MtL Lecture Series - Remaining Q&As
Check here for answers to the questions we ran out of time for during our lecture series! All posted questions come from lecture attendees and answers come from the listed lecturer.
SEPTEMBER 2, 2020
Sarah Köster: "A Minimal Cytoskeleton - Filament Mechanics and Interactions"
Q: What is vimentin?
A: An intermediate filament protein, intermediate filaments constitute one of the three filament systems in the cytoskeleton.
Q: The filaments you work on, are they purified from mammalian cells or are they artificially expressed in bacteria?
A: We express the proteins in E. Coli, which gives us the possibility to introduce targeted mutations.
Q: If IFs are associated with (plasma) membranes, does that alter their deformation and recovery properties?
A: I honestly do not know as we have not performed such experiments yet
Q: Are these 3 states are interchangeable?
A: You probably refer to the three states in our “three-state model”. Please have a look at https://doi.org/10.1021/acs.nanolett.9b02972 where we describe the model in detail. Transitions between the states are possible, of course.
Q: I can understand, reducing the complexity helps to understand better and perform in vitro experiments. However to compare with the invivo situation, it is important to test the combinatorinal effect of the mechanical forces (for example, compression and shear / tension at the same time). How challenge is to couple them and test in microfluidics setting?
A: Microfluidics is probably not the right approach here (we do not use microfluidics to exert forces, but for providing a controlled sample environment in our setup!). In my opinion it is a good idea to study different modes of force application (stretching, twisting, bending) separately, but of course emergent effects could occur when different modes are combined.
Q: Are IF mutations in human patients known for which altered IF mechanics is the disease mechanism?
A: There are certain mutations in keratin which cause severe skin diseases like EBS. And I know of design mutations which cause heart muscle issue and also aggregation of the protein in heart muscle cells, the two could be related.
Q: Would it not be more biological if MTs and IFs interacted orthogonally, not in parallel?
A: In our experiment, the MTs and IFs interact orthogonally (which is the more straight forward experimental realization), but it would also be very interesting to look at parallel components. Indeed, in migrating cells, MTs and IFs grow in parallel, however, it is not clear, how they interact there
Q: Did you study the filament mechanics with the change in temperature?
A: We did not, actually, all experiments were performed at room temperature.
Q: If the filaments get back to its alpha helical structure then what exactly do we mean by softening?
A: We show that they do not get back to the alpha-helical state, this is why they soften.
Q: Do you want to investigate the impact of epigenetics on the intermediate filaments?
A: This is not currently the plan, but certainly interesting from a biological point of view!
AUGUST 19, 2020
Jochen Guck: "Feeling for Cell Function"
Q: Do cell size and specific stage of cell affect the result of this RT-DC? If yes, how do you discriminate and select your desired type of cells from the rest of the population of cells?
A: Cell size determines how strongly cells are being deformed. The larger the cell, the closer it is to the channel walls, and the larger the deforming stress acting on it. That's why we use a numerical model to convert the size and deformation information into a Young's modulus, assuming that cells are homogeneous, isotropic elastic spheres (analogous to using the Hertz model to extract an apparent Young's modulus from AFM-indentation measurements). This Young's modulus will reflect many changes in cell function, such as a malignant transition from a healthy to a cancerous cells. Or duing the different stages of cell cycle progression. Whether the Young's modulus itself depends on the size of the cell is somewhat of an open question, with contradicting findings in the literature. Regarding the second part of the question, which relates probably to trying to discriminate between cell sub-populations in a heterogeneous sample (such as blood): the changes that occur based on cell cycle progression, for example, are often the reason for the spread in a sub-population. Quite often, the sub-populations are still distinct, even considering the spread in their properties that are typically occurring.
Q: Can this technique apply to measure the deformability of cells which are attached (adhered) to the walls of the microchannel?. Currently, I see from the paper the cells are exactly positioned in the center of the microchannel. For example, HUVECs (endothelial cells) which are present in the inner walls of blood vessels.
A: Only cells that are in suspension can be measured. Normally adherent cells would have to be detached from the surface first and then measured. This is similar to the requirements for measuring with FACS. Detachment which will most likely change their mechanical properties. Relative differences between different cell populations are often preserved though, if they exist in an adherent state, even after detachment.
Q: Cool stuff! Is it possible to use the deformability cytometer setup to induce deformability in model lipid systems in a controlled manner?
