Scholarly article on topic 'Purification of protein complexes of defined subunit stoichiometry using a set of orthogonal, tag-cleaving proteases'

Purification of protein complexes of defined subunit stoichiometry using a set of orthogonal, tag-cleaving proteases Academic research paper on "Biological sciences"

Share paper
Academic journal
Journal of Chromatography A
OECD Field of science
{"Affinity tag" / "Tag-removing protease" / "On-column cleavage" / "Protein complex" / "Controlled subunit stoichiometry"}

Abstract of research paper on Biological sciences, author of scientific article — Steffen Frey, Dirk Görlich

Abstract Tag-free proteins or protein complexes represent certainly the most authentic starting points for functional or structural studies. They can be obtained by conventional multi-step chromatography from native or recombinant tag-free sources. Alternatively, they can be expressed and purified using a cleavable N-terminal affinity tag that is subsequently removed by a site-specific protease. Proteolytic tag-removal can also be performed “on-column”. We show here that this not only represents a very efficient workflow, but also drastically improves the purity of the resulting protein preparations. Precondition for effective on-column-cleavage is, however, that the tag-cleaving protease does not bind the stationary phase. We introduce scAtg4 and xlUsp2 as very good and bdSENP1, bdNEDP1 as well as ssNEDP1 as ideal proteases for on-column cleavage at 4°C. Four of these proteases (bdSENP1, bdNEDP1, scAtg4, xlUsp2) as well as TEV protease display orthogonal, i.e. mutually exclusive cleavage specificities. We combined these features into a streamlined method for the production of highly pure protein complexes: Orthogonal affinity tags and protease recognitions modules are fused to individual subunits. Following co-expression or in-vitro complex assembly, consecutive cycles of affinity capture and proteolytic release then select sequentially for the presence of each orthogonally tagged subunit, yielding protein complexes of well-defined subunit stoichiometry.

Academic research paper on topic "Purification of protein complexes of defined subunit stoichiometry using a set of orthogonal, tag-cleaving proteases"


Contents lists available at ScienceDirect

Journal of Chromatography A

journal homepage

Purification of protein complexes of defined subunit stoichiometry using a set of orthogonal, tag-cleaving proteases

Steffen Frey *, Dirk Görlich *

Max-Planck-Institut für biophysikalische Chemie, Am Fassberg 11, D-37077 Göttingen, Germany


Tag-free proteins or protein complexes represent certainly the most authentic starting points for functional or structural studies. They can be obtained by conventional multi-step chromatography from native or recombinant tag-free sources. Alternatively, they can be expressed and purified using a cleavable N-terminal affinity tag that is subsequently removed by a site-specific protease. Proteolytic tag-removal can also be performed "on-column". We show here that this not only represents a very efficient workflow, but also drastically improves the purity of the resulting protein preparations. Precondition for effective on-column-cleavage is, however, that the tag-cleaving protease does not bind the stationary phase. We introduce scAtg4 and xlUsp2 as very good and bdSENPl, bdNEDPl as well as ssNEDPl as ideal proteases for on-column cleavage at 4 ◦ C. Four of these proteases (bdSENPl, bdNEDPl, scAtg4, xlUsp2) as well as TEV protease display orthogonal, i.e. mutually exclusive cleavage specificities. We combined these features into a streamlined method for the production of highly pure protein complexes: Orthogonal affinity tags and protease recognitions modules are fused to individual subunits. Following co-expression or in-vitro complex assembly, consecutive cycles of affinity capture and proteolytic release then select sequentially for the presence of each orthogonally tagged subunit, yielding protein complexes of well-defined subunit stoichiometry.

© 20l4 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND

license (


Article history:

Received 20 December 20l3

Received in revised form 4 February 20l4

Accepted 7 February 20l4

Available online l9 February 20l4

Keywords: Affinity tag

Tag-removing protease On-column cleavage Protein complex

Controlled subunit stoichiometry

1. Introduction

Purification of recombinant proteins has been greatly facilitated by the availability of affinity tags mediating specific high-affinity binding to dedicated affinity matrices [l —5]. Such affinity tags may, however, interfere with the function of a protein of interest, influence its structure or preclude crystallization. Therefore, non-tagged proteins are often preferred for functional or structural studies. The classical purification of tag-free proteins using various column chromatographic techniques is often tedious and requires detailed knowledge about the target protein's individual properties. Such purifications can therefore generally not be performed using standardized protocols. In most cases, however, untagged proteins can be produced using an elegant workaround: The target protein is first expressed as a fusion with an affinity tag and a linker presenting a recognition module for a site-specific protease. During or after the purification via the engineered affinity tag, the tag-free target protein is released using the cognate protease ([5,6]).

* Corresponding authors. E-mail addresses: (S. Frey), (D. Görlich).

Such proteolytic removal of affinity tags can be accomplished in solution after elution from the affinity resin. While allowing free access of the protease to its substrate, this procedure has the disadvantage that the cleaved affinity tag has to be separated from the target protein. This generally necessitates a buffer exchange (to remove the prior used eluent) and a "reverse affinity purification step", during which the tag and any non-cleaved fusion protein (still containing the tag) are re-bound to the affinity resin, while the tag-free target protein now remains in the non-bound fraction.

An alternative to such post-elution removal of affinity tags is on-column cleavage. Here, the target protein is released from the affinity resin by directly treating the loaded resin with a specific tag-cleaving protease [7,8]. This method offers several advantages. It not only makes purifications more time-efficient by avoiding any lengthy buffer exchange and reverse chromatography steps but also allows the target proteins to be specifically released from the resin under very mild conditions: As the elution buffer differs from the washing buffer only by a minute amount of protease, on-column cleavage bypasses more drastic elution conditions such as high concentrations of competitor, significant alterations in the buffer composition or pH changes.

