The importance of type 2 immunity in the pathogenesis of asthma was established in the 1990s, based on analyses of model mice.56 Thereafter, the research focus shifted to which signature cytokine in type 2 immunity—interleukin (IL)-4, IL-5, or IL-13—was important or to identifying the role of each of these cytokines in the pathogenesis of asthma. As it turns out, IL-13 plays a central role; it was shown that IL-13 alone was sufficient to cause asthma-like phenotypes in mice, whereas the blockage of IL-13 signals alone was sufficient to inhibit asthma-like phenotypes in ovalbumin-induced asthma model mice.78 Particularly in the pathogenesis of asthma, actions of IL-13 on airway epithelial cells have been shown to be important for inducing goblet cells and for enhancing airway hyper-responsiveness.910

To elucidate the effects of IL-13 on human airway epithelial cells, we comprehensively identified IL-13-inducible genes using the DNA microarray method.11 Consequently, we found that periostin is one of the highly expressed genes. IL-4, another cytokine sharing receptors and signal transduction pathways with IL-13, has the same ability to induce periostin.

We then investigated the expression of periostin in asthmatic patients using immunohistochemical analyses.1213 We found that periostin is deposited on the thickened basement membrane in asthmatic patients (Fig. 1). The localization of periostin overlapping with that of other ECM proteins composing thickened basement membrane—collagens I, III, and V, and tenascin-C—suggests that periostin contributes to generating subepithelial fibrosis in bronchial asthma by binding to other ECM proteins. Deposited periostin could be observed in the subepithelial areas of model mice in an IL-4- or IL-13-dependent manner.12 Woodruff et al.14 then confirmed that periostin is a gene highly expressed in the bronchial tissues of asthmatic patients. They showed that periostin expression by IL-13 is sensitive to corticosteroids and that expression of periostin is down-regulated with corticosteroid treatment in asthmatic patients.

The pathological role of periostin in asthma still remains controversial; several studies using periostin-deficient mice showed that periostin plays a protective role in airway allergic inflammation,1516 whereas another study with periostin-deficient mice and neutralizing antibodies against periostin showed that periostin accelerates it.17 The reason for this discrepancy is unclear. In contrast, Kanemitsu et al.18 followed up asthmatic patients for more than 20 years. They examined periostin expression in biopsy samples for more than 20 years ago and evaluated the change in FEV1. Consequently, it turned out that the more periostin was deposited in the lungs, the more pulmonary function decreased. Putting these findings together with our reports showing that periostin activates NF-κB by itself in keratinocytes and also activates NF-κB together with other inflammatory cytokines, such as TNFα or IL-1α in fibroblasts,1920 we assume that periostin exacerbates airway allergic inflammation.

It has been widely accepted that asthma is not a single disease, but rather a "syndrome."21 This concept is important for treating asthmatic patients, particularly for using molecularly targeted drugs. To support this concept, it is important to elucidate each "endotype" instead of each "phenotype" in heterogeneous subgroups comprising asthma.21 Although many trials have been performed to cluster asthmatic patients, the classification of asthmatic patients into "Th2-high" and "Th2-low" is important because it is potentially related to the choice of type 2 antagonists.21

Many molecularly targeted drugs against bronchial asthma are now under development. Approximately half of them are type 2 antagonists (Table 1). They target IgE, cytokines, chemokines, or prostaglandin D₂ receptors, all of which are involved in type 2 immunity. If we are to prescribe these agents for asthmatic patients, we have to select patients in which type 2 immunity is dominant in their pathogenesis and for whom we can expect that type 2 antagonists would show efficacy. Establishment of such "stratified" medicines is necessary in using molecularly targeted drugs, both to increase their efficacy and to decrease costs.222324 Although it was initially reported that anti-IL-5 antibodies were not effective overall for asthmatic patients,2526 they later showed good efficacies for patients with high eosinophils.2728 Several agents targeting IL-4 or IL-13 did not show sufficient efficacy either, so their development was stopped,293031 which might have been due in part to not stratifying the patients.

