Research Advances in Enhancing the Soluble Expression of Recombinant Heterologous Proteins

doi:10.13560/j.cnki.biotech.bull.1985.2025-1013

Abstract

Abstract:

Enhancing the soluble expression of recombinant heterologous proteins is a key challenge in bioengineering, pharmaceutical development, and industrial biotechnology, directly impacting protein functionality and large-scale production applications. This review systematically summarizes multi-level optimization strategies, from gene sequence to expression host, to improve the expression level, solubility, and stability of target proteins. At the codon level, strategies encompass host-preference-based codon adaptation, GC content adjustment, and the use of deep learning models to optimize mRNA structure and predict translation efficiency, thereby significantly increasing protein yield. At the amino acid level, rational design (e.g., hydrophobic core engineering, surface charge optimization), application of solubility-enhancing tags, and incorporation of non-standard amino acids can effectively improve protein folding, stability, and solubility. At the expression host level, the optimization of transcription and translation can be achieved through the selection and engineering of microbial strains (such as Escherichia coli, Bacillus subtilis, and yeast) and the precise design of various regulatory elements (e.g., promoters, ribosome binding sites, terminators) using artificial intelligence. The deep integration of artificial intelligence and biotechnology is driving the field of soluble protein expression from an empirical-driven approach towards a precise and predictable new “design-build-test” paradigm. This review aims to provide a comprehensive theoretical reference and technical outlook for the multidimensional and intelligent enhancement of soluble protein expression.

Key words: soluble protein expression, codons, amino acid sequence, expression elements, expression host

PANG Xin-li, ZHANG Hong-bing, LIU Xiao-qing, WANG Yuan, WU Ning-feng, TIAN Jian, GUAN Fei-fei. Research Advances in Enhancing the Soluble Expression of Recombinant Heterologous Proteins[J]. Biotechnology Bulletin, 2026, 42(4): 72-82.

