How do leucine zippers interact with dna

This definition can be directly translated into a simple algorithm, which is assessed in this paper. First the coiled coil prediction is computed for a given sequence. Assume a sequence contains either a strict or a relaxed leucine repeat. We demand a minimal overlap of 21 residues between the region predicted as coiled coil and the leucine repeat. We need to exclude long structural coiled coil proteins that happen to have a few leucines with the right spacing.

Depending on whether the basis of the prediction is a strict or a relaxed leucine repeat, this scheme will produce two classes of predicted LZs. We searched for sequences matching the mentioned criteria for leucine repeats and coil prediction. We recorded all frames of coil predictions for all annotated and undecided LZs as well as all other sequences containing a leucine repeat. First we inspected all frames with both a leucine repeat and a coiled coil prediction.

In the following, sequences containing at least one zipper according to the above-mentioned criteria are termed predicted zippers, while all others are called predicted non-zippers.

To keep figures consistent we decided in the statistics below only to count each such sequence once, even if more than one frame with a prediction was found. In an initial assessment we referred to the database annotations to verify our results. Table 1 summarizes the distribution of annotations among the predicted zippers and non-zippers. Overall sequences contain a leucine repeat frame, a zipper annotation or both.

From this base set, leucine repeat frames overlap sufficiently with a coil frame to be predicted as LZs. Nearly half of these require the relaxed leucine repeats to be detected. Based on annotation these would be the false positives when one combines predictions based on relaxed and strict leucine repeats.

Most of these false positives stem from admitting relaxed leucine repeats. A summary is given in Table 2. This makes it highly likely that our prediction is in fact correct. Since coupling of a LZ and a BR domain is frequently observed it appears that this protein too might indeed contain a LZ. The other sequences, 17 eukaryotic and 8 prokaryotic proteins, may well be real false positives since, to our knowledge, there is no reason to assume that any of them might have a LZ.

It is noteworthy that only two of the false negatives are short fibrous proteins missed by the filtering procedure. Some sequences that would have been very tempting to predict as containing a LZ are not in our false positive list. An example is DNA topoisomerase 2. It binds DNA and has both a leucine repeat and a coiled coil, but without overlap of the two corresponding frames Incidentally, this example was used by Hirst et al.

In our data set all four occurrences of DNA topoisomerase 2 are classified as non-zippers. Predicted and non-predicted zippers and their annotations see text for explanation. For nearly sequences our method does not predict a LZ. Most sequences are rejected by our method, in accordance with the annotation. For these sequences consideration of the coiled coil prediction really helps in rejecting a zipper hypothesis which is based on a leucine repeat alone.

There are 31 sequences where prediction failed. Table 3 gives a compilation of these instances that are, according to the annotations, false negatives. Interestingly, all of them lack a coil prediction. Only three had a strict repeat pattern; eight had a relaxed one.

All others had a repeat that was even more mutated, mostly with the variable residues from the relaxed pattern, Met, Ile and Val. Eight have a basic region which suggests that they are bZIP proteins and were missed by our procedure. On the other hand, in seven instances the BR does not have the right spacing such that coincidence or other functions cannot be excluded.

For three of the BR-containing sequences that were annotated as containing a LZ, this zipper was shorter than four heptads, which is commonly viewed as the minimum number for a stable dimer.

Four had two of the five leucines replaced by other residues and two had one substitution with non-canonical residues Ala or Tyr. Another five belong to the Myb class of eukaryotic transcription factors, where two of the five leucines are substituted.

Serious concerns must be raised about the zipper nature of their annotated regions. The proposed zipper region is far from the DNA binding domain and thus does not resemble the known architecture of other zipper-containing transcription factors.

Interestingly, for eight of the 31 sequences the corresponding references do not mention a LZ. For some putative instances it appears unlikely they are LZs because of their biological function or their sequences, or both. Others are either membrane proteins or their leucine repeat is at the very N-terminus, which is never observed for generally accepted LZs. Some are definitely annotated as not binding to DNA.

This of course does not exclude the existence of a LZ. Due to the problems with database annotations, we searched for other criteria to recognize whether a sequence might contain a LZ.

Therefore, we combined our criteria with the use of regular expressions to search for these regions adjacent to or in the N-terminus of transcription factors. When order and spacings of the motifs i. We refer to such zippers that co-occur with a basic region or a bHLH domain as confirmed zippers. LZs may also be associated with homeodomains in some homeobox families Since, however, these cases seem to be restricted to plants e.

Arabidopsis thalliana , hat1-thal, hat2-… and the structural role of the associated LZ is somewhat unclear, we do not refer to such motifs. Table 4 shows the distribution of sequences with a co-occurrence of both domains among the predicted zippers and predicted non-zippers.