Woodruff et al.32 stratified asthmatic patients into "Th2-high" and "Th2-low" based on the expression of IL-13 and IL-5. They then searched for signature molecules of "Th2-high" asthma, finding that 3 gene products—periostin, chloride channel regulator 1, and serpin peptidase inhibitor, clade B, member 2—correspond to these molecules, respectively.32

Based on this knowledge, Genentech33 applied serum periostin as a surrogate marker for Th2-high asthma and conducted a phase IIb study of anti-IL-13 antibodies (lebrikizumab) for steroid-resistant asthmatic patients. In this trial, lebrikizumab showed overall good efficacy in improving lung function for the patients. When the patients were divided into the high and low periostin groups based on serum periostin levels, lebrikizumab showed significant efficacy for the high periostin group, whereas it had no efficacy for the low periostin group. This study is a milestone in the field of asthma in that it showed for the first time that for asthma, periostin can be a target for a companion diagnostic, defined as one useful for predicting the efficacy of drugs following diagnosis.

Recently, it has been reported that clustering asthmatic patients into the high and low periostin groups is useful for predicting the efficacy of anti-IgE antibodies (omalizumab) as well.34 It is noteworthy that the target molecules of anti–IL-13 antibodies and anti-IgE antibodies are different; however, both IL-13 and IgE are type 2 immunity-related molecules, so serum periostin predicts the efficacy of both agents. Thus, periostin would be a surrogate biomarker for not only IL-13, but also type 2 immunity. Based on these findings, an algorithm for the treatment of asthma can be proposed (Fig. 2). The first line of anti-asthma drugs is inhaled corticosteroids (ICSs). Although ICSs are effective for most patients, 5%-10% are resistant or hyporesponsive to them.3536 Measurement of serum periostin is recommended for these patients. If some patients show high periostin levels, type 2 antagonists, such as lebrikizumab or omalizumab, should be added. If some show low periostin levels, since type 2 antagonists would be ineffective, other agents or therapies is recommended. However, the story is not so simple. Other biomarkers—eosinophils, fraction of nitric oxide (FeNO), and dipeptidyl peptidase-4 (DPP4)—are also used as surrogate biomarkers for type 2 immunity and as possible companion diagnostics for type 2 antagonists (Table 2). It remains to be addressed what these biomarkers have in common and how they differ, as well as which biomarker or combination is optimal for predicting the efficacy of each type 2 antagonist. Moreover, periostin is not a simple surrogate biomarker for type 2 immunity, as shown in the next section.

As mentioned earlier, asthma is a heterogeneous disease, and many trials have been reported for clustering asthmatic patients.3738 Serum periostin levels are diverse among asthmatic patients; Kanemitsu et al.18 reported that 37.9% of 224 asthmatic patients receiving ICS treatment showed higher periostin levels than the normal range (<95 ng/mL), whereas the rest remained within the normal range. Although periostin has appeared as a surrogate biomarker for type 2 immunity, it has not been simple to characterize it. Several studies have attempted to clarify the characteristics of periostin-high asthmatic patients.

Tissue remodeling of bronchial tissues, including fibrosis, is a typical histological characteristic of asthma. On the other hand, although ICSs are powerful and effective drugs for asthmatic patients and are used as a first-line drug for asthma, 5%-10% of asthmatic patients are resistant or hyporesponsive to them.35 It is assumed that hyporesponsiveness to ICSs is caused by many underlying mechanisms, such as tissue remodeling.49 Given that periostin is a component of fibrosis, a feature of tissue remodeling, it is reasonable to suppose that serum periostin could be a biomarker to predict hyporesponsiveness to ICSs.

Kanemitsu et al.18 evaluated hyporesponsiveness to ICSs by decline of FEV1 (ΔFEV1) in the course of treatment with ICSs; Kanemitsu et al. reported that serum periostin is correlated with the decline FEV1 in overall asthmatic patients. When the patients were divided into the rapid decliners (ΔFEV1 ≤-30 mL/year) and non-rapid decliners (ΔFEV1 >-30 mL/year), the rapid decliners showed a higher periostin level than the non-rapid decliners (104.6 vs 89.2 ng/mL), suggesting that periostin is a surrogate biomarker for ICS hyporesponsiveness. However, the difference in the averages of periostin levels between the rapid and non-rapid decliners was not significant, which may be attributed to the heterogeneous mechanisms of hyporesponsiveness to ICSs.