Within its cellular context, the physiologically relevant form of a protein is often not a single polypeptide but a complex

0021-9673/© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (

A S1-specific S2-specific modules and reagents modules and reagents

Specific_ protease 1 £ ---Target protein-___ Protease recognition site — "—-.___Affinity tag ___ Affinity resin with specific ligand 13 Specific —protease 2 (tagged)

B C Binding to D On-column cleavage E Elution from

Lysate affinity resin 1 with protease 1 affinity resin 1

\ » □ 1J t- L Jft /'■■"© AJft°n ( SI J ( SI Is !SI ] ( SI )a0 Protease 1 Affinity resin 1 ¿©i i > i £i A h Affinity resin 1

Affinity resin 1

F Binding to G On-column cleavage H Elution from affinity resin 2 with protease 2 affinity resin 2 | Protease removal

© ® Q (S1 ) ( S1 ) JuL Protease 2 © SP z* i.i Q (si) Is2l ! Affinity resin 3

Affinity resin 2 Affinity resin 2 Affinity resin 2

Fig. 1. Purification of a stoichiometric binary complex using two consecutive affinity purification steps with on-column cleavage. (A) Description of components used in the schemes (B)-(I). After co-expressing the two subunits S1 and S2 of the target complex (B), the binary complex is separated from host proteins and individual subunits S1 and S2 by two consecutive affinity chromatography steps using orthogonal tags and protease recognition modules on each of the proteins (C)-(I): The binary complex (and surplus subunit S1) is first bound to affinity resin 1 via the affinity tag on subunit S1 (C). After washing, specifically bound proteins are released from the resin using a site-specific protease recognizing the protease recognition module (D). All contaminant proteins not containing the proper protease recognition module will remain bound to the resin (E). In the second affinity chromatographic step (F)-(H), the binary complex S1«S2 is separated from surplus subunit S1 by binding to affinity resin 2 specifically recognizing the tag fused to component S2 (F) and similarly cleaved off with a component S2-specific protease (G and H). If desired, the protease can be removed e.g. via an adequate affinity resin (I). Protein complexes with more than two subunits can be purified in analogously using an appropriate number of orthogonal affinity matrices and orthogonal protease systems.

comprising two or even multiple subunits. Structural and functional characterization of such protein complexes thus critically relies on purification strategies that allow controlling the stoichi-ometry of subunits. Provided functional subunits can be produced in the absence of their binding partners, protein complexes can be assembled from individually pre-purified subunits. Alternatively, multiple subunits can be expressed and assembled in situ within the same host cell. In both cases, the assembled complex needs to be separated from excess non-assembled subunits and partially assembled sub-complexes. This can be a challenging task. We now propose a general and straightforward strategy for purification of protein complexes with defined subunit stoichiom-etry that exploits the combined discriminative power of multiple affinity matrices and proteases (Fig. 1). Briefly, by tagging individual subunits of a given protein complex with orthogonal affinity tags and orthogonal protease recognition modules, consecutive sequences of affinity capture and proteolytic release allow selecting for the presence of each tagged subunit individually. This strategy thus provides a streamlined purification scheme and a defined

stoichiometry of subunits alongside with a product purity conforming the highest standards. Although Fig. 1 only shows the purification of a binary complex, protein complexes with more than two subunits can be purified in an analogous manner.

Evidently, this strategy requires multiple proteases with orthogonal (i.e. mutually exclusive) specificities. In the accompanying paper [19] we introduced five new site-specific proteases (bdSENP1, bdNEDP1, ssNEDP1, scAtg4, xlUsp2) for tag-cleavage in solution. We show here that scAtg4, xlUsp2, bdSENP1 and bdNEDP1 are indeed fully orthogonal to each other and to TEV protease, even under conditions of an up to 10,000-fold over-digest. We further show that scAtg4, xlUsp2, bdSENP1 and the two NEDP1 proteases perform very well in on-column cleavage, even at the low temperatures (4 °C) that are preferred for gentle protein purification. For example, 30 nM bdSENP1 are sufficient to elute a >3000-fold molar excess (^100 ^M) of a His-SUMO-tagged target from a Ni2+ chelate matrix within one hour at 4 °C. For comparison, elution of an analogously tagged TEV substrate required a 300-fold higher TEV protease concentration and higher temperatures (>16 °C). Finally,

Table 1

Expression vectors designed for this study.

Name Expressed protein Resistance Origin

pSF2057 Hisi4-bdSUMO-GFP Kanamycin ColE1

pSF1466 His14-bdNEDD8-mCherry Kanamycin ColE1

pSF2301 Hisi4-xlUb-GFP Kanamycin ColE1

pSF2304 Hisi4-scAtg8-mCherry Kanamycin ColE1

pSF1438 Hisi4-TEV-GFP Kanamycin ColE1

pSF1837 Hisio-ZZ-TEV-GFP Kanamycin ColE1

pSF2106 ZZ-bdSUMO-DarpinMBP Spectinomycin p15A

pSF2325 Hisi4-ZZ-xlUb-GFP Kanamycin ColE1

pSF2326 Hisi4-ZZ-scAtg8-mCherry Kanamycin ColE1

we describe the successful implementation of the proposed tandem affinity capture and proteolytic release strategy for the purification of a stoichiometric protein complex.

2. Experimental

2.1. Expression vectors

Recombinant proteins were expressed in E. coli from appropriate expression vectors (see [l9] and Table l) essentially as described before (see [l9]). Plasmid sequences are provided on request.

2.2. Protease cleavage assays

Solution cleavage assays were performed as described [l9]. For on-column cleavage assays, 50 |il of the respective substrate-loaded affinity resin was extensively washed with LS-S buffer (250 mM NaCl, 40 mM Tris/HCl pH7.5, 2mM MgCl2, 250 mM sucrose, 2 mM DTT) in mini-spin columns (MoBiTec). For elution, the buffer was removed by mild centrifugation and the resins were mixed with l00 | l LS-S buffer containing the indicated protease or appropriate buffer controls, respectively. After elution, material released from the resin was collected by centrifugation. The resin was quickly washed with another l00 |il LS-S buffer. The wash fraction was combined with the elution fraction. Eluted GFP or mCherry proteins were quantified via their specific absorption at 488 and 585 nm, respectively. Resins and eluates were photographed while illuminated at 366 nm.

