Information Processing by cells

Cells have evolved complex signaling pathways to cope with environmental stress. Quantitative biology and interdisciplinary approaches can be used to probe the dynamics of signaling pathways in living cells. We studied the signal processing ability of yeast cells in response to osmotic stress ...[+]

Physical Limits of cell signaling

Whereas the genetic background is a key regulator of cellular functions, the physico chemical environment has long been considered as a major regulator of global cellular functions. Cells behaviours depend on temperature, pH, light... We show here that molecular crowding may be a major regulator of biochemical rates inside the cells... [+]

Cybergenetics: Real Time Control of gene expression

The development of cell - computer interfaces is very recent and yet bear fantastic potential applications. We are among the first to set up an experimentla platform for the real time control of gene expression in yeast at the single cell level... [+]

Synthetic Biology

Synthetic biology and, more generally, genome editing are fantastic tools and concepts for engineers who wants to explore, tinker, hack the functioning of living systems. Using and developping tools of genome editing is part of our primary research activities...[+]

Genetic Drift

Genetic drift is a passive selection process, usually at play in small population, where the probability of fixation of a neutral mutation is non null. We are interested in how this important effect in population dynamics is also at play in large, spatially structured population of yeast and bacteria... [+]

Ideal Cylindrical Colonies

We design a novel method to grow yeast and bacteria colonies into any arbitrary geometries. We are studying how efficiently cell assemblies grow when their shape is controlled. In particular we are studying cylindrical colony that are a perfect candidate for relating the metabolic yield to the growth of the population... [+]

Electrotaxis in worms

C. elegans Biomechanics

C. elegans is a small worm which moves by propagating a flexural wave along its body. We are studying how it adapts its locomotion pattern to its mechanical environment. Worms locomotion is indeed passively modulated by the resistance of its environment and possibly through proprioception...[+]

Mechano-Biology

Mamalian cells are known to both exert forces on their substrate and to react physiologically to forces exerted on them. We have studied various aspects of force generation and collective cell migration using microfabrication methods. We are now imaging with super resolution microscopy the structure of focal adhesion complexes, which are a key unit in mechano sensing and force transmission between the cytoskeleton and the extra cellular matrix... [+]

Cell Based Assay

Physics of Dunes

Sand dunes can adopt various shapes depending on the directionality of the local winds and the sand supply. The most simple, and studied dune is the barchan, a crescent shape dune that forms in desert blown by very directional wind. We were the first to develop a laboratory experiment to study downsized version of these structure in the lab, opening the way to several key findings on the dynamics of barchan dunes...[+]

Sand dune Collisions

Barchan dunes are crescent shaped sand dunes that migrate downwind. The smaller they are, the faster. To get an order of magnitude, sand dunes can go up to several tens of meters a year in aeolian desert for the smallest ones. Since not all dunes have the same size, they travel at different speed and can collide. The result of such events is shown to be a key regulator of the structure of deserts...[+]

Linear Dunes

Sand dunes can adopt the shape of long ridges which orient either transversally or longitudinally to the average wind direction. Such shapes form primarily when the wind direction alternates between two dominant orientations. We upgraded our experimental setup to study the stability and formation of such dunes...[+]

Polygonal vortices

It's well known that the surface of a liquid in a rotating cylinder will adopt a paraboloid like shape. Strikingly this is not always the case, and we observed that polygonal shapes can appear at the surface of the liquid...[+]

../Information Processing in Yeast/

Signal Processing by the HOG MAP Kinase pathway, PNAS, 2008.

Cells have evolved complex signal transduction pathways to detect, signal and adapt to the changes of their physico chemical environment. As for an electronic device, cells can process information and take decisions thanks to their many signaling pathways. Although the gene regulatory networks and the key proteins involved in signaling cascades are usually well known, the dynamical properties of signaling pathways are poorly described in the litterature. Their robustness, speed, frequency behavior, bandwidth are mostly not known, limiting our ability to predict cells dynamical behavior and our capability to use synthetic biology to design novel signaling cascade with well defined dynamical properties. We started to investigate this by adapting concepts from engineering and dynamical system theory to living cells. We used the HOG pathway, a well known, canonical signaling pathway in the budding yeast (S. cerevisiae). This cascade is activated upon hyper osmotic stress and triggers a large transcriptional response, including GPD1 dependent glycerol synthesis to perform volume homeostasis. We designed a simple and versatile microfluidic device to change periodically the osmotic environment of yeast cells while imaging the time dependent localisation patterns of Hog1p, a Mitogen Activated Protein Kinase (MAPK) and the key transcription factor of the osmotic response in yeast. Hog1p is indeed translocated in the nucleus upon osmotic stress. Varying the frequency of the variation of the osmotic environment, we found out that the HOG cascade behaves as a low pass filter. It cannot be activated by too fast fluctuations of its environment, but is able to faithfully follow a periodic osmotic input slower than typically 200 seconds. The inverse of this critical timescale is a measurement of the slowest in vivo kinetics at play in the HOG cascade. Importantly, we were also able to show that modifying the architecture of the pathway can lead to a different dynamical behaviour. More details about our work can be found here [PNAS 2005].