Figures/Tables 4

References 53

[1]	Buck CB, Trus BL. The papillomavirus virion: a machine built to hide molecular Achilles’ heels [M]//Viral Molecular Machines. Boston, MA: Springer US, 2011: 403-422.
[2]	Corbett KS, Edwards DK, Leist SR, et al. SARS-CoV-2 mRNA vaccine design enabled by prototype pathogen preparedness [J]. Nature, 2020, 586(7830): 567-571.
[3]	Kanaya S, Yamada Y, Kinouchi M, et al. codon usage and tRNA genes in eukaryotes: correlation of codon usage diversity with translation efficiency and with CG-dinucleotide usage as assessed by multivariate analysis [J]. J Mol Evol, 2001, 53(4): 290-298.
[4]	Musto H, Romero H, Zavala A. Translational selection is operative for synonymous codon usage in Clostridium perfringens and Clostridium acetobutylicum [J]. Microbiology, 2003, 149(4): 855-863.
[5]	Duret L, Galtier N. Biased gene conversion and the evolution of mammalian genomic landscapes [J]. Annu Rev Genom Hum Genet, 2009, 10: 285-311.
[6]	Lajoie MJ, Rovner AJ, Goodman DB, et al. Genomically recoded organisms expand biological functions [J]. Science, 2013, 342(6156): 357-360.
[7]	Martin RW, Des Soye BJ, Kwon YC, et al. Cell-free protein synthesis from genomically recoded bacteria enables multisite incorporation of noncanonical amino acids [J]. Nat Commun, 2018, 9: 1203.
[8]	Ding ZD, Guan FF, Xu GS, et al. MPEPE, a predictive approach to improve protein expression in E. coli based on deep learning [J]. Comput Struct Biotechnol J, 2022, 20: 1142-1153.
[9]	Yan ZL, Chu WR, Sheng YH, et al. Integrating deep learning and synthetic biology: a co-design approach for enhancing gene expression via N-terminal coding sequences [J]. ACS Synth Biol, 2024, 13(9): 2960-2968.
[10]	Watson JL, Juergens D, Bennett NR, et al. De novo design of protein structure and function with RFdiffusion [J]. Nature, 2023, 620(7976): 1089-1100.
[11]	Fu HG, Liang YB, Zhong XQ, et al. Codon optimization with deep learning to enhance protein expression [J]. Sci Rep, 2020, 10: 17617.
[12]	Cabrita LD, Gilis D, Robertson AL, et al. Enhancing the stability and solubility of TEV protease using in silico design [J]. Protein Sci, 2007, 16(11): 2360-2367.
[13]	Jiang SM, Li CH, Zhang WW, et al. Directed evolution and structural analysis of N-carbamoyl-D-amino acid amidohydrolase provide insights into recombinant protein solubility in Escherichia coli [J]. Biochem J, 2007, 402(3): 429-437.
[14]	Jung S, Park S. Improving the expression yield of Candida antarctica lipase B in Escherichia coli by mutagenesis [J]. Biotechnol Lett, 2008, 30(4): 717-722.
[15]	Jonet MA, Mahadi NM, Murad AMA, et al. Optimization of a heterologous signal peptide by site-directed mutagenesis for improved secretion of recombinant proteins in Escherichia coli [J]. Microb Physiol, 2012, 22(1): 48-58.
[16]	Ito Y, Ishigami M, Hashiba N, et al. Avoiding entry into intracellular protein degradation pathways by signal mutations increases protein secretion in Pichia pastoris [J]. Microb Biotechnol, 2022, 15(9): 2364-2378.
[17]	Skoczinski P, Volkenborn K, Fulton A, et al. Contribution of single amino acid and codon substitutions to the production and secretion of a lipase by Bacillus subtilis [J]. Microb Cell Fact, 2017, 16(1): 160.
[18]	Liu TY, Zhang YY, Li YJ, et al. Effective gene expression prediction and optimization from protein sequences [J]. Adv Sci, 2025, 12(8): 2407664.
[19]	Weber M, Burgos R, Yus E, et al. Impact of C‐terminal amino acid composition on protein expression in bacteria [J]. Mol Syst Biol, 2020, 16(5): MSB199208.
[20]	Norrild RK, Johansson KE, O'Shea C, et al. Increasing protein stability by inferring substitution effects from high-throughput experiments [J]. Cell Rep Meth, 2022, 2(11): 100333.
[21]	Sumida KH, Núñez-Franco R, Kalvet I, et al. Improving protein expression, stability, and function with ProteinMPNN [J]. J Am Chem Soc, 2024, 146(3): 2054-2061.
[22]	Studier FW. Use of bacteriophage T7 lysozyme to improve an inducible T7 expression system [J]. J Mol Biol, 1991, 219(1): 37-44.
[23]	İncir İ, Kaplan Ö. Escherichia coli as a versatile cell factory: Advances and challenges in recombinant protein production [J]. Protein Expr Purif, 2024, 219: 106463.
[24]	Makino T, Skretas G, Georgiou G. Strain engineering for improved expression of recombinant proteins in bacteria [J]. Microb Cell Fact, 2011, 10(1): 32.
[25]	Tegel H, Tourle S, Ottosson J, et al. Increased levels of recombinant human proteins with the Escherichia coli strain Rosetta(DE3) [J]. Protein Expr Purif, 2010, 69(2): 159-167.
[26]	Phillips TA, VanBogelen RA, Neidhardt FC. Lon gene product of Escherichia coli is a heat-shock protein [J]. J Bacteriol, 1984, 159(1): 283-287.
[27]	van Wijk DJS. Tuning Escherichia coli for membrane protein overexpression [J]. Proc Natl Acad Sci U S A, 2008, 105(38): 14371-14376.
[28]	Schlegel S, Rujas E, Ytterberg AJ, et al. Optimizing heterologous protein production in the periplasm of E. coli by regulating gene expression levels [J]. Microb Cell Fact, 2013, 12(1): 24.
[29]	Turner P, Holst O, Karlsson EN. Optimized expression of soluble cyclomaltodextrinase of thermophilic origin in Escherichia coli by using a soluble fusion-tag and by tuning of inducer concentration [J]. Protein Expr Purif, 2005, 39(1): 54-60.
[30]	Singh A, Upadhyay V, Upadhyay AK, et al. Protein recovery from inclusion bodies of Escherichia coli using mild solubilization process [J]. Microb Cell Fact, 2015, 14(1): 41.
[31]	Gu Y, Xu XH, Wu YK, et al. Advances and prospects of Bacillus subtilis cellular factories: From rational design to industrial applications [J]. Metab Eng, 2018, 50: 109-121.
[32]	Chen YZ, Li MM, Yan MC, et al. Bacillus subtilis: current and future modification strategies as a protein secreting factory [J]. World J Microbiol Biotechnol, 2024, 40(6): 195.
[33]	Rojas Contreras JA, Pedraza-Reyes M, Ordoñez LG, et al. Replicative and integrative plasmids for production of human interferon gamma in Bacillus subtilis [J]. Plasmid, 2010, 64(3): 170-176.
[34]	De S, Van D. Control of prokaryotic translational initiation by mRNA secondary structure [J]. Prog Nucleic Acid Res Mol Biol, 1990, 38: 1-35.
[35]	Looser V, Bruhlmann B, Bumbak F, et al. Cultivation strategies to enhance productivity of Pichia pastoris: a review [J]. Biotechnol Adv, 2015, 33(6): 1177-1193.
[36]	Mayson BE, Kilburn DG, Zamost BL, et al. Effects of methanol concentration on expression levels of recombinant protein in fed-batch cultures of Pichia methanolica [J]. Biotechnol Bioeng, 2003, 81(3): 291-298.
[37]	Çelik E, Çalık P, Oliver SG. Metabolic flux analysis for recombinant protein production by Pichia pastoris using dual carbon sources: Effects of methanol feeding rate [J]. Biotechnol Bioeng, 2010, 105(2): 317-329.
[38]	Arias CAD, de Araujo Viana Marques D, Malpiedi LP, et al. Cultivation of Pichia pastoris carrying the scFv anti LDL (-) antibody fragment. Effect of preculture carbon source [J]. Braz J Microbiol, 2017, 48(3): 419-426.
[39]	Azadi S, Mahboubi A, Naghdi N, et al. Evaluation of sorbitol-methanol co-feeding strategy on production of recombinant human growth hormone in Pichia pastoris [J]. Iran J Pharm Res, 2017, 16(4): 1555-1564.
[40]	Jahic M, Wallberg F, Bollok M, et al. Temperature limited fed-batch technique for control of proteolysis in Pichia pastoris bioreactor cultures [J]. Microb Cell Fact, 2003, 2(1): 6.
[41]	Tang HT, Wang SH, Wang JJ, et al. N-hypermannose glycosylation disruption enhances recombinant protein production by regulating secretory pathway and cell wall integrity in Saccharomyces cerevisiae [J]. Sci Rep, 2016, 6: 25654.
[42]	贺铁凡, 徐沙, 张阁元, 等. 重构酿酒酵母N-糖基化途径生产人源化糖蛋白 [J]. 微生物学报, 2014, 54(5): 509-516.
	He TF, Xu S, Zhang GY, et al. Reconstruction of N-glycosylation pathway for producing human glycoproteins in Saccharomyces cerevisiae [J]. Acta Microbiol Sin, 2014, 54(5): 509-516.
[43]	Liu DJ, Garrigues S, de Vries RP. Heterologous protein production in filamentous fungi [J]. Appl Microbiol Biotechnol, 2023, 107(16): 5019-5033.
[44]	Ha TK, Kim D, Kim CL, et al. Factors affecting the quality of therapeutic proteins in recombinant Chinese Hamster ovary cell culture [J]. Biotechnol Adv, 2022, 54: 107831.
[45]	Fu YS, Han ZM, Cheng WT, et al. Improvement strategies for transient gene expression in mammalian cells [J]. Appl Microbiol Biotechnol, 2024, 108(1): 480.
[46]	Kotopka BJ, Smolke CD. Model-driven generation of artificial yeast promoters [J]. Nat Commun, 2020, 11: 2113.
[47]	Bonde MT, Pedersen M, Klausen MS, et al. Predictable tuning of protein expression in bacteria [J]. Nat Meth, 2016, 13(3): 233-236.
[48]	Chu YY, Yu D, Li YP, et al. A 5' UTR language model for decoding untranslated regions of mRNA and function predictions [J]. Nat Mach Intell, 2024, 6(4): 449-460.
[49]	Zrimec J, Fu XZ, Muhammad AS, et al. Controlling gene expression with deep generative design of regulatory DNA [J]. Nat Commun, 2022, 13: 5099.
[50]	Munsamy G, Lindner S, Lorenz P, et al. ZymCTRL: a conditional language model for the controllable generation of artificial enzymes [J]. NeurIPS, 2022.
[51]	Yang KK, Wu Z, Arnold FH. Machine-learning-guided directed evolution for protein engineering [J]. Nat Meth, 2019, 16(8): 687-694.
[52]	Quax TEF, Claassens NJ, Söll D, et al. Codon bias as a means to fine-tune gene expression [J]. Mol Cell, 2015, 59(2): 149-161.
[53]	Rocklin GJ, Chidyausiku TM, Goreshnik I, et al. Global analysis of protein folding using massively parallel design, synthesis, and testing [J]. Science, 2017, 357(6347): 168-175.