2.3. Purification of a binary MBP»DarpinMBP complex

E. coli strain NEB Express (New England Biolabs) harboring plas-mids pSFl478 encoding Hisl4-bdNEDD8-MBP [l9] and pSF2l06 encoding ZZ-bdSUMO-DarpinMBP was grown in 200 ml TB medium with 50 |g/ml Kanamycin (Kan) and 50 |ig/ml Spectinomycin (Spect) to an OD600 of 6. The culture was diluted in 600 ml fresh medium containing Kan, Spect and 0.l mM IPTG and further shaken

at l8°C over night. After adding EDTA (5mM) and PMSF (l mM) directly to the culture, cells were harvested by centrifugation, resuspended in lysis buffer (290 mM NaCl, 45 mM Tris/HCl pH 7.5, 4.5 mM MgCl2, l0 mM DTT) supplemented with 25 mM imidazole, and lysed by sonication. The cleared lysate obtained by ultracen-trifugation (l h, 200.000 g) was incubated with l0ml Ni2+ chelate resin for one hour at 4 °C. The resin was extensively washed with lysis buffer containing l5 mM imidazole followed by lysis buffer containing 250 mM sucrose. Protein complexes bound to the resin were eluted by incubation with lysis buffer containing 250 mM sucrose and 500 nM bdNEDPl for l h at 4°C. For the second round of affinity purification, the eluted material was incubated for l h at 4 °C with l0 ml anti-ZZ resin and extensively washed with lysis buffer followed by buffer WB2 (l00mM NaCl, l0mM Tris/HCl pH 7.5, 5 mM DTT). The pure binary complex was eluted within l h at 4 °C using 30 nM bdSENPl in buffer WB2. Samples corresponding to 50 mOD of cells (cells or lysates) or l/l000 of the total purification (fractions during purification) were analyzed by SDS-PAGE and Coomassie staining.

3. Results

3.1. Identification of tag-removing proteases with orthogonal specificities

The primary aim of this and the accompanying paper [l9] was to identify new proteases that can be used for tag cleavage from recombinant proteins and thus for a selective proteolytic elution of appropriately tagged multi-subunit complexes during consecutive affinity purification steps (Fig. l). Specifically, we searched for highly specific and efficient proteases with orthogonal specificity to the S. cerevisiae scUlpl protease. To this end, we chose candidates according to two alternative rationales: First, assuming that a large evolutionary distance might have been sufficient to generate an orthogonal system, we searched for clearly identifiable scUlpl orthologs in organisms that diverged early from S. cerevisiae. We also selected paralogous substrate»protease pairs from other ubiquitin-like modification pathways, assuming that these might exhibit orthogonality to the SUMO-system even within one and the same type of cells. With these boundary conditions, five new proteases along with their potential substrates (Table 2) were identified, characterized in terms of their tag-cleaving properties in solution, and benchmarked against scUspl and a stabilized variant of TEV protease [l9].

3.2. SUMO-specificproteases, bdNEDPl, scAtg4, xlUsp2 and TEV protease represent five orthogonal groups ofproteases.

The proposed method for purifying stoichiometry-controlled protein complexes crucially depends on orthogonal substrate/protease systems. It is, however, virtually impossible

Table 2

Nomenclature of substrates and proteases used in this study.

Organism Substrate Protease Catalytic fragmenta Reference

Saccharomyces cerevisiae scSUMO (Smt3p) scUlp1 (Ulp1p) 403-621 [12-14]

Brachypodium distachyon bdSUMO bdSENP1 242-481b [19]

Brachypodium distachyon bdNEDD8 bdNEDP1 fl [19]

Salmo salar ssNEDD8 ssNEDP1 fl [19]

Saccharomyces cerevisiae scAtg8 (Atg8p) scAtg4 (Atg4p) fl [15]

Xenopus laevis xlUb xlUsp2 43-463 [19]

Tobacco etch virus TEV site TEV proteasec 1-236d [16-18]

Abbreviations: sc: Saccharomyces cerevisiae; bd: Brachypodium distachyon; ss: Salmo salar; xl: Xenopus laevis; fl: full-length. a Given are the amino acid numbers of the used protease fragments with respect to the respective full-length proteins. b A shorter fragment (amino acids 248-48l) is additionally used in Fig. 2.

c Throughout this paper, we used a solubility-enhanced and autocleavage-resistant variant of TEV protease (TEV(SH) [l7]) lacking the C-terminal autoinhibitory peptide [l8]. The catalytic activity of this protease was shown to be identical to the parent full-length enzyme [l9]. d 2038-2273 with respect to the viral poly-protein

Fig. 2. Both SUMO-specific proteases, bdNEDPl, scAtg4, xlUsp2 and TEV protease represent five orthogonal groups of proteases. A: Schematic representation of substrates used in (B) and (C). The general design of substrates containing bdSUMO, bdNEDD8, ssNEDD8, scAtg8 and xlUb is analogous to the scheme shown for scSUMO-MBP. All substrate proteins contain an N-terminal polyHis-tag to facilitate their purification using a Ni2+ chelate resin. B: l00 ||M of indicated substrates (l00 ||M) were incubated with l0 ||M of indicated proteases and protease fragments for 3 h at 25 °C. The reactions were stopped by dilution in hot SDS sample buffer and analyzed by SDS-PAGE. Shown are the full-length protein (fl) and the larger cleavage products (lcp). Note that in this assay the protease concentrations used are up to l0,000-fold higher than the concentrations required for efficient cleavage of their own substrates [l9]. Even at such drastic conditions, the SUMO-proteases, bdNEDPl, scAtg4, xlUsp2, and TEV protease represent five orthogonal groups of proteases. The ssNEDPl enzyme, however, shows a weak proteolytic activity also on the xlUb-MBP substrate and is therefore not orthogonal to xlUspl. Superscript numbers refer to the amino acid numbers of full-length bdSUMO or full-length bdSENPl, respectively. C: l00 ||M of the indicated substrates were incubated at 0 °C for one hour with l ||M of indicated proteases. Under these conditions, both SUMO proteases cleave substrates containing SUMOs of their own species more efficiently than substrates containing an orthologous extraspecies SUMO. The NEDPl enzymes do not show any significant species preference.