This work and its underlying concepts (synthetic biology, engineering and physics of dynamical systems) is the basis of both the long term goal of our lab and the interdisciplinary center of interest of its members. We are now interested in understanding what biological and physical parameters set the typical timescales and dynamical properties of cellular processes, from signaling pathways to gene expression and regulatory networks. To this end, we are developing and using microfluidics, single cell microscopy imaging, synthetic biology and engineers method to study the information processing ability of living systems. We recently studied the physical origin of cell signaling impairment by volume compression, designed an experimental platform implementing an in silico feedback loop to control gene expression in yeast, bacteria and mamalian cells, designed novel tools for microfluidics, image analysis and synthetic biology. Our main focus related to information processing is now oriented towards the development of cell - computers interfaces to drive cells automatically thanks to a computer controled feedback loop.

[Actual colloboration with S. Léon, Past collaboration with A. Miermont (Alumni, PhD), S. Bottani (Paris Diderot), M.N McClean (Princeton University), S. Ramanathan (Postdoc Advisor, Harvard University)]

../ Molecular crowding and cell signaling /

The cell interior is an exceptionnaly crowded environment; filled with proteins, cytoskeleton structures, organelles... The nucleus is equally, if not more, dense since it contains all chromosomes packed into higher order structures, waiting to be opened and processed by the transcription or the replication machinery. This crowded environment may impact the information processing ability of cells. Indeed, it is well known that diffusion of proteins is usually slower in the cellular context than in vitro. Moreover, we observed sometime ago that while the activation of the HOG cascade was obtained robustly and reproducibly after a 1M Sorbitol osmotic stress, it was greatly slowed down after a severe osmotic stress (2M Sorbitol). It was a striking observation since one would have expected a stronger/faster cellular response to help the cell to deal more efficiently with harsh conditions. We first checked our initial experiment and studied the activation of the HOG cascade over a range of osmotic stress using a basic fluidic device to change rapidly the cells environment and to observe their response over long time (> 1hr) by fluorescence microscopy. Both the dynamics of nuclear translocation of Hog1p and its level of phosphorylation were significantly slowed down when the osmotic stress increased. This slowdown could have various causes: for example a severe osmotic stress can alter the activity of a specific protein in the cascade, the closure of the nuclear pore or a loss of sensitivity of the osmosensors due to protein misfolding. We explored another lead, involving molecular crowding in the cytoplasm. Indeed, when the cell undergoes an hyper osmotic stress, it is mechanically compressed and its volume decreases. Maximum compression corresponds to a final volume of about 40% of the initial volume and is obtained for osmotic stress higher than 2M sorbitol. Thus, the protein density increases in the cytoplasm, and can hinder the mobility of all proteins. This effect is well known in colloidal solutions for which at high density one observe a glass transition and a pronounced slowdown of dynamic processes (the viscosity diverges exponentially fast). Since the cytoplasm of cells is already a dense environment in proteins (~300 g/L for a volume fraction of 20-30%), the volume reduction of the cytoplasm which results from a compression by hyper osmotic stress can lead to a dense enough state to significantly reduce the dynamic cellular processes. We measured the mobility of some signaling proteins fused with a GFP in yeast immediately after a more or less severe hyper osmotic stress. Using FRAP (Fluorescence Recovery After Photobleaching) and FLIP (Fluorescence Loss in Photobleaching) we observed that the distribution of proteins was hampered by severe osmotic stresses (> 1M Sorbitol). This effect was more and more marked as the osmotic compression increases and above 2M osmotic stress, the distribution of protein in the cytoplasm was frozen and no recovery was observed even after several minutes. We then analyzed the dynamic behavior of other signaling pathways and cellular processes as the cell volume was compressed by a severe osmotic stress. In all cases, we observed a slowdown, and even a complete arrest of dynamic processes for too severe stress. This confirmed our interpretation in terms of molecular crowding. Finally, we showed that the slowdown is consistent with a glass transition for a soft colloid – that is made of deformable object. In such a case, one can predict that the diffusion of proteins, and thus the rate of chemical reactions, are slowed down exponentially fast with the density of the colloid, in good agreement with our experimental results.

