Eukaryotic initiation factor 2B (eIF2B), a guanine nucleotide exchange factor (GEF), promotes protein synthesis by charging translation initiation factor 2 (eIF2) with GTP. Stress-induced phosphorylation of eIF2 on its α-subunit [eIF2(αP)] inhibits this reaction triggering a protective Integrated Stress Response (ISR). A DNA-encoded chemical library (DEL) screen for modulators of eIF2B, led to the identification of a chemical series that stabilises the inactive state of eIF2B, stimulating the ISR. Cryo-EM of compound-bound eIF2B reveals a conformational switch to the inactive state engaged by eIF2(αP). In cells, compound activity is sensitive to eIF2's phosphorylation state and to a competing eIF2B ligand (ISRIB) that activates the GEF allosterically. These findings establish the feasibility of targeting eIF2B with a drug-like allosteric inhibitor, that serves as an ISR activator (ISRAC), paving the way to explore the therapeutic potential of eIF2B-directed ISR activation.

The transcription factor ATF6α has a central role in adapting mammalian cells to ER stress via the unfolded protein response (UPR), prompting efforts to identify ATF6α modulators. Here, an unbiased genome-wide CRISPR-Cas9 screen performed in Chinese Hamster Ovary cells revealed that proteolytic processing of the ATF6α precursor to its active form was impaired in cells lacking the ER-resident solute carrier SLC33A1, a transporter previously implicated in acetyl-CoA import, sialylation, and Nε-lysine protein acetylation. Cells lacking SLC33A1 constitutively trafficked the ATF6α to the Golgi but exhibited impaired Golgi processing and activating proteolysis. IRE1α signalling was derepressed by SLC33A1 deficiency consistent with selective loss of ATF6α-mediated negative feedback in the UPR. Slc33a1 -deleted cells accumulated unmodified sialylated N-glycans, precursors to acetylated glycans, likely reflecting impaired glycan processing. Deletion of ER-localised acetyltransferases NAT8 and NAT8B, which catalyse protein Nε-lysine acetylation in the secretory pathway, did not replicate the ATF6α processing defects observed in Slc33a1 -deficient cells. Together, our findings highlight a role of SLC33A1-mediated metabolite transport in the post-ER ATF6α maturation, linking small-molecule metabolism to branch-specific signalling in the UPR.

The endoplasmic reticulum (ER) unfolded protein response (UPR) is tuned by the balance between unfolded proteins and chaperones. Reserve chaperones suppress UPR transducers via their stress-sensing luminal domains, but the underlying mechanisms remain unclear. The ER chaperone AGR2 is known to repress the UPR transducer IRE1β. Here, structural prediction, X-ray crystallography, and NMR spectroscopy identify critical interactions between an AGR2 monomer and a regulatory loop in IRE1β's luminal domain. However, in the repressive complex, it is an AGR2 dimer that binds IRE1β. Cryoelectron microscopy (cryo-EM) reconstruction explains this feature: one AGR2 protomer engages the regulatory loop, while the second asymmetrically binds IRE1β's luminal domain's C terminus, blocking IRE1β-activating dimerization. Molecular dynamic simulations indicate that the second, disruptive AGR2 protomer exploits rare fluctuations in the IRE1β dimer that expose its binding site. Thus, AGR2 disrupts IRE1β dimers to suppress the UPR, priming the system for activation by chaperone clients that compete for AGR2.

The transcription factor ATF6 has a central role in adapting mammalian cells to endoplasmic reticulum (ER) stress via the Unfolded Protein Response (UPR). This has driven efforts to identify modulators of ATF6 signalling. Here, an unbiased genome-wide CRISPR-Cas9 screen performed in Chinese Hamster Ovary (CHO) cells revealed that proteolytic processing of the ATF6α precursor to its active form was impaired in CHO cells lacking the ER-resident solute carrier SLC33A1, a transporter involved in acetyl-CoA import, sialylation and Nε-lysine protein acetylation. Cells lacking SLC33A1 constitutively trafficked the ATF6α precursor to the Golgi, but exhibit impaired subsequent Golgi processing, correlating with altered ATF6α Golgi glycosylation. SLC33A1 deficiency also deregulated activation of the IRE1 branch of the UPR, pointing to a selective loss of ATF6α-mediated negative feedback in the UPR. Notably, Slc33a1 -deleted cells accumulated higher levels of unmodified sialylated N-glycans, precursors to acetylated glycans, likely reflecting impaired glycan processing. By contrast, deletion of ER-localised acetyltransferases NAT8 and NAT8B, which catalyse protein Nε-lysine acetylation in the secretory pathway, did not replicate the ATF6α processing defects observed in Slc33a1 -deficient cells. Together, our findings highlight a role for SLC33A1-mediated metabolite transport in the post-ER maturation of ATF6α and point direct links between small-molecule metabolism and branch-specific signalling in the UPR. ### Competing Interest Statement The authors have declared no competing interest. Wellcome Trust, https://ror.org/029chgv08, 224407/Z/21Z Ministerio de Ciencia e Innovación, https://ror.org/05r0vyz12, RYC2022-035365-I

Protein phosphatase-1 catalytic subunit (PP1) joins diverse targeting subunits to form holophosphatases that regulate many cellular processes. Newly synthesized PP1 is known to be transiently sequestered in an inhibitory complex with Suppressor-of-Dis2-number-2 (SDS22) and Inhibitor-3 (I3), which is disassembled by the ATPases Associated with diverse cellular Activities plus (AAA+) protein p97. Here, we show that the SDS22-PP1-I3 complex also acts as a thermodynamic sink for mature PP1 and that cycles of SDS22-PP1-I3 formation and p97-driven disassembly regulate PP1 function and subunit exchange beyond PP1 biogenesis. Förster Resonance energy transfer (FRET) analysis of labeled proteins in vitro revealed that in the p97-mediated disassembly step, both SDS22 and I3 dissociate concomitantly, releasing PP1. In presence of a targeting subunit, for instance Growth Arrest and DNA Damage-inducible protein 34 (GADD34), liberated PP1 formed an active holophosphatase that dephosphorylated its substrate, eukaryotic translation initiation factor 2 alpha (eIF2α). Inhibition of p97 results in displacement of the GADD34 targeting subunit by rebinding of PP1 to SDS22 and I3 indicating that the SDS22-PP1-I3 complex is thermodynamically favored. Likewise, p97 inhibition in cells causes rapid sequestration of PP1 by free SDS22 and I3 at the expense of other subunits. This suggests that PP1 exists in a steady state maintained by spontaneous SDS22-PP1-I3 formation and adenosine triphosphate (ATP) hydrolysis, p97-driven disassembly that recycles active PP1 between different holophosphatase complexes to warrant a dynamic holophosphatase landscape.

Activating transcription factor 6 (ATF6) is one of three endoplasmic reticulum (ER) transmembrane stress sensors that mediate the unfolded protein response (UPR). Despite its crucial role in long-term ER stress adaptation, regulation of ATF6 alpha (α) signalling remains poorly understood, possibly because its activation involves ER-to-Golgi and nuclear trafficking. Here, we generated an ATF6α/Inositol-requiring kinase 1 (IRE1) dual UPR reporter CHO-K1 cell line and performed an unbiased genome-wide CRISPR/Cas9 mutagenesis screen to systematically profile genetic factors that specifically contribute to ATF6α signalling in the presence and absence of ER stress. The screen identified both anticipated and new candidate genes that regulate ATF6α activation. Among these, calreticulin (CRT), a key ER luminal chaperone, selectively repressed ATF6α signalling: Cells lacking CRT constitutively activated a BiP::sfGFP ATF6α-dependent reporter, had higher BiP levels and an increased rate of trafficking and processing of ATF6α. Purified CRT interacted with the luminal domain of ATF6α in vitro and the two proteins co-immunoprecipitated from cell lysates. CRT depletion exposed a negative feedback loop implicating ATF6α in repressing IRE1 activity basally and overexpression of CRT reversed this repression. Our findings indicate that CRT, beyond its known role as a chaperone, also serves as an ER repressor of ATF6α to selectively regulate one arm of the UPR.

