Mechanotransduction in collective cell migration and its synthetic mimic

Speaker: Joachim Spatz
Department Max Planck institute for medical research department of cellular biophysics
Subject: mechanotransduction in collective cell migration and its synthetic mimic
Location: TU Delft (BN Seminar)
Date: 08-09-2017
Author: Nemo Andrea

Joachim Spatz is a researcher from Germany who is working on a range of topics ranging from more traditional cell migration to microfluidics. We were informed that he was interested in the TU Delft’s BaSyC project. His talk today was a lecture in two parts with the first part focusing on cell migration and the latter on a new microfluidics-based approach to sequentially add components to a bilipid membrane.

 Part 1 – Cell migration

The talk started by introducing the motivation for his current work on migration: collective movement. Collective movement, as seen on the macro scale in flocks of birds and fish, is also seen in biology. Retinal cells are the textbook example of a collective form of cell migration during the formation of the eye. The two behaviors (on the macro and microscale) are very much the same behavior and are worth exploring from a physics point of view. Elucidating the principles that lead to collective motion is therefore of strong interest.

 To introduce his work on migration, we first had to be introduced to a few techniques, namely Traction Force Microscopy (TFM). This technique allows one to determine the forces exerted on the ECM by cells migrating on or through the material. From this, one can then also infer the stress between cells themselves (versus cells vs ECM). This can be done because the epithelial cells that they used exhibit a form of collective migration, where the cells are linked through various proteins. From these tractions and stresses, they were able to infer that each cell has a force correlation length that corresponds to about 10 cell lengths. This is then the maximum distance that cells can ‘’affect’’ other cells by exerting force/traction.

 The behavior they were interested in is related to wound healing. They spatially constrained the cells in a rectangular cutout, which could then be removed after which the cells would spread. Interestingly, they observed that cells do not spread out homogenously (equal dilation everywhere), but rather that certain cells would move outwards first and that these cells would drag others with them. These cells, dubbed ‘’leader cells’’, and their dynamics where studied. They wondered what the rules where for this system and how this leader cell first emerges. To do this, they observed the stress and force exerted by cells before the leader cell appears/can be identified. They found that the mean traction in the regions where a leader cell will appear is significantly higher than in cells that will not produce a leader cell, and similarly for stress. They then wished to find the rules for the spacing of leader cells, as they observed the spacing between leader cells (along the cell boundary) had a minimum value. This was not just an artifact, as when the aritificially patterned the cells in a way where the leader cells were spaced closer together, they would return to the natural spacing distance.

They found that the spacing of the leader cells depended on the ECM stiffness (which affects the traction exerted by each cell). This was verified by adding factors that increase or decrease internal actomyosin contraction, which similarly affects cell traction. An increase in traction was accompanied by an increase in leader cell spacing. They postulate is because with greater traction a leader cell can affect larger number of cells due to increased force correlation length.  

 Part 2 – Microfluidics

 In line with their work on migration, they wished to make bilipid membranes that had integrins in the membrane. Such vesicles could then adhere to the environment and could work as an artificial model for some aspects of cell migration. In their attempts to create these they ran into problems with the low mechanical strength and the low yield of the fabrication process. They then developed a new system: water droplets in oil with the water droplets being stabilized by a polymer shell. Through  this strong polymer shell (which can be moved using standard microfluidics platforms) they can inject many different proteins using a series of picoinjectors. As there is no limit to how many picoinjects can be used, and each picoinjector is placed after the other, one has the advantage of being able to sequentially add elements to the vesicle. This is a massive advantage, as various systems will not properly self-assemble if they are all ‘’’thrown in at once’’. Not only does this allow for precise control, the system also had very high throughput with the picoinjectors being able to handle roughly 500 cells per second.


image taken from a presentation by Marian Weiss titled ‘‘Droplet-Based Microfluidics for Sequential Bottom-Up Assembly of Functional Cell-Like Compartments

 Joachim showed us some beautiful demonstrations, with three differently labeled fluorescent actin types being injected into a single cell, and even adding actin and myosin into a vesicle. This then allowed the myosin to contract the actin, making the vesicles active in a real sense. As they did this they observed that the active actin vesicles were slowly rotating. They were able to turn this rotation into motion across a line using another bead attached to the outside of the vesicle, making essentially a very crude form of a migrating cell. (although mechanically completely different from natural cell migration).