Logarithmic plots of the signaling dynamics. This figure is based on data shown in Figs. 2F and 4J. A and C show log(Tmax) and the corresponding fits in function of the relative volume variation V/Vmax. B and D show log(Tmax) as a function of the relative protein density. The protein relative density, ϕ/ϕmax, is defined as the inverse of V/Vmax and normalized so that its maximum is 1 (arbitrary choice). Assuming that the cytoplasm behaves as a soft colloid, one expects to observe an exponential increase of the viscosity with the protein density: η ∼ exp(αϕ). Here, we do not have access to the viscosity. To link our measurements to the theoretical prediction of a soft colloid glassy transition, we assume that the diffusion coefficient of proteins vary inversely proportionally to the viscosity. Further assuming that the rates of diffusion limited kinetics scales as r ∼ Dϕ, we obtain Tmax ∼ 1/r ∼ 1/ϕ exp(αϕ) or, alternatively, Tmax ∼ Vexp(α′/V). The fit functions are in agreement with an exponential increase of the viscosity of the cytoplasm with the protein density as it would be the case for a soft colloid near a glassy like transition.

These results outline why it is so important for a cell to maintain its size and its protein content in order to function optimally. Ideally, the cells need to have a large number of proteins in a small volume both to minimize stochastic effect on signal transduction and gene expression and to maximize the likelihood of protein protein collisions. However, we showed that the total protein density is limited by the existence of a glass transition for too high protein densities. This is a sharp transition and if the cell density is not maintained, the cell can be almost instantaneously frozen - and thus oucompeted by other cells. Cells must maintain their volume (or their protein content) balanced to find a trade off between these two counter acting effects. We will continue to explore these issues in greater detail by looking at the effect of cellular compaction on various cellular processes.

See a comment on this work on the yeast SGD community portal and in the CNRS news.

These results were discussed in the PhD thesis of A. Miermont and published in “Severe osmotic compression triggers a slowdown of intracellular signaling, which can be explained by molecular crowding. PNAS 110 (14), 5725-30 (2013)” [In collaboration with A. Miermont, S. Bottani, M. McClean, S. Léon]

../ Real Time control of gene expression /

The ability to control the osmotic environment of yeast cells while imaging them in real time has allowed us to explore a novel, exciting field of research, on which the lab will focus his efforts for the coming years. Together with Grégory Batt (CR1, INRIA), we developed an automated experimental platform for the control of gene expression at the single cell level. We first used yeast and the modulation of the osmotic environment as a method to initiate transcription of a gene which level of expression can be measured in real time by fluorescence microscopy. The idea was to implement a feedback loop via a numerical model whose input is the current state of the cell and the output is the cell's environment to achieve the desired level of expression of the gene of interest. The platform includes a microfluidic device which liquid input could be selected by a computer, an automated fluorescence microscopy to image cells over long periods of time (> 15h), image analysis to segment and track individual the cells, and an algorithm for deciding how to change the environment to force the level of expression of a gene to follow a user-defined set-point. To induce gene expression we used osmotic stress which activates the cascade HOG and leads to the transcription of a fluorescent gene under the control of the promoter pSTL1. Everything takes place in real time, the cell state being updated every 6 minutes.

Platform for the control of gene expression in yeast. A microfluidic device is used to quickly change the osmotic environment of yeast cells. Cells are imaged by fluorescence microscopy and analyzed in real time (segmentation and monitoring). The measurement of their fluorescence state is used to determine via an algorithm of Model Predictive Control (MPC) the best way to change the environment in the future to force the level of fluorescence to follow a predetermined profile.

Since the HOG cascade is used by the cell to adapt to osmotic stress, we used the fact that it behaves like a low pass filter and sent short and well time-separated osmotic shocks(to avoid cell adaptation to their osmotic environment) and change the duration of the shocks (between 5 and 8 minutes) to stimulate more or less strongly the pSTL1 promoter. In this way, the cellular adaptation and cellular feedback loops are inoperative: the cells do not have time to adapt, but they can produce a small amount of protein at each shock. To close the feedback loop, we chose to use the so-called "Model Predictive Control" method. This approach requires a model of the system, here the expression of pSTL1 following osmotic shock. We used a minimal model, in the form of two coupled differential equations, whose parameters have been calibrated experimentally. This model, coupled with a state estimator (nonlinear Kalman filter) is used to find what would be the best combination of "osmotic stimuli" to use in the next 2 hours. The search for this profile (pulse duration, pulse spacing) is made every six minutes and the best profile is used by the micro-fluidic system. A major challenge of our approach lies in the time lag between cell stimulation and the production of fluorescent protein. In our case, this period is of the order of 20-30 min (transcription, translation and maturation of fluorescence). Despite this, we were able to show that it was possible to control gene to follow constant or time dependent profiles (sinusoidal, trapezoidal) at the level of populations of cells or for isolated cells. In other words we made an interface between a computer and living yeast cells that can drive their behavior in real time over several generations. We are now extending this platform to other organism (E. coli, mamalian cells) and building synthetic gene expression systems (inducible promoters, optogenetics) to gain control of cellular functions and gene regulatory networks. This is a very promising field of research, with several applications (bioproduction, cell reprogrammation, cell optimization) and a potential to enhance real applications of synthetic biology.