Dephosphorylation of pSer51 of the α subunit of translation initiation factor 2 (eIF2αP) terminates signaling in the integrated stress response (ISR). A trimeric mammalian holophosphatase comprised of a protein phosphatase 1 (PP1) catalytic subunit, the conserved C-terminally located ~70 amino acid core of a substrate-specific regulatory subunit (PPP1R15A/GADD34 or PPP1R15B/CReP) and G-actin (an essential cofactor) efficiently dephosphorylate eIF2αP in vitro. Unlike their viral or invertebrate counterparts, with whom they share the conserved 70 residue core, the mammalian PPP1R15s are large proteins of more than 600 residues. Genetic and cellular observations point to a functional role for regions outside the conserved core of mammalian PPP1R15A in dephosphorylating its natural substrate, the eIF2 trimer. We have combined deep learning technology, all-atom molecular dynamics simulations, X-ray crystallography, and biochemistry to uncover binding of the γ subunit of eIF2 to a short helical peptide repeated four times in the functionally important N terminus of human PPP1R15A that extends past its conserved core. Binding entails insertion of Phe and Trp residues that project from one face of an α-helix formed by the conserved repeats of PPP1R15A into a hydrophobic groove exposed on the surface of eIF2γ in the eIF2 trimer. Replacing these conserved Phe and Trp residues with Ala compromises PPP1R15A function in cells and in vitro. These findings suggest mechanisms by which contacts between a distant subunit of eIF2 and elements of PPP1R15A distant to the holophosphatase active site contribute to dephosphorylation of eIF2αP by the core PPP1R15 holophosphatase and to efficient termination of the ISR in mammals.

Here, we present a protocol for iterative enrichment of integrated single guide RNA (sgRNA) via derivative CRISPR-Cas9 from genomic DNA (gDNA) of phenotypically sorted fixed cells. We describe steps for high-scale lentiviral production, genome-wide screening, extracting gDNA from fixed cells, cloning of integrated sgRNAs, and high-scale transformation. This protocol introduces three key advantages: (1) applicability to fixed cells, (2) bypassing epigenetic drift, and (3) pause points lowering the contamination risk. We believe this approach will benefit researchers applying somatic cell genetics in cell biology. For complete details on the use and execution of this protocol, please refer to Ordoñez et al.1.

The endoplasmic reticulum (ER)-localized Hsp70 chaperone, BiP, undergoes a rapid, reversible and inactivating post-translational modification. This covalent modification complements the slower, conventional unfolded protein response (UPR) in matching the supply of active Hsp70 chaperone to the protein folding demand within the ER lumen. Long believed to be ADP-ribosylation, we now know this modification to be AMPylation (adenylylation) of BiP's threonine 518. Here, we review the discovery of the responsible enzyme (the Fic domain-containing protein FICD), the structural and biochemical basis of the inactivating modification and the discovery of FICD's dual role as the enzyme that both AMPylates and deAMPylates BiP. The structural basis of BiP recognition by FICD and recent in vitro insights into oligomeric state-mediated regulation of FICD's antagonistic enzymatic activities are also reviewed, the latter in the context of how such a regulatory system may arise in cells. Last, we consider the physiological significance of BiP AMPylation and speculate on the fitness benefits of this metazoan-specific adaptation.

