Skip to main content

Protein purification strategies must consider downstream applications and individual biological characteristics



Proteins are used as reagents in a broad range of scientific fields. The reliability and reproducibility of experimental data will largely depend on the quality of the (recombinant) proteins and, consequently, these should undergo thorough structural and functional controls. Depending on the downstream application and the biochemical characteristics of the protein, different sets of specific features will need to be checked.


A number of examples, representative of recurrent issues and previously published strategies, has been reported that illustrate real cases of recombinant protein production in which careful strategy design at the start of the project combined with quality controls throughout the production process was imperative to obtain high-quality samples compatible with the planned downstream applications. Some proteins possess intrinsic properties (e.g., prone to aggregation, rich in cysteines, or a high affinity for nucleic acids) that require certain precautions during the expression and purification process. For other proteins, the downstream application might demand specific conditions, such as for proteins intended for animal use that need to be endotoxin-free.


This review has been designed to act as a practical reference list for researchers who wish to produce and evaluate recombinant proteins with certain specific requirements or that need particular care for their preparation and storage.


Upstream recombinant protein processing in the industry is a highly-regulated practice performed following precise standard operating procedures. A similar consistency (in procedures) is difficult to observe at the academic level, primarily due to the misleading assumption that such a widely used technology does not require specific expertise and training to obtain usable proteins. The consequence of disregarding good practices and ignoring the necessity of checking the quality of the final products has been recently analyzed and prompted the publication of guidelines for the evaluation of purified recombinant proteins [1, 2] (for updates, see and The purpose of these guidelines is to improve the reproducibility of data obtained with protein reagents to make the whole process more transparent and encourage the use of good practices for protein preparation, data analysis, presentation, and reporting. Since the implementation of these recommendations requires not only their reliability but also their feasibility, the proposed guidelines focus on a small number of basic analyses, such as gel filtration and SDS-PAGE, that are easy to perform with widely available instrumentation.

The present review aims to relate the original indications/recommendations, that are generally useful and can be applied to any recombinant protein, to proteins with specific biochemical characteristics or the use of which needs particular requirements (Fig. 1). Such proteins (prone to aggregate, possessing disulfide bonds, stabilized by divalent cations, etc.) or proteins produced for particular biological applications require ad hoc assessment to achieve full functionality and enable high-quality downstream results (and avoid time-consuming troubleshooting). For each end-use, different quantities and levels of sample purity are required and this implies alternative purification strategies to avoid, for instance, the presence of endotoxins or other toxic components when the protein is used in cell assays or injected into animals, or of contaminant nucleic acids in proteins assessed for their interaction with such macromolecules. In Table 1, an overview can be found of the specific requirements and possible strategies for proteins intended for use in certain biological applications or with particular biochemical properties.

Fig. 1
figure 1

Recombinant proteins: useful reagents for many different applications. Proteins can function as the object of scientific research but can also be used as reagents and tool molecules. For example, mechanistic insights into protein function can be obtained by elucidating the 3D structure and studying interactions with other proteins, nucleic acids, or small molecules by determining affinities and specificities. Antibodies can be helpful tools to identify targets, whereas proteins such as cytokines and growth factors can be used as reagents in cell biology assays

Table 1 QC requirements and strategies for proteins that are used in specific biological applications or have particular intrinsic biochemical features

This article will describe a set of real-life examples to illustrate the most important issues to consider during the expression and purification process of such proteins and the strategies to overcome the most common pitfalls.

Demanding protein production—Examples of problem identification, mitigation strategy design, and corresponding output

Nucleic acid-binding proteins

For proteins that interact with nucleic acids, it is essential to introduce a nucleic acid removal step at some point in the purification workflow (Table 1). During the cell lysis process, nucleases such as Sm nuclease (benzonase®) and/or DNase or RNase can be added. However, this is often not sufficient to get rid of protein-bound nucleic acids and is also not recommended in all cases, as this could lead to the requirement to remove both the added nucleases and the contaminating nucleic acids. Therefore, extra or alternative nucleic acid removal steps such as a polyethyleneimine (PEI) or streptomycin sulfate precipitation, a heparin purification or an ion exchange chromatography step are usually required to obtain nucleic acid-free protein (Additional file 1: Examples S1 and S2). The presence of nucleic acids can be monitored throughout the purification process via the A260nm/A280nm ratio. If this ratio is below 0.6, it is a good indication of pure protein with minimal nucleic acid contamination. For specific cases of particularly strong nucleic acid binding, a more stringent procedure using a denaturant such as urea can be used, followed by protein refolding. Care should be taken to use good quality reagents, fresh sterile buffers, and clean columns (washed with 0.5 M NaOH before use).

Mouse Ferritin heavy chain 1

The goal of producing purified mouse Ferritin heavy chain 1 (mFth1) protein was to use it for high-resolution single-particle cryogenic electron microscopy (cryo-EM). The E. coli expression construct encoded untagged mFth1, and the initial protocol described cell lysis by sonication without adding any nucleases and a heating step at 70 °C followed by an ammonium sulfate precipitation, dialysis, and a size-exclusion chromatography step (work performed at Remans’ lab, according to the initial protocol published by Danev et al., 2021 [3]). However, this sample showed the presence of large amounts of contaminants in the cryo-EM images (Fig. 2A), making it wholly unsuitable for its purpose. To optimize the protocol and increase purity, some extra steps were introduced in the purification workflow and the presence of contaminating nucleic acids was monitored via the A260nm/A280nm ratio. Firstly, Sm nuclease was added both before the cell lysis process and during dialysis after the ammonium sulfate precipitation step. As the A260nm/A280nm ratio was still too high, the sample was subsequently subjected to an anion exchange chromatography step (HiTrap Q HP), leading to a reduction of the A260nm/A280nm ratio from 1.4 to 0.8. After the size-exclusion chromatography (HiLoad® 16/600 Superdex® 200 pg), the A260nm/A280nm ratio of the final mFth1 sample was 0.56, and the protein looked very pure on SDS-PAGE, indicating a proper removal of the contaminants, which was also confirmed in the cryo-EM images (Fig. 2B).

Fig. 2
figure 2

Mouse Ferritin heavy chain 1. a Cryo-EM image of mouse Ferritin heavy chain 1 (mFth1) before optimizing the purification process. The image clearly shows the presence of various contaminants. b Cryo-EM image of mFth1 after optimization of the purification process, resulting in pure protein without the presence of contaminants. (Original figure from Remans’ lab)

The chimeric protein human dsRBEC (dsRBD-EGF-chimera)

dsRBEC comprises the dsRNA binding domain (dsRBD) of human Protein Kinase R (hPKR), which serves as the polyIC (polyinosinic/polycytidylic acid) binding moiety, and it is fused to human EGF (hEGF), the targeting moiety. dsRBD shows an extremely high binding capacity for dsRNA when expressed in E. coli and such contaminants must be completely removed before polyIC addition and selective delivery into EGFR over-expressing tumor cells.

