Steroidal compounds, characterized by a cyclopentanoperhydrophenanthrene backbone, are natural products whose types and positions of functional groups on the molecule directly determine their physicochemical properties and biological activities[1], [2], [3]. Owing to their significant physiological roles in the body, these compounds are widely utilized in various therapeutic areas, including anticancer, anti-inflammatory, antiviral, and antibacterial treatments[1], [4], [5], [6], [7]. According to market research data, steroid hormone drugs occupy a substantial share of the global pharmaceutical market, with worldwide sales surpassing 100 billion USD in 2017—making them the second-largest drug category after antibiotics[8], [9]. This growth is largely attributed to the extensive use of steroid drugs in the treatment of chronic diseases, as well as the increasing demand from emerging markets.

In steroid drug synthesis, androst-1,4-diene-3,17-dione (ADD) serves as a crucial intermediate, typically synthesized from androst-4-ene-3,17-dione (AD) through a 1,2-dehydrogenation reaction[10], [11]. This reaction plays a vital role in the pharmacological activity of many steroid drugs. For example, when acetyl hydrocortisone (HA) is converted into prednisone acetate (PA) via this reaction, its anti-inflammatory activity increases three- to four-fold[12], [13], [14]. The enzyme 3-ketosteroid-Δ1-dehydrogenase (KstD), which catalyzes this reaction, is therefore considered a key catalyst in steroid drug synthesis[15], [16], [17]. However, natural enzymes exhibit significant limitations in practical applications[18]. KstD enzymes from various microbial sources differ markedly in catalytic activity and substrate specificity, and their overall catalytic efficiency is generally low. Additionally, natural enzymes often display poor substrate specificity and are susceptible to environmental factors such as temperature and pH fluctuations[17], [19], [20]. These limitations have severely restricted the industrial application of KstD enzymes.

Enzyme engineering techniques to offer effective solutions to overcome these limitations[21]. Among these, directed evolution generates enzyme variants through random mutations and high-throughput screening—a method that can yield a wide variety of variants, though its success depends on the establishment of an efficient screening system, and the inherent randomness may produce some ineffective or unstable mutants [22], [23], [24]. In contrast, rational design enables precise selection of mutation sites but relies heavily on structural information and computational power, making prediction and design challenging in complex enzyme systems[25], [26]. Semi-rational design, however, combines the advantages of molecular modeling with experimental screening by first predicting potential mutation sites through simulations and then optimizing them experimentally. This approach has significantly improved catalytic efficiency, substrate specificity, and stability. For example, Yujuan Shen et al. optimized sucrose phosphorylase (BlSPase) through semi-rational design and obtained the optimal mutant L341V/V346P, achieving a yield of 358.6 g/L AA-2G and a 75.7 % conversion rate of L-AA, providing important technical support for the efficient synthesis of AA-2G[27], [28], [29], [30]. These results highlight the substantial potential of semi-rational design in optimizing enzyme performance and advancing their industrial applications.

In recent years, multiple research teams have successfully modified KstD enzymes with demonstrated efficacy. Using site-saturation mutagenesis, Mao et al. optimized the catalytic performance of KstD₃ derived from Arthrobacter simplex, with the W299A mutation increasing the AD conversion rate by approximately 2-fold[28]. Zhang's team developed a focused iterative saturation mutagenesis (FSISM) technique that enhanced the specific activity of KstD toward hydrocortisone by 10-fold through blocking non-catalytic cavities adjacent to the substrate tunnel[31]. Through site-saturation mutagenesis, Wang et al. engineered KstD from Propionibacterium sp. (PrKstD), where the H135T/A356N double mutant exhibited a 16.7-fold improvement in catalytic efficiency for 6α-methyl-11β,17α-dihydroxy-4-pregnene-3,20-dione, achieving a space-time yield of 4.08 g·L⁻¹ ·h⁻¹ at high substrate concentration (60 g/L)[32].

Based on the aforementioned background, this study employed a semi-rational design strategy to optimize 3-ketosteroid-Δ1-dehydrogenase. In previous studies, the 3-ketosteroid-Δ1-dehydrogenase KstD2 from Mycobacterium neoaurum DSM 1381 was modified using error-prone PCR to obtain a mutant strain, KstD2ep, which exhibited significantly enhanced catalytic activity, achieving an AD conversion of 40 g/L. In this study, we first constructed and validated a homology model of KstD2ep, followed by molecular docking to identify key substrate-binding residues. We then applied alanine scanning and single-site saturation mutagenesis to probe functional hotspots, yielding a highly active triple mutant KstD2ep (V332E/L334T/G534V). Structural remodeling of the binding pocket and enhanced electrostatic complementarity were further confirmed through electrostatic surface analysis and molecular dynamics simulations. The mutant's improved stability and catalytic efficiency were verified under various pH and temperature conditions. Finally, the biocatalytic performance was evaluated and optimized at both whole-cell and fermenter scales, demonstrating the industrial applicability of the engineered enzyme.