Production of an Active, Human Membrane Protein in Saccharomyces cerevisiae: Full-Length FICD

The human Fic domain-containing protein (FICD) is a type II endoplasmic reticulum (ER) membrane protein that is important for the maintenance of ER proteostasis. Structural and in vitro biochemical characterisation of FICD AMPylase and deAMPylase activity have been restricted to the soluble ER-luminal domain produced in Escherichia coli. Information about potentially important features, such as structural motifs, modulator binding sites or other regulatory elements, is therefore missing for the approximately 100 N-terminal residues including the transmembrane region of FICD. Expressing and purifying the required quantity and quality of membrane proteins is demanding because of the low yields and poor stability often observed. Here, we produce full-length FICD by combining a Saccharomyces cerevisiae-based platform with green fluorescent protein (GFP) tagging to optimise the conditions for expression, solubilisation and purification. We subsequently employ these conditions to purify milligram quantities of His-tagged FICD per litre of culture, and show that the purified, detergent-solubilised membrane protein is an active deAMPylating enzyme. Our work provides a straightforward methodology for producing not only full-length FICD, but also other membrane proteins in S. cerevisiae for structural and biochemical characterisation.


Introduction
Fic (filamentation induced by cAMP) proteins are ubiquitous, structurally conserved enzymes that post-translationally modify hydroxyl-containing residues by cleaving highenergy phosphate bonds on small co-substrates, such as ATP, and transferring phosphatecontaining groups to substrates [1]. Fic proteins are prevalent in bacteria, and only a single fic gene has been identified in metazoans [2,3]. The human Fic domain-containing protein, FICD ( Figure 1A), is a type II transmembrane protein that localises to the endoplasmic reticulum (ER), and to a lesser extent to the nuclear envelope [4][5][6][7][8].
Membrane proteins constitute 25-30% of both the eukaryotic and prokaryotic proteomes [24][25][26], but are strongly underrepresented in the protein data bank [27,28]. Moreover, while single-pass membrane proteins such as FICD represent roughly half of all membrane proteins, structural studies of these are limited [24,29,30]. The currently available crystal structures of FICD include the catalytic domain and the adjacent tetratricopeptide repeats (TPRs) ( Figure 1B) [10][11][12]22]. In vitro investigations of FICD's structure and enzymatic mechanism have been conducted using N-terminally truncated (lacking the first approximately 100 residues), soluble protein variants produced in E. coli. Thus, the transmembrane domain and the membrane-adjacent regions (residues 1-102) have not been functionally or structurally characterised although they may harbour important determinants regulating, e.g., activity, localisation and protein-protein interactions.
Obtaining sufficient quantities of pure membrane protein can be both challenging and time consuming due to low expression levels, solubility issues and challenges in obtaining and maintaining uniformly folded protein. To overcome some of these challenges, we employed the previously developed PAP1500 S. cerevisiae-based membrane protein expression platform [31][32][33][34][35][36], and coupled this with GFP tagging to produce full-length human FICD (Figure 2). This platform has been successfully used to produce multi-spanning membrane proteins such as aquaporins, transmembrane ATPases and ion channels [32][33][34][35][37][38][39][40]. The approach bypasses traditional cloning procedures that can be time consuming and costly, and lead to the inclusion of cloning scars [41]. Instead, the platform utilises the ability of S. cerevisiae to recombine linear DNA fragments with overlapping segments, such as PCR products, into a mature expression vector [42,43]. The host strain overproduces the Gal4 transcriptional activator, which works in concert with the strong galactose-inducible CYC-GAL promoter in the pEMBLyex4 shuttle vector ( Figure 3A). FICD is comprised of a short Nterminal cytosolic tail, a transmembrane domain (TM; dark grey), two tetratricopeptide repeats (TPR; turquoise), a linker (yellow), and the catalytic Fic domain (FIC; blue). The segment for which crystal structures have been solved (approximately residues 102 to 445 depending on the particular study) is indicated [10][11][12]22]. The numbers denote residue positions that mark the boundaries of key regions in the structure of FICD. (B) Crystal structure of the N-terminally truncated, soluble FICD 103-433 dimer (PDB ID: 4U04) displays a helical conformation in which FICD forms an asymmetric, noncovalent dimer via the protomer Fic domains (I and II). Same colour scheme as Panel A. Created using PyMol [23].