A: In principle yes. If the lipid vesicles (the bilayers would have to form vesicles to be accommodated in RT-DC) have about the same size they will all be deformed to the same extent. However, vesicles are often easier to deform than entire cells so that they would be deformed a lot, or else you would have to reduce the flow speed, which would then also reduce the stress and the deformations. But this would also reduce the throughput at the same time. Or else, you could use a measurement buffer with lower viscorsity than we typically us.
Q: After injecting the Paam beads to the cells, what did you detect? the cells stiffness or the intercelluar forces? the deforamtion of the beads resulting from the mechanics of the cells or the intercellular forces？
A: The beads are not injected into the cells, but reside rather between the beads (maybe that is also what you meant). Then we detect the shape changes of the injected beads over time. The shape changes come from the intercellular forces between the attached cells and the deforming bead, but also from other stresses present in the tissue.
Q: Is there a threshold time before the mechanical properties of cells are affected by being out of ideal media conditions? If so, how long before cells start to become apoptotic?
A: The mechanical properties of the cell are changing as soon as you detach them from the surface (cells become softer). However, they then quickly stiffen over the course of about 10-15min and settle into a relatively stable state that doesn't change much over a few hours. If and when cells then become apoptotic depends on the cell type, whether they have the right medium, what the temperature is etc.
Q: How about measuring e.g. hollow (that is fluid filled) spheroids like blastula? To get local differences
A: You can certainly measure hollow objects, such as blastula. You just need an appropriately sized-channel so that they fit in, and a change in the imaging optics. However, we currently do not have a suitable model to turn the deformation and size into information about the mechanical properties of the shell (but this could be developed). Inferring local differences within the shell is much harder.
AUGUST 12, 2020
Tobias Erb: "Drop by Drop Towards an Artificial Chloroplast – Building Soft Matter Machines to Fix Carbon Dioxide"
Q: Do all employed enzymes of the CETCH cycle work best in the same environment, or did you make an adjustment for the best surrounding/environment leading to best performance?
A: Some of the enzymes do operate outside of their respective pH or temperature optimum. Part of the optimization strategy was to identify ideal buffer conditions and salt concentrations at which most enzymes would work well.
Q: Is it easier to put the CETCH cycle into other organisms/droplets compared to the natural cycle?
A: This is hard to answer. I would first argue that the implementation problem scales with the number of enzymes and pathways to be implemented into a given cell, independent of its origin (natural or synthetic). Compared to the CETCH cycle, the Calvin cycle needs only two more enzyme to be added to natural metabolism to be established, which is straight forward (see answer to question 15). Now comparing the implementation of similar complex natural or synthetic cycles, the problem of synthetic cycles, such as CETCH is that there are many non-native metabolites that a cell like E. coli might have never seen. These might inhibt the cell's native enzymes and thus directly affect metabolism.On the other hand, trying to (re-)implement natural cycles into other cells might come with the problem that these natural metabolites might get drained into the cells native metabolism and/or disturb the metabolic and regulatory networks of cells directly, which would also pose a big challenge.
Q: Is the H2O2 produced by Mcd not in general problematic to the other enzymes as well (and not only the thylacoids)?
A: Sure, but the photosynthetic membranes are present in much higher concentrations, so that the effects are more obvious.
Q: It seems like the efficiency of the light reaction in the droplets is decreasing, why is that so?
A: Over time, the photosynthetic efficiency of thylakoid membranes drops, mainly because the proteins of the (membrane-bound) photosytems are irreversible inactivated (e.g. through radical oxygen species formed during photosynthesis). One protein that is especially vuknerable is the D2 subunit of photosystem II that normally is constantly replaced in the living cell.
Q: Why are microcompartments used, as opposed to a larger vessel containing the artificial chloroplast membranes?
A: Because the microcomparments are easier to handle and analyze (saving time and material). The small size of microdroplets is also of advanatage when observing multiple experiment in parallel. On one microscopic slide we can monitor hundreds of droplets (= individual experiments) side-by-side.
Q: You mentioned that you altered an enzyme to directly accommodate O2 in the binding pocket. How easy was that? Are you using a lot of computational simulation beforehand for this kind of enzyme design tasks?
A: Most of the enzme re-engineering is done by manual inspection of the active site. This is very effective and does not require elaborate tools. The first shell mutations are very often very obvious to us and anything beyond (second/third shell is pretty hard to simulate).
Q: Why life didn't evolve to have the most efficient metabolic cycle like CETCH cycle?