Full text Figures and data Side by side Abstract Editor's evaluation Introduction Results Discussion Materials and methods Appendix 1 Data availability References Decision letter Author response Article and author information Metrics Abstract The ATPase p97 (also known as VCP, Cdc48) has crucial functions in a variety of important cellular processes such as protein quality control, organellar homeostasis, and DNA damage repair, and its de-regulation is linked to neuromuscular diseases and cancer. p97 is tightly controlled by numerous regulatory cofactors, but the full range and function of the p97–cofactor network is unknown. Here, we identify the hitherto uncharacterized FAM104 proteins as a conserved family of p97 interactors. The two human family members VCP nuclear cofactor family member 1 and 2 (VCF1/2) bind p97 directly via a novel, alpha-helical motif and associate with p97-UFD1-NPL4 and p97-UBXN2B complexes in cells. VCF1/2 localize to the nucleus and promote the nuclear import of p97. Loss of VCF1/2 results in reduced nuclear p97 levels, slow growth, and hypersensitivity to chemical inhibition of p97 in the absence and presence of DNA damage, suggesting that FAM104 proteins are critical regulators of nuclear p97 functions. Editor's evaluation This article reports on hitherto unrecognized adaptors of p97/VCP, which is a multifunctional ATPase that unwinds diverse protein substrates subserving important roles in cell physiology. The adaptors in question, members of the FAM104 family, direct p97 to the nucleus to enable unwinding events in that location. The findings, which are supported by solid experimental observations, are valuable and will inform the work of the sizable community that studies various aspects of p97/VCP. https://doi.org/10.7554/eLife.92409.sa0 Decision letter eLife's review process Introduction The abundant, highly conserved AAA+-type ATPase p97 (also known as VCP in mammals and Cdc48 in plants and lower eukaryotes) plays a central role in the maintenance of protein and organelle homeostasis as well as genome integrity (reviewed in Ahlstedt et al., 2022; Franz et al., 2016a; Papadopoulos and Meyer, 2017). The best-characterized function of p97 is the segregation of ubiquitin-modified proteins in various protein quality control pathways for their subsequent degradation by the 26S proteasome (Brandman et al., 2012; Dantuma and Hoppe, 2012; Tanaka et al., 2010; Ye et al., 2001). p97 is also involved in the lysosomal degradation of endocytic cargo (Ritz et al., 2011) and in the selective autophagy of protein aggregates, stress granules, as well as damaged mitochondria and lysosomes (Buchan et al., 2013; Ju et al., 2008; Kim et al., 2013; Papadopoulos et al., 2017; Tanaka et al., 2010; Turakhiya et al., 2018). Moreover, p97 possesses a number of crucial nuclear functions, for instance, in DNA replication and DNA damage repair (Acs et al., 2011; Davis et al., 2012; Franz et al., 2011; Maric et al., 2014; Meerang et al., 2011; Moreno et al., 2014; Mosbech et al., 2012; Mouysset et al., 2008; Raman et al., 2011). Importantly, mutational perturbation of p97 function causes the neuromuscular degenerative disease multisystem proteinopathy 1 (MSP1) (Johnson et al., 2010; Pfeffer et al., 2022; Watts et al., 2004), and several cancers as well as viral and bacterial pathogens rely on p97 activity, making p97 an attractive target for therapeutic intervention (Das and Dudley, 2021; Deshaies, 2014; Humphreys et al., 2009; Huryn et al., 2020). The molecular basis underlying the diverse cellular functions of p97 is the partial or complete unfolding of substrate proteins by threading through the ring-shaped p97 homohexamer in an ATP hydrolysis-driven process (Blythe et al., 2017; Bodnar and Rapoport, 2017; Buchberger, 2022). Since p97 itself lacks appreciable specificity for its physiological substrates, a large number of cofactor proteins control substrate binding, subcellular localization, and oligomeric state of p97 (reviewed in Buchberger et al., 2015). Some of these cofactors bind to p97 in a mutually exclusive manner and define functionally distinct, major p97 complexes: a heterodimer of UFD1 (also known as UFD1L) and NPL4 (also known as NPLOC4) recruits ubiquitylated substrates for p97-dependent unfolding and subsequent proteasomal degradation or processing (Bays et al., 2001; Jarosch et al., 2002; Rape et al., 2001; Ye et al., 2001), whereas UBXN6 (also known as UBXD1) controls p97 functions in lysosomal and autophagic degradation pathways (Papadopoulos et al., 2017; Ritz et al., 2011) and SEP domain-containing cofactors such as p47 (also known as NSFL1C) and UBXN2B (also known as p37) mediate the maturation of protein phosphatase 1 complexes (Weith et al., 2018). At least in the case of p97-UFD1-NPL4, auxiliary cofactors from the UBA-UBX family can fine-tune the subcellular localization and/or substrate specificity of a major p97 complex (Buchberger et al., 2015; Schuberth and Buchberger, 2008). For example, the yeast cofactor Ubx2 recruits Cdc48-Ufd1-Npl4 to the endoplasmic reticulum (ER) and mitochondria to promote ER- and outer mitochondrial membrane-associated protein degradation, respectively (Mårtensson et al., 2019; Neuber et al., 2005; Schuberth and Buchberger, 2005). Similarly, the metazoan cofactors UBXN7 and FAF1 direct p97-UFD1-NPL4 to nuclear chromatin to promote the removal of ubiquitylated substrates such as CDT1 and the CMG helicase (Franz et al., 2016b; Fujisawa et al., 2022; Kochenova et al., 2022; Tarcan et al., 2022). The majority of cofactors interact with p97 through a small number of defined domains/motifs, including the UBX(-like) domain, the PUB and PUL domains, the SHP box, the VCP interacting motif (VIM), and the VCP binding motif (VBM) (reviewed in Buchberger et al., 2015). However, some p97 interactors do not possess any of these canonical binding motifs, suggesting that the current inventory of p97 cofactors is incomplete and, hence, that the full scope of the p97–cofactor network is far from being understood. Examples illustrating this point include ZFAND1 (yeast Cuz1) and GIGYF1/2 (yeast Smy2), which were recently shown to regulate p97 functions in stress granule clearance (Turakhiya et al., 2018) and the transcription stress response (Lehner et al., 2022), respectively. Here, we report the identification of the previously uncharacterized, evolutionarily conserved FAM104 protein family as p97 cofactors. We show that the human FAM104 family members VCF1 and VCF2 bind p97 directly via a non-canonical helical motif and that they associate with different p97 complexes in cells. We demonstrate that VCF1/2 promote the nuclear localization of p97 and that their loss causes impaired growth and hypersensitivity to chemical inhibition of p97. Results FAM104 proteins are a conserved family of p97 interactors In our ongoing efforts to identify p97 binding partners, we isolated multiple clones encoding the hitherto uncharacterized proteins FAM104A (also known as FLJ14775) and FAM104B (also known as FLJ20434, CXorf44) in a yeast two-hybrid screen of a human testis cDNA library (Supplementary file 1). In the following, we will use the new official names VCF1 and VCF2 for human FAM104A and FAM104B, respectively, and the term FAM104 when addressing the entire protein family. VCF1 and 2 are both expressed in various isoforms originating from alternative transcript variants (Figure 1A). For VCF1, the two-hybrid hits matched isoforms 1, 2, and 5, whereas isoforms 3 and 4, which share just the N-terminal 73 residues with isoforms 1/2, were not isolated. For VCF2, the two-hybrid hits matched isoforms 3 and 4/5/6, respectively, which differ by the insertion of a single valine residue after residue 40. Compared to isoform 4, isoforms 5 and 6 of VCF2 possess slightly different N termini. Isoforms 1 and 2, which possess 30 divergent C-terminal residues, and isoform 7, which is truncated after 46 residues, were not isolated. Of note, those isoforms of VCF1 and 2 that were isolated in the two-hybrid screen possess significant sequence homology, suggesting that they are evolutionarily related. Indeed, FAM104-like proteins are present in vertebrates as well as in many invertebrates, including insects, octopuses, and echinodermata, whereas the occurrence of two distinct, VCF1- and VCF2-like homologs appears to be restricted to mammals (Figure 1B; Figure 1—figure supplement 1A). Basically, FAM104 family members possess a predicted mono- or bipartite classical nuclear localization signal (cNLS) at or close to the N terminus and a C-terminal, highly conserved helical region, which are separated by largely unstructured stretches of variable length and lower sequence conservation (Figure 1A and B and Figure 1—figure supplement 1B). Figure 1 with 1 supplement see all Download asset Open asset VCF1/2 bind to p97. (A) Schematic overview of human VCF1 and VCF2 isoforms isolated in a yeast two-hybrid screen. Relevant amino acid residue numbers are shown, and conserved N- and C-terminal sequence motifs are indicated by purple and orange boxes, respectively. Sequence identity outside these boxes is indicated by different shades of gray (light, medium, dark). Internal deletions in VCF1 isoform 2 and VCF2 isoform 4 are indicated by thin lines. (B) Multiple-sequence alignment showing representative members of the FAM104 family. Regions with predicted alpha-helical secondary structure are indicated at the top by 'H'. The most highly conserved residues in the C-terminal sequence motif are boxed in red. Other conserved residues are boxed in black or gray, according to the degree of conservation. Numbers in square brackets indicate the length of insertions. For human VCF1 and VCF2, the sequences of isoforms 2 and 4, respectively, are shown, and numbers in round brackets indicate a 21-residue insertion present in isoforms 1 and 5 of VCF1 and a 1-residue insertion present in isoform 3 of VCF2, respectively. (C) Yeast two-hybrid analysis. Yeast PJ69-4a reporter cells transformed with the indicated combinations of bait (BD-) and prey (AD-) plasmids were spotted onto agar plates containing synthetic complete medium lacking uracil and leucine (control) or uracil, leucine, and histidine (-His). Growth was monitored after 3 d. (D) Glutathione sepharose pulldown assay using wild-type p97 and GST fusions of the indicated VCF1/2 proteins. Binding of p97 was analyzed by SDS-PAGE followed by Coomassie brilliant blue staining. (E, F) Glutathione sepharose pulldown assays as in (D), using the indicated p97 variants and GST fusions of VCF1 isoform 5 (E) and VCF2 isoform 3 (F), respectively. Arrowheads mark the position of the p97 N domain. Figure 1—source data 1 Related to Figure 1D–F. https://cdn.elifesciences.org/articles/92409/elife-92409-fig1-data1-v2.zip Download elife-92409-fig1-data1-v2.zip To validate the results of the two-hybrid screen, we tested isoforms 1, 2, and 5 of VCF1 and isoform 3 of VCF2 in directed yeast two-hybrid assays and confirmed that all four full-length proteins interact with p97 (Figure 1C). Next, we performed glutathione sepharose pulldown experiments with purified GST fusions of these four VCF1/2 proteins and found that they were all able to efficiently bind recombinant p97 (Figure 1D). In order to determine the VCF1/2 binding region of p97, we performed pulldown assays with VCF1 isoform 5 and VCF2 isoform 3 using different truncated variants of p97 (Figure 1E and F). The ND1 variant of p97 lacking the D2 ATPase domain as well as the N domain alone bound efficiently to VCF1 and 2, whereas the ΔN variant lacking the N domain did not. Taken together, these results show that FAM104 proteins are a new family of evolutionarily conserved p97 interactors, and that the N domain is necessary and sufficient for the direct binding of p97 to VCF1/2. FAM104 proteins bind to p97 via their C-terminal helix As VCF1/2 do not possess any of the canonical p97 binding motifs found in other cofactors, we next sought to map their binding site for p97. The shortest VCF1 fragment isolated in the two-hybrid screen starts at glycine residue G131 of isoform 1 (equivalent to G110 and G64 of isoforms 2 and 5, respectively). Since the C-terminal alpha-helical region shows the highest sequence conservation among FAM104 family members (Figure 1B) and is missing in those isoforms of VCF1/2 that were not isolated in the two-hybrid screen, we truncated this region from the C-terminus to test its importance for p97 binding (Figure 2A). While deletion of the four C-terminal residues did not affect the two-hybrid interaction between VCF1 isoform 5 and p97, larger deletions of 7–26 residues completely abolished p97 binding (Figure 2B), even though the truncated proteins were expressed to similar levels in the reporter yeast strain (Figure 2C). Consistent with this, deletion of the C-terminal 26 residues impaired the ability of all four VCF1/2 isoforms to bind p97 in a pulldown experiment (Figure 2D), indicating that the C-terminal region contains the p97 binding site. We therefore tested whether a peptide spanning the conserved alpha-helical and flanking residues (residues C180 to G203 in VCF1 isoform 1) can bind to p97. To that end, we immobilized the N-terminally biotinylated peptide on streptavidin sepharose beads and performed a pulldown assay with p97 (Figure 2E). The peptide pulled down full-length p97 as well as the ND1 and N domain variants, demonstrating that the C-terminal conserved region of FAM104 family proteins is not only necessary, but also sufficient for p97 binding. We also noted some residual binding of the p97 ΔN variant, suggesting that the VCF1-derived peptide has some weak affinity for a p97 region(s) outside the N domain. Figure 2 with 1 supplement see all Download asset Open asset FAM104 proteins bind to p97 via their C-terminal helix. (A) Schematic overview of the C-terminal truncations of VCF1 isoform 5 used for yeast two-hybrid analysis. Labeling as in Figure 1A. (B) Yeast two-hybrid analysis of p97 binding to the C-terminally truncated VCF1 isoform 5 variants shown in (A). (C) Expression levels of the two-hybrid fusion proteins in (B) were analyzed by western blot (WB) using antibodies against p97 and the Gal4 transactivation domain (Gal4-TA). The asterisk in the p97 blot marks a cross-reactivity with endogenous Cdc48. (D) Glutathione sepharose pulldown assay using wild-type p97 and GST fusions of the indicated full-length or C-terminally truncated VCF1/2 proteins. (E) Streptavidin sepharose pulldown assay using the biotinylated peptide CQGLYFHINQTLREAHFHSLQHRG spanning the conserved C-terminal alpha-helix and flanking residues of VCF1 (residues C180–G203 in isoform 1) and the indicated p97 variants. p97 binding to the immobilized peptide was analyzed by SDS-PAGE, followed by