Membrane proteins constitute 25-30% of both the eukaryotic and prokaryotic proteomes [24][25][26], but are strongly underrepresented in the protein data bank [27,28]. Moreover, while single-pass membrane proteins such as FICD represent roughly half of all membrane proteins, structural studies of these are limited [24,29,30]. The currently available crystal structures of FICD include the catalytic domain and the adjacent tetratricopeptide repeats (TPRs) ( Figure 1B) [10][11][12]22]. In vitro investigations of FICD's structure and enzymatic mechanism have been conducted using N-terminally truncated (lacking the first approximately 100 residues), soluble protein variants produced in E. coli. Thus, the transmembrane domain and the membrane-adjacent regions (residues 1-102) have not been functionally or structurally characterised although they may harbour important determinants regulating, e.g., activity, localisation and protein-protein interactions.
Obtaining sufficient quantities of pure membrane protein can be both challenging and time consuming due to low expression levels, solubility issues and challenges in obtaining and maintaining uniformly folded protein. To overcome some of these challenges, we employed the previously developed PAP1500 S. cerevisiae-based membrane protein expression platform [31][32][33][34][35][36], and coupled this with GFP tagging to produce full-length human FICD ( Figure 2). This platform has been successfully used to produce multi-spanning membrane proteins such as aquaporins, transmembrane ATPases and ion channels [32][33][34][35][37][38][39][40]. The approach bypasses traditional cloning procedures that can be time consuming and costly, and lead to the inclusion of cloning scars [41]. Instead, the platform utilises the ability of S. cerevisiae to recombine linear DNA fragments with overlapping segments, such as PCR products, into a mature expression vector [42,43]. The host strain overproduces the Gal4 transcriptional activator, which works in concert with the strong galactose-inducible CYC-GAL promoter in the pEMBLyex4 shuttle vector ( Figure 3A). This promoter combines the constitutive cytochrome c1 (CYC1) promoter with the upstream activating sequence from the GAL1-GAL10 promoter [44]. This expression system ensures that the plasmid copy number can be increased from about 20 to 200-400 copies per cell prior to the induction of expression by starving the cells for leucine [37,45]. This platform has been shown to lead to exceptionally high amounts of pure membrane protein (up to 8% of the total membrane protein content in the yeast cell [32]), compatible with biophysical and functional characterisation, including X-ray crystallography [32,34,37,46,47], cryo-EM [48] and SAXS [49].
This promoter combines the constitutive cytochrome c1 (CYC1) promoter with the upstream activating sequence from the GAL1-GAL10 promoter [44]. This expression system ensures that the plasmid copy number can be increased from about 20 to 200-400 copies per cell prior to the induction of expression by starving the cells for leucine [37,45]. This platform has been shown to lead to exceptionally high amounts of pure membrane protein (up to 8% of the total membrane protein content in the yeast cell [32]), compatible with biophysical and functional characterisation, including X-ray crystallography [32,34,37,46,47], cryo-EM [48] and SAXS [49].
To identify the optimal conditions for obtaining solubilised, correctly folded FICD, we introduced a GFP tag at its cytosolic N-terminal tail [47]. This allowed us to use smallscale expression tests to quickly optimise the expression conditions and identify suitable detergents for solubilisation, before initiating large-scale expression and purification of FICD without the GFP tag. A single round of expression produced approximately 20 mg of FICD in a buffer that is compatible with enzymatic and biophysical characterisation, as supported by the demonstration of deAMPylation activity by the purified protein.    Table S1 for primer details). CG-P,  TEV-FICD and His 8 -FICD DNA fragments were constructed by PCR using appropriate cDNA templates and primers containing overlapping sequences (see Table S1 for primer details). CG-P, CYC-GAL promoter; leu2-d, a poorly expressing allele of the β-isopropylmalate dehydrogenase gene; URA3, selective marker gene encoding the orotidine-5 -phosphate decarboxylase enzyme; BLA, β-lactamase gene; 2µ, 2-µm origin of replication; pMB1, bacterial origin of replication. To identify the optimal conditions for obtaining solubilised, correctly folded FICD, we introduced a GFP tag at its cytosolic N-terminal tail [47]. This allowed us to use smallscale expression tests to quickly optimise the expression conditions and identify suitable detergents for solubilisation, before initiating large-scale expression and purification of FICD without the GFP tag. A single round of expression produced approximately 20 mg of FICD in a buffer that is compatible with enzymatic and biophysical characterisation, as supported by the demonstration of deAMPylation activity by the purified protein.

