Fine sampling of sequence space for membrane protein structural biology

Abstract

We describe an enhancement of traditional genomics-based approaches to improve the success of structure determination of membrane proteins. Following a broad screen of sequence space to identify initial expression-positive targets, we employ a second step to select orthologs with closely related sequences to these hits. We demonstrate that a greater percentage of these latter targets express well and are stable in detergent, increasing the likelihood of identifying candidates that will ultimately yield structural information.

Graphical Abstract

Introduction

It is well-known that even seemingly minor changes in the amino acid sequence of a protein, oftentimes just a difference of a single amino acid, can have profound effects on recombinant expression levels, as well as on stability and ultimately structure determination [4–7]. This observation is what prompted the onset of ‘structural genomics’ approaches, whereby one takes advantage of the natural sequence variation amongst proteins from different organisms. Over the past decade, structural genomics approaches have proven enormously successful, especially in the challenging realm of membrane proteins, leading to a significant increase in the number of structures determined [8–12]; (https://blanco.biomol.uci.edu/mpstruc/#Latest). Still, even when applying these genomics-based strategies, identifying a well-expressing membrane protein and determining the conditions necessary to maintain its proper stability, prerequisites for both structural and functional studies, have oftentimes been unsuccessful [13, 14]. Here we present a modification of the traditional structural genomics approach to include a second screen in sequence space around initial hits, and we show that it has the potential to increase the rate of success.

Results

We previously reported the x-ray crystal structures of the prokaryotic sulfate permease CysZ from three different organisms [1] (Supplementary Fig. 1). As described within, orthologs suitable for X-ray crystallography were identified using a structural genomics cloning and expression screening platform developed by the New York Consortium on Membrane Protein Structure [11, 15], now the Center on Membrane Protein Production and Analysis (COMPPÅ). Using the sequence of CysZ from E. coli (UniProt P0A6J3) as a seed, we identified, using predefined criteria [12], 43 orthologs from the NYCOMPS collection of genomes to be screened for expression testing – a subset of 36 from this initial screen from a total of over 350 sequence-identified CysZs (Fig. 1A) were used for the detailed study described below. This initial screen allowed us to identify three promising leads for scale-up expression and purification. These three targets, from Idiomarina loihiensis (UniProt Q5QUJ8), Pseudomonas syringae (UniProt Q887V6) and Pseudomonas aeruginosa (UniProt Q9I595) all yielded crystals, diffracting to different extents. While crystals obtained from P. syringae and P. aeruginosa diffracted poorly (> 6Å resolution; data not shown), crystals of CysZ from I. loihiensis diffracted well and the structure of the protein was determined to 2.1Å (Supplementary Fig. 1). Despite having solved the structure of CysZ from Idiomarina at high resolution, there remained several unanswered questions, such as its physiological oligomeric state, or clues to its functional and mechanistic properties and for this reason, additional structures were desired.

Figure 1. Microgenomic expansion of selected Pseudomonas sp. CysZ targets.

(a) Phylogenetic relationship between CysZ proteins across all Prokaryotes. Distances are plotted relative to E. coli CysZ (UniProt P0A6J3, indicated by black arrow) and was edited to ~400 sequences for simplicity. Protein sequence identities range from 40–100% identical. CysZ proteins from our ‘pan-genomic’ test set are indicated by blue dots and text; the E. coli ‘negative’ test set by red dots and text and the Pseudomonas test set by green dots and text. The phylogenetic tree was created using BlastP (https://blast.ncbi.nlm.nih.gov/), Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo/) and iTOL (https://itol.embl.de/, [3]). Proteins that produced crystals are indicated by asterisks. (b) Close-up of Pseudomonas targets from phylogenetic tree shown in Figure (a) and tested for expression as shown in Figure 2b. Proteins that expressed, were stable in at least two detergents, yielded well-diffracting crystals and whose structures were determined are indicated by circles, squares, diamonds and asterisks, respectively. (c) Open reading frames encoding twenty CysZ proteins closely related to CysZ from Idiomarina loihiensis (UniProt Q5QUJ8), Pseudomonas syringae (UniProt Q887V6) and Pseudomonas aeruginosa (UniProt Q9I595) were cloned into pNYCOMPS N-term and tested for expression. Expressed polyhistidine-tagged proteins were recovered by Ni²⁺ affinity and separated by SDS-PAGE. Resulting Coomassie blue stained gels of recovered proteins are shown. Neither of the CysZ clones from Idiomarina sp. yielded any detectable protein (lanes 1 and 2). Sixteen of the 18 Psuedomonas targets produced bands of the expected size that were clearly visible by Coomassie staining. Genome and species name of each target is shown. (d) Detergent screen results of CysZ from Pseudomonas denitrificans and Pseudomonas fragi. CysZ proteins were extracted and purified in DM, then sequentially injected on to a size exclusion column first equilibrated in DM, β-OG, LDAO and OM. Crystallization trials were set up in DM, β-OG and LDAO at two different temperatures. (e) Four proteins produced crystals in sparse matrix screens and were selected for repeated trials in optimization screens. Representative crystals produced by CysZ proteins from P. aeruginosa, P. denitrificans, P. fragi and P. syringae are shown.