After bacterial cell lysis, most of the recombinant protein is insoluble, and the contaminant RNA impairs binding of the recombinant protein present in the soluble fraction from binding to the first capture IMAC column. Conventional procedures like polyethyleneimine (PEI) or streptomycin sulfate precipitation, RNase treatment, or high salt buffer failed to eliminate the RNA. However, the addition of urea (4 M) allowed a complete release of the contaminant, as can be clearly detected in the figures reported in Additional file 1: Fig. 1 [4]. Therefore, the refolding conditions were optimized to find the most suitable buffers and additives. Finally, cell lysis was performed in buffer supplemented with 4 M urea and the supernatant was loaded on an IMAC column, followed by a slow overnight refolding gradient from 4 to 0 M urea in optimized refolding buffer (on-column refolding). The dsRBEC was eluted from the IMAC column with refolding buffer plus imidazole and then immediately applied to a preparative Superdex 75 (100 × 1.6 cm in tandem with 60 × 1.6 cm ~ 320 ml) gel filtration for final polishing [4, 5]. The resulting dsRBEC, charged with polyIC, effectively and selectively induced polyIC internalization into EGFR overexpressing cells, inducing cell death and cytokine secretion.

Proteins that bind to divalent cations or other co-factors

For proteins that interact with divalent cations such as Zn2+, Fe2+, Cu2+, or other co-factors, the addition of a specific divalent cation (or other co-factor) during the protein expression process can be crucial (Table 1). The presence of low quantities of the same divalent cations (or other co-factors) might be necessary during the purification workflow as well. In the case of divalent cations, it is imperative to avoid using any chelating agent such as EDTA, EGTA or reducing agents with chelating properties such as DTT or DTE. When working with divalent cation-binding proteins, care must be taken to only use chelating-free (commercial) protease inhibitor cocktails. The actual presence of the divalent cations (or other co-factors) in the purified protein should be confirmed spectroscopically when possible, for instance, by atomic absorption spectrophotometry (Additional file 1: Example S5).

Recombinant ferredoxin from thermophilic cyanobacterium Mastigocladus laminosus (rFd) containing a single [2Fe ± 2S] cluster

Ferredoxins are soluble iron-sulfur proteins involved in electron-transfer reactions. Plant-type ferredoxins, which carry a single [2Fe ± 2S] cluster, serve as electron acceptors of Photosystem I. The recombinant rFd protein was expressed in E. coli strain BL21 (DE3) (Novagen) grown in TB medium containing 10 mM FeCl3 and antibiotics. The elution from the IMAC capture column was performed with a buffer containing 10 mM histidine instead of imidazole since imidazole has been shown to disrupt the [2Fe ± 2S] cluster [6]. Further purification steps included an anion exchange and size-exclusion chromatography. Finally, the rFd protein was concentrated to 12 mg/ml for crystallization screening. Interestingly, a higher resolution was obtained when crystallizing the native ferredoxin purified from the cyanobacterium Mastigocladus laminosus instead of the recombinantly produced ferredoxin.

Proteins used as antigens

Proteins are often exploited as antigens for the recovery of target-specific ligands. The first point to consider is whether the proteins will be used for an in vivo immunization with the aim of obtaining conventional mono- or polyclonal IgG or if antigens will be exploited in a procedure involving camelid IgG and/or an in vitro selection process (Table 1).

Antigens used for conventional antibody production

When antigens are injected into mice to finally obtain hybridoma cells and produce monoclonal IgG antibodies, the correct folding of the protein is not crucial. This is because most antibodies will recognize linear epitopes that are not particularly affected by the antigen structure. Even denatured antigens, for instance recovered from inclusion bodies (Additional file 1: Figure S3), can generate a valid immune response ([7], Additional file 1: Example S3). However, in this case, the sample purity must be carefully evaluated because minor contaminations with highly immunogenic proteins can induce a strong response at the expense of the target antigen.

Panning of binder libraries

Additional care must be taken when dealing with libraries obtained from camelids immunized to produce collections of binders that underwent somatic maturation. It is known that a large number of camelid antibodies binds to structural epitopes that better fit their convex paratopes. This implies that the antigen used for immunization must conserve the tridimensional native protein conformation to induce the production of antibodies able to recognize the epitopes available in vivo. There is no straightforward explanation of how this selection works because the peptide presentation modality of the class II camelid major histocompatibility complex (MHC) is conserved among mammals [8]. The antigen is digested, and single peptides of usually 13–17 amino acids are then presented on the MHC of professional antigen-presenting cells. Apart from specific cases, in most species this process results in the availability of peptides with a limited tertiary structure that will promote the development of antibodies recognizing mostly linear epitopes. However, empirically it seems that the correct folding of antigens injected in camelids might substantially help in obtaining nanobodies with excellent binding properties [9].

Similarly, entirely in vitro selections obtained by panning large pre-immune (synthetic or naïve) libraries of antibody fragments or alternative ligand scaffolds require antigens with precisely the same structure as the target to be detected in the final application. The panning procedure has no bias, in the sense that the billions of different available clones have, at least theoretically, the same chances to bind to the epitope offering the best docking among the infinite configurations present on the antigen surface. It means that it is statistically possible to select binders towards both native and new spurious epitopes if these exist due to partial misfolding, aggregation, or different conformations (Additional file 1: Figure S4). Therefore, if the final aim is to detect the antigen in natural conditions, the protein used as the antigen for the panning should be in its native structure; otherwise, the procedure high effectiveness will result in the accumulation of binders specific for epitopes present in artifacts (Additional file 1: Example S4). This is particularly important when nanobody libraries are screened because such molecules prefer to bind to conformational epitopes that could be lost or newly formed due to antigen structural rearrangements.

Proteins that contain inter- or intramolecular disulfide bonds or free cysteines

Disulfide bonds are critical for stabilizing the native structures of proteins, particularly proteins exposed to oxidizing environments, like secreted proteins or protein domains exposed on the external cell surface. On the other hand, cytoplasmic proteins are mainly present in the reduced (-SH) form. The correct formation of disulfide bridges depends on the redox conditions and on the availability of disulfide isomerases able to supervise the accomplishment of the native cis and trans bonds (Table 1).

Since many (eukaryotic) proteins undergo further complex post-translational modifications as well, eukaryotic systems are often chosen to produce proteins harboring disulfide bonds. Nevertheless, several bacterial expression systems can provide good yields and sufficient quality in the case of proteins with a limited number of cysteines [10, 11], utilizing specialized E. coli strains that enable disulfide bond formation in the cytosol (e.g., SHuffle T7 Express, Origami 2(DE3), Rosetta-gami 2(DE3), etc.), co-expression of sulfhydryl oxidase, or periplasmic accumulation.

During the purification process, the addition of reducing agents must be avoided when purifying proteins that contain inter- or intramolecular disulfide bridges to avoid changes in protein conformation that may alter the function of the protein. In contrast, for proteins that contain free cysteines, the presence of reducing agents is essential during all steps of the purification process and, crucially, during storage. This is important to avoid the formation of artefactual disulfide bonds, which can eventually lead to aggregation.