宿主名称 Host name	优点 Advantages	缺点 Disadvantages	适用蛋白类型 Applicable protein types	常用优化策略举例 Example of common optimization strategies	参考文献References
大肠杆菌 E. coli	生长快、成本低、遗传背景清晰、表达量高	缺乏复杂翻译后修饰、易形成包涵体	非糖基化蛋白、酶、抗体片段	使用工程化菌株（BL21，Rosetta）、优化诱导条件（Tunner，Lemo21）、使用融合标签	［22-28］
枯草芽胞杆菌 B. subtilis	非致病、分泌能力强、无内毒素	胞外蛋白酶多、遗传稳定性有时不足	工业酶（淀粉酶、蛋白酶）、抗原	蛋白酶缺陷株、信号肽优化、强启动子	［30］
毕赤酵母 P. pastoris	高密度发酵、分泌能力强、具备真核修饰能力	糖基化模式与人类不同、甲醇利用有安全隐患	分泌型蛋白、工业酶、部分糖蛋白	AOX1启动子调控、温度控制、糖基化通路工程（ΔOCH1）、共底物喂养	［31-35］
酿酒酵母 S.cerevisiae	安全、遗传操作工具成熟	表达量相对较低、糖基化过度	疫苗抗原、食品酶、基础研究	糖基化通路改造（ΔOCH1，ΔALG3）、强启动子（GAL1等）	［36-38］
哺乳动物细胞（如CHO， HEK293）	能完成复杂翻译后修饰、产物最接近天然	成本高、周期长、培养复杂	治疗性抗体、复杂糖蛋白、病毒样颗粒	细胞系工程（GS/KO系统）、培养基优化、过程控制	［40-41］
丝状真菌（如木霉、曲霉）	强大的蛋白分泌能力、能进行某些复杂修饰	遗传操作相对复杂、背景分泌蛋白多	工业水解酶（纤维素酶、淀粉酶）	强启动子、蛋白酶缺陷株、分泌信号优化	［39］

