Developing a Coarse-Grained Model for Bacterial Cell Walls: Evaluating Mechanical Properties and Free Energy Barriers

The bacterial cell envelope of Gram-negative
bacteria is a complex
biological barrier with multiple layers consisting of the inner membrane,
periplasm of peptidoglycan, and the outer membrane with lipopolysaccharides
(LPS). With rising antimicrobial resistance there is increasing interest
in understanding interactions of small molecules with the cell membrane
to aid in the development of novel drug molecules. Hence suitable
representations of the bacterial membrane are required to carry out
meaningful molecular dynamics simulations. Given the complexity of
the cell envelope, fully atomistic descriptions of the cell membrane
with explicit solvent are computationally prohibitive, allowing limited
sampling with small system sizes. However, coarse-grained (CG) models
such as MARTINI allow one to study phenomena at physiologically relevant
length and time scales. Although MARTINI models for lipids and the
LPS are available in literature, a suitable CG model of peptidoglycan
is lacking. Using an all-atom model described by Gumbart et al. [PLoS Comput. Biol. 2014, 10, e1003475], we develop a CG model of the peptidoglycan network within
the MARTINI framework. The model is parametrized to reproduce the
end-to-end distance of glycan strands. The structural properties such
as the equilibrium angle between adjacent peptides along the strands,
area per disaccharide, and cavity size distributions agree well with
the atomistic simulation results. Mechanical properties such as the
area compressibility and the bending modulus are accurately reproduced.
While developing novel antibiotics it is important to assess barrier
properties of the peptidogylcan network. We evaluate and compare the
free energy of insertion for a thymol molecule using umbrella sampling
on both the MARTINI and all-atom peptidoglycan models. The insertion
free energy was found to be less than kBT for both the MARTINI and all-atom models. Additional
restraint free simulations reveal rapid translocation of thymol across
peptidogylcan. We expect that the proposed MARTINI model for peptidoglycan
will be useful in understanding phenomena associated with bacterial
cell walls at larger length and time scales, thereby overcoming the
current limitations of all-atom models.

革兰氏阴性菌（Gram-negative bacteria）的细菌细胞被膜是一层复杂的生物屏障，由内膜、肽聚糖（peptidoglycan）周质空间以及携带脂多糖（lipopolysaccharides，LPS）的外膜多层结构组成。随着抗菌耐药性问题日益严峻，学界对解析小分子与细胞膜的相互作用以助力新型药物研发的关注度不断提升。因此，为开展有意义的分子动力学模拟，亟需构建适配的细菌细胞膜表征模型。鉴于细胞被膜的复杂结构，采用显式溶剂的全原子描述方式进行细胞膜模拟在计算上成本极高，仅能实现小规模体系的有限采样。不过，粗粒化（coarse-grained，CG）模型（如MARTINI模型）则能够在生理相关的长度与时间尺度上开展相关现象研究。尽管现有文献中已报道了脂质与脂多糖的MARTINI模型，但适配的肽聚糖粗粒化模型仍较为匮乏。本研究基于Gumbart等人于2014年发表于《PLOS计算生物学》（PLoS Comput. Biol. 2014, 10, e1003475）的全原子模型，在MARTINI框架下构建了肽聚糖网络的粗粒化模型。该模型通过参数化拟合，以复现聚糖链的末端到末端距离。其结构性质——如聚糖链上相邻肽段间的平衡夹角、每个二糖所占面积以及空腔尺寸分布——均与全原子模拟结果吻合良好。面积压缩性、弯曲模量等力学性质也得到了精准复现。在研发新型抗生素的过程中，评估肽聚糖网络的屏障特性至关重要。我们通过伞形采样法，分别针对MARTINI与全原子肽聚糖模型，计算并比较了百里酚（thymol）分子插入该屏障的自由能。结果显示，无论是MARTINI模型还是全原子模型，该插入自由能均低于热运动能kBT。额外的无约束模拟结果表明，百里酚可快速跨肽聚糖屏障转运。我们期望，本研究提出的肽聚糖MARTINI模型，能够助力在更大长度与时间尺度上解析细菌细胞壁相关现象，从而突破全原子模型当前的局限性。