The most commonly used reducing agents are dithiothreitol (DTT), β-mercaptoethanol (β-ME) and tris-(2-carboxyethyl) phosphine hydrochloride (TCEP). TCEP is a non-thiol and odorless compound, stable in aqueous solutions and resistant to air oxidation. Moreover, unlike DTT, TCEP retains its reducing ability both at acidic pH and at pH above 7.5 [12]. For IMAC resins incompatible with DTT or TCEP, it is advisable to use 5–15 mM β-ME, which can then be replaced by other reducing agents in the following steps. Proteins containing a mixture of disulfide bonds and free cysteines represent a problematic intersection. As a compromise, it is recommended to either use no reducing agents at all or to include a very low concentration of β-ME (2 mM).

Disulfide bonds can be experimentally determined or predicted from the amino acid sequence using bioinformatic tools such as the Disulfide Bond Prediction server of Liu’s Lab (, DiANNA (, DISULFIND (, and diSBPred (

Recombinant antibody fragments

In the last years, interest in the production of recombinant antibody fragments has strongly increased. As they are used mainly as reagents, the required quality control checks are prototypic for disulfide bond-dependent proteins. Despite having a simplified structure with respect to IgGs, all antibody-derived fragments possess one or more disulfide bonds. There are several approaches for favoring the correct formations of such bonds, even in bacteria [11]. However, the yields of native, functional reagents are very case-dependent, with the accumulation of misfolded or partially folded proteins in both soluble and insoluble fractions. Furthermore, even correctly folded antibody fragments such as nanobodies might have the propensity to form soluble aggregates (colloidal aggregation) due to non-specific interactions mediated by surface patches [13]. Although removing insoluble fractions is straightforward, soluble aggregates are more challenging to identify if binder characterization is limited to functional assays such as ELISA since avidity can compensate for affinity loss. Soluble aggregates negatively affect all those applications, such as for example super-resolution microscopy techniques that rely on the small dimension of antibody fragments to maximize their performances. Therefore, sample monodispersity should be evaluated accurately with biophysical methods [2] and particular attention must be paid to the redox conditions of the cell compartment chosen for protein accumulation during recombinant expression and the optimization of storage buffers. In Fig. 3, the gel filtration profile of a nanobody sample that shows multiple peaks is reported. The first peak corresponds to the void volume and contains mainly aggregates, whereas the second peak corresponds to the monomeric protein. In other words, polymers of different complexity are present in the initial sample, and it is necessary to distinguish and separate these. Of course, this would not be possible by simple SDS-PAGE since apparently, the same target protein is present in all the elution fractions.

Fig. 3
figure 3

Separation of nanobody soluble aggregates by gel filtration. a Gel filtration profile of a nanobody sample. The first peak (blue arrow) corresponds to the void volume, the last (red arrow) to the monomeric molecule. b SDS-PAGE loaded with gel filtration elution fractions corresponding to first (blue), intermediate (no color), and last (red) peak. The fractions present different purity degrees, but the nanobody (green arrow) is always the major protein. (Original figure from de Marco’s lab)

Soluble fragment of the lymphocyte receptor LLT1

Stable and biologically active protein requiring disulfide bonds for reaching its native fold can be challenging to obtain recombinantly but biophysical analyses can help optimizing the process, as illustrated by the preparation of a soluble fragment of the lymphocyte receptor LLT1 [14]. The gel filtration profile of the wild-type LLT1 construct showed a broad peak corresponding to aggregated material in addition to a well-resolved peak corresponding to the expected non-covalent dimer of LLT1 (Fig. 4A). However, analysis of the disulfide bond pattern by mass spectrometry clearly showed misfolding due to the presence of an unpaired cysteine residue even in the well-resolved peak (Fig. 4B). Therefore, two LLT1 mutants were designed, one in which that particular cysteine residue was removed (leading to a minimal yield, Fig. 4A) and a second one in which another cysteine residue was introduced in order to reconstitute a cystic pair that is conserved in related members of this receptor family. This second mutant was stable, productive, was crystallized, and its 3D structure confirmed the presence of a correctly folded protein containing the reconstituted disulfide bond (Fig. 4C) [15].

Fig. 4
figure 4

Reconstruction of the disulfide bridge improves the folding and yield of soluble LLT1. a Gel filtration of wild-type recombinant soluble LLT1 (blue) and its C163S (orange) and H176C (green) mutants produced in HEK293T cell line. b Mass spectrometry analysis of disulfide bond pattern in wild-type LLT1 and its H176C mutant using samples from the peak at 16 ml position on SEC run shown in a corresponding to the LLT1 non-covalent dimer. The relative intensity of observed cystic peptides is shown. While in the H176C mutant both the expected two native disulfide bridges (Cys75-Cys86 and Cys103-Cys184) and the third reconstituted bond Cys163-Cys176 were formed (marked by asterisks), in the wild-type LLT1 the odd Cys163 residue paired randomly with other cysteines, leading to protein misfolding and aggregation. c Crystal structure of LLT1 (PDB 4QKI) non-covalent dimer (cyan and green) confirmed the expected disulfide bond pattern (in yellow with the reconstituted Cys163-Cys176 disulfide highlighted in red). (Original figure from Vaněk’s lab)

Prone-to-aggregation proteins

Many proteins we want to purify are completely insoluble, partially soluble, or aggregate for different reasons (Table 1). This condition can be alleviated at the expression level [16] but also poses a significant challenge during downstream purification steps requiring ad hoc strategies [17, 18].

Aggregation can be typically induced by the nucleation of a few proteins that form small and soluble aggregates, which serve as nucleation foci for the subsequent growth of larger insoluble aggregates. The nucleation-growth process can increase with time, temperature, protein concentration, or organic and inorganic contaminants. Notably, an extended lag phase can precede the abrupt formation of large insoluble aggregates [19].

Aggregation of recombinant proteins can start already during expression and may produce both inclusion bodies (IB) and soluble aggregates. Several options have been proposed to limit this problem [16,17,18,19,20,21]. These include the screening of different bacterial strains, decreasing culture temperature, using modified culture media or different solubility-enhancing fusion proteins such as maltose-binding protein, alternative expression systems such as baculovirus-mediated expression in insect cells, mammalian or cell-free expression, constructs with either amino- or carboxyl-terminal deletions, expressing homologs of the protein of interest, removing flexible loops or residues that affect solubility, and refolding of denatured proteins.

The quality of the overexpressed product must be evaluated by monitoring the ratio between monomeric and aggregated protein step-by-step in small-scale expression and purification tests coupled to SDS-PAGE and analytical gel filtration [22]. Conditions must be determined that lead to a minimal presence of aggregates (both soluble and insoluble) and maximal yield of the native target protein with the correct oligomeric conformation [17, 18].

