Abnormal TP53 Predicts Risk of Progression in Patients With Barrett’s Esophagus Regardless of a Diagnosis of Dysplasia

BACKGROUND AND AIMS: Barrett’s esophagus (BE) is the precursor to esophageal adenocarcinoma. A major challenge is identifying the small group with BE who will progress to advanced disease from the many who will not. Assessment of p53 status has promise as a predictive biomarker, but analytic limitations and lack of validation have precluded its use. The aim of this study was to develop a robust criteria for grading abnormal immunohistochemical (IHC) expression of p53 and to test its utility as a biomarker for progression in BE. METHODS: Criteria for abnormal IHC of p53 were developed in BE biopsies and validated with sequencing to assess TP53 mutations. The utility of p53 IHC as a biomarker for progression of BE was tested retrospectively in 561 patients with BE with or without known progression. The findings were prospectively validated in a clinical practice setting in 1487 patients with BE. RESULTS: Abnormal p53 IHC highly correlated with TP53 mutation status (90.6% agreement) and was strongly associated with neoplastic progression in the retrospective cohorts, regardless of histologic diagnosis (P < .001). In the retrospective cohort, abnormal p53 was associated with a hazard ratio of 5.03 (95% confidence interval, 3.88–6.5) and a hazard ratio of 5.27 (95% confidence interval, 3.93–7.07) for patients with exclusively nondysplastic disease before progression. In the prospective validation cohort, p53 IHC predicted progression among nondysplastic BE, indefinite for dysplasia, and low-grade dysplasia (P < .001). CONCLUSIONS: p53 IHC identifies patients with BE at higher risk of progression, including in patients without evidence of dysplasia. p53 IHC is inexpensive, easily integrated into routine practice, and should be considered in biopsies from all BE patients without high-grade dysplasia or cancer.


Development of p53 IHC scoring criteria
While much has been published regarding abnormal p53 expression in cancers and dysplasia, much less is known about expression patterns in NDBE. To further delineate the spectrum of p53 IHC staining in non-dysplastic biopsies, and delineate criteria for abnormal staining, we performed p53 IHC in 18 NDBE biopsies from patients with no known dysplasia, and in 115 NDBE biopsies that were concurrent with high grade dysplasia. The percentage of nuclei with positive staining was scored on an intensity scale of 0-3, with 0+ representing no staining and 3+ representing very strong staining. A consistent pattern of crypt staining was seen in all crypts in NDBE biopsies from patients with no known dysplasia, and in most crypts in NDBE biopsies from patients with concurrent high-grade dysplasia: crypt base positivity was always present, with most cells having 0-1+ nuclear positivity, scattered cells having 2+ nuclear positivity, and rare cells having 3+ nuclear positivity (Supplementary Figure S6). This crypt base positivity diminished towards the surface, which typically had 0 to scattered 1+ IHC positivity. This staining pattern is consistent with the physiologic normal expression of p53 in the proliferative zone of the crypt base.
In addition to this normal p53 crypt expression pattern, NDBE biopsies from patients with concurrent high-grade dysplasia harbored foci of distinctly abnormal staining comprised of two common patterns: markedly increased staining, and complete absence of staining (Supplementary Figure S6). Increased p53 expression is known to be associated with TP53 point mutations; these result in protein stabilization and nuclear accumulation. Absent p53 expression is known to be associated with TP53 truncation mutations, and other genomic aberrations that result in loss of p53 expression. To develop criteria for abnormally increased staining, we identified the foci that had the strongest staining, and scored the percentage of cells that had either 2+ and 3+ nuclear positivity. Increased staining extended onto the luminal surface in some cases, but was confined to the crypt base in others, in which case only the crypt base was evaluated. Foci as small as a single individual crypt base were considered adequate for scoring increased staining. Scoring using either 3+ positivity alone, or combined 2-3+ positivity, revealed that approximately 35% of NDBE biopsies from patients with concurrent high-grade dysplasia had foci of increased p53 expression that was greater than anything present in the NDBE biopsies without dysplasia (Supplementary Figure S7). We chose >50% 2-3+ nuclear positivity as a cut-off to define abnormal p53 expression because it had 100% specificity for an association with concurrent high-grade dysplasia, and could be rapidly assessed by a pathologist, potentially making it a practical biomarker for routine clinical use. Finally, occasional NDBE biopsies from patients with concurrent high-grade dysplasia also had focal marked p53 positivity present only within surface epithelium, but not adjacent crypts.
This finding was consistent with a mutant p53 clone extending onto the surface from a nearby crypt not present in the section. Therefore, the presence of 2-3+ p53 positivity in >50% of a contiguous focus of 20 surface cells was also considered abnormal.
While the vast majority of crypts from NDBE biopsies with no known dysplasia had <20% 2-3+ nuclear positivity in crypt bases, there were some crypts with 31-40% 2-3+ nuclear positivity. Although this degree of positivity was present at higher frequency in NDBE biopsies from patients with concurrent high-grade dysplasia, there was appreciable overlap. Therefore, 2-3+ nuclear positivity in 21-50% of crypt epithelial nuclei was considered equivocal, and while underlying TP53 mutations may be present in a subset of these cases, was classified as p53-NL in all analyses in this study.
The second common abnormal p53 expression pattern, complete absence of p53 expression, was identified in 20/115 (17%) NDBE biopsies from patients with concurrent high-grade dysplasia compared to none of the NDBE biopsies from patients with no known dysplasia. Because of the consistent normal p53 expression pattern present in crypt bases of NDBE biopsies from patients with no known dysplasia, this abnormal loss of staining could be reliably scored when identified in foci as small as a single crypt base.
In contrast to abnormally increased p53 expression, the very low (normal) p53 expression in Barrett's epithelium at the luminal surface makes it impossible to accurately diagnose absent p53 expression (mutation) when the involvement is limited to the surface epithelium of a NDBE biopsy.
Finally, during the blinded slide review of the retrospective cohort, we identified one additional distinctive rare abnormal p53 expression pattern: complete absence of nuclear p53 positivity in the presence of strong cytoplasmic positivity. This abnormal pattern has been seen for some other nuclear proteins as well as in p53 and has been attributed to mutations that result in loss of the nuclear localization domain 39 . This abnormal pattern was considered to be a variant of loss of expression and was combined with absent staining pattern in subsequent analyses. Supplementary Table S1 summarizes the p53 expression criteria used in this study.
The criteria defined above were further validated in an additional sample set of 50 NDBE biopsies from patients with no known dysplasia, and 50 BE-HGD biopsies. Abnormal p53 expression was found in 2/50 (4%) of the NDBE biopsies and 48/50 (96%) BE-HGD biopsies, providing additional validation that these scoring criteria are very sensitive for identification of abnormal p53 in advanced Barrett's neoplasms, while only being positive in a very small subset of unselected NDBE biopsies (which is to be expected for a possible biomarker).

