Expression of full-length p53 and its isoform Δp53 in breast carcinomas in relation to mutation status and clinical parameters

Background The tumor suppressor gene p53 (TP53) controls numerous signaling pathways and is frequently mutated in human cancers. Novel p53 isoforms suggest alternative splicing as a regulatory feature of p53 activity. Results In this study we have analyzed mRNA expression of both wild-type and mutated p53 and its respective Δp53 isoform in 88 tumor samples from breast cancer in relation to clinical parameters and molecular subgroups. Three-dimensional structure differences for the novel internally deleted p53 isoform Δp53 have been predicted. We confirmed the expression of Δp53 mRNA in tumors using quantitative real-time PCR technique. The mRNA expression levels of the two isoforms were strongly correlated in both wild-type and p53-mutated tumors, with the level of the Δp53 isoform being approximately 1/3 of that of the full-length p53 mRNA. Patients expressing mutated full-length p53 and non-mutated (wild-type) Δp53, "mutational hybrids", showed a slightly higher frequency of patients with distant metastasis at time of diagnosis compared to other patients with p53 mutations, but otherwise did not differ significantly in any other clinical parameter. Interestingly, the p53 wild-type tumors showed a wide range of mRNA expression of both p53 isoforms. Tumors with mRNA expression levels in the upper or lower quartile were significantly associated with grade and molecular subtypes. In tumors with missense or in frame mutations the mRNA expression levels of both isoforms were significantly elevated, and in tumors with nonsense, frame shift or splice mutations the mRNA levels were significantly reduced compared to those expressing wild-type p53. Conclusion Expression of p53 is accompanied by the functionally different isoform Δp53 at the mRNA level in cell lines and human breast tumors. Investigations of "mutational hybrid" patients highlighted that wild-type Δp53 does not compensates for mutated p53, but rather may be associated with a worse prognosis. In tumors, both isoforms show strong correlations in different mutation-dependent mRNA expression patterns.


Background
The tumor suppressor and transcription factor p53 (TP53) is a key regulator of cell integrity with impact on cell cycling, growth, DNA repair, cell cycle arrest, or apoptosis (see reviews [1][2][3][4]). Correct p53 signaling is essential for preventing tumor growth (see reviews [5][6][7]). The structure of the TP53 protein has been studied extensively and different conserved domains have been identified [8,9]: the transcription activation domain, the sequence-specific DNA-binding domain with a subdomain interacting with the 53bp2 SH3 domain, a non-structured spacer region containing a bipartite nuclear localization signal, a tetramerization domain with a nuclear export signal subdomain, and a C-terminal domain modulating DNAbinding [10][11][12]. The central core domain of p53 is built of highly conserved anti-parallel beta-sheet scaffolds assembling two alpha-helical loops interacting with the grooves in the DNA [13]. The functional unit of p53 is a tetramer, where the C-terminal ends of two carboxyl-terminal peptides form a dimer, and two dimers assemble to tetramers [14,15].
Several p53 isoforms have been described, but for most of them knowledge has been restricted due to unclear function, their expression only at certain conditions or at very low levels, or their detection in other organisms than humans (see reviews [16,17]). Initially, human p53 was shown to have only one promoter and two alternative splice forms, p53i9 [18] and Δ40p53 [19][20][21]. Commonly p53 alternative splice forms diverge from full-length p53 by altering the N-terminal [19,20,22] or the or the C-terminal domains [18,23], but preserve the central domain.
Recently, a new internal promoter together with four additional N-and C-terminal isoforms were found [22], and the first internal splice form Δp53 was discovered [24]. The novel alternative splice form Δp53 is unique due to its unusual splice sites and expression pattern. In addition, its activation profile differs from that of p53 [24]. The importance of regulatory features of p53 isoforms has likely been underestimated [16], in particular, whether mutations in the p53 gene in tumors have different effect on the various isoforms. The various functions associated with the novel p53 alternative splice forms have attracted attention and opened questions about possible other functions (see comments [17,25]), since differential expression of p53 isoforms represents an interesting option for promoter selectivity, tissue-specific activation, and selective activation of downstream targeting genes.
The p53 gene has the highest mutation frequency in human tumors [26,27], with large varieties in the positions of the alterations and in the mutation spectra due to environmental, geographical, racial and other factors [28][29][30][31]. Mutations in the p53 gene are found in 20-30% of breast carcinomas (see reviews [3,28]), most of them being missense point mutations mainly located in or close to the conserved DNA-binding region [32]. The p53 mutation status has been shown to be an independent prognostic marker for poor outcome in breast cancer [33,34]. All mutations are "loss-of-function" regarding the tumor suppressor functions of wild-type p53, but some reports also describe that at least some mutations exert a novel "gain of function" (see review [35]).
In this paper we have studied mRNA expression of fulllength p53 and its Δp53 isoform in both p53 wild-type and mutant tumors from 88 breast cancer patients. We used quantitative real-time PCR (qRT-PCR) and related the mRNA expression levels to clinical and biological data. We wanted to explore whether patients with mutations affecting both full-length and Δp53 differ with respect to clinical and molecular markers from patients with mutations affecting only the full-length and not the Δp53 isoform.

