Scientific statement

Prior Research

2013-2018 (PhD)

My PhD project led me to investigate on the ultrastructure of nanoscopic membrane-lined channels called plasmodesmata, which are embedded in the cell walls between plant cells, allowing them to communicate with each other. They establish a very special kind of communication, proper to the plant kingdom, called symplastic communication, by establishing a cytosolic and membrane continuum of the plasma membrane (PM) and the endoplasmic reticulum (ER) between the cells. While they are central to plants and actively studied, the relation between ultrastructure and function is not fully understood. When I started my PhD, no information was available on the 3-dimensional organization of these pores and the dynamics of their membranes.

The ambition of my PhD project was to pioneer a new perspective on plasmodesmata pores by providing a detailed understanding of how they were organised within the narrow space between the ER and the PM and how this was related to the control of cell-to-cell connectivity. The ultrastructural analysis of plasmodesmata was a formidable challenge given their nanoscopic size (they can be as narrow as 20 nm in diameter), location (embedded in the plant cell wall) and their dynamic nature. And because electron tomography had not been developed yet on the Bordeaux Imaging Centre (BIC), I took a great part in building this technic from the ground.          Using electron tomography of high pressure-frozen, near-native plasmodesmata, I provided for the first-time data on their 3D membrane architecture at an unprecedented resolution (see panel aside), forcing a reconsideration of the current model (https://www.nature.com/articles/nplants201782).


2018-2022 (Postdoc)

As a young microscopist experienced in 3D Electron microscopy on plastic section and given the lack of work pushing forward cryo-electron tomography (cryo-ET) in the plant biology field, the next relevant step towards near-native imaging was cryo-ET. Hence, I was appointed at Caltech (Pasadena, USA) in January 2018 as a joint post-doctoral scholar in two labs, the Meyerowitz lab, specialized in apical meristem development (plant biology), and the Jensen lab, specialized in cryo-ET method development. At the interface between these two labs, I started a project on the structural characterization cellulose synthesis.

The study of the arrangement of the cellulose fibers and their synthesis by the Cellulose Synthase Complexes in the plant cell wall by ECT

Because plant tissues are notoriously thick, cryo-EM requires thinning technics. Cryo-ET is only compatible with samples less than a micron thick at the most. Anything above requires thinning. For this aspect of my work, the white onion (Allium cepa) abaxial epidermal cell wall peels biological system was chosen. It is a common tissue used for the study of mechanical and structural properties of the plant cell wall and its thickness, around 8 microns thick is compatible with plunge-freezing and cryo-FIB milling. The latter allows the generation of ~100-200 nm thick lamellae, which are then transferred in the cryo-transmission electron microscope. I then had access to the whole depth of the cell wall. Such process allowed the high-resolution imaging, in near-native conditions, of the cellulose fibrils and other components within the cell wall. Combined with convolutional neural network based segmentation and data mining, I was able to describe how the plant cell lays out its successive cell wall layers over time.  This work won the 3rd place poster-award at the Southern California Cryo-EM symposium 2020 and the postdoctoral work award of the Microscopy and Microanalysis conference in July 2021. It was published in Current biology (https://www.sciencedirect.com/science/article/pii/S0960982222005930?via%3Dihub) and the protocol was published in Bioprotocol (https://en.bio-protocol.org/en/bpdetail?id=4559&type=0). The protocol paper made the cover of the issue.

The study of lipid droplet biogenesis in whole plant cells by cryo-ET

Despite the successful attempt at milling through onion cell walls, we did not have access to the full cells because the preparation methods involved stripping the cells away from the cell wall. My holy grail being looking inside whole plant cells, I started to investigate the type of cells that held the best chance to be compatible with plunge freezing and Gallium FIB milling. Pollen tube cells happen to be cylindrical cells 10 microns in diameter which is within the confines of plunge freezing and FIB milling. These cells emanate from the pollen grain and shuttle the male haploid nucleus from to the ovule deep in the carpel. I therefore set out to grow Arabidopsis pollen tubes from pollen grains, freeze them and FIB mill them. This allowed me to open 10-20um wide windows in different sections of the pollen tube showing intense vesicular trafficking. A subset of the data clearly displayed beautiful views of lipid droplet biogenesis so  I set out to characterize the life cycle of lipid droplets in pollen tubes. This project relied on extensive use of convolutional neural networks to exhaustively segment the tomograms in 3D in order to gather as much structural/morphological information as possible on vesicles, lipid droplets and other organelles in the cell. This work has yet to be published but a snapshot of it is available in a poster presented at the Microscopy and Microanalysis conference in August 2021 (https://www.researchgate.net/publication/373449087_Cryo-electron_tomography_methods_applied_to_plant_science).

The study of the cellulose synthesis and crystallization process in prokaryotes

It is thought that plants acquired the ability to make cellulose through early endosymbiosis events with cyanobacteria. Therefore, understanding the mechanism of cellulose synthesis in bacteria is a good proxy to gain insights into understanding this process in plants. Among several bacteria that produce cellulose as a component of their biofilms, only one genus can make crystalline cellulose just like in plants, Gluconacetobacter. I thus started to study the structure of this bacterium by cryo-ET to better understand the cellular determinants of the cellulose crystallization process. This also includes elucidating the structure of its cellulose synthase machinery, namely the Acetobacter Cellulose Synthase (ACS), a multimeric complex that spans the double bacterial membrane. This aspect of my work has been granted funds from the Center for Environmental Microbial Interactions of Caltech in October 2018. This project has allowed the discovery of a novel, unknown prokaryotic cytoskeletal element we named the cortical belt. The later piece of work has been published in Journal of Bacteriology in February 2021 and made the cover of volume 203, issue 3 (https://jb.asm.org/content/203/3.cover-expansion).

The study of the life cycle of predatory bacterium Bdellovibrio  bacteriovorus

My extensive experience in tomography, tomogram analysis and segmentation led me to collaborate heavily in this very fun project spearheaded by Mohammed Kapan, also postdoc in the Jensen lab at the time (now Assistant Professor at University of Chicago (https://profiles.uchicago.edu/profiles/display/36292832). Several members of the Jensen lab, including Mohammed, amassed several hundreds of tomograms of Bdellovibrio bacteriovorus for multiple projects, all archived on the internal database of the Jensen lab (a subset of it was made public: https://etdb.caltech.edu/). This peculiar bacterium latches on its bacterial victims, enters their periplasm and eats them alive from the inside is still unclear. Its life cycle is not well understood and the massive database we had allowed us to piece back together the life cycle, in a still image style. This led to several unique observations, updating the molecular mechanisms involved in different steps of the life cycle:

·        Hosts (victims) are detected by the use of pilli. Upon identification, the Bdellovibrio cell is brought closer to the host

·        Upon contact with host, the bacterium creates a polar attachment plaque.

·        Bdellovibrio’s flagellum is “reeled” inside the periplasm

·        The outer-membrane of the host is digested, an air-tight seal is created between Bdellovibrio and host cell and the former penetrates the periplasm of the latter

·        It will feed on the cytoplasm of the host cell until all cytoplasm is gone, divide and break loose, completing the life cycle.

This extensive work was published in Nature microbiology and made the cover of volume 8 issue 7 (https://www.nature.com/articles/s41564-023-01401-2).

Current Research

2022-present

I took on a cryo-EM lab manager position in the Howard Hughes Medical Institute (HHMI) Gonen lab at UCLA (Los Angeles). Aside from my managing duties, I have the pleasure to work on my own projects or collaborate on various lab projects involving microED methods development and membrane protein structures.

The amazing aspect of microED is how protein size is not limiting. Crystal size is, but with FIB milling, we can basically tackle any nano- to micro-sized crystals. This makes it the only EM technic that can solve structures of small proteins and small molecules, an in a record time!