A: This is simply a matter of time and serendipity. The ECR enzymes evolved relatively late in evolution, at a time, where the Calvin cycle was already invented, which made it pretty hard to compete with an existing process. Moreover, you should consider, how much it takes to evolve one or the other process. For the Calvin cycle of natural photosynthesis, it takes only two more enzymes to be added to the pentose phosphate cycle to "close" a functional cycle. Acquiring two genes (RubsiCO and phosphoribulokinase) is already sufficient to create the Calvin cycle in the background of natural cells (such as E. coli). The CETCH cylce on the other hand requires that 15 enzymes come together in time and space, which is highly unlikely. In other words, Nature has only sampled a tiny fraction of possible CO2 fixation patwhays and evolved probably the most likely one...
JULY 15, 2020
Ulrich Schwarz: "How Non-Muscle Cells Combine Signaling and Self-Assembly to Generate Force on Demand"
Q: What would you consider as the biggest gap of knowledge in the field of mechanobiology? Or what would you like to find out most in future?
A: Mechanobiology today is a mature research field and we now understand that many processes in the cell have a mechanical component. There is not the one magical molecule which regulates everything mechanical, but we deal with a network of mechanosensors and their downstream responses, e.g. for gene expression. We would like to develop models that cover all the relevant processes, not only force-induced dissociation, but also e.g. mechanosensitive assembly.
Q: It was shown in single myosin experiments, that the duty ratio (e.g. the time in which the myosin remains attached to the actin filament) differs quite alot between NM II A and NM II B. This was attributed to different modes of action: NM II A is only bound for short amounts of time, enabling powerstrokes. NM II B remains longer in the bound state and may therefore generate tension. These findings fit perfectly into your argumentation towards the end. What are your thoughts regarding the interplay of the single myosins and their characteristics for the stress generation of the whole cell?
A: We know that in biology, the properties of single molecules are key: even a single point mutation can lead to a disease. However, from the physics side, we also know that ensembles of molecules can behave differently from single molecules and in fact lead to emergent properties. For example, the force-velocity relation of a mixed minifilament depends on the mixing ratio of nonmuscle myosin II A versus B, and the physical basis is exactly the different duty ratio. In our work, we aim at predicting the large-scale effects of such microscopic differences, and this needs tools from statistical physics.
Q: There exist different drugs to inhibit myosin activity. Blebbistatin for example blocks the ADP release and thus should stall the myosin in the weakly bound state. How do you think would this affect the stress generation? As the blebbistatin treated myosins are weakly bound they may act as some kind of steric hindrance. What are your thoughts regarding this?
A: We can model the effect of blebbistatin in our crossbridge cycle model. The effect of calyculin is more upstream and we have incorporated this before in a model for the Rho-pathway (Besser and Schwarz, NewJournal of Physics 2007). These pharmacological interventions are standard practise and we understand them well.
Q: Are there diseases where NM Myosin II is affeted by mutations? If yes, is it possible to model them to figure out what happens to force generation depending on cell size?
A: There are many myosin mutations that lead to diseases, e.g. affecting the heart and our sense of hearing. Because our analysis of minifilament assembly starts from the sequence, we can in principle incorporate the effect of mutations.
Q: I have been wondering if the cell edge detach upon light induced myosin assembly and contraction, or if there is a maximum contraction reached which is not high enough to so so?
A: Indeed, the cell will detach if you opto-activate contractility at the rim. This is very physiological because lamellipodia and stress fibers often detach themselves during cell migration, both on a substrate and in gels.
Q: A quick question to your introduction: I have been wondering why you are using fluorescent beads in two colors for the TFM experiments with the Malaria parasite instead of one?
A: This is our version of super-resolution microscopy: by using two colors, we improve the spatial resolution of TFM, compare Sabass et al., Biophysical Journal 2008.
MAY 27, 2020
Helmut Grubmüller: "Nanomachines at Work: Atomistic Simulations of Biomolecular Systems"
Q: How does MD results compare with those from experiment?
A: Difficult to come up with a general answer other than ‘it depends’. The type of approximations that I mentioned and which are required to render the simulations computationally feasible are chosen such as to be accurate for those observables/experimental quantities that are most relevant to the biological questions we’re interested in, while other observables are, by construction, less accurate or cannot be calculated at all from the simulations. For example, I have shown the mechanical unbinding atomic force microscopy experiments which come out quite accurately. In contrast, if I were to calculate a heat capacity from the simulations, that would be super inaccurate, because it is strongly affected by quantum effects of the nuclear motions which we describe in classical approximation. On the positive side again, free energies of binding of ligands to proteins (affinities) usually are accurate to about 1 kcal/mol, sometimes even better, depending on the complexity of the system.
Q: How difficult is it to deal with long-range interactions in finite simulation sizes (such as periodic boundary conditions)?
A: Indeed, long range interactions used to be a considerable challenge in the nineties, because simple summation of the Coulomb-sum scales with N^2 (N being the number of atoms). At that time, the Coulomb forces were therefore simply ‘cut-off’ at about 1 nanometer, implying sometimes severe artefacts. Today, there are very efficient electrostatics methods available (scaling as N log N or even as N), such as Particle Mesh Ewald (PME) or the fast multipole methods (FMM), which can also handle periodic boundaries. In fact, PME can *only* handle periodic boundary systems. Still, the calculation of the electrostatics forces and energies comprises a large fraction of the total computer time required, but it’s not *that* much of a challenge any more.
Q: S. Grimme released recently a force field in Angewandte where there are terms that "allow" bond breaking and forming. Would you try such a force field, if it were to be accurate ?
A: Nothing wrong with that in principle, except if I want to study a process that does not involve bond formation or breaking, I would prefer one that does not allow for chemistry, simply because it is less complex and thus the calculations will be faster. There have in fact been many attempts to construct force fields that allow at least a certain type of chemical bonds to be formed and broken. To achieve full generality is, of course, super-challenging, and I am not aware that this has been satisfactorily achieved. It is definitely worth exploring further, though. One important point to consider: Besides the force field, there is a second main contributing factor to the accuracy of our results, which is sampling. In free energy calculations, for example, increasing the simulation length typically will improve accuracy. Now given a certain amount of computer time, and given the choice between a moderately accurate but simple and thus fast force field, and a super-accurate but complex and slow one, there is a trade-off: It may well be that the super-accurate force field only allows for very limited sampling, and so the gain in force field accuracy is outweighted by sampling error, and thus you would fare better with the sampler force field.
Q: If you try several simulations of the same protein with different resolution of structure e.g. from XRD or CryoEM methods do the results of simulations differ?
A: That can easily happen, particularly for structures at lower resolutions that, say, 3 Angstrom or so. In a typical simulation, you first minimize the system (i.e., relax steric strain, van der Waals overlaps etc.), and subsequently let it simply run at room temperature for a few 100 nanoseconds (‘equilibrate’) to see if it remains stable and also remains similar to the initial structure. This is all before the actual production runs in which you calculate the relevant observables. What you then typically observe is that not so well resolved structures tend to deviate more (and also faster) from the initial structure than higher quality structures. We have at some point even suggested to use this effect to assess the quality of experimentally determined structures.
Q: A question regarding the forces in the ligand unbinding simulations. There was a lower limit on the forces applied, around 400 pN in the 1990s, and a bit lower now. What causes this limitation, is it related to computational power?
A: Actually, it is not a lower limit on the applied forces, but on the speed with which we pull the ligand out of the binding pocket (referred to as ‘loading rate’ -- strictly, how fast we ramp up the pulling force). If you ramp it up too slowly, then, even after a year of computing time, there is still not enough force applied for the ligand to unbind. Simple as that!
Q: Do you consider carbohydrates in your systems ?
Q: How big is the efford difference beween a interaction simulation and a simulation of a complete chemical reaction?
A: I guess you’re referring to the difference between a force field based simulation and a QM or QM/MM simulation? That very much depends on the employed QM method. There are QM methods which are maybe only ten times more effort than force field based calculations, but are rather inaccurate. And there are super-accurate methods for which you simply can’t afford to simulate more than 10 atoms or so. The latter also scale much worse [e.g., ~ N^6 or even ~ exp(N)] than force field based simulations (~ N). As a practical rule of thumb, expect an increased effort by two or three orders of magnitude for a typical solvated enzyme for QM/MM with the active site (like 100 atoms) being described quantum-mechanically and the rest by the force field.
Q: What kind of validation parameters e.g. reproducibility etc we can use for simulation methods?
A: Reproducibility is of course key: If the quantities of interest come out different for repeated simulations, that’s not good. Of note, every simulation system is what is called ‘chaotic system’, meaning that with initial conditions that differ only by a super-tiny amount, two simulations will diverge in phase space (i.e., in the details of the individual atomic motions) rather quickly, in a few picoseconds. But this is also true for the real systems in nature, so it can’t be a problem per se. What would be a problem, of course, is if you calculate a binding free energy, and in a second identical calculation you get quite a different value (see my comment on the sampling error above). So that should not happen. Of course, getting similar results in repeated simulations is good, but strictly does not guarantee an accurate result: you can also get reproducibly wrong results (e.g., due to an inaccurate force field). To me, therefore, the best validation is always to calculate something that can be measured, then measure it, then compare to see if the simulation agrees with the experiment.