Selection of functional Barnase Inhibitors from
"Barstars" with randomized hydrophobic Cores

Robert W. Hartley*

Published on-line: March 12, 2008

Key words

Protein folding, protein design, hydrophobic core, barstar


Barstar, the specific inhibitor of the bacterial ribonuclease barnase, is a small protein with a relatively large and compact hydrophobic core. The core is composed of the sidechains of 22 of its 89 residues and is almost completely surrounded by elements of secondary structure, a three-stranded parallel β-sheet and four α-helices. To what extent does the arrangement of the secondary structure depend on the precise composition of this core? Selection from a large synthetic gene library, with the entire core replaced by a random selection of hydrophobic sidechains, has yielded a number of functional barstars. This suggests that designed novel secondary structure frameworks could be filled in similar fashion.


Does the limited number of basic folds found in natural proteins imply that no other folds are possible, or as useful? Or is it simply that evolutionary success has not required the invention of other folds? A clear answer to this question could come from the design and construction of well-behaved proteins with folds unlike anything found in nature. How hard will this be? To start, for a well structured soluble protein, we want mostly hydrophilic sidechains on the outside and hydrophobics on the inside. This requirement implies that the distribution of hydrophobic residues along the amino acid chain must be one of the major determinants of the fold, as the precise fold determines which sidechains are pointed inwards from the backbone. To design a new protein, one could build an external frame of secondary structural units, helices and sheets, each with one hydrophobic side, borrowed from known structures but arranged in a novel fold. With the interior hydrophobic residues randomized at the level of the synthetic gene, the interior three-dimensional jigsaw puzzle is then solved by screening a large library of peptide products for well-folded molecules. Screening methods could include phage or ribosome display, by binding to monoclonal antibodies raised against elements of the exterior secondary structure as they appeared on the structures from which the were borrowed, or simply by selecting for protease resistance. Matsuura and Pluckthun [1] recently reported the start of a systematic program to answer the questions posed at the start of this paragraph, using similar procedures, starting with simple structures and building up to more complexity. The results reported here support the idea that novel protein creation is feasible.

According to Axe et al. [2], when all of the core residues of barnase were replaced by random hydrophobic residues, 20% of the products were, like the wild type, lethal when expressed in E. coli. Gassner et al [3] have reported that all ten residues of the major hydrophobic core of T4 lysozyme can be changed to methionine one at a time without ever losing all of its activity and the completely substituted molecule retained the basic fold of the wild type. In a methods paper [4], I reported briefly on the isolation by phage display of genes for functional barstar with a compact eight-residue portion of its core so randomized. Of these, a substantial fraction (22/71), when replacing the wild type in an expression plasmid, produced titers of activity comparable to the wild type. 200 unselected clones had been all negative. Nor was I able to find any active barstars using the two-plasmid selection system described below. We have, then, three reports of successful refilling of hydrophobic cores of known structures. For the lysozyme, yield was 100% (one out of one), for barnase, 20%, and for barstar, less than 1%. For the lysozyme experiments, methionine had been chosen because of its flexibility and its size. In a preliminary experiment, replacement of the entire 22-residue core of barstar with methionines yielded no activity.

The choice of barstar [5] for this investigation was suggested by its small size, the relatively large size and compactness of its core, Fig.1, and a report [6], that many of its core residues have some freedom of motion. An alignment of barstar with nine homologous proteins [7] confirms that the hydrophobicity of the residues that align with barstar's core is strongly conserved, although at no position does the same amino acid occur in all ten.