to predict a priori in how far such proteases can discriminate between their cognate substrate and other related modifiers. It was therefore crucial to stringently test for cross-reactivity between the proteases. We did that by incubating each protease substrate (Fig. 2A) with a high concentration (l0 |M) of each protease for three hours at 25 °C (Fig. 2B). A truncated version of bdSUMO lacking the first 20 amino acids (bdSUMO2l~97) and a further N-terminally shortened version of bdSENPl (bdSENPl248-4Sl) were also included in the analysis. Strikingly, even under these

drastic conditions (representing an up to l0,000-fold overdigestion), both SUMO proteases (i), the bdNEDPl enzyme (ii), the scAtg4 protease (iii), xlUsp2 (iv), and TEV protease (v) separated into five groups of proteases (i-v) that did not cleave substrates of the respective other groups. These five groups can therefore indeed be regarded as truly orthogonal. An interesting case is ssNEDPl. This protease is truly orthogonal to scUlpl, bdSENPl, scAtg4 and TEV protease. At very high concentrations, however, ssNEDPl can cleave a substrate containing xlUb, although with low efficiency.

Fig. 3. BothSUMO-specific proteases show a clear species preference for their respective SUMO substrates, but are not fully orthogonal. l00 ||M of indicated substrates were cleaved at various conditions with either scUlpl or bdSENPl. The green bars mark lanes with the lowest protease concentration required for efficient digestions of cognate protease/substrate pairs. Similarly, the red bars highlight the lowest protease concentration sufficient for efficient digestion of substrates by the extraspecies protease. A: One hour incubation at 0 °C with varying concentrations of protease. A «40-fold increased concentration of bdSENPl is needed for efficient cleavage of scSUMO-MBP as compared to scUlpl. In contrast, efficient cleavage ofbdSUMO-MBP requires «l0 times higher concentration of scUlpl as compared to bdSENPl. B: Time course at 0 °C with fixed concentration (300 nM) of protease. bdSENPl needs >l50 times longer than scUlpl for cleaving >95% of the scSUMO substrate. In contrast, scUlpl needs only «l5 times longer than bdSENPl to cleave >95% of the orthologous Brachypodium distachyon substrate.

The Brachypodium ortholog bdNEDPl, in contrast, does not show this cross-reactivity.

3.3. scUlp1 and bdSENP1 each prefer their natural SUMO substrates, but they are not fully orthogonal

Under such conditions of drastic over-digestion, none of the SUMO- or NEDD8-specific proteases discriminated visibly between the orthologous "intraspecies" and "extraspecies" substrates (Fig. 2B). Lower protease concentrations, however, revealed a clear preference of the two SUMO-specific proteases for their corresponding intraspecies SUMO variants (Fig. 2C). In this setup, the fraction of remaining full-length SUMO substrates was consistently larger when using the corresponding extraspecies protease. A more detailed analysis revealed that a >5-fold increased scUlpl concentration or a l5 times longer incubation was needed to cleave the bdSUMO substrate as efficiently as its authentic scSUMO substrate (Fig. 3). The species preference of the Brachipodium enzyme was even more pronounced. Efficient cleavage of scSUMO by bdSENPl required «40 times more bdSENPl or a l50 times longer incubation than the natural intraspecies bdSUMO fusion (Fig. 3). No such species preference became apparent for any of the two NEDPl enzymes. Each of them cleaved the two orthologous NEDD8 substrates equally well (Fig. 2C).

3.4. On-column cleavage of immobilized substrates

To address the applicability of the characterized proteases for on-column cleavage, we tested if bdSENPl and bdNEDPl can specifically cleave off their respective fluorescent target proteins (Fig. 4A) from a Ni2+ chelate resin loaded with Hisl4-bdSUMO-GFP and Hisl4-bdNEDD8-mCherry. Indeed, 20-30 nM bdSENPl efficiently released its specific target (GFP) from the resin within one hour at 4 °C (Fig. 4B), while NEDD8-tagged mCherry remained firmly bound

to the resin even at much higher SENPl concentrations. Conversely, bdNEDPl did not recognize the bdSUMO-GFP substrate but specifically and efficiently released mCherry from the resin (Fig. 4C). Together, these experiments corroborate the orthogonal specificities of these two proteases. Strikingly, the protease concentrations needed for efficient release of the specific substrates were similar to the concentrations needed for substrate cleavage in solution [l9], even though substrate immobilization is expected to limit the diffusion and accessibility of the substrate. Similar to bdNEDPl, also ssNEDPl could be used for highly efficient on-column cleavage (Fig. Sl). As already suggested by the high extraspecies promiscuity of the NEDPl enzymes in solution (Fig. 2), ssNEDPl cleaved resin-bound ssNEDD8- or bdNEDD8-tagged target proteins with similar efficiencies (Fig. Sl).

On-column cleavage reactions might be influenced by interactions between the protease and the affinity resin, because an immobilized protease can only access substrates in its immediate vicinity. To address this issue, we compared the efficiency of substrate release from a Ni2+ chelate resin by non-tagged and polyHis-tagged bdNEDPl (Fig. 4D). While the non-tagged protease efficiently released its substrate, the polyHis-tagged version, which directly interacts with the Ni2+ chelate resin, failed to release its specific target protein even at rather high enzyme concentration (l |M) and at elevated temperature (Fig. 4D, upper panel). This effect was not caused by a general interference of the polyHis-tag with the protease activity, as both the tagged and the non-tagged enzymes efficiently cleaved a soluble substrate (Fig. 4D, lower panel).