Bioinformatic analyses of exon transition, structural domain organization, and prediction of three-dimensional protein folding for the alternative splice form Δp53
Δp53 is a novel alternative splice form that differs from the full-length p53 form by lacking parts of exon 7, complete elimination of exon 8 and partial removal of exon 9 [24]. The uncommon splice mechanism involves two 7 base pair long cassettes with an identical sequence in exon 7 and exon 9, of which one is retained in the isoform (Figure 1A; for sequence details see Additional file 1). We estimated the splice site plausibility by analyzing the cassette sequence for exonic splicing enhancer (ESE) motifs using the ESEfinder [36]. Exonic enhancers are potential binding sites for splicing factors of the highly conserved serine/ arginine-rich (SR) protein family. The cassette motif gives a high score (3.5) for the splice factor SF2/ASF protein and has some similarity with known signal sequences for alternative splicing (see review [37]). According to Swiss-Prot/ TrEMBL structural protein domain classification [12] fulllength p53 consists of a transcription activation domain (aa 1-44), a DNA-binding domain (aa 102-292), an unstructured spacer containing a bipartite nuclear localization signal (aa 305-321; a bipartite nuclear localization signal domain is defined as two adjacent basic amino acids with a spacer region of any 10 residue and at least three basic residues (Arg or Lys) in the five positions following the spacer region [38]), a tetramerization domain (aa 325-356), and a C-terminal regulatory domain (aa 368-387) ( Figure 1B). In Δp53, this domain organization is modified by the removal of 66 amino acid (residues 257 to 322), which mainly disturbs the DNA-binding domain, and eliminates the spacer with the bipartite nuclear localization signal. The DNA-binding domain (aa 102-292) is truncated, but the 53bp2 SH3 domain remains intact, while the spacer with the bipartite nuclear localization signal domain (aa 305-321) is entirely removed.
To elucidate the differences in three-dimensional structure between p53 and Δp53 the complete sequences of p53 and Δp53 were submitted to predictive CPH-modelling using the web-based service of the Technical University of Denmark [39]. The core domains of the proteins were modelled based on available structures in the database: p53 was predicted from aa 94 to 297 with an identity of 100% (557.0 bits score) and Δp53 from aa 94 to 274 with 93.4% identity by a 451.5 bits score. (The removal of residues 257 to 322 by the Δp53 specific splice process changes the position numbers of the Δp53 predicted protein relative to the full-length protein: residue 274 in Δp53 corresponds to 340 in the full-length p53 protein).
The isoforms clearly vary in both their secondary and their calculated three-dimensional structure, even though the prediction identity for Δp53 was limited. The threedimensional structure prediction, illustrated by using the RasMol program (Version 2.7.2.1.1.), reveals a modified structure for Δp53 in comparison to p53, resulting in a position shift of an alpha-helical structure and thus condensing the structure ( Figure 2). In Δp53 a major alphahelical structure at the C-terminal end is deleted, altering its further orientation. The structural data are supported by functional studies in cell lines, where the isoforms have different transactivation activities for the p21, mdm2, 14-3-3-sigma, bax and PIG3 promoters [24].