Although protein solubility during expression is an essential prerequisite before purification, this does not prevent aggregation problems from arising at later stages of the protein production process. In general, factors such as buffer composition, pH value, kosmotropes, or chaotropes should be screened (see Fig. 5). Moreover, the procedures require rapid manipulation at 4 °C, as well as avoiding protein overcrowding during purification steps. The classical strategy of affinity chromatography (e.g., IMAC) followed by protease cleavage to remove the fusion tag and reverse affinity chromatography and size-exclusion chromatography (SEC) [19] could be replaced by affinity chromatography followed immediately by SEC to quickly eliminate soluble aggregates that can serve as aggregation nuclei (see Additional file 1: Example S6). Emphasis should be put on designing a “rapid strategy of purification” for producing and storing the target protein as fast as possible. To achieve this goal, optimizing each purification step and storage conditions should be performed prior to production scale-up.

Fig. 5
figure 5

Strategies to alleviate protein aggregation. The factors leading to protein aggregation during purification can be mitigated. Operational time should be reduced as much as possible and samples should be kept at low temperatures throughout the entire purification process. Oxidating conditions, high protein concentrations and destabilizing pH values and salt concentrations should be avoided. Mechanical stress needs to be limited and, when necessary, stabilizing molecules should be added. Some proteins may also have specific requirements for reducing agents and detergents

Although recommended additive concentrations can be found in the literature [17, 20], the optimal range for each protein is highly specific, and the buffer conditions must be fine-tuned for each project. Moreover, there could be a synergistic effect between some of these agents that could prevent different aggregation mechanisms (Additional file 1: Examples S6-S8).

Several experimental methods are routinely used to determine the protein quality and the most suitable buffers, such as visual observation of turbidity, SEC, circular dichroism (CD), light scattering (LS), differential scanning fluorimetry (DSF) and fluorescence-based thermal shift (thermofluor) assays [2]. Screening a pool of additives covering most aggregation mechanisms can considerably reduce cost and efforts [16,17,18].

Crystallization of human CLK1 kinase (how to obtain reproducible batches)

Cdc-like kinase 1 (CLK1) is a dual-specificity kinase capable of autophosphorylation on tyrosine residues and Ser/Thr phosphorylation of its substrates. Its three-dimensional X-ray structure was solved in a previous communication [23], but the procedure was difficult to reproduce. This is not uncommon when working with kinases since their preparation for structural studies can result in non-specific or partial phosphorylation(s) leading to heterogeneous protein populations. Consequently, the recovery of homogeneous preparations of CLK1 without heterogeneous phosphorylation is instrumental to achieve reproducible results in crystallization and co-crystallization experiments with potential inhibitors [24]. First, CLK1 was co-expressed with λ-phosphatase in the same E. coli host strain to cleave all possible phosphates from the phosphorylation sites, despite the fact that this approach significantly decreased the final CLK1 yield. His-tagged CLK1 was initially purified by IMAC, after which the eluate was immediately supplemented with 50 mM arginine, 50 mM glutamate and 10 mM DTT to avoid aggregation [25], concentrated and incubated overnight with TEV protease at 4 °C. The correct oligomeric conformation was separated from soluble aggregates by SEC in a buffer containing 5% glycerol and 5 mM β-ME. The last polishing step with anion-exchange chromatography was crucial to separate two different CLK1 populations: the first one, containing a consistent crystallizable non-phosphorylated CLK1 and a second one, consisting of a non-crystallizable partially phosphorylated protein.

It is important to realize that aggregation problems can occur at different stages of the purification process. For example, in the case of CLK1, the change of the cell disruption method to a more aggressive system based on higher pressure and temperature resulted in an unstable CLK1 protein with a higher tendency to aggregate, which was also unsuitable for crystallization assays.

Production of stable, ligand-binding-competent Galectin-1

Galectins, a family of carbohydrate-binding proteins with affinity for β-galactosides, are characterized by the presence of a highly conserved carbohydrate recognition domain (CRD) and a shared consensus of amino acid sequences. Galectin-1 is a prototypical lectin with an affinity for β-galactosides that plays a vital role in numerous physiological and pathological processes. It exists physiologically as a non-covalent homodimer, spontaneously dissociating into monomers at a low micromolar range. The carbohydrate ligand-binding (lectin) activity of Galectin-1 depends on three factors: (1) the correct folding and conformation of the carbohydrate recognition domain; (2) the maintenance of the reduced state of the cysteine residues either by addition of thiol-reducing agents or mutation of specific or all cysteine residues present; and (3) the preservation of the ligand-binding site in its free, unbound state.

Reports in the literature describing Galectin-1 recombinant production are commonly based on protocols that do not consider maintaining its lectin activity properly, even in studies where carbohydrate recognition is essential. It is common to find reports where there is no selection of properly folded proteins during purification, where changes in redox state compromise the lectin activity of Galectin-1, or even to find rather incomplete descriptions of how remnants from the purification are removed (i.e., reducing agents or lactose from affinity chromatography).

In order to study protein–protein interactions of Galectin-1, four different constructs were prepared—tag-less and His-tagged wild-type Galectin-1, and tag-less and His-tagged cysteine-less mutant Galectin-1 (described previously in [26]). The constructs were expressed overnight at 20 °C in the E. coli T7 Express strain followed by cell lysis using sonication in the presence of protease inhibitors and centrifugation. The His-tagged constructs were purified by IMAC and eluted with imidazole, whereas the tag-less constructs were purified by affinity chromatography over a lactose-Sepharose column and eluted with lactose. The lack of selection of properly folded proteins, as is the case in IMAC where selection occurs through a His-tag, resulted in protein aggregation; therefore, it is critical to purify Galectin-1 based on its lectin activity to ensure fully functional folded protein preparation. One drawback of this purification method is the presence of lactose in the binding site, which hinders further ligand binding studies where the CRD is involved. Therefore, the next step in the preparation of active, fully binding-competent Galectin-1 was the removal of lactose, where two methods were employed—gel filtration or extensive dialysis. The former proved not efficient enough; only thorough, extensive dialysis against HEPES buffer led to fully restored binding capacity. We can thus conclude that it is superior to purify Galectin-1 over a lactose-based column and remove lactose through extensive dialysis. A different problem arose with the prolonged preparation of the protein—in the absence of any reducing agent, the wild-type Galectin-1 construct completely oxidized following purification, and even in its presence, the protein remained reduced for a short time only (Fig. 6). Knowing that the preparation of recombinant Galectin-1 was in our particular case intended for ligand binding studies with proteins containing disulfide bonds, the presence of reducing agents, even at a lower concentration, could not be considered to avoid destabilizing the binding partner. It became evident that the best candidate for a folded, functional, and stable protein was the cysteine-less mutant of Galectin-1 purified by affinity chromatography over a lactose-based column and dialyzed for removal of lactose. Quality controls were performed by SDS-PAGE in reducing and non-reducing conditions, DLS, nanoDSF, AUC, label-free MST, and ITC, where the redox conditions, protein dispersity, thermal stability, and lectin activity were thoroughly assessed (Abreu et al., manuscript in preparation). In all instances, the cysteine-less mutant of Galectin-1 performed equally to wild-type Galectin-1 but with enhanced long-term stability, thus proving to be a valuable candidate for ligand binding studies where Galectin-1 ought to be correctly folded and active.

