ARTICLE Expression of soluble heterologous proteins via fusion with NusA protein Roger G. Harrison — School of Chemical Engineering and Materials Science, University of Oklahoma, Norman, Oklahoma he production of proteins by ge- to the N-terminus of the target protein (5). interleukin-3 (hIL-3) compared to the netically engineered Escherichia Other carrier proteins that have been widely fusion of TRX to hIL-3. NusA expressed as coli bacteria cells has become used in fusion proteins in E. coli are maltose a fusion with hIL-3 gave the highest level of well established in biotechnol- binding protein (MBP) and glutathione S- solubility and hIL-3 expression of the car- Togy research and large-scale production. transferase (GST). MBP and GST were rier proteins tested. NusA fused to human Cloning and expression of proteins in chosen as carriers because they enable the interferon-γ (hIFN-γ), bovine growth hor- E. coli are favored in many instances be- fusion protein to be affinity purified: MBP mone (bGH), or tyrosinase from Rhizobium cause E. coli has relatively simple genetics, is binds to amylose, while GST binds to im- meliloti was also expressed in nearly com- well characterized, and has a relatively rapid mobilized glutathione. While these carrier pletely soluble form and at high levels. This growth rate. One disadvantage, however, of proteins have resulted in the successful over- article elaborates on these new findings. expressing heterologous proteins in E. coli is expression of many heterologous proteins in Recombinant protein solubility that the proteins are frequently expressed as E. coli, each was discovered empirically and modeling insoluble aggregated folding intermediates, certainly may not possess maximal solubiliz- known as “inclusion bodies.” In order to re- ing characteristics. In the original model of Wilkinson and cover an active protein expressed in inclu- The new fusion protein systems that we Harrison (6) for the solubility of recombi- sion bodies, the protein that is insoluble describe here are based on a systematic eval- nant proteins expressed in E. coli, five must be solubilized with denaturants such uation to identify E. coli proteins that have amino acid-based parameters were corre- as 8 M urea or 6 M guanidine hydrochlo- ride, and then the denaturant must be re- Table 1. Predicted solubilities of carrier and target proteins moved under conditions that are optimal Protein MW (kDa) Amino Acid Length Probability of for protein folding. While progress has been Solubility or Insolubility1 made in the refolding of recombinant pro- NusA 54.8 495 95% soluble teins (1, 2), the specific folding conditions BFR 18.5 158 95% soluble regarding buffer composition, protein con- GrpE 21.7 197 92% soluble centration, temperature, etc. must be opti- thioredoxin 11.7 109 73% soluble hIL-3 15.1 133 73% insoluble mized for every protein. Even for refolding bGH 21.6 189 60% insoluble processes that have been optimized, the hIFN-γ 17.1 146 58% insoluble yield of renatured protein may still be rela- tyrosinase 54.1 494 51% soluble 1. The revised Wilkinson-Harrison solubility model involves calculating a canonical variable (CV) or composite parameter for the protein for which the solubility is tively low, requiring large volumes and sig- being predicted. The canonical variable in the two-parameter model is defined as: CV=λ1 N+G+P+S +λ2 (R+K)–(D+E) nificant cost for the preparation of large ( ) |( -0.03)| nn quantities of protein. n = number of amino acids in protein N,G,P,S = number of Asn, Gly, Pro, or Ser residues, respectively. One approach to expressing heterolo- R,K,D,E = number of Arg, Lys, Asp, or Glu residues, respectively. λ1, λ2 = coefficients (15.43 and -29.56, respectively) gous proteins in soluble form has been to The probability of the protein solubility is based on the parameter CV-CV', where CV' is the discriminant equal to 1.71. If CV-CV' is positive, the protein is predicted to be insoluble, while if CV-CV' is negative, the protein is predicted to be soluble. The probability of solubility or insolubility can be predicted from the following equation: co-express molecular chaperones which aid Probability of solubility or insolubility = 0.4934 + 0.276 |(CV-CV')| - 0.0392 (CV-CV')2 in protein folding. Although there have been several instances of E. coli chaperones aiding the folding of heterologous proteins, the highest potential for solubility when lated with inclusion body formation. It has unfortunately it is a trial and error process overexpressed. A modified version of the since been discovered, however, that only to determine the specific match between the solubility model of Wilkinson and Harrison two of the parameters are critical in distin- target protein and the chaperone that will (6) for recombinant proteins expressed in E. guishing soluble versus insoluble protein