p53 and Δp53 mutational status in relation to clinical, pathological and biological factors
We analyzed the existence and expression of the Δp53 isoform in breast tumors from patients with advanced disease at the mRNA level and asked whether mutations present in both isoforms had different effects on clinical and molecular parameters compared to mutations found only in the full-length form. These patients have previously been analyzed by whole genome expression microarrays and grouped according to their expression profile ( [40][41][42][43] for cohorts A, B and D, and unpublished results for cohort C). Gel electrophoresis followed by qRT-PCR confirmed that Δp53 together with full-length p53 was present in all tumor samples. Patients with tumors harbouring mutations residing inside the spliced out region of the Δp53 isoform represented the rare mutational genotype of mutated full-length p53 and wild-type Δp53 and were termed "mutational hybrids" (Wild type Δp53/ Mutant p53 = WtM). These patients where grouped (11 patients) and compared to patients with mutations before and after the spliced region affecting both isoforms (27 patients), and to patients without mutations (50 patients). Of the 27 tumors with mutations affecting both isoforms 14 were missense mutations, two were in frame mutations, one a nonsense mutation, four were splice mutations, and six had frame shift mutations. (For a full  description of the various mutations see Additional file  2). Based on the mutation type the p53/Δp53 double mutant group (MM) was further subdivided into the MI group with missense and in frame mutations, and the MII group with nonsense, frame shift and splice mutations (Table 1). Kruskal Wallis rank tests were performed for differences in clinical and molecular parameters in three (WtWt-WtM-MM) or four classes (WtWt-WtM-MIMI-MIIMII) and the Mann-Whitney test was used to test for independent association between subgroups. A slightly higher frequency of patients with distant metastasis at time of diagnosis was observed in the WtM group compared to the MIMI group (p < 0.07). No other significant differences were observed for the "mutational hybrid" group weighted against the other groups. We analyzed whether the "mutational hybrid" genotype had an effect on patient survival time in a subset of patients from two prospective studies [44,45] of comparable treatment and uniformity. In the Kaplan-Meier plot ( Figure 3) survival data for a total of 50 patients without distant metastases at time of diagnosis are shown. The survival rates in patients with the "mutational hybrid" genotype (Wt Δp53 -M p53) was similar to the survival rate in patients with mutations in both p53 and Δp53 (M Δp53 -M p53), and the survival rates in these two groups were significantly different compared to patients with wild-type Δp53 and p53 (Wt Δp53 -Wt p53) (p < 0.05).

qRT-PCR analysis of p53 and Δp53 mRNA expression levels using Universal Human Reference cell lines, mutated, and non-mutated human breast tumors
Using qRT-PCR, the expression level of both full-length p53 and Δp53 mRNA were determined. Both isoforms were present in the Universal Human Reference of 10 human cell lines mixture and in the human breast tumor samples. The p53 and Δp53 mRNA levels were determined by standard curve measurements in a 1.5 orders of linear dynamic dilution range performed on the same plate. Both splice forms showed similar slopes and accordingly have equivalent target efficiencies ( Figure 4). Under the presupposition of equivalent efficiencies the Comparative Ct (Cycle threshold) ΔΔCt method can be selected to compare normalized expression levels of different samples relative to a calibrator sample.
Using the comparative Ct method we compared the linearized (2 -ΔΔCt ) expression levels of the standard curves of the two alternative splice forms relative to each other, whereas for the various tumor samples the more robust standard curve method was applied. An investigation of different housekeeping genes revealed two independent genes, PMM1 and RPL32 [46], suitable for mRNA expression level determination in human breast tumors (for Schematic representation of the full-length p53 and the alternative splice form Δp53

