Designing Proteins by Using Randomly Generated Genes Studies of the amino acid sequence and corresponding three-dimensional structure of wild-type or mutant proteins have led to significant insights into the principles that govern protein folding. An important test of this understanding would be to design a protein based on these principles and see whether it folds as expected.
Kamtekar and colleagues (1993) used aspects of the genetic code to generate random protein sequences with defined patterns of hydrophilic and hydrophobic residues. Their clever approach combined knowledge about protein structure, amino acid properties, and the genetic code to explore the factors that influence protein structure.
The researchers set out to generate a set of proteins with the simple four-helix bundle structure shown below, with $\alpha$ helices (shown as cylinders) connected by segments of random coil (light red).
Each $\alpha$ helix is amphipathic-the $R$ groups on one side of the helix are exclusively hydrophobic (yellow), and those on the other side are exclusively hydrophilic (blue). A protein consisting of four of these helices separated by short segments of random coil would be expected to fold so that the hydrophilic sides of the helices face the solvent.
(a) What forces or interactions hold the four $\alpha$ helices together in this bundled structure?
Figure $4-4$ a shows a segment of $\alpha$ helix consisting of 10 amino acid residues. With the gray central rod as a divider, four of the $R$ groups (purple spheres) extend from the left side of the helix, and six extend from the right.
(b) Number the $R$ groups in Figure $4-4 a,$ from top (amino terminus; 1 ) to bottom (carboxyl terminus; 10 ). Which $R$ groups extend from the left side and which from the right?
(c) Suppose you wanted to design this 10 amino acid segment to be an amphipathic helix, with the left side hydrophilic and the right side hydrophobic. Give a sequence of 10 amino acids that could potentially fold into such a structure. There are many possible correct answers.
(d) Give one possible double-stranded DNA sequence that could encode the amino acid sequence you chose for (c). (It is an internal portion of a protein, so you do not need to include start or stop codons.)
Rather than designing proteins with specific sequences, Kamtekar and colleagues designed proteins with partially random sequences, with hydrophilic and hydrophobic amino acid residues placed in a controlled pattern. They did this by taking advantage of some interesting features of the genetic code to construct a library of synthetic DNA molecules with partially random sequences arranged in a particular pattern.
To design a DNA sequence that would encode random hydrophobic amino acid sequences, the researchers began with the degenerate codon NTN, where $\mathrm{N}$ can be $\mathrm{A}, \mathrm{G}, \mathrm{C}$ or T. They filled each N position by including an equimolar mixture of $\mathrm{A}, \mathrm{G}, \mathrm{C},$ and $\mathrm{T}$ in the DNA synthesis reaction to generate a mixture of DNA molecules with different nucleotides at that position (see Fig. $8-32$ ). Similarly, to encode random polar amino acid sequences, they began with the degenerate codon NAN and used an equimolar mixture of A, $G,$ and $C$ (but in this case, no
T) to fill the N positions. (e) Which amino acids can be encoded by the NTN triplet? Are all amino acids in this set hydrophobic? Does the set include all the hydrophobic amino acids?
(f) Which amino acids can be encoded by the NAN triplet? Are all of these polar? Does the set include all the polar amino acids?
(g) In creating the NAN codons, why was it necessary to leave T out of the reaction mixture?
Kamtekar and coworkers cloned this library of random DNA sequences into plasmids, selected 48 that produced the correct patterning of hydrophilic and hydrophobic amino acids, and expressed these in $E$. coli. The next challenge was to determine whether the proteins folded as expected. It would be very time-consuming to express each protein, crystallize it, and determine its complete three-dimensional structure. Instead, the investigators used the $E$. coli protein-processing machinery to screen out sequences that led to highly defective proteins. In this initial screening, they kept only those clones that resulted in a band of protein with the expected molecular weight on SDS polyacrylamide gel electrophoresis (see Fig. $3-18$ ).
(h) Why would a grossly misfolded protein fail to produce a band of the expected molecular weight on electrophoresis?
Several proteins passed this initial test, and further exploration showed that they had the expected four-helix structure.
(i) Why didn't all of the random-sequence proteins that passed the initial screening test produce four-helix structures?