Utility of major basic protein, eotaxin-3, and mast cell tryptase staining for prediction of response to topical steroid treatment in eosinophilic esophagitis: analysis of a randomized, double-blind, double dummy clinical trial

Summary

Inflammatory factors in eosinophilic esophagitis (EoE), including major basic protein (MBP), eotaxin-3 (EOT3) and mast cell tryptase (TRP), may predict treatment response to topical corticosteroids (tCS). We aimed to determine whether baseline levels of these markers predict response to tCS for EoE. To do this, we analyzed data from a randomized trial comparing two topical steroids for treatment of newly diagnosed EoE (NCT02019758). A pretreatment esophageal biopsy was stained for MBP, EOT3, and TRP to quantify tissue biomarker levels (cells/mm²). Levels were compared between histologic responders (<15 eos/hpf) and nonresponders (the primary outcome), and endoscopic responders (EREFS<2) and nonresponders. Complete histologic response (<1 eos/hpf) was also assessed, and area under the receiver operator characteristic curve (AUC) was calculated. We also evaluated whether baseline staining predicted symptom relapse in the trial’s off-treatment observation phase. Baseline samples were evaluable in 110/111 subjects who completed the randomized trial. MBP levels were higher in nonresponders (n = 36) than responders (704 vs. 373 cells/mm²; P = 0.007), but EOT3 and TRP levels were not statistically different. The combination of all three stains had an AUC of 0.66 to predict response. For complete histologic response, baseline TRP levels were higher in nonresponders (n = 69) than responders (370 vs. 268 mast cells/mm²; P = 0.01), with an AUC of 0.65. The AUC for endoscopic response was 0.68. Baseline staining did not predict symptom recurrence after remission. Pretreatment MBP, EOT3, and TRP levels were not strongly or consistently associated with histologic or endoscopic response to topical steroids. While elevated TRP levels may be associated with nonresponse compared with complete response, the magnitude and predictive utilities were modest. Novel methods for predicting steroid response are still required.

INTRODUCTION

Eosinophilic esophagitis (EoE) has rapidly emerged as a major cause of upper gastrointestinal (GI) morbidity with a rapidly increasing incidence and substantial healthcare burden.^1–3 Diagnosis is based on symptoms of esophageal dysfunction and demonstration of an esophageal eosinophilic infiltrate, in the absence of competing causes of eosinophilia.^4,5 After diagnosis, first-line pharmaceutical treatments consist of either proton pump inhibitors (PPIs) or swallowed topical corticosteroids (tCS).⁶ There has been substantial research interest in assessing the efficacy of tCS, which has been demonstrated in multiple observational studies and clinical trials,^7,8 and more recently in developing esophageal-specific tCS formulations.^9–11 As of yet, there are still no treatments approved for EoE in the United States, but a tCS formulation has recently been approved for use in Europe.¹⁰

A recent meta-analysis has found that while the two major forms of topical steroids, a slurry or budesonide of fluticasone dispensed from an asthma inhaler, have good efficacy,⁸ in many studies nonresponse to steroids can be seen in more than a third of patients.¹² Nonresponse may be even more common in the real-world, nonclinical trial setting.^13,14 Understanding which patients would most benefit from topical steroid therapy would allow these treatments to be individualized or targeted, and patients who are likely to be nonresponders could be directed to other treatments such as dietary elimination, or perhaps biologics, which are in development for EoE.¹⁵ However, few predictors of steroid response have yet to be identified, and none can be applied to a wide population of EoE patients. We previously conducted a retrospective analysis of tCS nonresponders and assessed biomarkers related to allergic inflammation in EoE.¹⁴ These included major basic protein (MBP), an eosinophil granule protein,¹⁶ eotaxin-3 (EOT3), a potent cytokine and the most up-regulated gene in the EoE transcriptome,¹⁷ and tryptase (TRP), a mast cell marker.¹⁸ Initial results suggested that lower EOT3 and TRP levels were associated with nonresponse,¹⁴ but it was necessary to validate these results in a larger prospective sample before applying them clinically.