lead to correct folding (3, 4). coli was used to determine the solubility po- expression (7). The critical parameters are Another approach that has achieved suc- tential of the more than 4000 E. coli pro- the approximate charge average, which cess in recent years in producing soluble teins in the SwissProt protein database. accounts for the differences in the numbers heterologous proteins in E. coli has been the Based on this solubility model, we identi- of Asp plus Glu versus Lys plus Arg use of gene fusions. The first fusion protein fied three E. coli proteins--BFR, GrpE, and residues, and the turn-forming residue con- system specifically aimed at increasing solu- NusA--that gave a significantly higher level tent, which accounts for the total number bility of a target protein was the one in of solubility when expressed as a fusion of Asn, Gly, Pro, and Ser residues. which E. coli thioredoxin (TRX) was fused with the target heterologous protein human Incorporation of the two-parameter solubil- inNovations 11 4 ity model into a custom C computer pro- 18.4%; GrpE, 7.4%; BFR, 13.7%; and able to express high levels of hIL-3. gram allowed the rapid evaluation of all thioredoxin, 8.5% (see fig. 1). The highest In order to determine if the hIL-3 present known E. coli protein sequences. As input, soluble expression level was achieved by the in each of the fusion proteins was biologi- this program used a single text file created by NusA/hIL-3 protein (97% soluble), while cally active, indicating that hIL-3 was prop- the SwissProt server (http://www.expasy.ch), the thioredoxin/hIL-3 fusion protein was erly folded, a cell proliferation assay was which contained approximately 4000 expressed almost completely in the inclu- performed on each fusion protein in soluble E. coli protein sequences. sion body fraction (8% soluble). Thus, the cell lysate. hIL-3 activity was found to be E. coli proteins identified by the solubil- percentage solubility of the NusA/hIL-3 present in all fusion proteins, with 67% of ity model which possessed a solubility prob- fusion protein was over 12 times that of the the hIL-3 determined to be active for the ability of greater than 90% were considered thioredoxin/hIL-3 fusion protein. NusA/hIL-3 fusion protein, which was the for fusion protein experiments. The list of The Western blot (see fig. 1, lower highest of the fusion proteins tested (7). possible carrier proteins was then reduced panel) shows, with more clarity, the distrib- For the expression of the NusA/bGH to those which were stable when expressed ution of hIL-3 among the soluble and insol- and NusA/hIFN-γ fusion proteins (see fig. in E. coli as shown in previous studies. uble fractions. BFR and GrpE are distrib- 2), it is clear from SDS-PAGE and Western Next, choices were further restricted to uted between the soluble and insoluble blot results that the vast majority of these those containing more than 100 amino fractions. Upon centrifugation of BFR/hIL- fusion proteins are expressed in the soluble acids to ensure that predicted solubility 3 cultures at the end of induction periods, it fraction (estimated as 89% for NusA/bGH characteristics would not be simply limited was noticed that the cell pellet had a slightly and 87% for NusA/hIFN-γ). To our knowl- by the molecular weight of the carrier pro- red tint compared to uninduced cultures. edge, this is the first evidence of soluble tein relative to the target protein chosen. This is presumably due to the iron binding expression of bGH in E. coli. Three E. coli proteins, NusA, BFR (bac- properties of BFR (12) and serves as a con- The fusion of NusA with the relatively terioferritin), and GrpE, were identified by venient marker for confirmation of protein large tyrosinase (54 kDa) was expressed in the solubility model as having a very high expression. One striking finding from the almost completely soluble form (>90%) probability (> 90%) of being soluble when Western blot data (fig. 1) is that the level of based on SDS-PAGE analysis (see fig. 3). expressed in E. coli. All three of these pro- expression of hIL-3 in the soluble fraction Thus, NusA may be a very good carrier pro- teins were evaluated as N-terminal fusions was higher in the NusA fusion than in the tein for solubilizing large target proteins. In with human interleukin-3 (hIL-3), which other fusions. Thus, the large size of NusA has been found in inclusion bodies both (55 kDa) was not at all limiting in being continued on page 6 when expressed alone (8) and as a fusion protein with thioredoxin (5). Bovine growth hormone (bGH) and human inter- NusA fusion GrpE fusion BFR fusion thioredoxin fusion M u i sol ib u i sol ib u i sol ib u i sol ib feron-γ (hIFN-γ) were selected for fusion kDa with NusA at their N-terminus because 80 – they had previously been expressed alone in 50 – inclusion bodies (9, 10).