Figure 1: "Front" and "back" views of barstar. Sidechains of the hydrophobic core are represented as yellow space-filling spheres, the rest of the molecule as red ball and stick. (Rasmol)

Methods and Results

All of the methods outlined here have been described in detail elsewhere [4, 8]. Total synthesis of the barstar gene was carried out with the codons for all 22 of the hydrophobic core residues randomized to yield Leu, Val, Ile, Met or Phe. In brief, preparation of the gene library began with the synthesis, using a Beckman 1000M synthesizer, of two long oligonucleotides of opposite sense and overlapping 3' ends (Fig.2) which were annealed together and extended with the Klenow polymerase. This product was amplified by polymerase chain reaction (PCR), using oligonucleotides (Fig.2) which completed the barstar sequence and supplied EcoR1 and HindIII restriction sites. The number of independent sequences originally produced was estimated by semi-quantitative PCR to be on the order of 108.

Long oligonucleotides for initial synthesis

                           3'- CTCATGGGCNASCAANASCTTNASTCCGTC

Oligonucleotides to complete sequence by PCR



Figure 2. Long oligonucleotides prepared for synthesizing the gene for barstar, with the codons for residues of the hydrophobic core randomized to yield only leucine, isoleucine, valine, phenyladenine or methionine, showing overlap of their 3' ends. Also shown are the oligonucleotides used for PCR.

N = A,T,C or G S = C or G.

The library was first ligated into the plasmid pMT3020, which carries chloramphenicol resistance, and transformed into E. coli. The chloramphenicol resistant bacteria were tested by transformation by the compatible plasmid pMT816 [4,8]. This plasmid carries ampicillin resistance and an active barnase gene, making it lethal without concomitant synthesis of an active inhibitor. No successful transformants were found. This implies that functional barstars in the library are rare, estimated at less than about one in 104.

Genes for functional barstars were then selected by phage display. The EcoR1 and HindIII restriction sites allowed insertion of the library into phage T7Select 10 (Novagen, Inc.) This phage displays multiple copies of the protein coded by the inserted gene on the phage surface. The product phage DNA was packaged into phage according to the supplier and amplified in E. coli. Selection of phage displaying active barnase inhibition was achieved by passing the spent E. coli medium through a small metal chelate column carrying barnase bound by a 10 histidine peptide attached to its N terminus. Eluted in 1% sodium dodecyl sulfate, the phage were still fully infective. Several cycles of this selection process were necessary to usefully concentrate phage carrying genuine barstar from a one in a million mix with wild type phage. The library was carried through eight such cycles.

The prospective barstar genes from the selected phage were then transferred to the plasmid vector in E. coli as before and tested individually for barstar activity and by transformation with the compatible barnase-bearing plasmid. About half of these transformations were successful, and only the transformable clones produce barnase inhibitor measurable in extracts (9). The yields from these clones, however, were about two orders of magnitude less than obtained from the wild type barstar gene. Activity of barnase decreased linearly to zero with addition of each of the active extracts, an indication of stable complex formation. Stability of three of the extracted inhibitors, numbers 13, 14 and 15 in Fig.3 and a wild-type barstar control, were tested by protease digestion. 50 ng/ml of each inhibitor was mixed with 20 ng/ml of Proteinase K in 0.2 M TrisHCl buffer, pH7.0 at 30° C. Barnase inhibition in each of the four disappeared with a half life between 95 and 105 minutes.

The inhibitor gene was sequenced in 33 of the plasmids which had tested positive. The identities of the 22 core residues of the twenty different sequences found are shown in Fig.3. Eleven sequences were found twice and three, three times, suggesting that the library contained no more positive sequences than a small multiple of the twenty found. Since the number of possible sequences is 522 = 2.4 x 1015, the total number of sequences that would yield functional inhibitors is on the order of 108 or 109.


