【作者】 刘元花；

【导师】 袁景淇；

【作者基本信息】 上海交通大学 ， 控制理论与控制工程， 2008， 博士

【摘要】 现代基因工程药物主要来自于重组微生物和动物细胞培养,如单克隆抗体、细胞生长因子和生长素等。动物细胞培养的突出优势在于在很多情况下能分泌出具有正确折叠和后修饰的、具有生命活性的蛋白质。然而,采用动物细胞培养获取的生物药品产率相当低。主要的原因有:1)培养条件的专一性;2)在培养条件的微小扰动下动物细胞的产率变化和细胞死亡的敏感性增加。欲实现培养过程的仿真、优化和控制,最终达到提高产率的目的,模型化是先导。本文以杂交骨髓瘤细胞培养为例,根据对动物细胞代谢机理和细胞周期机理的现有认知,研究了在不同应用层面上的动物细胞模型化方法,给出了模型的具体数学描述,并进行了初步验证。详细描述如下:首先,本文研究了杂交骨髓瘤细胞培养的改进的集总参数动力学模型(Macrokinetic Model with Lumped Parameters)。传统的集总参数动力学模型,尤其是Monod模型,形式简单并取得了广泛应用。然而,众所周知,Monod模型不能描述接种或补料操作后的细胞生长迟滞现象。为此,本文根据细胞代谢酶系调节机理,提出了模仿细胞内酶调节的代谢调节模型,该代谢调节模型与Monod模型相结合,可以更好的描述细胞比生长速率。另外,鉴于细胞培养过程中氨基酸的限制性影响,本文建立了重要氨基酸的代谢模型。通过这两方面的改进,使得集总参数动力学模型的实用性得到了增强,如模型能够描述氨基酸耗尽后细胞维持短期生长的现象。利用上述改进模型对杂交骨髓瘤细胞间歇培养和脉冲式补料间歇培养进行仿真,较好描述了活细胞和死细胞的密度以及葡萄糖、谷氨酰胺、赖氨酸和乳酸的浓度的动态变化。论文提供了仿真结果与实验值的对比。论文建立的第二个模型是基于细胞周期机理的杂交骨髓瘤细胞培养的群体平衡模型(Population Balance Model)。真实的细胞系统往往是一个异质系统,细胞群体中的个体表现出不同的细胞行为或细胞性状。获得细胞群体关于这些细胞行为或细胞性状的分布信息,对于与细胞周期相关的重组蛋白(如单抗)的产率优化至关重要。文献中平衡模型的优势正是在于描述细胞系统的各类分布信息。然而,目前动物细胞培养的平衡模型研究还有很大的提升空间,如细胞周期特性信息的充实和模型的验证。为此,本文根据细胞周期和周期调控机理的生物学信息,把细胞体积和DNA含量作为细胞周期时相以及时相内细胞与细胞之间的区分标志,进而建立了杂交骨髓瘤细胞群体平衡模型。这个模型能够直接仿真DNA和体积分布的动态变化。论文利用DNA分布实验数据验证了上述细胞群体平衡模型。实验验证是文献中现有平衡模型未能做到的事。细胞群体平衡模型还能用来计算细胞周期各时相的细胞分率(各时相细胞数目占总细胞数量的百分比)的动态变化,以及活细胞和死细胞密度、底物(葡萄糖和氨基酸)浓度以及副产物(乳酸)的浓度。对此,论文也给出了实验验证结果。与集总参数动力学模型相比,平衡模型能提供更多的用于细胞培养过程优化和控制的信息,如细胞周期各时相的细胞分率和DNA分布的动态变化,可以为细胞群体周期动力学的控制提供理论依据。同时,DNA分布的模型仿真有助于更好的设计细胞周期控制过程。然而,由于模型本身数学描述的复杂性,实现平衡模型仿真的计算负担远比集总参数动力学模型大得多,在实际应用中有一定的难度。论文建立的第三个模型是基于酶系调控的杂交骨髓瘤细胞培养控制论模型(Cybernetic Model)。控制论模型是从动物细胞内部代谢和代谢酶调节的机理分析出发,研究细胞生长、消耗底物和生成副产物的情况。控制论模型的前提假设为生物系统经过长期的进化,已经形成了一套自我优化的策略。基于这个前提,细胞可以通过对酶水平和酶活性的控制来调节细胞代谢网络生化反应的进行。在控制论代谢网络中,酶系根据最优化原则竞争利用胞内物质资源。根据文献中代谢机理的分析,谷氨酰胺的利用过程中存在转氨和脱氨作用的竞争,这种竞争直接影响到副产物氨和丙氨酸的生成,以往的控制论模型没有考虑到这一竞争关系。另外,以往的控制论模型往往忽略了氨基酸的(谷氨酰胺除外)竞争性利用。本文提出的控制论模型重点考虑了这两方面因素的影响。该模型不但能描述细胞生长、底物(葡萄糖和谷氨酰胺)代谢和副产物(乳酸、氨和丙氨酸)形成等宏观变量的动态变化,还能仿真细胞内部物质(代谢中间产物和酶)的动态变化,为细胞培养提供了胞内水平的控制变量。控制论模型考虑细胞代谢酶调节机理,因而能描述细胞生长对营养扰动的延迟响应。从这一点来看,控制论模型相比其它类型的代谢机理模型更具优势。以上三种模型化方法各有优势,适用于不同的场合:集总参数动力学模型结构简单,易于实现,适于环境扰动不太明显、不确定因素较少的培养过程;群体平衡模型侧重于提供细胞群体关于细胞周期特性的动力学信息,适用于重组蛋白对细胞周期特性比较敏感的细胞培养过程;控制论模型的细胞代谢和代谢调节机理性强,模型参数具有明确的物理意义,适用于环境或其它扰动比较明显、不定因素较多的培养过程。除此之外,本文还初步探讨了模型在细胞培养过程中补料策略优化和控制中的应用,给出了仿真研究结果。更多还原