Recombinant GFP-FICD Is Expressed to a High Density in S. cerevisiae
To generate the expression vectors encoding human His 8 -GFP-TEV-FICD (GFP-FICD), including a TEV protease cleavage site, or His 8 -FICD (His-FICD), we utilised the ability of S. cerevisiae to recombine linear DNA fragments by homologous recombination ( Figure 3A, Table S1). We chose to perform FICD expression at 15 • C, as previous experiments have shown that although S. cerevisiae grows optimally at 30 • C, membrane protein folding and maturation is often compromised at this temperature [32,34]. To determine the appropriate induction time, the total cell GFP fluorescence was measured at 24 h intervals ( Figure 3B). In total, 120 h after induction, fluorescence accumulation slowed down significantly, and the cells were harvested. GFP-FICD was expressed to a concentration of approximately 180 pmol/mg membrane protein, corresponding to approximately 1.5% of total membrane protein content. The optimal expression conditions obtained from the small-scale experiments were subsequently applied for large-scale production of the His-FICD construct in 15-litre bioreactors (see below).

GFP-FICD Likely Localises to the Golgi
In human cells, FICD localises to the ER membrane, where its C-terminal catalytic domain faces the ER lumen. We used fluorescence microscopy to ensure that heterologously expressed GFP-FICD was integrated into a membrane in our S. cerevisiae expression system. Here, the GFP signal did not display a pattern characteristic of ER localisation [50]. We compared the FICD signal to that of two GFP-tagged yeast membrane proteins of known localisation, the vacuolar Pmc1 Ca 2+ -ATPase and the Golgi-located Pmr1 Ca 2+ , Mn 2+ -ATPase [35,51,52] ( Figure 3C). GFP-FICD showed a distribution similar to that of Pmr1-GFP, suggesting that in yeast this FICD fusion protein is most likely integrated into the Golgi membrane.
To investigate whether the GFP-FICD protein accumulated as a full-length protein, we separated the membrane proteins by SDS-PAGE and visualised the fusion protein using in-gel fluorescence ( Figure 3D). The large majority of GFP-FICD migrated according to expectation at around 62-67 kDa, including a 10-15 kDa contribution from the correctly folded denaturation-resistant GFP tag [37,53]. A faint fluorescent band was visible at around 28 kDa, the molecular weight of GFP, indicating that the tag was removed from a very small fraction of the fusion protein.

Zwitterionic Detergents Recover More GFP-FICD Than Non-Ionic Detergents
We next carried out a screen to identify detergents (see Table 1) that could solubilise GFP-tagged FICD from the crude membranes, while preserving protein integrity ( Figure 4). Detergents were either used alone or in combination with cholesteryl hemisuccinate (CHS), which may have a stabilising effect during membrane protein solubilisation and purification [33,35]. Most of the harsher ionic detergents were only tested in the presence of CHS. Overall the solubilisation efficiency varied substantially, ranging from about 5% to 100% solubility, with the zwitterionic Fos-cholines (FC12 and FC13) and n-dodecyldimethylamine oxide (LDAO) demonstrating higher recovery (70-100%) than the non-ionic detergents ( Figure 4A,B). Fos-cholines recovered over 90% of the fluorescence when used in combination with CHS, whereas LDAO solubilised approximately 60% of GFP-FICD. For all zwitterionic detergent combinations, a 2:1 detergent to protein ratio yielded the best results. Table 1. Detergents used in the screening phase.

Abbreviation
Detergent Name Properties Among the non-ionic detergents, the maltoside detergents DM, DDM and TDM solubilised an acceptable 30-40% of GFP-FICD, while HDM performed poorly ( Figure 4A,B). The glucopyranoside detergents OG and NG only solubilised around 10% and 25% of the fusion protein, respectively. The maltopyranoside detergent MNG and the cyclohexyl maltoside detergent PCCM performed similarly to NG. The addition of CHS markedly lowered the yield in all detergent combinations excluding DM. The solubilisation efficiency of the non-ionic detergents could be improved slightly but consistently by increasing the detergent concentration ( Figure 4A,B).

