Allergic diseases affect approximately 20 % of the population worldwide [1] and are recognized by the World Health Organization as a major health problem in the world [2]. The prevalence of allergic diseases has increased dramatically in recent years due to the changes in people’s lifestyles and the aggravation of environmental pollution [3]. Allergic diseases are chronic non-infectious inflammatory reactions mainly mediated by immunoglobulin-E (IgE) after exposure to allergens in atopic individuals, which can involve multiple systems and organs [4].

Human allergy sensitized to cat allergen is common worldwide. In China, about 16.4 % of patients with allergic diseases are allergic to cat dander, especially among adolescents aged 12–18 years [5]. Currently, clinical diagnosis of cat-allergic patients is mainly based on cat dander extracts [6]. However, the composition of cat dander extracts is complex and ill-defined. These cat dander extracts may contain other cat proteins from dander or fur as well as non-allergic proteins, and may also be contaminated with other allergens (e.g. dust mites) [7]. Among all cat allergens, Felis domesticus allergen 1 (Fel d 1) is a major one [8] and 60 %–90 % of cat-allergic patients were sensitized to it [8], [9]. Furthermore, a recent study showed that Fel d 1 was at least as sensitive for in vitro diagnostics of cat allergy as current extract-based tests and the increased Fel d 1-specific IgE levels may be a potential risk factor for allergic asthma in children [10]. Therefore, Fel d 1 is a good candidate to replace cat dander extracts for molecular diagnosis of cat allergy. Accordingly, the preparation of high-quality Fel d 1 is particularly important for allergy diagnostic methods and for the standardization of desensitizing vaccines for cat allergy.

Fel d 1 is a 35 kDa tetrameric protein formed by two heterodimers, each of which in turn consists of two independent genetically encoded polypeptide chains (polypeptide chain 1 and polypeptide chain 2) [11], [12]. The structure of chain 1 and chain 2 as well as the disulfide bonds are now well known, but the preparation of the two chains of Fel d 1 separately often results in inactive or low active products during prokaryotic expression [13]. The previous study reported the direct fusion strategy in connecting the two chains of Fel d 1, but it was purified from inclusion bodies and required an additional refolding step [8]. Although another study obtained soluble direct-fused Fel d 1 by co-expression with an auxiliary folding factor, but no serum IgE binding study was performed to verify the reactivity [14]. Moreover, previous study showed that unproperly folded allergen would greatly affect the IgE-reactivity [15], it is unclear whether the direct connection of two Fel d 1 chains would affect the local conformation or folding. Considering the correctly folded allergens are important for diagnostic purposes [16], it is necessary to investigate the connection mode about the two polypeptide chains of Fel d 1.

Artificial Intelligence (AI) technology has been developing rapidly in recent years [17], [18], which greatly facilitates protein design based on computed structure models and deep learning. Empowered by AI technology, the success rate and rationality of protein design have greatly improved [19]. For example, the machine learning-aided engineering of Poly(ethylene-terephthalate) hydrolases exhibited superior hydrolytic activity relative to the wild-type [20], the deep-learning based de novo design of luciferases with much higher substrate specificity [21], generation of artificial proteins with similar catalytic efficiencies as its natural counterpart [22]. Protein design by AI avoids relatively random mutation and provides a guiding design blueprint based on the biophysical and biochemical principles of proteins, bringing great convenience and potential to protein research [23], [24], [25]. However, the AI-engineering strategy has not been applied to design allergens with improved activity till date.

In this study, we used the AI-derived AlphaFold2 and quality assessment program to improve the rational folding of the cat major allergen Fel d 1 by optimizing the fusion strategy of the two polypeptide chains. We then expressed these fusion constructs and the direct fusion Fel d 1 as described in the previous study [8]. We also performed the Circular Dichroism (CD), High Performance Liquid Chromatography-Size Exclusion Chromatography (HPLC-SEC) and reducing/non-reducing SDS-PAGE comparison of the optimized fusion Fel d 1 and the previously well-defined direct fusion Fel d 1 to confirm their folding status and heterodimer form. The improvement of IgE-reactivity of the optimized Fel d 1 was investigated. It lays the foundation for the highly efficient and large-scale production of Fel d 1, which promotes the development of detection methods for cat allergen components and quality assessment of molecular desensitization vaccines.