University of Central Florida Undergraduate Research Journal - Computational Analysis of Broad Complex Zinc-Finger Transcription Factors
US tab

Computational Analysis of Broad Complex
Zinc-Finger Transcription Factors

By: Barbara Mascareno-Shaw
Mentor: Dr. Thomas Selby


"Homology modeling is therefore the best available tool to help analyze differences in proteins.  However, the question of reliability and biological relevance always follows any type of homology analysis." 

With regard to the computational analysis, these results demonstrate significant interaction energy differences for most of the proteins.  For the Z1 and Z3 proteins, both the Z1:R57C and Z3:T58M showed increases in the interaction energy (more stability) relative to the wild type models.  In the case of Z3:S55L there was very little difference, which can be rationalized by a structural analysis that shows amino acid position 55 is outside the DNA binding region.   In the case of Z2/Z4, the model appeared to be somewhat unreliable due to a decrease in binding energy of nearly 50% relative to the template interaction energy.  In this case a more suitable template structure is needed to better analyze this sequence. 

Modeling sequences are certainly not the best means of analyzing mutations in proteins.  Ideally, an experimentally determined structure would be available for every protein sequence known, but this is obviously not possible due to experimental constraints.  Homology modeling is therefore the best available tool to help analyze differences in proteins.  However, the question of reliability and biological relevance always follows any type of homology analysis.  Our study answers these questions, but in a slightly indirect manner.   

In the case of the Z1 wild type and mutant, the greatest energy difference in the binding capacity of the protein to the substrate is caused by a substitution of a longer carbon chain to a shorter one.  However, our data suggest that this change is not due directly to the amino acid change alone, but due to changes in surrounding amino acids that are also repositioned.  Our findings correlate to the fruit fly genetic function; the Z1 mutant lacks the rbp function and deters the ability to reduce the bristles on the palp of the fruit fly.  An interesting point to make is that the interaction energy increases (providing a more stable structure with DNA), and still causes loss of rbp function.  This interaction may initially appear counter intuitive, but consider the following:  the Z1:R57C mutation causes the interaction energy to nearly double.  At the same time, the Z1 model shows only a 7% change relative to the template, demonstrating that the Z1 model is very reliable.  These two facts show that tighter binding may be the cause of loss of function.

Transcription factor binding is not meant to be permanent.  Once the transcriptional machinery is assembled, the proteins involved in RNA synthesis must move along the DNA template to allow transcription events to occur efficiently.   Thus if a transcription factor is binding too tightly, transcription may be inhibited.  This appears to be the case for Z1:R57C mutant.  Based on the available data, we propose that this mutant inhibits transcription, not through loss of DNA binding, but by binding too tightly and not allowing the required RNA synthesis to occur. 
By contrast, the Z3 mutants do not show the same effect.  The Z3:S55L mutant does not show a decrease in the interaction energy with DNA.  In relation to the morphology of the fruit fly, mutations in this protein would not be expected to alter the 2Bc genetic function.  However, our results only consider the DNA binding interactions and not the structure of the entire protein. 
The entire Z3 protein is over 700 amino acids in length, but our DNA binding analysis only considers the region from amino acid 583 to 640 (based on Z3 numbering).  With our calculations showing no change in the DNA interaction energy, but a change in morphology of the fruit fly, it appears that this particular mutant is most likely affected by an abnormally folded 3D structure that inhibits transcription from occurring.  This mutation would therefore be classified as one causing misfolding of the Z3 protein, or loss of binding affinity for other proteins involved in the formation of the transcriptional complex. 

In the case of Z3:T55M, the mutation appears to increase DNA binding, but not as significantly as Z1:R57C.  Additionally, the effect of the mutation does not appear to alter the locations of neighboring side chains.  In this case, the mutation may have a loss of function due to both causes mentioned above: increase in DNA affinity and a change in the overall protein fold.  It is also important to reiterate that the Z3 model’s overall interaction energy was altered by 27% relative to the 2DRP template structure.  This value represents the level of uncertainty in the model and these conclusions are not going to be as reliable as those found for the Z1 model/template combination. 

The Z2/Z4 proteins served two purposes in this study.  First of all, the sequences verified that modeling approaches are reliable based on the fact that both models were built independently, but still converged to energy levels that were very similar.  Secondly, the two proteins demonstrated a significant energy loss (~50%) relative to the template used.  This loss of energy, with all other factors being equal, showed that there is significant error associated with this model/template combination and a more suitable template structure is required.  The overall interaction energy of -428 kJ/mol, was the lowest observed during this study and brings into question the level of error in models with such low homology.  Our study demonstrates that these values can be used as a type of “internal standard” for modeling that allows the reliability of the model to be more accurately assessed.   

Finally, this study demonstrates that computational methods, combined with structural and functional analysis, can be used to determine the effect of mutations at the molecular level.  These computational tools, when applied to available biological databases, would allow the classification of all of mutations that occur in model organisms and well as humans.  Such information would be extremely valuable in the effective diagnosis and treatment of disease.

References >>