Nagasaki et al.50 explored the possibility that if asthmatic patients are clustered into several groups, some groups may show better correlation between serum periostin and hyporesponsiveness to ICSs than the overall groups. They applied blood eosinophils and neutrophils for clustering patients, finding that the patients could be subdivided into 4 clusters: Cluster 1 (low eosinophils and low neutrophils), late-onset and non-atopic; Cluster 2 (moderate eosinophils and low neutrophils), early-onset and atopic; Cluster 3 (high eosinophils and low neutrophils), late-onset and eosinophil-dominance; and Cluster 4 (moderate eosinophils and high neutrophils), poor control and high IL-6. The patients in Clusters 1 and 2 were good responders to ICSs irrespective of serum periostin, whereas the patients in Cluster 4 were poor responders, irrespective of serum periostin. It is noteworthy that the difference in ΔFEV1 between the high periostin (>95 ng/mL) and low periostin (≤95 ng/mL) groups is significant in Cluster 3 (-23.0 vs -1.42 mL/year). As mentioned earlier, Cluster 3 would greatly overlap with the subgroup manifesting active airway inflammation, fixed airflow limitation, male sex, and longer duration, as reported by Haldar et al.37 and Moore et al.38 These results suggest the ability of serum periostin to predict the hyporesponsiveness to ICSs, particularly in Cluster 3 patients, namely the patients with adult-onset and eosinophil-dominant asthma.

Kato et al.51 examined the ability of serum periostin to predict the hyporesponsiveness to ICSs from a different point of view. They enrolled 25 asthmatic patients well controlled by ICSs. They observed the patients for 12 weeks after they tapered off ICSs and divided them into the stable (n=20) and unstable (n=5) groups, based on the occurrence of acute exacerbation. When they compared serum periostin levels in the patients before tapering ICSs, the unstable group showed a higher periostin level than the stable group (141.9 vs 91.5 ng/mL). These results suggest that high periostin levels entail the risk of acute exacerbation by tapering ICSs. Although examined subjects and endpoints are different among the above studies, serum periostin appears to be a good biomarker in patients who cannot maintain lung functions with ICSs, and it can also be a good biomarker in patients who will not be able to taper off ICS treatment when they are stable.

In contrast to adult asthma, the usefulness of periostin in childhood asthma is still under discussion. Song et al.52 demonstrated high periostin levels in patients with childhood asthma; however, the difference between the patients and control subjects was subtle (76.0 vs 71.0 ng/mL). They also showed the correlation of serum periostin with other type 2 immunity biomarkers, blood eosinophils, and FeNO, as in adult asthma. In contrast, Konradsen et al.53 did not detect correlations of serum periostin with blood eosinophils or FeNO in childhood asthma patients. Moreover, Inoue et al.54 did not find any difference in serum periostin levels between patients with childhood asthma and control subjects. One reason for the inconsistent results with serum periostin levels in patients with childhood asthma could be high baseline levels of serum periostin in children. Inoue et al.54 showed that serum periostin levels in elementary school-age children were higher than adults (6-11 years; mean: 125.0 ng/mL) and become higher as the children grew older (Fig. 3). High serum periostin levels in the teenagers dropped after puberty. This was probably because serum periostin in teens is mostly derived from bones and thereafter drops because bone growth stops. Another reason could be that there is less tissue remodeling in childhood asthma than in adult asthma. As mentioned earlier, serum periostin has characteristics as a surrogate biomarker for tissue remodeling in addition to type 2 immunity. Although the pathogenesis of childhood asthma is mostly type 2 immunity, type 2 immunity may not be sufficient to enhance serum periostin as in adult asthma. The finding that adult patients with early-onset and atopic type of asthma do not show high periostin levels as those with late-onset type asthma may support this idea.50 Moreover, it is noteworthy that we should be careful in evaluating serum periostin levels in children because normal ranges vary among different age groups.