Also TEV protease can be used for on-column cleavage of a target protein containing a TEV protease recognition site (Fig. 5). Due to the generally lower activity of TEV protease, however, an efficient one hour on-column cleavage required a far higher enzyme concentrations (>l0 |M) and incubation at 25 °C. Consistent with the results obtained with polyHis-tagged bdNEDPl protease, cleavage

Fig. 4. bdSENPl and bdNEDPl can be used for highly efficient on-column cleavage. A: Schematic representation of substrates used for (B) and (C). B, C: A Ni2+ chelate resin was pre-loaded with similar amounts of His14-bdSUMO-GFP and His14-bdNEDD8-mCherry. 50 |xl aliquots were treated with indicated concentrations bdSENPl (B) or bdNEDPl (C) for 1 h at 4 °C. Control incubations were performed with buffer or with buffer containing 400 mM imidazole. Resins and eluates were photographed while illuminated at 366 nm. GFP and mCherry in the eluate fractions were quantified via their absorption at 488 nm and 585 nm, respectively. Quantification results are given below the respective eluate fractions. Efficient on-column cleavage (>95% elution) occurred with 20 nM bdSENPl and 300 nM bdNEDPl, respectively. The cleavage was specific as even at a >30-fold higher protease concentration, no significant elution of the non-specific target protein was evident. D: l |xM of either Hisl4-tagged or non-tagged bdNEDPl protease was incubated for l h at 25 °C with a Ni2+ chelate resin pre-loaded with Hisl4-bdSUMO-GFPand Hisl4-bdNEDD8-mCherry. Efficient release of mCherry (the NEDPl-specific target protein) from the resin was only evident with a polyHis-tag-free protease (upper panel). In parallel, the activity of the protease preparations used was assayed in solution using bdNEDD8-MBP as a substrate (lower panel). In solution, polyHis-tagged and non-tagged proteases were equally active.

from the Ni2+ chelate column was severely compromised when the TEV protease was polyHis-tagged and consequently immobile on the affinity resin (Fig. 5A and B). In the absence of resin-interactions, however, polyHis-tagged and non-tagged TEV protease cleaved with similar efficiencies (Fig. 5C).

The scAtg4 and xlUsp2 proteases were tested for their on-column cleavage properties on two different affinity resins, a Ni2+ chelate resin and an anti-ZZ resin. While «0.6-l |M of each protease was sufficient to efficiently release within one hour at 4 °C their respective target proteins from the anti-ZZ resin (Fig. 6B), on-column cleavage from the Ni2+ chelate resin required significantly higher protease concentrations (3-l0 |M; not shown). We assume that this difference can be attributed to weak direct interactions between the proteases and the Ni2+ chelate resin. A summary of specific conditions recommended for on-column cleavage is given in Table 3.

3.5. Comparison between on-column cleavage and protease cleavage after elution from an affinity resin

To assess the power of on-column cleavage in comparison to post-elution tag removal, we compared the protein purities obtainable by Ni2+ chelate chromatography using the two different purification schemes (Fig. 7). For this purpose, a Ni2+ chelate resin was loaded with a limiting amount of Hisl4-bdSUMO-tagged target protein (MBP), washed and then split in two identical quantities. When imidazole was applied to the first half of the resin,

Table 3

Suggested protease concentrations and temperatures for a near quantitative (>95%) on-column substrate cleavage.

Protease Protease concentration a Temperature

scUlp1 100 nM 0-4 °C

bdSENP1 30 nM 0-4 °C

bdNEDP1 0.3-0.5 |xM 0-4 °C

ssNEDP1 0.5-1 |xM 0-4 °C

scAtg4 0.5-1 |xM 0-4 °C

xlUsp2 1-2 |xM 0-4 °C

TEV protease > 10 |xM 16-25°C

a Concentrations are based on the results shown in Figs. 4-6 and Sl as well as numerous routine purifications performed in the lab. They refer to standard conditions: l00|xM on-column-immobilized PlAia or PlGiy substrates, l h, LS-S buffer (250 mM NaCl, 40mM Tris/HCl pH7.5, 2mM MgCh, 250mM sucrose, 2mM DTT). Correction factors for deviating conditions are given in Table 3 of the accompanying paper [l9].

several contaminants from the bacterial host co-eluted with the MBP fusion protein (Fig. 7, lane 4). After treating the eluate fraction with bdSENPl, the final preparation (Fig. 7, lane 5) thus contained not only the target protein, but also these contaminants along with the Hisl4-bdSUMO tag, traces of non-cut Hisl4-bdSUMO-MBP and the protease that was used for cleavage. Removal of these impurities would require additional steps. In contrast, when the second half of the charged resin was on-column treated with 30 nM bdSENPl without changes in buffer composition, the eluate fraction (Fig. 7, lane 6) contained almost exclusively the target MBP along with some minor degradation products and a minute amount

Fig. 5. On-column cleavage using polyHis-tagged and non-tagged TEV protease. A Ni2+ chelate resin was separately loaded with His14-TEV-GFP (A) or His10-ZZ-TEV-GFP (B). Similarly, IgG Sepharose was loaded with His10-ZZ-TEV-GFP (C). 50 |xl aliquots of loaded resins were treated with indicated concentrations of polyHis-tagged or non-tagged TEV protease for 1 h at 25°C. Control incubations were performed with buffer or 500 mM imidazole. Resins and eluates were photographed with illumination at 366 nm. GFP in the eluate fractions was quantified via its absorbance at 488 nm. Numbers are given below the eluate fractions. Release of the fluorescent target proteins from the affinity matrices was less efficient when TEV protease directly interacted with the affinity resin: PolyHis-tagged TEV protease matched the efficiency of the non-tagged enzyme in cleaving soluble or IgG Sepharose-immobilized substrates (C), but performed far worse in releasing substrates from a Ni2+ chelate resin (A and B).

