Protein solubility and folding monitored in vivo by ...

© 2001 Nature Publishing Group http://biotech.nature.com

RESEARCH ARTICLES


Protein solubility and folding monitored in vivo by structural complementation of a genetic marker protein W. Christian Wigley1, Rhesa D. Stidham1,2, Nathan M. Smith1, John F. Hunt3, and Philip J. Thomas1,2* 1Department

of Physiology and 2Graduate Program in Molecular Biophysics, The University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX 75235. 3Department of Biological Sciences, Columbia University, New York, NY 10027. *Corresponding author ([email protected]). Received 3 October 2000; accepted 7 December 2000

Protein misfolding is the basis of a number of human diseases and presents an obstacle to the production of soluble recombinant proteins. We present a general method to assess the solubility and folding of proteins in vivo. The basis of this assay is structural complementation between the α− and ω− fragments of β-galactosidase (β-gal). Fusions of the α-fragment to the C terminus of target proteins with widely varying in vivo folding yield and/or solubility levels, including the Alzheimer’s amyloid β (Aβ) peptide and a non-amyloidogenic mutant thereof, reveal an unambiguous correlation between β-gal activity and the solubility/folding of the target. Thus, structural complementation provides a means of monitoring protein solubility/misfolding in vivo, and should find utility in the screening for compounds that influence the pathological consequences of these processes. Keywords: aggregation, folding, solubility, complementation, protein-folding diseases

There are many potential applications for a genetic system enabling rapid and efficient characterization of protein solubility in vivo. One of the cornerstones of biotechnology is the ability to heterologously express functional proteins. Unfortunately, many important target proteins are not efficiently expressed in a folded and/or soluble form as a result of the complexity of folding1 and the limited solubility of many folded domains. The yield of soluble protein can often be improved by optimizing the primary sequence of the target protein2 or the genetic background or growth conditions of the expressing cells3–8. However, extant biochemical means of assessing solubility are tedious, making screening for constructs and/or conditions yielding improved solubility inefficient and genetic selection impossible. Protein-folding diseases represent a second area in which solubility characteristics are of vital importance9,10. These diseases, which have proved particularly refractory to pharmaceutical development, are caused by misfolding4,11,12 or by aberrant processing leading to the formation of aggregation-prone protease-resistant products13–26. The ability to screen for folding and solubility in vivo could be used to assay for compounds that promote the folding or inhibit the aggregation of disease-associated proteins. The data presented here demonstrate that a genetic system based on structural complementation27–37 of a selectable marker protein can be used to assess in vivo solubility (here, the term solubility is used to denote both chemically defined solubility of a folded domain and the absence of aggregation due to misfolding). Structural complementation involves the division of a protein into two component segments that combine to form a functional structure38. The specific implementation presented is an adaptation of the α-complementation system of β-gal (ref. 28). Each monomer of the homotetrameric enzyme can be divided into two fragments, the small α-fragment (typically 50–90 residues in length), and the larger ω-fragment, composing the remainder of the 135 kDa monomer. In the presence of NATURE BIOTECHNOLOGY VOL 19 FEBRUARY 2001

http://biotech.nature.com

α-fragment, dimers of ω-fragments achieve a dynamic equilibrium to form a tetramer with enzymatic activity28,39. Thus, redistribution of the α-fragment to the insoluble fraction in cells would be predicted to lead to a reduction in the level of β-gal activity. Three extant systems for monitoring protein misfolding in vivo rely on the ability of misfolding proteins to co-translationally induce improper folding of a C-terminally fused marker protein40–42. As such, slow misfolding or aggregation events may escape detection by these schemes. Here we show that fusion of the α-fragment to the C terminus of a target protein produces a chimera with solubility properties that reflect those of the target alone (Fig. 1A). Thus, β-gal activity levels have been engineered to report the solubility of the target. Such an assay should facilitate efficient, high-throughput screening of compounds for promoters of protein folding or inhibitors of protein aggregation. Results To test the ability of α-fragment chimeras to complement the ω-fragment and report target protein solubility, model polypeptides were fused to the N terminus of the α-fragment (Fig. 1B). Initial experiments focused on the maltose-binding protein (MBP) of Escherichia coli. Although MBP is normally secreted into the periplasm of E. coli, the construct used here folds in the cytoplasm where the ω-fragment is located. Escherichia coli harboring the expression constructs were plated on isopropyl-β-D-thiogalactoside (IPTG)/X-gal indicator plates, and the development of blue color was monitored. The most intensely blue colonies (Fig. 2) were pUC19-transformed43 DH5α E. coli, which express a 54-residue α-fragment (residues 6–59 of β-gal). This represents the level of β-gal complementation attributable to the α-fragment. The MBP/α fusion protein (MBP, residues 1–366; α, residues 7–58 of β-gal) also yields significant α-complementation, though less than the α-fragment alone. Mutation of two residues, G32D and I33P, 131