Fig. 6
figure 6

Recombinant Galectin-1 hydrodynamic characterization discloses its instability. a Analytical ultracentrifugation shows that wild-type Galectin-1 purified in non-reducing conditions sediments as a mixture of monomer, dimer, and various disulfide-crosslinked oligomeric species (purple curve). In contrast, purification in reducing conditions yields the expected non-covalent dimer (blue curve), which can also be revived from the oxidized sample by transferring it to the reducing conditions (cyan curve). However, the cysteine-less mutant of Galectin-1 is perfectly stable in non-reducing conditions and sediments solely as the expected non-covalent dimer. b SEC analysis of wild-type and cysteine-less Galectin-1 purified in non-reducing and reducing conditions complementing the AUC data. Apart from being freshly reduced, wild-type Galectin-1 is not monodisperse, even when purified in reducing conditions where self-oxidation still slowly occurs upon storage. On the contrary, the cysteine-less mutant is perfectly monodisperse in non-reducing conditions. (Original figure from Vaněk’s lab)

Endotoxin removal

Endotoxin contamination has been the bane of the bioprocessing industry since its inception. Endotoxins are everywhere: They are toxic and/or interfere with every type of therapeutic, diagnostic, and research product; they are indestructible within the limits of product tolerance and they are difficult to remove [27,28,29]. Proteins produced in E. coli are contaminated by endotoxins, which are lipopolysaccharides (LPS) of the outer cell membrane of Gram-negative bacteria. LPS binds to and activates toll-like receptors (TLR), thereby inducing the secretion of pro-inflammatory cytokines and systemic inflammation [30]. High LPS blood concentrations, namely above 0.6 endotoxin activity units/ml, are associated with a high risk of severe sepsis, septic shock, and mortality in humans [31]. Thus, endotoxin removal is fundamental when proteins are used for animal and human studies or for several in vitro analyses using cells in culture (Table 1).

Endotoxin removal methods have been based on positively charged chromatography (anion exchange) and affinity chromatography using poly-cation ligands such as poly-L-lysine (PLL) or immobilized polyamine (polymyxin B). Alternatively, the addition of surfactant Triton X-114 exploits the hydrophobicity of the endotoxin lipid A moiety [32]. The protocol involves an incubation step at room temperature in the presence of the surfactant to promote association with lipid A, followed by refrigeration to cause the detergent to gelatinize, facilitating its removal with still-associated endotoxin. The protocol can be adapted for simultaneous metal chelate affinity purification and endotoxins clearance: once the protein target is bound to IMAC (Ni) resin, it is washed with 50 column volumes of buffer containing 0.1% (v/v) of Triton X-114 followed by 20 column volumes of the buffer without detergent at 4 °C before imidazole elution [33]. One percent (v/v) of Triton X-114 is used for IB washing before solubilization [34]. A preliminary ion exchange step is beneficial as LPS is highly negatively charged and strongly binds to anion exchange columns (at pH 8.0, it should elute around 500 mM NaCl), whereas the protein of interest is recovered in the flow-through.

To avoid endotoxin contaminations originating from prior E. coli purifications on the same chromatography workstation, it is opportune to incubate the chromatographic pump/fluidic system overnight in 0.5 M NaOH or 4 h in 1.0 M NaOH and flush all the valves and lines before starting the purification process. Ideally, buffers should be prepared in LPS-free water (or in tap water, which usually is LPS-free when collected after 4–5 min of tap purge). Be aware that many commercial salt stocks also contain LPS; therefore, LPS-free (cell culture-grade) compounds should be purchased for these applications.

For the purification, use new columns or columns that have only been used in other LPS-free purifications. Some resin material contains endotoxin as well (e.g., StrepTactin, as it is produced in E. coli), so this needs to be washed carefully beforehand, too. A Limulus amoebocyte lysate (LAL) assay is required to assess the final amount of LPS in samples and verify that it is below the limit required for the specific applications.

Purification of protein “X”

Protein “X” from M. tuberculosis was expressed in E. coli BL21(DE3) Star as a His6-TrxA fusion protein and purified from the soluble fraction through immobilized-metal affinity chromatography (IMAC). The TrxA-tag was cleaved off using the TEV protease and the target protein was recovered by an additional IMAC step (Additional file 1: Fig. 2A). The endotoxin removal was carried out using a spin column (Column 1: ThermoFisher, cat. #88,277) containing a resin made of poly-ε-lysine covalently attached to porous cellulose beads following the standard manufacturer procedures. However, this standard protocol was not successful in bringing the endotoxin concentration in the protein sample to the set threshold (< 10 EU/ml; < 0.01 EU/μg) (Additional file 1: Fig. 2B). Hence, to have a final product suitable for mouse immunization, several approaches were tested to abate endotoxin level significantly. Firstly, an improvement of the “Column 1” protocol was explored. For Protein “X”, the optimized conditions included a longer incubation time of 2–3 h (up to overnight) at room temperature on a rotating mixer, a sample recovery step from the spin column at 500 × g for 3 min, and the usage of Tris–HCl as elution buffer. With this method, the final endotoxin concentration was around 8 EU/ml, namely tenfold lower than the one obtained in standard conditions. Alternatively, additional chromatographic steps such as anion exchange chromatography (AIEX) and SEC were introduced considering Protein “X” properties such as isoelectric point (9.36) and molecular weight (29.10 kDa). The AIEX polishing step was performed using buffers at pH 8.0, thus allowing the negatively charged lipopolysaccharide (endotoxin) molecules to bind to the resin and collect the unbound Protein “X” in the flow-through and/or during washing. As a result, the endotoxin concentration ranged between 20–30 EU/ml with a protein recovery of 80% from the total protein present in the preceding step (2nd IMAC). The third method exploited size-exclusion chromatography and led to a final endotoxin concentration < 10 EU/ml. However, a considerable sample loss was observed (protein recovery: 20%), probably due to protein precipitation or unspecific binding to the column. Both AIEX and SEC can be combined to an endotoxin removal process with spin Column 1, thus having a final product that reaches an acceptable endotoxin level. Importantly, to avoid any external endotoxin contaminations [35], all the solutions were freshly made using endotoxin-free water. The FPLC system was washed thoroughly with 1 M NaOH after every purification process, and when possible, sample handling was done under a sterile working environment. The most effective strategies that led to the best endotoxin removal with minor sample loss were the usage of spin Column 1 (optimized conditions, O.C.) and the combination of AIEX with spin Column 1 (standard condition, S.C.), where the endotoxin concentration is reduced to level < 0.01 EU/μg and the protein recovery is > 70% and > 50%, respectively.