Fig. 6. On-column cleavage using scAtg4 and xlUsp2. A: Schematic representation of substrates used for (B) and (C). B, C: Similar to the experiments shown in Fig. 4, the release of ZZ-scAtg8-mCherry or ZZ-xlUb-GFP from an anti-ZZ resin was analyzed using the indicated concentrations of scAtg4 (B) and xlUsp2 (C) for 1 h at 4 °C. Efficient cleavage required 0.6-1 |xM of scAtg4 and xlUsp2, respectively.

stoichiometry. For that, we chose the well-characterized complex between the E. coli maltose binding protein (MBP) and the MBP-specific designed ankyrin-repeat protein "off7" (anti-MBP DARPin, here called DarpinMBP) [9]. The two binding partners were co-expressed in E. coli as His14-bdNEDD8-MBP and ZZ-bdSUMO-DarpinMBP and purified by two consecutive capture-and-release steps (Fig. 8, for a schematic representation also see Fig. 1): His14-bdNEDD8-MBP and its binary complex with ZZ-bdSUMO-DarpinMBP were captured via their His14-tags by a Ni2+ chelate resin and - after washing off lysate proteins - they were specifically released by bdNEDP1. This first eluate fraction was then applied to a ZZ-specific affinity resin. Excess of MPB remained un-bound, while a highly pure and stoichiometric binary complex could be eluted with bdSENP1. Most importantly, the two target proteins were cleaved off from their tags during the purification procedure, thereby yielding a non-tagged complex.

of protease (^1 ^g/ml). In this setup, the cleaved His14-bdSUMO tag, the major contaminant proteins and non-cut target protein remained firmly bound to the column until post-elution with imidazole (lane 7). A direct comparison of the final protein preparations obtained with both schemes (lanes 8 and 9) clearly illustrates the far higher purity of the target purified by the affinity capture and proteolytic release strategy.

3.6. Purification of a stoichiometric protein complex using an orthogonal protease pair

We next wanted to provide a proof of principle that the orthogonal specificities of bdSENP1 and bdNEDP1 and their excellent performance in on-column cleavage can be exploited in the purification of a protein complex and the selection for a precise subunit

4. Discussion

4.1. Specificity of proteolytic processing

The primary aim of this study was to develop a generally applicable method for the purification of protein complexes. This method relies on multiple site-specific proteases with mutually exclusive cleavage specificities. In the accompanying paper [19], we analyzed and compared the cleavage properties of seven different site-specific proteases. Here, we show that six of these seven proteases can indeed be divided into five groups of orthogonal specificities, i.e. a given substrate is efficiently cleaved by a protease of its cognate class, but is resistant even to excessive concentrations of any orthogonal protease. These groups are (i) the SUMO-specific proteases scUlp1 and bdSENP1, (ii) the NEDP1-enzyme from Brachypodium,

Fig. 7. Comparison of protein purities obtained by Ni2+ chelate chromatography with imidazole elution and on-column cleavage, respectively. An E. coli lysate (lane 2) containing «4 mg of HisM-bdSUMO-MBP in LS-buffer (250 mM NaCl, 40 mM Tris/HCl pH7.5, 2mM MgCl2, 250mM sucrose, 2mM DTT) supplemented with l0 mM imidazole was applied to l ml of Ni2+ chelate resin. Non-bound material was washed off using the same buffer (lane 3). Half of the matrix was eluted with LS-buffer containing 400 mM imidazole (lane 4); the eluate was then cleaved in solution using 30 nM bdSENPl (lane5). The second half of the loaded resin was treated on-column with 30 nM bdSENPl for l h at 4 °C before elution with LS-buffer (lane 6). Material remaining on the column was released with LS-buffer+ 400 mM imidazole (lane 7). The relevant eluate fractions of both purification schemes were loaded again in neighboring lanes to allow for a direct comparison (lanes 8 and 9). 50mOD of lysate samples (lanes l-3) or l/800 of the total material (lanes 4-9) was resolved by SDS-PAGE and stained with colloidal Coomassie. The stars (*) mark protein contaminants that elute from the Ni2+ chelate resin with imidazole but are absent when using on-column cleavage. Note that on-column cleavage yields a significantly purer final protein preparation.

(iii) the yeast scAtg4 protease, (iv) Xenopus laevis xlUsp2 and (v) TEV protease.

The seventh protease characterized here, ssNEDPl, showed an unexpected specificity profile: Similar to bdNEDPl, it is fully orthogonal to the protease groups (i), (iii) and (v). In contrast to its Brachypodium ortholog, however, it can also cleave an xlUb-containing substrate - although with low efficiency. This example illustrates that a priori the prediction of orthogonality based just on sequence or evolutionary distance is not reliable. The differences between bdNEDPl and ssNEDPl in respect to their cross-reactivity with the xlUb-containing substrate also illustrate that extrapolating a protease's specificity straight from one species to another might result in incorrect predictions.

Brachypodium NEDPl as well as salmon NEDPl cannot discriminate between ssNEDD8- and bdNEDD8-containing substrates and cleaved them with virtually identical efficiencies (Fig. 2 and Sl). This species promiscuity can easily be explained by the striking conservation of the NEDD8 proteins: There are as few as l2 amino acid exchanges between salmon and Brachypodium, and only 5 of them are non-conservative [l9].The two exchanges at the putative interface with the proteases appear non-critical for the recognition by the protease. The observed species promiscuity of NEDPl enzymes has, however, interesting practical implications: As a given NEDD8

substrate can be cleaved by both, bdNEDPl and ssNEDPl (Fig. 2 and Sl), the protease used for cleavage in solution or on-column can be chosen freely. While bdNEDPl is slightly more active under standard conditions, ssNEDPl is remarkably insensitive towards high salt or a suboptimal residue in the substrate's Pl '-position [l9]. The salmon enzyme might thus be the protease of choice when cutting suboptimal substrates or cleaving at special buffer conditions -as long as no strict orthogonality to any ubiquitin-specific protease is required.