Cholesteryl Hemisuccinate Improves GFP-FICD Homogeneity
We next tested the best performing detergents for their ability to retain GFP-FICD in a folded conformation. For this purpose, we used FSEC, in which a monodisperse (symmetric) elution profile is indicative of folding and a unique oligomeric state [54][55][56][57]. The zwitterionic FC12 combined high solubilisation efficiency with high protein integrity, as shown by the monodisperse elution profile in the absence and presence of CHS ( Figure 4C). GFP-FICD that was solubilised in the non-ionic DM or DDM eluted asymmetrically in the absence of CHS, whereas the addition of CHS led to a monodisperse elution profile ( Figure 4D,E). In all conditions, only a small amount of fluorescent protein eluted in the void volume, likely representing higher molecular weight aggregates ( Figure 4C-E). Although FC12 provided the highest overall yield, we decided to proceed with DM and DDM in conjunction with CHS, since these detergents are best suited for downstream applications such as biochemical characterisation and crystallography [58].   Table 1 for detergent names and properties). When noted, solubilisation was performed in the presence of CHS. Fluorescence intensity was measured in the soluble fraction after ultracentrifugation, and solubility was quantified as the proportion of fluorescence in the soluble fraction relative to total membrane fluorescence. Solubility in three different detergent to protein ratios (1:1, 2:1 and 3:1) is shown. (B) GFP-FICD solubility measured as in A, but in two detergent to protein ratios (1:1 and 3:1). (C-E) FSEC profiles of GFP-FICD solubilised in the indicated detergents in the highest yielding detergent to protein ratio (FC12; 2:1 (C), DM; 3:1 (D), DDM; 3:1 (E)). An in-line fluorescence detector was used to record the elution profile of GFP-FICD. Blue line, without CHS; red line, with CHS. Fluorescent material eluting in the void volume is denoted by an asterisk (*).

Cholesteryl Hemisuccinate Improves GFP-FICD Homogeneity
We next tested the best performing detergents for their ability to retain GFP-FICD in a folded conformation. For this purpose, we used FSEC, in which a monodisperse (symmetric) elution profile is indicative of folding and a unique oligomeric state [54][55][56][57]. The zwitterionic FC12 combined high solubilisation efficiency with high protein integrity, as shown by the monodisperse elution profile in the absence and presence of CHS ( Figure  4C). GFP-FICD that was solubilised in the non-ionic DM or DDM eluted asymmetrically in the absence of CHS, whereas the addition of CHS led to a monodisperse elution profile ( Figure 4D,E). In all conditions, only a small amount of fluorescent protein eluted in the void volume, likely representing higher molecular weight aggregates ( Figure 4C-E). Although FC12 provided the highest overall yield, we decided to proceed with DM and DDM in conjunction with CHS, since these detergents are best suited for downstream applications such as biochemical characterisation and crystallography [58].  Table 1 for detergent names and properties). When noted, solubilisation was performed in the presence of CHS. Fluorescence intensity was measured in the soluble fraction after ultracentrifugation, and solubility was quantified as the proportion of fluorescence in the soluble fraction relative to total membrane fluorescence. Solubility in three different detergent to protein ratios (1:1, 2:1 and 3:1) is shown. (B) GFP-FICD solubility measured as in A, but in two detergent to protein ratios (1:1 and 3:1). (C-E) FSEC profiles of GFP-FICD solubilised in the indicated detergents in the highest yielding detergent to protein ratio (FC12; 2:1 (C), DM; 3:1 (D), DDM; 3:1 (E)). An in-line fluorescence detector was used to record the elution profile of GFP-FICD. Blue line, without CHS; red line, with CHS. Fluorescent material eluting in the void volume is denoted by an asterisk (*).

