Plasma P-selectin is inversely associated with lung function and corticosteroid responsiveness in asthma

Abstract

Background:

Severe asthma has multiple phenotypes for which biomarkers are still being defined. Plasma P-selectin reports endothelial and/or platelet activation.

Objective:

To determine if P-selectin is associated with features of asthma in a longitudinal study.

Methods:

Plasmas from 70 adult patients enrolled in the Severe Asthma Research Program (SARP) III at the University of Wisconsin-Madison were analyzed for concentration of P-selectin at several points over the course of three years, namely at baseline (BPS), after intramuscular triamcinolone acetonide (TA) injection, and at 36 months post-baseline. Thirty-four participants also came in during acute exacerbation and six weeks post-exacerbation.

Results:

BPS correlated inversely with forced expiratory volume in 1 s (FEV₁) and with residual volume/total lung capacity (RV/TLC), an indicator of air trapping. BPS was inversely associated with FEV₁ change after TA, by regression analysis. FEV₁ did not change significantly after TA if BPS was above the median, whereas patients with BPS below the median had significantly increased FEV₁ after TA. BPS was higher in and predicted assignment to SARP phenotype cluster 5 (“severe fixed-airflow asthma”). P-selectin was modestly but significantly increased at exacerbation but returned to baseline within three years.

Conclusions:

High BPS is associated with airway obstruction, air trapping, the “severe fixed-airflow” cluster, and lack of FEV₁ improvement in response to TA injection. P-selectin concentration, which is a stable trait with only modest elevation during exacerbation, may be a useful biomarker for a severe asthma pheno- or endotype characterized by low pulmonary function and lack of corticosteroid responsiveness.

Introduction

Severe asthma has multiple phenotypes for which biomarkers are still being defined (1–13). One potential biomarker is plasma P-selectin (14–21). P-selectin is a type I membrane protein that is sequestered in Weibel-Palade granules of endothelial cells and α-granules of platelets and is mobilized in response to various inflammatory and thrombogenic mediators, resulting in cell-surface display and subsequent proteolytic release into blood plasma (14–18, 22–26). It is found soluble in plasma in normal subjects, with reported median values ranging from about 10 to 40 ng/ml or more, depending on anticoagulant, plasma preparation protocol, and detection platform (14, 27–29). Plasma P-selectin is elevated under certain circumstances in asthma (14, 16–19). It has been reported to be increased up to three-fold in other diseases, including hemolytic uremic syndrome, thrombotic thrombocytopenic purpura, acute myocardial infarction, aspirin-exacerbated respiratory disease, aspirin-intolerant urticaria, chronic obstructive pulmonary disease, and in human immunodeficiency disease, and to be associated with features of interstitial lung disease (14, 20–22, 27, 30–32).

In the Severe Asthma Research Program (SARP), phase II, which was an observational study in which the cohort was classified in five asthma phenotypes using unsupervised hierarchical cluster analysis (33), plasma P-selectin analyzed in University of Wisconsin-Madison patients was higher in a pool of the three more severe phenotype clusters (16). Further, higher plasma P-selectin levels correlated with greater area of low density on chest computed tomography (CT) at total lung capacity (TLC), whereas the platelet-specific marker platelet factor 4 did not correlate with the CT signal, indicating that P-selectin in asthma likely mainly reports endothelial activation (16). Following whole-lung allergen challenge in (non-SARP) patients with mild allergic asthma, a human model of asthma exacerbation (34) known to cause platelet activation (17), there was a transient modest increase in P-selectin (16), but the variation in P-selectin among time-points was less than the variation among subjects (16). Overall, these findings indicate that asthma patients have variable levels of plasma P-selectin that correlate with disease features and are perturbed only transiently by exacerbation.

