Emerging Biomarkers for Managing Barrett’s Esophagus

Introduction

In 2023, there were approximately 16,120 deaths from esophageal cancer and 21,560 new cases diagnosed in the US.¹ Esophageal adenocarcinoma (EAC) is a major form of esophageal cancer in Western countries including the United States. Since the 1970’s, the incidence of EAC has risen exponentially for reasons that remain unclear.² EAC has a poor 5-year survival rate of only 20%, largely due to its asymptomatic nature until in advanced stages. Barrett’s esophagus (BE), the known precursor to EAC, is characterized by the replacement of normal squamous epithelium with specialized metaplastic intestinal epithelium in the esophagus, likely as a consequence of chronic gastroesophageal reflux disease (GERD). BE is diagnosed endoscopically by the appearance of columnar mucosa extending 1 cm or more above the gastroesophageal junction, confirmed histologically by the presence of goblet cells. The prevalence of BE in the US general population is estimated to range from 0.04% for long-segment BE to 5% for BE of any length.^2,3 The progression of BE into EAC follows a metaplasia-dysplasia-carcinoma sequence, with nondysplastic BE progressing into low-grade dysplasia, then high-grade dysplasia, and ultimately to EAC. Patients with BE have an estimated annual risk of 0.04% and a lifetime risk of 3% to 5% for developing EAC.^2,4

Current BE/EAC screening recommendations

Current guidelines from several professional societies generally recommend against screening the general population for BE. Instead, screening is recommended for patients with specific risk factors, including chronic GERD symptoms, age over 50 years, male sex, white race, central obesity, smoking history, or a family history of BE or EAC.^5–8 For at-risk patients, the current “gold standard” for screening is an upper endoscopy (EGD) with sampling of any salmon-colored mucosa proximal to the gastro-esophageal junction. Once BE has been diagnosed, the ‘Seattle protocol’ is recommended for surveillance, involving four-quadrant biopsies every 2 cm of BE mucosa in the absence of dysplasia, or every 1 cm if dysplasia is present.⁹ Subsequent endoscopic surveillance intervals are determined by baseline histology, ranging from 3–5 years for non-dysplastic BE (NDBE) to every 3–12 months for those with indefinite dysplasia (IND), low-grade dysplasia (LGD), and high-grade dysplasia (HGD), depending on the guidelines.¹⁰

However, such screening/surveillance strategy has significant limitations. First, many individuals with BE are asymptomatic, and a substantial proportion of patients who develop EAC have no prior history of BE, chronic GERD, or of any reflux symptoms, suggesting that current screening guidelines miss a significant proportion of the at-risk population.¹¹ Furthermore, approximately 25% of patients with NDBE or LGD are reported to progress to HGD or EAC before their next scheduled surveillance EGD, indicating that a subset of BE patients are being undersurveilled.^12,13 At the same time, the annual incidence of EAC among patients with BE and LGD is relatively low (~0.54%), and most patients with BE die from unrelated causes, raising concerns about oversurveillance.¹⁴ The dual problems of over- and under-surveillance, combined with the cost, risks, and access barriers associated with repeated EGDs, underscore the urgent need for novel biomarkers to improve BE risk stratification and management.

Recent advances in molecular technologies have led to the development of promising biomarkers for BE and EAC, including ones that assess DNA methylation, copy number alterations, genomic instability, non-coding RNAs, as well as protein-based markers. Many of these markers have been combined with a growing array of non-endoscopic esophageal mucosa sampling systems, offering cost-effective and minimal invasive alternatives to conventional endoscopic sampling. In this review, we will survey the current landscape of emerging biomarkers for the early detection, risk prediction, and surveillance of BE and EAC, and discuss their potential to improve BE management.

Genomic DNA alterations

Comprehensive analysis of somatic mutations in EAC using whole-exome sequencing, whole-genome sequencing (WGS), and RNA sequencing has been performed across large cohorts of EAC samples. In 2019, Frankell and colleagues analyzed 551 EAC samples and identified a series of significantly mutated genes, including “classical” cancer driver genes such as TP53, CDKN2A, SMAD4, ARID1A and PIK3CA, as well as candidate driver genes, such as SPG20, ERBB2, TLR4, ELMO1, RNF43 and DOCK2.¹⁵ Chromosomal instability and copy number alterations are also common in BE and EAC, including 9p loss encompassing the p16/CDKN2A locus in BE, HGD and EAC cases, losses of chromosomes 5q, 13q and 18q in HGD and EAC, and high-level amplification of ERBB2 on chromosome 17q in EAC.^16–19

Over the last decade, studies of BE have revealed a number of pathogenic variants in oncogenes and tumor suppressor genes, present in approximately 10–20% of BE cases. An analysis of 25 matched cases of BE and EAC revealed frequent mutations in tumor suppressor genes, particularly TP53 and SMAD4, but relatively few oncogene mutations or large-scale genomic alterations.²⁰ This study suggested two major pathways of BE progression: one involving TP53 mutations and genomic doubling, which may account for over 60% of EAC cases, and another involving the serial accumulation of mutations enriched for SMAD4 and CDKN2A alterations. With the exception of TP53 and SMAD4, mutation rates for bona fide tumor suppressors such as CDKN2A (30%) and ARID1A (15%), KMT2D, MYO18B, UNC13C, FBXW7, ATM, FAT2, LRP1B, and SMARCA4 (each <5%) are similar between BE and EAC. Notably, TP53 mutations are detected in fewer than 5% of nondysplastic BE; but in 70% of HGD and EAC.^20,21 Moreover, Stachler et al found that TP53 mutations were more frequent in NDBE that progressed to EAC compared to those that did not (46% vs 5%), although biallelic inactivation of TP53 appears to be necessary for progression to HGD/EAC.^22,23 These results suggest that, while genetic alterations beyond TP53 and SMAD4 may have limited utility for NDBE risk stratification, biallelic TP53 deregulation remains a promising risk marker of BE progression if a reliable assay to assess for this can be developed.