The protein sequences of all three of the CysZs that were crystallized are considerably divergent from the protein sequence of E. coli CysZ that was initially used to identify related sequences from other species in our collection. CysZ from I. loihiensis is 49% identical to E. coli CysZ and P. aeruginosa and P. syringae are only 44% identical. Moreover, while CysZ from P. aeruginosa and P. syringae are ~70% identical to one another as might be expected, they are only ~41% identical to CysZ from I. loihiensis (Supplementary Table 1). These seemingly distant relationships support the initial genomics-based strategy and in particular, the value of screening sequence space broadly (Fig. 1A).

We hypothesized that focusing on the sequence space immediately surrounding our initial successes in generating crystals might increase our likelihood of identifying other CysZ orthologs suitable for structural studies. We therefore used the protein sequences of CysZ from I. loihiensis, P. syringae and P. aeruginosa to search for other closely related orthologs, not limiting ourselves to the NYCOMPS/COMPPÅ genome collection but exploring all prokaryotic sequences available in the NCBI database (https://www.ncbi.nlm.nih.gov/). The genus Idiomarina is rather small, with complete genomic sequence information for only three species in the cluster. A search returned only two closely related CysZs, from Idiomarina baltica and Idiomarina xiamenensis, which are 62% and 59% identical to their ortholog from I. loihiensis, respectively. On the other hand, the Pseudomonas genus is well represented and from the sequences of more than 200 different Pseudomonas species, we arbitrarily selected 18 orthologs evenly dispersed along the branch, with sequence identities ranging from 59–97% (Fig. 1B, Supplementary Table 2 and Supplementary Fig. 2).

These 20 genes – 2 from Idiomarina and 18 from Pseudomonas – were synthesized commercially, with their nucleotide sequences optimized for expression in E. coli and cloned into the same expression vector coding for an N-terminal purification tag as used for the initial screen. Each was then tested for expression and single step purification by extraction in n-decyl-β-D-maltopyranoside (DM) under the same conditions. While neither Idiomarina proteins expressed and purified well, 16 of the 18 Pseudomonas targets were well-expressed, showing a convincing band after initial purification (Fig. 1C).

Next, each of these 16 promising proteins from Pseudomonas was screened for stability in several different detergents, a necessary albeit not sufficient prerequisite for successful structure determination [16–18]. After extraction and purification in DM, these were tested for stability in detergents by the NYCOMPS/COMPPÅ platform [11, 15], which in this case included a set of six different detergents screened by automated HPLC exchange on a size-exclusion chromatography (SEC) column. Overall, 13 of the 16 well-expressed CysZs Pseudomonas were stable in at least two different detergents (Supplementary Table 2). For each of these, crystallization trials were set up in (DM), n-octyl-β-D-glucopyranoside (β-OG) and n-dodecyl-N,N-dimethylamine-N-oxide (LDAO), at two different temperatures, and seven produced crystals. CysZs from P. fragi and P. denitrificans were part of this cohort. They were stable in every detergent tested (Fig. 1D) and produced well-diffracting crystals in β-OG (Fig. 1E) allowing their structures to be determined (Supplementary Fig. 1, Supplementary Table 3, [1]).