宿主名称 Host name	优点 Advantages	缺点 Disadvantages	适用蛋白类型 Applicable protein types	常用优化策略举例 Example of common optimization strategies	参考文献References
大肠杆菌 E. coli	生长快、成本低、遗传背景清晰、表达量高	缺乏复杂翻译后修饰、易形成包涵体	非糖基化蛋白、酶、抗体片段	使用工程化菌株（BL21，Rosetta）、优化诱导条件（Tunner，Lemo21）、使用融合标签	［22-28］
枯草芽胞杆菌 B. subtilis	非致病、分泌能力强、无内毒素	胞外蛋白酶多、遗传稳定性有时不足	工业酶（淀粉酶、蛋白酶）、抗原	蛋白酶缺陷株、信号肽优化、强启动子	［30］
毕赤酵母 P. pastoris	高密度发酵、分泌能力强、具备真核修饰能力	糖基化模式与人类不同、甲醇利用有安全隐患	分泌型蛋白、工业酶、部分糖蛋白	AOX1启动子调控、温度控制、糖基化通路工程（ΔOCH1）、共底物喂养	［31-35］
酿酒酵母 S.cerevisiae	安全、遗传操作工具成熟	表达量相对较低、糖基化过度	疫苗抗原、食品酶、基础研究	糖基化通路改造（ΔOCH1，ΔALG3）、强启动子（GAL1等）	［36-38］
哺乳动物细胞（如CHO， HEK293）	能完成复杂翻译后修饰、产物最接近天然	成本高、周期长、培养复杂	治疗性抗体、复杂糖蛋白、病毒样颗粒	细胞系工程（GS/KO系统）、培养基优化、过程控制	［40-41］
丝状真菌（如木霉、曲霉）	强大的蛋白分泌能力、能进行某些复杂修饰	遗传操作相对复杂、背景分泌蛋白多	工业水解酶（纤维素酶、淀粉酶）	强启动子、蛋白酶缺陷株、分泌信号优化	［39］

