Engineering and purification of a thermostable, high-yield, variant of PfCRT, the Plasmodium falciparum chloroquine resistance transporter

Protein Expression and Purification 141 (2018) 7e18

Contents lists available at ScienceDirect

Protein Expression and Purification journal homepage: www.elsevier.com/locate/yprep

Engineering and purification of a thermostable, high-yield, variant of PfCRT, the Plasmodium falciparum chloroquine resistance transporter David J. Wright, Marc O'Reilly, Dominic Tisi* Astex Pharmaceuticals, 436 Cambridge Science Park, Milton Road, Cambridge CB4 0QA, UK

a r t i c l e i n f o

a b s t r a c t

Article history: Received 18 March 2017 Received in revised form 16 June 2017 Accepted 10 August 2017 Available online 17 August 2017

Historically chloroquine was used to treat the most deadly form of malaria, caused by the parasite Plasmodium falciparum. The selective pressure of chloroquine therapy led to the rapid emergence of chloroquine resistant parasites. Resistance has been attributed to the Plasmodium falciparum Chloroquine Resistance Transporter (PfCRT), an integral membrane protein of unknown structure. A PfCRT structure would provide new insights into how the protein confers chloroquine resistance and thereby also yield novel opportunities for developing anti-malarial therapies. Although PfCRT is an attractive target for characterisation and structure determination, very little work has been published on its expression and purification. Here we present a medium throughput protocol, employing Sf9 insect cells, for testing the expression, stability and purification yield of rationally designed PfCRT mutant constructs and constructs of a PfCRT orthologue from Neospora caninum (NcCRT). We have identified a conserved cysteine residue in PfCRT that results in elevated protein stability when mutated. Combining this mutation with the insertion of T4-lysozyme into a specific surface loop further augments PfCRT protein yield and thermostability. Screening also identified an NcCRT construct with an elevated purification yield. Furthermore it was possible to purify both PfCRT and NcCRT constructs at milligram-scales, with high purities and with size exclusion chromatography profiles that were consistent with monodispersed, homogeneous protein. Crown Copyright © 2017 Published by Elsevier Inc. All rights reserved.

Keywords: PfCRT Membrane protein Membrane transport Recombinant protein expression Protein purification Construct screening Protein stability Drug resistance Malaria Transporter Plasmodium falciparum Neospora canium NcCRT

1. Introduction Malaria is a major threat to worldwide health and was responsible for over 400,000 deaths in 2015 alone [1]. Although numerous anti-malarial treatments have been developed, none of these have been successful in eradicating this disease. Chloroquine is one such treatment and is thought to act by preventing the polymerisation and detoxification of haem degradation products [2]. Chloroquine initially provided a highly effective treatment for malaria; however, parasitic strains less sensitive to chloroquine quickly emerged. It was shown that the resistant strains contained mutations in the gene pfcrt [3]. The product of this gene, PfCRT, resides in the membrane of the parasite's digestive vacuole and, when mutated, results in reduced accumulation of chloroquine in this subcellular compartment [4]. Heterologously expressed PfCRT has been shown to transport chloroquine and other antimalarial drugs [5e7], but

* Corresponding author. E-mail address: [email protected] (D. Tisi). http://dx.doi.org/10.1016/j.pep.2017.08.005 1046-5928/Crown Copyright © 2017 Published by Elsevier Inc. All rights reserved.

the native substrate for this transporter remains unknown. PfCRT has broader clinical relevance as malarial strains harbouring a knockout of this gene are non-viable [8], suggesting that even the wild type protein might be a putative drug target. Despite years of study, little is known about the molecular mechanism of PfCRT: in fact, there is still debate as to whether PfCRT acts as a passive or active transporter [9]. Therefore the characterisation and structure of this transporter would be of interest to both academia and the pharmaceutical industry. Integral membrane proteins are typically challenging targets for structure determination. Although it is estimated that 20e30% of all open reading frames (ORFs) code for membrane proteins [10], less than 2% of structures in the Protein Data Bank (PDB) are of membrane proteins [11]. Many factors contribute to this lack of structural knowledge, but the low stability and purification yield of membrane proteins are perhaps the main reasons. Researchers have adopted two main approaches to enhance the likelihood of being able to solve the three-dimensional structure of a membrane protein. The first is to search for orthologous proteins with

8

D.J. Wright et al. / Protein Expression and Purification 141 (2018) 7e18

increased stability, or purification yield, and the second is to engineer the target membrane protein to have enhanced stability. Indeed, it has been shown that alanine scanning mutagenesis coupled with a radioligand binding, thermostability, assay can be used to engineer a GPCR [12e14], or transporter [15e17], to have increased thermostability and that this correlates with improved crystallisation properties. An alternative strategy to increase the probability of membrane protein crystallisation is to increase the hydrophilic surface area available for the formation of crystal lattice contacts. This can be achieved by binding a soluble partner protein, such as an antibody [18] or camelid nanobody [19,20], or via a fusion with a soluble partner protein such as T4-lysozyme. There have been particular successes using N-terminal, or intra-cellular loop 3, T4-lysozyme insertions [21e23] or, more recently, modified apocytochrome B562 insertions [24e26] with GPCRs. Inserted soluble proteins have been shown to form the majority of crystal contacts in some crystals and this method has been particularly helpful when combined with lipidic cubic phase (LCP) crystallisation (reviewed in Ref. [27]). In order to identify effectively membrane protein mutations that improve the biophysical properties of a protein it is critical to have a suitably sensitive assay. Although the thermostability and radioligand binding assay approach has been very successful, it is necessary to have access to a high affinity (nM), radiolabelled, ligand for each protein tested. This can be challenging for membrane protein transporters which typically bind ligands, or substrates, with mM-mM affinities [28e30]. Therefore there is a clear requirement for alternative profiling strategies for membrane proteins for which no high affinity ligand is known and for which a structure determination is sought. One method that has been used to estimate the stability of membrane proteins in the absence of a ligand or substrate is fluorescence-detection size exclusion chromatography (FSEC). This technique measures the fluorescence of a green fluorescent protein (GFP) tagged protein in a lysate or complex mixture [31]. A heating step can be added to use this technique to rank protein variants by apparent thermostability [32]. Additionally, there has been significant success using C-terminally appended GFP tags to indicate if a membrane protein has been correctly processed and inserted into the plasma membrane of yeast, or prokaryotic, cells [33]. It also has been shown that GFP is SDS resistant and remains folded when analysed by SDS-PAGE [34]. Although there is much academic interest in PfCRT, there is very little literature precedence for its purification. At the time of publishing, PfCRT had only been heterologously expressed and purified from Pichia pastoris [7] or as a double fusion protein in E. coli [5]. There are no published protocols for insect cell based purifications and no evidence of construct screening. Given the relatively sparse literature available on PfCRT we also elected to investigate expression and purification of the closest, non-plasmodial, orthologue of PfCRT from Neospora canium (NcCRT). In this work we have employed a medium throughput insect cell expression screen to identify a set of rationally designed PfCRT and NcCRT constructs that have higher purification yield and increased protein thermal stability. We have used a combination of FSEC analysis and measurement of purification yield to further engineer and triage a set of PfCRT and NcCRT constructs for crystallisation screening. Further we show that some of the mutations and truncations are additive in terms of increasing protein purification yield and stability. We have used FSEC to profile the stability of a PfCRT mutant in a range of detergents and have used these data to identify which detergents might be most suitable for further, large-scale, purification of the PfCRT variants. These data were also used to inform the parameters employed during milligram-scale purifications of high quality PfCRT and NcCRT samples that were subsequently

progressed for characterisation.

