNCG,TMOD3,TRIM28,TS C22D1,UBA1,UBE2M,UBL4A,WH SC1L1,WHSC2,XRCC5,XRCC6,Z NF295 GO.0043232 intracellular non- 24 BANF1,CDKN2AIP,CHAF1A,FOX membrane- K1,GTF2E1,IGF2BP1,KIAA0020, bounded MBNL1,MKI67IP,OGT,PPP4R2,PQ organelle BP1,PWP1,RCC2,RPP30,RPP38,RP S21,SNCG,SNW1,TMOD3,TRIM28 ,WHSC1L1,XRCC5,XRCC6 GO.0010494 cytoplasmic stress 3 IGF2BP1,MBNL1,PQBP1 granule

174

GO.0005655 nucleolar 2 RPP30,RPP38 ribonuclease P complex GO.0070419 nonhomologous 2 XRCC5,XRCC6 end joining complex GO.0000228 nuclear 7 CHAF1A,MKI67IP,SIN3A,SIN3B, chromosome TRIM28,XRCC5,XRCC6 GO.0005730 nucleolus 10 CDKN2AIP,FOXK1,KIAA0020,M KI67IP,PWP1,RCC2,RPP30,RPP38, SIN3A,XRCC5 GO.0005694 chromosome 9 BANF1,CHAF1A,MKI67IP,RCC2, SNW1,TRIM28,WHSC1L1,XRCC5 ,XRCC6 GO.0000118 histone 3 SAP18,SIN3A,SIN3B deacetylase complex GO.0044424 intracellular part 52 BANF1,CDKN2AIP,CHAF1A,CRT C3,DCAF7,DGCR14,DHX38,DHX 8,FAM50A,FAM98B,FOXK1,GTF2 A2,GTF2E1,HNRNPA2B1,HSPA4,I GF2BP1,KDM3B,KIAA0020,MBN L1,MCM6,MKI67IP,MLLT10,MLL T6,MNAT1,NUCKS1,OGT,PFKP,P OLR2B,PPP4R2,PRPF38A,PSME3, PWP1,QRICH1,RBM12,RPP30,RP P38,RPS21,SAP18,SARNP,SF3A1, SNCG,TMOD3,TRIM28,TSC22D1, UBA1,UBE2M,UBL4A,WHSC1L1, WHSC2,XRCC5,XRCC6,ZNF295 GO.0016580 Sin3 complex 2 SIN3A,SIN3B GO.0000783 nuclear telomere 2 XRCC5,XRCC6 cap complex GO.0044452 nucleolar part 3 CDKN2AIP,RPP30,RPP38 GO.0044454 nuclear 6 CHAF1A,SIN3A,SIN3B,TRIM28,X chromosome part RCC5,XRCC6

175

Supplementary data 8.1 (C): Functional enrichment (molecular function) of potential BAG6 substrates.

Pathway ID Pathway Observe Matching proteins in the network description d gene (labels) count GO.0044822 poly(A) RNA 28 DHX38,DHX8,FAM50A,FAM98B, binding GPATCH8,HNRNPA2B1,IGF2BP1, KIAA0020,MBNL1,MKI67IP,NUC KS1,PABPN1,POLR2B,PRPF38A, RBM12,RCC2,RPP30,RPS21,RSRC 2,SAP18,SARNP,SF3A1,SNW1,TR IM28,UBA1,WBP11,XRCC5,XRC C6 GO.0003723 RNA binding 28 CDKN2AIP,DHX38,DHX8,FAM50 A,FAM98B,GPATCH8,IGF2BP1,K IAA0020,MBNL1,MKI67IP,NUCK S1,PABPN1,POLR2B,PRPF38A,RB M12,RCC2,RPP30,RPS21,RSRC2,S AP18,SARNP,SF3A1,SNW1,TRIM 28,UBA1,WBP11,XRCC5,XRCC6 GO.0003676 nucleic acid 36 BANF1,CCDC75,CDKN2AIP,DHX binding 38,DHX8,FAM50A,FAM98B,FOX K1,GPATCH8,GTF2E1,IGF2BP1,K IAA0020,MBNL1,MCM6,MKI67IP ,MLLT10,NUCKS1,PABPN1,POLR 2B,PQBP1,PRPF38A,RBM12,RCC 2,RPP30,RPS21,RSRC2,SAP18,SA RNP,SF3A1,SIN3A,SNW1,UBA1, WBP11,XRCC5,XRCC6,ZNF295 GO.0042162 telomeric DNA 3 HNRNPA2B1,XRCC5,XRCC6 binding GO.0097159 organic cyclic 34 BANF1,CCDC75,CDKN2AIP,DHX compound 38,DHX8,FAM50A,FAM98B,FOX binding K1,GPATCH8,GTF2E1,HSPA4,IGF 2BP1,KIAA0020,MBNL1,MLLT10, NUCKS1,PFKP,PQBP1,PRPF38A, RCC2,RPP30,RPS21,RSRC2,SAP1 8,SARNP,SF3A1,SIN3A,SNW1,UB A1,UBE2M,WBP11,XRCC5,XRCC 6,ZNF295 GO.1901363 heterocyclic 34 BANF1,CCDC75,CDKN2AIP,DHX compound 38,DHX8,FAM50A,FAM98B,FOX binding K1,GPATCH8,GTF2E1,HSPA4,IGF 2BP1,KIAA0020,MBNL1,MLLT10, NUCKS1,PFKP,PQBP1,PRPF38A, RCC2,RPP30,RPS21,RSRC2,SAP1 8,SARNP,SF3A1,SIN3A,SNW1,UB A1,UBE2M,WBP11,XRCC5,XRCC 6,ZNF295