The relationships of plasma P-selectin with asthma clusters and airway structural change described above were determined at single points in time, and the whole-lung challenge study lacks the complexity of a natural exacerbation (15, 16). Associations between aspects of asthma and the stability of P-selectin levels over time, including exacerbations in asthma, have not been examined. We therefore undertook the present investigation to address these unknowns in Wisconsin patients participating in the longitudinal SARP III, which included an intramuscular triamcinolone acetonide (TA) injection and visits during and after exacerbations.

Materials and Methods

Patients

Patients were recruited at the University of Wisconsin-Madison to participate in the National Heart, Lung, and Blood Institute’s SARP III (35–37), which was a Health Insurance Portability and Accountability Act (HIPAA)-compliant study. Participants included adult patients with non-severe or severe asthma (Table 1), according to a modification of the European Respiratory Society/American Thoracic Society definition (35). The study subjects are a well-characterized population (35–39).

Table 1.

Patient characteristics at baseline

Variable	Value
n	70
Asthma severity; severe, n (%); non-severe, n (%)	46 (66), 24 (34)
Sex; female, n (%); male, n (%)	45 (64), 25 (36)
Age; years, median (quartiles)	48 (33, 58)
Race/ethnicity; white non-Hispanic, black, Asian American, Hispanic, n (%) for each group	60 (86), 9 (13), 1 (1), 3 (4)
Pre-BD FEV1, % predicted, mean ± SD	78 ± 20
median (quartiles)	78 (62, 92)
Pre-BD FEV1/FVC, % predicted, median (quartiles)	91 (84, 96)
Pre-BD FVC, % predicted, mean ± SD	86 ± 17
Pre-BD RV/TLC, % predicted, mean ± SD	116 ± 18
Post-BD FEV1, % predicted, mean ± SD	86 ± 20
median (quartiles)	84 (73, 102)
Post-BD change in FEV1, % predicted, median (quartiles)	8 (5, 11)
On ICS, n (%)	60 (86)
On OCS, n (%)	6 (9)
ICS dose, μg/day, mean ± SD	740 ± 420

Data are baseline data and presented as mean ± standard deviation (SD) (if data are normally distributed) or median (25^th, 75^th percentiles) (if data are not normally distributed). BD, bronchodilation; FEV₁, forced expiratory volume in 1 s; FVC, forced vital capacity; ICS, inhaled corticosteroid; OCS, oral corticosteroid; RV, residual volume; TLC, total lung capacity.

Study visits and assessments

Patients were characterized at baseline (35, 36). Patients received a single dose of 40 mg intramuscular tricamcinolone acetonide (TA) given deep in the gluteal region as described (36). The response was evaluated at a post-TA visit 18 ± 3 days after TA administration (36). Following the baseline and TA-response characterizations, some participants also came in during a subsequent exacerbation. The acute exacerbation visit was within five days of a subject reporting ≥ two days of increased symptoms compared to baseline and could occur any time after the subject had completed the baseline and post-TA visits and at least six weeks prior to the final visit (37), which was 36 months post-baseline (38). The recovery visit was six weeks after the acute exacerbation visit. Assessments were performed as before and described (15, 16, 36, 40, 41).

ELISA

Plasma was collected by a procedure that minimizes artifactual platelet activation as before (15, 16). Briefly, blood was collected in vacuum tubes containing CTAD (citrate, theophylline, adenosine, and dipyridamole) anticoagulant solution (BD Vacutainer Systems, Franklin Lakes, NJ, USA), the sample was double-spun, and the final supernatant was stored in aliquots at −80°C until tested (15). The concentration of soluble P-selectin was determined using sandwich enzyme-linked immunosorbent assay (ELISA) (R&D Systems, Minneapolis, MN, USA), according to the manufacturer’s instructions, as before (15), except that samples were diluted 1:20, 1:30, and 1:40. The values within the range of the standards were averaged (15). The absorbance of the colored product was measured at 450 nm, with wavelength correction at 620 nm, in a SpectraMax M5 plate reader (Molecular Devices, Sunnyvale, CA, USA). Each dilution was run in duplicate (15). The detection limit was 5 ng/ml (15). Values were not adjusted for the dilution by the anticoagulant solution present in the tube, for the reason given before (15). An aliquot of pooled plasma collected from eight subjects and handled the same as test plasma was analyzed alongside the test sample to ensure the stability of the ELISA over time. The concentration of P-selectin in the pooled sample measured on 26 occasions was 30.7 ± 1.7 ng/ml (mean ± standard error of the mean).