His-FICD Can Be Purified in a Folded Conformation
Structural and biochemical characterisation can be hampered by the presence of the large and bulky GFP tag. Since we were unable to efficiently remove the tag by TEV protease cleavage ( Figure S1), we generated a His-tagged variant of our FICD construct, devoid of GFP. His-FICD was expressed and solubilised according to our optimised protocol, before subjecting the soluble fraction to immobilised metal affinity purification (IMAC). The identity of the purified protein was confirmed by SDS-PAGE and immuno -blotting ( Figure 5A). The analysis revealed the presence of monomeric and oligomeric FICD species. Despite this apparent heterogeneity, during the subsequent size exclusion chromatography (SEC) analysis the protein eluted in a symmetric peak ( Figure 5B), indicating that the His-tagged FICD adopts a folded, unique oligomeric state in solution.
The purified sample was further analysed by reducing and non-reducing SDS-PAGE ( Figure 5C). A few high molecular weight species observed on the gel but not in the SEC analysis ( Figure 5B) were partially sensitive to DTT, whereas the FICD dimer band was DTT resistant, suggesting that the purified protein might form non-covalent dimers. This would not be surprising, since N-terminally truncated FICD adopts a dimeric form in vitro [10,12], and non-covalent, SDS-resistant oligomerisation is often observed for oligomeric membrane proteins during SDS-PAGE analysis [37,59,60]. Overall, the chromatography analysis indicated that the GFP tag is not required for FICD stabilisation, and that the His-tagged variant remained folded upon purification.
2.6. Purified Full-Length His-FICD Is an Active deAMPylase FICD's switch from AMPylation to deAMPylation activity is governed by dimerisation at a low nanomolar concentration (K D = 1.2 nM), as shown for the N-terminally truncated enzyme, which exists exclusively as a dimer in solution and possesses very weak AMPylation activity towards FICD's natural substrate, BiP [9,10,12]. We therefore hypo-thesised that full-length His-FICD functions as a deAMPylase.
Structural and biochemical characterisation can be hampered by the presence of the large and bulky GFP tag. Since we were unable to efficiently remove the tag by TEV protease cleavage ( Figure S1), we generated a His-tagged variant of our FICD construct, devoid of GFP. His-FICD was expressed and solubilised according to our optimised protocol, before subjecting the soluble fraction to immobilised metal affinity purification (IMAC). The identity of the purified protein was confirmed by SDS-PAGE and immunoblotting ( Figure 5A). The analysis revealed the presence of monomeric and oligomeric FICD species. Despite this apparent heterogeneity, during the subsequent size exclusion chromatography (SEC) analysis the protein eluted in a symmetric peak ( Figure 5B), indicating that the His-tagged FICD adopts a folded, unique oligomeric state in solution.
The purified sample was further analysed by reducing and non-reducing SDS-PAGE ( Figure 5C). A few high molecular weight species observed on the gel but not in the SEC analysis ( Figure 5B) were partially sensitive to DTT, whereas the FICD dimer band was DTT resistant, suggesting that the purified protein might form non-covalent dimers. This would not be surprising, since N-terminally truncated FICD adopts a dimeric form in vitro [10,12], and non-covalent, SDS-resistant oligomerisation is often observed for oligomeric membrane proteins during SDS-PAGE analysis [37,59,60]. Overall, the chromatography analysis indicated that the GFP tag is not required for FICD stabilisation, and that the His-tagged variant remained folded upon purification.  To test for deAMPylation activity, we first AMPylated BiP with radioactively labelled α-[ 32 P]-ATP using the soluble L258D variant of FICD ∆102 (produced in E. coli), which is rendered monomeric by a mutation in the dimer interface and is known to efficiently AMPylate BIP [9,10,12,61]. AMPylated BiP (BiP-AMP) was then purified, and subsequently incubated in the presence or absence of His-FICD. The time-dependent removal of radiolabelled [ 32 P]-AMP was quantified by phosphorimaging ( Figure 5D,E). His-FICD readily removed~80% of the AMP from BiP within 60 min with a half-time for deAMPylation of approximately 15 min, demonstrating that the enzyme functions as a deAMPylase in vitro. In the absence of the enzyme, the AMPylation level remained largely constant, indicating that the removal of AMP was due to the enzymatic activity of His-FICD, and not due to spontaneous hydrolysis. The minor increase in BiP AMPylation observed at the 2 min timepoint is within the experimental error. Alternatively, it may be explained by the potential presence of residual amounts of ATP and monomeric FICD in the sample, carried over from the preceding AMPylation reaction. Regardless, full-length His-FICD deAMPylated BiP with comparable kinetics as observed for the N-terminally truncated enzyme purified from E. coli (see Discussion). Overall, the protocol presented here allowed the purification of full-length human FICD from yeast membranes in an active form, in a buffer that supports both structural and functional analysis.