名称 Model name	类型 Type	应用层面/目标Application level/Objective	核心优势功能/原理 Key feature/Mechanism	优势 Advantages	局限性 Limitations	参考文献References
CNN启动子模型	预测	宿主/转录调控	使用卷积神经网络分析启动子序列，预测其强度	实现对启动子活性的高精度（R²=0.79-0.84）预测和精细调控	依赖大量标注数据	［46］
EMOPEC	预测	宿主/翻译起始	使用随机森林算法，定量揭示SD序列与蛋白表达水平的关系	可理性设计RBS以提升异源蛋白表达量	对真核系统适用性有限	［47］
UTR-LM	预测	宿主/翻译效率	基于Transformer架构，通过5'UTR序列预测平均核糖体负载量（MRL）	仅凭序列即可准确预测，实验验证可将表达量提升32.5%	模型可解释性弱	［48］
MPEPE	预测	密码子/协同效应	深度学习预测高表达倾向的氨基酸替换，再经保守性分析筛选位点	综合考虑氨基酸替换和密码子使用，揭示协同效应	优化流程相对复杂	［8］
MLD-NCS	预测	密码子/翻译起始	结合LSTM与注意力机制，优化mRNA 5'端前30个密码子	有效避免翻译停滞，在枯草芽胞杆菌中表达量提升5.41倍	主要优化翻译起始阶段	［9］
MPB-EXP/MUT	预测	氨基酸/可溶性设计	基于蛋白质语言模型，从序列学习高可溶性特征并设计突变	能有效改善多种蛋白的可溶性，实现从几乎不表达到可溶的跃迁	生成序列的可靠性需验证	［18］
C-terminal composition 分析	预测	氨基酸/表达丰度	统计分析C末端氨基酸组成（带电荷残基）与蛋白表达水平的关联	规则简单明确，易于应用（如在C端引入Lys/Arg）	普适性有待考察	［19］
GMMA	预测	氨基酸/突变组合	从多突变体数据中推断单点突变效应，并组合有利突变	能够发现突变间的叠加效应，实现表达量的数十至数百倍提升	依赖于大量的实验数据	［20］
ZymCTRL	生成	宿主/全序列设计	条件语言模型，根据酶学性质定向生成酶序列及配套调控元件	实现“按需定制”的全序列人工设计，整合性强	长序列功能验证挑战大	［50］
DeepCodon	生成	密码子/翻译优化	深度学习模型优化密码子使用频率，提高翻译效率	在多种宿主中均实现2-10倍的表达量提升，通用性强	对mRNA高级结构考虑不足	［11］
RFdiffusion	生成	氨基酸/结构设计	通过迭代去噪过程，从随机噪声中生成具有目标结构的蛋白质骨架	能够从头设计复杂蛋白结构（如对称寡聚体），突破性强	计算资源需求极高	［10］
ProteinMPNN	生成	氨基酸/结构设计	基于给定的蛋白质骨架结构，逆折叠生成兼容的氨基酸序列	序列回收率高，设计的序列可溶性和表达量显著提升（最高20倍）	严重依赖输入结构的准确性	［21］