Statistical analysis

Paired t test or Wilcoxon matched-pairs signed-ranks test was used to compare data between visits, if the variable was normally distributed (passed Prism’s normality test) or was not normally distributed, respectively. Unpaired t test or Mann-Whitney U test was used to compare ordinal data between two groups, if the variable was or was not normally distributed. One-way analysis of variance (ANOVA) or Kruskal-Wallis test was used to compare ordinal data among groups, if the variable was or was not normally distributed. Fisher’s exact probability test was used in contingency table analysis to compare nominal data between two groups. Spearman rank test was used to analyze correlations. In some cases, Pearson test was used to analyze correlations between normally distributed or log-transformed variables. Unadjusted or adjusted linear regression models were used to analyze the effects of P-selectin as independent predictor variable on dependent outcome variables. A level of probability (p) ≤ 0.05 was considered significant. Analyses were performed using Prism (GraphPad, San Diego, CA, USA) or SAS (Cary, NC, USA). Group data are reported as mean ± standard deviation (SD) if the variable was normally distributed and as median with 25^th and 75^th percentiles if the variable was not normally distributed.

Results

Patients

ELISA for soluble plasma P-selectin was performed on samples from 70 University of Wisconsin-Madison SARP III patients, of whom 66% had severe asthma and 34% had non-severe asthma (Table 1). Other patient characteristics at baseline are also described in Table 1. Criteria for severe and non-severe asthma were previously described (35).

Baseline P-selectin (BPS) correlates with FEV₁ or RV/TLC

The coefficient of variation (CV) for plasma P-selectin concentration among subjects at baseline (BPS) was 59%, i.e., there was considerable patient variability (Fig. 1a, b). Two patients had a very high baseline P-selectin (BPS) of > 70 ng/ml that decreased in subsequent samples (Fig. 1c and not shown). BPS was not normally distributed (Fig. 1a), but log transformation resulted in a normal distribution (Fig. 1b). BPS did not correlate with age (Spearman rank correlation coefficient [r_s] = −0.00, p = 0.97). Also, BPS was not significantly different between the sexes (median [quartiles] in females = 21.8 ng/ml [14.5, 34.0], in males = 25.9 ng/ml [16.6, 34.5], p = 0.37). BPS correlated inversely with baseline FEV₁ (pre-bronchodilation [BD], as percentage of the predicted value) in all patients or in patients with severe asthma (Table 2), and was associated with baseline (pre-BD) FEV₁ by regression analysis, with somewhat lower p value after adjustment for age and severity (Table 3). Similarly, BPS also correlated inversely with baseline post-BD FEV₁ (r_s = −0.25, p = 0.04), including patients with severe asthma (r_s = −0.29, p = 0.05). Air trapping, like decreased FEV₁, is a feature of severe asthma (40). BPS correlated and was associated with percentage of predicted residual volume as fraction of total lung capacity (RV/TLC), an indicator of air trapping, regardless whether adjusting for age and severity (40) (Tables 2 and 3). However, BPS was no longer associated with RV/TLC after additional adjustment for FEV₁ (Table 3), presumably because FEV₁ and RV/TLC correlate with each other (r_s = −0.79, p < 0.001). Taken together, these data indicate that high BPS is associated with airway obstruction or decreased pulmonary function, as reported by FEV₁, and with air trapping, as reported by RV/TLC.