Discussion
Most published crystal structures of membrane proteins are of prokaryotic origin [27]. Since approximately 85% of human membrane proteins do not have a prokaryotic ortholog, our knowledge of their structure and function is limited [62]. One such protein is the human ER membrane protein FICD, which has been studied in vitro using only its truncated, soluble form. Here, we have developed a protocol to express, solubilise, and purify functional, full-length human FICD using the S. cerevisiae-based PAP1500 membrane protein production platform [31][32][33][34][35]. Similar to bacteria, yeasts are inexpensive to cultivate and require relatively simple culture conditions. As a eukaryote, yeasts are more suitable for the expression of human proteins that require specific post-translational modifications for function or maturation. For our purposes, S. cerevisiae was an even more attractive host because yeast does not carry any known fic genes, preventing the inadvertent purification of heterodimers of yeast and human FICD. To streamline the expression, solubilisation and purification conditions, we coupled the expression platform with GFP tagging of FICD [47].
In human cells, FICD localises to the ER membrane in a type II orientation [4,12,17]. However, as FICD does not contain an arginine-based ER-retention motif [63,64], it is presently unclear how the protein is retained in the ER in human cells. In our S. cerevisiae expression system, the GFP-tagged FICD construct most likely localised to the Golgi membrane ( Figure 3C). We and others have previously reported that some heterologously expressed proteins can be mislocalised in yeast [37,65], potentially caused by inherent differences in secretory system function or due to the strong expression of the proteins. It is also possible that the GFP tagging overrides or interferes with an ER retention signal. Despite the unexpected subcellular localisation, the fusion protein was not degraded. By quantifying the GFP fluorescence, we could estimate that the exogenous GFP-FICD was expressed to a density of~181 pmol/mg membrane protein in crude membranes, over a 120 h expression period. This represents about 1.5% of the total membrane protein content ( Figure 3B), which is sufficient for biochemical characterisation and structure determination.
GFP tagging ensured that we could easily test the ability of various detergents to solubilise and retain FICD in the folded conformation (as judged from FSEC elution profiles) [47,66]. By screening various detergents, FICD solubility was shown to be far greater in the relatively harsher zwitterionic detergents than in the milder non-ionic detergents ( Figure 4A,B). The non-ionic DM and DDM also retained the protein in a folded conformation, and are generally more compatible with crystallography [58]. Due to its ten-fold lower critical micelle concentration, DDM is generally a more economical choice than either DM or FC12. On the other hand, the use of DM might be preferable over DDM in some cases due to the significantly smaller micelle size of DM [67].
Biophysical characteristics of the detergents used in the present study are shown in Table S2. Due to large differences in CMC values and costs, detergents are not used at the same relative CMC value. This is also apparent from Table S2, which details that detergents were used at concentrations ranging from 0.4× CMC to 17,667× CMC. An important issue that is usually not addressed in practical membrane protein purification is the dimension of each micelle. A membrane-spanning α-helix consists of approximately 20 amino acid residues since the hydrophobic part of a eukaryotic membrane is 30 Å. As is apparent from Table S2, the tested detergents have a radius of gyration that matches this value.
Here, we used the knowledge gained from the initial screening phase to scale up the expression and purification of His-FICD (solubilised in DDM), resulting in the purification of 1-2 mg of protein per litre of yeast culture ( Figure 5A-C). As alternatives to detergents, it may be worth pursuing styrene-maleic acid copolymers (SMALPs) for solubilisation [68] and insertion of the detergent solubilised FICD protein into phospholipid nanodiscs [69] to stabilise the protein. These approaches have the advantage that the membrane protein is maintained in a native lipid environment or at least a phospholipid environment.
Solubilised His-FICD adopted a single oligomeric state as judged by SEC analysis (Figure 5B), as is the case for N-terminally truncated FICD, which forms a highly stable homodimer in the solution and in all published structures, e.g. [10][11][12]. To assess the oligomeric state of His-FICD experimentally, we employed native mass spectrometry, but were unfortunately not able to detect the protein by this method (Drs. C. Sahin and M. Landreh, Karolinska Institutet, Stockholm, Sweden, personal communication). Still, based on the strictly dimeric behaviour of the protein produced in E. coli, we find it likely that His-FICD is also a dimer. Similar to dimeric, N-terminally truncated FICD, full-length His-FICD readily deAMPylated its natural substrate BiP in vitro ( Figure 5D,E). The kinetics of BiP deAMPylation displayed by His-FICD were similar to published data for BiP deAMPylation by N-terminally truncated FICD [9], with both showing deAMPylation half-times of 10-15 min with around 80% deAMPylation within 60 min, when conducting experiments under comparable conditions. We conclude that the conditions chosen for detergent solubilisation of His-FICD are well suited to preserve the structural integrity and function of the protein.
The expression platform developed in this study allowed the production of detergentsolubilised human FICD displaying deAMPylation activity. To our knowledge, this is the first time that FICD purified to homogeneity has been investigated in vitro in its full-length form. Of note, the transmembrane domain and juxta-membrane regions of single-pass membrane proteins are increasingly being recognised as functionally important, modulating for example protein-protein interactions, signalling and cellular localisation [70][71][72][73][74]. Investigating the previously uncharacterised N-terminal region of FICD may thus reveal the presence of unidentified structural motifs, modulator binding sites or other regulatory elements that would be impossible to identify when using a truncated variant of the enzyme. Our work provides a straightforward method for the recombinant production of singlepass transmembrane proteins to allow biochemical and biophysical investigation of the full-length proteins, and will help elucidate the structure and function of full-length FICD.

Construction of S. cerevisiae Expression Plasmids
A cDNA encoding human FICD (Integrated DNA Technologies, Coralville, IA, USA; Gene ID: 11153), codon-optimised for expression in S. cerevisiae, was PCR amplified using Phusion DNA polymerase (Thermo Fisher Scientific, Waltham, MA, USA)) and the primers detailed in Table S1 to generate two constructs: FICD N-terminally tagged with either His 8 -yEGFP [75], including a TEV protease cleavage site (His 8 -GFP-TEV-FICD), or the His 8 tag alone (His 8 -FICD). The plasmids were generated by in vivo homologous recombination in S. cerevisiae by transforming the PAP1500 strain with the appropriate PCR fragments and BamHI/HindIII-digested pEMBLyex4 vector [35,44] (Figure 3A) according to [76]. PAP1500 transformants were selected on synthetic defined medium plates supplemented with 30 mg/L leucine and 20 mg/L lysine [34,77].

Protein Expression in S. cerevisiae
A single colony of transformed yeast cells was propagated initially in synthetic defined medium supplemented with 30 mg/L leucine and 20 mg/L lysine, and then in the same medium supplemented only with lysine [34]. For small-scale expression of His 8 -GFP-TEV-FICD, a leucine-starved culture was used to inoculate 2 L yeast extract-peptone-glycerol medium supplemented with 0.5% (w/v) glucose. The cultivation was continued at 25 • C with shaking until OD 450 reached 1.0. Cultures were thermo-equilibrated to 15 • C and protein expression was induced by adding galactose to a final concentration of 2% (w/v). The large-scale expression of His 8 -FICD was carried out in 15 L Applikon Bioreactors (Getinge, Göteborg, Sweden) equipped with ADI 1030 Bio Controller as described previously [34]. Briefly, 1 L of leucine-starved preculture was used to inoculate 10 L of minimal medium supplemented with 3% glucose (w/v), 3% glycerol (w/v) and all amino acids except leucine and isoleucine. During growth the temperature was maintained at 20 • C, and the pH at 6.0 using automated titration with 20% NH 4 OH. When the glucose was used up, identified as the time point where the rate of NH 4 OH addition started to level off, the bioreactor was supplemented with an additional 2% (w/v) glucose. At the time when the rate of NH 4 OH addition started to level off again, the bioreactor was cooled to 15 • C and protein production was induced by the addition of galactose to a final concentration of 2% (w/v). The cells were harvested 120 h after induction by centrifugation at 1500× g for 10 min at 4 • C. The cells were washed with 0.9% (w/v) NaCl (His 8 -GFP-TEV-FICD) or 1 M glycerol (His 8 -FICD) to ensure isosmotic conditions, and the cell pellets were stored at −80 • C until use.

Detergent Screen
Membranes purified from yeast expressing His 8 -GFP-TEV-FICD were diluted to 2 mg/mL in Lysis Buffer 1 and supplemented with a detergent (see Table 1) at a final concentration of 2, 4 or 6 mg/mL. When indicated, CHS was included to one third of the detergent concentration. The mixtures were incubated at 4 • C for 1 h with gentle agitation. Insoluble material was collected by centrifugation at 160,000× g for 30 min at 4 • C. The solubilisation efficiency was estimated by measuring the GFP fluorescence (Fluoroskan Ascent, Thermo Fisher Scientific, Waltham, MA, USA; excitation at 485 nm and emission at 520 nm) in the soluble fraction before and after the centrifugation [32]. For FSEC, 350 µL of the solubilised membrane proteins were separated on a Superose 6 Increase 10/300 GL (Cytiva, Marlborough, MA, USA) column connected to an ÄKTA purifier (Cytiva, Marlborough, MA, USA) in SEC Buffer 1 (25 mM Tris-HCl pH 7.6, 500 mM NaCl, 10% glycerol, 0.3 mg/mL DDM). An in-line fluorescence detector (Shimadzu RF-20A prominence, Shimadzu Corporation, Kyoto, Japan) was used to record the elution of the GFP-tagged FICD. All detergents used for detergent screening were purchased from Affymetrix (Santa Clara, CA, USA) or GlyconBiochemicals (Luckenwalde, Germany) and are listed in Table 1.

Whole-Cell Fluorescence and Live Cell Bioimaging
Whole-cell fluorescence was measured in 1 OD 450 unit of cells resuspended in 200 µL of 0.9% (w/v) NaCl using a Fluoroskan Ascent microplate reader (Thermo Fisher Scientific, Waltham, MA, USA) with 485 nm excitation and 520 nm emission wavelengths. For bioimaging, 100 µL of S. cerevisiae expressing His 8 -GFP-TEV-FICD were harvested, washed in 0.9% NaCl and resuspended in 5 µL of 0.9% NaCl. In total, 1 µL was pipetted onto a glass slide (Delta Lab, Barcelona, Spain) and whole-cell GFP fluorescence was imaged at 1000 times magnification using a Nikon Eclipse E600 microscope (Nikon, Tokyo, Japan) coupled to an Optronics Magnafire model S99802 camera (Hensoldt, Taufkirchen, Germany). GFP-tagged versions of the endogenous Pmr1 [51] and Pmc1 [52] proteins were used to assess the localisation of recombinant GFP-tagged FICD. The localisation of Pmc1-GFP and Pmr1-GFP was established previously [35,50].