See a comment on this work in the CNRS webnews, the CNRS magazine; in the french press (L'Humanité) and on the web.

These results were discussed in the PhD thesis of J. Ulhendorf and published in details in “Long-term model predictive control of gene expression at the population and single-cell levels. PNAS, 109 (35), 14271-6, (2012)". [In collaboration with A. Miermont, S. Bottani, F. Deveaux, T. Delaveau, F. Fages, G. Charvin, G. Batt]

../ Synthetic Biology/

We use genome engineering and synthetic biology as a tool to synthetically add functions to cells. We routinely use yeast and bacteria transformation protocols to delete genes or fuse sequence that codes for fluorescent taggs. More recently we have benefited from our collaboration with X. Duportet to develop more advanced genome editing tools in the lab. In particular we are now able to perform genome edition in Mamalian cells. The method we use relies on integration at a well define locus (AASV1 in human cell line) of preengineered cell lines by using the BxB1 recombinase activity. Integration is highly specific and efficient. X. Duportet and coworkers were able to integrate several transcription units with minimal proof testing. We combine this integration method with the Modular Cloning (MoClo) method which allows us to quickly generate a plasmid containing several transcription unitsincluding 3'UTRs, promoters, gene, fluorescent taggs, 5' UTRs and to do this in a combinatorial way (different promoters, different fluorescent proteins, different genes, etc...). A similar method can be used for yeast and bacteria to speed up strains production and find the best construct in terms of fluorescence output, degradation taggs and promoters. In particular we used this method to generate several Heck cell lines bearing an antibiotic resistance cassette, a nuclear tagg (NLS signal) and a Tet TREtight induction system driving a fluorescent protein. We are using this cell line (and its derivative) to study the propagation of phenotypic variability from single cells in mamalian cells.

We are actively developing this aspect of our research, as a tool to improve the control of cellular functions by computer aided feedback loop, but also for our other projects on cell signaling, noise in gene expression and cooperation / competition in muticellular systems. We are now implementing other tools such as optogenetics, CRISPR/CAS and TALEs/TALENs to get a complete genome engineering toolbox at our disposal. Several of our PhDs or undergraduate students have participated to the iGEM competition and we are glad to train future students from all discipline to synthetic biology coupled to quantitative methods in microscopy and modeling.

[In collaboration with V. Peschetola (MSC), JB Lugagne (MSC/INRIA), Z. Marinkovic (MSC), A. Llamosi (MSC/INRIA), X. Duportet (PhD, MIT/INRIA/MSC), G. Batt.]

../ Genetic Drift/

Genetic drift is a passive selection process, usually at play in small population, where the probability of fixation through bottleneck sampling of a neutral mutation is high. Said differently, if one take a sample of a large population containing two strains A and B with a neutral point mutation difference. Although the strain will share the same fitness, when a small sample of a population is taken out to start a new population, the proportion of A/B is given by a poisson's law. After several sampling, the probability to get only A or only B is non null and actually fixation can occur in a finite amount of time (sampling). We showed that this effect can also be important in the case of freely expanding colonies of yeast or bacteria which counts billions of cells. Indeed, only cells near the edge are able to access nutrients, divide, and eventually carry out their genes. And this is a small number of cells, so that genetic drift is an important mechanism at the edge of expanding microbial colonies. We tested E. Coli and S. Cerevisiae populations made with a mix of two strains that are isogenic but for the constitutive expression of either YFP or CFP. Both strains have the same fitness, the point mutation difference between the sequence of YFP and CFP can thus be considered as a neutral mutation. Thus, both strains can be distinguished with an epifluorescence binocular and sectors become clearly visible with the lateral expansion of the colony. Moreover, the frontier between sectors is a trace of the random selection process that occur locally at the expanding edge. This study has been beautifully developped by D. Nelson (Harvard University) and co workers on both theoretical and experimental aspects. In the lab we have an ongoing project to study experimentally the relation between the shape and size of the expanding edge and the probability of fixation. To do so we have designed a microfluidic system to constrain the growth of microbial strains in channels of known dimensions.