Protein complexes

The expression and purification of protein complexes can be challenging, and the optimal conditions will vary on a case-to-case basis (Table 1). In some cases, the individual subunits can be expressed separately and then assembled into a protein complex in vitro. However, individual subunits are often not stable or unable to fold into their native 3D structure in the absence of their interaction partners. In these particular cases, simultaneous co-expression of individual subunits is required to form functional protein complexes. Fortunately, over the years, various heterologous multiprotein expression tools have been developed for standard host organisms such as E. coli [36, 37], insect [38,39,40], and mammalian cells [41]. Some methods rely on co-transformation, co-infection, or co-transfection, whereas others allow simultaneous expression of various complex subunits from a single plasmid or virus. For example, an exciting strategy developed to produce protein complexes in insect cells relies on the expression of a single polypeptide chain containing the TEV protease followed by all complex subunits, each of them (including the TEV) being separated by the TEV cleavage site. This approach has the advantages of expressing all subunits under the same promoter and having them in close proximity for correct complex assembly [42, 43].

Construct design is of utmost importance for recombinant expression of multiprotein complexes, as care needs to be taken that tags introduced for purification and/or detection purposes do not hinder proper complex assembly in the cells. When little information is available about the exact complex composition, stoichiometry, and structural arrangement of the separate subunits, this can take several rounds of optimization. It is essential to carefully assess complex integrity and stability throughout the various steps of the purification workflow. Many different methods can be used for this, the simplest of which is SDS-PAGE visualization of the individual subunits. Homogeneity can be evaluated by SEC, while SEC coupled to multi-angle light scattering (SEC-MALS) offers the additional benefit of also being able to assess the molar mass. Additionally, mass photometry [44] or native mass spectrometry [45, 46] can be used as well to check the masses of the entities present in the purified sample. When using methods such as SEC, SEC-MALS, or mass photometry, it is crucial to keep in mind that disassembly of the complex into various subunits can be observed when working at concentrations below the KD of the individual interactions. If complex disassembly occurs during certain steps of the purification workflow, an adaptation of the chromatographic methods and/or the purification buffers might be necessary. For buffer optimization, techniques such as thermofluor or differential scanning fluorimetry (DSF) can be very helpful [47]. In some cases, limiting the number of steps in the chromatographic workflow might also be more effective to avoid losing some of the (more weakly interacting) subunits by “over-purification”.

Antigen–antibody complex purification

An antibody bound to its antigen represents a specific example of a protein complex. Antibody-antigen complexes are often co-crystallized to study the interaction modalities and then such structural information can be used for engineering the antibody into a variant with improved binding features. The increasing availability of small antibody fragments has facilitated both the structural characterization of protein complexes and the modification of recombinant binders. Conventionally, the antigen and antibody are purified separately and then mixed together to form a stoichiometric protein complex. However, it has been demonstrated that they can also be co-expressed and co-purified [48], which implies only a single purification. Furthermore, co-expression can be exploited to stabilize labile antigens by means of antibody binding and also allows the expression of one of the partners without any tag. In this particular case, SDS-PAGE and gel filtration are generally sufficient to evaluate the purification procedure and the quality of the protein complex.


Purified protein reagents are used in a large variety of scientific fields. Many reviews exist for protein production on an academic scale, detailing popular strategies (for examples, see references [49, 50]). These strategies represent good starting points for producing pure, soluble proteins in close-to-milligram quantities and are implemented on a daily basis by professionals in research groups and core facilities with strong expertise in (recombinant) protein expression, purification, and standard characterization techniques. However, these are general indications that might require specific adaptations when downstream applications require particular conditions, such as endotoxin-free proteins for animal experiments or nuclease-free proteins for nucleic acid-interaction studies (see Fig. 1 and Table 1). Similarly, certain proteins possess intrinsic biochemical features that need to be considered when designing the expression and purification strategies, such as proteins that are prone to aggregation, possess disulfide bonds, or have a high affinity for nucleic acids. In this work, we presented experimental cases describing ad hoc strategies for producing such types of proteins. In these specific cases, a proper strategy needs to be designed upfront, starting from the choice of the expression system, design of the constructs (choosing appropriate tags and domain boundaries when necessary) and development of the protein purification workflow, including quality controls at various steps (Fig. 7). Some specific analyses not included in the initial set of guidelines for protein quality control assessment [2] might become essential in such specific experimental contexts. This proactive process of strategy design is comparable to Quality Assurance (QA) in pharmacological and biotechnological industries. QA is defined as the ensemble of activities required to ensure good quality in the processes used for product development and includes many methods, all aiming at preventing defects in the final product. With the introduction of quality control guidelines [1, 2] and by generating a better understanding of the specific requirements for proteins to be used as reagents in biological applications, we hope to develop proper standards for purifying high-quality proteins in a broad range of academic labs as well.

Fig. 7
figure 7

Workflow of a protein production process. Each protein production process starts with the strategy design. The biochemical characteristics of the protein of interest and the intended downstream applications, as indicated in Table 1 and Fig. 1, need to be considered when deciding which expression host organism to use and how to design the expression construct. After cloning the gene(s) of interest into a suitable expression plasmid, the sequence must be verified. Next, the most optimal expression conditions in the host organism of choice (usually E. coli, yeast, insect, or mammalian cells) are determined. This includes the screening of various parameters such as the expression strain, the expression medium, growth temperature, time, etc. Once the best condition to obtain soluble protein(s) has been found, one can proceed to large-scale protein purification. At this step, it is important to decide on the chromatographic methods that will be used (affinity chromatography, ion exchange, size-exclusion chromatography, …) and to find buffer conditions in which the protein remains in a soluble, properly folded state. Appropriate quality controls throughout the entire process are important to make sure the protein of interest is stable, non-aggregated (Fig. 5), and in a native state. The purified protein can then be used in various downstream applications (Table 1), such as biophysical characterization, interaction studies, structural analysis, immunization, cell assays, etc.

Availability of data and materials

Not applicable.



Analytical Ultracentrifugation




Circular Dichroism


Capillary Electrophoresis


Cryo-Electron Microscopy


Dynamic Light Scattering


5,5'-Dithio-bis-(2-nitrobenzoic acid)




Enzyme-Linked ImmunoSorbent Assay


Electrophoretic Mobility Shift Assay


Fast Protein Liquid Chromatography


Inclusion Body


Immunoglobulin G


Immobilised Metal Affinity Chromatography




Isothermal Titration Calorimetry




Mass Spectrometry


MicroScale Thermophoresis


Nano-Differential Scanning Fluorimetry


Nuclear Magnetic Resonance


Quality Control


Sodium Dodecyl Sulfate-PolyAcrylamide Gel Electrophoresis


Size Exclusion Chromatography


Size Exclusion Chromatography coupled to Multi-Angle Light Scattering


Tris-(2-carboxyethyl) phosphine hydrochloride

TEV protease:

Tobacco Etch Virus protease


  1. Berrow N, de Marco A, Lebendiker M, Garcia-Alai M, Knauer SH, Lopez-Mendez B, Matagne A, Parret A, Remans K, Uebel S, Raynal B. Quality control of purified proteins to improve data quality and reproducibility: results from a large-scale survey. Eur Biophys J. 2021;50(3–4):453–60.

