Introduction

The endoplasmic reticulum (ER), and its associated chaperones, constitutes the major cellular compartment for the synthesis, folding, and quality control of secretory proteins (Sun and Brodsky, 2019). An imbalance between synthesis and folding can lead to ER stress, potentially resulting in ER dysfunction and pathological conditions (Wang and Kaufman, 2016). To restore cellular homeostasis, adaptive pathways, collectively known as the unfolded protein response (UPR), are activated. The UPR is coordinated by three known ER stress transducers: IRE1, PERK and ATF6 (Walter and Ron, 2011). Of the three arms of the UPR, the mechanisms that govern ER stress-dependent ATF6 activation remain the least well understood, even though ATF6 mediates much of the ER-stress-induced changes in gene expression observed in vertebrates. Thus, ATF6 is specialised in the regulation of quality control proteins in the ER (Adachi et al., 2008) and loss of ATF6 impairs survival upon ER stress in cellular and animal models (Wu et al., 2007).

ATF6 is a type II transmembrane ER glycoprotein containing an N-terminal portion comprising its basic leucine zipper (b-ZIP), DNA-binding and transactivation domains and a C-terminal ER luminal domain (LD) that senses stress and regulates ATF6⍺ activity. Vertebrates have two isoforms: ATF6α and ATF6β. The ATF6α isoform dominates the regulation of UPR target genes in mammalian cells and is the focus of this study. Unlike IRE1 and PERK, which remain in the ER after UPR activation, ATF6 translocates from the ER to the Golgi, where it is sequentially cleaved by two serine protease, site-1 (S1P) and site-2 (S2P) proteases (Haze et al., 1999) to release its soluble active N-cytosolic domain (N-ATF6⍺) that is further translocated into the nucleus. There, N-ATF6α acts as a potent but short-live transcription factor (George et al., 2020) that binds ER stress response elements (ERSE-I and -II) in the promoter regions of UPR target genes in complex with the general transcription factor NF-Y and YY1 to upregulate the expression of a large number of ER chaperones (Li et al., 2000; Yoshida et al., 2000). Among those, the gene encoding the ER Hsp70 chaperone BiP (GRP78) has been well documented as an important ATF6α target (Adachi et al., 2008).

At ER level, it has been proposed that ATF6⍺ is retained by its association with BiP (Shen et al., 2002), yet the existence of additional factors influencing its ER localisation remains uncertain. Redox regulation has also been proposed to keep ATF6⍺ in the ER as a mixture of monomeric and multimeric disulphide-linked forms (Nadanaka et al., 2007) that upon ER stress are reduced by protein disulphide isomerase (PDI) family members (such as ERp18 or PDIA5), promoting the transit of a reduced monomeric ATF6⍺ population (Higa et al., 2014; Oka et al., 2019). COPII-coated vesicles have been implicated in the trafficking of ATF6⍺ (Schindler and Schekman, 2009; Lynch et al., 2012) and the ATF6⍺_LD’s N-glycosylation state has also been proposed to influence its ER trafficking (Hong et al., 2004), yet the details of these processed remain incompletely understood. Furthermore, it is widely recognised that a key step of ATF6⍺ activation occurs in the Golgi, the site of proteolytic processing. This is reflected in the inability of cells lacking S2P to trigger BiP transcription in response to ER stress (Ye et al., 2000). Hence, although there is a relatively clear understanding of the events triggering ATF6⍺ cleavage and activation upon reaching the Golgi, much remains unknown about the factors influencing ATF6⍺ regulation in the ER.

ATF6⍺ signalling also involves crosstalk with the IRE1 pathway. This is partly mediated by direct heterodimerisation of N-ATF6⍺ with XBP1s the active, spliced form (Yamamoto et al., 2007, 2004). However, the relationship between the ATF6⍺ and IRE1 pathway is complex: (i) XBP1 and IRE1 genes are transcriptional targets of activated ATF6⍺ (Yoshida et al., 2001), (ii) XBP1s and IRE1 have some different target genes (Shoulders et al., 2013), (iii) and ATF6⍺ activity can also repress IRE1 signalling (Walter et al., 2018).

Here, we established an ATF6⍺/IRE1 dual UPR reporter cellular model to conduct an unbiased genome-wide CRISPR/Cas9 screen focussed on ATF6⍺. Complemented by bioinformatic analysis, post-screen hit validation using targeted gene editing and in vitro biophysical assays we report on a comprehended analysis of a complex signalling pathway.

Results

A dual UPR reporter cell line to monitor ATF6⍺ and IRE1 activity

In vertebrates, the transcriptional activation of BiP is predominantly governed by ATF6⍺ (Adachi et al., 2008; Adamson et al., 2016). To dynamically monitor ATF6⍺ activity in live cells, a “landing pad” cassette, containing recombination sites and expressing the Cricetulus griseus (cg) BiP promoter region fused with a super folder green fluorescent protein (BiP::sfGFP), was targeted to the ROSA26 safe-harbour locus of Chinese hamster ovary cells (CHO-K1) using CRISPR/Cas9 (Gaidukov et al., 2018) (Supplemental S1A). These cells also carried an integrated XBP1s::mCherry reporter transgene monitoring the IRE1 pathway (Harding et al., 2019). Flow cytometry analysis of these ATF6⍺/IRE1 dual UPR reporter cells (referred as XC45-6S) treated with ER stress-inducing agents such as tunicamycin (which inhibits N-glycosylation), 2-Deoxy-D-glucose (which inhibits glycolysis), or thapsigargin (which depletes ER calcium), revealed time-dependent activation of both UPR arms with a broad dynamic range (Fig 1A, top panel, and Supplemental S1B).

Fig 1.

Disruption of endogenous cgATF6⍺ in XC45-6S cells by CRISPR/Cas9 genome editing abolished the responsiveness of the BiP::sfGFP reporter to ER stress and derepressed IRE1 signalling, both basally and in response to ER stress. These observations are consistent with ATF6⍺’s obligatory role in BiP expression and with findings that IRE1 signalling is repressed by ATF6⍺ activity (Walter et al., 2018).

To further validate these UPR reporters, cells were subjected to tunicamycin treatment in the presence of Ceapin-A7 (a small molecule that blocks ATF6⍺ activation by tethering it to the lysosome) (Gallagher et al., 2016) or an S1P inhibitor (PF429242) (Hawkins et al., 2008). Ceapin-A7 effectively blocked the activation of the ATF6⍺ fluorescent reporter, whereas the S1P inhibitor partially attenuated the BiP::sfGFP signal in stressed cells (Supplemental S1C). The selective IRE1 inhibitor 4μ8C (Cross et al., 2012), did not affect the ATF6⍺-dependent BiP::sfGFP reporter under ER stress conditions, but blocked the IRE1-dependent XBP1s::mCherry reporter (Supplemental S1D). Collectively, these observations confirm that the dual UPR reporter XC45-6S cell line enables independent monitoring of ATF6⍺ and IRE1 activities, making it well-suited for identifying additional modulators of ATF6⍺ signalling by high-throughput techniques.

Genome-wide CRISPR/Cas9 knock-out screens to profile ATF6⍺ signalling

To identify regulators of ATF6⍺, a library of pooled lentiviral single guide RNAs (sgRNAs) targeting 20,680 predicted protein-coding genes in CHO cells, with ∼ 6 sgRNAs per gene, was used (Ordonez et al., 2020). Transduction of the XC45-6S ATF6⍺/IRE1 dual UPR reporter cells was performed at a coverage of ∼ 640X, with a multiplicity of infection of 0.3, intentionally set low to disfavour acquisition of more than one sgRNA by any one cell (Shalem et al., 2014). Non-transduced cells were purged by puromycin selection, and the pool of transduced cells was expanded for 10 days to favour gene editing and the establishment of loss-of-function phenotypes.

