Introduction

Single nucleotide polymorphisms (SNPs) are the simplest and most frequent type of DNA sequence variation among individuals. Human genome is composed of more than three billion base pairs and ten million SNPs (Cargill et al. 1999). SNPs are not only important as markers for constructing genetic maps but also have potential as direct functional polymorphic variants involved within common and genetically complex diseases as well as drug response because (1) SNPs within the coding regions (cSNPs) of functional genes introduce biological variations directly into the gene products through the creation of missense substitutions or premature termination codon, (2) SNPs present in noncoding regions have effects on gene expression by affecting regulatory elements and (3) some intronic SNPs activate cryptic splice sites, leading to alternative splicing (Goto et al. 2001). SNPs in coding and regulatory regions may be implicated in disease themselves. Nonsynonymous SNPs that lead to an amino acid change in the protein product are of major interest because amino acid substitutions currently account for approximately half of the known gene lesions responsible for human inherited disease (Krawczak et al. 2000).

Several different SNP assays have been applied for SNP prioritization experimentally, most involving target amplification following allelic discrimination and a detection step (Chen and Sullivan 2003; Kwock 2003). Though experimental-based approach will provide the strongest evidence for the functional role of a genetic variant, these studies are difficult for characterizing all human genetic variants and also their results might not always reflect in vivo genotype function in humans. On the other hand, computational algorithms are a high priority for characterizing variants because they have the ability to be employed on a scale that is consistent with the large number of variants being identified in systematic screening of representatives of human populations for variation. Computational analysis undertaken for an in silico investigation of nsSNPs in TP53 gene is scarce. In this work, we have analyzed the SNPs that can alter the expression and function of transcriptional factor TP53 as a pipeline and for providing a guide to experimental work. We have developed a computational strategy to identify potential and functionally significant SNPs (Fig. 1). We applied different computational algorithms namely Sorting Intolerant From Tolerant (SIFT), Polymorphism Phenotyping (PolyPhen), PupaSuite, UTRscan, and FASTSNP to identify the candidate SNPs that are likely to affect the protein and subsequently the cellular functions. We identified the possible mutations, proposed modeled structures for the mutant proteins, and compared them with the native protein. Our computational study demonstrates the presence of deleterious mutations in TP53 gene that affect the expression and function of proteins with possible roles in human cancers.

Fig. 1
figure 1

Flow chart for computational analysis of functional SNPs

Full size image

Suites of software for assessing SNPs

Functional analysis of coding nsSNPs by SIFT

SIFT is a sequence homology-based tool that sorts intolerant from tolerant amino acid substitutions and predicts whether an amino acid substitution in a protein will have a phenotypic effect. SIFT is based on the premise that protein evolution is correlated with protein function. We used the program SIFT (Pauline and Henikoff 2003) available at http://blocks.fhcrc.org/sift/SIFT.html to detect the deleterious coding nonsynonymous SNPs and submitted the query in the form of either SNPids or as protein sequences. The underlying principle of this program is that SIFT takes a query sequence and uses multiple alignment information to predict tolerated and deleterious substitutions for every position of the query sequence. SIFT is a multistep procedure in the sense that, given a protein sequence, (1) searches for similar sequences, (2) chooses closely related sequences that may share similar function, (3) obtains the multiple alignment of these chosen sequences, and (4) calculates normalized probabilities for all possible substitutions at each position from the alignment. Substitutions at each position with normalized probabilities less than a chosen cutoff are predicted to be deleterious and those greater than or equal to the cutoff are predicted to be tolerated (Pauline and Henikoff 2001). The cutoff value in SIFT program is tolerance index of ≥0.05. The higher a tolerance index, the less functional impact a particular amino acid substitution is likely to have.

Simulation for functional change in coding nsSNPs

PolyPhen is an automatic tool for prediction of possible impact of an amino acid substitution on the structure and function of a human protein available at http://coot.embl.de/PolyPhen/. This prediction is based on straightforward empirical rules which are applied to the sequence, phylogenetic, and structural information characterizing the substitution. Input options for PolyPhen server (Ramensky et al. 2002) is protein sequence or SWALL database ID or accession number together with sequence position with two amino acid variants. We submitted the query in the form of protein sequence with mutational position and two amino acid variants. Basically, PolyPhen searches for 3D protein structures, multiple alignments of homologous sequences, and amino acid contact information in several protein structure databases, calculates position-specific independent counts (PSIC) scores for each of two variants, and then computes the PSIC score difference of two variants. The higher a PSIC score difference, the higher functional impact a particular amino acid substitution is likely to have. A PSIC score difference of 1.5 and above is considered to be damaging.