In contrast to the NEDPl enzymes, the two SUMO-specific proteases clearly prefer their natural intraspecies substrates over substrates containing the orthologous extraspecies SUMO variant (Figs. 2C and 3): While the yeast enzyme scUlpl significantly cleaves bdSUMO-containing substrates, only «5% of the scSUMO substrate is cleaved by the Brachypodium enzyme at conditions required for «95% cleavage of its own bdSUMO substrate. In kinetic assays, bdSENPl cleaves the corresponding Brachypodium substrate even >l50-fold more efficiently than the substrate containing scSUMO (Fig. 3B). These effects can most likely be attributed to a significant number of amino acid exchanges within the substrate»enzyme interfaces of the respective yeast and Brachypodium substrate»protease complexes [l9].

4.2. On-column cleavage

Most critical for practical applications, all new proteases described here can be used for highly efficient on-column cleavage from various affinity resins at 4 °C. Compared to common elution schemes that require high concentrations of competitor, significant alterations in the buffer composition or drastic pH changes, such on-column cleavage allows for the elution of target proteins without changing the buffer composition: The elution buffer differs from the washing buffer only by minute amounts of protease (e.g. ^l |ig/ml for bdSENPl). Most importantly, however, on-column cleavage potentiates the efficiency of protein purifications by elegantly combining the specificities of the affinity resin and the protease: Only proteins containing the proper affinity tag and the proper protease recognition module will be bound and consecutively released from the resin. In contrast, contaminant proteins non-specifically interacting with the resin-and thus lacking the specific protease recognition module - will remain bound to the affinity resin during such elution step (Fig. 7). As most contaminant proteins, non-cleaved target protein and the tag remains bound to the resin, on column cleavage also circumvents time-consuming buffer exchange or "reverse chromatography" steps generally required for removal of the cleaved-off tag and co-eluted contaminant proteins. On-column cleavage (as any other affinity purification strategy) does not preclude co-purification of contaminants interacting with the target protein itself. In such cases, an adaptation of the washing strategy might be considered. Altering the ionic strength, for example, may help to wash off contaminant proteins or nucleic acids. Co-purifying chaperones can often be released by washing the resin with l mM Mg2+/ATP in the presence of l00 mM potassium ions - conditions known to stimulate substrate dissociation from Hsp70-like chaperone [l0].

The new proteases described here cleave their substrates with an extraordinary efficiency (see also the accompanying paper [l9]). Therefore only minute amounts of protease have to remain as in the final protein preparation after on-column cleavage (e.g. <0.l% of the target protein if bdSENPl is used). In practice, the eluate will even contain far less protease than added to the cleavage buffer because most of the added protease appears to be retained through the cleaved tag on the resin. If ultimate purity is aspired, we recommend using a protease with an orthogonal tag that allows protease-removal with an additional affinity resin.

Fig. 8. Purification of a tag-free binary complex with controlled subunit stoichiometry. A: Schematic representation of the complex purified in (B). B: His14-bdNEDD8-MBP and ZZ-bdSUMO-DarpinMBP were co-expressed in E. coli (lane1-4). In the first affinity purification step (lanes 5-8), the cleared lysate (lanes 4 and 5) containing the binary complex along with an excess His14-bdNEDD8-MBP and binding-incompetent ZZ-bdSUMO-DarpinMBp was applied to a Ni2+ chelate resin. The binary ZZ-bdSUMO-DarpinMBp«His14-bdNEDD8-MBP complex and the excess His14-bdNEDD8-MBP bound efficiently to the resin. Most lysate proteins and non-complexed DarpinMBp species (*) remained in the flow-through fraction (lane 6). Treatment for 1 h at 4 °C with 0.5 ||M of bdNEDP1 efficiently released the ZZ-bdSUMO-DarpinMBp«MBP complex along with free MBP (lane 7); the His14-bdNEDD8-tag and most contaminant lysate proteins remained bound to the resin (lane 8). In the second affinity purification step (lanes 9-12), the ZZ-bdSUMO-DarpinMBP«MBP complex was bound to a ZZ-specific affinity resin, while all MBP not complexed to MBP as well as the bdNEDP1 enzyme were washed out. Incubation with 30 nM of bdSENP1 for one hour at 4 °C released the tag-free binary DarpinMBP«MBP complex from the resin (lane 11). Samples corresponding to 50 mOD of cells (cells or lysates) or 1/1000 of the total purification (fractions during purification) were analyzed by SDS-PAGE and Coomassie staining.

Using bdNEDP1 and TEV protease as examples, we showed that a direct interaction between the protease and the affinity resin limits the efficiency of on-column cleavage reactions. This conclusion is most probably generally true and needs to be considered for the intended experimental design. For example, as most commercial protease preparations contain polyHis-tags, such proteases cannot be used for efficient on-column cleavage from Ni2+ chelate resins. We reason that the comparably low activity of scAtg4 and xlUsp2 in our on-column cleavage assays on Ni2+ chelate resin might be explained by intrinsic interactions between the Ni2+ chelate resin and these proteases. Consistent with this interpretation, both proteases efficiently released their respective substrates from an unrelated anti-ZZ resin.

4.3. Purification of monomeric proteins

For purification of monomeric proteins using a single affinity chromatography step, we routinely elute our target proteins directly from various affinity resins using 30-50 nM bdSENP1 or 300-500 nM bdNEDP1 (depending on the tag used) within one hour at 4°C. In the vast majority of cases, we observe an efficient release of the target protein, generally yielding target protein concentrations between 100 and 300 |M (up to 120 mg/ml), probably mostly limited by the binding capacity of the resin. This on-column cleavage procedure is exceptionally robust: In our lab, from several hundreds of target proteins, only one could so far consistently not be processed by the protease. In this special case, the

known structure suggests that the residue in Pl '-position is already part of a large folded domain that effectively prevents proteolytic processing.