    Article  CAS  PubMed  Google Scholar 

  2. de Marco A, Berrow N, Lebendiker M, Garcia-Alai M, Knauer SH, Lopez-Mendez B, Matagne A, Parret A, Remans K, Uebel S, Raynal B. Quality control of protein reagents for the improvement of research data reproducibility. Nat Commun. 2021;12(1):2795.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  3. Danev R, Yanagisawa H, Kikkawa M. Cryo-EM performance testing of hardware and data acquisition strategies. Microscopy (Oxf). 2021;70(6):487–97.

    Article  Google Scholar 

  4. Edinger N, Lebendiker M, Klein S, Zigler M, Langut Y, Levitzki A. Targeting polyIC to EGFR over-expressing cells using a dsRNA binding protein domain tethered to EGF. PLoS ONE. 2016;11(9): e0162321.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  5. Langut Y, Edinger N, Flashner-Abramson E, Melamed-Book N, Lebendiker M, Klein S, Levitzki A. (2017) PSMA-homing dsRNA chimeric protein vector kills prostate cancer cells and evokes anti-tumor immunity. Oncotarget. 2017;8(15):24046–62.

    Article  PubMed  PubMed Central  Google Scholar 

  6. Fish A, Lebendiker M, Nechushtai R, Livnah O. Purification, crystallization and preliminary X-ray analysis of ferredoxin isolated from thermophilic cyanobacterium Mastigocladus laminosus. Acta Crystallogr D Biol Crystallogr. 2003;59(4):734–6.

    Article  PubMed  Google Scholar 

  7. Rivera F, Espino AM. Adjuvant-enhanced antibody and cellular responses to inclusion bodies expressing FhSAP2 correlates with protection of mice to Fasciola hepatica. Exp Parasitol. 2016;160:31–8.

    Article  CAS  PubMed  Google Scholar 

  8. Plasil M, Mohandesan E, Fitak RR, Musilova P, Kubickova S, Burger PA, Horin P. The major histocompatibility complex in Old World camelids and low polymorphism of its class II genes. BMC Genomics. 2016;7:167.

    Article  CAS  Google Scholar 

  9. Pardon E, Laeremans T, Triest S, Rasmussen SG, Wohlkönig A, Ruf A, Muyldermans S, Hol WG, Kobilka BK, Steyaert J. A general protocol for the generation of Nanobodies for structural biology. Nat Protoc. 2014;9(3):674–93.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  10. de Marco A. Strategies for successful recombinant expression of disulfide bond-dependent proteins in Escherichia coli. Microb Cell Fact. 2009;8:26.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. de Marco A. Recombinant expression of nanobodies and nanobody-derived immunoreagents. Prot Expr Purif. 2020;172: 105645.

    Article  CAS  Google Scholar 

  12. Willis MS, Hogan JK, Prabhakar P, Liu X, Tsai K, Wei Y, Fox T. Investigation of protein refolding using a fractional factorial screen: a study of reagent effects and interactions. Protein Sci. 2005;14(7):1818–26.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Soler MA, de Marco A, Fortuna S. Molecular dynamics simulations and docking enable to explore the biophysical factors controlling the yields of engineered nanobodies. Sci Rep. 2016;6:34869.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Bláha J, Pachl P, Novák P, Vaněk O. Expression and purification of soluble and stable ectodomain of natural killer cell receptor LLT1 through high-density transfection of suspension adapted HEK293S GnTI(-) cells. Prot Expr Purif. 2015;109:7–13.

    Article  CAS  Google Scholar 

  15. Skálová T, Bláha J, Harlos K, Dušková J, Koval T, Stránský J, Hašek J, Vaněk O, Dohnálek J. Four crystal structures of human LLT1, a ligand of human NKR-P1, in varied glycosylation and oligomerization states. Acta Crystallogr D Biol Crystallogr. 2015;71(3):578–91.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. de Marco A, Vigh L, Diamant S, Goloubinoff P. Native folding of aggregation-prone recombinant proteins in Escherichia coli by osmolytes, plasmid- or benzyl alcohol-overexpressed molecular chaperones. Cell Stress Chaperones. 2005;10(4):329–39.

    Article  PubMed  PubMed Central  Google Scholar 

  17. Lebendiker M, Danieli T. Production of prone to aggregate proteins. FEBS Lett. 2014;588(2):236–46.

    Article  CAS  PubMed  Google Scholar 

  18. Lebendiker M, Maes M, Friedler A. A screening methodology for purifying proteins with aggregation problems. Meth Mol Biol. 2015;1258:261–81.

    Article  CAS  Google Scholar 

  19. Philo JS, Arakawa T. Mechanisms of protein aggregation. Curr Pharm Biotechnol. 2009;10(4):348–51.

    Article  CAS  PubMed  Google Scholar 

  20. Leibly DJ, Nguyen TN, Kao LT, Hewitt SN, Barrett LK. Van Voorhis WC Stabilizing additives added during cell lysis aid in the solubilization of recombinant proteins. PLoS ONE. 2012;7(12): e52482.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Esposito D, Chatterjee DK. Enhancement of soluble protein expression through the use of fusion tags. Curr Opin Biotechnol. 2006;17(4):353–8.

    Article  CAS  PubMed  Google Scholar 

  22. Sala E, de Marco A. Screening optimized protein purification protocols by coupling small-scale expression and mini-size exclusion chromatography. Prot Expr Purif. 2010;74(2):231–5.

    Article  CAS  Google Scholar 

  23. Fedorov O, Huber K, Eisenreich A, Filippakopoulos P, King O, Bullock AN, Szklarczyk D, Jensen LJ, Fabbro D, Trappe J, Rauch U, Bracher F, Knapp S. Specific CLK inhibitors from a novel chemotype for regulation of alternative splicing. Chem Biol. 2011;18(1):67–76.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. Dekel N, Einsenberg-Domovich Y, Karlas A, von Kries J, Lebendiker M, Danieli T, Livnah O. Expression, purification and crystallization of CLK1 kinase - A potential target for antiviral therapy. Prot Expr Purif. 2020;176: 105742.

    Article  CAS  Google Scholar 

  25. Golovanov AP, Hautbergue GM, Wilson SA, Lian LY. A simple method for improving protein solubility and long-term stability. J Am Chem Soc. 2004;126(29):8933–9.

    Article  CAS  PubMed  Google Scholar 

  26. Nishi N, Abe A, Iwaki J, Yoshida H, Itoh A, Shoji H, Kamitori S, Hirabayashi J, Nakamura T. Functional and structural bases of a cysteine-less mutant as a long-lasting substitute for galectin-1. Glycobiology. 2008;18(12):1065–73.

    Article  CAS  PubMed  Google Scholar 

  27. Ongkudon CM, Chew JH, Liu B, Danquah MK. Chromatographic Removal of Endotoxins: A Bioprocess Engineer’s Perspective. ISRN Chromatogr. 2012.

    Article  Google Scholar 

  28. Karkar A, Rees A. Influence of endotoxin contamination on anti-GBM antibody induced glomerular injury in rats. Kidney Intl. 1997;52(6):1579–83.

    Article  CAS  Google Scholar 

  29. Watanabe J, Miyazaki Y, Zimmerman GA, Albertine KH, McIntyre TM. Endotoxin contamination of ovalbumin suppresses murine immunologic responses and development of airway hyper-reactivity. J Biol Chem. 2003;278(43):42361–8.