RESEARCH ARTICLES


A

A

B

B Figure 1. An in vivo solubility assay based on structural complementation. (A) Schematic depicting the complementation solubility assay. p (gray squares) represents the target protein, and α (white triangles) and ω (black trapezoids) represent each of the complementing fragments of the tetrameric β-galactosidase (β-gal). Brackets indicate the concentration dependence of complementation reflecting the availability of soluble (folded) target/α fusion. Kd is indicated solely to highlight the concentration-dependent equilibrium association/dissociation reaction. (B) The target protein/α-fragment C-terminal fusion expression construct (α-fragment, residues 7–58 from full-length β-gal). HA indicates the position of the inserted influenza hemagglutinin (HA) epitope tag (residue sequence YPYDVPDYA) present in some of the constructs examined (see text).

decreases the folding yield of MBP by more than 100-fold44. This double mutation was introduced into the MBP/α fusion construct, and monitored for α-complementation (Fig. 2). Consistent with the effect of these mutations on the in vivo solubility of MBP, the G32D/I33P double mutation reduced solubility (Fig. 3) and complementation of β-gal activity (Fig. 2). To test the generality of the assay system, we generated additional α-fusion constructs. Expression of fusions of α-fragment to either thioredoxin (TRx) or glutathione S-transferase (GST), two highly soluble proteins that can improve the solubility of ill-behaved partners45,46, results in blue color similar to the MBP/α fusion construct (Fig. 2). Nucleotide-binding domains (NBD) from ATP-binding cassette (ABC) transporters were examined, including the first NBD of the cystic fibrosis transmembrane conductance regulator (CFTR): NBD1-B (residues 404–644), and NBD1-D (residues 419–655). Mutations within this domain prevent proper folding of full-length CFTR in vivo and, thus, lead to cystic fibrosis. Another NBD, MJ1267, is a subunit of the branched-chain amino acid transporter from the hyperthermophilic archaeon Methanococcus jannaschi. It has been shown that CFTR NBD1 forms insoluble inclusion bodies when expressed in E. coli (Fig. 3C)47 unless fused to a soluble protein such as wild-type MBP (ref. 48) or GST (ref. 46). However, MJ1267 yields ∼30% soluble protein from a T7 expression system in BL21 E. coli (Fig. 3C). When expressed in DH5α on indicator plates, both CFTR NBD/α fusions result in little blue color after 48 h of growth (Fig. 2), although the NBD1-D/α fusion complements more than NBD1-B/α. Expression of the MJ1267/α fusion results in a significantly elevated level of blue color when compared to either of the CFTR NBD/α fusion proteins (Fig. 2). As a group, the NBD/α 132

C

Figure 2. Colony color correlates with fusion protein solubility. (A) A representative colorimetric agar plate assay of E. coli DH5α induced to express (see text) each of the indicated α-fusion proteins. “Stop” indicates the presence of an in-frame stop codon between MJ1267 and the α-fragment in the negative (no α-fragment) control construct. The asterisk denotes colonies incubated for 48 h at 37°C (all others were grown 18 h). (B) High-resolution, magnified sectors from the plates presented in (A) are shown to highlight the color differences produced by each of the constructs in the agar plate assay at the single-colony level. (C) The complementation assay is adaptable to a 96-well plate assay format, a configuration well suited to rapid-throughput screening for compounds that affect target solubility.