A.
Predicted structural organization of the p53 and Δp53 core domains illustrated by three-dimensional models Figure 2 Predicted structural organization of the p53 and Δp53 core domains illustrated by three-dimensional models.
Illustrated are CPH predicted models [39] of the p53 and Δp53 isoforms using the RasMol program. The p53 core domain of 204 aa (total length of wild-type p53 is 393 aa) is predicted from aa 94 to 297 with a prediction identity of 100% (557.0 bits score) and the Δp53 core domain is predicted from aa 94 to 274 (total length of Δp53 is 327 aa) with a 93.4% identity and 451.5 bits score (protein position 274 in the Δp53 accordingly corresponds to protein position 340 in the full-length p53). Models and prediction structures for p53 (A) and Δp53 (B) are shown, and the variations are colored by secondary structures as follows: alpha helices in magenta, beta sheets in yellow, turns in pale blue, and all other residues are colored white. Differences between the isoform predictions are indicated with arrows and the N-terminal starting and C-terminal end points are marked in the figure. Due to the alternative splicing a major alpha-helical structure is missing in Δp53 (in A red arrow) and the tertiary protein structure of Δp53 is slightly more compact, as can bee seen from the moved alpha helix, indicated by green arrows. Below: uploaded protein core for the three-dimensional structure prediction query and received structural template.    Comparison of the mRNA levels of the two isoforms p53 and Δp53 in 88 tumors samples revealed a high correlation in both mutated and non-mutated tumors with a correlation coefficient of r 2 = 0.86 for wild-type and r 2 = 0.85 for mutant tumors, respectively ( Figure 5A and 5B). The expression levels of Δp53 mRNA in the two tumor samples with in frame (FU27), or with a frame shift (FU08) mutation located in the splice cassette were very low (Figure 6), further confirming the existence of Δp53 in the other human tumor samples.
Interestingly, the expression level of both mutant and wild-type p53 varied by more than 3-fold for both full-length and Δp53, and the different mutation types showed a large variation in mRNA expression for both isoforms ( Figure 6). Tumors with wild-type p53 have an average mRNA expression level of 0.770 arbitrary units (a.u.), tumors with missense or in frame mutations (MI) showed elevated mRNA abundance of 1.310 a.u., while nonsense, frame shift or splice mutations (MII) had lower mRNA levels with an average of 0.496 a.u. (Figure 7 and Table 2). The wild-type p53 isoform mRNA level were significantly different (p < 0.00002) from the levels in both mutation groups and the same was the case for the mRNA level of the Δp53 isoforms (p < 0.004) ( Figure 7 and Table 2).

mRNA expression levels of mutated and non-mutated human p53 and Δp53 in human breast tumors in relation to clinical and biological parameters
The wild-type full-length p53 mRNA expression levels display a wide range. We explored whether mRNA expression levels were associated with particular clinical and/or molecular parameters. Therefore, we divided the wildtype p53 mRNA expression profiles into three classes of quartiles, merging the two midst quartiles to one group: (Q1) <25%, (Q2&Q3) 25-75% and (Q4) >75%. Kruskal Wallis tests were used to analyze for differences between the three groups and the Mann-Whitney test was used for tests for differences between any two groups (  tumors with MII mutations the ERBB2 subtype (46%) was most highly represented (Table 4).
Beside mutational status and despite relatively low numbers of tumor samples, the mRNA expression levels for both p53 and Δp53 mRNA mutant and wild-type demonstrated interesting results: The Luminal A subtype revealed scattered p53 mRNA expression levels in wild-type p53, but rather high mRNA expression in the mutated p53 tumors, while the Luminal B subtype had either a very low or very high mRNA expression rate in the wild-type p53, but normal expression distribution in mutant p53 tumors. The ERBB2 molecular breast cancer subtype indicates low p53 mRNA levels in mutant p53, but normal scattering in wild-type p53 tumor samples ( Figure 6).