Therefore, the aim of this study was to determine whether baseline levels of MBP, EOT3, or TRP predict response to tCS treatment for EoE, using pretreatment samples collected during the course of a randomized controlled trial. Based on our prior data, we hypothesized that lower levels of EOT3 and TRP, but not MBP, would predict histologic response.

MATERIALS AND METHODS

Study design, patients, and parent study

We conducted an analysis of samples collected during the conduct of a randomized double-blind, double dummy clinical trial that compared two tCS preparations, oral viscous budesonide (OVB) and fluticasone dispensed from a multidose inhaler (MDI), for treatment of newly diagnosed EoE cases (NCT02019758).^19,20 Full details of the parent study and protocol have previously been published. The study was approved by the University of North Carolina Institutional Review Board, and all patients provided informed consent for participating in the study as well as for future use of biospecimens.

In brief, patients aged between 16 and 80 were eligible if they had a new diagnosis of EoE according to published consensus guidelines at that time. Key exclusions were: concomitant diagnosis of eosinophilic gastroenteritis or other non-EoE eosinophilic GI disorder; swallowed tCS or systemic steroids for any condition in the 4 weeks prior to baseline endoscopy; inability to pass a standard upper endoscope due to esophageal narrowing or stricture; previous esophageal surgery; esophageal or gastric cancer; or any other medical condition or instability that precluded endoscopy. After enrollment, patients were randomized 1:1 to treatment with either OVB (1 mg BID) + a placebo inhaler or fluticasone (MDI 880 mcg BID) + placebo slurry. After 8 weeks of treatment upper endoscopy was repeated to assess post-treatment outcomes. In the second phase of the study, histologic responders (<15 eosinophils per high-power field [eos/hpf]) could enter a 52-week observation phase where tCS treatment was stopped and patients were monitored until symptom recurrence, at which time an additional upper endoscopy was performed.

Biosamples and immunohistochemistry

During the screening endoscopy, dedicated esophageal research biopsies were obtained from the distal, mid, and proximal esophagus for biobanking. For this study, an archived baseline (pretreatment) formalin-fixed, paraffin-embedded esophageal biopsy from the area of peak eosinophilic inflammation as determined in the primary study was selected and masked as to histologic responder status. This was then sectioned (5-micron thickness) and stained for immunohistochemistry (IHC) analysis of the three markers of interest as previously described.^21–24 Specifically, IHC was performed using a high volume fully-automated slide staining system (Bond Autostainer, Leica Microsystems, Inc., Norwell, MA) according to the manufacturer’s protocol. Slides were deparaffinized (Bond Dewax solution, AR9222) and hydrated (Bond Wash solution, AR9590). Heat-induced antigen retrieval was done for MBP and TRP (Bond Epitope Retrieval solution 1 pH –6.0, AR9961) as well as EOT3 (Bond Enzyme 1, 1:10, AR9551). After incubation (30 minutes) with the primary antibodies of interest, chromogenic antibody detection was performed (Bond Intense R Detection Kit, DS9263) supplemented with the Novocastra Post Primary and Novolink Polymer (Leica, #RE7260-K) for MBP and TRP, and with immPRESS anti-goat (Vector Labs, #AI-5000) for EOT3. Slides counterstained with hematoxylin, dehydrated, and coverslipped. The primary antibodies were: anti-MBP (mouse, #MCA5751, 1:1,000 dilution, AbD Serotec, Raleigh, NC); anti-eotaxin-3 (goat, #500-P156G, 1:50 dilution, PeproTech, Rocky Hill, NJ), and anti-mast cell tryptase (mouse, #M7052; 1:3,000 dilution, Dako, Carpinteria, CA). Stock samples of lung, skin, and esophageal tissue were used as controls, as per manufacturer recommendations. For each run, positive control slides were incubated with primary antibody while only antibody diluent was added to the negative control slides.