fusions express (under pTac promoter control) at a lower level than the MBP/α fusion proteins, and thus, produce less activity. It should be noted that relative levels of α-complementation, as evidenced by blue color on indicator plates, can be observed at the single-colony level for each of the constructs tested (Fig. 2B), providing a measure that is independent of plated cell density. The constructs were also analyzed in a 96-well plate; the levels of blue color obtained in the microtiter plate assay (Fig. 2C) agrees with that obtained in the agar plate assay (Fig. 2A, B), indicating that the assay is adaptable to a high-throughput format. To verify the hypothesis that the blue color reports target protein solubility, the amount of soluble versus insoluble protein was measured in fractionation experiments. Escherichia coli expressing wild-type or G32D/I33P MBP/α fusions were disrupted and fractionated. Analysis of the soluble and insoluble fractions for each fusion protein (Fig. 3A) revealed a strong correlation between solubility and blue color on X-gal plates (Fig. 2). The wild-type MBP/α fusion fractionates primarily to the supernatant, whereas the double mutant (G32D/I33P) fractionates primarily to the pellet (Fig. 3A). These results indicate that the agar plate β-gal assay is most NATURE BIOTECHNOLOGY VOL 19 FEBRUARY 2001

http://biotech.nature.com



RESEARCH ARTICLES sensitive to changes from insoluble to partial solubility, the range of greatest practical utility. The fraction of MBP/α fusions that are soluble is in agreement with the previously published stability and folding yield of these mutants without the α-fragment44, suggesting that fusion of the α-fragment does not appreciably influence the relative solubility characteristics of the MBP fusion proteins. Similarly, the blue color observed for the GST/α and TRx/α fusions (Fig. 2) correlates well with the partitioning of these proteins to the soluble fraction (Fig. 3B). A correlation between solubility and α-complementation was also demonstrated for the NBD/α fusion constructs. Both of the CFTR NBD/α fusion proteins exhibit little blue color (Fig. 2), and all measurable fusion protein partitions to the insoluble fraction whether expressed with (DH5α expression) or without (BL21 expression) the α-fragment (Fig. 3C, and data not shown). In contrast, MJ1267, when expressed as an α-fragment fusion, produces a significant blue color relative to either of the CFTR-NBD/α fusions (Fig. 2). This correlates with the partial solubility of MJ1267 either with (DH5α expression) or without (BL21 expression) the α-fragment (Fig. 3C). Taken together, these results suggest that in these cases, the α-fragment does not have large effects on the target’s solubility; it neither increases that of the otherwise insoluble CFTRNBDs nor decreases that of the partially soluble MJ1267. Four MBPs were utilized to establish the quantitative relationship within a target system between β-gal activity in cell lysates and biochemical solubility. Table 1 summarizes the results of in vitro enzyme assays. Activity correlates well with the blue color observed for the constructs in Figure 2. The plate assay is, however, unable to distinguish soluble targets from those of intermediate solubility (MBP single mutants), most likely because of integration of the signal during growth of the colonies. Figure 4 shows the linear relationship between the activity (Table 1) and both the soluble fraction of the MBP/α fusions as assessed by densitometry of Coomassiestained gels (Fig. 3A) and the reported periplasmic folding yields for the unfused MBPs44. To test whether the structural complementation assay has application to proteins involved in disease-related misfolding/aggregation9,13,23, the Alzheimer’s Aβ (1–42) peptide, which forms insoluble fibrils in the brains of affected individuals, was selected as an additional test case. When fused to the α-fragment and expressed in E. coli, the fusion protein is unable to complement β-gal activity (Fig. 5A). In contrast, mutation of Aβ phenylalanine 19 to proline (F19P), a mutation that retards fibril formation in vitro49, results in a clear and measurable increase in blue color (Fig. 5A), approximately a threefold increase in β-gal activity (Fig. 5B), and increased fusion protein in the soluble fraction (Fig. 5C). Also analyzed was a highly aggregation-prone tandem repeat of Aβ (ref. 50) as a fusion with the α-fragment (Aβ-rpt). Colonies expressing the Aβ-rpt/α fusion proTable 1. Assay of β-galactosidase complementation Target protein

β-Gal activity (units/cell)a

MBP wild type G32D I33P G32D/I33P GST TRx CFTR NBD1-B CFTR NBD1-D MJ1267 NBD

102 ± 19 94 ± 21 46 ± 12 14 ± 3 134 ± 8 159 ± 14 5±1 6±2 12 ± 6

aA unit of β-gal activity is defined as the amount of enzyme required to hydrolyze 1 µmol of ONPG to o-nitrophenol and D-galactose per minute. Note that the polylinker between MBP (and mutants thereof) and the α-fragment is 36 residues in length. This polylinker was reduced to 9 residues during construction of the CFTR-, MJ1267-, GST-, and TRx-α fusion constructs.