Central Pathology Review
When there were discrepancies between the original diagnosis and reviewed diagnosis, a third independent reviewer assessed the slides. Any remaining discrepancies were assessed by consensus at a multi-headed microscope. Primary analysis was performed using the central review pathologic diagnosis. A secondary analysis using the original pathologic diagnosis was also performed to validate the utility of p53 IHC in the context of the original routine clinical diagnosis. For the prospective cohort, to determine how the p53 IHC protocol would work in routine practice, we utilized the original clinical diagnosis except all diagnoses of dysplasia or cancer were confirmed by at least one additional pathologist.
For the retrospective testing cohorts, central pathology review confirmed the presence of Barrett's mucosa, columnar dysplasia or adenocarcinoma in 3243 biopsies. IHC for p53 was not attempted in 353 of these cases because there was no tissue available, or the histologic focus of interest was not present in remaining tissue. Of the remaining 2890 biopsies, a p53 IHC result was obtained in 2853 (98.7%). Figure 1b shows examples of histologic findings and p53 IHC staining from both non-progressors and progressors, across the complete spectrum of precursor diagnoses in this study.
For the prospective validation cohort, to determine how frequent p53 IHC staining was performed, a random sampling of 500 cases of BE-IND and BE-LGD each revealed that p53 IHC was performed on 68.2% and 59.4% respectively. For NDBE, p53 IHC was utilized exclusively by a group of 7 pathologists and was performed in 57.2% of cases.
IHC for p53 was performed on all lesional blocks in a given case after histopathologic review by a pathologist. To further evaluate for bias based upon reading pathologists, we calculated the rate of p53-ABNL at the individual locations for BE-LGD.

Supplementary Figure Legends
Supplementary Figure S1. Cartoon of endoscopy and biopsy terminology. Diagram illustrates determination of baseline or later surveillance endoscopies as positive (p53-ABNL) or negative (p53-NL) by biopsies. When looking at the endoscopy level, if any biopsy/pathology block at a given endoscopy is p53-ABNL the endoscopy is considered to be positive/abnormal.