名称 Model name	类型 Type	应用层面/目标Application level/Objective	核心优势功能/原理 Key feature/Mechanism	优势 Advantages	局限性 Limitations	参考文献References
CNN启动子模型	预测	宿主/转录调控	使用卷积神经网络分析启动子序列，预测其强度	实现对启动子活性的高精度（R²=0.79-0.84）预测和精细调控	依赖大量标注数据	［46］
EMOPEC	预测	宿主/翻译起始	使用随机森林算法，定量揭示SD序列与蛋白表达水平的关系	可理性设计RBS以提升异源蛋白表达量	对真核系统适用性有限	［47］
UTR-LM	预测	宿主/翻译效率	基于Transformer架构，通过5'UTR序列预测平均核糖体负载量（MRL）	仅凭序列即可准确预测，实验验证可将表达量提升32.5%	模型可解释性弱	［48］
MPEPE	预测	密码子/协同效应	深度学习预测高表达倾向的氨基酸替换，再经保守性分析筛选位点	综合考虑氨基酸替换和密码子使用，揭示协同效应	优化流程相对复杂	［8］
MLD-NCS	预测	密码子/翻译起始	结合LSTM与注意力机制，优化mRNA 5'端前30个密码子	有效避免翻译停滞，在枯草芽胞杆菌中表达量提升5.41倍	主要优化翻译起始阶段	［9］
MPB-EXP/MUT	预测	氨基酸/可溶性设计	基于蛋白质语言模型，从序列学习高可溶性特征并设计突变	能有效改善多种蛋白的可溶性，实现从几乎不表达到可溶的跃迁	生成序列的可靠性需验证	［18］
C-terminal composition 分析	预测	氨基酸/表达丰度	统计分析C末端氨基酸组成（带电荷残基）与蛋白表达水平的关联	规则简单明确，易于应用（如在C端引入Lys/Arg）	普适性有待考察	［19］
GMMA	预测	氨基酸/突变组合	从多突变体数据中推断单点突变效应，并组合有利突变	能够发现突变间的叠加效应，实现表达量的数十至数百倍提升	依赖于大量的实验数据	［20］
ZymCTRL	生成	宿主/全序列设计	条件语言模型，根据酶学性质定向生成酶序列及配套调控元件	实现“按需定制”的全序列人工设计，整合性强	长序列功能验证挑战大	［50］
DeepCodon	生成	密码子/翻译优化	深度学习模型优化密码子使用频率，提高翻译效率	在多种宿主中均实现2-10倍的表达量提升，通用性强	对mRNA高级结构考虑不足	［11］
RFdiffusion	生成	氨基酸/结构设计	通过迭代去噪过程，从随机噪声中生成具有目标结构的蛋白质骨架	能够从头设计复杂蛋白结构（如对称寡聚体），突破性强	计算资源需求极高	［10］
ProteinMPNN	生成	氨基酸/结构设计	基于给定的蛋白质骨架结构，逆折叠生成兼容的氨基酸序列	序列回收率高，设计的序列可溶性和表达量显著提升（最高20倍）	严重依赖输入结构的准确性	［21］