NATURE BIOTECHNOLOGY VOL 19 FEBRUARY 2001 http://biotech.nature.com

A

I

S

P MBP wild type G32D I33P G32D/I33P

B

I

S

P TRx GST

U

C BL21 DH5α BL21 DH5α

I

S

P

} CFTR NBD1-D } MJ1267

Figure 3. Target protein solubility. (A) Representative Coomassiestained gels showing the induced (I), the soluble (S, supernatant), and the insoluble (P, pellet) fractions from each of the indicated MBP/α fusion constructs. Cells were lysed and subjected to separation of the soluble and insoluble fractions as described in the text. Analysis of the fractions was performed by Tricine SDS–PAGE through 10% gels. (B) DH5α induced to express either the TRx/ or GST/α-fragment fusion proteins were analyzed by biochemical fractionation and gel electrophoresis as in (A). Representative western blots of HA immunoreactivity visualized by ECL are presented. (C) Representative results from biochemical fractionation experiments for two NBD/α fusions (CFTR-NBD1-D and MJ1267) are presented. For each construct, a western blot of HA immunoreactivity derived from biochemical fractionation of HAtagged proteins expressed in DH5α under control of the pTac promoter, and a Coomassie-stained gel of untagged proteins expressed in BL21 cells under control of the stronger T7 promoter exhibit similar fractionation. Results similar to NBD1-D were obtained upon fractionation of expressed NBD1-B (data not shown). Lanes are as in (A), and the uninduced lanes (U) are included to indicate the presence (CFTR NBD1-D) or absence (MJ1267) of DH5α background bands at the same mobility as the fusion proteins.

tein exhibit no detectable blue color on indicator plates (Fig. 5A), in vitro β-gal activity less than that observed for the wild-type Aβ/α fusion (Fig. 5B), and no detectable protein in the soluble fraction (Fig. 5C). Interestingly, the Aβ-rpt protein forms a ladder of highmolecular-weight insoluble species (Fig. 5C), perhaps reflective of the disease condition. Discussion One of the most vexing problems facing a “genomics” approach to structural biology is protein misfolding and aggregation in recombinant expression systems. Continued growth of the class of diseases caused by cellular protein misfolding further motivates the development of new methods for the assessment of protein solubility in vivo. For maximum utility, such a method should provide an easily measured signal, be sensitive to subtle changes in the solubility of the target protein over a practical concentration range, allow phenotypic selection of the soluble protein, and have only systematic effects (similar both in the presence and absence of either drug or mutation) on the solubility of the target protein. The α-complementation fusion system described here seems to satisfy all of these criteria. The system reliably reports on the solubility of fused target proteins from several different classes: MBP and mutants thereof, highly soluble proteins GST and TRx, homologous yet distinct NBDs of ABC transporters, and the wild type, a mutant, and a tandem repeat of the pathogenic Aβ (1–42) peptide. The signal produced by the 133


RESEARCH ARTICLES

25

0.5

20

0.4

15

0.3

10

5

0.2

Folding yield ( )

0.6

from Betton and Hofnung, 1996

from this study

Fraction soluble (o)


A

0

0.1 0

20

40

60

80

100

120

β-gal activity

B

(units/cell)

Figure 4. Correlation of β-gal activity with fusion protein solubility and folding. The in vitro β-gal activity, measured in cell lysates (see Table 1), exhibits a linear correlation with the fraction soluble in vivo (, see Fig. 3A) and the reported periplasmic folding yield in vivo44 (), for each of the MBP/α fusion proteins.

fusions correlates well with the solubility of targets expressed without the α-fragment. Thus, the relatively small α-polypeptide (52 residues) does not have dramatic effects on the solubility of the targets. Indeed, fusion of the α-fragment to Aβ (42 residues) produces an insoluble chimera whose solubility is dominated by the smaller fusion partner (Fig. 5). Previously reported systems for monitoring in vivo protein folding rely on coupling the misfolding of an N-terminal target to the misfolding of a C-terminal marker protein such as β-gal (ref. 40), green fluorescent protein (GFP) (ref. 41), or chloramphenicol acetyltransferase (CAT)42. Therefore, these methods report on events that occur during the process of folding40, perhaps even cotranslationally41. It is unclear if these methods will detect the slow aggregation reactions that are hallmarks of many degenerative diseases. Notably, expanded polyglutamine tracts fused to the N terminus of GFP form fluorescent inclusions51,52, indicating that slow, post-translational aggregation neither prevents formation of the GFP fluorophore, nor induces its inactivation. In contrast, the α-complementation assay relies on an equilibrium association reaction, which should reflect the concentration of α-fusion available at any given time (see Fig. 1A). Indeed, immediately following induction (