Epigenomic alterations

Epigenetic alterations, including DNA methylation alterations, histone modifications, and chromatin remodeling, are common in BE and EAC. Aberrant DNA methylation alterations are detected in the majority of BE and EAC cases.^24–26 Environmental factors including aging, smoking and obesity may play a role in the formation of these epigenetic alterations.²⁷ Frequently hypermethylated genes include known tumor-suppressor genes, such as APC, CDKN2A (p16INK4a), RUNX3, MGMT, CDH1, and members of the SFRP family.²⁴ While a subset of the hypermethylated genes likely plays a driver role in EAC pathogenesis, many appear to be passenger alterations.^24,26 Aberrant methylation of classic tumor suppressor genes such as CDKN2A and MGMT has been correlated with loss of mRNA and protein expression during the metaplasia-dysplasia-carcinoma sequence, possibly driving the tumorigenesis of EAC.^28,29 Yu et al identified four methylation subtypes of EAC and BE through genome-wide DNA methylation profiling, including a high methylator (HM) subtype enriched for ERBB2 activating events and a higher global mutation burden.²⁶ This study also uncovered a novel epigenetic mechanism by which EAC cells activate the oncogenic ERBB2/EGFR signaling pathway via epigenetically silencing the tyrosine phosphatase non-receptor 13 (PTPN13), specifically in the HM subtype. Of particular relevance to biomarker discovery, a large number of genes and loci have been identified as frequent targets of aberrant methylation in BE and EAC.²⁶ Although the functional significance of the DNA methylation alterations affecting these genes remains unclear, DNA methylation events are promising as biomarkers of BE because of their high frequency and robustness of methods to assess DNA methylation.

Noncoding RNA alterations

Non-coding RNAs (ncRNAs) are a heterogeneous group of RNA molecules that do not encode proteins but regulate gene expression at multiple levels and affect various hallmark behaviors of cancer cells.³⁰ Among the various types of ncRNAs, microRNAs (miRNAs), long non-coding RNAs (lncRNAs), circular RNAs (circRNAs), and exosomal RNAs have been implicated in the development of BE and EAC.³¹ Notably, miRNAs have shown potential as early detection markers for BE.³² However, the reproducibility and clinical translation of these findings have been hampered by inconsistencies across studies, largely due to the wide spread use of non-validated, non-standardized analytical methods.³¹

Protein alterations

In addition to genomic, epigenome, and microRNA alterations, aberrant protein expression has also been observed in BE and EAC. While the functional significance of many dysregulated proteins in the pathogenesis of BE and EAC remains to be uncovered, several have demonstrated utility as biomarkers for BE. Notably, immunohistochemical assays for trefoil factor 3 (TFF3) and TP53, have emerged as robust markers for identifying non-dysplastic BE and advanced dysplasia, respectively, and are discussed in detail below.³³

Extrachromosomal DNA

ecDNA is a distinct type of circular DNA that does not reside on a chromosome and is rarely observed in normal cells.³⁴ ecDNA has been theorized to form following chromothripsis or errors in DNA repair of double- or single-stranded DNA breaks, and often contain highly-expressed oncogenes or drug-resistance genes.³⁴ ecDNA particles lack a functional centromere, leading to unequal segregation of ecDNA during replication, and have been shown to reintegrate into chromosomal DNA in the setting of cellular stress, leading to chromosomal instability and significant intratumoral genetic heterogeneity.³⁴ The presence of ecDNA across multiple different cancer types has been associated with significantly worse survival, including in esophageal cancers.³⁵ Computational analyses of whole-genome sequencing have allowed for improved detection and analysis of ecDNA as compared to previous low-throughput cytogenic methods, and contributed to the potential of ecDNA to serve as a BE/EAC biomarker.^35,36

Early Detection Markers of Barrett’s Esophagus

(see Table1)

Table 1:

Biomarkers for Early Detection of Barrett’s Esophagus

Esophageal Tissue Sampling Method	Biomarker	Study Type	Training Cohort Size	Validation Cohort Size	Test Characteristics	References
EsoCheck / EsoGuard (Lucid Diagnostics, New York, NY) - Swallowable, inflatable balloon-based esophageal sampling device	methylated vimentin (mVIM) and methylated cyclin A1 (mCCNA1)	Prospective multicenter clinical trial	243 participants	NA	Overall sensitivity for detecting all cases (newly diagnosed BE, patients undergoing surveillance of BE, or a new diagnosis of EAC) was 85% (95% CI 0.78–0.93) specificity for controls (patients with no endoscopic evidence of BE and no histologic evidence of intestinal metaplasia on any previous clinical biopsy) was 85% (95% CI 0.79–0.90)	38686933
		Single center, cohort study	124 participants	NA	sensitivity for BE/EAC: 92.9% (95% CI 66.1–99.8) specificity: 72.2% (95% CI 62.1–80.8)	38989889
Esophacap (CapNostics, Concord, NC) - Sponge-on-a-string device	5 methylated DNA marker panel (ZNF682, VAV3, NDRG4, FER1L4, ZNF568)	Multicenter, case-control	199 participants (110 BE and 89 controls)	89 participants (60 BE and 29 controls)	Sensitivity of 93% at 93% specificity and an AUC of 0.97 in the validation cohort for identifying BE	33857451
Cytosponge (Medtronic, Minneapolis, MN) - Sponge-on-a-string device	Methylated TFP12	Retrospective analysis of multicenter, case-control	30 participants (20 BE, 10 controls)	278 participants (149 BE and 129 controls)	Sensitivity of 82.2% for identifying BE and specificity of 95.7%	29084829
	Immunohistochemical staining of TFF3	Multicenter, randomized controlled trial	Usual care (6388) vs intervention group (6834)	NA	Over an average of 12 months of follow-up, 140 (2%) of 6834 participants in the intervention group and 13 (<1%) of 6388 participants in the usual care group were diagnosed with BE	32738955

BE, Barrett’s Esophagus; CI, Confidence Interval; EAC, Esophageal Adenocarcinoma; NA, Not Applicable

Methylated DNA based biomarkers

Methylated DNA based biomarkers are being used in a commercially available multi-target esophageal cytology DNA assay (MT-ecDNA) for screening for Barrett’s esophagus (BE), named EsoGuardÒ. The MT-ecDNA assay, performed on samples collected via a non-endoscopic esophageal cytology collection device (EsoCheckÒ, Lucid Diagnostics, New York, NY), uses next-generation sequencing to detect aberrantly methylated vimentin (mVIM) and methylated cyclin A1 (mCCNA1).³⁷ Moinova et al. demonstrated that mVIM has a sensitivity of 91–100% and specificity of 100% for detecting nondysplastic BE (NDBE) in both formalin fixed paraffin embedded (FFPE) tissue and esophageal brushings.⁴⁰ A positive MT-ecDNA test result suggests the presence of cellular changes associated with BE or EAC and warrants a confirmatory EGD. Of note, the assay does not distinguish BE at different stages along the metaplasia-to-neoplasia sequence.

In a prospective multicenter study, 243 participants underwent sampling with the EsocheckÒ device followed by a MT-ecDNA test prior to a scheduled endoscopy. The sensitivity for detecting NDBE was 84%, with an overall specificity of 85% among patients without endoscopic or histologic evidence of intestinal metaplasia.⁴¹ These findings were consistent with results from an independent single-center study at a Veteran’s Administration, which evaluated 124 participants with symptomatic GERD who met current guideline recommendations for endoscopic BE screening, and underwent mucosal sampling with the EsocheckÒ device followed by an EGD.⁴² The MT-ecDNA test on samples collected via EsocheckÒ demonstrated a sensitivity of 92.9% (95% CI, 66.1–99.8) and specificity of 72.2% (95% CI, 62.1–80.8) for detecting either NDBE or early EAC, using EGD with pathology as the reference standard.⁴²

Another methylated DNA biomarker panel for early detection of NDBE was developed using esophageal tissue samples collected using the EsophaCap system (PAVMed, New York, NY), a non-endoscopic “sponge-on-a-string” device.⁴³ Iyer et al. validated a five-gene methylation panel (ZNF682, VAV3, NDRG4, FER1L4, ZNF568) across a training cohort (N=199) and an independent validation cohort (N=89), achieving a sensitivity of 93% (95% CI, 84%–98%), specificity of 93% (95% CI, 77%–99%), and an area under the curve (AUC) of 0.97 (95% CI, 0.95–1.0).⁴⁴ Secondary analyses indicated that several three-gene combinations also provided comparable accuracy.⁴⁴ Kalra et al developed a separate biomarker panel based on EsophaCap tissue samples, combining seven novel methylated DNA markers with five previously identified markers, resulting in a 12-marker candidate assay.⁴⁵ In an independent cohort (N=24), a model incorporating three biomarkers (USP44, TBCD1D30, and NELL1), along with age and sex, yielded an AUC of 0.776 (95% CI, 0.583–0.968), although the small sample size and non-screening population limit the generalizability of these findings.⁴⁵

In a multicenter case-control BEST2 study, Chettouh et al. identified four hypermethylated genes (TFPI2, TWIST1, ZNF345 and ZNF569) in Cytosponge samples.^46,47 In a validation cohort, TFP12 alone achieved an AUC of 0.88 (95% CI, 0.84–0.91), with a sensitivity of 78.5% and specificity of 96.9% for distinguishing NDBE from controls.

Protein-Based Biomarkers

Trefoil factor 3 (TFF3), a mucin-associated peptide expressed on goblet cells, has emerged as a promising biomarker for non-endoscopic detection of Barrett’s Esophagus (BE). In esophageal cytology samples collected with the Cytosponge, a minimally invasive esophageal cell collection device, TFF3 immunohistochemical (IHC) staining could detect patients with BE. In an initial study, TFF3-Cytosponge achieved a sensitivity of 78% and specificity of 94% for BE detection.⁴⁸ In the prospective BEST1 trial, TFF3 had a sensitivity of 73% and specificity of 94% for detecting BE segments ≥1 cm.⁴⁹ In the larger BEST2 case-control study (N=1042), TFF3-Cytosponge showed a sensitivity of 80% and specificity of 92% for identifying BE.⁵⁰ The randomized BEST3 trial further validated the clinical utility of the Cytosponge-TFF3. Screening with Cytosponge-TFF3 in patients on chronic acid suppression therapy significantly increased BE detection rates compared to conventional patient care (2% vs <1%, p<0.0001).^51,52 These findings suggest that TFF3-Cytosponge could be an effective, scalable, and less invasive alternative to traditional endoscopic screening, particularly for identifying patients who may benefit from confirmatory endoscopy and surveillance.

microRNA based biomarkers

MicroRNAs (miRNA) are small, non-coding RNAs that can regulate gene expression and exhibit distinct expression profiles in BE and EAC compared to normal tissues.⁵³ Due to their structural stability and ease of detection in blood and archived tissue samples, miRNA have been an attractive biomarker candidate.

The EMERALD trial, conducted a large, international, multicenter evaluation of serum-derived miRNA biomarkers for detection of BE and EAC.⁵⁴ A six-miRNA panel was developed and validated using serum samples, achieving an AUROC of 91.9%, with a sensitivity of 82.5% and specificity of 90.5% for distinguishing BE/LGD/HGD/EAC from controls. Although the EMERALD study demonstrates the potential of miRNA biomarkers for minimally invasive detection of BE and EAC, it is important to note that the enrolled subjects were identified based on clinical presentation rather than through prospective screening. As a result, further validation studies are needed in asymptomatic, screening-relevant populations to determine the broader clinical utility of this approach.

Alternative class of biomarkers—volatile organic compounds

Another emerging class of biomarkers proposed for early BE detection is volatile organic compounds (VOC). In a study by Peters et al., an “electronic nose” device was used to analyze breath samples from 402 participant undergoing a clinically indicated EGD. A distinct VOC profile associated with BE (N=129) was identified, achieving a sensitivity of 91% (95% CI, 84% to 95%) and specificity of 74% (95% CI 69% to 79%).⁵⁵ While these early findings are encouraging, both miRNA and VOC technologies need validation in large, independent cohorts of screen-relevant individuals to demonstrate reproducibility across independent studies.

HGD/EAC Early Detection Markers

(See Table 2)

Table 2:

Biomarkers for Early Detection of High-Grade Dysplasia or Esophageal Adenocarcinoma

Esophageal Tissue Sampling Method	Biomarker	Study Type	Training Cohort Size	Validation Cohort Size	Test Characteristics	References
EsoCheck / EsoGuard (Lucid Diagnostics, New York, NY) - Swallowable, inflatable balloon-based esophageal sampling device	methylated vimentin (mVIM) and methylated cyclin A1 (mCCNA1)	Prospective multicenter clinical trial	243 participants	NA	Sensitivities of 81% for LGD and IND, 78% for HGD, and 100% for EAC	38686933
Esophageal brushings	4-marker methylated DNA panel (Cg6522, YPEL3, Up35-1, and Up10)	Nonrandomized observational study	194 tissue samples	129 tissue samples	Sensitivity of 80% for HGD and EAC Specificity of 67.6% for NDBE and 96.3% for normal squamous tissue.	35705525
Micro-dissected sections of FFPE tissue samples	DNA content abnormality as quantified by DNA flow cytometry	Observational study	153 tissue samples	NA	Sensitivity of 95% and a specificity of 85% for HGD	28642331
Esophageal brushings	Aneuploidy as measured by RealSeqS and classification according to the Barrett’s aneuploidy decision (BAD) tool	Cross-sectional study	79 participants	268 participants	Classification of “Very-BAD” or “Maybe-BAD” was reported to have a sensitivity of 71.4% for HGD and 98.8% for EAC, with a specificity of 96.4% for unaffected controls.	33493502
Cytosponge (Medtronic, Minneapolis, MN) - Sponge-on-a-string device	Three protein biomarkers (P53, c-Myc, and Aurora kinase A), two methylation markers (MYOD1 and RUNX3), glandular atypia, and TP53 mutation status	Multi-center cohort study	468 participants	65 participants	In the validation cohort, 25 were classified as low-risk (24 NDBE, 1 HGD), 35 as moderate-risk (27 NDBE, 8 HGD), and 5 as high-risk (0 NDBE, 5 HGD) – with a classification of low-risk being associated with a 96% probability of having NDBE	28404010
	Model consisting of the presence of atypia, p53 overexpression, and 17 clinical demographic variables	Retrospective analysis of a multi-center cohort study	557 participants	334 participants	Sensitivity of 80% and specificity of 87% for the combined biomarker and clinical data model for detecting HGD/EAC	35030332

BAD, Barrett’s aneuploidy decision tool; BE, Barrett’s Esophagus; CI, Confidence Interval; EAC, Esophageal Adenocarcinoma; FFPE, Formalin-Fixed Paraffin Embedded; HGD, High-grade dysplasia; IND, Indefinite dysplasia; NA, Not Applicable; NDBE, Non-Dysplastic Barrett’s Esophagus

Methylated DNA based assays

In addition to being used to detect NDBE, methylated DNA has also been shown to be used for the detection of HGD/EAC. In a recent study, Yu et al. conducted DNA methylation profiling on endoscopic biopsies samples and identified five candidate methylation markers (Cg6522, YPEL3, POU3F1, Up35–1, and Up10) that were significantly hypermethylated in HGD/EAC samples.⁵⁶ A four-marker subset panel (Cg6522, YPEL3, Up35-1, and Up10) was validated in an independent cohort of esophageal brushings and yielded a sensitivity of 80% for HGD (95% CI, 57.4%–100.0%) and 82.5% (95% CI, 61.0%–100.0%) for EAC, and a specificity of 67.6% (95% CI, 49.2%–85.9%) for NDBE and 96.3% (95% CI, 89.0%–100.0%) for normal squamous tissue.⁵⁶

The methylated DNA biomarkers of mVIM and mCCNA1, when assessed in samples collected with the Esocheck device, have demonstrated the ability to detect HGD and EAC. In a prospective clinical trial consisting of 243 patients with or without a history of BE who successfully underwent EsocheckÒ esophageal mucosal sampling prior to a scheduled EGD, the test achieved sensitivities of 81% for LGD and IND (N=16), 78% for HGD (N=23), and 100% for EAC (N=18) at a specificity of 85% (95% CI 79–90) for normal squamous esophagus.⁴¹

Genomic DNA based markers

Genomic instability and copy number abnormalities are one of the important genomic alterations to drive the progression of BE to EAC, and have been investigated as potential biomarkers dysplasia and malignancy.^15,19,57 In a retrospective, single-center study, Choi et al. reported normal diploid DNA histograms for BE patients without dysplasia and increasing frequencies of DNA content abnormality as BE progressed from low to high grade dysplasia, with abnormal DNA content identified in 21.1% of LGD (N=38) and 95% of HGD cases (N=80).⁵⁸ Using DNA content abnormality as a diagnostic marker of HGD yielded a sensitivity of 95% and a specificity of 85%, though no independent validation cohort was reported. This study demonstrated the feasibility of this aneuploidy as a biomarker for HGD and EAC but was of little clinical significance since it was done using esophageal biopsies, for which histology is accurate when performed by two GI pathologists.

More recently, Markowitz, Douville and colleagues developed a method called repetitive element sequencing (RealSeqS) to assess for aneuploidy in esophageal brushings collected as part of a cross-sectional multicenter study.⁵⁹ RealSeqS uses massively parallel sequencing of ~350,000 genome-spanning regions with a single primer pair to interrogate specific chromosome arm alterations.⁶⁹ A classifier was developed based on a “genome-wide aneuploidy score (GAS).” and specific chromosomal changes, resulting in a Barrett’s aneuploidy decision (BAD) tool categorizing BE samples as “Not-BAD,” “Maybe-BAD”, and “Very-BAD.” In an independent validation set of 268 participants, a classification of “Very-BAD” or “Maybe-BAD” demonstrated a sensitivity of 71.4% for HGD (N=28) and 98.8% for EAC (N=83), with a specificity of 96.4% for unaffected controls (N=84).

Mixed Biomarker Panels

Several studies have also demonstrated the utility of combining multiple biomarkers to identify HGD and EAC. In the multicenter cohort BEST2 study, a panel of biomarkers including three protein biomarkers (TP53, C-MYC, and Aurora kinase A), two methylated DNA markers (MYOD1 and RUNX3), glandular atypia, and TP53 mutation status were used to develop a model to categorize Cytosponge samples as low-, medium-, or high-risk for dysplasia with the goal of more accurately selecting candidates for endoscopy.⁶⁰ Building on this study, Pilonis et al. conducted a retrospective analysis of esophageal mucosal samples collected as part of the BEST2 and BEST3 studies to develop an endoscopy risk stratification model including the presence of atypia (by H&E staining), p53overexpression (by IHC staining), and 17 clinical-demographic variables.⁶¹ In a validation cohort of 334 participants, this mixed model achieved a sensitivity of 80% and specificity of 87% for detecting HGD/EAC, yet did not outperform the biomarker model alone (sensitivity 89%, specificity 84%).

Potential alternative biomarkers

As previously described, a liquid biopsy assay was developed as part of the EMERALD study, consisting of a panel of six miRNA biomarkers and a corresponding machine learning algorithm. It has demonstrated promise in early detection of HGD/EAC, achieving sensitivity of 84% for identifying EAC and a specificity of 90.5%.⁵⁴

BE to EAC Risk Prediction Markers/Assays

(See Table 3)

Table 3:

Biomarkers Associated with Progression of Barrett’s Esophagus to Esophageal Adenocarcinoma

Esophageal Tissue Sampling Method	Biomarker	Study Type	Training Cohort Size	Validation Cohort Size	Test Characteristics	References
FFPE blocks from esophageal biopsies	Esopredict (Previse, Halethorpe, MD) - four methylated DNA biomarkers (p16, HPP1, RUNX3, and FBN1) + age	Retrospective multicenter study	99 participants	110 participants	In the validation cohort, 11 were high-risk (4 non-progressors and 7 progressors) and 29 were low-risk (26 non-progressors and 3 progressors). High-risk level patients were 15.2 times more likely to progress as compared to low-risk patients	39140473
FFPE blocks from esophageal biopsies	p53 IHC	Retrospective case-control study + prospective cohort study for validation	672 participants	1,449 participants	In the prospective cohort, p53 abnormality was associated with a hazard ratio of 12.5 (95% CI, 2.05–5.29) for progression	34757142
FFPE blocks from esophageal biopsies	AGRN IHC	Retrospective multicenter study	129 participants	161 participants	Using a ≥1% cutoff value for AGRN loss across two biopsies was able to separate progressors and non-progressors with a sensitivity of 80.2% and a specificity of 54.8%	34785582
FFPE blocks from esophageal biopsies	Genome-wide copy number instability as measured using shallow WGS	Retrospective case-control study	88 participants	76 participants	In the validation cohort, 78/142 (55%) non-progressors were classified as low-risk and 55/71 (77%) progressors were classified as high-risk.	32895572
Micro-dissected FFPE blocks from esophageal biopsies	BarreGEN (Interpace Diagnostics, Pittsburgh, PA) - panel of 10 genomic loci for 24 DNA markers to subsequently calculate a mutational load (ML) score by PCR and quantitative capillary electrophoresis	Case-control study	69 patients	NA	Sensitivity of 100% at a ML ≥ 0.5 and specificity of 96% at ML ≥1.5	26010308
		Single-center retrospective pilot	28 patients	NA	Using a ML threshold of ≥1.5 yielded a risk of progression to HGD of 33% and a sensitivity of 100% and specificity of 85%.	30815274
Micro-dissected FFPE blocks from esophageal biopsies	DNA content abnormality, quantified by DNA flow cytometry	Observational study	153 tissue samples	NA	DNA content abnormality on DNA flow cytometry was a significant risk factor for progression to HGD or EAC, with a hazard ratio of 7.0 and 20.0, respectively	28642331
Esophageal brushings	Aneuploidy as measured by RealSeqS and classification according to the Barrett’s aneuploidy decision (BAD) tool	Cross sectional study	79 participants	268 participants	4 patients were identified to progress to HGD/EAC within 3 years of study entry — of which three had antecedent brushings classified as “Very-BAD” and one as “Maybe-BAD”, and no patients with samples classified as “Not-BAD” were reported to have disease progression	33493502
FFPE blocks from esophageal biopsies	Extrachromosomal DNA (ecDNA)	crosssectional study + prospective case-control cohort	206 participants (cross sectional surveillance cohort) 80 participants (prospective case-control)	NA	ecDNA was identified at a significantly higher rate in progressors than non-progressors (33% vs 3%, p=3.3×10–4)	37046089
FFPE blocks from esophageal biopsies	TissueCypher (Castle Biosciences, Friendswood, TX) - IHC labeling of FFPE samples + processing by image analysis to generate a risk score that classifies samples as low-, intermediate-, or high-risk.	Single-blinded, case-control study	268 patients	NA	The sensitivity and specificity of the TissueCypher system using a 3-tier classification (low, intermediate, high-risk) was 29% and 86% respectively Sensitivity and specificity of TissueCypher with a 2-tier classification (low, intermediate/high-risk combined) was 40% and 86%, respectively	32079863

BAD, Barrett’s aneuploidy decision tool; BE, Barrett’s Esophagus; CI, Confidence Interval; EAC, Esophageal Adenocarcinoma; ecDNA, extrachromosomal DNA; FFPE, Formalin-Fixed Paraffin Embedded; HGD, High-grade dysplasia; IHC, Immunohistochemistry; ML, Mutational Load; NA, Not Applicable; NDBE, Non-Dysplastic Barrett’s Esophagus

Methylated DNA based assays

An algorithm incorporating four methylated DNA biomarkers (CDKN2A, HPP1, RUNX3, and FBN1) with patient age has been developed to estimate the risk of disease progression in patients with established BE and is now available as a laboratory-developed test under the name Esopredict (Previse, Halethorpe, MD). In a retrospective case-control study, Laun et al. evaluated participants with an initial histologic diagnosis of NDBE, IND, or LGD who either progressed to HGD/EAC or had at least five-years of follow-up without progression. The initial tissue biopsy was stratified into low risk, low-moderate risk, high-moderate risk, or high risk.^62,63 After training the model on 99 participants, the authors validated the panel in an independent cohort of 110 participants, where 11 patients were classified as high-risk (4 non-progressors and 7 progressors) and 29 as low-risk (26 non-progressors and 3 progressors). High-risk patients were found to be 15.2 times more likely to progress to HGD/EAC compared to low-risk patients.⁶³

Protein based assays

IHC staining for p53 is one of the earliest biomarkers studied for BE progression, though its clinical utility has historically been limited by interobserver variability.⁶⁴ In a recent study, Redston et al. standardized the p53 IHC grading criteria.⁶⁵ In the retrospective case-control cohort of patients with NDBE, IND, and LGD, abnormal p53 classification had a sensitivity of 50.8% and a specificity of 98.3% for progression to HGD/EAC. In the prospective cohort, p53 abnormality was noted to successfully stratify patients as progressors and non-progressors regardless of their baseline histology of NDBE, IND, or LGD, and p53 abnormality was associated with a hazard ratio of 12.5 (95% CI, 2.05–5.29) for progression. These findings are supported by prior studies showing that pathogenic TP53 mutations can be detected up to 9 years before histologic progression.²³ Stachler et al have also reported that some of these pathogenic TP53 variants persist through longitudinal samples collected at pre- and post-progression timepoints, suggesting that these early TP53 pathogenic variants may serve as early markers of neoplastic transformation.²³

Genomic DNA based assays

Killcoyne et al. applied shallow WGS to evaluate genome-wide copy number instability as a predictor of progression from NDBE to HGD/EAC.⁶⁷ Using a discovery cohort of 88 patients (43 non-progressors and 45 progressors), they developed a model incorporating copy number instability and a measure of overall complexity, which was validated in an independent cohort of 76 patients (with 213 samples) undergoing surveillance for BE. Notably, risk classification was independent of histopathology, and 78% of progressors had least one high-risk sample one to two years prior to HGD/IMC diagnosis, suggesting potential for risk stratification.

In another study, Paulson et al. performed deep WGS of 427 BE tissue samples collected longitudinally across 80 patients -- 40 progressors to early EAC and 40 non-progressors.²² Through a multi-step analysis, the authors developed a six-feature model incorporating mono- vs bi-allelic TP53 inactivation, gains in chromosome 18:18–25 Mb gains, a functional mutation in PCDH17, and their interactions. This model accurately classified 38 of 40 progressors and 38 of 40 non-progressors. Although promising, the findings need to be validated in an independent cohort to confirm the reproducibility of the results if the assay were used in clinical care.

A commercial assay known as BarreGENÒ (Interpace Diagnostics, Pittsburgh, PA) has been developed to quantify genomic instability in BE. BarreGENÒ analyzes 24 DNA markers across a panel of 10 genomic loci from micro-dissected samples of FFPE tissue to calculate a mutational load (ML) score, in order to assess the presence of loss of heterozygosity and of microsatellite instability at each locus.⁶⁸ In a case-control study with 69 patients (46 non-progressors and 23 progressors), an ML score ≥1.5 predicted progression to high-grade dysplasia (HGD) or EAC with 69.6% sensitivity and 96% specificity.⁶⁸ A follow-up study in 28 patients with indefinite dysplasia (IND) (20 non-progressors, 8 progressors) reported a 33% risk of progression to HGD of with 85% specificity using the same ML score (≥1.5).⁶⁹

Other methods for detecting DNA content abnormalities such as DNA flow cytometry⁵⁸ and “RealSeqS” analysis with Barrett’s Aneuploidy Decision (BAD) classification⁵⁹, may also have a role in helping to risk stratify patients for progression to advanced disease. Choi et al. reported that among patients with baseline LGD/IND, DNA content abnormality identified by flow cytometry was a significant predictor of progression to HGD or EAC, with a hazard ratio of 7.0 and 20.0, respectively.⁵⁸ Similarly, Douville et al. reported that among 92 NDBE/LGD cases, four individuals progressed to HGD/EAC within three years of study entry: three had antecedent brushings classified as “Very-BAD” and one as “Maybe-BAD”.⁵⁹ However, both of these findings have yet to be validated in larger, prospective cohorts.

Extrachromosomal DNA based marker

In addition to abnormalities in genetic, epigenetic, and protein expressions that are associated with BE progression, ecDNA has emerged as a promising biomarker.³⁵ Luebeck et al. analyzed WGS data from tissue samples from a cross-sectional surveillance cohort of 206 patients with BE and 20 EAC tumor samples, and found that ecDNA was significantly enriched in EAC and associated with increasing tumor stage.³⁶ The authors also detected ecDNA in 33% of progressors versus 3% of non-progressors (p=3.3 × 10⁻⁴) in a prospective, case-control cohort of 80 BE patients with surveillance biopsies taken from at least two different points in time (40 progressors to EAC and 40 non-progressors). Interestingly, ecDNA is often found to co-exist with TP53 alterations, evolve during malignant progression, and harbor EAC-associated oncogenes genes such as ERBB2, KRAS, and MYC as well as a variety of immunomodulatory genes.³⁶ While these findings highlight the potential of ecDNA as a novel biomarker for risk stratification in BE, further validation in larger, prospective cohorts is needed to determine if an ecDNA based assay could be used clinically to predict the risk of NDBE progressing to HGD or EAC.

Mixed Biomarker Panels

Another commercial assay that has been developed for risk stratification of patients with NDBE, IND, or LGD is TissueCypher^™ (Castle Biosciences, Friendswood, TX). This assay is based on immunohistochemical assays of a panel of proteins, including p16, AMACR, p53, HER2, CK20, CD68, COX-2, HIF-1α, and CD45RO, together with Hoechst staining for fluorescent detection of DNA in the tissue.⁷⁰ Stained slides are analyzed using an image analysis algorithm, and the resulting quantified features are integrated into a composite risk score that classifies samples as low-, intermediate-, or high-risk for progression.

In a single-blinded, case-control study, Davison et al. evaluated the performance of the TissueCypher assay in a cohort of 58 patients with NDBE/IND/LGD who progressed to HGD/EAC and 210 patients without progression.⁷¹ The sensitivity and specificity of the assay using a three-tier classification (low, intermediate, and high-risk) was 29% and 86% respectively. When the intermediate- and high-risk categories were combined into a two-tier classification, sensitivity improved to 40% while specificity remained at 86%. Among the 51 patients who progressed within five years, 14 (27%) were classified as high-risk, 6 (12%) as intermediate-risk, and 31 (61%) as low-risk. In contrast, among the 210 non-progressors, 13 (6%) were classified as high-risk, 18 (9%) as intermediate-risk, and 179 (85%) as low-risk.

Most recently, Khoshiwal et al. conducted a blinded cohort study comparing the accuracy of the TissueCypher assay with that of 30 pathologists from five countries in risk-stratifying BE.⁷² Using a cohort of 154 BE patients with LGD, including 24 who progressed to HGD/EAC within five years, the TissueCypher assay demonstrated a higher sensitivity for identifying progressors compared to pathologists, although the difference was not statistically significant (70.8% vs 63.2%, p=0.12).

Overall, TissueCypher offers an assay for predicting progression from NBDE/LGD that performs at least as accurately as expert pathologist interpretation, with the added benefits of consistency and elimination of interobserver variability. However, the assay’s significant costs, combined with its performance being comparable to histologic examination, are barriers to its widespread clinical adoption.

In a separate study, Timmer et al. prospectively evaluated clinical variables and six molecular markers (p16, p53, Her-2/neu, 20q, MYC, and aneusomy) using brush cytology specimens collected from 428 NDBE patients to develop a multivariable prediction model for progression to HGD/EAC.⁷³ The authors reported that the most robust model included age, circumferential Barrett’s length, p16 loss, MYC gain, and aneusomy, yielding a sensitivity of 86% and a specificity of 54%.⁷³ When used to classify NDBE patients into high-risk or low-risk groups, the model demonstrated significantly different 5-year progression-free survival rates: 98.6% in the low-risk group versus 92.8% in the high-risk group (HR 8.7, 95% CI 2.6–29.8; log-rank p<0.001).

Discussion/Summary

Despite advances in endoscopic screening and surveillance, clinical management of BE faces significant challenges including both over- and under-diagnosis, covert progression and suboptimal risk stratification. Emerging molecular biomarker assays have the potential to enable early detection, improve risk prediction and personalize surveillance strategies. Among these, DNA methylation-based biomarkers, particularly when detected by non-endoscopic devices, have shown substantial promise as assays that can improve the detection NDBE, HGD and high-risk NDBE. Moving forward, it will be critical to refine the complex, multifaceted biomarker models and conduct large, randomized validation studies to ensure that these emerging biomarker assays show similar performance across a large population. It will also be important to address cost and accessibility barriers that may limit their widespread adoption. Future advances in biotechnology and artificial intelligence are poised to accelerate biomarker discovery, ensuring that the future assays are robust and clinically impactful. The integration of molecular biomarkers with minimally invasive sampling approaches holds the promise to transform BE management—ushering in a new era of screening, surveillance, and early detection that could substantially reduce morbidity and mortality from EAC.

Synopsis:

Comparative analyses of normal esophageal tissue, Barrett’s Esophagus and esophageal adenocarcinoma have identified distinctive molecular alterations in genomic DNA, epigenetics, non-coding RNA, and proteins that have shown promise as disease-specific biomarkers. Some of the best characterized and promising biomarkers for managing BE are multi-target esophageal cytology DNA assays based on methylated DNA and immunohistochemical staining of TFF3. Additional biomarkers including microRNA, measures of genomic instability, mixed-method panels, and other novel biomarkers (e.g. volatile organic compounds and saliva microbiome) remain in early development and will need further validation in large, prospective studies prior to clinical use.

Key Points:

• Emerging biomarkers for Barrett’s Esophagus consist of a wide array of molecular alterations in genomic DNA, epigenetic alterations, non-coding RNA, and proteins.

• Some of the most characterized and promising biomarkers for BE are multi-target esophageal cytology DNA assays based on methylated DNA and immunohistochemical staining of TFF3.

• Other proposed biomarkers including microRNA, measures of genomic instability, volatile organic compounds, saliva microbiome, systemic bile acids, and mixed-method panels remain early in development and will need further validation in large, prospective studies

Clinics Care Points.

• Comparative analyses of normal esophageal tissue, Barrett’s Esophagus (BE), and high-grade dysplasia (HGD) / esophageal adenocarcinoma (EAC) have identified distinctive molecular alterations in genomic DNA, epigenomic features, non-coding RNA, and proteins that have shown promise as disease-specific biomarkers.

• Several multi-target esophageal cytology DNA assays (MT-ecDNA) based on methylated DNA biomarkers have demonstrated potential in detecting BE and HGD/EAC including one commercially available assay (EsoGuardÒ). Several additional MT-ecDNA assays that have been paired with non-invasive cytology sampling devices (e.g. EsophaCap and Cytosponge), though these still remain in development.

• Global measures of genomic instability, copy number alterations, aneuploidy, and extrachromosomal DNA have demonstrated promise in predicting progression of BE to EAC, but still need validation in larger, prospective cohorts.

• MicroRNA biomarker targets have shown promise as a non-invasive, blood-derived biomarker for differentiating patients with BE and EAC from non-disease controls, though larger studies based on a prospective, screening-based cohort are still needed.

• There are many promising molecular markers that have been developed or are in development for BE and HGD/EAC screening and surveillance and for risk stratification for BE progression to HGD or EAC, but to date none have been subjected to a prospective cohort study in a screen-relevant population, which is needed to determine accurately the sensitivity and specificity of the tests when used in clinical care.