[Actual collaboration with C. Vulin (PhD), Z. Marinkovic (PhD), JM Di Meglio (Paris Diderot); Past collaboration with O. Hallatscheck, S. Ramanathan, D. Nelson.]

../ Ideal Colonies/

A cylindrical colony, several millimeter high growing on a YPD agar plate

The growth of yeast and bacteria colonies on solid substrate is a conventional tool for biologists and a fascinating, hard to capture, morphogenesis process. Actually, little is known on colony morphogenesis. This contrasts with the fact that microbial colonies are routinely grown on solid agar gels for strains maintenance, genetic manipulations and large scale screening assays. Colonies expand by sucking out nutrients from the solid, moist substrate on top of which they grow. Compared to the homogeneous case of planktonic culture in well mixed liquid media, here, single cells are locally competing for nutrients. If we come back to the case of a yeast population in a homogeneous environment (agitated culture liquid medium), all cells feel and have access to the same resources. It is then enough to know the average rate of division to predict the evolution of the population size (number of cells) over time. In the case of a colony spreading freely on an agar gel, the mathematical description of the rate of growth of the colony is more complex. Indeed, cells at the edge of the colony and the cells in the center of the colony have access to very different nutrient environments. Gradients of metabolites and chemicals generated from the degradation of the metabolites form under and inside the colony. Each cell sees a different environment depending on its position in the colony. We have designed a simple tool to constrain the growth of colonies on a modified porous substrate. Using a technique known as "contact printing", we print the surface of a membrane with a biocompatible polymer (PDMS). We can selectively block the pores of the membrane according to any desired geometry. This allows to create porous geometries with a typical resolution of 100 µm. Once modified, the membrane can be placed on top of a agar gel containing nutrients. Nutrients can diffuse through the membrane but the micro-organisms cannot pass through the membrane as long as its pores are sufficiently small (0.02 microns is a classic standard filtration pore size in microbiology). The growth of micro-organisms (E. coli, S. cerevisiae) on these membranes is very peculiar: colonies form shapes that faithfully respect the geometry of the printed pattern and then rise vertically. For example and in contrast with the conventional case of a yeast colony which extends isotropically and form a flat conical structure, we were able to grow cylindrical colonies. Cylindrical geometry can greatly simplify the study of the growth of an assembly of yeast. Indeed there is no more lateral extension and metabolic gradients are only due to absorption of cells located at the base of the cylinder. We have shown experimentally that these cylindrical colonies grow at a constant rate and that their growth rate can be directly related to the metabolic yield of the colony. In the case where glucose is the limiting nutrient, the growth rate of the colony is not directly dependent on the growthrate of single cells, but rather on the ratio between the average growth rate and the average specific uptake of glucose. The important idea is that in a vertical colony, the cumulative vertical growth rate is equivalent if there is a large number of cells that divide slowly or a small number of cells that divide rapidly at the base of the cylinder, and thus deprives the cells of the upper layers to have access to nutrients. This idea essentially involves rethinking how the "fitness" is defined in teh context of multicellular assembly and link it to the geometry of the colonies. As a side result, using cylindrical colonies it becomes possible to estimate the metabolic efficiency of a colony by simply measuring its vertical growth rate. This vertical growth rate increases linearly with the concentration of glucose at low concentrations (111 mM). For higher concentrations, there is a saturation of the rate of growth, probably related to a significant decrease in the performance of each cell under conditions or glucose is abundant. We are now developing a microfluidic approach to better understand how cells adapt their gene expression levels according to their position in a gradient of glucose. This will allow us to more accurately model the expansion of a colony and to better understand the evolutionary mechanisms at play in setting the shapes of wild colonies.

See the gallery to get insights into the growth of geometrical colonies.