With one exception, all predictions of LZs in confirmed zippers were correct when the BR motif was used for evaluation. These results constitute further confirmation of our prediction method and strengthen the view that combining a relaxed leucine repeat and a coiled coil prediction are a good strategy to identify LZs.

Confirmed and predicted zippers: occurrences of sequences with additional domains, indicative of eukaryotic transcription factors bZIP, bHLH-LZ with the correct spacing another eight annotated zippers had a BR but not with the correct spacing. In the following, we evaluate the positional coil probability for some of the confirmed LZs.

This has important structural implications because it was reported that the basic region of bZIP proteins shows a slight coiled coil probability 4. This suggests that at least the hydrophobic interface would be similar in a bundle and a zipper and the border between the domains difficult to detect. Finally, several zipper motifs were reported to be relatively unstable.

For the case of Myc-Myc dimers, neither the complete dimer nor the zipper fragments alone are stable and also the dimer of the Myc-Max LZs alone is less stable than the LZ dimer of bZIP proteins alone Most probably the LZ motif serves primarily for recognition and not so much as a stabilizing domain 9. There are, however, significant deviations for Jun and Myc.

The BR of Jun shows a very high coil probability. This is interesting considering that Jun, unlike Fos, can form homodimers and thus the coil region may be extended in the absence of DNA. Further, it can be clearly seen that the LZs of Jun and Fos are in fact six heptads long instead of four as is mostly annotated. In both cases the motif is characterized by a His instead of the sixth Leu.

With Fos the drop-off sharpens because of a Pro at the very end of the six heptads. Myc sequences indeed show widely varying coil probabilities. This may be due to a smaller contribution to dimer formation and strengthens the view that the LZ in bHLH-LZ proteins mediates specificity rather than stability. All frames are aligned such that the leucine repeats within a family overlap. Averages are given as thick dashed lines. We have presented a simple computational approach for identification of LZs by combining a standard coiled coil prediction algorithm with an approximate search for the characteristic leucine repeat.

Another goal of our study was a systematic investigation of the co-occurrence of leucine repeats and a detectable coiled coil in LZs, in particular for eukaryotic transcription factors. To avoid the pitfalls of wrong annotations we use additional biological signals, such as DNA binding motifs, for verification.

In summary, we find the following conclusions of both practical relevance and general interest. First of all we designed a fast and easily applicable strategy to predict LZs. Specifically, for eukaryotic transcription factors the method has excellent accuracy.

Also, for sequences with no or few known homologues it will prove useful to decide whether the protein may dimerize through a LZ or not. Hirst et al. We show that such a criterion is a major source of false negatives. Finally, it is interesting to observe that the biological needs for flexibility or alternative dimerization are well reflected by different coil probabilities. This is an intriguing observation because there are no thermodynamic or other biophysical considerations directly included in the coil prediction heuristics.

Our work also pinpoints some basic problems in the fields of structure prediction and motif recognition. The need for flexibility in biological activity frequently results in marginal stability, which in turn leads to fuzzy rules of recognition.

This undermines efforts at clear and precise definitions and sharp discrimination, e. Thus one may also conclude that there is no specific code for a LZ, but LZs are simply short parallel dimeric coiled coils with as much additional regularity as needed for proper function, such as orientation and flexibility.

We conclude that since a coiled coil prediction seems to be a more reliable indicator for a LZ, the hallmark of a LZ is rather the coiled coil than the leucine repeat. A prediction strategy making use of both features has a surprisingly high success rate, yet an ultimate classification of such proteins can only be achieved by homology comparison and structural information. Google Scholar. Google Preview. Oxford University Press is a department of the University of Oxford.

It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide. Sign In or Create an Account.

Sign In. Advanced Search. Search Menu. Article Navigation. Close mobile search navigation Article Navigation. Volume Article Contents Abstract. Materials and Methods. Computational approaches to identify leucine zippers. Deutsches Krebsforschungszentrum, Theoretische Bioinformatik. Oxford Academic.

Eric Rivals. Martin Vingron. Select Format Select format. Acta cryst. Johnson, P. Biochem 58 , — McCormack, K. Nature , White, M. Buckland, R. Turner, R. Schuermann, M. Cell 56 , — Hollis, M. Murre, C. Cell 58 , — Wright, W. Ptashne, M. Creighton, T. Proteins Freeman, New York, Google Scholar. Download references. You can also search for this author in PubMed Google Scholar. Reprints and Permissions. Abel, T. Action of leucine zippers. Nature , 24—25 Download citation.

Issue Date : 07 September Anyone you share the following link with will be able to read this content:. Sorry, a shareable link is not currently available for this article. Provided by the Springer Nature SharedIt content-sharing initiative.

BMC Biochemistry Cell and Tissue Research Nature Advanced search. Skip to main content Thank you for visiting nature. Access through your institution.