Quantification of His 8 -GFP-TEV-FICD Expression
The total protein concentration in the crude membranes was determined by the Lowry assay [79] as described previously [32]. To estimate the amount of FICD in the crude membranes, the GFP fluorescence was converted to pmol/mg membrane protein using a standard curve generated from purified GFP mixed with yeast membranes as previously established [32,35]. The fluorescence intensity was measured in the crude membranes, and converted into pmol/mg crude membranes using the equation: where RFU denotes relative fluorescence units.

SDS-PAGE, in-Gel Fluorescence, and Immunoblotting
Proteins were separated on 10% SDS-PAGE gels and visualised by Coomassie staining or in-gel fluorescence on a ChemiDoc MP imaging system (Bio-Rad, Hercules, CA, USA) using excitation at 492 nm and emission at 510 nm. For immunoblotting, the proteins were transferred onto a polyvinylidene fluoride membrane (Immobilon-P, Merck Millipore, Darmstadt, Germany) in a Mini Trans-Blot (Bio-Rad) transfer system. A polyclonal rabbit antiserum (BioGenes, Berlin, Germany) raised against the entire ER-luminal region (residues 45-458) of human FICD produced in E. coli was used in combination with a horseradish peroxidase-conjugated α-rabbit secondary antibody (Pierce, Thermo Fisher Scientific, Waltham, MA, USA). The primary antibody was diluted 1:1000 in 2% skim milk dissolved in Tris-buffered saline with 0.1% Tween ® 20 detergent (TBST) before incubation of the membrane overnight at 4 • C. Secondary antibody was diluted 1:100,000 in 2% skim milk-TBST and incubated with the membrane for 50 min at room temperature before thorough washing with TBST and a final wash with TBS. Chemiluminescence detection was performed using ECL Select Peroxide and Luminol solutions (Cytiva, Marlborough, MA, USA) according to the manufacturer's directions.

Purification of Soluble Proteins
The soluble proteins were purified using an ÄKTA Start chromatography system (Cytiva, Marlborough, MA, USA) mounted with a 5 mL HiFliQ Ni-nitrilo triacetic acid column (ProteinArk, Sheffield, UK) pre-equilibrated with Lysis Buffer 3. The column was washed with 15 CV of Lysis Buffer 3 containing 25 mM imidazole, and eluted with an imidazole gradient from 25 to 400 mM over 20 CV. For His 6 -BiP ∆18 T229A V461F the histidine tag was cleaved using TEV protease and removed by affinity purification on a Ni-NTA column [88]. The purified proteins were dialysed into Storage Buffer (50 mM NaH 2 PO 4 pH 7.4, 300 mM NaCl) and stored at 4 • C.

DeAMPylation Assay
To investigate His 8 -FICD deAMPylation activity, 2.8 µg AMPylated BiP ∆18 T229A V461F (a mixture of 2.52 µg non-radioactively AMPylated BiP and 0.28 µg radioactively AMPylated BiP) was incubated in deAMPylation Buffer (25 mM HEPES-KOH pH 7.4, 100 mM KCl, 5 mM MnCl 2 , 2 mM DTT, 0.2 mg/mL DDM) in the presence or absence of 0.26 µg His 8 -FICD in a final reaction volume of 15 µL. The negative control was used to ensure that the loss of AMP signal was not due to spontaneous hydrolysis or deAMPylation by Ub-His 10 -FICD ∆102 L258D [10]. Samples were incubated at room temperature and quenched at specified time points by boiling in reducing SDS-PAGE loading buffer. Proteins were separated by SDS-PAGE and analysed by Coomassie staining and phosphorimaging using a Typhoon FLA 700 phosphorimager (Cytiva, Marlborough, MA, USA). Protein band intensities were quantified using Fiji [89]. The radioactivity signal was related to the Coomassie signal, and the resulting values were normalised using the equation: y norm = (y − y min )/(y max − y min ), where y norm is the normalised value, y max the maximum value of the dataset and y min is assumed to be 0. The standard error of the mean (SEM) was calculated using the equation: where SD is the standard deviation and n is the number of experiments.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.