Analyzing the molecular phenotypic effects of human nonsynonymous SNPs

The SNPeffect (Reumers et al. 2006) and PupaSuite (Conde et al. 2006) are now synchronized to deliver annotations for both noncoding and coding SNP, as well as annotations for the SwissProt set of human disease mutations. In this approach, the input consists of a list of genes (genes belonging to a given pathway, involved in a particular biological function, etc). The user must specify the type of gene identifiers by selecting either Ensembl or an external database (which include GenBank, Swissprot/TrEMBL, and other gene ids supported by Ensembl). PupaSuite is a unique and more integrated interface of PupaSNP (Conde et al. 2004) and PupasView (Conde et al. 2005) accessible at http://pupasuite.bioinfo.cipf.es and through http://www.pupasnp.org. PupasView retrieves SNPs that could affect conserved regions that the cellular machinery uses for the correct processing of genes (intron–exon boundaries or exonic splicing enhancers), predicted transcription factor binding sites, and changes in amino acids in the proteins for which a putative pathological effect. PupaSuite finds all the SNPs mapping in locations that might cause a loss of functionality in the genes.

Scanning of noncoding SNPs

Functional significance of each SNP in untranslated region (UTR) was determined by UTRscan (Pesole and Liuni 1999) available at (http://www.ba.itb.cnr.it/BIG/UTRScan). UTResource, which is an internet resource of sequence analysis of 5’ and 3’ UTR of eukaryotic mRNAs which are involved in many posttranscriptional regulatory pathways that control mRNA localization, stability, and translation efficiency (Sonenberg 1994; Nowak 1994). Briefly, two or three sequences of each UTR SNP that have a different nucleotide at an SNP position were analyzed by UTRscan, which looks for UTR functional elements by searching through user-submitted sequence data for the patterns defined in the UTR site and UTR databases. If different sequences for each UTR SNP are found to have different functional patterns, this UTR SNP is predicted to have functional significance. The internet resources for UTR analysis were UTRdb and UTRsite. UTRdb contains experimentally proven biological activity of functional patterns of UTR sequence from eukaryotic mRNAs (Pesole et al. 2002). The UTRsite has the data collected from UTRdb and also is continuously enriched with new functional patterns.

We used the FASTSNP (Yuan et al. 2006) to identify the polymorphism involving the intron which may lead to defects in RNA and mRNA processing. The FASTSNP server (http://fastsnp.ibms.sinica.edu.tw) follows the decision tree principle with external Web service access to TFSearch, which predicts whether a noncoding SNP alters the transcription factor binding site of a gene. The score will be given on the basis of levels of risk with a ranking of 0, 1, 2, 3, 4, or 5. This signifies the levels of no, very low, low, medium, high, and very high effect, respectively.

Modeling nsSNP locations on protein structure and their RMSD difference

Structure analysis was performed for evaluating the structural stability of native and mutant protein. We used the web resource SAAPdb (Cavallo et al. 2005) and dbSNP to identify the protein coded by TP53 gene (PDB id 1TSR). We also confirmed the mutation positions and the mutation residues from this server. These mutation positions and residues were in complete agreement with the results obtained with SIFT and PolyPhen programs. The mutation was performed by using SWISSPDB viewer and energy minimization for 3D structures was performed by NOMAD-Ref server (Lindahl et al. 2006). This server use Gromacs as default force field for energy minimization based on the methods of steepest descent, conjugate gradient, and L-BFGS methods (Delarue and Dumas 2004). We used conjugate gradient method for optimizing the 3D structures. Deviation between the two structures was evaluated by their RMSD values.

Using the human transcription factor TP53 as a test case

The human gene TP53 (tumor suppressor gene) is mutated in more than 50% of human cancers and p53 dysfunction is caused through a direct mutation within the DNA binding domain of the gene (Vogelstein et al. 2000). The p53 tumor protein is essential for regulating cell division; it has been nicknamed the “guardian of the genome.” The gene is located on the short (p) arm human chromosome 17 (17p13.1; Strachan et al. 1999). The human p53 protein consists of 393 amino acids with five evolutionarily conserved domains (I to V). Domains II to V correspond to the DNA binding domain which is the target for p53 mutations. Ninety percent of p53 mutations occur in the central region (101–306) which harbors four of the five highly conserved evolutionary domains and this region is essential for the p53–DNA interaction. The p53 tumor suppressor is a transcription factor that coordinates cellular responses to DNA damage and stress, initiating cell cycle arrest or triggering apoptosis. There are a variety of stresses that have been shown to activate p53 including DNA damage, cell cycle aberrations, hypoxia, and aberrant growth signals resulting from expression of oncogenes (Monica et al. 2007). The spectrum of TP53 mutations in various types of cancers is mainly due to GC→AT transition at the CpG nucleotide [colon, ovary brain, or leukemia] and GC→TA transversion [lung, head and neck, or HCC] (Hollstein et al. 1991). Incidence of p53 mutations is highest in ovarian cancer (48.3%), followed by colorectal cancer (43.6%), esophageal cancer (42.6%), head-and-neck cancer (41.5%), and lung cancer (38.4%; Petitjean et al. 2007). Our survey shows that there is a wide choice of literature in mutations of the TP53 gene associated with various forms of human cancers.