To systematically identify genes regulating the ATF6⍺ transcriptional programme, two positive selection screens were conducted in parallel: a “loss-of-ATF6⍺ function screen” under ER stress conditions to identify activators/enhancers of ATF6⍺, and a “gain-of-ATF6⍺ function screen” under basal conditions to identify ATF6⍺ repressors (Fig 1B). In the ATF6⍺ activator screen, cells were stressed by exposure to 2-Deoxy-D-glucose for 18 h before fluorescent-activated cell sorting (FACS). Cells at the lowest 2% of BiP::sfGFP signal, with preserved IRE1 signalling (BiP::sfGFP^low; XBP1s::mCherry^high), were selected for further analysis. The search for genes whose inactivation de-represses ATF6⍺ (ATF6⍺ repressor screen) was conducted in the absence of ER stress. Here, cells exhibiting a constitutively active ATF6⍺ reporter but only basal levels of the IRE1 reporter (BiP::sfGFP^high; XBP1s::mCherry^low) were selected for further analysis. Sorted cells from both screens underwent two rounds of progressive enrichment through cell culture expansion and phenotypic selection by sorting. As a result, a progressive increase in the proportion of BiP:sfGFP^low or BiP:sfGFP^high cells occurred in the respective screen (Supplemental S2A and S2B).

The integrated sgRNAs were amplified from genomic DNA, and their abundance in the enriched populations after one or two rounds of sorting was estimated by deep sequencing and compared to a reference sample of transduced and unsorted population. MAGeCK bioinformatics analysis (Li et al., 2014) was used to determine sgRNA sequence enrichment, establishing the corresponding ranking of targeted genes (Table S1; see GEO accession number: GSE254745) and quality controls (QC) measures (Supplemental S2C and S2D). Both screens were conducted at least in duplicate for all conditions.

Expected and new candidate genes that contribute to ATF6⍺ activation

MAGeCK-based analysis of sgRNAs enriched in cells with attenuated ATF6⍺ signalling and intact activity of the IRE1 reporter, revealed that those targeting S1P and S2P proteases (encoding by the MBTPS1 and MBTPS2 genes) were the two most enriched genes (Fig 1C), validating the experimental methodology. Gene Ontology (GO) cluster analysis of the 100 most significantly enriched targeted genes revealed up to 19 enriched pathways (Supplemental S3A), with the top two pathways highlighted in Figure 1C being closely related to ATF6⍺ signalling.

We focused our attention on targeted genes that could act upstream to find early regulatory events of ATF6⍺. Therefore, targeted genes that were likely to act downstream and impact on the cleaved ATF6⍺ form (N-ATF6⍺) were dismissed, including the SP2 transcription factor, which shares the binding site with NF-Y (Völkel et al., 2015), and also genes involved in nuclear import, such as NUP50 and SEH1L (Fig 1C and Supplemental S3A). Similar consideration was applied to PRMT8, as the family member PRMT1 has been shown to enhance ATF6⍺ transcriptional activity (Baumeister et al., 2005). Additionally, guides targeting IGFR1 and CXXC1 were also dismissed as likely affecting ER proteostasis indirectly via cellular growth (Pfaffenbach et al., 2012). Guides targeting the genes encoding the ER resident proteins SEC61A1, SEC61B (both components of the translocon machinery) and the DNAJB9/ERDj4 co-chaperone were highly enriched. However, given their roles in selectively regulating IRE1 signalling (Adamson et al., 2016; Shoulders et al., 2013), we deemed their identification to reflect a bias towards IRE1 activators arising from the selection scheme for ATF6⍺^low and IRE1^high cells.

One of the genes that could plausibly play a proximal role and act upstream on ATF6⍺ activation was the ubiquitous Golgi-localised protease FURIN (Figure 1C). Like S1P, FURIN belongs to the subtilisin-like proprotein convertase family (Nakayama, 1997)and is thus poised to act at the level of ATF6⍺ proteolytic processing. All six sgRNAs targeting FURIN were enriched in the BiP:sfGFP^low cells, a level of enrichment similar to that observed to the established ATF6⍺ regulators MBTPS1 (S1P) and MBTPS2 (S2P) (Supplemental S3B). The genotype-phenotype relationship suggested by the screen was also observed upon targeting FURIN in ATF6⍺/IRE1 dual reporter cells, as both pools of targeted cells and an individual clone had diminished responsiveness of BiP:sfGFP to ER stress (Supplemental S4A left and middle panels). Furthermore, the S1P inhibitor more effectively attenuated the BiP::sfGFP reporter in FURINΔ cells compared to wild-type cells (Supplemental S4A right panel, S4B and S4C).

These observations raised the possibility of redundant roles of FURIN and S1P in activating ATF6⍺ by cleaving its luminal domain. However, ATF6⍺_LD expressed and purified from mammalian cells, failed to serve as a substrate for FURIN in vitro (Supplemental S4D), suggesting instead that FURIN’s role in regulating ATF6⍺ signalling is indirect.

Calreticulin and an interconnected gene network repress ATF6⍺

To identify genes that basally repress ATF6⍺, a similar MAGeCK-based data analysis was applied to the targeted genes enriched in cells that selectively activated the ATF6⍺ reporter under basal conditions. GO analysis of the 100 most significantly enriched genes established an interconnected network highlighting signalling processes including the calreticulin (CRT)/calnexin (CNX) cycle, N-glycan biosynthesis, ERAD and COPI-dependent Golgi-to-ER retrograde traffic (Fig 1D and Supplemental S3C). Among these, the top 10 genes included CRT, components of the dolichol-phosphate mannose (DPM) synthase complex, such as DPM3, DPM2 and DPM1 involved in the N-glycosylation process, and ERAD components, such as UBE2G2, SEL1L (Fig 1D).

Because these four pathways potentially activate all three branches of the UPR, genes previously identified in genome-wide screens to modulate IRE1 or PERK functionality were excluded from further analysis (e.g. TMED10, SEL1L, FAB2, SYVN1, PDI6) (Fig 1H) (Adamson et al., 2016; Panganiban et al., 2019; Yang et al., 2018). Of the remaining genes, we focused on those encoding proteins localised in the ER lumen or Golgi, which could act as specific proximal regulators of ATF6⍺. Among these, CRT, an abundant luminal ER lectin chaperone, caught our attention (Fig 1H). Notably, although CRT promotes the folding of ER-synthesised glycoproteins via the CRT/CNX cycle (Lamriben et al., 2016), its ER transmembrane homolog CNX was not identified in the screen. This observation was further supported by the fact that sgRNAs targeting CNX were not enriched, whereas five out of six sgRNAs targeting CRT showed significant enrichment during the selection process (Fig 2A).

Fig 2.

Constitutive activation of ATF6⍺ in cells lacking calreticulin

To explore CRT’s role in ATF6⍺ regulation, CRT was targeted by CRISPR/Cas9 in wild-type XC45-6S dual UPR reporter cells. CRT-targeted cells exhibited large populations of BiP::sfGFP^high; XBP1s::mCherry^low cells, mirroring the results of the initial screen. This phenotype was stable in single clones from the targeted pool (Fig 2B and 2C). Targeting CRT repressed the IRE1 reporter basally and left little room for further induction of the ATF6⍺ reporter upon ER stress (Fig 2B and 2C). De-repression of the ATF6⍺ reporter in CRTΔ clones correlated with a slight increase in BiP protein levels (Fig 2D). By contrast, depletion of CNX had no impact on the BiP::sfGFP reporter (Supplemental S5A-S5C), consistent with the findings of the screen.