    Article  CAS  PubMed  Google Scholar 

  30. Lu YC, Yeh WC, Ohashi PS. LPS/TLR4 signal transduction pathway. Cytokine. 2008;42(2):145–51.

    Article  CAS  PubMed  Google Scholar 

  31. Yaroustovsky M, Plyushch M, Popov D, Samsonova N, Abramyan M, Popok Z, Krotenko N. Prognostic value of endotoxin activity assay in patients with severe sepsis after cardiac surgery. J Inflamm. 2013;10(1):8.

    Article  Google Scholar 

  32. Aida Y, Pabst M. Removal of endotoxin from protein solutions by phase separation using Triton X-114. J Immunol Meth. 1990;132(2):191–5.

    Article  CAS  Google Scholar 

  33. Zimmerman T, Petit Frère C, Satzger M, Raba M, Weisbach M, Döhn K, Popp A, Donzeau M. Simultaneous metal chelate affinity purification and endotoxin clearance of recombinant antibody fragments. J Immunol Meth. 2006;314(1–2):67–73.

    Article  CAS  Google Scholar 

  34. Chen Z, Cui Y, Leong YA, Beddoe T, Yu D. Efficient production of recombinant IL-21 proteins for pre-clinical studies by a two-step dilution refolding method. Int Immunopharmacol. 2013;16(3):376–81.

    Article  CAS  PubMed  Google Scholar 

  35. Endotoxins and Cell Culture. Corning Application Note TC-305.

  36. Bieniossek C, Nie Y, Frey D, Olieric N, Schaffitzel C, Collinson I, Roier C, Berger P, Richmond TJ, Steinmetz MO, Berger I. Automated unrestricted multigene recombineering for multiprotein complex production. Nat Methods. 2009;6(6):447–50.

    Article  CAS  PubMed  Google Scholar 

  37. Tan S, Kern RC, Selleck W. The pST44 polycistronic expression system for producing protein complexes in Escherichia coli. Prot Expr Purif. 2005;40(2):385–95.

    Article  CAS  Google Scholar 

  38. Berger I, Tölzer C, Gupta K. The MultiBac system: a perspective. Emerg Top Life Sci. 2019;3(5):477–82.

    Article  PubMed  Google Scholar 

  39. Weissmann F, Petzold G, VanderLinden R, Huis PJ, Brown NG, Lampert F, Westermann S, Stark H, Schulman BA, Peters JM. biGBac enables rapid gene assembly for the expression of large multisubunit protein complexes. Proc Natl Acad Sci USA. 2016;113(19):E2564–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  40. Gradia SD, Ishida JP, Tsai MS, Jeans C, Tainer JA, Fuss JO. MacroBac: New technologies for robust and efficient large-scale production of recombinant multi-protein complexes. Meth Enzymol. 2017;592:1–26.

    Article  CAS  Google Scholar 

  41. Mansouri M, Bellon-Echeverria I, Rizk A, Ehsaei Z, Cianciolo Cosentino C, Silva CS, Xie Y, Boyce FM, Davis MW, Neuhauss SC, Taylor V, Ballmer-Hofer K, Berger I, Berger P. Highly efficient baculovirus-mediated multigene delivery in primary cells. Nat Commun. 2016;7:11529.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Vijayachandran LS, Viola C, Garzoni F, Trowitzsch S, Bieniossek C, Chaillet M, Schaffitzel C, Busso D, Romier C, Poterszman A, Richmond TJ, Berger I. Robots, pipelines, polyproteins: enabling multiprotein expression in prokaryotic and eukaryotic cells. J Struct Biol. 2011;175(2):198–208.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Reich S, Guilligay D, Pflug A, Malet H, Berger I, Crépin T, Hart D, Lunardi T, Nanao M, Ruigrok RW, Cusack S. Structural insight into cap-snatching and RNA synthesis by influenza polymerase. Nature. 2014;516(7531):361–6.

    Article  CAS  PubMed  Google Scholar 

  44. Sonn-Segev A, Belacic K, Bodrug T, Young G, VanderLinden RT, Schulman BA, Schimpf J, Friedrich T, Dip PV, Schwartz TU, Bauer B, Peters JM, Struwe WB, Benesch JLP, Brown NG, Haselbach D, Kukura P. Quantifying the heterogeneity of macromolecular machines by mass photometry. Nat Commun. 2020;11(1):1772.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Leney AC, Heck AJ. Native Mass Spectrometry: What is in the Name? J Am Soc Mass Spectrom. 2017;28(1):5–13.

    Article  CAS  PubMed  Google Scholar 

  46. Tamara S, den Boer MA, Heck AJR. High-Resolution Native Mass Spectrometry. Chem Rev. 2021.

    Article  PubMed  Google Scholar 

  47. Chari A, Haselbach D, Kirves JM, Ohmer J, Paknia E, Fischer N, Ganichkin O, Möller V, Frye JJ, Petzold G, Jarvis M, Tietzel M, Grimm C, Peters JM, Schulman BA, Tittmann K, Markl J, Fischer U, Stark H. ProteoPlex: stability optimization of macromolecular complexes by sparse-matrix screening of chemical space. Nat Methods. 2015;12(9):859–65.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  48. Bossi S, Ferranti B, Martinelli C, Capasso P, de Marco A. Antibody-mediated purification of co-expressed antigen-antibody complexes. Protein Expr Purif. 2010;72(1):55–8.

    Article  CAS  PubMed  Google Scholar 

  49. Gräslund S, Nordlund P, Weigelt J, Hallberg BM, Bray J, Gileadi O, Knapp S, Oppermann U, Arrowsmith C, Hui R, Ming J. Protein production and purification. Nat Methods. 2008;5(2):135–46.

    Article  PubMed  Google Scholar 

  50. Oliveira C, Domingues L. Guidelines to reach high-quality purified recombinant proteins. Appl Microbiol Biotechnol. 2018;102(1):81–92.

    Article  CAS  PubMed  Google Scholar 

Download references


The authors thank Dr. Arne Börgel from the EMBL Protein Expression and Purification Core Facility for the production of mouse Ferritin heavy chain 1 and Dr. Felix Weiss from the EMBL Structural and Computational Biology Unit for grid preparation, cryo-EM imaging and data analysis. Figure 1 in this publication was created with


A. de Marco was supported by the CRP—ICGEB research grant SVN20-01, O. Vaněk and C. Abreu were supported by the Czech Science Foundation (18-10687S), the MEYS CR (LTC20078), and the Charles University (GAUK 1436220).

Author information

Authors and Affiliations



KR, ML, and AdM conceived and organized the work, KR, ML, AdM, MM, and OV wrote the text and prepared the figures; all the authors contributed to the example collection. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Ario de Marco.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interest.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1.

Supplementary Materials

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Remans, K., Lebendiker, M., Abreu, C. et al. Protein purification strategies must consider downstream applications and individual biological characteristics. Microb Cell Fact 21, 52 (2022).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: