Eosinophil Peroxidase Staining Enhances the Diagnostic Utility of the Cytosponge in Eosinophilic Esophagitis

Abstract

INTRODUCTION:

We aimed to assess the diagnostic utility of eosinophil peroxidase (EPX) staining on Cytosponge (CS) samples in eosinophilic esophagitis (EoE).

METHODS:

Esophageal biopsy (BX) samples from adult subjects with EoE were assessed using peak eosinophils per high-power field (eos/hpf), EPX, and the EoE histologic scoring system. EPX staining and eos/hpf were compared (BX vs CS).

RESULTS:

CS EPX positivity correlated with eos/hpf (CS [r = 0.82, P < 0.0001]; BX [r = 0.65, P < 0.0001]) and EoE histologic scoring system (grade [r = 0.62, P < 0.00001]; stage [r = 0.61, P < 0.0001]). CS EPX identified subjects with active EoE (area under the curve = 0.86, P < 0.0001).

DISCUSSION:

The correlation of CS EPX with eosinophilic inflammation and histologic disease severity supports its diagnostic utility in EoE.

INTRODUCTION

Eosinophilic esophagitis (EoE) is an allergic disease characterized by eosinophilic inflammation, tissue remodeling, fibrosis, and esophageal dysfunction. Diagnosis and monitoring of EoE requires multiple invasive endoscopic biopsies (BX) for histopathological assessment of eosinophilic infiltrate (1). Furthermore, manual quantification of tissue eosinophils is a laborious and subjective process (2).

We have shown that automated image analysis of eosinophil peroxidase (EPX) immunohistochemistry (IHC) is an accurate tool for histologic diagnosis of EoE (3) and non-EoE eosinophilic gastrointestinal diseases (4). EPX staining correlates strongly with tissue eosinophil density and is a marker of treatment response (3). Image analysis of EPX staining decreases time, effort, and subjectivity as compared with manual quantification of eosinophils. Importantly, as a marker that incorporates both tissue eosinophilia and eosinophil degranulation, EPX assessments correlate better with clinical symptoms than eosinophils per high-power field (eos/hpf) (5).

The Cytosponge (CS) is a minimally invasive, well-tolerated, and accurate method to assess eosinophil infiltration in EoE (6,7). Although previous work has noted that CS eos/hpf and BX eos/hpf are strongly correlated in subjects with EoE, a pathologist must still manually quantify eosinophil density and normalize counts using a correction factor estimating the area of hpf occupied by tissue. We sought to automate this assessment by evaluating CS EPX IHC for histologic diagnosis of EoE.

METHODS

Analysis was performed on banked tissue blocks from a prior study conducted at Mayo Clinic and the University of North Carolina at Chapel Hill (6). This study was approved by the Institutional Review Board at Mayo Clinic (No. 11-006429; No. 17-007025).

Matched esophageal BX and CS specimens from subjects with EoE were stained with hematoxylin and eosin. BX and CS peak eos/hpf was determined and quantified by a gastrointestinal pathologist. CS eos/hpf was normalized based on the estimated percentage of tissue occupying a hpf (6). The pathologist scored BX specimens using the EoE histologic scoring system (EoEHSS) (8). In addition, slides were stained by EPX IHC using a mouse monoclonal anti-EPX antibody (Lee, Jacobsen, MM.85.2.11) (9). To eliminate the need for a correction factor, EPX staining within a hpf (0.307 mm²) was quantified using automated image analysis (3,10) and reported as EPX positivity (EPX-positive pixels/total tissue pixels). The hpf with the most EPX staining was selected manually. For slides with numerous areas of EPX deposition, multiple fields were analyzed.

Median eosinophil counts and EPX quantifications were compared using the Mann-Whitney U test. Correlations between eos/hpf and EPX staining were assessed using Spearman rho. Receiver operating characteristic curves for EPX staining were generated, and the Youden index was used to determine an EPX positivity cutoff for EoE diagnosis based on the established histologic threshold of 15 eos/hpf. Statistical comparisons and plots were made with GraphPad Prism (version 9; GraphPad software, San Diego, CA).

RESULTS

Demographic information and clinical information are presented in the Supplementary Digital Content (see Supplementary Figure 1, http://links.lww.com/CTG/A880). Matched BX and CS from 62 subjects with EoE (42 active and 20 inactive) were assessed for eos/hpf, EPX, and EoEHSS ratios (grade and stage). EPX staining of BX and CS specimens is shown in Figure 1a from representative subjects with active and inactive EoE. Subjects with active EoE had significantly elevated eos/hpf and EPX in BX and CS specimens (Figure 1a-c).

Figure 1.

EPX staining increases in BX and CS specimens in active EoE. (a) H&E and EPX IHC stains on matched BX and CS specimens from representative subjects with active and inactive EoE. Scale bars = 200 μm. Median peak eosinophils per high-power field (eos/hpf) and EPX positivity in subjects with active vs inactive EoE measured in BX (b) and CS (c) specimens. ****P < 0.0001. BX, biopsy; CS, Cytosponge; EoE, eosinophilic esophagitis; EPX, eosinophil peroxidase; H&E, hematoxylin and eosin.

EPX staining (CS) correlated well with eos/hpf (BX) (r = 0.65 [0.48–0.78] 95% confidence interval [CI], P < 0.0001) and EoEHSS ratios for grade and stage (Figure 2; see Supplementary Figure 1, http://links.lww.com/CTG/A880). Consistent with our previous findings (3), EPX staining correlated strongly with eos/hpf within specimens {BX (r = 0.83 [0.72–0.89 95% CI], P < 0.0001); CS (r = 0.82 [0.71–0.89 95% CI], P < 0.0001)} and across specimens {EPX BX vs EPX CS (r = 0.64, [0.46–0.77 95% CI], P < 0.0001)}. As previously reported (6), eos/hpf correlated in BX and CS (r = 0.68 [0.52–0.80 95% CI], P < 0.0001). CS eos/hpf correlated with EoEHSS ratios {grade (r = 0.54 [0.32–0.70 95% CI], P < 0.0001) and stage (r = 0.54 [0.33–0.70 95% CI], P < 0.0001)}. Notably, EoEHSS correlations were stronger for CS EPX {grade (r = 0.62 [0.43–0.76 95% CI], P < 0.0001) and stage (r = 0.61 [0.42–0.75 95% CI], P < 0.0001)} than CS eos/hpf. Correlations between CS EPX and individual components of the EoEHSS are presented in the Supplementary Digital Content (see Supplementary Figure 2, http://links.lww.com/CTG/A880).

Figure 2.

EPX staining correlates with peak eosinophil counts and histologic disease severity. Correlation matrix showing Spearman rho values for comparisons of eosinophils per high-power field (eos/hpf), EPX positivity, and eosinophilic esophagitis histologic scoring system (EoEHSS) (grade and stage ratios) for matched BX and CS specimens. P < 0.0001 for all comparisons. BX, biopsy; CS, Cytosponge; EPX, eosinophil peroxidase.

We next compared the diagnostic accuracy of CS eos/hpf and CS EPX (see Supplementary Table 2, http://links.lww.com/CTG/A880). Receiver operating characteristic analysis of CS eos/hpf yielded an area under the curve (AUC) = 0.82 ([0.72–0.92 95% CI], P < 0.0001) (Figure 3). CS EPX staining also identified subjects with active EoE with high diagnostic accuracy (AUC = 0.85 [0.75–0.95 95% CI], P < 0.0001). The positive and negative predictive values were 97% and 68%, respectively. The overall agreement was 84% (κ = 0.66 [0.483–0.849 95% CI]). The optimal positive detection threshold for active EoE corresponded to an EPX positivity value of 0.215 (CS) and identified 33 of 42 subjects with active EoE (79% sensitivity) and 19 of 20 subjects with inactive EoE (95% specificity). In summary, we found that CS EPX staining provided comparable diagnostic accuracy with CS eos/hpf.

Figure 3.

CS peak eosinophils per high-power field (eos/hpf) and EPX positivity identify subjects with active EoE. ROC curves for CS eos/hpf and EPX positivity. Diagnosis of active EoE is based on ≥15 peak eos/hpf (BX). BX, biopsy; CS, Cytosponge; EPX, eosinophil peroxidase; ROC, receiver operating characteristic.

DISCUSSION

The current standard of care for diagnosing EoE and monitoring response to therapy requires manual quantification of peak eos/hpf by histological assessment of endoscopic biopsy. In this study, we demonstrate that EPX staining of CS specimens correlates with multiple markers of disease activity, including peak eos/hpf and EoEHSS scores. Moreover, we show that digital assessment of EPX CS staining has similar diagnostic accuracy to CS eos/hpf.

One subject with inactive EoE (BX eos/hpf = 8, BX EPX = 0.2) had 54 CS eos/hpf and markedly elevated EPX staining (CS EPX = 0.63). This instance highlights how CS EPX overcomes sampling bias inherent to a patchy disease. Fifteen of 42 subjects with active EoE had <15 eos/hpf on CS (false negatives). Of note, 8 of these subjects were identified as active by CS EPX. Seven subjects with active EoE were considered inactive by CS eos/hpf (<15 eos/hpf) and CS EPX (<0.215). Some of these subjects revealed elevated BX eosinophils and EPX in the lamina propria, largely sparing the epithelial layer sampled by the CS. For others, the CS failed to collect an adequate tissue sample. Future studies incorporating patient-reported outcomes are needed to investigate the utility of CS EPX in monitoring treatment responses in EoE. In conclusion, we have shown that EPX IHC can be applied to CS. Image analysis of EPX can facilitate efficient histologic analysis with comparable diagnostic accuracy to peak eos/hpf.

CONFLICTS OF INTEREST

Guarantor of the article: Benjamin L. Wright, MD.

Specific author contributions: M.Y.M., S.M.B., and B.L.W.: designed the study. J.A.A., D.A.K., and E.S.D.: recruited the subjects and collected the human specimens. S.G.: analyzed the pathology specimens. M.Y.M., S.M.B., A.P., and W.E.L.: performed the data collection. M.Y.M., A.D.D., and B.L.W.: analyzed the data and wrote the initial draft of the manuscript. J.L.H., J.A.A., D.A.K., E.S.D., and H.K.: provided critical feedback during the revision process. All authors have reviewed and approved the final version of the manuscript.

Financial support: This work was supported by Donald R. Levin Family Foundation, Phoenix Children's Hospital Foundation, and Mayo Clinic Foundation. M.Y.M. is a member of the Immunology Graduate Program and is supported by the Mayo Clinic Graduate School of Biomedical Sciences. J.A.A. serves a consultant for Meritage Pharmacia and has stock ownership in Meritage Pharmacia. D.A.K. serves as a consultant for Celgene and Regeneron. E.S.D. reports research funding from Adare/Ellodi, Allakos, Arena, AstraZeneca, GSK, Meritage, Miraca, Nutricia, Celgene/Receptos/BMS, Regeneron, Revolo, Shire/Takeda. He serves as a consultant for Abbott, Abbvie, Adare/ Ellodi, Aimmune, Akesobio, Alfasigma, ALK, Allakos, Amgen, Arena, Aslan, AstraZeneca, Avir, Biorasi, Calypso, Celgene/Receptos/BMS, Celldex, Eli Lilly, EsoCap, Eupraxia, Ferring, GSK, Gossamer Bio, Holoclara, Invea, Landos, LucidDx, Morphic, Nextstone Immunology, Nutricia, Parexel/Calyx, Phathom, Regeneron, Revolo, Robarts/Alimentiv, Salix, Sanofi, Shire/Takeda, Target RWE, and Upstream Bio. He has also received educational grants from Allakos, Banner, Holoclara, and Invea. H.K. has received grants from the NIH, R37AI71106, R01AI128729, and R01HL117823, and from Mayo Foundation. J.H.S. reports research funding from Allakos, Celgene, and Regeneron. B.L.W. also reports funding from the Consortium of Eosinophilic Gastrointestinal Disease Researchers U54AI117804 (CEGIR), which is part of the Rare Disease Clinical Research Network (RDCRN), an initiative of the Office of Rare Disease Research (ORDR). CEGIR is also supported by patient advocacy groups including American Partnership for Eosinophilic Disorders (APFED), Campaign Urging Research for Eosinophilic Diseases (CURED), and Eosinophilic Family Coalition (EFC). As a member of the RDCRN, CEGIR is also supported by its Data Management and Coordinating Center (DMCC) (U2CTR002818).

Potential competing interests: None to report.

IRB approval: This study was approved by the Institutional Review Board at Mayo Clinic (No. 11-006429; No. 17-007025).

Supplementary Material

ct9-13-e00534-s001.docx ^{(1.8MB, docx)}

ct9-13-e00534-s002.pdf ^{(441KB, pdf)}

ACKNOWLEDGMENTS

We acknowledge Debra Geno and Crystal Lavey for their assistance with subject recruitment, regulatory approvals, and data coordination. We thank Elizabeth Jacobsen, PhD, for providing the anti-eosinophil peroxidase (EPX) antibody. We thank Kelly Shim for her assistance with cataloguing the specimens and optimization of EPX staining quantification.