Coomassie brilliant blue staining. Arrowheads mark the position of the p97 N domain. (F, G) AlphaFold Multimer model of the C-terminal alpha-helix of VCF1 (turquoise) bound to the N domain of p97. (F) Overview showing binding to the subdomain cleft of the N domain. (G) Close-up view showing the interaction of the four most highly conserved residues with the N domain (green). Residue numbers refer to isoform 1 of VCF1. (H) Glutathione sepharose pulldown assay using wild-type p97 and GST fusions of the indicated full-length (wildtype, NL->AA, NL->RR) or C-terminally truncated (Cdel26) variants of VCF1 isoform 5. NL->AA, N188A/L191A double mutant; NL->RR, N188R/L191R double mutant. Figure 2—source data 1 Related to Figure 2C. https://cdn.elifesciences.org/articles/92409/elife-92409-fig2-data1-v2.zip Download elife-92409-fig2-data1-v2.zip Figure 2—source data 2 Related to Figure 2D, E, and H. https://cdn.elifesciences.org/articles/92409/elife-92409-fig2-data2-v2.zip Download elife-92409-fig2-data2-v2.zip Using AlphaFold Multimer, we next generated a structural model of the C-terminal region of VCF1 bound to the N domain of p97 (Figure 2FG). The central part of the C-terminal region was predicted as an alpha-helix that binds to the subdomain cleft of the p97 N domain (Figure 2F), which is the major binding site for other N domain binding motifs including the UBX(-like) domain and the VIM and VBM linear motifs (Buchberger et al., 2015). Closer inspection of the modeled interface revealed that the four most strongly conserved residues of the helix, Y184, N188, L191, and H195, all contact the N domain (Figure 2G; Figure 1B). The predicted interactions are centered around L191, which occupies a predominantly hydrophobic pocket in the p97 subdomain cleft. Adjacent residues contribute to a hydrogen bonding network with p97, while Y184 is predicted to exhibit an orthogonal pi-stacking with Y138 of p97. To test the importance of this interface, we mutated the two central residues N188 and L191 to either alanine or arginine. We reasoned that the NL->AA double exchange should eliminate key contacts with the N domain, whereas the NL->RR double exchange should introduce steric clashes. Intriguingly, both double mutations abolished the interaction of VCF1 with p97 as efficiently as the deletion of the entire C-terminal region (Figure 2H). Consistent with a conserved essential role of residues N188 and L191 for p97 binding, an AlphaFold Multimer model of the Drosophila melanogaster homologs of VCF1/2 (CG14229) and p97 (TER94) showed a highly similar binding interface with basically identical positions of the equivalent residues N90 and L93 (Figure 2—figure supplement 1). Together, our data identify the C-terminal alpha-helix of FAM104 proteins as a novel p97 binding motif and show that the strictly conserved residues N188 and L191 are of central importance for the interaction. VCF1/2 bind to several p97 complexes in cells To investigate whether FAM104 proteins associate with p97 in cells, we ectopically expressed N-terminally FLAG epitope-tagged VCF1/2 in HEK293T cells and performed anti-FLAG immunoprecipitations (Figure 3A). Endogenous p97 was efficiently co-precipitated with all four wild-type VCF1/2 proteins tested, despite the fact that VCF1 isoform 5 was less and VCF2 isoform 3 was much less well expressed than VCF1 isoforms 1 and 2, and that the levels of soluble p97 in the input were strongly reduced upon ectopic expression of VCF1 isoforms 1 and 2. Importantly, UFD1 and NPL4 were co-precipitated at levels roughly proportional to those of p97, strongly suggesting that VCP1/2 form complexes with p97-UFD1-NPL4. This conclusion is further supported by the finding that VCF1 Cdel26 variants lacking the C-terminal helix failed to co-precipitate not only p97, as expected, but also UFD1 and NPL4. (The expression level of the VCF2 isoform 3 Cdel26 variant was below the detection limit, precluding its analysis in this experiment.) We also detected a robust co-precipitation of the SEP domain cofactor UBXN2B with the wild-type, but not Cdel26 VCF1/2 proteins (Figure 3A and Figure 3—figure supplement 1A and B), indicating the existence of ternary p97-UBXN2B-VCF1/2 complexes and thereby confirming and extending recent data from high-throughput interaction studies (Huttlin et al., 2017; Luck et al., 2020). By contrast, we were unable to confirm the recently reported co-immunoprecipitation of the related SEP domain cofactor p47 with VCF1 (Figure 3—figure supplement 1A; Raman et al., 2015), even though our data do not exclude the possibility of a transient, conditional or weak interaction with p47. Together, our data show that FAM104 proteins can associate via their C-terminal helix with distinct cellular p97 complexes including p97-UFD1-NPL4 and p97-UBXN2B, but not with all p97 complexes. Figure 3 with 1 supplement see all Download asset Open asset VCF1/2 form complexes with p97 and several p97 cofactors in cells. (A) HEK293T cells ectopically expressing the indicated N-terminally FLAG-epitope-tagged wildtype or C-terminally truncated (Cdel26) VCF1/2 proteins were subjected to anti-FLAG immunoprecipitation (IP). Input and IP samples were immunoblotted for the FLAG epitope tag, p97, and the indicated p97 cofactors. The central empty lane had been loaded with a marker. The asterisks label marker bands cross-reactive with the FLAG and FAF1 antibody, respectively. (B) Glutathione sepharose pulldown assay, using the indicated p97 cofactors and GST-VCF1 isoform 1; U-N, UFD1-NPL4. (C) Glutathione sepharose pulldown assay, using FAF1 and GST fusions of the indicated VCF1/2 proteins. Upper panel: 2,2,2-trichloroethanol-stained gel; lower panel: immunoblot with FAF1 antibody. Figure 3—source data 1 Related to Figure 3A. https://cdn.elifesciences.org/articles/92409/elife-92409-fig3-data1-v2.zip Download elife-92409-fig3-data1-v2.zip Figure 3—source data 2 Related to Figure 3B. https://cdn.elifesciences.org/articles/92409/elife-92409-fig3-data2-v2.zip Download elife-92409-fig3-data2-v2.zip Figure 3—source data 3 Related to Figure 3C. https://cdn.elifesciences.org/articles/92409/elife-92409-fig3-data3-v2.zip Download elife-92409-fig3-data3-v2.zip We also determined the impact of ectopically expressing wild-type VCF1/2 on p97–cofactor complexes by immunoprecipitating endogenous p97 (Figure 3—figure supplement 1C). Intriguingly, the strongly expressed VCF1 isoforms 1 and 2 efficiently outcompeted the major cofactors UFD1-NPL4 and p47 as well as high molecular weight ubiquitin species presumably representing polyubiquitylated substrate proteins. By contrast, the more weakly expressed VCF1 isoform 5 and VCF2 isoform 3 did not interfere with binding of UFD1-NPL4 and p47 and even appeared to increase the association of ubiquitylated substrates with p97. These results suggest that FAM104 proteins bind very tightly to p97 and that their effect on cofactor and substrate binding depends on their expression level. Interestingly, FAF1 and UBXN7, two auxiliary cofactors important for nuclear functions of p97-UFD1-NPL4, were also co-precipitated with VCF1/2 (Figure 3A). Whereas FAF1 bound to isoforms 1 and 2 of VCF1, UBXN7 bound to all four VCF1/2 isoforms. Of note, the amounts of co-precipitated FAF1 and UBXN7 did not correlate with those of p97, UFD1, and NPL4, suggesting that the interaction of FAF1 and UBXN7 with VCF1/2 may not strictly depend on the p97-UFD1-NPL4 complex. To directly address this possibility, we performed pulldown experiments with purified p97 cofactors and found that FAF1, but not p47, UFD1-NPL4, or UBXN7, bound to VCF1 isoforms 1 and 2, whereas no binding of FAF1 to VCF1 isoform 5 and VCF2 isoform 3 was detected (Figure 3B and C). These results suggest that the N-terminal extension shared by VCF1 isoforms 1 and 2 mediates a direct interaction with FAF1. In summary, our data indicate a complex interplay of VCF1/2 and other cofactors upon binding to p97 in living cells. Ectopic expression of VCF1/2 increases nuclear p97 levels All FAM104 proteins possess a bona fide cNLS at or close to the N terminus (Figure 1A and B and Figure 1—figure supplement 1B and C). To analyze their subcellular localization, we ectopically expressed N-terminally FLAG epitope-tagged VCF1/2 in HeLa cells and performed confocal immunofluorescence microscopy using anti-FLAG antibodies (Figure 4A). Consistent with the very high score calculated by cNLS mapper (Kosugi et al., 2009), all four VCF1/2 proteins strongly accumulated in the nucleus. Intriguingly, endogenous p97 co-accumulated with VCF1/2 in the nuclei of transfected cells, as evident from the comparison with neighboring non-transfected cells or with the vector control (Figure 4A). We next explored the effect of deleting the C-terminal helix or the cNLS of VCF1 using C-terminally truncated isoforms 1 and 2 as well as cNLS-deleted isoforms 1, 2, and 5, respectively (Figure 4B). The corresponding variants of the other VCF1/2 isoforms under study were poorly expressed below the detection limit of the immunofluorescence experiments, precluding their analysis. The C-terminally truncated VCF1 Cdel26 variants localized to the nucleus but failed to effectuate a significant nuclear accumulation of p97, as expected from the loss of their p97 binding site. The cNLS-deleted VCF1 variants showed a nuclear and cytoplasmic distribution, in accordance with their small size of less than 40 kDa, and a much less pronounced nuclear accumulation of p97 (see Figure 4—figure supplement 1A for higher intensity images of full-length versus cNLS-deleted VCF1 isoform 1). Figure 4 with 1 supplement see all Download asset Open asset Ectopic expression of VCF1/2 increases nuclear p97 levels. (A) HeLa cells ectopically expressing full-length, N-terminally FLAG epitope-tagged VCF isoforms 1, 2, and 5, VCF2 isoform 3, and empty vector control, respectively, were analyzed by confocal immunofluorescence microscopy using antibodies against endogenous p97 and the FLAG epitope. Scale bars, 50 µm. All images were taken with identical acquisition settings and processed identically. (B) HeLa cells ectopically expressing N-terminally FLAG epitope-tagged VCF1 isoforms 1 and 2 lacking the C-terminal conserved helix (Cdel26) or the classical nuclear localization signal (cNLS) (delNLS), and VCF1 isoform 5 lacking the cNLS, were analyzed as in (A). Scale bars, 50 µm. (C) Quantification of the ratio of nuclear to cytoplasmic p97 signals in panels (A) and (B). Except for the vector control where all imaged cells were included (>80 cells per replicate and condition), only transfected cells (as judged by the FLAG channel) were included in the quantification. Shown is the mean ± SD; n = 2 biological replicates with 15–30 transfected cells per replicate and condition; one-way ANOVA. *p<0.05; **p<0.01; ***p<0.001; the differences between the Cdel26 and delNLS constructs and the vector control are all not significant (p>0.87). (D) Quantification of the ratio of nuclear to cytoplasmic FLAG signals in panels (A) and (B) was performed as described in (C). Figure 4—source data 1 Related to Figure 4C and D. https://cdn.elifesciences.org/articles/92409/elife-92409-fig4-data1-v2.zip Download elife-92409-fig4-data1-v2.zip To quantify the microscopy results, we used CellProfiler (McQuin et al., 2018) for segmentation of the images into nuclei and cytoplasm (see Figure 4—figure supplement 1B for an example) and determined the ratio of nuclear to cytoplasmic p97 signal in control-transfected versus VCF1/2-transfected cells (Figure 4C). The ratio increased strongly from 2.08 ± 0.01 for control-transfected cells to up to 15.46 ± 3.21 for cells transfected with wild-type VCF1 isoform 1, reflecting the robust nuclear accumulation of p97 in the presence of VCF1/2. By contrast, no significant nuclear accumulation of p97 was observed for cells transfected with the Cdel26 or the cNLS-deleted VCF1 variants. Quantification of the FLAG signals showed that the wild-type and Cdel26 VCF1/2 proteins strongly accumulated in the nucleus, whereas the cNLS-deleted variants were detected in both cytoplasm and nucleus (Figure 4D). Together, these results suggest that VCF1/2 target p97 via a piggy-back mechanism to the nucleus. Ectopic expression of VCF1 isoforms 1 and 2 induces chromatin binding of p97 As the overexpression of VCF1 isoforms 1 and 2 had caused a noticeable reduction in the p97 input levels in the immunoprecipitation experiments described above (Figure 3A and Figure 3—figure supplement 1A and C), we hypothesized that a subpopulation of p97 may have become insoluble under these conditions. To test this possibility, we performed a biochemical fractionation of HEK293T cells ectopically expressing VCF1 isoform 1 or 2 (Figure 5A). Both isoforms strongly accumulated in the soluble nucleoplasmic and insoluble chromatin fractions and were only weakly detected in the cytoplasmic fraction, in full agreement with the immunofluorescence experiments. Intriguingly, the overexpression of both VCF1 isoforms did not affect the total amount of p97, as judged by the signal in SDS-denatured total cell extracts, but caused a strong increase of p97 in the chromatin fraction at the expense of the cytoplasmic pool. Because the other two VCF1/2 isoforms under study were only weakly expressed, their potential to cause similar effects could not be addressed in these experiments. To confirm that the reduced solubility of p97 is indeed caused by its association with chromatin, we used an alternative fractionation protocol including a benzonase treatment step to solubilize chromatin-bound proteins (Figure 5B). Comparing cells overexpressing VCF1 isoform 1 with control cells, we found that the majority of p97 was released into the solubilized chromatin fraction upon benzonase treatment, similar to VCF1 isoform 1 itself and the chromatin-associated proteins MYC, MCM7, and Ub-H2B. Figure 5 with 1 supplement see all Download asset Open asset Ectopic expression of VCF1 isoforms 1 and 2 promotes the association of p97 with chromatin. (A) HEK293T cells were transfected with plasmids encoding the indicated N-terminally FLAG-epitope-tagged VCF1 proteins or with empty vector (control). Total protein extracts were prepared by direct boiling part of the cells in SDS-PAGE sample buffer (SDS). The remaining cells were processed to cytoplasmic, soluble nuclear (nucleoplasmic), and chromatin fractions as indicated. Tubulin and ubiquitylated histone H2B (Ub-H2B) served as markers for the cytoplasmic and chromatin fractions, respectively, whereas the Coomassie staining of the membrane (total protein stain) served as loading control. (B) Fractionation of lysates from HEK293T cells ectopically expressing VCF1 isoform 1 or control cells using an alternative protocol including benzonase treatment that a