crystallisation

screening

and

biophysical

2. Results Initial construct design - A full length, wild type, construct for PfCRT (Hb3) was synthesised with a C-terminal thrombin cleavage site, enhanced GFP (eGFP) and deca-histidine tag (Fig. 1 - construct A) and cloned into the pFastBac1 vector. Viruses were produced and PfCRT expression in Sf9 insect cells was evaluated by in-gel GFPfluorescence. Wild type (construct A) protein from 10 ml of Sf9 insect cells could be purified using TALON resin and visualised on a Coomassie stained gel (Fig. 1). Mutation screening- By using the TMHMM [35] transmembrane prediction and RONN disorder prediction [36] servers and an alignment of PfCRT plasmodial orthologues it was possible to design rationally a set of PfCRT constructs to be tested for improved purification yield and thermal stability. The variants selected included deletions, insertions and point mutations (Table 1). These variants were synthesised in the wild type full length PfCRT (Hb3) background, with a C-terminal thrombin cleavage site, enhanced GFP (eGFP) and deca-histidine tag as described previously (Fig. 1 e construct A). This set of mutants involved modification of regions predicted to be disordered, non-transmembrane, loop regions. Of particular interest were the putative loops between transmembrane helices 2 and 3, or 7 and 8. The only disordered intramembrane region predicted by the RONN disorder prediction server [36] was found between helices 2 and 3 (residues 108e122 inclusive with a >0.5 probability of disorder). Sequence analysis indicated that the region between transmembrane helices 7 and 8 should be ordered and that it contained a cys rich motif that is highly conserved across even distant orthologues (Supplementary Table 1). Thus it was deemed interesting to examine the role that this conserved region plays in maintaining protein stability. A possible glycosylation site (Asn88), predicted by the NetNGlyc server [37], was also targeted with the mutation N88D. T4 lysozyme

Fig. 1. Purifications of PfCRT truncations. PfCRT constructs were purified from 10 ml of Sf9 cells and analysed by SDS-PAGE by in-gel fluorescence (left) and Coomassie staining (right). Left of each image: Full length PfCRT followed by a thrombin cleavage site, eGFP and His10 (Fig. 1econstruct A) Right of each image: Same construct with D2-50 and D406-424 (construct B). Lanes for each construct correspond to the flow through, 50 mM imidazole wash and elution in 1M imidazole. Marker positions on the left are BenchMark Fluorescent Protein Standard and right are Amersham ECL FullRange Rainbow Molecular Weight Markers. The positions of “full length” construct are marked with a star (*).

D.J. Wright et al. / Protein Expression and Purification 141 (2018) 7e18

9

Table 1 PfCRT construct summary. Construct number

Construct

Signs that Sf9 cells were infected Purification band intensity (% wild type) FSEC Survival (% ± Standard deviation)

A3 A8 A7 n/a A4 A A9 A2

T4-lysozyme replacing T230-D241 C289/301/309/312A C312A D115-116 C289A Wild type D293-305 T4-lysozyme replacing G113N127 N88D C289S D(115e116, 121e125) C312S C309S C309A D399-424 D286-313 D305-314 C301S D266-293 D295-304 D308-313 T4-lysozyme replacing R294-G304 D266-278 T4-lysozyme replacing N114-N127 Negative control D270-275 D266-314 D403-424 C301A C289/301/309/312S T4-lysozyme replacing A79-S89 T4-lysozyme replacing T149-S157 T4-lysozyme replacing E198-S210 T4-lysozyme replacing F268-T281 T4-lysozyme replacing T281-R294 T4-lysozyme replacing F268-G314

þ þ þ þ þ þ þ þ

350 142 112 109 103 100 100 92

59 67 69 e 55 54 e 59

þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ þ            

89 89 87 84 83 77 69 69 59 44 43 42 42 39 24 21 0 e e e e e e e e e e e

e 57 ± 2 56 ± 8 e e e e e e e e e e e e e e e e e e e e e e e e e

n/a A4S A1 n/a n/a A6 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a

±1 ±0 ±0 ±3 ±7 ±4

Shown is a summary of all initial screening data, including: whether an infection was successful, the purification yield compared to wild type estimated by in-gel fluorescence and the survival ratio. Construct numbers refer to those used in Fig. 2 and are ordered according to purification yield. Constructs in bold were those of at least 85% the yield of wild type PfCRT and so were selected for analysis by Fluorescence-detection Size Exclusion Chromatography (FSEC).

insertions were made in the regions predicted to be loops between consecutive helices [(1&2), (2&3), (3&4), (5&6), (6&7) and (7&8)] (Fig. 3), as each of these loops was predicted to contain at least 10 amino acids that could be replaced. Insertions were also made at multiple positions in loops 2e3 and 7e8 for the above reasons. The initial screen also included a single construct of NcCRT (285e676) (Fig. 2 construct N1) that was designed using the TMHMM and RONN disorder prediction servers and a sequence alignment with PfCRT (Supplementary Table 2). It was hoped that this construct would provide a ‘proof of concept’ as to whether recombinant NcCRT could be expressed in insect cells. Viral production and protein purification - A medium throughput baculovirus production and insect cell expression screen was developed. This involved P0 viral generation by in suspension transfection of Sf9 insect cells in 96-well deep well blocks [39]. This P0 stock was then used to generate the P1 stock in a 24-well deep well block format. P1 virus stocks were used for expression in 50 ml of Sf9 cells in 250 ml flasks. Cell viability was measured with trypan blue [40] and wells that showed no sign of infection were discarded (11 of 35). Small scale protein purification, using the pellet from a 10 ml Sf9 culture, was performed in dodecylmaltoside detergent (DDM) and using TALON resin. The protein eluted from the TALON was analysed by SDS-PAGE (Fig. 4). The majority of the variants for which a viable virus was obtained yielded amounts of purified protein that

could be visualised by in-gel fluorescence or Coomassie stained SDS-PAGE. The NcCRT construct showed the highest purification yield of any construct in this screen (>5-fold higher overall yield than wild type PfCRT) and therefore this construct was fast-tracked for purification optimisation and characterisation (Fig. 4). In addition, some PfCRT variants exhibited significantly higher yields than the wild type protein (Table 1). FSEC analysis was performed on 8 of the 10 PfCRT variants that had purification yields that were at least 85% of wild type levels. The construct containing a deletion of residues 115 and 116 was discarded at this stage as the related construct containing deletion of residues 115, 116 and 121e125 (Fig. 2 - construct A1) resulted in a similar yield and removed a greater number of residues from the loop predicted to be disordered between transmembrane helices 2 and 3. The point mutant N88D was not analysed further, as there was very little change in purification yield and no obvious shift in molecular weight, which would be indicative of a loss of glycosylation, suggesting that this residue was not glycosylated. The constructs that were selected for further profiling were D115-116 þ D121-125 (Fig. 2 - construct A1), T4113-127 (construct A2), T4230-241 (construct A3), C289A (construct A4), C289S (construct A4S), C312A (construct A7), C289/301/309/312A (construct A8) and D293-305 (construct A9). The reproducibility of the virus production and expression protocol was checked by repeating the purification experiments