 After this, they also added lipids to the vesicle, which, in the right concentration, allowed for the formation of a lipid bilayer in the polymer vesicle. This a remarkable feat, as making a membrane isn’t easy with traditional methods. Next to this, they also demonstrated ester formation and dynamics inside these vesicles in some truly breathtaking video fragments. Importantly, they also demonstrated that is was very easy to remove the polymer shell and just leave the bilipid membrane. To get back to their original goal, they managed to add integrins into this membrane using this technique. Depending on the concentration, they could modulate the extent to which cells adhered to fibrinogen. They are currently working on making artificial mitochondria, and were able to get ATPase and bacterial rhodopsin to work in this artificial context, which I consider to be a remarkable achievement.

 I found this to be the most excited seminar I have attended to date. The first part was interesting to me because I knew a fair amount about cell migration due to my Honours Programme Project, and the second part was extremely fascinating due to the wonderful things that they were able to do. I am very excited for work on the BaSyC project of the TU, so seeing a new promising method like this makes me very, very excited about what is to come.


Information, from the structure of the universe to life

Speaker: Robbert Dijkgraaf
Department: Kavli institute
Subject: Information, from the structure of the universe to life
Location: TU Delft
Date: 07-09-2017
Author: Renée van der Winden

For the Kavli Day on September 7th Robbert Dijkgraaf gave the opening lecture. In this seminar he did not talk about a particular research subject, but rather tried to give us an insight in the assumptions and dimensions we work with in science. In order to do this he discussed several ways in which we can distinguish different types of science and gave examples of many scientists and the different views they have on this matter.

The first thing Dijkgraaf discussed was the scale on which we operate. The smallest things in life we can explore are on the 10-35 m scale, while the biggest structures are about 1025 m. The things we are studying, life and cells, are right in the middle at around 10-5 m. When considering beauty and elegance in science, you can find it at the ends of this spectrum. Those are the places where the messy, everyday life structures crystallize into simple and clear-cut laws and phenomena. The fact that you see this at each end is a reflection of two ways to approach science: reductionism and emergence. Reductionism starts at bigger phenomena and tries to uncover the laws that underlie it. For example, going from a tree to photosynthesis. Emergence starts out at laws of nature and tries to see what they bring about. For instance, going from collisions of water molecules to fluid dynamics. Whichever approach you choose, the end product seems to be more elegant from that point of view, though not from the other, which is interesting. This realization is of the same order as one of the quotes Dijkgraaf used in his presentation, this particular one coming from John Wheeler: “Every law of physics is approximate, not precise”. Dijkgraaf used this quote to illustrate that no matter how elegant a theory might seem, it is still something we came up with to explain our data.Seminar 9

Figure 1: Exerpts from the movie ‘Powers of Ten’

Another thing Dijkgraaf illustrated is that what we consider to be truths, might not in fact be true completely and always. He pointed out the many theories Einstein has come up with to explain how relativity emerges from quantum theory, all while ignoring the possibility that things might be the other way around. Moreover, while discussing black holes, Dijkgraaf mentioned that something strange happens when you move within a certain zone around it, called the horizon. After the horizon, time flips and stops being endless, rather it moves to the center of the black hole, where time stops. What we consider as a given (the flow of time), turns out to be false under certain conditions.

Dijkgraaf ended his talk by saying that the scale of the powers of ten that he started his talk with, might not be the endpoint. He stated that it might be possible there is an even smaller layer below it, a so-called basement layer, which may be concerned with quantum information.

I had hoped to get a chance to see Robbert Dijkgraaf speak for a while, so I was very excited when that opportunity came around. His talk was easy to follow and interesting. I think Dijkgraaf is good at explaining even difficult topics concisely and clearly, so I would very much like to hear him speak about his own ongoing research as well at some point.

Information from the structure of the universe to life 

Aïsha Mientjes 


Seminar 6:  

Speaker:  Robbert Dijkgraaf               

Department: Institute for advanced study (Princeton) 

Subject: Information from the structure of the universe to life 

Location: TU Delft            

Date: 07-09-2017         

On the 7th of September 2017 I attended a seminar given by Robbert Dijkgraaf in de new AS building. Robbert Dijkgraaf started by giving us some information on scales and ‘the power of ten’. He explained that 10 to the power of –35 is the smallest (Planck) scale and the largest scale is 10 to the power of 25 (the Hubble scale). He also elaborated that life occurs right in the middle of this ‘scale line’. Which he finds quite remarkable. He also said that we have discovered approximately three quarters of this ‘line’. 

powers of ten

He continued by explaining a little bit about reductionism and emergence, and that the concepts that we work with are not fundamental but emergent. He also explained the phenomenon of rescaling in physics, where we take a building block of something as a new base. Zooming in and zooming out is a very interesting and intuitive way to look at physics. Life simplifies at both ends. 

Then we arrived at the more technical part of the lecture, which concerned quantum theory and black holes. We were given an explanation about the Einstein-Rosen bridge, which connects spacetimes. We were also told a bit about the horizon of a black hole, where the direction of time is reversed. The amount of information a black hole contains is determined by the size of its horizon. Robbert Dijkgraaf also explained to us that black holes collide and form larger black holes.  

The lecture ended with a small not on quantum information, which is even beyond the smallest scale. It is quickly becoming a universal ‘currency’ in science. 

I found it very nice to attend a lecture of such a famous figure in the world of science. The building was very full which was extra confirmation for me that attending something like this is truly a special experience. I found the talk very interesting and generally fairly easy to follow. There were some more complicated bits, but most of the talk I could understand well. The topic was very appealing to me as well, even though I found it a little abstract at some places. All in all I really enjoyed attending this seminar.

Linear and nonlinear cascade structures as auditory filter models and machine hearing front ends

Speaker: Richard F. Lyon
Department: Neuroscience
Subject: Linear and nonlinear cascade structures as auditory filter models and machine hearing front ends
Location: Erasmus MC
Date: 04-09-2017
Author: Renée van der Winden

In this seminar Richard Lyon came to talk to us about his work on machine hearing. When proposed with the problem of speech recognition, Lyon realized he did not even know how humans performed that task. So he went and studied how the human auditory system works in order to try and replicate that in machines.

The human auditory system contains a structure called the cochlea, which has a spiral-like form. The sound travels along this spiral as a waveform and the structure vibrates in specific places depending on the frequency of the sound. Lyon found the cochlea can be modelled as a cascade of filters through which the sound passes. If these filters are linear, that means the system is not signal dependent. If they are non-linear, the system is signal dependent and that is thus the more accurate representation of the real-life situation. Now on to the type of filter. It turns out that the best model for the human auditory system is a pole-zero filter cascade. You can still distinguish between different scenarios here; you can have either stable or unstable zero crossings. When there are stable zero crossings, the zero crossings of signals with different frequencies align, while for unstable zero crossings they do not. The model that was eventually chosen by Lyon is termed CARFAC and is somewhere in between the pole-zero models with stable and unstable zero crossings. It turned out to be an efficient cochlear model and also an efficient machine front end. Moreover, CARFAC can adapt the gain of the filters to model desensitizing.

Seminar 8

Figure 1: Model of waveforms travelling along the cochlea (Lyons, 2017)

There is a big drawback to CARFAC though, since it is very expensive. Lyon discussed two solutions to this problem. The first one is to use a linear model instead of a non-linear one. This would then be named CARL (Cascade of Asymmetric Resonators, Linear). The second solution is an idea that Lyon came up with. He proposed to think of sound not as a waveform, but as particles. He called these particles ‘sound atoms’. If you combine several of the sound atoms you can also get sound molecules. You can then model these sparse events instead of the waveforms and that is also cheaper. Lyon termed this model CARLA (Cascade of Asymmetric Resonators, Linear with Atoms).

I thought Lyon’s approach  to the speech recognition problem is a great one. It is a very logical approach when you think about it, but apparently people around him had not thought of it. I could not quite understand the details of the model and its workings, but I was interested in how to model a biological system to fit machines. I would like to know a bit more about how it was determined that a filter cascade was a good starting point for a model in this case.