Fig. 1.

a Distribution of baseline P-selectin (BPS) in patients grouped by decades of concentration in ng/ml. b Distribution of BPS in patients grouped by log concentration. c Plasma P-selectin concentration over time in the 25 patients with an asthma exacerbation, from whom samples were available from all five visits. Symbols for visits: open circle, baseline; filled circle, post-triamcinolone acetonide (TA) administration; open square, acute exacerbation; filled square, recovery post-exacerbation; open triangle, final visit 36 months post-baseline. d Plasma P-selectin as percentage of baseline in the same patients as in c.

Table 2.

Correlations between BPS concentration and FEV₁ (% predicted) or RV/TLC (% predicted) at baseline

	FEV1 (% predicted)		RV/TLC (% predicted)
Group (n)	rs	p	rs	p
All (70)	−0.30	0.01	0.28	0.03
Severe asthma (46)	−0.34	0.02	0.25	0.12
Non-severe asthma (24)	−0.20	0.34	0.31	0.17

Log P-selectin correlated with FEV₁ or RV/TLC (which were normally distributed) with Pearson correlation coefficient (r) = −0.30, p = 0.01 or r = 0.30, p = 0.02 in all patients, respectively. FEV₁ = pre-bronchodilation FEV₁. BPS, baseline P-selectin; FEV₁, forced expiratory volume in 1 s; r_s, Spearman rank correlation coefficient; RV, residual volume; TLC, total lung capacity.

Table 3.

Linear regression models for the effect of BPS on FEV₁ (% predicted), RV/TLC (% predicted) at baseline, or FEV1 change (% predicted) after intramuscular triamcinolone acetonide (TA) administration

	FEV1 (% predicted)		RV/TLC (% predicted)		FEV1 (% predicted) change after TA
Model	β (95% CI)	p	β (95% CI)	p	β (95% CI)	p
Unadjusted	−6.1 (−10.4, −1.8)	0.007	6.0 (2.2, 9.9)	0.003	−1.4 (−3.5, 0.6)	0.17
Adj. for age and severity	−6.1 (−9.9, −2.2)	0.003	5.7 (1.9, 9.5)	0.004	−1.8 (−3.8, 0.2)	0.09
Adj. for age, severity, FEV1	-		1.2 (−1.5, 3.8)	0.40	−2.6 (−4.6, −0.4)	0.02

β estimate for the association with BPS showing the amount of difference in a lung function variable or in the FEV₁ change variable that would be expected for a 1 SD (15 ng/ml) increase in BPS, n = 62 (with complete RV/TLC data) for FEV₁ and RV/TLC, 55 for FEV₁ change. FEV₁ = pre-bronchodilation FEV₁. With n = 70, β was −6.9 (−11.3, −2.5), p = 0.002, for FEV₁. Adj., adjusted; BPS, baseline P-selectin; CI, confidence interval; FEV₁, forced expiratory volume in 1 s; r_s, Spearman rank correlation coefficient; RV, residual volume; TLC, total lung capacity.

BPS is inversely associated with FEV₁ change in response to TA

The baseline evaluation included an injection with TA, after which the participants returned for repeat assessments 18 ± 3 days later. After the TA administration, there was a trend to lower P-selectin compared to baseline (Table 4, Fig. 1c,d) (p = 0.12). BPS correlated with change in P-selectin after TA in the current study (r_s = −0.74, p < 0.001), i.e., the largest decreases in P-selectin in response to TA occurred in patients with highest BPS. In consistency with the network-wide SARP III population, in which the TA injection led to variable change among subjects in FEV₁ (36), variable changes in FEV₁ were observed in the present subjects. BPS was inversely associated with (pre-BD) FEV₁ change after TA by regression analysis, after adjustment for age, severity and baseline FEV₁ (Table 3). When dividing the patients according to median BPS, BPS-high patients did not have a significant increase in FEV₁ after TA compared to baseline, whereas BPS-low patients had significantly higher FEV₁ after TA (Table 5). Dividing patients by BPS quartiles, the 1^st quartile (with the lowest BPS) had significantly higher FEV₁ after TA, and the other quartiles had no significant change in FEV₁ after TA (Online Supplement Table 1). Further, there was a significant difference in FEV₁ among the quartiles at baseline (p = 0.006, ANOVA) and after TA (p = 0.002), with the 4^th quartile (with the highest BPS) having significantly lower FEV₁ than the other quartiles at both visits (Supplemental Table 1). BPS did not correlate significantly with post-BD change in FEV₁ (i.e., change from baseline pre-BD FEV₁ to baseline post-BD FEV₁, r_s = 0.01, p = 0.91). Overall, these data indicate that high BPS is associated with lack of response to injected steroid, whereas low BPS is associated with significant improvement in FEV₁ after injected steroid.

Table 4.

Plasma P-selectin concentration at baseline, post-intramuscular triamcinolone acetonide (TA) administration, during exacerbation and recovery, and 36 months post-baseline

	Visit
Variable	Baseline	Post-TA	Acute exacerbation	Recovery post-exacerbation	36 months post-baseline
P-selectin, ng/ml	23.2 (15.6, 34.4)	22.4 (16.9, 29.4)	26.4 (19.8, 33.0)	30.0 (21.3, 34.8)*††	23.2 (18.2, 29.4)
-“-, % of baseline	100	92 (66, 139)	117 (79, 152)*	127 (84, 174)*†#§	89 (75, 122)

Data are presented as median (25^th, 75^th percentiles), n = 70, 70, 26, 26, and 60 for the different visits, respectively.

^*p ≤ 0.05 versus the 36 months visit;

^††p ≤ 0.01 versus the acute exacerbation visit.

^†p ≤ 0.05 versus the acute exacerbation visit;

^#p ≤ 0.05 versus the baseline visit;

^§p ≤ 0.05 versus the post-TA visit.

Table 5.

FEV₁ (% predicted) at baseline and post-intramuscular triamcinolone acetonide (TA) administration in BPS-low (below median BPS) and BPS-high (above median) patients

	Visit
Group	Baseline	Post-TA	p
Below median/BPS-low (27)	84 ± 19	87 ± 20	0.002
Above median/BPS-high (27)	74 ± 22	75 ± 22*	0.43

Data are presented as mean ± SD. FEV₁ = pre-bronchodilation FEV₁. BPS, baseline P-selectin; FEV₁, forced expiratory volume in 1 s.

^*p ≤ 0.05 versus BPS-low (< 23.1 ng/ml).

BPS is not associated with type 2 immunity, is higher in asthma phenotype cluster 5, and predicts assignment to this cluster

Asthma can be analyzed and categorized in different ways, including severity, degree of type 2 immunity, and by cluster analysis. We therefore placed our subjects in such subcategories including investigating whether BPS is associated with type 2 immunity inflammation, which is present in many patients with asthma (1–3, 5–13). Fraction of exhaled nitric oxide (FENO) or sputum eosinophils are associated with corticosteroid responsiveness (36) and blood eosinophils are associated with reported exacerbations in the past year (37) in the general SARP III population. BPS did not correlate significantly with any of the type 2 immunity variables, blood and sputum eosinophils, FENO, or serum immunoglobulin (Ig) E (r_s = 0.00, p = 0.98 for blood eosinophil percentage; r_s = 0.02, p = 0.86 for blood eosinophil concentration; r_s = −0.11, p = 0.39 for sputum eosinophil percentage; r_s = 0.04, p = 0.75 for FENO; and r_s = 0.01, p = 0.92 for IgE). To characterize further the patients with high BPS, values for the type 2 immunity variables as well as for blood and sputum neutrophils in the BPS-high group (divided by median as above) compared to the BPS-low group are shown in Table 6. BPS-high patients were not significantly different from the BPS-low ones with respect to these variables (Table 6). Baseline FENO and sputum eosinophils correlated positively with (pre-BD) FEV₁ change after TA (r_s = 0.44 and 0.39, p = 0.001 and 0.006, respectively), consistent with associations in the general SARP III cohort (36). In the current population, blood eosinophil percentage and concentration, and IgE also correlated with FEV₁ change after TA (r_s = 0.30, 0.28, and 0.34, p = 0.03, 0.04, and 0.01, respectively). In the general SARP III cohort post-TA FEV₁ improvement was also associated with baseline BD response (36).

Table 6.

Characteristics of BPS-high and -low patients at baseline

Variable	BPS-high	BPS-low	p
Age, years, median (quartiles)	48 (32, 56)	48 (35, 62)	0.56
Sex: female, n (%); male, n (%)	20 (57), 15 (43)	25 (71), 10 (29)	0.32
Serum IgE, IU/ml, median (quartiles)	80 (30, 250)	120 (40, 200)	0.74
Blood eosinophils, %, median (quartiles)	4 (2, 5)	4 (2, 6)	0.68
Blood eosinophils, per μl, median (quartiles)	250 (130, 370)	220 (150, 400)	0.95
Blood neutrophils, %, median (quartiles)	56 (52, 62)	60 (47, 64)	0.91
Blood neutrophils, per μl, median (quartiles)	3700 (3000, 4900)	3800 (2600, 4400)	0.58
Sputum eosinophils, %, median (quartiles)	0.2 (0.0, 1.2)	0.9 (0.0, 3.9)	0.20
Sputum neutrophils, %, mean ± SD	61 ± 22	66 ± 21	0.40
FENO, ppb, median (quartiles)	22 (12, 36)	20 (12, 34)	0.77

Data are presented as mean ± standard deviation (SD) (if data are normally distributed) or median (25^th, 75^th percentiles) (if data are not normally distributed). BPS-high ≥ median, BPS-low < median. BPS, baseline P-selectin; FENO, fraction of exhaled nitric oxide; Ig, immunoglobulin.

To investigate whether there was any relationship between plasma P-selectin and the five asthma phenotype clusters identified in SARP II (33), we examined BPS levels in the current Wisconsin SARP III patients, who have been assigned at baseline to the five clusters using a 11-variable model (data provided by the SARP III Data Coordinating Center [DCC]). There was a significant difference in BPS concentration among the five clusters (p = 0.03, Kruskal-Wallis test), such that BPS was higher in cluster 5 (called “severe fixed-airflow asthma” (33)) than in the other clusters (Fig. 2a). Receiver operating characteristic (ROC) curve analysis demonstrated that BPS significantly predicted assignment to cluster 5 (Fig. 2b), indicating that BPS is a potential biomarker for cluster 5 assignment.

Fig. 2.

a Baseline P-selectin (BPS) in patients assigned to different asthma phenotype clusters. Using the five phenotype groups identified in SARP II (1), present SARP III subjects have been assigned at baseline to the five clusters using a 11-variable model (data provided by the SARP III DCC). Cluster 1 = “mild allergic asthma”, 2 = “mild-moderate allergic asthma”, 3 = “more severe older-onset asthma”, 4 = “severe variable allergic asthma”, 5 = “severe fixed-airflow asthma” (1); n = 70 (10 + 29 + 9 + 8 + 14). Bar = median, **p < 0.01 versus cluster 1 or 2, *p < 0.05 versus cluster 3 or 4. b ROC curve for the ability of BPS to predict assignment to cluster 5. Area under curve = 0.77 (p = 0.002); for the statistically optimal criterion of > 34.5 ng/ml, specificity = 86% and sensitivity = 57%.

P-selectin is modestly increased at a natural asthma exacerbation but otherwise is stable over time

Of the 70 subjects, 26 (37%) reported at least one exacerbation during the three years of the study and came in for an acute exacerbation visit and six weeks later for a recovery visit (Table 4, Fig. 1c,d). P-selectin was modestly but significantly increased at the acute exacerbation visit and remained significantly elevated at the recovery visit but not at the 36 months post-baseline visit. At the 36 months visit P-selectin was not significantly different from the baseline value, both in patients who experienced and were studied at an exacerbation or in the overall population (Table 3, Fig. 1c,d). These results demonstrate that a natural exacerbation was associated with a modest increase in P-selectin that persisted for at least six weeks but later returned to baseline level. For a given individual, the mean CV for P-selectin among the five visits within these subjects was 27% and among the three non-exacerbation visits 22%, i.e., the variation among visits in a subject was less than the variation among subjects at baseline.

Discussion

We found that high plasma soluble P-selectin concentration at baseline (BPS) was associated with low FEV₁ and with RV/TLC, an indicator of air trapping, in patients with asthma. These associations remained after adjustment for age and disease severity. Unlike FENO and sputum eosinophils, which are positively associated with clinical response, i.e., improvement in FEV₁, to corticosteroid (36), BPS was negatively associated with FEV₁ change after systemic corticosteroid injection. That is, high BPS was associated with lack of improvement in FEV₁ in response to systemic corticosteroid, whereas low BPS was associated with significant FEV₁ improvement in response to corticosteroid. Elevated serum chitinase-like protein YKL-40 has also recently been associated with lack of responsiveness to a four-week asthma treatment regimen of inhaled corticosteroid (ICS) plus long-acting β-agonist (42). Importantly, lack of corticosteroid responsiveness has recently been shown to be associated with and predict FEV₁ decline in a SARP III network-wide study (Denlinger LC et al., unpublished).

BPS was highest in patients assigned to phenotype cluster 5 (called “severe fixed-airflow asthma” (33)) and predicted assignment to this cluster. BPS was not associated with indicators of type 2 immunity. In addition, neutrophils were not significantly different in BPS-high patients compared to BPS-low ones.

P-selectin was modestly but significantly elevated above baseline during a natural exacerbation and for at least six weeks afterwards but later returned to baseline. The median increase was about 15–30% over the baseline value. As a comparison, the mean temporary rise in plasma P-selectin after whole-lung antigen challenge, which causes platelet activation (17), ranged from about 20% to 60% at 2–6 h in two studies (16, 17). This mean increase was likely higher, because it was at the defined same times after challenge in each patient compared to the more modest median increase recorded here within five days after a natural exacerbation, which presumably was registered at somewhat varying times after the actual time of exacerbation in different patients. Few other molecular blood biomarkers have been found to be altered during and after a natural asthma exacerbation. Plasma monocyte chemotactic protein-4 (chemokine [C-C] motif ligand 13 [CCL13]) level was higher in patients with an acute exacerbation than in those with stable asthma (43).

There are several limitations of the present study. First, we used an in-house assay of P-selectin, albeit with reagents, protocol, and standard provided by a leading commercial provider. The stated cut-off concentrations, therefore, should be considered specific for our assay. Second, there is uncertainty of how to apply the results to the totality of asthma patients. Because our subjects were more than 80% white non-Hispanic, to generalize our findings, a larger and more diverse cohort should be examined. Further, the ability to discern biomarkers that correlate significantly with features of asthma will depend upon the mix and diversity of asthma patients being studied. As an example, P-selectin correlated with area of low density on chest CT at TLC in 59 Wisconsin patients in SARP II (16). However, this correlation did not hold up in the current study of SARP III patients; BPS and low density area did not correlate (r_s = −0.02, p = 0.90) in the 62 patients on whom CT scans were done. The makeup of the Wisconsin SARP II and III cohorts was different. In the SARP II cohort 33% of the patients had severe asthma and 26% were in the pooled clusters No. 3–5 (16), whereas in the current SARP III cohort 66% of the patients were classified as severe (Table 1) and 44% were assigned to clusters 3–5 with 20% in cluster 5 alone (Fig. 2a).

In summary, we propose that P-selectin is a stable biomarker of a non-type 2 immunity asthma phenotype or endotype, characterized by airway obstruction/low FEV₁, air trapping, lack of corticosteroid responsiveness, and the “severe fixed-airflow” cluster. The reasons for persistently high P-selectin in this group of asthma patients are not known but likely include systemic activation of endothelial cells and platelets.