To test whether de-repression of the BiP::sfGFP reporter in CRTΔ cells was ATF6⍺ dependent, endogenous ATF6⍺ was also inactivated in CRTΔ cells. ATF6⍺ inactivation restored BiP::sfGFP reporter to its low baseline levels and derepressed the IRE1 reporter to levels observed in XC45-6S parental cells (Fig 2E). Overall, these findings confirmed that BiP::sfGFP reporter activity correlates with the transcriptional activity of ATF6⍺ in CRTΔ cells.

Calreticulin represses Golgi-trafficking and processing of ATF6⍺

Considering the crucial role of trafficking in ATF6⍺ activation, we next investigated whether depletion of CRT affected the stress-induced localisation of ATF6⍺ by comparing the subcellular localisation of ATF6⍺ in parental and CRTΔ cells. As trafficking is closely linked to proteolytic processing, wild-type ATF6⍺ does not accumulate in the Golgi of stressed cells (Shen et al., 2002). To stabilise ATF6⍺ molecules reaching the Golgi, a CHO-K1 cell line stably expressing a GFP-tagged version of cgATF6⍺_LD lacking S1P and S2P cleavage sites (GFP-ATF6⍺_LD^S1P,S2Pmut) was engineered (Fig 3A). In non-stressed cells, GFP-ATF6⍺_LD^S1P,S2Pmut was distributed in a prominent reticular ER pattern with only a minor fraction co-localising with the Scarlet-tagged Giantin Golgi localisation marker. Exposure to the rapidly acting ER stress-inducing agent Dithiothreitol (DTT), increased the Golgi pool at the expense of the ER pool (Fig 3A); an expected outcome validating GFP-ATF6⍺_LD^S1P,S2Pmut as a sentinel for the ATF6⍺ protein. Compared to parental cells, GFP-ATF6⍺_LD^S1P,S2Pmut showed significantly more Golgi co-localisation in CRTΔ cells under basal conditions (Fig 3A). These results were in line with our previous flow cytometry observations and suggested that CRT contributes to ER retention of ATF6⍺, either directly or indirectly.

Fig 3.

GFP-ATF6⍺_LD^S1P,S2Pmut reporter cannot be proteolytically processed. Consequently, to gauge the effect of CRT on ATF6⍺ processing we sought to create a CHO-K1 stable cell line to track endogenous ATF6⍺. For that, the cgATF6⍺ locus was tagged with a 3xFLAG and a monomeric GreenLantern (mGL) fluorescent proteinClick or tap here to enter text., generating a cell line referred to as 2K: 3xFLAG-mGL-ATF6⍺ (Fig 3B). The functionality of this probe was confirmed by treatment with DTT that resulted in the appearance of a prominent faster-migration band in SDS-PAGE, compatible with the processed S2P-cleaved ATF6⍺ (N-ATF6⍺) form, and depletion of the precursor form (ATF6⍺-P) (Fig 3B.1, lane 8). Subsequently, 2K cells were targeted to generate 2K-CRTΔ derivatives cells, enabling a comparison of ATF6⍺ processing. 2K_CRTΔ clonal cells exhibited elevated levels of the processed N-ATF6⍺ form under both basal and stress-induced conditions (Figure 3B.1 and 3B.2) that correlated with a higher nuclear ATF6⍺-mGL signal observed by live cell microscopy (Fig 3C). These findings indicated that CRT represses both ER-to-Golgi trafficking and processing of ATF6⍺ into its active form.

Calreticulin interacts with the ATF6⍺ luminal domain in vitro

To investigate the underlying mechanism of CRT’s repressive effect on ATF6⍺, we explored the physical interaction between the two purified proteins in vitro using BioLayer Interferometry (BLI). The biotinylated ATF6⍺_LD, expressed in mammalian cells, was immobilised on the BLI probe and used as the ligand, and bacterially expressed CRT was used as the analyte. Considering possible regulatory roles of N-glycosylation and disulphide bonds in ATF6⍺ activation (Hong et al., 2004; Oka et al., 2019), three different versions of ATF6⍺_LD were used: 1) fully glycosylated ATF6⍺_LD (WT), 2) non-glycosylated ATF6⍺_LD (ΔGly) and 3) ATF6⍺_LD lacking its cysteines (ΔC) (Fig 4A). BLI showed reversible binding of CRT to ATF6⍺_LD^WT that was characterised by slow association (kon) and dissociation (koff) rates (Fig 4B). Both the association and dissociation phases were fitted to a single exponential association-dissociation model, yielding a KD within the range of 1 µM, a concentration that was consisted with the estimated CRT concentration in the ER of CHO-K1 cells obtained through quantitative immunoblotting (∼7 µM) (Supplemental Fig S5D). Similar interactions were also observed with ATF6⍺_LD^ΔGly and ATF6⍺_LD^ΔC.

Fig 4.

To examine a possible interaction between CRT and ATF6⍺ in cells, HEK293T cells were transfected with either GFP-ATF6⍺_LD^WT or GFP-ATF6⍺_LD^ΔGly, followed by selective recovery using GFP-Trap Agarose and subsequent immunoblotting with an anti-CRT antibody. The co-immunoprecipitation results showed that both ATF6⍺_LD^WT and, to a lesser extent, ATF6⍺_LD^ΔGly immunoprecipitated endogenous CRT (Fig 4C). Re-probing the blot with an anti-GFP antibody revealed that GFP-ATF6_LD^ΔGly was expressed at lower levels than ATF6⍺_LD^WT (Fig 4C), potentially accounting for the reduced recovery of CRT in complex with ATF6⍺_LD^ΔGly. Overall, these observations suggested that CRT binds to ATF6⍺_LD both in cells and in isolation in vitro, supporting the notion of a direct interaction between the two proteins.

A genetic platform to study endogenous ATF6⍺ signalling

To circumvent the potentially corrupting effect of protein overexpression, we sought to express ATF6⍺ variants from the endogenous locus, by replacing the ATF6⍺_LD coding sequence with mutants of our design. To enable CRISPR-Cas9 mediated homology-directed repair (HDR) that replaces the wild-type sequence with the mutants, we first created a non-functional ATF6⍺ allele that could be subsequently restored to function by homologous recombination, offering wild-type or mutant repair templates (Fig 5A). As expected, the deletion of ATF6⍺ resulted in the loss of the BiP::sfGFP signal in stressed cells (and enhanced XBP1s::mCherry signal in basal conditions, Fig 5A). This ATF6⍺Δ clone was re-targeted with a unique sgRNA/Cas9, alongside ATF6⍺_LD^WT, ATF6⍺_LD^ΔGly or ATF6⍺_LD^ΔC repair templates (Fig 5B).

Fig 5.

Unlike the ATF6⍺_LD^WT, neither the ATF6⍺_LD^ΔGly nor the ATF6⍺_LD^ΔC repair templates fully restored BiP::sfGFP responsiveness to stress (Fig 5B top panel), consistent with the importance of N-glycosylation and disulphide bond formation to ATF6⍺ functionality. Notably, whereas ATF6⍺_LD^ΔC knock-in cells exhibited basal activation of the IRE1 reporter comparable to levels observed in the ATF6⍺Δ parent (Fig 5B lower panel), the IRE1 reporter signal in the ATF6⍺_LD^ΔGly knock-in cells shifted to that observed in wild-type cells, suggesting that ΔGly allele retained some functionality. The partial ATF6⍺ functionality of the ATF6⍺_LD^ΔGly knock-in allele aligned with the lower expression levels of ATF6⍺_LD^ΔGly compared to ATF6⍺LD^WT in transfected cells, as previously observed in Figure 4C. However, this feature complicated interpretation of experiments designed to establishing the role of ATF6⍺_LD glycans on CRT-mediated repression of ATF6⍺: failure of CRT deletion to derepress the BiP::sfGFP signal in ATF6⍺_LD^ΔGly knock-in cells (Supplemental Fig S6) could be interpreted as either an obligatory role for glycans in the repression or as lack of dynamic range of the signal needed to detect a residual effect of CRT deletion in ATF6⍺_LD^ΔGly knock- in cells.