We elected to call this more targeted structural genomics approach ‘microgenomic expansion’, to distinguish it from the traditional broad-spanning ‘pan-genomic expansion’ and next, we set forth to probe its effectiveness quantitatively. To this end, we created three test sets: (1) a ‘pan-genomic’ set consisting of 36 targets from the initial broad screen; (2) a ‘negative’ set made up of 19 targets closely related to E. coli CysZ, a target that failed to express to acceptable levels in our initial screen and (3) a ‘positive’ test set consisting of the 18 from Pseudomonas selected as described above (Figs. 2A and 2B).

Figure 2. Expression and quantification of CysZ test sets and schematic of microgenomic method.

CysZ targets from each test set were expressed in triplicate in E. coli. Bands on Coomassie stained gels corresponding to CysZ recovered by Ni²⁺ affinity were quantified by densitometry using ImageJ [2]. Relative expression levels from of each sample were normalized to the expression level P. syringae included on each gel to permit comparison between each of the test sets. (a) Representative Coomassie stained SDS-PAGE gel and relative expression levels of ‘pan-genomic’ test set. (b) Representative Coomassie stained SDS-PAGE gel and relative expression levels of ‘positive’ Pseudomonas test set. (c) Representative Coomassie stained SDS-PAGE gel and relative expression levels of ‘negative’ E. coli test set. * = by one-way ANOVA means are statistically different (p=0.01). *** = by Tukey’s HSD test Pseudomonas mean is statistically different from E. coli mean (p=0.01) and Pan-genomic mean (p=0.02). (d) Schematic representation of microgenomic expansion approach. Targets related to the protein of interest are identified across a broad sequence (blue dots). Expression screening tests are used to identify promising targets (yellow stars). In the second, microgenomic round of screening, targets more closely related (ie. >70–75% identity) to the those that expressed in the initial screen (green dots) are tested for expression. These targets can then be scaled up and tested for stability in detergent(s). Ultimately, one or more of these targets may yield additional structures.

These 72 targets were tested for expression in E. coli and subsequently extracted and purified in DM, in triplicate, using the same small-scale expression and purification methods utilized previously in the NYCOMPS/COMPPÅ pipeline [11, 15]. We found that the overall mean normalized CysZ expression levels of the E. coli ‘negative’ test set (0.35) is lower than the overall expression level of the initial pan-genomic set (0.43) screened for expression (p=0.01). Furthermore, the expression levels of the Pseudomonas test set (0.75), as a whole, is significantly higher than both the pan-genomic and the E. coli test sets (p=0.02 and p=0.01, respectively), with seven proteins expressed at levels even higher than the P. syringae CysZ identified in the pan-genomic, initial broad screen (Fig. 2).

While protein expression and initial purification are important determinants for structural and functional studies of membrane proteins, it is also critical that the sample be properly folded and stable in detergent over time. When separated on a SEC column, well-folded, detergent-stable proteins will produce a single, monodisperse peak within the resolved portion of the column. To further probe the effectiveness of microgenomic expansion, we selected the six best-expressing targets from each test set for scale-up purification, removed the purification tag by treatment with TEV protease and finally, SEC analysis. In our pan-genomic test set, only one of the six targets yielded substantial tag-less protein, as determined by Coomassie staining, in the E. coli ‘negative’ test set there were none, while five of the six targets from the Pseudomonas ‘positive’ test set were efficiently cleaved (Fig. 3, upper panels). We have observed that within a given family of proteins, inefficient proteolytic cleavage of the affinity tag most often correlates with protein instability, denaturation and/or aggregation. Indeed, for this example, as the structures reveal, the N-terminus of CysZ is exposed and should be readily accessible to the protease [1]. Consistent with this observation, only the six orthologs that could be efficiently cleaved by TEV yielded a monodisperse peak on SEC, in agreement with their classification as properly folded and well-suited for structural studies (Fig. 3, lower panels).

Figure 3. Scale-up expression tests. The six top expressing targets from each test set shown in Figure 2 were selected for large scale expression and purification.

Proteins were extracted and purified in DM, treated with TEV protease to remove the affinity tag used for purification and the resulting protein loaded on to a Superdex S200 gel filtration column equilibrated in buffer containing DM and protein elution monitored by A₂₈₀. Coomassie stained SDS-PAGE gels of Ni²⁺ purified CysZ from each sample. Coomassie stained gels and size exclusion chromatography results of purified targets from scale-up expression of (a) the pan-genomic test set (b) the Pseudomonas ‘positive’ test set and (c) the E. coli ‘negative’ test set are shown. The left side of each panel shows the protein recovered immediately following elution from the Ni²⁺ resin and the right side of each panel shows the protein recovered following treatment with TEV protease and rebinding to Ni²⁺ resin. Only one protein from the pan-genomic test set produces a good elution profile and none from the E. coli negative test set. Five of the six CysZ proteins from Pseudomonas produce good size exclusion elution profiles, indicating that they are well folded.

