Underestimation of Luminal Eosinophilia by Quantitative Sputum Cytometry

Rationale: On Wright-stained sputum cytospins, eosinophil differential of >1.2% is considered abnormal and ≥ 3% clinically relevant. We hypothesized that failure to consider free eosinophil granules (FEG), and the re-emergence of eosinophilia (at ≥ 1.2 and ≥ 2.3%) underestimate the prevalence of the eosinophilic phenotype. Methods: This is a retrospective analysis of our Institutional Review Board-approved clinical sputum database. Of the 24,176 examinations, 17,693 had viable cell counts from 9570 patients (6604 on one occasion, 2967 with two or more) with various airway diseases. In all samples, FEGs were semi-quantied to estimate the prevalence of eosinophilia at <1.2 and <2.3%. In those patients with sputum examined on more than one occasion, subsequent results identied re-emergence of eosinophilia at ³1.2 and ³2.3%. Results: Of those with intact cell counts (n=15,278), 8562 (56.0%) and 9690 (63.4%) would have been classied as clinically non-relevant eosinophilia of <1.2% and <2.3% respectively. Among those with two or more intact cell counts, 1142 (38.5%) and 1083 (36.5%) of samples had re-emergence of eosinophilia (at 1.2 and 2.3%), either when a previous neutrophilia resolved (n= 218/19.1%, n=312/28.8%) or due to a reduction in the dose of corticosteroids (n=634/21.4%, n=666/22.4%) with no signicant differences between groups (p=0.018). There was a signicant difference between the proportions of eosinophilic versus non-eosinophilic at 1.2 and 2.3%, using the presence of FEG and the non-intact cell count (p<0.001). Conclusions: A total of 3180 (32.8%) samples identied as non-eosinophilic, had clinically relevant airway eosinophilia that would not have been identied if the phenotypic classication was limited to ≥ 2.3% intact eosinophils on one occasion. This underestimation is likely to be signicantly more if other means of airway eosinophilic activity were included and if mast cell activity and lymphocyte numbers, (not routinely quantied by sputum cytometry), could be contributing to the under-appreciation of airway T2 inammatory processes.


Introduction
Quantitative cytometry on Wrights (or Giemsa) stained sputum cytospin provides total cellularity and white cell differential count (1). The normal values have been established. For eosinophil differential %, the mean is 0.4%, and the upper limit of normality is 1.1 % (90th centile) or 2.2% (when 2 standard deviations are considered) (2). An eosinophil endotype of asthma is identi ed when the eosinophil % is greater than 2.2%, and a measurement > 3% is often considered clinically relevant (3). However, these cutoffs take into consideration only intact eosinophils, and not degranulated eosinophils. Free eosinophil granules (FEG) in tissue is an indicator of eosinophilic activity (4). In sputum, this correlates with eosinophil cationic proteins such as eosinophil peroxidase and the presence of moderate and many free granules correlate with clinical severity (5). Additionally, eosinophil % may change over time when any associated neutrophilia may resolve (6), with exacerbations (7), or when the dosage of corticosteroids are reduced (8).
We hypothesized that the eosinophilic endotype may be underestimated using the current practices. The objective of this study was to evaluate the prevalence of eosinophilia in sputum (using a cut-off of both 1.2% and 2.3%) in a large clinical database when eosinophil granules, multiple sputum samples, and the effects of corticosteroids are taken into consideration.

Methods
This is a retrospective study of those patients who attended the Firestone Institute for Respiratory Health for assessment of sputum cell counts between July 2004 to September 2020. The results were entered into FileMaker Pro Version 16 (Claris International, Cupertino, CA), a clinical database maintained for this purpose. The database contained demographics, physician diagnosis, reason for the test, whether the patient was stable or exacerbated (clinical diagnosis), and dose of inhaled and oral corticosteroids. In addition, baseline and post-induction spirometry were also recorded. The sputum differential cell count parameters included total cell count (x10 6 /g), and percentages of viable cells, squamous cell contamination, neutrophil, eosinophil, macrophage, lymphocyte, and bronchial epithelial cells. The presence of free eosinophil granules was enumerated by quanti cation of degranulated eosinophil clumps per eld of view under 400x magni cation. They were recorded as none, 1-2 as few, 2-3 as moderate and > 3 as many. Sputum induction and examination of cell counts were performed as described by Pizzichini et al (9).

Statistical analysis
Statistical analysis was performed using GraphPad Prism Version 8.3.1 (GraphPad Software, USA) and R 2020 (R Core Team, Austria). Continuous variables are presented as mean with standard deviation and categorical variables as percentage. Fisher's exact test was used to compare the proportions between eosinophilic and non-eosinophilic groups based on 1) 1.2 and 2.3% without consideration of FEG, 2) 1.2% with few/moderate/many FEG and 2.3% with moderate/many FEG and 3) addition of non-intact differential count based on speci cations in 2). In patients with two or more viable cell counts, detailed database mapping assessed for emergent eosinophilia at both thresholds (≥ 1.2% and presence of few/moderate/many FEG and ≥ 2.3% and presence of moderate/many FEG) on subsequent samples from the same patient. One-way ANOVA was used to compare the degree of responses for emergent eosinophilia based on resolution of previous neutrophilia (when total cell count ≥ 9.7x10 6 /g and neutrophil ≥ 64.4%), persistent/new neutrophilia, change in dose of corticosteroids (inhaled and/or oral) and disease are when there was no change in total cell count, neutrophil %, and corticosteroid dose.

Results
There were 24,176 individual records that include both spontaneously expectorated and induced sputum.
Of these, 6483 did not possess adequate sputum volume for analysis. The remaining 17,693 comprise 6604 patients with only cell count and 2967 had two or more cell counts for the remaining 11,088 results. 2415 had excessive cell degeneration and therefore a differential cell count could not be obtained, but 1188 of these exhibited FEGs. The demographic, clinical and sputum characteristics for patients that possessed adequate sputum for analysis is shown in Table 1. The relationship between proportions classi ed as eosinophilic and non-eosinophilic between cut-off thresholds of 1.2 and 2.3% when considering presence/absence of FEG including non-intact cell differential is shown in Table 2. Using the presence of FEG, the proportion between groups changes by 25.6% with eosinophil cut-off of 1.2% as compared to 11.6% when cut-off is 2.3% (p < 0.001). When a non-intact cell count is included, the proportion between groups with FEG changes to 19.6% for 1.2% threshold and 13.1% for 2.3% threshold (p < 0.001).

Discussion
This retrospective study using a large clinical database demonstrates that eosinophilia in sputum may be underestimated by as much as 28% if free eosinophil granules are not taken into consideration. Limiting examination to a single time point assessment that does not consider the changes in eosinophilia associated with exacerbations, resolution of concomitant neutrophilia or change in corticosteroid dosages may lead to further underestimation by as much as 60%. This has important clinical relevance, not only to endotype for selection of patients into clinical trials, but also to make therapeutic decisions about escalating or decrease steroid dosage and initiation of eosinophil-speci c biologic therapies.
It is important, for number of reasons, to recognize free eosinophil granules in sputum. Eosinophil cytolysis and release of cationic proteins are a marker of severity and contribute to bronchial epithelial injury and impairs repair (10). Eosinophil peroxidase contributes to the bromination of tyrosine residues (11) and is associated with epithelial dysfunction. Peroxidase activity also triggers an autoimmune reactive process in the airway (12) that has implications for disease severity and response to treatment of asthma with biologics (13). Further, cytolysis and release of extracellular traps may lead to crystal formation within the airways (14,15). Indeed, recent evidence suggests that Charcot Leyden Crystals, that are the products of auto-crystallization of galectin-10 is regulated independent of IL-5 (16), and the failure to recognize this in sputum may further underestimate the eosinophilic activity in the airways. This needs further investigation. A further underestimation of airway eosinophilic activity may result from failure to recognize these granular proteins within airway macrophages (17).
A second important reason for underestimating airway eosinophilia is when conclusions are drawn from a single time point assessment. The three most clinically relevant factors that might affect eosinophil % are whether sputum was sampled during an exacerbation, the dose of corticosteroids at the time of sampling, and if there was a concomitant neutrophilia that may mask an underlying eosinophilia. Certain airway infections may also directly induce an eosinophilic response that may resolve over time (18).
Particularly in patients with COPD, neutrophilic and eosinophilic exacerbations may interchange over time (19).
A limitation of our study is its retrospective nature that is associated with all the inherent biases of such a study design. However, the strengths include the large sample size, robust laboratory methods, and regular stringent external quality control. Another limitation is that we did not characterize patients into speci c disease categories such as asthma, COPD, overlap, bronchiectasis etc. This is because of the likely imprecision in the data coding based on a physician diagnosis. Given the large sample size, scrutiny of individual charts to con rm the physician diagnosis was not possible and we did not have approval from our Research Ethics Board for chart review. However, the purpose of this manuscript is not to relative eosinophilia to a particular disease state, rather to report fallacies in estimation of eosinophilia in sputum analysis. Although our analysis was limited to assessment of eosinophilic activity, there might be other aspect of cellular in ammation that are often not taken into consideration leading to inaccurate characterization as a non-T2 endotype. These include lymphocyte (20) and mast cell numbers (21) in sputum that can be identi ed particularly with more advanced microscopy, ow cytometry, and mass cytometry. They could also be markers of steroid responsiveness and it remains to be seen how often these endotypes may occur in the absence of eosinophilia.
In summary, we highlight the relevance of recognizing free eosinophil granules, and the importance of multiple examinations to identify sputum eosinophilia. This is likely relevant to select patients for anti-eosinophil clinical trials, to interpret treatment responses, and to guide the use of corticosteroids to treat eosinophil-responsive airway diseases.