Impact of calreticulin variants on ATF6⍺ reporter activity

CRT is a lectin chaperone that interacts with the protein disulphide isomerase ERp57 to stabilise CRT-substrate binding (Oliver et al., 1999). Therefore, we tested the ability of human CRT^WT and two mutants that compromised glycan binding (Y92A) or Erp57 co-chaperone binding (W244A) (Del Cid et al., 2010) to reverse the CRTΔ phenotype. Parental and CRTΔ clones were transfected with CRT^WT, CRT^Y92A and the double mutant CRT^Y92A,W244A- expressing plasmids, and cells were evaluated by flow cytometry 48 and 72-h post-transfection (Fig 6A). Expression of CRT^WT restored basal levels of both BiP::sfGFP and XBP1s:mCherry reporters in CRTΔ cells (Fig 6A purple panels, and Fig 6B), confirming the repressive role of CRT on ATF6⍺. By contrast, expression of the double CRT^Y92A,W244A mutant failed to rescue the CRTΔ phenotype (Fig 6A green panels, and Fig 6B), whereas CRT^Y92A, selectively incapable of glycan binding, retained some ability to repress BiP::sfGFP and modestly rescuing the CRTΔ phenotype (Fig 6A orange panels, and Fig 6B). Immunoblots targeting CRT raised concerns about the integrity of theCRT^Y92A,W244A protein, precluding a separate analysis of Erp57’s role in ATF6⍺ repression (Fig 6C). However, CRT^WT and CRT^Y92A proteins were expressed similarly (Fig 6C). Together, our observations pointed to an important role for interactions between CRT and ATF6⍺_LD glycans but also left room for some lectin-independent repression of ATF6⍺ by CRT.

Fig 6.

Calreticulin depletion exposes a negative feedback loop by which ATF6⍺ represses IRE1

Given the well-established interplay between the ATF6⍺ and IRE1 branches of the UPR, we sought to assess whether the constitutive activation of ATF6⍺ observed in CRTΔ clones could be a contributing factor to the conspicuous downregulation of IRE1 activity. Through RT-PCR and gel electrophoresis, IRE1-mediated XBP1 mRNA splicing in parental cells and CRTΔ clones was compared. Under basal conditions, unspliced mRNA levels (XBP1u) were significantly higher in CRTΔ clones (Fig 7A left bar graph), consistent with the role of N-ATF6⍺ in the transcriptional activation of XBP1 (Yoshida et al., 2001). In addition, compared to parental cells, CRTΔ cells exhibited a smaller fraction of spliced XBP1 mRNA (XBP1s) (Fig 7A right bar graph), consistent with the lower IRE1 activity measured by a XBP1s::mCherry reporter.

Fig 7.

IRE1 activity is closely tied to its phosphorylation state (Shamu and Walter, 1996), therefore IRE1⍺-phosphorylation (IRE1⍺-P) levels were compared in parental and CRTΔ clones. Accordingly, CRTΔ cells exhibited lower levels of stress-induced phosphorylation of their endogenous IRE1 protein and slightly reduced total IRE1 levels (Fig 7B). These results supported earlier findings that IRE1 signalling is repressed by ATF6⍺ activity (Walter et al., 2018). However, the defect in IRE1 activity wrought by CRT depletion was partial, as in the presence of ER stress, all genotypes exhibited a comparable increase in XBP1s (Fig 7A).

Discussion

This study reports on a comprehensive and unbiased genome-wide CRISPR/Cas9-based knockout screen dedicated to discovering regulators of ATF6⍺. Whereas previous high-throughput studies have broadly studied modulators of the UPR (Adamson et al., 2016; Jonikas et al., 2009; Panganiban et al., 2019) we designed an ATF6⍺/IRE1 dual UPR reporter cell line to selectively detect regulators of ATF6⍺. Although a connection between CRT and ATF6⍺ was proposed previously based on an interaction between overexpressed proteins (Hong et al., 2004), our study utilises unbiased genetic tools and manipulation of endogenous proteins to establish CRT’s role as a selective repressor of endogenous ATF6⍺ signalling, mapping its function in retaining ATF6⍺ in the ER.

It is notable that, though the screen identified well-known regulators of ATF6⍺ processing (including S1P, S2P and NF-Y, and thus validating the experimental approach), only a limited number of new genes that contribute to ATF6⍺ signalling emerged from the screen. Despite the biochemical evidence for the role of COPII-coated vesicles and the oligomeric state of ATF6⍺ in its activation, no cluster of genes involved in COPII vesicular transport or PDIs were found among the most enriched pathways. Nor were any counterparts to SCAP or INSIG, regulators of ER-to-Golgi trafficking of the related SREBP proteins (Yang et al., 2002), identified. This might reflect essentiality of the genes involved or redundancy amongst them. Cell type specificity may also be implicated in failure to identify components that are unimportant to ATF6⍺ activation in CHO-K1 cells.

The screen implicated two components of the nuclear pore complex (NUP50 and SEH1L) in ATF6⍺ pathway activity. While it might be tempting to consider a role for these components in the nuclear translocation of the processed soluble ATF6⍺ N-terminal domain, there is no reason to think it would be regulatory, as it operates downstream of ER stress-regulated trafficking and processing events.

In addition to the identified S1P and S2P proteases, the screen for ATF6⍺ activators identified FURIN, a Golgi-localised protease belonging to the subtilisin-like proprotein convertase family (like S1P protease) responsible for proteolytic activation of a wide array of precursor proteins within the secretory pathway (Braun and Sauter, 2019). FURIN deletion synergised with S1P inhibitors in attenuating ATF6⍺ signalling. Additionally, ATF6⍺_LD has a conserved region (RTKSRR) related to the FURIN cleavage motif (generally described as R-X-R-X-[K/R]-R↓ (Nakayama, 1997) (Supplemental Fig S4E). It has been suggested that S1P’s role in activation is to reduce the size of ATF6⍺’s_LD, thereby promoting S2P cleavage (Shen et al., 2002; Ye et al., 2000). FURIN cleavage could contribute to such a process and account for partial redundancy between S1P inhibition and FURIN depletion observed here. Nonetheless, in vitro FURIN failed to cleave the ATF6⍺_LD. These findings collectively suggest that FURIN may function as an indirect enhancer of ATF6⍺ signalling, perhaps by setting up conditions in a post-ER compartment that favours ATF6⍺ cleavage and activation.

The screen for ATF6⍺ repressors revealed a substantial enrichment of genes involved in glycoprotein metabolism. Among that cluster, the most notable hit, not previously identified in other high-throughput UPR screens, was CRT, a well-characterised soluble ER chaperone with a key role in protein quality control. Intriguingly, its ER-membrane counterpart, CNX, did not emerge in our screen, despite the inclusion of six sgRNA targeting CNX in the CRISPR library. This suggests a special role of CRT in the regulation of ATF6⍺, at least in CHO-K1 cells. CRT may not be the only repressor of ATF6⍺; this role may be shared by BiP, which has been widely suggested to suppress all the three branches of the UPR under basal conditions. Nevertheless, because of BiP’s pervasive effects on ER proteostasis and its role in direct repression of the IRE1 branch, its role in regulating ATF6 might have been missed by this screen.

Disruption of CRT in the XC45-6S dual ATF6⍺/IRE1 UPR reporter cell line selectively activated the ATF6⍺ pathway reporter under basal conditions. This activation was accompanied by an increased baseline trafficking of ATF6⍺ from the ER to the Golgi, enhanced processing of ATF6⍺ to its active form and increased endogenous BiP protein levels. Loss of CRT also resulted in a conspicuous basal downregulation of the XBP1s::mCherry reporter that was associated with a reduced activity of IRE1 in CRTΔ cells. These findings align with previous evidence that N-ATF6⍺ suppresses IRE1 signalling (Walter et al., 2018), and highlights the importance of ATF6⍺/IRE1 crosstalk.