【Abstract】 Nowadays, the recombinant microbial and mammalian cells are used to produce genetic engineering pharmaceuticals, such as monoclonal antibodies, growth factors and hormones. Mammalian cells are characterized by the ability to produce high-value biopharmaceutical products that require post-translation modification in order to become biologically active. These high-value products are produced in relatively small quantities due to the highly specialized culture conditions and their susceptibility to the reduced productivity or cell death as a result of slight deviations in the culture conditions. Models of cell processes are particularly useful for simulating, optimizing and controlling the cell culture processes, and eventually contribute to efforts to increase productivity. In this PhD thesis, three models with different purposes are proposed and validated with myeloma cell cultivations. These three models will be demonstrated as follows:Firstly, we developed an improved macrokinetic model with lumped parameters for myeloma cell cultivations.The traditional macrokinetic models with lumped parameters, especially the Monod model, are simple and therefore, are easy to be developed. They are widely used for applications. However, Monod-type kinetics is not able to describe the lag phase after inoculation or after pulse feeding. In order to deal with the this growth lag phase, a model based on the metabolic regulation mechanisms is developed. This metabolic regulation model together with the Monod-type model is better to describe the specific growth rate. In addition, the amino acid limitations are considered in modeling to account for the short-term period of cell growth after amino acid depletion.The improved macrokinetic model is validated with batch culture and fed-batch culture with pulse feedings. It shows that the model is able to simulate the concentrations of glucose, glutamine, lysine and lactate, as well as the densities of total, viable and dead cell with a suitable accuracy.Secondly, we developed a population balance model based on cell cycle mechanisms for myeloma cell cultivation. Mammalian cell cultures comprise heterogeneous cells that differ according to their size and intracellular levels of DNA, protein and other cell properties. Cell population distributions with respect to different cell properties are important for the control of the cell-cycle specific products. Population balance model is the most rigorous approach for describing the dynamics of the distributions of different cell properties. However, futher studies are still needed to improve this model by considering the cell-cycle specific properties. Moreover, few population balance models developed in the literature have been validated by experiments. Therefore, a population balance model here is developed based on cell-cycle mechanisms for myeloma cell cultivations. In this model, both cell volume and DNA content are used to differentiate individual cells in the cell population. Thus, the model can be used to simulate the distributions of cell volume and DNA content. The model simulations of DNA distribution are validated with the DNA measurements from flowcytometer with a reasonable accuracy. The dynamics can also be described by this model in terms of the cell fraction of each phase, concentrations of glucose, glutamine, lysine, ammonia, lactate and alanine, as well as the cell densities of total, viable and dead cells. These are also validated by experimental data.Compared with the macrokinetic model with lumped parameters, population balance model can provide more information, such as the dynamics of DNA distribution and phase fractions, for the control of the cell cycle dynamics. Furthermore, the DNA simulation results are helpful for the design of cell-cycle controlling process. Nevertheless, there is a challenge that population balance models may easily suffer from computational intensity.Thirdly, we developed a cybernetic model based on the regulation of enzyme systems for myeloma cell cultivations.Cybernetic models are structured on the basis of the regulation of the metabolic network and the cybernetic principles. Metabolic network is regulated by enzyme synthesis and enzyme activity that are represented by the cybernetic variabes‘u’and‘v’, respectively. Enzyme systems are hypothesized to compete for optimizing resource utilization, and different pathways can be up and down regulated depending on the outcome of these competitions. The competition between deamination pathway and transamination pathway to use glutamine to produceα-KG is considered in this study. This is because these two pathways lead to different level of glutamine utilization and different quantity of byproducts (i.e. ammonia and alanine). In addition, the competition between lysine utilization for protein formation and energy supply is also involved due to lysine limitation observed in our experiments. In addition to the densities of viable and dead cells, concentrations of substrates, the byproducts, for example ammonia, lactate and alanine are simulated by this model. Model simulations also provide the levels of enzyme and other intracellular species, which are useful for the control of the mammalian cell cultures. Cybernetic models are powerful for capturing the dynamic response to environmental changes because they take into account the metabolic regulation of the network.It can be concluded that these three models will be selected for different applications according to their features. The macrokinetic models with lumped parameters have simple mathematical formulations. They are the most suitable alternative if the cell culture processes with slight perturbation and uncertainty are in question. Population balance models are dominant in the applications with the control of cell-cycle specific products involved. Contrary to the macrokinetic model with lumped parameters, cybernetic models show advantages when the cell culture processes with strong perturbation and uncertainty are considered.Finally, the models are used for further study of the simulation, optimization and control of the bioprocesses.更多还原