SNP dataset from dbSNP

SNP dataset for TP53 gene investigated in this work was retrieved from dbSNP (Sherry et al. 2001) http://www.ncbi.nlm.nih.gov/SNP/ for our computational analysis. We selected (1) nonsynonymous coding SNPs, (2) 5’ and 3’ UTR SNPs, and (3) introns for our investigation and distribution of SNPs are shown in (Fig. 2). Out of 209 SNPs, 20 were nonsynonymous SNPs (nsSNPs) and five SNPs in coding synonymous. Noncoding region comprises of one SNP in 5’ UTR, 18 SNPs in 3’ UTR, and 165 SNPs were in the intron region. Since the numbers of SNPs in 5’ UTR are much less as compared to the SNPs in the 3’ UTR and in the coding regions, it may be presumed that TP53 gene shows no functional significance in the 5’ UTR.

Fig. 2
figure 2

Distribution of SNPs in nsSNPs, 3’ UTR, 5’ UTR and intron regions

Full size image

Deleterious nsSNP by SIFT program

SIFT, a sequence homology-based tool, was used to identify the conservation level of a particular position in a protein. Protein sequences of 20 nsSNPs were submitted independently to SIFT program and 14 nsSNPs (70%) were identified to be deleterious having the tolerance index score of ≤0.05. Thirteen nsSNPs showed a highly deleterious tolerance index score of 0.00 and one nsSNP had a tolerance index score of 0.03, respectively.

Damaged nsSNP by PolyPhen server

The structural levels of alteration were determined by applying PolyPhen program. All the 20 protein sequences of nsSNPs submitted to SIFT were also submitted as input to the PolyPhen server. Fourteen nsSNPs (70%) were considered to be damaging (Table 1) and exhibited a range of PSIC score difference between 1.74 and 3.14. Interestingly, we found significant correlation between SIFT and PolyPhen tools. So, we could infer from the results obtained using the SIFT and PolyPhen tool that these nsSNPs may disrupt both the protein function and structure and mutations occurring in these nsSNPs could be of significant importance in causing of various forms of cancer.

Table 1 List of SNPs that were predicted to be of functional significance by SIFT and PolyPhen
Full size table

Predicting the effect of coding nonsynonymous SNPs

The PupaSuite server aims to provide a platform for predicting the effect of coding nonsynonymous SNPs on the structure and function of the affected protein. We submitted TP53 gene in Pupa Suite and specified the type of gene identifiers by selecting either Ensembl or an external database. Out of 20 nsSNPs in TP53 gene, 11 nsSNPs disrupted the exonic splicing enhancers, three nsSNPs disrupted the exonic splicing silencers, and four nsSNPs (pathological SNPs) were involved in cellular processing and two nsSNPs in functional sites and two nsSNPs showed selective constraints prediction. In noncoding region of TP53 gene, out of eight SNPs in mRNA regions, four SNPs disrupted the exon splicing enhancer, four SNPs disrupted the exon splicing silencer. The results of nsSNPs obtained by PupaSuite (Table 2) correlates with the nsSNPs affected by SIFT and PolyPhen tools.

Table 2 List of SNPs that were predicted to be of functional significance by PupaSuite
Full size table

Functional SNPs in noncoding SNPs

Polymorphism in the 3’ UTR affects the gene expression by affecting the ribosomal translation of mRNA or by influencing the RNA half-life (Deventer 2000). Among 18 SNPs in 3’ UTR, four SNPs were related to the functional pattern change of IRES by UTRscan (Table 3). Internal ribosome entry site (IRES) is bound by internal mRNA ribosome. It is an alternative mechanism of translation initiation compared to the conventional 50-cap dependent ribosome scanning mechanism (Becky and Anne 2005). Out of 165 SNPs in intron region of TP53 gene (Table 4), 73 SNPs were predicted to be functionally significance, with a risk ranking of 1–2 (71 SNPs) and 3–4 (2 SNPs), respectively, by FASTSNP. SNP with an id (rs11575997) present in intron region showed functionally significance in both FASTSNP and PupaSuite. Similarly, SNPs, namely with id rs17881366 and rsl6956880, in mRNA region showed functional significance in both UTRscan and PupaSuite. SNPs with id rs1625895 and 12951053 present in intronic region showed risk of 1–2 by FASTSNP, which is in good correlation with experimental studies (Brian et al. 2007 ).