176

GO.0005488 binding 49 BANF1,CCDC75,CDKN2AIP,CHA F1A,CRTC3,DHX38,DHX8,FAM50 A,FAM98B,FOXK1,GPATCH8,GT F2A2,GTF2E1,HSPA4,IGF2BP1,K DM3B,KIAA0020,MBNL1,MLLT1 0,MLLT6,MNAT1,NUCKS1,OGT, PFKP,PPP4R2,PQBP1,PRPF38A,PS ME3,RCC2,RPP30,RPS21,RSRC2, SAP18,SARNP,SF3A1,SIN3A,SIN3 B,SLC9A3R1,SNW1,TMOD3,TRI M28,UBA1,UBE2M,UBL4A,WBP1 1,WHSC1L1,XRCC5,XRCC6,ZNF2 95 GO.0004386 helicase activity 5 DHX38,DHX8,MCM6,XRCC5,XR CC6

177

Supplementary data 8.2: Part of the output file for hydrophobicity prediction of potential BAG6 substrates. Highlighted is the maximum length of continuous hydrophobic region.

Name: O75531 NumHydroRegions: 35 0.393258426966292 HHHHHHHHHHHHHHH------HH--H HHHHHHHHHH-HHHHHHH------

Name: Q9BW85 NumHydroRegions: 249 0.770897832817337 HHHHHHHHHHHHHH-HHHHHHHHHHH------H------HHHHHHHHHHHHHHHH HHH------HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH HHHHHHHHHHHHH HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH----- HHHHHHHHHHHH HHHHHHHHHHHH------H-----HH- HHHHHHHHHHHHHHHHHHHHHHHHHHHH H------HHHHHHHHHHH

Name: Q6UUV7 NumHydroRegions: 363 0.586429725363489 --HHHHHHHH-HHHHHHHHHHHHHHHHHHHHHHH------H-- HHHHHHHHH-H H-H-HHHHH- HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH HHH HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH---- HHHHHHHHHHHHHHHHH HHHH----HHHHHHHHH-H-H-HHHHHHH------HHHHHHHHHHHHHHHHHHHHHHHHH H--HHHHHHHH------HHHH-H------H-H ------HHHH-H-HHHH--H------H-HHHH---HHHHH------HHHHHHHHHHHHHHHHHHHH-HH-HHHHHHHHHHHHH------HH ------H-H-H--HHHH----HHHHHHH-HHHHHHHHHHHHHHHHHHHHHHHHHH- H-H------H-HHHH-HHHHHH--HHHHHHHHHHHHHHHHHHHHHHHHHHH- -- --HH------H------HHHHH

Name: P61962 NumHydroRegions: 94 0.274853801169591 HHHHHHHHHHHHHHH----HHHHHH------H-HHHHH------HHHHH--H---H-HH HHHHHH---HHH--HH------HH----HHHHHHHHHH-----HH HHHH------H HH------HHHHHHHH-HHHHHHHHHHHH------HH------HHHH------

178