These results will be discussed in the PhD thesis of C. Vulin (MSC). [In collaboration with Z. Marinkovic (MSC), JM Di Meglio(MSC), A. Lindner (INSERM), A. Murray (Harvard)]

../Electrotaxis /

C. elegans, a 1 mm long worm, displays several interesting behavior, that are regulated by its simple, yet efficient, neural network. It is able to forage for food, respond to mechanical stimulus, adapt its locomotion pattern from crawling to swimming depending on the mechanical resistance of its environment, etc... Importantly, it is also able to decide where to move to depending on environmental cues. Such prefered directional locomotion is classicaly studied in the presence of chemicals (chemotaxis). But C. elegans has been shown to perform other tactic behavior such as thermotaxis (ability to move towards a certain range of temperatures) and electrotaxis (ability to move in the direction of an electric field). Yet, the mechanisms by which worms are able to sense and orient in the direction of an electric field are far from begin understood. Interestingly, when worms perform electrotaxis, they tend to move very directionally and as fast as they can. We used this behavior to setup a simple and efficient sorting method for worms based on their locomotory abilities. It only requires an electrophoresis device. Worms are set at the surface of an agar gel which is surrounded by a liquid buffer and a strong electric field is applied. Then, worms move to the opposite side of the gel by crawling. Faster worms will move away from slower ones. Fast and slow worms can easily be collected afterwards. We published this method as a proof of principle in 2011. We are now interested in understanding the behavior of worms in time varying electric field. We designed an automated system where the electric field direction can be reversed periodically and we can now monitor the behavior and locomotion of single worms at high resolution in such conditions. We will use this device to quantify the reorientation behavior for different genotypes with the hope to better describe and understand electrotaxis.

Cumulated trajectories of several worms moving in presence of an electric field. Worms move from left to right in approximately straight trajectories. The small wavelength on the trajectories is the trace of the undulatory locomotion of worms and is of typically 0.5 mm.

This method was published in "Running Worms: C. elegans self sorting by electrotaxis", PLoS One, 2011. [Collaboration with JM di Meglio (MSC)]

../ C. elegans locomotion /

Illustration showing a worm (~1 mm long) and its two gaits, swimming and crawling. Cover of Biophysical Journal.

The mechanics of locomotion of the nematode C. elegans is a perfect example of information processing by a higher, multicellular organism. Such nematodes are indeed able to detect and process a lot of information from their physico-chemical environment. They are capable of measuring gradients (chemical, temperature, osmotic pressure), which allow them to search for food (bacteria) and move to areas suitable for their life (temperature, osmolarity, pH). They are also able to detect mechanical stress: a little pressure on the head of a nematode leads to an immediate backward motion. Nematodes are able to move in various medium and use primarily two types of gaits: swimming and crawling. These two modes of locomotion are very different and are observed in different mechanical environment. Nematodes swim in a liquid medium, but crawl using an undulatory motion when they are moving at the surface of an agar gel - on which they are pinned by capilarity. Our goal was to understand this difference in behavior and in particular to determine whether these two modes of locomotion were the result of a passive adaptation to two different mechanical environments or rather the result of a change in patterns of muscle excitation through a neural processing. To do this we built an experimental system to force the nematode to swim or crawl dynamically. The idea is very simple: a nematode is immersed in liquid between a glass plate and an agar surface. By varying the distance between these two surfaces the nematode locomotion switches from a swimming mode to a crawling mode when pressed against the agar surface. We have shown that this transition is continuous, and that the kinematic parameters of the nematodes evolve continuously with the distance between the plates. This indicates, that there is no binary locomotion mode (swimming / crawling), but rather an infinite number of modes of locomotion for the nematode when it moves in varying mechanical environments. Furthermore, we showed the existence of a power law relationship between the propagation velocity of the curvature wave along the worm body and the period of the oscillation of the curvature wave. This relationship is particularly robust as it is obtained for mutants whose locomotion is impaired; it is also obtained for backward locomotion or movements in different mechanical environments. This seems to be some sort of state equation for the locomotion of the worm that we are now trying to understand in relation with the power used by the worms to move in different mechanical environment.

This study was published in 2011 in Biophysical Journal and detailed in the PhD of Félix Lebois (2011, MSC). [In collaboration with Félix Lebois, Charlotte Py (MSC), Jean Marc Di Meglio (MSC)]

../Mechano - Biology/

[In collaboration with Shiqiong Hu (Postdoc, MBI/NUS), R. Zaidel-Bar (MBI/NUS) and A. Bershadsky(MBI/NUS & Weizman Institute).]

../Cell Based Assay/

Cell Based Assays are an interesting alternative to biochemical assays performed on live animals. The first advantage is to be able to screen for drugs and measure quantitatively the toxicity of chemicals directly at the single cell level, using automated fluorescence microscopy and image analysis. The second advantage is to avoid subjecting animals (often mice) to all kind of harsh tests. Several microfluidic based systems to perform highthroughput cell based assays have been proposed recently. They all lack the (important) ability to perform sequential, independent treatment to an array of cells. We demonstrated in a recent article (Lab on Chip, 2012), that it was possible to do such sequantial assays using a novel method to modify porous filtration membrane. We contact- printed filtration membrane with PDMS using a stamp with sub-millimetric features. The PDMS entered the pores of the membrane and clogged them after a curing step. This way, we could obtained an array of small (1 mm) disks that were porous and surrounded by non porous regions.