Discussion

Structural genomics approaches have proven instrumental to the acquisition of high-resolution structural information of integral membrane proteins ([8–12]; https://blanco.biomol.uci.edu/mpstruc/#Latest). Early pioneering work demonstrated the potential of this strategy for identifying targets suitable for in-depth structural investigation [19, 20]. Since then, several developments, not least as a result of the National Institutes of Health’s Protein Structure Initiative [21, 22], coupled to an increased availability of sequence information, helped further develop and strengthen such approaches. These include, but are not limited to, the use of automated liquid handling and other high throughput-enabling technology, streamlined methods for cloning, expression screening and target evaluation [11, 13, 23–25], developments in membrane protein detergents and membrane mimetics [26–28] and of course, the advent of single-particle cryo-electron microscopy [29].

The continuously increasing number of genomes for which sequence information is available, coupled to the decrease in price of synthetic genes, offers unique opportunities for membrane protein structural biology. While with a robust high-throughput screening platform (and an unlimited budget) every ortholog could in theory be screened for expression and stability in detergents, this is neither cost nor time effective, arguing for a more astute utilization of available data and resources. Here we show how a significant increase in success at the expression and purification stages, which then propagates to detergent stability and ultimately to the number of structures determined can be obtained by coupling a broad macrogenomic expansion screen to identify sequence hotspots with a second – fine tuning – microgenomic step, sampling around these initial regions (Fig. 2). Here, we tested and validated the hypothesis that orthologs that have a high sequence identity will be more likely to share expression and detergent stability profiles than those that are distantly related. Applied to CysZ, this approach resulted in numerous crystal forms, and two additional structures, substantially increasing our molecular-level functional understanding of this protein [1] (Supplementary Fig. 1). While broader applicability of microgenomic expansion has yet to be demonstrated, at least in the case of CysZ, it has proven successful. We intend to utilize this approach for other challenging projects and anticipate that it will be of general applicability and benefit to the membrane protein community.

Materials and Methods

Cloning and Small-scale Expression and Purification

Target sequences were selected from the NYCOMPS genome library using a bioinformatics approach [11, 12], or synthesized commercially (GenScript), with all synthetic genes codon optimized for expression in E. coli. Open reading frames encoding each of the selected CysZ targets were amplified by PCR using genomic DNA, or the synthetic gene fragment, as a template. Amplified products were cloned into the expression vector pNYCOMPS N-term, which carries a 10x polyhistidine tag for purification and a TEV protease cleavage recognition site for removal of the affinity tag [15, 30, 31]. Constructs were verified by DNA sequencing. Expression plasmids were used to transform BL21(DE3)pLysS cells and grown to saturation overnight in 1 ml 2xYT, supplemented with kanamycin and chloramphenicol, in 96 well deepwell blocks. Cultures for expression were prepared by diluting overnight saturated cultures in the fresh selective media and incubated at 37°C, shaking at 700 rpm for 90 minutes prior to the addition of IPTG (Gold Biotechnology) to a final concentration of 1 mM. Following the addition of IPTG, the temperature was reduced to 22°C, and cultures grown for an additional ~15 hours. Cells were harvested by centrifugation and immediately used for purification or stored at −80°C for later processing. Cell pellets were resuspended in buffer containing 50 mM HEPES pH 7.8, 300 mM NaCl, 20 mM imidazole, 5% (v/v) glycerol, 0.5 mM TCEP-HCl (Gold Biotechnology), DNase I, and AEBSF (BioResearch Products, Inc.) at a final concentration of 1.2 mg/ml. Cells were lysed by sonication and solubilized by the addition of n-Dodecyl-β-D-Maltopyranoside (DDM, Anatrace) to 2% (w/v) final concentration. Solubilized lysates were combined with Ni²⁺-NTA resin (G Biosciences) in filter plates (Thomson Instrument Company) and incubated for ~16 hours at 4°C with shaking. Unbound sample was removed by vacuum filtration and the resin was washed twice with buffer containing 25mM HEPES pH 7.8, 500 mM NaCl, 75 mM imidazole, 5% (v/v) glycerol, 0.1 mM TCEP-HCl, and 0.2% (w/v) DDM. Bound proteins were eluted with buffer containing 25 mM HEPES pH 7.8, 200 mM NaCl, 500 mM imidazole, 5% (v/v) glycerol, 0.1 mM TCEP-HCl, and 0.02% (w/v) DDM. Equal portions of each affinity purified sample were separated on denaturing polyacrylamide gels (Bio-Rad Laboratories) and visualized by staining with Coomassie Brilliant Blue R-250 (Sigma Aldrich). CysZ targets from each of the three test sets were expressed and purified in triplicate. The quantity of CysZ recovered in each sample was determined by densitometry and quantified with ImageJ software [2]. Statistical analysis of expression levels was performed in Microsoft Excel using the using the Real Statistics Resource Pack software (Release 6.8). Copyright (2013 – 2020) Charles Zaiontz. www.real-statistics.com.