After staining was complete, the slides, still blinded to response status, were digitized using the Aperio ScanScope XT and a 20× objective (Leica Biosystems, Buffalo Grove, IL). Image analysis was conducted using the Aperio Image Scope (Aperio Technologies, Vista, CA), as previously described.^21–24 In brief, the maximum density of cells that met the preset threshold of signal intensity for staining positive for each antibody of interest in the esophageal epithelial layer was quantified (cells/mm²) in five microscopy fields using the Aperio Positive Pixel Count Algorithm (version 9.1, Aperio Technologies). Of note, the same settings to define a positive cell in the algorithm were used for all samples and all stains to minimize variability.

Outcomes and statistical analysis

The primary outcome assessed was histologic response to an 8-week course of topical steroids, defined as <15 eos/hpf (hpf area = 0.24 mm²).^25,26 As there were no statistical differences in outcomes between the two topical steroids tested (budesonide and fluticasone),¹⁹ we analyzed all subjects together regardless of treatment allocation. We also assessed alternative histologic response thresholds, including: (i) <5 eos/hpf, vs. ≥5 eos/hpf; (ii) < 1 eos/hpf vs. ≥1 eos/hpf; and (iii) <1 eos/hpf vs. <50% decrease in eosinophil count (a definition of ‘complete nonresponse’ that we have previously reported²⁷). For all of these, the eosinophil counts for the parent study were quantified using our previously validated protocol by the study pathologist.^28,29 An additional outcome was endoscopic response. In the parent study, all patients had endoscopic findings assessed with the validated EoE Endoscope Reference Score (EREFS), which quantified five key features: exudates (scored 0–2), rings (scored 0–2), edema (0–1), furrows (0–2), and stricture (0–1).³⁰ Based on the prior work, we defined an endoscopic response to be EREFS<2.³¹

The main analysis for this study was a prespecified component of the parent clinical trial. To test the hypothesis that decreased levels of EOT3 and TRP, but not MBP, would predict histologic nonresponse, we compared mean baseline staining for MBP, EOT3, and TRP between histologic nonresponders and responders using a two-sample t-test. We repeated this analysis for the three different histologic response thresholds and for endoscopic response, as well as by medication compliance versus noncompliance (defined as taking <70% of the study medication¹⁹). We then used multivariate logistic regression to calculate the area under the receiver operator characteristic curve (AUC) to evaluate the utility of the stains to predict response. We performed models with all three stains, even if there were not differences for individual stains, in order to take into account patterns of all patient data simultaneously. Because a sizeable proportion of patients entered the 52-week post-treatment study extension, we also assessed whether baseline staining levels predicted eventual symptom relapse in the off-treatment observation phase.²⁰

RESULTS

Patient characteristics

Baseline samples were evaluable in 110/111 subjects who completed treatment in the randomized phase of the parent trial. Overall, the mean age was 39.2 ± 15.2 years, 74 (67%) were male, 106 (96% were white), and 81 (74%) had at least one concomitant atopic condition. Though all cases were newly diagnosed with EoE, mean length of dysphagia symptoms prior to diagnosis was 10.8 ± 9.8 years. The baseline mean total EREFS score was 4.75 ± 1.87, with individual component scores of 0.92 for exudates, 1.29 for rings, 0.82 for edema, 1.23 for furrows, and 0.60 for stricture. The mean stricture diameter was 12.5 ± 2.8 mm, with 60 subjects (55%) requiring dilation (mean size achieved after initial dilation: 14.7 ± 2.3 mm). The baseline peak eosinophil count was 75.1 ± 54.5 eos/hpf. The overall esophageal levels of MBP, EOT3, and TRP were 485, 142, and 331 cells/mm², respectively, with no differences between the two steroid groups (Table 1). Illustrative staining examples are presented in Figure 1.

Table 1.

Overall immunohistochemistry levels, and levels stratified by initial topical steroid allocation in the parent trial

	All subjects (n = 110)	Budesonide (n = 55)	Fluticasone (n = 55)	P *
Major basic protein density (cells/mm2)	485 ± 524	471 ± 520	498 ± 533	0.81
Eotaxin-3 density (cells/mm2)	142 ± 405	148 ± 468	136 ± 335	0.89
Tryptase density (mast cells/mm2)	331 ± 203	350 ± 222	313 ± 183	0.35

^*means for the budesonide and fluticasone groups compared with a two-sample t-test.

Fig. 1.

Illustrative histopathologic images for (A) major basic protein (MBP), (B) eotaxin-3 (EOT3), and (C) mast cell tryptase (TRP) immunohistochemistry (IHC) staining.

Histologic response

There were 74 subjects (67%) who achieved histologic response with <15 eos/hpf. In this group, the post-treatment peak eosinophil count was 2.0 ± 3.2 eos/hpf compared with a count of 50.5 ± 38.8 in the nonresponders. Baseline MBP levels were higher in nonresponders (704 ± 678 vs. 373 ± 387 cells/mm²; P = 0.007), but EOT3 and TRP levels were not statistically different (Fig. 2A). There were 60 subjects (55%) who achieved histologic response with <5 eos/hpf, with a post-treatment peak eosinophil count of 0.7 ± 1.2 eos/hpf compared with 38.6 ± 38.1 in the nonresponders. With this threshold, baseline MBP levels were also higher in nonresponders (621 ± 627 vs. 379 ± 403 cells/mm²; P = 0.04), with no differences seen for EOT3 and TRP (Fig. 2B). Forty-one subjects (37%) achieved histologic response with <1 eos/hpf, and at this threshold TRP levels were higher in nonresponders than responders (370 ± 208 vs. 268 ± 178 cells/mm²; P = 0.01), with no differences in MBP or EOT3 (Fig. 2C). For histologic response defined as <1 eos/hpf compared with <50% decrease in eosinophil count for ‘complete nonresponse’, there were no differences in any baseline staining levels (data not shown). These results were unchanged after a sensitivity analysis accounting for study drug compliance (data not shown).

Fig. 2.

IHC levels of MBP, eotaxin-3 (EOT3), and mast cell tryptase (TRP) by histologic and endoscopic response status. (A) Histologic response defined as <15 eos/hpf. (B) Histologic response defined as <5 eos/hpf. (C) Histologic response defined as <1 eos/hpf. (D) Endoscopic response defined as an EREFS score < 2. In all graphs, black bars represent nonresponders and gray bars are responders. Mean levels were compared with a two-sample t-test.

Endoscopic response

A total of 43 subjects (39%) achieved endoscopic response of EREFS <2. Endoscopic nonresponders had higher baseline levels of MBP compared with responders (585 ± 590 vs. 327 ± 351 cells/mm²; P = 0.03), but there were no differences for EOT3 or TRP (Fig. 2D). We also assessed a combined histologic (<15 eos/hpf) and endoscopic (EREFS <2) response definition, which 40 subjects (36%) achieved. There were trends towards higher baseline levels of MBP (567 ± 586 vs. 341 ± 358 cells/mm²; P = 0.06) and EOT3 (194 ± 493 vs. 49 ± 101 cells/mm²; P = 0.08) in nonresponders compared with responders, but these were not significant and there was no trend for TRP (P = 0.23). There were also no differences in baseline staining levels for patients who did or did not require esophageal dilation (Table 2).

Table 2.

Staining levels stratified by requirement for esophageal dilation and by symptom recurrence before 1 year in the observation phase of the study

	No dilation (n = 50)	Dilation performed (n = 60)	P *
Major basic protein density (cells/mm2)	529 ± 513	449 ± 536	0.50
Eotaxin-3 density (cells/mm2)	156 ± 486	131 ± 330	0.76
Tryptase density (mast cells/mm2)	319 ± 208	341 ± 200	0.60
	Symptom recurrence (n = 32)	No symptom recurrence at 1 year (n = 25)
Major basic protein density (cells/mm2)	445 ± 476	408 ± 457	0.80
Eotaxin-3 density (cells/mm2)	92 ± 291	189 ± 633	0.42
Tryptase density (mast cells/mm2)	311 ± 190	357 ± 238	0.43

^*means compared with a two-sample t-test.

Diagnostic utility and prediction of symptom recurrence

The combination of MBP, EOT3, and TRP staining had an AUC of 0.66 to predict histologic response at the <15 eos/hpf threshold (Fig. 3A), an AUC of 0.62 for the <5 eos/hpf threshold, and an AUC of 0.67 at the <1 eos/hpf threshold. At this most restrictive histologic response definition, baseline TRP staining alone had an AUC of 0.65. For endoscopic response, the combination of all three stains had an AUC of 0.68 (Fig. 3B). Baseline staining did not predict eventual symptom recurrence after remission (Table 2).

Fig. 3.

Area under the receiver operator characteristic curves (AUC) assessing the utility of all three stains (MBP, EOT3, and TRP) to predict (A) histologic response at the <15 eos/hpf threshold and (B) endoscopic response defined as EREFS <2.

DISCUSSION

As a first-line pharmacologic option for treatment of EoE, tCS are commonly used. Our previous data suggested that staining for markers of inflammation, including MBP, EOT3, and TRP, could potentially predict steroid nonresponse.¹⁴ In that study, both TRP and EOT3 were significantly elevated in those with steroid response compared with nonresponders (244 vs. 157 mast cells/mm², P = 0.04, and 2,425 vs. 239 cells/mm², P = 0.02, respectively), but the number of patients assessed was modest (n = 40). In the current study, an analysis of a randomized trial of budesonide versus fluticasone for initial treatment of EoE,^19,20 we stained baseline samples in an attempt to validate our prior findings. While we hypothesized that levels of EOT3 and TRP, but not MBP, would predict histologic response, we did not confirm that result. Overall, we saw no consistent pattern, though baseline staining levels of MBP were higher in nonresponders (at the 15 eos/hpf response threshold) and baseline levels of TRP were lower in complete responders (at the 1 eos/hpf threshold). There was only mild utility for predicting either histologic response (AUC 0.66) or endoscopy response (AUC 0.68) with these stains. These are important, if somewhat frustrating findings, indicating these markers will not have the utility we had hoped for guiding medical therapy, and that other predictive methods are still needed.

There have been a number of attempts to identify predictors of response to topical steroids in EoE, though no metrics are currently clinically used. A study of a model of five clinical features (age, food allergies, baseline dilation, esophageal edema, and hiatal hernia) had some predictive utility, but this was limited to complete responders compared with patients defined as complete nonresponse (<50% decrease in eosinophil count).²⁷ Two prior studies suggested that patients requiring dilation at baseline were less likely to respond to topical steroids,^13,14 but this was not confirmed in the parent clinical trial for the present study.¹⁹ Related to this, patients with what has been termed an extreme narrow caliber esophagus were only a third as likely to respond to topical steroid treatment than EoE patients who did not have this phenotype, but <10% of EoE patients fit into this category, leaving the large majority without a response predictor.³² One clinical trial of fluticasone identified a set of genes that were differentially expressed at baseline between responders and nonresponders,³³ but this study was small and these findings were not replicated in a larger study.^27,34 Similarly, a less steroid-responsive genetic endotype of EoE has been recently described,³⁵ and an panel of epigenetic markers also showed promise,³⁶ but both of these remain to be validated for clinical use. A single nucleotide polymorphism (CT/TT vs. CC genotype) in the transforming growth factor-β gene has been associated with poor steroid response in children,³⁷ but this has also not reached clinical use. While the present study found that elevated MBP was associated with some histologic response thresholds, this was in contrast to our prior study,¹⁴ and the bulk of prior literature has not identified baseline eosinophil count to be associated with subsequent histologic response.¹² Therefore, while it is possible that some findings could be due to type II error (such as differences in MBP staining with threshold of <15 and < 5 eos/hpf but not with <1 eos/hpf), we do not believe that measurement alone of MBP would have clinical utility for response prediction.

The current treatment landscape for EoE is rapidly evolving, particularly for pharmacologic therapies.¹⁵ New esophageal specific steroid formulations are being tested,^9,11 with one now approved in Europe.¹⁰ Additionally, biologic agents, typically monoclonal antibodies targeting inflammatory mediators related to EoE pathogenesis, are also under investigation.^38,39 It would be ideal to understand who would be most likely to respond and benefit from steroid use, such that potential nonresponders could be directed to nonsteroid treatments, either dietary elimination or biologics. However, in the context of the previous literature and in light of our results, additional investigations are needed to discover and validate these predictors of response.

This study has a number of strengths. It was a prespecified analysis of a randomized, double-blind, double dummy clinical trial, in which samples were prospectively collected and banked with rigorous and standardized protocols. Subjects were newly diagnosed with EoE and thus receiving their first course of topical steroids, precluding any response bias from previous steroid treatments. All IHC was done in automated fashion, with samples and analysis blinded as to treatment response status. Furthermore, staining levels were quantified with an image analysis system, decreasing the subjectivity of these assessments. There are also some limitations to acknowledge. This was a single center study in adult subjects with a long symptom duration prior to diagnosis, so it is not known if the same results and staining patterns would be seen in children or patients with a short symptom duration. While patients were newly diagnosed with EoE, they were PPI nonresponders, as required by the EoE diagnostic guidelines available at the time of study design and conduct. Therefore, we do not know what the results would be in PPI responders or naïve subjects. While we used automated staining and standardized slide analysis techniques, these were optimized to focus on cell counts so assessments of extracellular areas of staining and specific quantification of areas of degranulation were not done. Finally, there is debate in the literature as to what constitutes as ‘treatment response’ in EoE.⁴⁰ However, to counter this, we performed sensitivity analysis of multiple histologic response thresholds.

In conclusion, analysis of baseline samples collected during the course of a comparative efficacy trial of topical steroids for initial treatment of EoE showed that pretreatment staining for MBP, EOT3, and TRP was not consistently or strongly associated with subsequent histologic or endoscopic response to tCS therapy, or to relapse of symptoms following cessation of therapy. While some trends were noted, the magnitude and predictive utility of these stains was modest and not likely to be clinically relevant as a test prior to inception of therapy in EoE. Because of the importance of finding markers of steroid response, future investigations should focus on novel methods of stratifying response to therapy, allowing personalization of therapeutic approached in EoE.

Funding: This study was supported by NIH R01 DK101856 and used resources from UNC Center for GI Biology and Disease (NIH P30 DK034987). The UNC Translational Pathology Lab, which is supported in part by grants from the NCI (2-P30-CA016086-40), NIEHS (2-P30ES010126-15A1), UCRF, and NCBT (2015-IDG-1007).

Disclosures: None of the authors have potential conflict of interest related to this paper. However, Dr. Dellon has received research funding from Adare, Allakos, GSK, Meritage, Miraca, Nutricia, Celgene/Receptos, Regeneron, and Shire/Takeda. Dr. Dellon has received consulting fees from Abbott, Adare, Allakos, Arena, AstraZeneca, Biorasi, Calypso, Celgene/Receptos, Eli Lilly, EsoCap, GSK, Gossamer Bio, Regeneron, Robarts, Shire/Takeda, and educational grants from Allakos, Banner, and Holoclara.

Specific author contributions: All authors approved the final draft submitted; guarantor is Evan S. Dellon. Evan S. Dellon: project conception, study design, data collection, data analysis/interpretation, manuscript drafting, critical revision, obtained funding. John T. Woosley: data collection, study pathologist, data interpretation, critical revision. Sarah J. McGee and Susan E. Moist: data collection and management, critical revision. Nicholas J. Shaheen: project conception, study design, data interpretation, critical revision.