4.4. Purification protein complexes of defined stoichiometry

Combined, the described proteases can be of great benefit for purifying protein complexes with controlled subunit stoichiome-try: We here introduced and tested a generally applicable strategy that combines multiple consecutive affinity purification steps with specific on-column cleavage. The purification of binary complexes according to this scheme is detailed in Fig. l. In practice, we prefer in the first affinity chromatography step a resin that allows for a quick and highly efficient capture of target complexes. For this purpose, we routinely use a Ni2+ chelate resin. The protease used for on-column cleavage must therefore not contain a polyHis-tag (as discussed in Section 4.2). For elution at this initial step, bdNEDPl is preferred, because the slightly higher amount of protease needed for efficient cleavage (in comparison to bdSENPl) can be efficiently removed during the following purification step. For the second purification step, several well-established affinity matrices can be used, such as IgG Sepharose for binding ZZ-tags, chitin-beads for capturing chitin-binding domains, beaded cellulose for cellulose-binding domains, or any resin directed against peptide tags [3-5,ll]. For elution at this step it is advisable to use a protease with the highest possible specific activity, because then less protease has to be added and hence will remain in the elu-ate. Therefore, we recommend bdSENPl for elution at this step. A tagged variant of bdSENPl can be used if a complete removal of trace amount of the protease is required after on-column cleavage.

It is important to note that the introduced procedure is not limited to binary complexes. A purification scheme employing three or more orthogonal tags and proteases can analogously be used for a straightforward purification of triple or higher order complexes with defined subunit stoichiometry. Also, the method allows for the purification of complexes with uneven stoichiometry (e.g. a trimeric complex with two individual components in l:2 stoi-chiometry). In such cases, however, we recommend to introduce distinct tags and protease recognition modules on the otherwise identical subunits.

In all applications, it is crucial to consider that most of the so-far commercially available preparations of site-specific proteases contain affinity tags and can therefore not be used for efficient on-column cleavage from the respective affinity resins. In addition, it has to be taken into account that some proteases may possess suboptimal features with respect to their efficiency, specificity, or special requirements concerning temperature and buffer. We are confident that the proteases presented here are largely devoid of such drawbacks. They will therefore be of great practical use for

labs routinely purifying recombinant proteins and protein complexes from prokaryotic hosts. The proposed purification schemes for single proteins and protein complexes are highly robust and can therefore be used in standardized protocols. Most importantly, due to the efficient cleavage of our proteases even at low temperatures and their tolerance towards various buffer conditions [19], the schemes can be adapted to the needs of the target proteins or complexes over a wide range of conditions.


We wish to thank Heike Behr for excellent technical support, Matthias Samwer and Bastian Hülsmann for comments on the manuscript, and the Max-Planck-Gesellschaft and the Deutsche Forschungsgemeinschaft (SFB860) for funding.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at 2014.02.030.


[1] M. Uhlen, G. Forsberg, T. Moks, M. Hartmanis, B. Nilsson, Curr. Opin. Biotechnol. 3(1992)363.

[2] S.C. Makrides, Microbiol. Rev. 60 (1996) 512.

[3] J.J. Lichty, J.L. Malecki, H.D. Agnew, D.J. Michelson-Horowitz, S. Tan, Protein Expr. Purif. 41 (2005) 98.

[4] D.S. Waugh, Trends Biotechnol. 23 (2005) 316.

[5] C.L. Young, Z.T. Britton, A.S. Robinson, Biotechnol. J. 7 (2012) 620.

[6] J. Arnau, C. Lauritzen, G.E. Petersen, J. Pedersen, Protein Expr. Purif. 48 (2006) 1.

[7] P.A. Walker, L.E. Leong, P.W. Ng, S.H. Tan, S. Waller, D. Murphy, A.G. Porter, Biotechnology (N Y) 12(1994) 601.

[8] C. Dian, S. Eshaghi, T. Urbig, S. McSweeney, A. Heijbel, G. Salbert, D. Birse, J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 769 (2002) 133.

[9] H.K. Binz, P. Amstutz, A. Kohl, M.T. Stumpp, C. Briand, P. Forrer, M.G. Grutter, A. Pluckthun, Nat. Biotechnol. 22 (2004) 575.

[10] B. Feifel, E. Sandmeier, H.J. Schonfeld, P. Christen, Eur. J. Biochem. 237 (1996) 318.

[11] J. Nilsson, S. Stahl, J. Lundeberg, M. Uhlen, P.A. Nygren, Protein Expr. Purif. 11

(1997) 1.

[12] S.J. Li, M. Hochstrasser, Nature 398 (1999) 246.

[13] T.R. Butt, S.C. Edavettal, J.P. Hall, M.R. Mattern, Protein Expr. Purif. 43 (2005) 1.

[14] M.P. Malakhov, M.R. Mattern, O.A. Malakhova, M. Drinker, S.D. Weeks, T.R. Butt, J. Struct. Funct. Genomics 5 (2004) 75.

[15] T. Kirisako, Y. Ichimura, H. Okada, Y. Kabeya, N. Mizushima, T. Yoshimori, M. Ohsumi, T. Takao, T. Noda, Y. Ohsumi, J. Cell Biol. 151 (2000) 263.

[16] R.B. Kapust, J. Tozser, J.D. Fox, D.E. Anderson, S. Cherry, T.D. Copeland, D.S. Waugh, Protein Eng. 14 (2001) 993.

[17] S. van den Berg, P.A. Lofdahl,T. Hard, H. Berglund,J. Biotechnol. 121 (2006) 291.

[18] C.M. Nunn, M.Jeeves, M.J.Cliff,G.T.Urquhart, R.R.George, L.H.Chao,Y.Tscuchia, S. Djordjevic, J. Mol. Biol. 350 (2005) 145.

[19] S. Frey, D. Görlich, J. Chromatogr. A 1337 (2014) 95.