CRT engages monoglycosylated glycoproteins in the ER via its glycan-binding lectin domain, retaining them in the CRT/CNX cycle (Hebert et al., 1996; Ou et al., 1993). A role for CRT’s interaction with ATF6⍺_LD glycans in repressing ATF6⍺ is suggested by the markedly diminished ability of a lectin mutant CRT^Y92A to repress the ATF6⍺ pathway compared to the wild-type. However, it is worth noting that CRT has been observed to bind proteins also in a glycan-independent manner (Wijeyesakere et al., 2013). Both BLI data and immunoprecipitation from cell extracts revealed that CRT can associate with both wild-type and unglycosylated ATF6⍺ (ΔGly). In the former, direct interaction is characterised by slow kinetics, consistent with previous measurements of glycosylation-independent binding of CRT to its clients (Wijeyesakere et al., 2013, Table 1). These biophysical features, coupled with the residual ability of the lectin mutant CRT^Y92A to repress endogenous ATF6⍺, lead us to suggest that CRT repression of ATF6⍺ may involve both glycosylation-dependent and independent interactions.

Currently, the structural basis for ATF6⍺_LD interactions with CRT remains unknown, and its role in the direct or indirect retention of ATF6⍺ in the ER is a subject for speculation. Nonetheless, the observations presented here support previous studies indicating that the stable interaction of CRT with cellular substrates combines both lectin-dependent and lectin-independent interactions into a hybrid mode. Moreover, our observations raise the possibility that the lectin activity of CRT is dispensable for its interaction with ATF6⍺_LD but implicated in recruitment of a third component required for repression. Either way titration of CRT by competing ligands is a plausible regulatory mechanism coupling ATF6⍺ activity to the burden of unfolded (glycol)proteins in the ER (Fig 8).

Fig 8.

Material and Methods

Plasmids construction and reagents

Standard cloning techniques were used to create the recombinant DNA vectors listed in the Table S2. Single guide RNAs (sgRNAs) and oligonucleotides used in this study are listed in Table S3. Antibodies, reagents, and software are listed in Table S4.

Mammalian cell culture, transfections, and treatments

Adherent Chinese Hamster Ovary (CHO-K1) cells (ATCC CCL-61) were maintained in Ham’s nutrient mixture F12 (Sigma) supplemented with 10% (v/v) serum (FetalClone-2, HyClone), 2 mM L-glutamine (Sigma), and 1% penicillin/streptomycin (Sigma). HEK293T cells (ATCC CRL-3216) were cultured in DMEM cell media (Sigma) supplemented as above. All cells were grown in a humidified 37°C incubator with 5% CO₂ and passed every 2-3 days. When indicated, cells were treated with 5 μM Ceapin A7 (SML2330; Sigma), 15 μM S1P inhibitor (PF429242 dihydrochloride; Tocris), 2.5 μg/ml Tunicamycin (Tm, Melford), 4 mM 2-Deoxy-D-glucose (2DG; ACROS Organics), 0.5 μM Thapsigargin (Tg; Calbiochem) and 16 μM 4μ8C (Sigma) for the indicated times. All compounds were diluted in pre-warmed culture medium and immediately added to the cells by medium exchange. Untreated cells were treated with DMSO solvent vehicle control. Transfections in CHO-K1 cells were performed using Lipofectamine LTX (Thermo Fisher Scientific, USA) at 1:3 DNA (µg) to LTX (µl) ratio, as for HEK293T using TransIT-293 Transfection Reagent (MIR2704, Mirus), according to the manufacturer’s instructions. Typically, cells were seeded at a density of 2.5 x 10⁵ cells/well in 6-well plates a day before transfection and analysed 48 - 72 h after transfection.

Mammalian cell lysates and immunoblotting

CHO-K1 or HEK293T cells were cultured in 6-, 12-well plates or 10-cm dishes until reaching 95% confluence. Cells were washed twice with prechilled PBS on ice, and whole-cell extracts were scrapped out in 1mM EDTA-PBS, pelleted at 370xg for 10 min at 4°C and incubated in Nonidet-lysis buffer (150 mM NaCl, 50 mM Tris-HCl pH 7.5, 1% (v/v) NP-40) supplemented with 2x protease inhibitor cocktail (Roche Applied Science) for 30 min. Next, the samples were clarified at 21,130xg for 20 min at 4°C and the supernatants were transferred to fresh tubes. Protein concentration was determined using Bio-Rad protein assay. To assess the interaction between CRT and ATF6α through either Bio-Layer Interferometry (BLI) or co-immunoprecipitation, the lysis buffer was supplemented with 5mM CaCl2. The protein samples were separated on 8-12.5% SDS-PAGE under reducing conditions and transferred onto PVDF membranes as described previously (Ordóñez et al., 2013). Membranes were probed with the following primary antibodies: rabbit polyclonal anti-calreticulin (1:1000) (ab19261, abcam; to detect human CRT); chicken polyclonal anti-calreticulin (1:1000) (PA1-902A, Invitrogen; to detect Chinese hamster CRT); rabbit polyclonal anti-calnexin (1:1000) (SPA-860, Enzo Life Sciences); rabbit polyclonal anti-FURIN (1:1000) (ab3467, abcam); mouse monoclonal anti IRE1_LD serum (NY200) (1:1000) (Bertolotti et al., 2000); rabbit monoclonal anti p-IRE1 (1:1000) (Genetech); rabbit polyclonal anti GFP (NY1066) (1:1000) (Marciniak et al., 2004) made against purified bacterially expressed GFP following removal of the GST affinity tag via thrombin cleavage site; chicken anti-BiP IgY (1:1000) (Avezov et al., 2013); mouse monoclonal anti-FLAG M2 (1:1000) (F1804, Sigma); mouse monoclonal anti-actin (1:1000) (clone C4-691002, Fisher Scientific); rabbit monoclonal anti-GAPDH (1:2000) (14C10; Cell Signalling Technology); rabbit polyclonal anti-α/β-Tubulin (1:1000) (CST2148S, Cell Signalling Technology). For secondary antibody, IRDye fluorescently labelled antibodies or horseradish peroxidase (HRP) antibodies were used. Membranes were scanned using an Odyssey near infrared imager (LI-COR) and signals were quantified using ImageJ software. Signal quantification and analysis were performed using Prism 10 (GraphPad).

Flow Cytometry and fluorescence-activated cell sorting (FACS)

To analyse UPR reporter activities, cells were grown on 6- or 12-well plates until reaching 80-90 % confluence and then treated as indicated. For flow cytometry analysis, cells were washed twice in PBS, collected in PBS supplemented with 4mM EDTA, and 20,000 cells/sample were analysed by multi-channel flow cytometry on a LSRFortessa cell analyser (BD Biosciences). For FACS, cells were collected in PBS containing 4mM EDTA and 0.5% BSA and then sorted on a Beckman Coulter MoFlo cell sorter. Sorted cells were either collected in fresh media as a bulk of cells or individually sorted into 96-well plates and then expanded. Gating for live cells was based on FSC-A/SSC-A and for singlets was based on FSC-W/SSC-A. BiP::sfGFP and mGreenLantern (mGL) fluorescent signals were detected with an excitation laser at 488 nm and a 530/30 nm emission filter; XBP1s::mCherry fluorescence with an excitation laser 561 nm and a 610/20 nm emission filter, while mTurquoise and BFP fluorescence with an excitation laser 405 nm and a 450/50 nm filter emission. Data analysis was performed using FlowJo V10, and median reporter analysis was conducted using Prism 10 (GraphPad).

Lentiviral production

Lentiviral particles were produced by co-transfecting HEK293T cells with the library plasmid (UK2561), the packaging plasmids psPAX2 (UK1701) and pMD2.G (UK1700) at 10:7.5:5 ratio using TransIT-293 Transfection Reagent (MIR2704, Mirus) according to the manufacturer’s instructions. Eighteen hours after transfection, medium was changed to medium supplemented with 1% BSA (Sigma). The supernatant containing the viral particles was collected 48 h after transfection, filtering through a 0.45 μm filter, and directly used to infect CHO-K1 cells seeded in 6-well plates for viral titration to calculate the amount of virus to use aiming a low multiplicity of infection (MOI) around 0.3 to ensure a single integration even per cell.

Generation of the XC45-6S double UPR reporter cell by genome editing in ROSA26 locus of CHO-K1 cells

The putative ROSA26 locus in CHO cells has been recently identified and described as a “safe-harbour” site for heterologous gene expression and stable long-term expression of a transgene (Gaidukov et al., 2018). We used the previously described XC45 CHO-K1 cell line bearing a XBP1s::mCherry reporter (Harding et al., 2019), for targeting the ROSA26 locus for knock-in with a landing pad cassette, comprised the cgBiP promoter region containing ERSE-I elements, fused to the superfolded (sf) GFP into the CHO genome by CRISPR/Cas9 integration to make this reporter cell line also sensitive to ATF6⍺ activity. The landing pad donor vector was constructed by appending homology arm sequences of ROSA26 locus (440 bp predefined sequence from Gaidukov’s paper) to the 5’ promoter region of cgBiP (1,000 bp), along with the first 8 amino acids of the BiP CDS fused to the sfGFP. Homology arms were PCR amplified from CHO genomic DNA using predefined primers from Gaidukov’s paper. Subsequently, the cgBiP promoter, a geneblock containing sfGFP, and the flanking homology arms were fused by Gibson assembly cloning method, which allows multiple DNA fragments joining in a single, isothermal reaction (1 h at 50°C). Targeted integration was performed by co-transfection with the circular landing pad donor vector (UK2846) and a single sgRNA/Cas9 targeting cgROSA26 locus (UK2847). Transfected cells were selected with puromycin (8 μg/ml) for 3 days and then treated with 2-Deoxy-D-glucose (4 mM) for 20 h to induce ER stress prior FACS. Cells were phenotypically sorted into 96-well plates by their activation of both UPR reporters (mCherry^high; GFP ^high). Phenotypically selected clones were subsequently confirmed by correct genetic integration. For stability integration, clones were maintained and expanded every 2-3 days for 1 month and weekly analysed with flow cytometry for mCherry and GFP expression under treatment with different ER stressors agents. For the wide-genome CRISPR/Cas9 screen, Cas9 was ultimately introduced by transduction with lentiviral particles prepared from packaging Lenti-Cas9 plasmid (UK1674). Cas9 activity was confirmed by targeting the exon 18 of IRE1 (ERN1 locus) (UK2929) followed by induction of ER stress with 2-Deoxy-D-glucose (4 mM) for 20 h (Supplemental S1E).

CRISPR/Cas9 screen

The screen was performed following established protocols using a pooled CHO-K1 knockout CRISPR library containing 125,030 sgRNAs targeting 21,896 genes, with six guides per gene, as well as 1,000 non-targeting sgRNAs as a negative control cloned into the lentiviral sgRNA expression vector pKLV-U6gRNA(BbsI)-PGKpuro2ABFP (Shalem et al., 2014; Ordonez et al., 2020). Approximately 2.1 × 10⁸ CHO-K1 Cas9 stable cells carrying the dual UPR reporter ATF6⍺/IRE1 (XC45-6S cells) were infected at a multiplicity of infection (MOI) of 0.3 to favour infection with a single viral particle per cell. Two days after infection, transduced cells were selected with 8 µg/ml puromycin for 14 days. Afterwards, transduced cells were split into two subpopulations: one was treated with 2-Deoxy-D-glucose to induce ER stress, while the other one remained untreated. For the ATF6⍺ activator screen, conducted under ER stress conditions, less than 2% of cells that showed a XBP1s::mCherry^high; BiP::sfGFP^low phenotype were selected for analysis. Similarly, for the ATF6⍺ repressor screen, conducted under basal (untreated) conditions, less than 2% of total sorted cells that showed a XBP1s::mCherry^low; BiP::sfGFP^high phenotype were selected for analysis. Rounds of enrichment were carried out through cellular recovery, expansion, and sorting. Equal number of transduced cells, both untreated or 2-Deoxy-D-glucose treated, were passed without sorting as a control group in each round of sorting. To prepare samples for deep sequencing, genomic DNA from enriched and sorted populations as well as unsorted cells (to represent the entire library) was extracted from ∼3.6× 10⁷ and ∼1-3× 10⁶ cells, respectively, by incubation in proteinase K solution [100 mM Tris-HCl pH 8.5, 5 mM EDTA, 200 mM NaCl, 0.25% (w/v) SDS, 0.2 mg/ml Proteinase K] overnight at 50°C. Integrated sgRNA sequences were amplified by nested PCR and the necessary adaptors for Illumina sequencing were introduced at the final round of amplification. The purified products were quantified and sequenced on an Illumina NovaSeq 6000 by 50-bp single-end sequencing. Downstream analyses to obtain sgRNA read counts, gene rankings, and statistics were performed with MAGeCK software (Li et al., 2014). Gene ontology analyses were performed using Metascape with default parameters.

Knockout cells using CRISPR/Cas9 technology

sgRNAs targeting exon regions of Cricetulus griseus CRT, CNX, FURIN and ATF6⍺ were designed using CRISPR-Design tools from CCTop (https://cctop.cos.uni-heidelberg.de:8043/index.html) and CRISPy-Cas9 target finder for CHO-K1 (http://staff.biosustain.dtu.dk/laeb/crispy/) databases. sgRNAs were then cloned into the pSpCas9(BB)-2A-mTurquoise (UK2915) or pSpCas9(BB)-2A-Puro plasmids (UK1367) as previously reported (Ran et al., 2013). The integrity of the constructs was confirmed by DNA sequencing. To generate the knockout cells, 0.75-1 µg of sgRNA/Cas9 plasmids was transfected into XC45-6S cells using Lipofectamine LTX (Thermofisher) in a 6-well plate format. After 48 h post-transfection, mTurquoise^high cells were sorted into 96-well plates at 1 cell/well using a MoFlo Cell Sorter (Beckman Coulter). For cells transfected with sgRNA/Puro plasmids, selection was performed in the presence of 8 μg/ml puromycin, and serial dilution was used to isolate single clones. The knockouts cells were confirmed by Sanger sequencing and immunoblotting.

ATF6⍺ tagged cells by CRISPR/Cas9 genome-editing

To investigate potential modulators of ATF6⍺ activity following the CRISPR/Cas9 high-throughput screen using biochemistry and microscopy-compatible tools, two mammalian cell lines were engineered. These cell lines either harboured a knock-in of the endogenous cgATF6⍺ locus or carried a stable transgene encoding ATF6⍺_LD tagged with a bright GFP. The tags were positioned at the N-cytosolic terminus, known to be well tolerated.

3xFLAG-mGreenLantern_ATF6⍺ knock-in cells (2K cells)

Protein alignment of hsATF6⍺ and cgATF6⍺ (UniProt database) indicated that translation initiation predominantly occurs at the second Met at the beginning of exon 2 (MESP). Instead of tagging at the first Met in the CHO genome, tags were introduced before the third Met. CHO-K1 plain cells were offered with a unique pSpCas9(BB)-2A-mCherry targeting exon 2 of cgATF6⍺ locus (UK2942) and a donor plasmid containing homology arms flanking a minigene encoding a 3xFLAG-mGreenLantern(mGL)-TEV sequence geneblock (UK2955) to create an endogenous tagged ATF6⍺. After 48 h post-transfection, mGL^high cells were sorted as single cells into 96-well plates. Clones that preserved stable mGL fluorescent signal by flow cytometry and showed expected ATF6⍺ processing upon ER stress induction (DTT) were subsequently confirmed by correct genetic integration. Microscopy characterisation was performed under basal conditions or after treatments with ER stressors (DTT), Ceapin A7 and S1P inhibitor. A clone referred to as 2K was expanded and selected for functional analysis as a 3xFLAG-mGL tagged endogenous ATF6⍺.