Scale-Up Expression and Purification

Expression plasmids with each of the CysZ open reading frames were used to transform BL21(DE3)pLysS as above and grown to saturation overnight at 37°C with shaking. Cultures for expression were inoculated with the saturated overnight culture and grown at 37°C until the A₆₀₀ was ~1.0. IPTG was added to a final concentration of 1 mM and growth allowed to continue overnight at 22°C. Cells were harvested by centrifugation and immediately used for purification or stored at −80°C for later processing. Cells were resuspended in the same buffer as above and lysed by sonication. n-Decyl-β-D-Maltopyranoside (DM, Anatrace) was added to each lysate at a final concentration of 1% (w/v) and rotated end-over-end for one hour at 4°C. Insoluble material was removed by centrifugation at 25,000xg for one hour. Ni²⁺-NTA resin was added to the clarified supernatants and rotated end-over-end for ~16 hours at 4°C. The slurries were loaded into disposable polypropylene columns (Bio-Rad) and drained by gravity. The resin was washed with buffer containing 25 mM HEPES pH 7.8, 500 mM NaCl, 75 mM imidazole, 5% (v/v) glycerol, 0.1 mM TCEP-HCl, and 0.2% (w/v) DM. Bound protein was eluted from the resin with buffer containing 25 mM HEPES pH 7.8, 200 mM NaCl, 500mM imidazole, 5% (v/v) glycerol, 0.1 mM TCEP-HCl, and 0.05% DM. A portion of each eluate was separated on denaturing polyacrylamide gels and visualized by staining with Coomassie blue. TEV protease was added to the remainder of each purified protein and dialyzed overnight at 4°C against 25 mM HEPES pH 7.8, 150 mM NaCl and 0.05% (w/v) DM. Following dialysis, each sample was incubated with Ni²⁺-NTA resin to remove any uncleaved protein, the protease and any contaminants. The flow throughs were collected, and their volumes reduced using centrifugal concentrators with a 10 kDa molecular weight cutoff (Sartorius). Fifty microliters of each purified CysZ protein was injected on to a 5_150 Superdex S200 size exclusion column (GE Healthcare) in 25 mM HEPES pH 7.8, 150 mM NaCl and 0.05% (w/v) DM. Migration of protein over the column was monitored by A₂₈₀. The resulting chromatograms for presentation were prepared in Microsoft Excel.

Abbreviations:

NYCOMPS

New York Consortium on Membrane Protein Structure

COMPPÅ

Center on Membrane Protein Production and Analysis

E. coli

Escherichia coli

SEC

size exclusion chromatography

DDM

n-Dodecyl-β-D-Maltopyranoside

DM

n-decyl-β-D-maltopyranoside

β-OG

n-octyl-β-D-glucopyranoside

LDAO

n-dodecyl-N,N-dimethylamine-N-oxide

TEV

tobacco etch virus

IPTG

Isopropyl β-d-1-thiogalactopyranoside

AEBSF

4-benzenesulfonyl fluoride hydrochloride

Ni²⁺-NTA

nickel nitrilotriacetic acid

HEPES

2-[4-(2-hydroxyethyl) piperazin-1-yl] ethanesulfonic acid

TCEP-HCl

Tris (2-Carboxyethyl) phosphine Hydrochloride