Discussion
Recently, several new p53 isoforms have been detected, but their functional roles, particular in tumorigenesis, remain unclear and require further investigations (see perspective [17]). Δp53 is one of these novel isoforms, it arises by an uncommon alternative splice mechanism, exhibits a p53-independent transcriptional activity, and a gene activation pattern different from that of p53 in cell lines [24]. Recently, it was shown that cells from patients with acute myeloid leukemia induction of chemotherapy modulates the p53/Δp53 protein ratio pattern [47].
In this study we have investigated Δp53 in-silico and at the mRNA expression level in relation to wild-type and mutated full-length p53 in order to determine possible correlations to biological and clinical parameters in human breast tumors. Our bioinformatic analysis revealed a high score for exonic splicing enhancers for the cassette sequence motif. We performed three-dimensional predictions for the structure of the Δp53 isoform, and confirmed its mRNA expression in human tumors. Although the model prediction identity for Δp53 was lower than for p53, we were able to identify that a major alpha-helical structure is missing at the C-terminal end, changing the further orientation of the protein resulting in a more compacted structure. These structural changes may explain the inability of Δp53 to form hetero-tetramers with fulllength p53, and the different transcriptional activity of Δp53 independent from full-length p53 [24]. Activity differences have also been observed in other p53 isoforms: The Δ40p53 isoform is not activated in response to genotoxic stress [19], p53i9 is defective in transcriptional activity [18], and the p53AS isoform in mice displays different DNA-binding efficiencies [48].
The three-dimensional predictions and the functional analysis of Δp53 in cell lines encouraged us to investigate Δp53 function in relation to full-length p53 in human tumors. The special alternative splice process removes all mutations inside the spliced-out sequence and, as a consequence, the mutational statuses of the tumors are affected differentially for p53 than for Δp53. These "mutational hybrid" tumors have a mutated full-length p53 and a non-mutated Δp53. We investigated whether patients containing "mutational hybrid" tumors had biological and clinical parameters different from other types of mutations. No significant differences to specific changes in survival rate, in clinical or biological parameters were observed, although a slightly higher frequency of patients with distant metastasis at time of diagnosis was found. Thus, wild-type Δp53 does not seem to compensate for mutated p53, but possibly exerts adverse effects in tumors expressing mutant p53. One may speculate that a correct balance between full-length p53 and Δp53 is required to full-fill specific patterns in control of cell-cycle regulation p53 and Δp53 standard curve by qRT-PCR   and thus influences the overall survival in patients with a disturbed Δp53/p53 phenotype. Since this dataset is small, larger cohorts are necessary to confirm these findings.
Full-length p53 and Δp53 mRNA expression levels were measured by qRT-PCR. We could confirm that Δp53 mRNA is expressed in a mixture of different human cancer cell lines and in tumors, with Δp53 being expressed at a lower level than full-length p53. The mRNA levels of the two isoforms p53 and Δp53 are highly correlated. A 25-30% reduction in expression levels has also been reported for the mouse ASp53 isoform [49]. It may be a general feature that N-or C-terminal end truncated isoforms rather modulate p53 functions than abrogate it completely [16]. The existence of Δp53 mRNA is further confirmed by the observation that an in frame mutation located in the splice cassette led to high full-length p53 but reduced Δp53 mRNA expression, distinguishable from all others in frame or missense mutations. These and other results from different isoforms [22] form a picture according to which full-length p53 is the most highly expressed form during genotoxic stress.
In a series of 88 advanced primary breast tumors we investigated whether certain clinical parameters are related with the expression patterns of p53 and Δp53. It has previously been shown [22] that various p53 isoforms are expressed in human breast tumors, but correlation of expression levels to clinical data or mutational status was missing in that study. Steady-state amounts of mRNAs in genes related to breast cancer have very rarely been meas-Correlation between mRNA expression level of full-length p53 versus Δp53 in breast carcinomas with various p53 mutations  ured [50]. In the examined breast tumors we recognized for both p53 and Δp53 that the different mutation types show particular mRNA expression patterns. In comparison to wild-type mRNA expression, tumors with missense and in frame mutations had significantly increased amounts of mRNA, while in tumors with nonsense, frame shift and splice mutations mRNA levels were significantly reduced for both isoforms. Our study is thus one of the first confirming that mRNA expression of both the fulllength p53 and the Δp53 form are elevated in tumors with missense and in frame mutations. The high level of p53 protein seen in mutated tumors has previously been explained by accumulation of the protein due to lack of degradation of the mutated protein and not by overexpression at the mRNA level. After DNA damage p53 is activated and Mdm2-p53 interaction decreases [4,51]. In a situation with increased levels of mutated p53, the disturbed dynamics of this fine-balance may result in an unsatisfied request for functional p53 activity inside the cell.
The molecular breast cancer subtypes [41,42] differ significantly with respect to frequencies of p53 mutations. The majority of wild-type tumors is classified as Luminal A subtype, while the majority of tumors with missense and in frame mutations belong to the Basal subtype and the ERBB2 subtype has mainly nonsense, frame shift or splice mutations. The different subgroups have different survival with the poorest survival rates for the class with highest p53 mutation rate [33,43,52].
We observed that the expression of full-length wild-type p53 was widely scattered, despite the significant differences between mutation groups as described above. To explore this unexpected result we divided the tumors into four p53 wild-type expression groups by quartiles. The tumors in the lowest quartile group were significantly associated with estrogen-negative receptor staining, high grade and the Luminal B and ERBB2 breast cancer subtypes, while tumors in the two middle quartiles showed significant association with a high fraction to estrogen receptor positive tumors, low grade and tumors of the Luminal A subtype. Tumors in the highest quartile of p53 abundance were associated with negative estrogen receptor status, low grade and the Luminal A breast cancer subtype. These findings are interesting and warrant further investigations in order to elucidate the molecular basis for these expression "extremes" in the wild-type p53 gene. Explanations may comprise technical reasons, like difficulties in detection of all types of mutations by standard Histogram of p53 and Δp53 relative mRNA expression Figure 7 Histogram of p53 and Δp53 relative mRNA expression. Mean and SEM of the relative p53 and Δp53 mRNA expression in a.u. (arbitrary units) for the different mutation classes: missense, in frame, nonsense, frame shift, splice or splice cassette (splice cass). p53 and Δp53 mRNA expression levels are proportion adjusted according to the comparative Ct of 2.64. The number of cases (n) is given by 1 for p53 and by 2 for Δp53, respectively. Differences in case numbers for p53 and Δp53 are due to some mutations which are located inside the area, but removed by the alternative splicing process of Δp53. The expression level for both full-length p53 and Δp53 between the various mutational classes were highly significant ( Table 2).  techniques, or a biological basis due to mutations in genes other than p53 itself, disrupting the p53-signaling pathway. Future studies with an enlarged cohort size are required to determine whether the different distributions of low and high p53 mRNA level in the molecular subgroups are the cause or effect of the tumor.