From Fig.3 it can be seen that, except for leucine in the penultimate position 88, the residues found in the wild type are not strongly favored. Indeed, almost all positions can be occupied by a residue of any size, from the smallest, valine, to the largest, phenylalanine. For the twenty sequences, the range of occurrence for each is 5 to 13 for leucine, 2 to 10 for valine, 0 to 5 for isoleucine, 0 to 5 for phenylalanine and 1 to 6 for methionine. The serine in number 10 is presumably from a PCR copy error. The NTS codon (Fig.2) represents eight actual codons, three for leucine, two for valine and one each for isoleucine, phenylalanine and methionine. If we ignore residue at position 88 which is 90% leucine, and add up each of the other residues in the twenty sequences, the ratios (L : V : I : F : M) are 3.00 : 2.08 : 0.91 : 0.95 : 1.03, surprisingly close to the 3 : 2 : 1 : 1 : 1 expected for a completely random selection. Note also, however, the numbers in the right hand column of Fig.3. These represent the number of non-hydrogen (C or S) atoms in the sidechains of each core, approximately equivalent to the relative volumes they take up. The smallest, at 78, is 15% less than the wild type's 92. Unless the cores can maintain much more empty space than seems likely, this suggests that considerable distortion of the outside framework must be allowed. The three apparently much larger cores each have five phenylalanines and the flipping out of one of these would reduce the effective core volume. For barstar function, what must be maintained is only the configuration of the helix and adjacent loop that actually form the barnase binding site. It seems likely, however, that this would require something fairly close to the natural fold of wild type barstar.


Figure 3: Compositions of the hydrophobic cores of the twenty functional barstars found in this study. The right hand column lists the number of non-hydrogen atoms in each.

It is not clear why the yields of the functional barnase inhibitors were all so small. The Proteinase K experiment suggests that the stability of the inhibitors is not at fault but perhaps slower folding makes these molecules vulnerable during synthesis. It is disappointing that no inhibitors were obtained with yields easily compatible with crystalography or other physical studies. Such success might come from a similar effort using the larger libraries obtainable by the more difficult technique of ribosome selection. Application of this kind of study to membrane proteins and insoluble structural proteins will require more thought.


Functional barnase inhibitors can be selected from a library of synthetic barstars in which all 22 residues of the hydrophobic core have been replaced at random by any of five hydrophobic residues. It appears that the twenty independent solutions to the three dimensional jigsaw puzzle are quite different, suggesting that any sizable arbitrary space, as in a designed novel protein, could be filled in similar fashion.


Matsuura T and Pluckthun A: Strategies for selection from protein libraries composed of de novo designed secondary structure modules. Orig. Life Evol. Biosp. 2004, 34:151-157.

Axe DD, Foster, NW and Fersht AR: Active barnase variants with completely random hydrophobic cores. Proc. Natl. Acad. Sci. USA 1996, 93:5590-5594.

Gassner, N.C., Baase, W.A. and Matthews, B.W. 1996. A test of the "jigsaw puzzle" model for protein folding by multiple methionine substitutions within the core of T4 lysozyme. Proc. Natl. Acad. Sci. USA 93: 12155-12158.

Hartley RW: Barnase-barstar interaction. Methods Enzymol. 2001, 341:599-611.

Hartley RW: Barnase and barstar. in Ribonucleases: Structures and Functions. Edited by D'Alassio G and Riordan JF. New York: Academic Press,. 1997:51-100.

Wong K-B and Daggett V: Barstar has a highly dynamic hydrophobic core: evidence from molecular dynamics simulations and nuclear magnetic resonance relaxation data. Biochemistry 1998. 37:11182-11192.

Krajcikova D and Hartley RW: A new member of the bacterial ribonuclease inhibitor family from Saccharopolyspora erythraea. FEBS Letters 2004, 557:164-168.

Jucovic M and Hartley RW: Protein-protein interaction: A genetic selection for compensating mutations at the barnase-barstar interface. Proc. Natl. Acad. Sci. USA 1996, 93:2343-2347.

Hartley RW, Rogerson DL and Smeaton J R: Production and purification of the extracellular ribonuclease (barnase) and its intracellular inhibitor (barstar). II. Barstar. Prep. Biochem. 1972, 2:243-250.

An author's note on the web publication of this paper.

Rejection letters and referees' comments from several journals.