Cells could be grown as a monolayer on one side of the filtration membrane, while the other side could be used to deliver drugs and chemical. We demonstrated using DAPI and Calcein-GFP staining that it was possible to adress independently each disks on an 8x8 arrays of 1mm porous disks, separated by 1mm non porous regions.We also successfully treated at a selected location cells with the actin inhibitor, cytochalasin D and observed local depolymerization of the cytoskeleton, showing the potency of our method to deliver drugs locally to an assembly of cells. As shown on the abovepicture, we even managed to perform spatially overlapping tests on the same monolayer of cells, by stacking filtration membrane with overlapping printed patterns. We patented the resulting substrate and called them, micro Patterned Membranes (µPM). Such substrates can be used in other configurations to create a complex spatial delivery of chemicals. We are now exploring its use for microbiology and microfludiics.

../Physics of Dunes/

Snapshot of a subaqeous barchan dunes. Its width is of typically 5 cm and it is made with ceramic beads of 200 µm.

Sand dunes can adopt various shapes depending on the directionality of the local winds and the sand supply. The most simple, and studied dune is the barchan, a crescent shape dune that forms in desert blown by very directional wind and low sand supply. They are pushed downwind by constant erosion on their back and deposition of sand in their slipface. They are considered as the prototypical dune model. The seminal work of R. A. Bagnold in the forties have led several physicist to try to understand the origin and dynamics of such barchans. Unfortunately, experimental field study of barchan dynamics are very limited. Indeed, dunes move and transform very slowly (a few meters per year, for dunes typically 20 meters long...). Also, meteorological conditions can change suddenly (storms, rain) and alter the dynamics. We implemented two methods to access the dynamics of barchan dunes. First we designed a simple and elegant experimental setup to create small dunes in the laboratory. The trick was to form dunes under water. We showed that subaqueous barchans shared the same morphologies and motion than aeolian one, but at the centimeter scale! It was then possible to perform several key quantitative experiments in controlled conditions. We measured the morphologies and speed of subaqueous barchans and showed that they were comparable to field data of aeolian barchans, once rescaled by the saturation length (distance on which the sand flux reaches its maximum). This length is 1000 times smaller under water, explaining the down sizing of dunes under water. This demonstrated that the saturation length is the relevant lengthscale to describe the physics of dunes. We further studied the formation of dunes by using numerical computation of an elementary model that contained only the basic physical ingredients : erosion, deposition and transport by the flow (wind). We showed that barchans adopt a stationnary shape that is the trace of lateral sand redistribution and sand transport in the wind direction. Taken together, our experimental and numerical results gave a comprehensive view of the physics of barchan.

See the gallery, visit the dedicated web presentation or the articles that i published on the subject during my PhD. You can also have a look at the numerous press articles popularizing our work. http://physicsworld.com/cws/article/news/2002/dec/18/sand-dune-science-creeps-ahead

[PhD thesis done in the "Laboratoire de Physique Statistique", Ecole Normale Supérieure, Paris, France. PhD advisor : S. Douady]

../Sand dune Collisions/

Barchan dunes are crescent shaped sand dunes that migrate downwind. The smaller they are, the faster. To get an order of magnitude, sand dunes can go up to several tens of meters a year in aeolian desert for the smallest ones. Since not all barchans of one location have the same dimensions, one can expect the small, upwind dunes to collide with larger, downwind ones. Aerial pictures give us hints of such events, but it is hardly possible to observe and study the dynamics of a collision event in the field. It is however easy to observe collisions event of centimeter long subaqueous barchans in the lab. We used our experimental setup to produce binary collisions with varying impact parameter, that is for dunes that were not strictly aligned. We observed that several outcomes were possible, from the simple dunes merging, to dune splitting and mutiple emissions of very small barchans in the wake of one of the barchan's horn. Collisions is thus a complex process and its modeling is far from trivial. Similar results were obtained by solving numerically an elementary model of barchan.

Snapshot of a collision between two subaqueous dunes made with green and red glass beads to distinguish them. This image shows how the small, fast green dune destabilizes the larger, slower red dune. The downwind horn of the red dune will separate and form a new barchan downwind while the green dune will merge with the remaining part of the red one.