Conclusion
The tumor suppressor and transcription factor p53 is accompanied by different alternative splice forms. In silico analysis indicate three-dimensional and functional differences of the Δp53 and full-length p53 isoforms. Quantitative real-time PCR confirmed Δp53 mRNA expression with strong correlation between the two isoforms, however, with 2.64 higher levels for the p53 fulllength form, both in mutated and non-mutated tumors. If at all, "mutational hybrid" patients had a slightly worse prognosis than patients with p53 mutations in both isoforms, indicating that wild-type Δp53 does not replace the p53 function lost by mutation, but rather might exert an adverse effect. The mRNA expression of p53 and Δp53 level showed a wide range in p53 wild-type tumors, with significant association to molecular breast cancer subtype distribution. In tumors, different mutation-dependent mRNA expression patterns were found with significant higher mRNA expression of both isoforms from missense or in frame p53 mutated genes compared to the wild-type p53 gene. A significant association was found for the distribution of breast cancer subtypes for wild-type and mutated Δp53 and the scattering of p53 mRNA expression levels revealed differences in wild-type p53 or mutated p53 tumors among the various subtypes.

Patients
A total of 88 breast tumors samples from patients with advanced disease were selected from 4 different cohorts (A, B, C, and D): Fifty-six of the patients were part of a prospective study at the Haukeland University Hospital Bergen (Norway) on locally advanced breast cancer (T3/T4  [54]. All breast carcinoma samples were frozen immediately after surgery and stored at -70°C to -80°C. Total RNA was isolated from snap frozen tumor tissue using TRIzol ® solution (Invitrogen™). The concentration of total RNA was determined using an HP 8453 spectrophotometer (Hewlett Packard) and the integrity of the RNA was assessed using a 2100 Bioanalyzer (Agilent) for series C and D.

Biological and clinical factors
The estrogen receptor (ER) and progesterone receptor (PR) status were analyzed using both immunohistochem-istry (IHC) and biochemical/ligand-binding assay (Abbott Diagnostics). Lymph node status, grade, and distant metastasis at time of diagnosis were available for patients in cohorts A, B and D, ERBB2/HER status were available for patients in cohorts A and D, and response to chemotherapy were available for patients in cohorts A and B [33,44,45]. For patients in cohort C, the lymph node status and histological grade were obtained from histological reports. Hormone receptors status and ERBB2/HER status were performed using IHC and scored using a semiquantitative evaluation. Age, menopausal status, lymph node status, tumor histology, or p53 LOH were available for patients in all cohorts ( [44,45,53,54] and unpublished). Breast cancer subtype classification was based on variation in gene expression derived from microarray experiments. Each sample was assigned to its subclass by the correlation to the centroid for the subclass [43]. Breast cancer subtype assignment for patients in the cohorts A, B, and D have previously been published [40][41][42][43] and are unpublished for patients in cohort C.

