Dynamic Architecture of Mycobacterial Outer Membranes Revealed by All-Atom Simulations

Reviewer #1 (Public review):

Disclaimer:

This reviewer is not an expert on MD simulations but has a basic understanding of the findings reported and is well-versed with mycobacterial lipids.

Summary:

In this manuscript titled "Dynamic Architecture of Mycobacterial Outer Membranes Revealed by All-Atom 1 Simulations", Brown et al describe outcomes of all-atom simulation of a model outer membrane of mycobacteria. This compelling study provided three key insights:
(1) The likely conformation of the unusually long chain alpha-branched beta-methoxy fatty acids, mycolic acids in the mycomembrane, to be the extended U or Z type rather than the compacted W-type. (2) Outer leaflet lipids such as PDIM and PAT provide regional vertical heterogeneity and disorder in the mycomembrane that is otherwise prevented in a mycolic acid-only bilayer.
(3) Removal of specific lipid classes from the symmetric membrane systems leads to significant changes in membrane thickness and resilience to high temperatures.

Strengths:

The authors take a step-wise approach in building the complexity of the membrane and highlight the limitations of each of the approaches. A case in point is the use of supraphysiological temperature of 333 K or even higher temperatures for some of the simulations. Overall, this is a very important piece of work for the mycobacterial field, and will help in the development of membrane-disrupting small molecules and provide important insights for lipid-lipid interactions in the mycomembrane.

Weaknesses:

(1) The authors used alpha-mycolic acids only for their models. The ratios of alpha, keto, and methoxy-mycolic acids are known in the literature, and it may be worth including these in their model. Future studies can be aimed at addressing changes in the dynamic behavior of the MOM by altering this ratio, but the inclusion of all three forms in the current model will be important and may alter the other major findings of the current study.

(2) The findings from the 14 different symmetric membrane systems developed with the removal of one complex lipid at a time are very interesting but have not been analysed/discussed at length in the current manuscript. I find many interesting insights from Figures S3 and S5, which I find missing in the manuscript. These are as follows:

a) Loss of PDIM resulted in reduced membrane thickness. This is a very important finding given that loss of PDIM can be a spontaneous phenomenon in Mtb cultures in vitro and that this is driven by increased nutrient uptake by PDIM-deficient bacilli (Domenech and Reed, 2009 Microbiology). While the latter is explained by the enhanced solute uptake by several PE/PPE transporter systems in the absence of PDIM (Wang et al, Science 2020), the findings presented by Brown et al could be very important in this context. A discussion on these aspects would be beneficial for the mycobacterial community.

b) I find it interesting that loss of PAT or DAT does not change membrane thickness (Figure S3). While both PAT and PDIM can migrate to the interleaflet space, loss of PDIM and PAT has a different impact on membrane thickness. It is worth explaining what the likely interactions are that shape membrane thickness in the case of the modelled MOM.

c) Figure S5: Is the presence of SGL driving PDIM and PAT to migrate to the inter-leaflet space? Again, a discussion on major lipid-lipid interactions driving these lipid migrations across the membrane thickness would be useful.

Reviewer #2 (Public review):

Summary:

The manuscript reports all-atom molecular dynamics simulations on the outer membrane of Mycobacterium tuberculosis. This is the first all-atom MD simulation of the MTb outer membrane and complements the earlier studies, which used coarse-grained simulation.

Strengths:

The simulation of the outer membrane consisting of heterogeneous lipids is a challenging task, and the current work is technically very sound.

The observation about membrane heterogeneity and ordered inner leaflets vs disordered outer leaflets is a novel result from the study. This work will also facilitate other groups to work on all-atom models of mycobacterial outer membrane for drug transport, etc.

Weaknesses:

Beyond a challenging simulation study, the current manuscript only provides qualitative explanations on the unusual membrane structure of MTb and does not demonstrate any practical utility of the all-atom membrane simulation. It will be difficult for the general biology community to appreciate the significance of the work, based on the manuscript in its current form, because of the high content of technical details and limited evidence on the utility of the work.

Major Points:

(1) The simulation by Basu et al (Phys Chem Chem Phys 2024) has studied drug transports through mycolic acid monolayers. Since the authors of the current study have all atom models of MTb outer membrane, they should carry out drug transport simulations and compare them to the outer membranes of other bacteria through which drugs can permeate. In the current manuscript, it is only discussed in lines 388-392. Can the disruption of MA cyclopropanation be simulated to show its effect on membrane structure ?

(2) In line 277, the authors mention about 6 simulations which mimic lipid knockout strains. The results of these simulations, specifically the outcomes of in silico knockout of lipids, are not described in detail.

(3) Figure 5 shows PDIM and PAT-driven lipid redistribution, which is a significant novel observation from the study. However, comparison of 3B and 3D shows that at 313K, the movement of the PDIM head group is much less. Since MD simulations are sensitive to random initial seeds, repeated simulations with different random seeds and initial structures may be necessary.

(4) As per Figure 1, in the initial structure, the head group of PAT should be on the membrane surface, similar to TDM and TMM, while PDIM is placed towardsthe interior of the outer membrane. However, Figure 5 shows that at t=0, PAT has the same Z position as PDIM. It will be necessary to provide Z-position Figures for TMM and TDM to understand the difference. Is it really dependent on the chemical structure of the lipid moiety or the initial position of the lipid in the bilayer at the beginning of the simulation?

Minor Point:

In view of the complexity of the system undertaken for the study, the manuscript in its current form may not be informative for readers who are not experts in molecular simulations.