Full text Figures and data Side by side Abstract eLife assessment eLife digest Introduction Results Discussion Materials and methods Data availability References Peer review Author response Article and author information Abstract A long-standing goal of amyloid research has been to characterize the structural basis of the rate-determining nucleating event. However, the ephemeral nature of nucleation has made this goal unachievable with existing biochemistry, structural biology, and computational approaches. Here, we addressed that limitation for polyglutamine (polyQ), a polypeptide sequence that causes Huntington’s and other amyloid-associated neurodegenerative diseases when its length exceeds a characteristic threshold. To identify essential features of the polyQ amyloid nucleus, we used a direct intracellular reporter of self-association to quantify frequencies of amyloid appearance as a function of concentration, conformational templates, and rational polyQ sequence permutations. We found that nucleation of pathologically expanded polyQ involves segments of three glutamine (Q) residues at every other position. We demonstrate using molecular simulations that this pattern encodes a four-stranded steric zipper with interdigitated Q side chains. Once formed, the zipper poisoned its own growth by engaging naive polypeptides on orthogonal faces, in a fashion characteristic of polymer crystals with intramolecular nuclei. We further show that self-poisoning can be exploited to block amyloid formation, by genetically oligomerizing polyQ prior to nucleation. By uncovering the physical nature of the rate-limiting event for polyQ aggregation in cells, our findings elucidate the molecular etiology of polyQ diseases. eLife assessment The authors investigate the mechanism of amyloid nucleation in a cellular system using novel ratiometric measurements, providing fundamental insight into the role of polyglutamine length and the sequence features of glutamine-rich regions in amyloid formation. The problem addressed by this study is very significant and the ability to assess nucleation in cells is of considerable value. The data, as presented and analyzed, are mostly convincing. https://doi.org/10.7554/eLife.86939.3.sa0 About eLife assessments eLife digest Diseases that typically occur later in life, such as Alzheimer’s, are often caused by specific proteins clumping together into structures known as amyloids. Once the process starts, amyloids will continue to form, leading to worse symptoms that cannot be cured. The best way to treat these diseases is therefore to stop amyloids from arising in the first place. Amyloids initially develop by proteins coming together to create an unstable structure referred to as the nucleus. The instability of the nucleus means it cannot be observed directly, making it hard to study this nucleation process. To overcome this, Kandola, Venkatesan et al. investigated the simplest protein known to form an amyloid – polyglutamine, which is made up of a chain of repeating building blocks known as amino acids. Polyglutamine forms only one type of amyloid which is associated with nine neurodegenerative diseases, including Huntington’s disease. However, it only does this when its chain of amino acids exceeds a certain length, suggesting that a specific structure may be required for nucleation to begin. Kandola, Venkatesan et al. made alternative versions of the polyglutamine protein which each contained slightly different sequences of amino acids that will alter the way the protein folds. They then tested how well these different variants could form amyloids in yeast cells. This revealed that in order to join together into a nucleus, polyglutamine needs to be able to fold into a zipper shape made up of four interlocking strands. The length of the protein required to form this shape is also the same length that causes the amyloid associated with neurodegenerative diseases. Kandola, Venkatesan et al. also found that polyglutamine tends to bind to nuclei that have already formed in a way that hinders their growth. This ‘self-poisoning’ affect could potentially be exploited as a way to pre-emptively stop amyloids from initially arising. These findings have uncovered a potential therapeutic strategy for blocking amyloid formation that could eventually benefit people with or at risk of developing neurodegenerative diseases linked to polyglutamine. Additionally, this approach provides a blueprint for understanding how other proteins undergo amyloid nucleation, including those responsible for Alzheimer’s, Parkinson’s, and other diseases. Introduction Amyloids are highly ordered protein aggregates with self-templating activity. This activity drives the progression of multiple incurable diseases of aging, such as Alzheimer’s (Chiti and Dobson, 2017; Huang et al., 2019). Understanding how amyloids start, or nucleate, is therefore fundamental to preventing these diseases. However, despite decades of intense research, we still lack a clear picture of even the gross anatomy of an amyloid nucleus. Polyglutamine (polyQ) is an amyloid-forming sequence common to eukaryotic proteomes (Mier et al., 2020). In humans, it is responsible for nine invariably fatal neurodegenerative diseases, the most prevalent of which is Huntington’s Disease. Cells exhibiting polyQ pathology, whether in Huntington’s disease patients, tissue culture, organotypic brain slices, or animal models, die independently and stochastically with a constant frequency (Clarke et al., 2000; Linsley et al., 2019). Two observations suggest that this kinetic feature emerges directly from the amyloid nucleation barrier. First, polyQ aggregation itself occurs stochastically in cells (Colby et al., 2006; Kakkar et al., 2016; Sinnige et al., 2021). Second, polyQ disease onset and progression are determined almost entirely by an intramolecular change, specifically, a genetically encoded expansion in the number of sequential glutamines beyond a protein-specific threshold of approximately 36 residues (Lieberman et al., 2019). Unlike other amyloid diseases (Book et al., 2018; Kim et al., 2020; Selkoe and Hardy, 2016), polyQ disease severity generally does not worsen with gene dosage (Cubo et al., 2019; Lee et al., 2019; Wexler et al., 1987), implying that the rate-determining step in neuronal death occurs in a fixed minor fraction of the polyglutamine molecules. It is unclear how, and why, those molecules differ from the bulk. Hence, more so than for other proteopathies, therapeutic progress against polyQ diseases awaits detailed knowledge of the very earliest steps of amyloid formation. Amyloid nuclei cannot be observed directly by any existing experimental approach. This is because unlike mature amyloid fibrils, which are stable and amenable to structural biology, nuclei are unstable by definition. Their structures do not necessarily propagate into or correspond with the structures of mature amyloids that arise from them (Auer et al., 2008; Buell, 2017; Erdemir et al., 2009; Hsieh et al., 2017; Levin et al., 2014; Liang et al., 2018; Li et al., 2010; Phan and Schmit, 2020; Sil et al., 2018; Yamaguchi et al., 2005; Zanjani et al., 2020). Moreover, amyloid nucleation by a full length polypeptide occurs far too infrequently, and involves far too many degrees of freedom, to simulate with high resolution from a naive state (Barrera et al., 2021; Kar et al., 2011; Strodel, 2021). Unlike phase separation, which primarily concerns a loss of intermolecular entropy, amyloid formation additionally involves a major loss of intramolecular entropy. That is, nucleation selects for a specific combination of backbone and side chain torsion angles (Khan et al., 2018; Vitalis and Pappu, 2011; Zhang and Schmit, 2016c). As a consequence, amyloid-forming proteins can accumulate to supersaturating concentrations while remaining soluble, thereby storing potential energy that will subsequently drive their aggregation following a stochastic nucleating event (Buell, 2017; Khan et al., 2018). Due to the improbability that the requisite increases in both density and conformational ordering will occur spontaneously at the same time (i.e. homogeneously), amyloid nucleation tends to occur heterogeneously. In other words, amyloids tend to emerge via a progression of relatively less ordered but more probable, metastable intermediates of varying stoichiometry and conformation (Auer et al., 2008; Buell, 2017; Hsieh et al., 2017; Levin et al., 2014; Liang et al., 2018; Li et al., 2010; Serio et al., 2000; Sil et al., 2018; Vekilov, 2012; Vitalis and Pappu, 2011; Yamaguchi et al., 2005; Zanjani et al., 2020). A major body of current opinion holds that such heterogeneities tend to involve liquid-liquid phase separation, whereby the protein first forms disordered multimers that then facilitate productive conformational fluctuations (Borcherds et al., 2021; Camino et al., 2021; Crick et al., 2013; Crick et al., 2006; Dignon et al., 2020; Fisher et al., 2021; Halfmann et al., 2011; Peskett et al., 2018; Posey et al., 2018; Vitalis and Pappu, 2011; Yang and Yang, 2020). Heterogeneities divide the nucleation barrier into smaller, more probable steps. Therefore, the occurrence of heterogeneities implies that the nature of the actual nucleus for a given amyloid can depend not only on the protein’s sequence but also its concentration and cellular factors that influence its conformation (Bradley et al., 2002; Buell, 2017; Collinge and Clarke, 2007; Phan and Schmit, 2022; Sanders et al., 2014; Törnquist et al., 2018), as illustrated in Figure 1A. Figure 1 with 2 supplements see all Download asset Open asset A hidden pattern of glutamines governs amyloid nucleation. (A) Reaction coordinate diagrams schematizing the energy barriers governing amyloid nucleation. Because amyloid involves a transition from disordered monomer to ordered multimer, the nucleation barrier results from a combination of high energy fluctuations in both density (blue) and conformation (red). The present study manipulates either component of the nucleation barrier by changing the protein’s concentration (middle panel, illustrating a higher concentration surpassing a phase boundary for condensation) or the presence/absence of a conformational heterogeneity (right panel, with the amyloid template, [PIN+]). Nucleation then depends on the remaining fluctuation, as illustrated by vertical arrows. Cartoon images illustrate the relevant reactant species (naive monomers, condensate, [PIN+]), products (amyloid), and transition states (nuclei). Note that while the relative heights of the blue and red barriers are arbitrary, we illustrate the latter as higher in keeping with the findings of Khan et al., 2018 for prion-like amyloid nucleation. (B) Cellular volumes quantize amyloid nucleation. Amyloid nuclei occur at such low concentrations that fewer than one exists in the femtoliter volumes of cells. This causes amyloid formation to be rate-limited by stochastic nucleation in individual yeast cells (right) but not in the microliter volumes of conventional in vitro kinetic assays (left). Taking a population-level snapshot of the extent of protein self-assembly as a function of concentration in each cell reveals heterogeneity attributable to the nucleation barrier. (C) DAmFRET plots, showing the extent of de novo self-assembly (AmFRET) as a function of protein concentration for polyglutamine (Q60) or polyasparagine (N60) in yeast cells, lacking endogenous amyloid ([pin-]). Cells expressing Q60 partition into distinct populations that either lack (no AmFRET) or contain (high AmFRET) amyloid. The bimodal distribution persists even among cells with the same concentration of protein, indicating that amyloid formation is rate-limited by nucleation. The nucleation barrier for N60 is so large that spontaneous amyloid formation occurs at undetectable frequencies. Insets show histograms of AmFRET values. AU, arbitrary units. (D) Bar plot of the fraction of [pin-] cells in the AmFRET-positive population for the indicated length variants of polyQ, along with the pathologic length thresholds for polyQ tracts in the indicated proteins. Shown are means +/- SEM of biological triplicates. CACNA1A, Cav2.1; ATXN2, Ataxin-2; HTT, Huntingtin; AR, Androgen Receptor; ATXN7, Ataxin-7; ATXN1, Ataxin-1; TBP, TATA-binding protein; ATN1, Atrophin-1; ATXN3, Ataxin-3. (E) Bar plot of the fraction of [pin-] cells in the AmFRET-positive population for the indicated sequences, showing that amyloid is inhibited when Q tracts are interrupted by a non-Q side chain at odd-numbered intervals. Shown are means +/- SEM of biological triplicates. Heterogeneities may be responsible for amyloid-associated proteotoxicity. Partially ordered species accumulate during the early stages of amyloid aggregation by all well-characterized pathological amyloids, but generally do not occur (or less so) during the formation of functional amyloids (Otzen and Riek, 2019). In the case of pathologically expanded polyQ, or the huntingtin protein containing pathologically expanded polyQ, amyloid-associated oligomers have been observed in vitro, in cultured cells, and in the brains of patients (Auer et al., 2008; Hsieh et al., 2017; Legleiter et al., 2010; Levin et al., 2014; Liang et al., 2018; Li et al., 2010; Olshina et al., 2010; Sathasivam et al., 2010; Sil et al., 2018; Takahashi et al., 2008; Vitalis and Pappu, 2011; Yamaguchi et al., 2005; Zanjani et al., 2020), and are likely culprits of proteotoxicity (Kim et al., 2016; Leitman et