Importantly, barchan dunes are fundamentally unstable objects. They capture sand carried by the wind on their back and lose sand from their horns, but not in equal proportion for all sizes. As a result, barchans can only grow or shrink and disappear when considered as isolated objects. This may appear as a paradox, since barchans can actually be found in many deserts. However, barchans are very rarely found not as isolated objects. They usually are part of large barchan assembly, that are called corridors. Such corridors can contain several thousand barchans extending several kilometers wide and several tens of kilometes along the wind direction. We showed, using a meand field theory approach, that collisions of barchans within such assembly can behave as a stabilizing mechanism, redistributing sand mass from large dunes to smaller ones. Thus collisions may be a key interactive process explaining why corridors are stable while its barchans are unstable when considered isolated.

Recently we pushed this idea and setup a numerical study of thousands of barchans in interactions using an off-latice approach. Every dune is represented by its size and its position. Collisions are modeled through a set of elementary rules which take into account dune merging, dune splitting and emission of smaller barchan. Mass conservation is used to redistribute sand between dunes. We then analyze the properties of a randomly seeded dune field with known properties (initial distribution of sizes, density) in the absence of an incoming sand flux between sand dune. We observed that there exist a stationnary regime in terms of the distribution of dunes as well as the emergence of an internal structure within the corridor, mimicking what is observed in nature. Our numerical results evidenced the role of collisions in stabilizing the corridor. Importantly the size of dunes within a cluster is function of the local dune density. Clusters with high dune density are made of small dunes, whereas clusters with low dune density are made of larger dunes. This is a direct evidence that the collisions between dunes is a key process that sets the size of dunes within a corridor.

See the gallery, visit the dedicated web presentation or the articles that we published on dune collisions. You can also have a look at a recent press article . The website from Pieter Vermeesch also shows evidence of collisions of dunes in real deseart viewed from space.

[This work has been started during my PhD thesis and continued recently in collaboration with M. Génois during his PhD (MSC) and with G. Grégoire (MSC) and S. Courrech du Pont (MSC).]

../Linear Dunes/

Formation and stability of transverse and longitudinal transverse sand dunes, Geology, 2010.

Sand dunes can adopt the shape of long ridges which orient either transversally or longitudinally to the average wind direction. Such shapes form primarily when the wind direction alternates between two dominant orientations. Given the difficulty of field experimentation - such dunes are several kilometer long - not much was known on their dynamics of formation and their stability regarding a change of the wind regime. We improved our lab experimental setup to reproduce the conditions of a wind that periodically switch from one direction to another. We sucessfully formed subaqueous transverse and linear dunes. Interestingly, transverse dunes were fundamentally unstable. This was not the case for longitudinal dunes. We further showed that linear dunes can grow from an isolated patch of sand by the progressive extension of its tips. This extension mechanism requires that the angle between the two dominant wind direction is larger than 90°. Otherwise, the initial sand pile transform into a barchan. This illustrates that the two attractive shapes in a bimodal wind regime are the barchan dune and the longitudinal ones. We are now studying the properties of linear dunes, and in particular their rate of migration and extension.

You can fin here a press release from the CNRS , and in Lacroix (french magazine).

This work was done during the PhD of E. Reffet and published primarily in Geology. [In collaboration with S. Courrech du Pont, S. Douady].

../Polygonal vortices/

A triangular shape precessing around the rotation axis. The water tank is typically one meter large.

It's well known that the surface of a liquid in a rotating cylinder will adopt a paraboloid like shape. The level of water is higher near the edge than in the center of the cylinder. If the rotation speed becomes too large, then there will a dry circular area in the center of the cylinder, all the water being pushed by centrifugal forces on the cylinder walls. This is actually not always the case. We showed that in conditions where only the bottom of the cylinder is rotating and put the fluid into motion, the center dry area can adopt polygonal shapes. We observed triangles, squares, pentagon, hexagon ... Such polygons are not static and rotate at a slower speed than the bottom of the cylinder. Interestingly, such polygonal structure were also recently observed at very large scale in atmospheric flows on earth and mars. We show that the circular state and the unstable manifold connecting it with the polygon solution are universal in the sense that very different initial conditions lead to the same circular state and unstable manifold. We further measured the surface flows and discussed how such shapes can be modeled by a set of three rotating point vortices. Importantly, if the surface flow is blocked, by the addition of floating particles, the polygonal shapes are lost and ony the circular, symetrical case can be observed. This demonstrates the importance of surface flows in determining the shape of the free surface.

These experiments were started during a short stay in T. Bohr's laboratory at the Danish Technical University. They have been then improved and developped by Tomas's team. More details can be found in our two publications "Polygons on a rotating fluid surface", PRL, 2006 and "Polygon formation and surface flow on a rotating fluid surface", JFM, 2011.

Lab 513