Analytical Method
Briefly, in real time PCR the exponential increase in fluorescence signal during cycling is measured to generate a quantitative relation to the target amount at reaction start. Two analytical methods are established for qRT-PCR measurement, the standard curve method and the comparative Ct method (ΔΔCt). The benefit of the standard curve method is its independence of variations in amplification efficiencies between different genes, different splice variants, or between the target gene and the endogenous controls. The comparative Ct method was used to compare the linearized (2 -ΔΔCt ) expression levels of the standard curves of the two alternative splice forms relative to each other, whereas for the various tumor samples the more robust standard curve method was applied.

Primer and probes
Primers and probes for the p53 and Δp53 mRNA sequence were designed with the assistance of the Primer Express ® software (Applied Biosystems). We tested several primer pairs, highest specificity for Δp53 was reached with a forward primer covering exon 7, the splice cassette and the exon 9 junction. For the full-length p53 the forward primer covers the exon 6/7 junction, respectively. The sequence of the various primers were: RPL32-PP 5-FAM-CGCCAGTTACGCTTAA-MGB-3 (modified after [46]).

Endogenous control
A general problem for gene expression measurements using qRT-PCR is to adjust for differences in loading, PCR inhibition or degradation. Photometric measurements are unreliable for sensible methods like qRT-PCR, as they do not take into account differences in the RT-reaction or changes in the mRNA/total RNA ratio. Many technical papers and reviews have addressed the problem of proper endogenous control genes for the normalization of qRT-PCR performance [46,[56][57][58][59][60][61]. However, these studies show often contradictory results and do not come up with a "golden rule". We have carefully investigated most of the commonly used "housekeeping" genes, like the human 18S ribosomal RNA (18S), β-actin, cyclophilin, glyceraldehydes-3-phosphatase dehydrogenase (GAPDH), β2microglobulin, β-glucronidase (GUS), hypoxanthine ribosyl transferase (HPRT), and the transcription factor IID TATA binding protein (TBP). We have rejected most of these commonly used endogenous controls based on the existence of retropseudogenes for some of these genes (βactin, GADH, HPRT [62][63][64]) or on their previous observed tumor or cancer type specific expression alteration patterns (β-actin, GAPDH, TBP, cyclophilin [58,60,[65][66][67]). Especially GAPDH correlates with clinical and molecular parameters in human breast cancer and should not be used as control RNA [68]. We followed the recommendation of [60] to use at least two endogenous controls. Based on the investigations of [46,59,61], and our own survey of these genes in microarray experiments, we choose the two endogenous controls LP32 and PMM1 for this study. The traditional loading control 18S rRNA was included as reference to previous studies and thus, low, medium and highly expression ranges were covered by our endogenous controls. However, 18S rRNA is overexpressed in mammary gland and colon cancer [66,69], affected by various biological factors and drugs [70,71], and shows an imbalance between mRNA and the rRNA content in mammalian mammary tumors [72]. Comparison of expression variation in our study showed that the highest divergence was found for 18S and consequently it was omitted. All relative mRNA quantity values of p53 and Δp53 were normalized to the average levels of the two independent endogenous control references PMM1 and RPL32.
Software applications and statistical analysis p53 protein domain classification and their locations along the protein were specified according to Swiss-Prot/ TrEMBL [12]. Exonic splicing enhancers (ESEs) binding sites for splicing factors of the serine/arginine-rich (SR) protein family were analyzed using the ESEfinder software [36]. For three-dimensional structural prediction the complete sequences of p53 and Δp53 were submitted to the CPHmodels 2.0 server [73], a web-based CPH-modeling service of the Technical University of Denmark [39]. The bits scores of the model prediction indicate the precision of a match to a HMM profile. Alignments scores are commonly reported as bits scores: The likelihood that the query sequence is a plausible homologue of the database sequence is compared to the likelihood that the sequence was instead generated by a "random" model. The log2 of this likelihood ratio gives the bits score. The three-dimensional structure prediction models were illustrated using the RasMol program (