10

D.J. Wright et al. / Protein Expression and Purification 141 (2018) 7e18

Fig. 2. PfCRT and NcCRT construct summary. The wild type Hb3 PfCRT construct with a C-terminal thrombin cleavage site, GFP and deca-histidine tag (A) is shown in relation to other constructs used in this work. Construct N1 shows the initial NcCRT construct identified. The middle panel contains the equivalent B constructs (D2-50 þ D406-424) used in this work, in addition to the combination mutants. Large scale purification constructs (C for PfCRT and D for NcCRT) are shown in the bottom panel.

several months later using fresh batches of bacmid DNA and Sf9 cells. Experiments were repeated for all 23 of the constructs that originally resulted in a viable virus, except D115e116, C312S, D266278 and a T4 replacement of residues 114e127. Each of the 8 mutants that were chosen in the initial screen again produced over 85% of the yield of the wild type (construct A) (Supplementary Fig. 3). The top expressing mutant from the initial screen, containing a T4 lysozyme replacement of residues 230e241 (construct A3), showed the highest purification yield in the repeat purification screen (in both experiments 350% of construct A). Additionally construct A8 (C289/301/309/312A) also showed reproducible elevated expression levels (140e200% of A). These data therefore suggest that small-scale expression and purification from insect cells is a reliable and reproducible means of identifying yield enhancing mutations of PfCRT constructs.

Fluorescence-detection size exclusion chromatography - FSEC uses the fluorescence of GFP, or another fluorophore, to measure the elution profile of a target without the need for a purified protein sample. DDM was used to solubilise whole cell samples, which were then applied to a 5/150 S200 column and the elution profile was measured at 509 nm by excitation of GFP at 488 nm. The thermal stability of each variant was assessed using a 40  C heating step and calculating the ratio of the fluorescence between the heated sample and an identical sample incubated on ice. All samples underwent ultra-centrifugation to remove any precipitated material prior to loading. A 40  C temperature step was chosen from preliminary work with construct A which showed that 40  C was the temperature at which the height of the eluting peak was reduced to half the maximal peak height observed at 4  C (data not shown). The fluorescence value at the equivalent volume to the 4  C

D.J. Wright et al. / Protein Expression and Purification 141 (2018) 7e18

11

Fig. 3. PfCRT construct design. A brief summary of PfCRT constructs generated for initial work. The position of the 10 predicted transmembrane helices (TM) are shown. In red are the N- and C-terminal truncations that reduce PfCRT degradation and result in construct B. The positions of T4 lysozyme loop replacements are shown. Red boxes show the loops that were targeted with deletions. In blue are residues targeted by point mutations, either the predicted glycosylation site Asn88 (circled in red) between predicted transmembrane helices 1 and 2 or the highly conserved cysteine residues (circled in blue). The bottom of the image is cytoplasmic and the top the digestive vacuole. Figure prepared with TOPO2 [38]. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

maximum was used rather than the 40  C peak maximum, or area under the curve, as this would take in to account any peak shift or broadening due to aggregation. Of the mutants with similar, or greater, purification yield than wild type, five showed thermostabilities indistinguishable from wild type: D(115e116 þ 121e125) (construct A1), T4113-127 (construct A2), T4230-241 (construct A3), C289A (construct A4) and C289S (construct A4S) (Fig. 5 and Supplementary Fig. 2). However the cysteine-less loop mutant containing C289A, C301A, C309A and C312A (construct A8) and a variant containing the C312A mutation alone (construct A7) showed elevated stabilities (approximately 10% greater survival at 40  C p < 0.05). The mutant D293-305 (construct A9) resulted in a very low signal and so was not analysed further (data not shown). Truncation mutagenesis e Purified, full length, PfCRT (Hb3) (construct A) gave three discrete bands on SDS-PAGE. The RONN disorder prediction server [36] predicted that residues 2e50 and 406e424 should be disordered. Deletion of these residues gave construct B (Fig. 1), which ran as a discrete, higher intensity, band on SDS-PAGE. This was attributed to the truncated construct being less susceptible to proteolytic degradation in insect cells (Fig. 1). Combinations of PfCRT mutants- Since several mutations in construct A had led to increases in either purification yield or thermal stability, it was decided to explore if combinations of these mutations would be additive, in terms of enhancing yields and/or thermal stability, and could make the constructs amenable to large scale purification. As construct B (D2-50 þ D406-424) appeared to be less susceptible to proteolytic degradation and gave a discrete band on SDS-PAGE it was decided to introduce the combination mutants into the construct B background. The construct A mutations that increased yield were C289/301/309/312A (construct A8) and T4 lysozyme replacement of residues 230e241 (construct A3). Mutations that increased thermal stability of construct A were C312A (construct A7) or C289/301/309/312A (construct A8). The only loop predicted to have significant disorder was between putative transmembrane helices 2 and 3 (residues 114e125). It was hoped that alterations in this loop might increase the probability of

these constructs crystallising by both increasing the hydrophilic surface area of constructs and reducing the number of disordered residues. Although deletion of residues 115e116 and 121e125 (construct A1) or a T4 lysozyme insertion for residues 113e127 (construct A2) did not significantly affect either the yield or thermal stability of the PfCRT it was decided to include these loop mutations in the combination screen (Figs. 2 and 6). Although in the construct A background the T4 lysozyme replacement of residues 230e241, between predicted transmembrane helices 6 and 7, (construct A3) resulted in an elevated purification yield (350%), in the construct B background the same replacement (B3) resulted in a much lower yield (<40%) compared to construct B. Similarly T4 lysozyme replacement of residues 113e127, between predicted transmembrane helices 2 and 3 (construct B2), also resulted in a lower yield than construct B. Additionally, when both of these T4 lysozyme loop replacements were combined in a single construct, the purification yield was much lower (<50%) than with construct B. The cysteine to alanine point mutations (C312A or C289/301/309/312A) previously shown to increase PfCRT stability in constructs A7 and A8 were also added to the T4 lysozyme loop replacement mutants. Combining the C312A mutation (construct B7) with a T4-lysozyme replacement of residues (230e241) (construct B3), to give construct (B3þB7), resulted in 150% of the yield of construct B. Similarly combining the C289/301/309/312A mutant (construct B8) with the T4-lysozyme construct B3, to give construct (B3þB8), also resulted in 150% of the yield of construct B. Thus introduction of the B7 and B8 mutations effectively rescued the expression of the B3 construct. The cysteine substitution mutants B7 or B8 were, however, unable to rescue the poor yield of the double T4 lysozyme replacement construct (construct B2þB3). The triple combination constructs (B2þB3þB7) or (B2þB3þB8) both had less than 50% of the yield of construct B. Since a T4 lysozyme replacement of the residues in the loop between predicted transmembrane helices 2 and 3 did not enhance protein yield or stability it was decided to investigate the impact of

12

D.J. Wright et al. / Protein Expression and Purification 141 (2018) 7e18

Fig. 4. Purifications of initial constructs. PfCRT and NcCRT constructs were purified from 10 ml of Sf9 cells and analysed by SDS-PAGE by in-gel fluorescence (top) and Coomassie staining (bottom). Red arrows indicate the location of full length PfCRTHb3-Th-eGFP-His10 and stars indicate the location of NcCRT on each image. Red bands are over-exposed. Gel 1 1. Wild type (Fig. 2 - construct A); 2. D115-116; 3. D(115e116 þ 121e125) (construct A1); 4. D266-278; 5. D266-293; 6. D295-304; 7. D305-314; 8. D286-313; 9. D293-305 (A9); 10. D308-313. Gel 2 - 1. Wild type (A); 2. D399-424; 3. N88D; 4. C289A (A4); 5. C309A (A6); 6. C312A (A7); 7. C289/301/309/C312A (A8) 8. C289S (A4S); 9. C301S; 10. C309S; 11. C312S. Gel 3 - 1. Wild type (A); 2. T4 replacing 113e127 (A2); 3. T4 replacing 230e241 (A3); 4. T4 replacing 294e304; 5. T4 replacing 114e127; 6. NcCRT(285e676 construct N1). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

the loop deletion mutant (D(115e116 þ 121e125), present in constructs A1 or B1) on the (B3þB7) or (B3þB8) combination mutant constructs. The resulting, triple mutant, constructs (constructs B1þB3þB7 or B1þB3þB8) were purified, but resulted in yields that were comparable with construct B and slightly lower than those obtained for the B3þB7 or B3þB8 double mutations alone. Each of the above constructs that could be purified with a yield similar to, or higher than, construct B (D2-50 þ D406-424) were then analysed by FSEC and the thermal stabilities compared to construct B. In comparison with construct B, additional T4 lysozyme replacement of residues 230e241 (construct B3) resulted in a lower maximal signal, but the survival percentage was increased by approximately 20%. Combining construct B3 with either the C312A single point mutant, to give construct (B3 þ B7) or with the tetracysteine substitution mutant (C289/301/309/312A), to give

construct (B3 þ B8), resulted in constructs that combined both increased purification yield and thermal stability (150% yield relative to construct B and thermal stability equivalent to construct B3). The triple mutants (B1þB3þB7) or (B1þB3þB8) had FSEC profiles essentially unaltered from constructs B3þB7 and B3þB8. Detergent screen with construct B (D2-50 þ D406-424) PfCRTFSEC and small-scale purification were used to test whether alternative detergents might be more suitable for large-scale purification of the PfCRT variants (Fig. 7). Construct B (D2-50 þ D406-424) was purified using a panel of detergents. It proved possible to isolate construct B in each of the detergents tested (see materials & methods); however the yield in octyl glucoside neopentyl glycol (OGNG) was very low. Every detergent resulted in a lower yield than DDM except for a combination of DDM and cholesterol hemisuccinate (CHS). In addition, FSEC studies suggested that the

D.J. Wright et al. / Protein Expression and Purification 141 (2018) 7e18

Fig. 5. FSEC survival of PfCRT initial constructs. PfCRT constructs were solubilised in DDM and aliquots heated to 40  C, or incubated on ice, and analysed by FSEC. The elution volume corresponding to the maximum peak height at 4  C was compared to the peak height observed at the identical elution volume at 40  C and expressed as a percentage. Each mutant was analysed twice and compared to wild type (Fig. 2 construct A) in triplicate using GraphPad Prism 6. Standard deviations are shown. For representative trace data see Supplementary Fig. 1. Each mutation was made in the construct A background (Fig. 2).

thermal stability of PfCRT in each detergent was similar to, or lower than, that in DDM except for the DDM/CHS combination. Construct B purified in DDM/CHS exhibited significantly higher survival (approximately þ25% p < 0.05) and a narrower peak elution profile. Large scale purification of PfCRT and NcCRT e PfCRT and NcCRT constructs were re-cloned without GFP tags to give constructs C and D respectively. The GFP-free NcCRT construct contained additional mutations that were introduced in an attempt to further enhance protein homogeneity; 20 amino acids were truncated from the N-terminus, and 6 from the C-terminus, of construct N1 and an N555Q mutation was also introduced to remove a predicted glycosylation site (Fig. S3). Constructs C and D could be purified on a milligram-scale from 12 L of insect cells (Fig. 8). The purifications employed a protocol adapted from the small-scale, high throughput, purification protocol (see materials and methods). Samples were solubilised using DDM and CHS. Cobalt affinity resin purification was followed by desalt and SEC. Coomassie stained gels and SEC analysis of the purified samples show that the proteins have high purity and run close to their expected elution volumes. Intact mass spectrometry confirmed that both the PfCRT and NcCRT proteins had the correct mass and that they do not carry any posttranslational modifications. Thus we have demonstrated that it is possible to obtain sufficient quantities of the PfCRT and NcCRT proteins to enable further biophysical characterisation and/or crystallisation screening.

3. Discussion This screen has revealed several mutants of PfCRT that show improved purification yield and thermostability and which may be more amenable to large scale purification, characterisation and crystallisation screening. 34 constructs were profiled and of these 11 did not produce infectious virus using standard Bac-to-Bac protocols, even when repeat transfections were performed (data not shown). The inability to produce a virus does not prove that it is impossible to express these mutants in insect cells; however it is possible that many of these mutants may be inherently unstable or mis-folded and hence may not be tolerated in insect cells. For example, 6 of the 11 failed mutants were designed to contain T4 lysozyme replacements of 11e47 residue spans that were predicted

13

to be loops between transmembrane helices. If these transmembrane helices have not been accurately predicted, or are not adjacent in the three dimensional structure, then it is possible that these insertion mutants will not fold correctly and will result in toxicity upon over-expression. The fact that it was possible to express and purify even two of the T4 lysozyme insertions was surprising, as successful T4 lysozyme fusion has only been demonstrated with GPCRs and to some extent with lactose permease [41]. These data therefore suggest that fusion with T4 lysozyme, or another soluble partner protein, may represent a generic method for increasing the purification yield and crystallisation probability of a wide range of membrane protein families. The data from this screen indicate that, although the majority of mutations had very little impact on the purification yield of PfCRT, there were several constructs that showed significantly altered yields. This suggests that this screening methodology represents an effective way to distinguish between mutations that increase the yield of a membrane protein target from those that do not. It is likely that a screen of several hundred mutants would be needed to identify more than one mutant with both elevated yield and thermal stability. For instance, if these screening methodologies were combined with alanine scanning mutagenesis then this technique could be used analogously to the thermostability screening used successfully with GPCR radio-ligand binding assays. Many modifications, including point mutations, deletions and insertions were made in the region between predicted transmembrane helices 7&8, which, unlike many intra-helical regions, was predicted to be ordered. Deletions in this region appear to be detrimental to protein expression and stability, as purification was not possible. In the initial screen only one of four insertions in this region produced a visible band on a Coomassie stained gel and this band had less than half the intensity of the wild type. However, the highly conserved cysteine residues within this region are apparently not required for stability, as mutating all four to alanine resulted in improved recombinant protein yield and thermal stability. The C312A mutation appears to be dominantly responsible for the increase in stability, as there was no notable difference in stability between the C312A mutant alone and the tetra-mutant C289/301/309/312A. Glycosylation does not appear to play a role in the correct folding of PfCRT in insect cells, as the only predicted glycosylation site, Asn88, could be mutated (N88D) without any loss in purification yield or any change in apparent molecular weight. Although the T4 lysozyme replacement of residues 230e241 between predicted transmembrane helices 6 and 7 reproducibly resulted in a large increase in purification yield in the full length background (construct A3), we did not observe the same increase when this variant was introduced into the D2-50 þ D406-424 background (construct B3). This suggests that both these mutations may either increase expression or reduce proteolytic degradation of the protein, but in a non-additive fashion. When the stabilising mutations C312A or C289/301/309/312A were added into the construct B3 background, to give combination mutants (B3þB7) or (B3þB8), we observed an increase in purification yield, although the C312A mutation appears to be dominant. In addition these combinations were more stable than the B construct and could be further combined with the loop deletion mutant, D(115e116 þ 121e125) to give B1þB3þB7 and B1þB3þB8, with essentially the same stability. The FSEC methodology was also successfully used to instruct our choice of the most suitable preliminary purification detergent. It was possible to purify PfCRT (construct B) in a range of single detergents and the data suggest that DDM was the most suitable for the purification of this construct in terms of overall stability and purification yield. The addition of cholesterol hemiscuccinate (CHS)

14

D.J. Wright et al. / Protein Expression and Purification 141 (2018) 7e18

Fig. 6. FSEC profiles and purification of PfCRT mutant combination constructs. Top: FSEC profiles of mutant combinations in the construct PfCRTHb3-Th-eGFP-His10 after incubation on ice or at 40  C. Arrows signify the calculated void volume for the S200 5/150 column. Construct B is D2-50 þ D406-424 (Fig. 2); B3 is the same with an additional T4 lysozyme replacing residues 230e241, B3þB7 contains the T4 insertion and C312A point mutation; B3þB8 contains the T4 insertion and C289/301/309/312A point mutations; B1þB3þB7 contains the T4 insertion, C312A point mutation and D(115e116 þ 121e125) deletion; and B1þB3þB8 contains the T4 insertion, C289/301/309/312A point mutations and D(115e116 þ 121e125) deletion. For clarity curves of each construct have been offset by 1 ml and 10 mV. Bottom left: survival of combinations, SEM is shown of two repeats of all but construct B, where 3 repeats were performed. Bottom centre: stained gel of purifications of PfCRT variants e 1: Construct B, 2: B3, 3: B3þB7, 4: B3þB8, 5: B1þB3þB7, 6: B1þB4þB8, 7: B2þB3 (containing a dual T4 loop replacement at 113e127 and 230e241), 8: B2þB3þB7 (as previous plus C312A) 9: B2þB3þB8 (dual T4 replacement plus C289/301/ 309/312A). Bottom right: same gel as bottom centre but imaged with in-gel fluorescence. Supplementary Table 3 summarises the above data.

further increased the stability of PfCRT. CHS has been shown to stabilise a large range of membrane proteins either by specific interactions or, more generally, by increasing micelle size [42] and hence providing more shielding to the transmembrane helices from aqueous solvent. High throughput purification was also successful in identifying a high-yield construct of the Neospora canium orthologue of PfCRT. We propose that a combination of FSEC and medium throughput purification represents a useful way to characterise protein variants and orthologues and to provide a method by which purification parameters can be optimised. To the authors' knowledge this is the first published work where constructs of PfCRT or any orthologue have been screened for increased purification yield and/or thermal stability. This is the first published example of the heterologous

expression and purification of PfCRT or NcCRT on a milligram-scale that is compatible with crystallisation screening.

4. Experimental procedures Cloning- Full length codon optimised PfCRT (Hb3) followed by a thrombin cleavage site, enhanced GFP (eGFP) [43] and a decahistidine tag was synthesised by Genscript in pFastBac1 between EcoRI and NotI sites. All variants and orthologue constructs (between restriction sites NdeI and NotI), except the double PfCRT truncation D2-50 þ D406-424, N-terminal truncation of NcCRT and point mutation N555Q of NcRT were also synthesised by Genscript. These mutants were generated using degenerate primers lacking the regions to be omitted or with the desired point mutation and,

D.J. Wright et al. / Protein Expression and Purification 141 (2018) 7e18

15

Fig. 7. FSEC profiles and purification of construct B (D2-50 þ D406e424) PfCRT in a range of detergents. Top: FSEC profiles of PfCRT (Fig. 2 - construct B) solubilised in different detergents after incubation on ice or at 40  C. An arrow shows the calculated column void volume. Bottom left: survival in each detergent, calculated from peak height at 4  C maximum, SEM is shown of three repeats in DDM and DDM/CHS only. For clarity curves in each detergent have been offset by 1 ml and 10 mV. Bottom centre: stained gel of purifications of PfCRT in different detergents e 1: dodecylmaltoside (DDM) 2: decylmaltoside (DM) 3: lauryl maltose neopentyl glycol (LMNG) 4: decyl maltose neopentyl glycol (DMNG) 5: octyl glucoside neopentyl glycol (OGNG) 6: cymal-5 7: foscholine-12 8: DDM/CHS (10:1). Bottom right: same gel viewed with in-gel fluorescence.

following PCR using KOD Hot Start Polymerase (Merck Millipore), the templates were digested with DpnI (New England Biolabs) and transformed into XL1 Blue supercompetent cells (Agilent). Constructs in the absence of GFP were cloned in house and ligated between the restriction sites EcoRI and XhoI in pFastBac1. Each construct produced in house was then verified by sequencing by Beckman Coulter Genomics. Virus production- Bacmids were prepared according to Bac-toBac methodologies (Life technologies). Transfections were then performed in suspension in a 96-well format. In order to improve efficiency, two transfections per construct were performed. 2 or 4 ml of midiprep bacmid DNA (Qiagen plasmid plus kit) or no DNA for the negative control was incubated with 2 ml of Cellfectin-II reagent (Life Technologies) in 40 ml of Sf900 II SFX medium (Thermofisher Scientific) for 30 min at room temperature in a sterile Nunc 96 deep well block. 500 ml of Sf9 cells at 5  105/ml were then added and the

mixture was incubated at 27  C for 5 h shaking at 300 rpm on a Rotamax 120 shaker (Heidolph). 500 ml of fresh media was then added and plates were incubated for 1 week at 27  C at 300 rpm. Transfections were harvested by centrifugation at 4000 rpm for 5 min (Megafuge 1.0R - Heraeus 2704 rotor) and 30 ml of each supernatant (from 2 transfections) was added to 3 ml of Sf9 cells at 1.0  106/ml in sterile 24-well blocks (Axygen) and incubated for 1 week at 27  C at 300 rpm. Infections were harvested by centrifugation at 4000 rpm for 5 min and 500 ml of each supernatant was added to 50 ml of Sf9 cells at 2.0  106/ml in 250 ml Erlenmeyer flasks (Corning) and grown for 48 h at 27  C and 90 rpm in an Innova 4230 incubator. Viability was measured using trypan blue solution (Thermofisher Scientific), two 10 ml aliquots of these infected cells were harvested by centrifugation in 15 ml centrifuge tubes (Thermofisher Scientific) at 4000 rpm for 15 min and the pellets were frozen.

16

D.J. Wright et al. / Protein Expression and Purification 141 (2018) 7e18

Fig. 8. PfCRT (top panel: construct C) and NcCRT (bottom panel: construct D) constructs can be purified to give samples that have a ‘mono-dispersed’ SEC profile. Left e S200 16/ 600 SEC traces of samples purified by cobalt affinity chromatography. The positions of molecular weight standards are shown below and the void volume above. Right e Coomassie stained gels of SEC fractions with molecular weight markers shown on the right. Top - lane 1 is the concentrated sample post affinity chromatography prior to loading on the gel filtration column, 2 and 3 are samples from the void peak, 4e10 are samples from the main peak and 11 is the protein standard ladder. Bottom - lane 1 is the concentrated sample, 2 is the protein standard ladder, 3 is a sample from the void peak and 4e11 are samples from the main peak.

Purification screening- Small-scale purification of each variant in a single detergent was used as an indicator for large scale purification yield. A 10 ml aliquot was thawed and solubilised in 1 ml of solubilisation buffer (1% DDM (Generon), 500 mM NaCl, 20 mM Tris pH 7.5) by rotation for 1 h at 15 rpm and 4  C. Insoluble material was separated by centrifugation in a TLA 100.3 rotor in an Optima Max Ultracentrifuge (Beckman Coulter) at 80,000 rpm for 30 min at 4  C. The supernatant was incubated with 100 ml bed volume of TALON resin (Clontech) pre-equilibrated in solubilisation buffer for 2 h rotating at 15 rpm and 4  C. This was then applied to a Pierce microfuge spin column and spun at 5000 rpm at 4  C for 2 min in an Eppendorf F45-3-11 rotor in an Eppendorf 5417R centrifuge. The beads were washed twice with 800 ml of 500 mM NaCl, 20 mM Tris pH 7.5 containing 0.03% DDM, twice with 800 ml of 50 mM Imidazole, 500 mM NaCl, 20 mM Tris pH 7.5 containing 0.03% DDM with the flow through being discarded each time. Samples were spun at 5000 rpm at 4  C for 2 min to remove any residual wash buffer, before 100 ml of 1 M Imidazole, 500 mM NaCl, 20 mM Tris pH 7.5 containing 0.03% DDM was added and incubated for 2 min on ice. The protein was eluted by centrifugation at 5000 rpm at 4  C for 2 min. A 20 ml sample was added to 20 ml 2 x SDS-PAGE loading buffer (Novex) and 15 ml was applied to a 4e20% Tris-glycine gel (Novex) without first heating the sample. Gels were run and then visualised by in-gel fluorescence (Gel Doc EZ reader e Blue plate)

and Coomassie staining (Gel Doc EZ reader e White plate). Constructs were ranked, using in-gel fluorescence, by measuring the intensity (Image Lab software) of the dominant PfCRT band corresponding to the correct molecular weight. Detergent screening by purification- This purification protocol was identical to that above; however purifications used different detergents: solubilisation used 1% detergent and all subsequent steps used a concentration of 3 x critical micelle concentration (CMC). This was the case for all detergents except DDM/CHS, where DDM was at 1% or 3 x CMC and cholesteryl hemisuccinate (CHS) was at 1/10 the percentage of DDM. All detergents were purchased from Generon and CHS was purchased from Sigma Aldrich. Fluorescence-detection size exclusion chromatography- A 10 ml aliquot was thawed, resuspended in 500 ml of 500 mM NaCl, 20 mM Tris pH 7.5 and sonicated briefly. This was then solubilised by addition of 500 ml of 2% DDM, 500 mM NaCl, 20 mM Tris pH 7.5 and rotation for 1 h at 15 rpm and 4  C. Insoluble material was pelleted by centrifugation in a TLA 100.3 rotor in an Optima Max Ultracentrifuge (Beckman Coulter) at 80,000 rpm for 30 min at 4  C. Each supernatant was added to 10 ml of 0.03% DDM 500 mM NaCl, 20 mM Tris pH 7.5 and 1 ml aliquots were heated at 40  C for 30 min in a dry bath FB15103 heat block (Fisher Scientific) or incubated on ice. Precipitated material was separated by centrifugation in a TLA 100.3 rotor in an Optima Max Ultracentrifuge at 80,000 rpm for

D.J. Wright et al. / Protein Expression and Purification 141 (2018) 7e18

30 min at 4  C and the supernatants were loaded in a 50 ml loop, € which was then injected using an AKTA Pure (GE Healthcare) on a Superdex S200 Increase 5/150 GL column (GE healthcare) preequilibrated with 0.03% DDM, 500 mM NaCl, 20 mM Tris pH 7.5. Fluorescence with excitation at 488 nm and emission at 509 nm was measured with a VWR Hitachi Chromaster 5440 FL Detector. The maximum peak height of the sample left at 4  C was measured and the fluorescence signal at the equivalent volume for 40  C was measured. The ratio of fluorescence at 4  C and 40  C was used to calculate the % survival, which was performed in duplicate and compared to wild type in triplicate using an unpaired T test (GraphPad Prism). For FSEC in alternative detergents 1% of the selected detergent (or 1% DDM þ 0.1% CHS) was used for solubilisation: these were then treated identically to the DDM samples shown; however only 200 ml of the supernatant was added to 2 ml of 500 mM NaCl, 20 mM Tris pH 7.5 plus detergent at 3 x CMC. 1 ml samples were heated and centrifuged as above, which were then loaded on a Superdex S200 Increase 5/150 GL column (GE healthcare) pre-equilibrated with 0.03% DDM, 500 mM NaCl, 20 mM Tris pH 7.5. Large scale expression and purification of PfCRT and NcCRT- Sf9 cells at 2.0  106 cells/ml were infected with 500 ml/L high titer virus and grown at 27  C for 72 h. Cells were harvested by centrifugation in a JLA-8.1000 rotor (Beckman Coulter) at 5000 rpm for 15 min at 4  C and frozen. Purification was typically performed with batches of 6 L of cells. Pellets were thawed and resuspended in 225 ml of 555 mM NaCl, 111 mM Tris pH 7.5, and 11% glycerol containing protease inhibitors (Calbiochem) and DNaseI (Roche). After sonication, 25 ml of 10% DDM (Generon) and 1% CHS (Generon) was added and the cells were solubilised by rotation for 1 h at 15 rpm at 4  C. Insoluble material was separated by centrifugation in a Ti45 rotor in an Optima XE centrifuge (Beckman Coulter) at 35,000 rpm for 1 h at 4  C. The supernatant was incubated with 5 ml bed volume of pre-washed TALON resin (Clontech) for 16 h rotating at 15 rpm and 4  C. This was then applied to a Biorad Econocolumn and the flow through collected. The column was washed with 250 ml of 500 mM NaCl, 100 mM Tris pH 7.5 and 10% glycerol containing 0.03% DDM and 0.003% CHS followed by 250 ml of 10 mM Imidazole, 500 mM NaCl, 100 mM Tris pH 7.5 and 10% glycerol containing 0.03% DDM and 0.03% CHS. The protein was eluted by addition of 500 mM Imidazole, 500 mM NaCl, 100 mM Tris pH 7.5 and 10% glycerol containing 0.03% DDM and 0.003% CHS, incubation for 5 min and collection of the subsequent flow through. The eluate was then applied to a desalt column pre-equilibrated in 500 mM NaCl, 100 mM Tris pH 7.5 and 10% glycerol containing € 0.03% DDM and 0.003% CHS using an AKTA Pure and the peak fractions were pooled. These were then concentrated with a 15 ml 100 kDa cutoff concentrator (Corning Spin X -UF) to approximately 3 ml. This sample was then loaded on an S200 16/60 preequilibrated in 500 mM NaCl, 100 mM Tris pH 7.5 and 10% glycerol containing 0.03% DDM and 0.003% CHS and the fractions containing purified PfCRT or NcCRT were pooled and further concentrated as required.

Acknowledgements The authors would like to thank Andrew Quigley and Jackie Ang at the SGC for performing intact mass analysis. We would also wish to thank Liz Carpenter of the SGC Oxford for useful scientific discussion during this work and feedback on the manuscript when in preparation. DJW would also like to thank Chris Tate of the MRC LMB for his continued guidance. DJW is wholly funded by Astex Pharmaceuticals as a Sustaining Innovation Postdoctoral Training Fellow.

17

Author contributions DJW conceived the idea for the project, performed all of the experiments and prepared the manuscript with guidance from DT and MOR. Conflict of interest The authors declare that they have no conflicts of interest with the contents of this article. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.pep.2017.08.005. References [1] WHO, World Malaria Report 2015, 2015. [2] A. Dorn, S.R. Vippagunta, H. Matile, C. Jaquet, J.L. Vennerstrom, R.G. Ridley, An assessment of drug-haematin binding as a mechanism for inhibition of haematin polymerisation by quinoline antimalarials, Biochem. Pharmacol. 55 (1998) 727e736. [3] D.A. Fidock, T. Nomura, A.K. Talley, R.A. Cooper, S.M. Dzekunov, M.T. Ferdig, , K.W. Deitsch, X.-z. Su, J.C. Wootton, L.M.B. Ursos, A. bir Singh Sidhu, B. Naude P.D. Roepe, T.E. Wellems, Mutations in the P. falciparum digestive vacuole transmembrane protein PfCRT and evidence for their role in chloroquine resistance, Mol. Cell 6 (2000) 861e871. [4] A.B.S. Sidhu, D. Verdier-Pinard, D.A. Fidock, Chloroquine resistance in Plasmodium falciparum malaria parasites conferred by pfcrt mutations, Science 298 (2002) 210e213. [5] N. Juge, S. Moriyama, T. Miyaji, M. Kawakami, H. Iwai, T. Fukui, N. Nelson, H. Omote, Y. Moriyama, Plasmodium falciparum chloroquine resistance transporter is a H(þ)-coupled polyspecific nutrient and drug exporter, Proc. Natl. Acad. Sci. U. S. A. 112 (2015) 3356e3361. , J.A. Brzostowski, A.R. Kimmel, T.E. Wellems, Dictyostelium Dis[6] B. Naude coideum expresses a malaria chloroquine resistance mechanism upon transfection with mutant, but not wild-type, Plasmodium falciparum transporter PfCRT, J. Biol. Chem. 280 (2005) 25596e25603. [7] W. Tan, D.M. Gou, E. Tai, Y.Z. Zhao, L.M.C. Chow, Functional reconstitution of purified chloroquine resistance membrane transporter expressed in yeast, Arch. Biochem. Biophys. 452 (2006) 119e128. [8] K.L. Waller, R.A. Muhle, L.M. Ursos, P. Horrocks, D. Verdier-Pinard, A.B.S. Sidhu, H. Fujioka, P.D. Roepe, D.A. Fidock, Chloroquine resistance modulated in vitro by expression levels of the Plasmodium falciparum chloroquine resistance transporter, J. Biol. Chem. 278 (2003) 33593e33601. [9] C.P. Sanchez, W.D. Stein, M. Lanzer, Is PfCRT a channel or a carrier? Two competing models explaining chloroquine resistance in Plasmodium falciparum, Trends Parasitol. 23 (2007) 332e339. [10] E. Wallin, G. von Heijne, Genome-wide analysis of integral membrane proteins from eubacterial, archaean, and eukaryotic organisms, Protein Sci. 7 (1998) 1029e1038. [11] H.M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T.N. Bhat, H. Weissig, I.N. Shindyalov, P.E. Bourne, The protein Data Bank, Nucleic Acids Res. 28 (2000) 235e242. [12] M.J. Serrano-Vega, F. Magnani, Y. Shibata, C.G. Tate, Conformational thermostabilization of the beta 1-adrenergic receptor in a detergent-resistant form, Proc. Natl. Acad. Sci. U. S. A. 105 (2008) 877e882. [13] Y. Shibata, J.F. White, M.J. Serrano-Vega, F. Magnani, A.L. Aloia, R. Grisshammer, C.G. Tate, Thermostabilization of the neurotensin receptor NTS1, J. Mol. Biol. 390 (2009) 262e277. [14] J.L. Miller, C.G. Tate, Engineering an ultra-thermostable beta(1)-adrenoceptor, J. Mol. Biol. 413 (2011) 628e638. [15] S. Abdul-Hussein, J. Andrell, C.G. Tate, Thermostabilisation of the serotonin transporter in a cocaine-bound conformation, J. Mol. Biol. 425 (2013) 2198e2207. [16] A. Penmatsa, K.H. Wang, E. Gouaux, X-ray structure of dopamine transporter elucidates antidepressant mechanism, Nature 503 (2013), 85-þ. [17] J.A. Coleman, E.M. Green, E. Gouaux, X-ray structures and mechanism of the human serotonin transporter, Nature 532 (2016) 334e339. [18] C. Hunte, H. Michel, Crystallisation of membrane proteins mediated by antibody fragments, Curr. Opin. Chem. Biol. 12 (2002) 503e508. [19] S.G.F. Rasmussen, H.J. Choi, J.J. Fung, E. Pardon, P. Casarosa, P.S. Chae, B.T. DeVree, D.M. Rosenbaum, F.S. Thian, T.S. Kobilka, A. Schnapp, I. Konetzki, R.K. Sunahara, S.H. Gellman, A. Pautsch, J. Steyaert, W.I. Weis, B.K. Kobilka, Structure of a nanobody-stabilized active state of the beta(2) adrenoceptor, Nature 469 (2011) 175e180. [20] S.G.F. Rasmussen, B.T. DeVree, Y.Z. Zou, A.C. Kruse, K.Y. Chung, T.S. Kobilka, F.S. Thian, P.S. Chae, E. Pardon, D. Calinski, J.M. Mathiesen, S.T.A. Shah,

18

[21]

[22]

[23] [24]

[25]

[26]

[27] [28]

[29]

[30] [31]

D.J. Wright et al. / Protein Expression and Purification 141 (2018) 7e18 J.A. Lyons, M. Caffrey, S.H. Gellman, J. Steyaert, G. Skiniotis, W.I. Weis, R.K. Sunahara, B.K. Kobilka, Crystal structure of the beta(2) adrenergic receptor-Gs protein complex, Nature 477 (2011), 549eU311. S.G.F. Rasmussen, H.-J. Choi, D.M. Rosenbaum, T.S. Kobilka, F.S. Thian, P.C. Edwards, M. Burghammer, V.R.P. Ratnala, R. Sanishvili, R.F. Fischetti, G.F.X. Schertler, W.I. Weis, B.K. Kobilka, Crystal structure of the human beta(2) adrenergic G-protein-coupled receptor, Nature 450 (2007) 383eU384. D.M. Rosenbaum, V. Cherezov, M.A. Hanson, S.G.F. Rasmussen, F.S. Thian, T.S. Kobilka, H.-J. Choi, X.-J. Yao, W.I. Weis, R.C. Stevens, B.K. Kobilka, GPCR engineering yields high-resolution structural insights into beta(2)-adrenergic receptor function, Science 318 (2007) 1266e1273. Y.Z. Zou, W.I. Weis, B.K. Kobilka, N-terminal T4 lysozyme fusion facilitates crystallization of a G Protein coupled receptor, PLoS One 7 (2012). W. Liu, E. Chun, A.A. Thompson, P. Chubukov, F. Xu, V. Katritch, G.W. Han, C.B. Roth, L.H. Heitman, A.P. Ijzerman, V. Cherezov, R.C. Stevens, Structural basis for allosteric regulation of GPCRs by sodium ions, Science 337 (2012) 232e236. G. Fenalti, P.M. Giguere, V. Katritch, X.P. Huang, A.A. Thompson, V. Cherezov, B.L. Roth, R.C. Stevens, Molecular control of delta-opioid receptor signalling, Nature 506 (2014) 191e196. C. Wang, Y. Jiang, J. Ma, H. Wu, D. Wacker, V. Katritch, G.W. Han, W. Liu, X.P. Huang, E. Vardy, J.D. McCorvy, X. Gao, X.E. Zhou, K. Melcher, C. Zhang, F. Bai, H. Yang, L. Yang, H. Jiang, B.L. Roth, V. Cherezov, R.C. Stevens, H.E. Xu, Structural basis for molecular recognition at serotonin receptors, Science 340 (2013) 610e614. V. Cherezov, Lipidic cubic phase technologies for membrane protein structural studies, Curr. Opin. Chem. Biol. 21 (2011) 559e566. C.W. Sikora, R.J. Turner, Investigation of ligand binding to the multidrug resistance protein EmrE by isothermal titration calorimetry, Biophys. J. 88 (2005) 475e482. D.J. Wright, C.G. Tate, Isolation and characterisation of transport-defective substrate-binding mutants of the tetracycline antiporter TetA(B), Biochim. Biophys. Acta (BBA) - Biomembr. 1848 (2015) 2261e2270. J.L. Parker, S. Newstead, Molecular basis of nitrate uptake by the plant nitrate transporter NRT1.1, Nature 507 (2014) 68e72. T. Kawate, E. Gouaux, Fluorescence-detection size-exclusion chromatography for precrystallization screening of integral membrane proteins, Structure 14

(2006) 673e681. [32] M. Hattori, R.E. Hibbs, E. Gouaux, A fluorescence-detection size-exclusion chromatography-based thermostability assay for membrane protein precrystallization screening, Structure 20 (2012) 1293e1299. [33] D. Drew, M. Lerch, E. Kunji, D.J. Slotboom, J.W. de Gier, Optimization of membrane protein overexpression and purification using GFP fusions, Nat. Methods 3 (2006) 303e313. [34] I.A. Saeed, S.S. Ashraf, Denaturation studies reveal significant differences between GFP and blue fluorescent protein, Int. J. Biol. Macromol. 45 (2009) 236e241. [35] E.L. Sonnhammer, G. von Heijne, A. Krogh, A hidden Markov model for predicting transmembrane helices in protein sequences, Proc. Int. Conf. Intelligent Syst. Mol. Biol. 6 (1998) 175e182. [36] Z.R. Yang, R. Thomson, P. McNeil, R.M. Esnouf, RONN: the bio-basis function neural network technique applied to the detection of natively disordered regions in proteins, Bioinformatics 21 (2005) 3369e3376. [37] R. Gupta, E. Jung, S. Brunak, Prediction of N-glycosylation Sites in Human Proteins. http://www.cbs.dtu.dk/services/NetNGlyc/. [38] S.J. Johns, TOPO2, Transmembrane Protein Display Software. http://www.sacs. ucsf.edu/TOPO2/. €inen, H.P. Lesch, A.I. Ma €€ €ho €nen, [39] H.-R. K€ arkka att€ a, P.I. Toivanen, A.J. Ma €-Herttuala, A 96-well format M.M. Roschier, K.J. Airenne, O.H. Laitinen, S. Yla for a high-throughput baculovirus generation, fast titering and recombinant protein production in insect and mammalian cells, BMC Res. Notes 2 (2009) 63. [40] W. Strober, Trypan blue exclusion test of cell viability, Curr. Protoc. Immunol. (2001), 21:3B:A.3B.1eA.3B.2. [41] C.K. Engel, L. Chen, G.G. Prive, Insertion of carrier proteins into hydrophilic loops of the Escherichia coli lactose permease, Biochim. Biophys. Acta-Biomembranes 1564 (2002) 38e46. [42] A.A. Thompson, J.J. Liu, E. Chun, D. Wacker, H. Wu, V. Cherezov, R.C. Stevens, GPCR stabilization using the bicelle-like architecture of mixed steroldetergent micelles, Methods 55 (2011) 310e317. [43] G. Zhang, V. Gurtu, S.R. Kain, An enhanced green fluorescent protein allows sensitive detection of gene transfer in mammalian cells, Biochem. Biophys. Res. Commun. 227 (1996) 707e711.