Table 3 List of SNPs (mRNA) that were predicted to be of functional significance by UTRscan
Full size table
Table 4 List of SNPs (intron) that were predicted to be of functional significance by FASTSNP
Full size table

Modeling of mutant structure

Mapping the deleterious nsSNPs into protein structure information was obtained from dbSNP and SAAPdb. The available structure for TP53 gene is reported to be having a PDB id (1TSR).

According to this resource, the mutations mainly occurred for ITSR with 14 SNP rs ids, namely, rs11540654, rs28934873, rs28934875, rs28934874, rs28934578, rs28934573, rs28934572, rs28934575, rs11540652, rs28934571, rs28934577, rs28934576, rs17849781, and rs28934574. The mutation for 1TSR at the corresponding positions were performed by SWISSPDB viewer independently to get modeled structures. Then, energy minimizations were performed by NOMAD-Ref server for the native-type protein (1TSR) and the mutant-type structures.

From the modeled structures, RMSD values between the native and mutant amino acid residues were calculated (Table 5) and exhibited a range from 0.14 to 2.14 Å, respectively. It can be seen from the Table 5 that total energy for the native-type structure (1TSR) after energy minimization is found to be 13,161.882 kcal/mol and the mutant-type structures (1TSR) exhibited a minimization energy range of −12,792.573 to −13,157.354 kcal/mol, respectively. The total energy after energy minimization is almost similar or high for the mutant-type structures as compared to the native-type structure (1TSR). The highest and least RMSD values of superimposed structures of the native protein (1TSR) with mutant-type proteins 245(G→S) and 241(S→F) are 2.14 and 0.14 Å, respectively, and shown in (Figs. 3 and 4). All the deleterious nsSNPs which were mapped by SAAPdb and dbSNP in the native protein (1TSR) of TP53 gene should be considered functionally significant based on both sequence homology (SIFT), structural homology (PolyPhen), and PupaSuite for causing various forms of cancer. Our results obtained through computational algorithms are in good correlation with the experimental results (Soussi et al. 2006).

Fig. 3
figure 3

Superimposed structure of native protein 1TSR (orange) with mutant protein (cyan) 245(G→S)

Full size image
Fig. 4
figure 4

Superimposed structure of native protein 1TSR (orange) with mutant protein (cyan) 241(S→F)

Full size image
Table 5 RMSD of native-type protein (1TSR) and mutant modeled proteins
Full size table

Conclusions

Given that hundreds of thousands of SNPs are estimated to exist in the human population. However, only a small subset of variants that affect the phenotype will confer to disease risk. The effect of many nsSNPs will probably be neutral as natural selection will have removed mutations on essential positions. Assessment of nonneutral SNPs is mainly based on phylogenetic information (i.e., correlation with residue conservation) extended to a certain degree with structural approaches (PolyPhen). Much attention has been focused on modeling by different methods the possible phenotypic effect of SNPs that cause amino acid changes and only recently has interest focused on functional SNPs affecting regulatory regions or the splicing process. However, there is increasing evidence that many human disease genes are the result of exonic or noncoding mutations affecting regulatory regions. Study of the molecular basis of diseases by experimental methods is laborious and time-consuming, and at the structural level often nearly impossible, especially in cases where there are several missense mutations causing the disease. By contrast, precise and useful information about the effects of mutations on protein structure and function can be readily obtained by theoretical methods.

Out of 20 nsSNPs in TP53 gene, 14 of them were found to be deleterious (SIFT) and damaging (PolyPhen); 16 nsSNPs and eight SNPs in mRNA showed molecular phenotypic variations by PupaSuite. Four SNPs in the 3’ UTR were found to be functionally significant by UTRScan and 73 SNPs in the intronic region were found to be functionally significant by FASTSNP. Our results from this study suggests that the application of computational algorithms namely SIFT, PolyPhen, PupaSuite, UTResource, and FASTSNP might provide an alternative approach to select target SNPs by understanding the effect of SNPs on the functional attributes or “molecular phenotype” of a protein. Some detailed laboratory work has been published already, but our analysis reveals that further structural and functional information can be derived using computer-assisted methods. The models built in this work would be applicable for predicting the deleterious SNPs and their functions in gene regulation which would be helpful for further genotype–phenotype researches as well as the pharmacogenetics study.