Development of efficient fluorescence-guided FIB milling methodologies ("FIBucial" method)

In the Gonen lab, a lot of focus is put on solving structures of difficult membrane protein crystals that require special growth solvants and conditions. Namely, several require the use of a tooth-paste like substance called lipidic cubic phase (LCP). When LCP matrix loaded with crystals is loaded on a grid, it looks like amorphous piles resembling hills or mountains (sometimes several hundreds or nanometers high). It is impossible to see where the crystals are, and it can take a lot of trial and error to make a lamella right through the crystal.

We therefore use fluorescence guided plasma-FIB milling to target as best we can the crystals. While the X-Y correlation is quite straightforward and could be done by eye, the Z correlation, taking into account the different imaging perspectives of the SEM, FIB and our in-chamber fluorescence light microscope (iFLM), is impossible to do by eye.

When I arrived into the lab, the process of x,y,z correlation was not streamlined at all and the success rate was low. Additionally, the process took time, not because of milling time but correlation time. I thus set out to establish the best correlation pipeline to do the correlation on-the-fly, allowing for rapid targeting. Correlation methods often require landmarks, or fiducial markers that are electron-dense fluorescent spheres that can be spotted in all imaging modalities. Unfortunately, LCP is very hard to mix with aqueous fiducial marker mixes. I developed a method where the fiducial markers are directly “drilled” into the sample. I call them “FIBucials” and it allows quick x,y,z registration of the targets of interests allowing to focus the session time on the milling.

Although the streamlining of this method is still ongoing, it was used in two published manuscripts in Nature Communications (https://www.nature.com/articles/s41467-023-36733-4) and IUCrJ (https://journals.iucr.org/m/issues/2023/04/00/fq5021/index.html).
A focused paper on the FIBucial method is in preparation.

Development of a high-throughput microED data acquisition pipeline


There is a growing need in the lab to acquire large amounts of data on crystals. I took advantage of my experience operating the Krios and scripting with SerialEM to work on microED data acquisition automation. The goal being to choose which grid to acquire on, identify the grid squares of interest and from there, let the computer do all the detection and data acquisition. Currently, several of my python and SerialEM scripts are routinely used to acquire all grid atlases, detect lamellae and crystals on the grids, using in-house trained YOLOv8 models, designate large amounts of targets and run automatic data acquisition on all targets. All these tasks, when performed in a non-assisted, require very accurate eucentric and target recall. The large amounts of data now being acquired are then fed into a semi-automatic in-house data processing pipeline written by my colleague Mike Martynowycz (https://scholar.google.com/citations?user=9c1BCqMAAAAJ&hl=en). I find it very useful to have a plethora of users with different levels of expertise to test out these scripts. This drives me to keep these scripts as easy to use as possible, which is a great constraint to thrive for.

De-novo structure of Vasopressin-1 receptor (V1B) by micro-ED

I am actively collaborating in solving the structure of the human Vasopressin-1 B (V1b) receptor, a G-Protein Coupled Receptor (GPCR) central in the regulation of vasoconstriction, which controls blood pressure and osmotic homeostasis, which balances the body's water and electrolyte levels.

Mike Martynowycz and Anna Shiriaeva provided me with the grids with the LCP loaded with V1b crystals, and I performed the FIB milling, data acquisition and helped in data processing, applying the various tools Mike and I created to automate the process. Essentially V1b, along with A2A, has been a testing ground for fluorescence targeted plasma-FIB milling and energy-filtered data acquisition of microED datasets on a electron-counting detector. As of now, this work has been published as a preprint (https://www.biorxiv.org/content/10.1101/2023.07.05.547888v1.abstract).

De-novo structure of a Tetraspanine protein by microED

Our lab has a special interest for a ~20 kDa Tetraspanine (I cannot disclose the identity of the protein yet…), way too small to even hope looking at it with cryo-EM. Along with Mike and Anna, we are applying a similar strategy to V1B to solve its structure. The paper is being written up, so stay tuned in!