al., 2013; Lu and Palacino, 2013; Matlahov and van der Wel, 2019; Miller et al., 2011; Takahashi et al., 2008; Wetzel, 2020). In contrast, mature amyloid fibers are increasingly viewed as relatively benign or even protective (Arrasate et al., 2004; Kim et al., 2016; Leitman et al., 2013; Lu and Palacino, 2013; Mario Isas et al., 2021; Matlahov and Matlahov and van der Wel, 2019; Takahashi et al., 2008; Wetzel, 2020). The fact that polyQ disease kinetics is governed by amyloid nucleation, but the amyloids themselves may not be responsible, presents a paradox. A structural model of the polyQ amyloid nucleus can be expected to resolve this paradox by illuminating a path for the propagation of conformational order into distinct multimeric species that either cause or mitigate toxicity. Structural features of amyloid nucleation may be deduced by studying the effects on amyloid kinetics of rational mutations made to the polypeptide (Thakur and Wetzel, 2002). The occurrence of density heterogeneities confounds this approach, however, as they blunt the dependence of amyloid kinetics on concentration that could otherwise reveal nucleus stoichiometry (Michaels et al., 2023; Vitalis and Pappu, 2011). Resolving this problem would require an assay that can detect nucleation events independently of amyloid growth kinetics, and which can scale to accommodate the large numbers of mutations necessary to identify sequence-structure relationships. Classic assays of in vitro amyloid assembly kinetics are poorly suited to this task, both because of their limited throughput, and because experimentally tractable reaction volumes are typically too large to observe discrete aggregation events (Michaels et al., 2017). We recently developed an assay to circumvent these limitations (Khan et al., 2018; Posey et al., 2021; Venkatesan et al., 2019). Distributed Amphifluoric FRET (DAmFRET) uses a monomeric photoconvertible fusion tag and high throughput flow cytometry to treat living cells as femtoliter-volume test tubes, thereby providing the large numbers of independent reaction vessels of exceptionally small volume that are required to discriminate independent nucleation events under physiological conditions (Figure 1B). Compared to conventional reaction volumes for protein self-assembly assays, the budding yeast cells employed in DAmFRET increase the dependence of amyloid formation on nucleation by nine orders of magnitude (the difference in volumes), allowing amyloid to form in the same nucleation-limited fashion as it does in afflicted neurons (Colby et al., 2006; Wetzel, 2006). DAmFRET employs a high-variance expression system to induce a hundred-fold range of concentrations of the protein of interest. The large intramolecular fluctuation required for amyloid nucleation manifests as a discontinuity in the relationship of ratiometric FRET (‘AmFRET’) to concentration, and in this way allows DAmFRET to distinguish amyloid formation from finite oligomerization and liquid-liquid phase separation. AmFRET values report the FRET intensity per molecule of protein; hence they plateau with concentration as the fraction of molecules in a given self-assembled state approaches unity. Unlike in classical in vitro kinetic experiments, where the protein concentration does not vary with time, in typical DAmFRET experiments, the concentration of protein increases monotonically with time in each cell (Li et al., 2000; Posey et al., 2021) over the hours-long induction prior to analysis. Nucleation occurs semi-stochastically during this time, such that the measured fraction of cells with amyloid at a given concentration reflects the cumulative probability of nucleation while transiting all lower concentrations. One can therefore qualitatively infer how nucleation depends on concentration from single DAmFRET plots. In short, by taking a snapshot of the protein’s concentration and extent of self-association in each cell, at a point in time appropriate for the kinetics of amyloid formation (16 hr), DAmFRET probes the existence and magnitude of critical density fluctuations that may govern nucleation. The yeast system additionally allows for orthogonal experimental control over the critical conformational fluctuation. This is because yeast cells normally contain exactly one cytosolic amyloid species—a prion state formed by the low abundance Q-rich endogenous protein, Rnq1 (Kryndushkin et al., 2013; Nizhnikov et al., 2014). The prion state can be gained or eliminated experimentally to produce cells whose sole difference is whether the Rnq1 protein does ([PIN+]) or does not ([pin-]) populate an amyloid conformation (Derkatch et al., 2001). [PIN+] serves as a partial template for amyloid formation by compositionally similar proteins including polyQ (Alexandrov et al., 2008; Duennwald et al., 2006a; Meriin et al., 2002; Serpionov et al., 2015). Known as cross-seeding, this phenomenon is analogous to, but much less efficient than, amyloid elongation by molecularly identical species (Keefer et al., 2017; Khan et al., 2018; Serio, 2018). By evaluating nucleation frequencies as a function of concentration in both [pin-] and [PIN+] cells, we can uncouple the two components of the nucleation barrier and thereby relate specific sequence features to the nucleating conformation. Sequence changes that are specific to the sequence-encoded nucleus will more strongly influence nucleation in [pin-] cells than in [PIN+] cells. PolyQ is an ideal polypeptide with which to deduce the physical nature of a pathologic amyloid nucleus. It has zero complexity, which simplifies the design and interpretation of sequence variants. Unlike other pathogenic protein amyloids, which occur as different structural polymorphs each plausibly with their own structural nucleus, polyQ amyloids have an invariant core structure under different assembly conditions and in the context of different flanking domains (Boatz et al., 2020; Galaz-Montoya et al., 2021; Lin et al., 2017; Schneider et al., 2011). This core contains antiparallel β-sheets (Buchanan et al., 2014; Matlahov and Matlahov and van der Wel, 2019; Schneider et al., 2011), while most other amyloids, including all other Q-rich amyloids, have a parallel β-sheet core (Eisenberg and Jucker, 2012; Margittai and Langen, 2008). This suggests that amyloid nucleation and pathogenesis of polyQ may follow from an ability to spontaneously acquire a specific nucleating conformation. Any variant that decelerates nucleation can therefore be interpreted with respect to its effect on that conformation. The experiments and results reported herein reveal that the polyQ amyloid nucleus is a steric zipper encoded by a pattern of approximately twelve Q residues in a single polypeptide molecule. We find that the clinical length threshold for polyQ disease—approximately 36 residues—is the minimum length that encompasses the pattern. Consistent with a monomeric nucleus, amyloid formation occurred less frequently at high concentrations or when polyQ was expressed in oligomeric form. We further found that the nucleus promotes its own kinetic arrest by templating competing dimensions of Q zipper ordering—both along the amyloid axis and orthogonally to it. This leads to the accumulation of partially ordered aggregates. Results A hidden pattern of glutamines governs amyloid nucleation To validate the use of DAmFRET for pathologic polyQ, we first surveyed the length-dependence for amyloid formation by polyQ tracts expressed as fusions to mEos3.1 in nondividing [pin-] yeast cells. We genetically compromised protein degradation in these cells to prevent differential turnover of potential polyQ heterogeneities, which could otherwise obscure the relationship of aggregation to concentration. Our data recapitulated the pathologic threshold—Q lengths 35 and shorter lacked AmFRET, indicating a failure to aggregate or even appreciably oligomerize, while Q lengths 40 and longer did acquire AmFRET in a length and concentration-dependent manner (Figure 1C–D; Figure 1—figure supplement 1A). The values of AmFRET at high concentrations decreased with polyQ length, as expected because longer molecules have a lower mass fraction of fluorophore. The cells partitioned into two discontinuous populations: one with high AmFRET and the other with none (top and bottom populations in Figure 1C). The two populations occurred with overlapping concentration ranges (red dashed box in Figure 1C, top). This suggests that a kinetic barrier associated with conformational ordering limits the acquisition of the high-FRET state (Khan et al., 2018). The two populations were well-resolved at high concentrations, where relatively few cells populated intermediate values of AmFRET. These features together indicate that aggregation of polyQ is rate-limited by a nucleation barrier associated with a rare conformational fluctuation. Sixty residues proved to be the optimum length to observe both the pre- and post-nucleated states of polyQ in single experiments, and corresponds clinically to disease onset in early adulthood (Kuiper et al., 2017). We therefore focused on this length moving forward. The frequency of polyQ nucleation greatly exceeds that of other Q-rich amyloid-forming proteins of comparable length (Khan et al., 2018; Posey et al., 2021). To determine if the protein’s exceptional amyloid propensity can be attributed to a property of the Q residues themselves, as opposed to its extremely low sequence complexity, we next tested a 60-residue homopolymer of asparagine (polyN), whose side chain is one methylene shorter but otherwise identical to that of glutamine. PolyN is known to form amyloid (Halfmann et al., 2011). We found that polyN did populate the high FRET state, but at much lower frequencies than polyQ even at the highest concentrations (Figure 1C; Figure 1—figure supplement 1B), approximately 200 µM (Khan et al., 2018). The apparently much larger nucleation barrier for polyN—despite its physicochemical similarity to polyQ—led us to consider whether substituting Qs for Ns at specific positions in polyQ might reveal patterns that uniquely encode the structure of the polyQ nucleus. We reasoned that a random screen of Q/N sequence space would be unlikely to yield informative patterns, however, given that there are more than one trillion combinations of Q and N for even a 40-residue peptide. We therefore devised a systematic approach to rationally sample Q/N sequence space. Specifically, we characterized by DAmFRET a series of related sequences with a single N residue inserted after every q Q residues over a total length of 60, for all values of q from 0 through 9 (designated Q1N, Q2N, etc. with the value of q subscripted; Supplementary file 1). The resulting dataset revealed a shockingly strong dependence of amyloid propensity on the exact sequence of Q residues, and specifically the following two determinants. First, we observed amyloid formation for all values of q ≥ 6 (Figure 1—figure supplement 1B). Second, amyloid formation for values of q<6 was limited to odd numbers 1, 3, and 5 (Figure 1E). That is, the relationship of amyloid formation to the number of consecutive Qs exhibited an ‘odd-even effect’ that saturated at six. This simple pattern could be explained by a local requirement for Q at every other position along the sequence (i.e. i, i+2, i+4, etc.). To test this, we inserted a second N residue into the repeats for two such sequences. Indeed, these failed to aggregate (Figure 1—figure supplement 2A). To determine if this ‘odd’ dependency results from a general preference of amyloid for homogeneity at every other position, irrespective of their amino acid identity, we created an analogous ‘polyN’ series with a single Q placed after every first, second, third, or fourth N. This series showed no preference for even or odd values (Figure 1—figure supplement 2A), confirming that Qs, specifically, are required at every other position. We next asked if the odd-even effect concerns the Q side chains themselves, rather than their interaction with Ns, by replacing the Ns with residues of diverse physicochemistry—either glycine, alanine, serine, or histidine. Again, nucleation was much more frequent for Q3X and Q5X than for Q4X (Figure 1E; Figure 1—figure supplement 2B), confirming that the odd-even effect resulted from the Q side chains interacting specifically with other Q side chains. The identity of X did, however, influence nucleation frequency particularly for Q4X. Glycine proved just as detrimental as asparagine; histidine slightly less so; and alanine lesser still. Serine was relatively permissive. The relative indifference of Q3X and Q5X to the identity of X suggests that these sequences form a pure Q core that excludes every other side chain. In other words, the nucleus is encoded by segments of sequence with Qs at every other position. To reduce the contribution of the conformational fluctuation to nucleation, we next performed DAmFRET experiments for all sequences in [PIN+] cells. Amyloid formation broadly increased (Figure 1—figure supplement 1A–B; Figure 1—figure supplement 2A–B), with polyQ now nucleating in most [PIN+] cells even at the lowest concentrations sampled by DAmFRET (~1 µM, Khan et al., 2018). Importantly, amyloid formation again occurred only for polyQ lengths exceeding the clinical threshold. Additionally, nucleation in [PIN+] retained the structural constraints of de novo nucleation, as indicated by (a) the relative impacts of different X substitutions, which again increased in the order serine

Effector mechanisms of the unfolded protein response (UPR) in the endoplasmic reticulum (ER) are well-characterised, but how ER proteostasis is sensed is less well understood. Here, we exploited the beta isoform of the UPR transducer IRE1, that is specific to mucin-producing cells in order to gauge the relative regulatory roles of activating ligands and repressing chaperones of the specialised ER of goblet cells. Replacement of the stress-sensing luminal domain of endogenous IRE1α in CHO cells (normally expressing neither mucin nor IRE1β) with the luminal domain of IRE1β deregulated basal IRE1 activity. The mucin-specific chaperone AGR2 repressed IRE1 activity in cells expressing the domain-swapped IRE1β/α chimera, but had no effect on IRE1α. Introduction of the goblet cell-specific client MUC2 reversed AGR2-mediated repression of the IRE1β/α chimera. In vitro, AGR2 actively de-stabilised the IRE1β luminal domain dimer and formed a reversible complex with the inactive monomer. These features of the IRE1β-AGR2 couple suggest that active repression of IRE1β by a specialised mucin chaperone subordinates IRE1 activity to a proteostatic challenge unique to goblet cells, a challenge that is otherwise poorly recognised by the pervasive UPR transducers.

Dysfunction of the endoplasmic reticulum (ER) in insulin-producing beta cells results in cell loss and diabetes mellitus. Here we report on five individuals from three different consanguineous families with infancy-onset diabetes mellitus and severe neurodevelopmental delay caused by a homozygous p.(Arg371Ser) mutation in FICD. The FICD gene encodes a bifunctional Fic domain-containing enzyme that regulates the ER Hsp70 chaperone, BiP, via catalysis of two antagonistic reactions: inhibitory AMPylation and stimulatory deAMPylation of BiP. Arg371 is a conserved residue in the Fic domain active site. The FICDR371S mutation partially compromises BiP AMPylation in vitro but eliminates all detectable deAMPylation activity. Overexpression of FICDR371S or knock-in of the mutation at the FICD locus of stressed CHO cells results in inappropriately elevated levels of AMPylated BiP and compromised secretion. These findings, guided by human genetics, highlight the destructive consequences of de-regulated BiP AMPylation and raise the prospect of tuning FICD's antagonistic activities towards therapeutic ends.

J-domain proteins (JDPs) are the largest family of chaperones in most organisms, but much of how they function within the network of other chaperones and protein quality control machineries is still an enigma. Here, we report on the latest findings related to JDP functions presented at a dedicated JDP workshop in Gdansk, Poland. The report does not include all (details) of what was shared and discussed at the meeting, because some of these original data have not yet been accepted for publication elsewhere or represented still preliminary observations at the time.

A causal relationship between endoplasmic reticulum (ER) stress and the development of neurodegenerative diseases remains controversial. Here, we focused on Seipinopathy, a dominant motor neuron disease, based on the finding that its causal gene product, Seipin, is a protein that spans the ER membrane twice. Gain-of-function mutations of Seipin produce non-glycosylated Seipin (ngSeipin), which was previously shown to induce ER stress and apoptosis at both cell and mouse levels albeit with no clarified mechanism. We found that aggregation-prone ngSeipin dominantly inactivated SERCA2b, the major calcium pump in the ER, and decreased the calcium concentration in the ER, leading to ER stress and apoptosis in human colorectal carcinoma-derived cells (HCT116). This inactivation required oligomerization of ngSeipin and direct interaction of the C-terminus of ngSeipin with SERCA2b, and was observed in Seipin-deficient neuroblastoma (SH-SY5Y) cells expressing ngSeipin at an endogenous protein level. Our results thus provide a new direction to the controversy noted above.

Guanine nucleotide-exchange factors (GEFs) activate the function of guanine nucleotide-binding proteins (G-proteins) by promoting the exchange of GDP for GTP on the latter. Here, we describe a protocol for in vitro measurements of the GEF activity of eukaryotic translation initiation factor 2B, eIF2B, toward its substrate eIF2. This protocol provides a relatively simple method for determining the eIF2B's GEF activity in crude cell extracts. The eIF2 heterotrimeric substrate, with phosphorylated or unphosphorylated eIF2α, is prepared by immunoprecipitation, following subsequent loading of a fluorescent BODIPY-FL dye-attached GDP. The exchange of the bound fluorescent GDP molecule for an unlabeled one on eIF2 promoted by eIF2B is monitored kinetically using a fluorescence microplate reader.

ABSTRACT Dysfunction of the endoplasmic reticulum (ER) in insulin-producing beta cells results in cell loss and diabetes mellitus. Here we report on 5 individuals from three different consanguineous families with infancy-onset diabetes mellitus and severe neurodevelopmental delay caused by a homozygous p.(Arg371Ser) mutation in FICD . The FICD gene encodes a bifunctional Fic domain-containing enzyme that regulates the ER Hsp70 chaperone, BiP, via catalysis of two antagonistic reactions: inhibitory AMPylation and stimulatory deAMPylation of BiP. Arg371 is a conserved residue in the Fic domain active site. The FICD R371S mutation partially compromises BiP AMPylation in vitro but eliminates all detectable deAMPylation activity. Overexpression of FICD R371S or knock-in of the mutation at the FICD locus of stressed CHO cells result in inappropriately elevated levels of AMPylated BiP. These findings, guided by human genetics, highlight the destructive consequences of de-regulated BiP AMPylation and raise the prospect of tuning FICD’s antagonistic activities towards therapeutic ends.