GFP-cgATF6⍺_LD^S1P,S2Pmut (UK3122 cells)

ATF6 has a half-life of about 3 h (George et al., 2020). To increase the proportion of intact, active N-ATF6⍺ molecules reaching the Golgi, a genetic construct featured the TM and LD of ATF6⍺, deliberately lacking S1P and S2P cleavage sites, and further strategically appended with an Avi-tag at the C-terminal, was engineered. This construct was then used for a dual purpose: as a tool for both microscopy studies as well as to be biotinylated and purified to be used as a ligand in BLI experiments. Shen and Ye’s research provided the basis for the design of S2P and S1P mutations (N391F & P394L for S2P mutant, and LL408/9VV for S1P mutant) (Shen and Prywes, 2004; Ye et al., 2000). A DNA-plasmid encoding eGFP_cgATF6⍺_TEV_LD410-659_S1-2PMUT_AviTag_pCEFL_pu (UK3122) was introduced into CHO-K1 plain cells via Lipofectamine LTX. Transfected cells were selected for resistance to puromycin (8 µg/ml), and the isogenic clones were selected for confocal microscopy assays and in vitro BLI interaction assays.

A genetic platform to generate cgATF6⍺_LD variants knock-in cells

Atf6

Two sgRNA sequences targeting exon 9 and 10 of cgATF6⍺ were incorporated into pSpCas9(BB)-2A-mTurquoise plasmid (UK2915) to generate two sgRNA/Cas9 expression plasmids (UK3021 and UK3024) according to the published procedure (Ran et al., 2013). These plasmids were then transfected into CHO-K1 XC45-6S double UPR reporter cells. After 72 h, mTurquoise^high cells were sorted as single cells into 96-well plates. Genomic DNA from clones showing no response to ATF6⍺ activation (BiP::sfGFP^low) upon tunicamycin treatment (0.5 µM, 20h) was extracted for Sanger sequencing. Clones with frame shifts causing deletions in both alleles of the DNA region flanking by both sgRNAs were selected for further functional assays.

cgATF6⍺_LD “minigenes” knock-in

A unique sgRNA sequence targeting the exon 10 of ⍺ATF6⍺Δ cells was inserted into the pSpCas9(BB)-2A-mTurquoise plasmid to create the sgRNA/Cas9 plasmid UK3025. Three repair “minigenes” templates were constructed: A wild-type, glycan free (ΔGly) or a cysteine free (ΔC) LD to be reconstituted in ATF6⍺Δ cells by CRISPR/Cas9-mediated homology-directed repair (HDR).

For the construction of the wild-type LD repair template (WT; UK3054), a minigene block that contains exons 10 -17 of cgATF6⍺ (785bp; GenScript) was digested with AgeI and AflII and ligated into AgeI/ AflII-digested pUC57 plasmid. To create a Glycan free LD (ΔGly; UK3086) template, a minigene block containing exon 10 -17 (GenScrip) with the three N-glycosylation sites were mutated by replacing Thr 464, 575 and 634 with Ala, was cloned into pUC57 plasmid as previously described. Similarly, to generate a Cysteine-free LD (ΔC; UK3147) template, a minigene block containing exon 10 -17 (GenScrip) with Cys457 and Cys607 replaced with Ser, was cloned into pUC57 plasmid as previously performed. The integrity of the constructs was confirmed by DNA sequencing.

Immunoprecipitation using GFP-Trap Agarose

Equal volumes of the cleared and normalised lysates were incubated with 15-20 μl GFP-Trap Agarose (ChromoTek) pre-equilibrated in lysis buffer. The mixture was incubated overnight at 4 °C with rotation. The beads were then recovered by centrifugation (845xg, 3 min) and washed four times with washing buffer (50 mM Tris-HCl, 1 mM EDTA, 0.1% Triton X-100) and one more with PBS. Proteins were eluted in 35 μl of 2×SDS-DTT sample for 5 min at 95°C. Equal sample volumes were analysed by SDS-PAGE and immunoblotting, as described above. Normalised cell lysates (15-30 μg) were loaded as an ‘input’ control.

Bio-layer Interferometry (BLI), protein expression and in vivo biotinylation

BLI studies were undertaken using on the Octet RED96 System (Pall FortéBio) using ATF6⍺_LD variants as biotinylated ligands that were immobilised on streptavidin (SA)-coated biosensors (Pall FortéBio) and rat calreticulin (rCRT) as the analyte. Prior to use, SA-coated biosensors were hydrated in a BLI buffer (150 mM NaCl, 20 mM Tris, 1mM CaCl2) for 10 min. BLI reactions were prepared in 200 μl volumes in 96-well microplates (greiner bio-one), at 30°C. Ligand loading was performed for 600 sec until a binding signal of 1 nm displacement was reached. Following ligand immobilisation, the sensor was baselined in the reaction buffer for at least 200 seconds, after which it was immersed in wells containing the analyte for association for 1200 seconds. After each association, the sensor was dipped into a buffer-containing well to allow dissociation for 2400 sec.

Biotinylated ATF6⍺LD constructs [ATF6⍺LD^WT (UK3122); ATF6⍺LD^ΔGly (UK3141) and ATF6LD^ΔC (UK3157)] were expressed in HEK293T cells, each one encoding a N-terminal GFP tag followed by a TEV protease cleavage site just after the TM and an Avi-tag at C-terminal. S1P and S2P cleavage sites were also mutated to increase expression yield. Biotinylation was conducted in vivo, where cells were co-transfected with a BirA-encoding plasmid (biotin ligase, UK2969) in a 1:0.06 ratio. After 24 h, media was replaced with fresh media containing 50 μM biotin dissolved in DMSO and cells were harvested and lysed the following day. The lysis buffer was supplemented with 5 mM CaCl₂. Post immunoprecipitation using GFP-Trap Agarose, proteins were eluted by incubation with an excess of TEV protease (UK759, 16 μg) in 400 ul TEV cleavage buffer (25 mM Tris–HCl pH 7.4, 150 mM NaCl, 0.05% NP40, 0.05mM CaCl₂, 0.05 mM TCEP) for 16 h at 4°C with orbital rotation. Cleavage products were snap-frozen in liquid nitrogen and stored at -80°C for BLI experiments.

Wild-type full-length rCRT (UK3142) was expressed in E.Coli cells (BL21 C3013). Bacterial cultures were grown in LB medium with 100 µg/ml ampicillin at 37°C to an OD_600nm of 0.6-0.8. Expression was induced with 1 mM IPTG, and the cells were further grown for 16 h at 18°C. After cell sedimentation by centrifugation, the pellets were lysed with a high-pressure homogeniser (EmulsiFlex-C3; Avestin) in bacterial lysis buffer (50 mM Tris-HCl pH 7.4, 500 mM NaCl, 2mM CaCl₂, 20mM imidzole) containing protease inhibitors (1 mM PMSF, 2 µg/ml pepstatin, 8 µg/ml aprotinin, 2 µM leupeptin) and 0.1 mg/ml DNaseI. Typical yields from 2 litres of bacterial culture were obtained.

Immunoprecipitation using GFP-Trap Agarose

To investigate the interaction between CRT and ATF6⍺ in cells, HEK293T cells were cultured in 10-cm dishes (3-4 plates per sample) and transfected with constructs expressing either GFP-ATF6⍺_LD^WT (UK3122) or GFP-ATF6⍺_LD^ΔGly (UK3141). Equal volumes of the cleared and normalised lysates were incubated with 15-20 μl GFP-Trap Agarose (ChromoTek) pre-equilibrated in lysis buffer. The mixture was incubated overnight at 4 °C with rotation. The beads were then recovered by centrifugation (845xg, 3 min) and washed four times with washing buffer (50 mM Tris-HCl, 1 mM EDTA, 0.1% Triton X-100). Proteins were eluted in 35 μl of 2×SDS-DTT sample for 5 min at 95°C. Equal sample volumes were analysed by SDS-PAGE and immunoblotting, as described above. Normalised cell lysates (15-30 μg) were loaded as an ‘input’ control.

Analysis of endogenous ATF6⍺ processing in CRT-depleted cells

To investigate the impact of CRT depletion on ATF6⍺ cleavage, CRT was depleted in 2K cells (3xFLAG-mGL-ATF6⍺) by CRISPR/Cas9 gene editing (UK2991). 2K-CRTΔ derivative clones were confirmed by Sanger sequencing and selected for further experiments. Both 2K parental cells and 2K-CRTΔ derivative cells were cultured in 10 cm-dishes for 2 days and then treated with 2-Deoxy-D-glucose (4 mM) for a time course up to 6 h before harvested. Following treatment, the cells were lysed in Nonidet-lysis buffer for extracting the cytosolic fraction, as previously described. For nuclear extraction, high-salt lysis buffer (500 mM NaCl, 50 mM Tris-HCl pH 7.5, 1% (v/v) NP-40) supplemented with 2x protease inhibitor cocktail was added to the nuclear pellets, followed by rotation at 4°C for 2 h. Subsequently, samples were centrifuged at 15,000 rpm at 4°C for 15 min, and the soluble portion containing the nuclear extract was combined with the cytosolic fraction. Equal volumes of clarified and normalised cytosolic and nuclear extracts were incubated with 15-20 μl GFP-Trap Agarose (ChromoTek) for overnight at 4 °C with rotation. Afterwards, beads were recovered by centrifugation, washed three times in washing buffer and eluted in 35 μl of 2×SDS-DTT sample for 5 min at 95°C.

Live cell confocal microscopy

Cells were grown on 35 mm live cells imaging dishes (MatTek), transfected as required, and maintained in culture for 48 h after transfection. Live cell microscopy was conducted using an inverted confocal laser scanning microscope (Zeiss LSM880) in a 37°C/5% v/v chamber with a 64x oil immersion objective lens with 1 Airy unit pinhole using the 488 nm (GFP; mGL), 405 nm (DAPI) and/or 594 nm (mCherry) lasers as appropriate. Co-localisation analysis between fluorescence channels (Pearson correlation coefficient) within individual cells was performed using Volocity software, version 6.3 (PerkinElmer).

RNA isolation and reverse transcription

To extract total RNA, cells were treated with TRIzol™ Reagent (Invitrogen) for 10 min and then transferred to a new tube. Chloroform (200 µl) was added into each sample, vortexed for 1 min, and then centrifuged at 13,500xg, at 4°C for 15 min. The upper clear phase was collected and mixed with equal volume of 70% ethanol. Afterward, PureLink™ RNA mini kit (Invitrogen) was used for washing and elution. RNA yield was assessed using NanoDrop, and 2 µg of RNA from each sample was evaluated for integrity via electrophoresis on a 1.2% agarose gel.

For reverse transcription (RT)-PCR, 2 µg of RNA was initially heated up at 70°C with RT buffer (Thermo Scientific; Cat #EP0441), 0.5 mM dNTP and 0.05 mM Oligo dT18 for 10 min. Subsequently, the reaction was supplemented with 0.5 µl RevertAid Reverse Transcriptase (Thermo Scientific) and 100 mM DTT, followed by incubation at 42°C for 90 min. The resulting cDNA product was diluted up to 1:4.

PCR analysis of XBP1 mRNA splicing

XBP1s and XBP1u fragments were amplified from cDNA by PCR using primers flanking the splicing site identified by IRE1 with NEB Q5® High-Fidelity 2X Master Mix, as previously reported (Neidhardt et al., 2023). Briefly, XBP1u (255bp) and XBP1s (229bp) were separated in a 3 % agarose gel by electrophoresis and stained with SYBR Safe nucleic acid gel stain (Invitrogen). Additionally, a XBP1^hybrid band (280 bp), was observed. The percentage of XBP1s was quantified by determining the band intensity with Fiji, v1.53c. For quantification, it was assumed that 50% of the XBP1^hybrid signal represented XBP1s, while the remaining 50% represented XBP1u.

Immunoprecipitation and FURIN cleavage assay

To investigate whether ATF6⍺ is directly cleaved by FURIN, GFP-ATF6⍺_LD^S1P,S2Pmut (UK3122) construct was expressed and purified from a FURINΔ clone (#2.2.G). Equal volumes of clarified and normalised cell lysates were incubated with 15-20 μl GFP-Trap Agarose (ChromoTek) for overnight at 4 °C with rotation. Afterwards, beads were recovered by centrifugation, washed three times in washing buffer and then incubated with 75 μl of FURIN cleavage buffer (20 mM HEPES, 1 mM CaCl₂, 0.2 mM β-mercaptoethanol, 0.1% Triton X-100; pH 7.5) supplemented with 1 U FURIN (#P8077S, NEB) for 6 h at 25°C. As a negative control, calcium-depleted FURIN buffer was used. Additionally, a FURIN control substrate, MBP5-FN-paramyosin-ΔSal (#E8052S, NEB) containing a FURIN site, was included in parallel reactions and incubated for 6h at 25°C. The eluted protein was then analysed either via immunoblot or Coomassie blue staining.

Statistics

Data groups were analysed as described in the figure legends using GraphPad Prism10 software. Differences between groups were considered statistically significant if * p < 0.05; ** p < 0.01; *** P < 0.001; and **** p < 0.0001. All error bars represent mean ± SEM.

Data availability

The raw and processed high-throughput sequencing data reported in this paper have been uploaded in NCBI’s Gene Expression Omnibus (GEO, accession number: GSE254745).

Acknowledgements

We thank the CIMR flow cytometry core facility team (Reiner Schulte, Chiara Cossetti and Gabriela Grondys-Kotarba), the microscopy team (Matthew Gratian and Mark Bowen) for technical support and the Huntington lab for access to the Octet machine. We also thank Marcella Ma (CRUK) for assistance with NGS and Avi Ashkenazi (Genentech) for the monoclonal antibody against human IRE1_LD. This work was supported by Wellcome Trust Principal Research Fellowship to DR (Wellcome 224407/Z/21Z),

Author’s contributions

AO: co-led, conceptualisation; conceived and initiated the project; designed and conducted the CRISPR/Cas9 screen; conducted in vivo experiments and biophysical experiments; analysed and interpreted the data; prepared figures and wrote the original draft. JT: designed and conducted most of the in vivo experiments; analysed and interpreted the data; contributed to discussion. LH: conducted in vivo experiments; analysed and interpreted the data; GG: conducted in vivo experiments; interpreted the data; contributed to discussion. HPH: constructed the XCdel45 precursor cell line; interpreted the data; contributed to discussion. DR: conceptualisation; supervision; funding acquisition; investigation; interpreted data; co-wrote the manuscript. All authors read and approved the final manuscript.

Supplemental Figures

Supplemental Fig S1.

Supplemental Fig S2.

Supplemental Fig S3.

Supplemental Fig S4.

Supplemental Fig S5.

Supplemental Fig S6.

Supplemental Tables

Table S2:

Table S3:

Table S4: