Meeting report – New York Symposium on Quantitative Biology of the Cell

In the era of ‘big data’, diverse groups across the Greater New York City area aim to interrogate cellular functions at different scales – from the movements of single molecules to orchestrated multicellular development – using varied quantitative approaches. However, in order to better probe how cells work, a coordinated local meeting was necessary to bring together quantitative understanding at different levels. Co-organized by graduate students Chenshu Liu and Carmen Taveras (Columbia University, New York, NY), and postdoctoral fellow Anita Kulukian (The Rockefeller University, New York, NY), the New York Symposium on Quantitative Biology of the Cell (NYQBoC) created an otherwise rare occasion for interaction among local scientists with different expertise but who share a common interest in the quantitative interpretation of the cellular system. The symposium brought together 180 scientists from 24 institutions, including 14 speakers, to discuss the latest advances over the spectrum of quantitative biology of the cell. Highlighting some of the most exciting developments in various fields, the symposium was organized into three sessions with focuses that are interdependent upon one another – quantitative imaging over space and time, quantitative engineering from molecules to cells, and quantitative modeling of complex cellular and multicellular systems (Fig. 1).

Ever since the invention of the microscope by Antonie van Leeuwenhoek, biologists have been obsessed with observing cells in action. To date, with the advent of ever more sophisticated microscopy and image analysis algorithms, cells have started to reveal some of the deepest secrets of their lives, in ways unimaginable even a few years ago. Jennifer Zallen (Howard Hughes Medical Institute and Sloan Kettering Institute, New York, NY) is particularly interested in deciphering the positional code underlying collective cell rearrangement, a process that is essential to build the head-to-tail body axis during embryogenesis. Using live imaging and automated image analysis, Zallen and colleagues are able to track the motion of hundreds of cells at the surface of a developing Drosophila embryo, where epithelial cells form dynamic ‘rosette’ structures (Blankenship et al., 2006). Nearby cells are brought together, first converging along the vertical axis and then resolving along the horizontal axis, so contributing to the extension of body axis. Although local polarization and contractile forces can explain cell intercalation, the long-range coordination of these rearrangements requires global spatial cues. In the fly embryo, the spatial cues are based on contact-dependent signals, which require the pair-rule transcription factors Eve and Runt. To understand the downstream effectors of transcriptional inputs from striped patterns of Eve and Runt activity, Zallen and colleagues performed an RNA-seq screen, and found three candidate targets of Eve and Runt – Toll-2, Toll-6 and Toll-8 – which are also expressed in overlapping transverse stripes. Using genome engineering, Zallen and colleagues discovered that Toll-2, Toll-6 and Toll-8 work in combination to guide polarized cell rearrangement during concerted elongation, and showed that ectopic Toll receptor expression is sufficient to cause polarized contractility (Paré et al., 2014). Powered by quantitative imaging and computational image analysis, the Zallen lab has thus identified the missing piece that links positional information to the cell movements that produce body axis extension.

Moving from multicellular tissue to single cell dynamics, Tarun Kapoor (The Rockefeller University, New York, NY) presented ongoing work from his lab dissecting the mechanisms that build the central spindle in anaphase mammalian cells. One outstanding question in cell biology today is how micron-sized subcellular structures can assemble from nanometer-sized ‘building blocks’, such as motors and regulatory proteins. Over the years, using a powerful in vitro reconstitution approach, the Kapoor lab has discovered that the non-motor microtubule-binding protein PRC1 can tag microtubule plus-ends, with the size of the tag scaling with microtubule length (Subramanian et al., 2013). Enabled by the latest progress in light sheet microscopy, Kapoor and colleagues directly tested their hypotheses inside living cells by imaging at extremely high temporal resolution (hundreds of images per second) over the entire three-dimensional (3D) stack with minimal photo-toxicity. With this powerful tool, they have been able to examine transient and dynamic events in unprecedented detail in 3D space, linking in vitro data to the in vivo understanding of the regulated formation of antiparallel microtubule arrays that are essential for cytokinesis.

Remaining at the single-cell level, Jeeyun Chung (Yale University, New Haven, CT) spoke about her work studying membrane lipid homeostasis using total internal reflection fluorescence microscopy (TIRF). It has been known that the majority of lipids are manufactured at the endoplasmic reticulum (ER); however, it remains unclear how the plasma membrane and various membranous organelles can have different phospholipid compositions. Chung and her colleagues have identified members of a protein family – the oxysterol-binding protein related proteins (ORPs), as essential regulators of membrane heterogeneity. By using TIRF microscopy to look at specific focal planes for specific junctions between membranes, ORP5 and ORP8 (also known as OSBPL5 and OSBPL8, respectively) were found to be enriched at the contact sites between the ER and the plasma membrane. ORP5 and ORP8 tether the contact sites to promote the exchange of phosphatidylinositol 4-phosphate and phosphatidylserine between the ER and the plasma membrane, thus providing a mechanism to deliver newly synthesized phosphatidylserine to the plasma membrane, where this phospholipid is selectively enriched (Chung et al., 2015). Again, in combination with gain- and loss-of-function analyses, quantitative imaging generated insights into the molecular underpinnings of cells.

Besides allowing us to simply peek into cells or tissues, when some engineering principles are applied, a microscope can greatly extend our ability to actively modulate our subjects of interest and to ask questions that are otherwise difficult to address. Five speakers modulated both the intensive and extensive properties of cellular systems to illuminate some of their most deeply buried mechanisms. Temperature directly influences the thermodynamics and kinetics of biochemical reactions, and thus can become a quantitatively tunable probe when appropriately used. Using fast-acting temperature-sensitive mutants of Caenorhabditis elegans embryos, Julie Canman (Columbia University, New York, NY) presented recent work from her laboratory studying the relationship between cell polarity and cytokinesis during asymmetric cell division. In the C. elegans single-cell zygote, anterior–posterior polarity is established just after fertilization as different PAR proteins target to opposite sides of the cortex. Using a fluid device called ‘Therminator’ that allows rapid shifting of sample temperature (within 17 s) while simultaneously imaging cell division at high spatiotemporal resolutions (Bioptechs; Davies et al., 2014), Canman and colleagues made an unexpected finding when they mapped the precise time window required for formin- and/or myosin-II based cytokinesis. At semipermissive temperatures, which support successful cytokinesis in temperature-sensitive formin and myosin-II single mutants, a temperature-sensitive formin and myosin-II double mutant displayed a synthetic cytokinesis failure defect. Strikingly, this defect only happens when temperature is shifted during polarity establishment and maintenance, well before cytokinesis occurs. Combining RNA interference and temperature-sensitive mutants, Canman and colleagues further demonstrated that PAR-based cortical polarity is required for ordinary cytokinesis in the presence of a weakened actomyosin contractile ring. Remarkably, anterior–posterior polarity maintains proper F-actin levels at the contractile ring by promoting the anterior cortical localization of anillin and septin – two actomyosin-binding and crosslinking proteins that unexpectedly inhibit F-actin accumulation at the ring (Jordan et al., 2016). With seminal temperature-sensitive mutants and the Therminator, Canman and colleagues have addressed a longstanding mystery in the field by providing direct genetic evidence that cortical polarity actively regulates the contractile machinery of cytokinesis during asymmetric cell division.

A common daily phenomenon, especially during January in NYC, is the observation that as temperature drops, water vapor condenses into liquid droplets, which can further turn into ice (at freezing point). Similar phase-transition phenomena happen inside the cell and have recently emerged as a fundamental mechanism for intracellular organizations, including the assembly of nonmembrane-bound organelles. Often containing RNAs and proteins that have multivalent interaction domains and low-complexity sequences, the membrane-less organelles exist as liquid-like droplets and participate in a variety of physiological processes. Physiological phase separation takes the form of neuronal granules and germ granules in the cytoplasm, or Cajal bodies and nucleoli in the nucleoplasm, whereas aberrant phase separation from liquid droplet into solid aggregate might be associated with amyloid pathogenesis and amyotrophic lateral sclerosis (Weber and Brangwynne, 2012). Focused on the question ‘what are the molecular interactions that drive phase separation?’, Shana Elbaum-Garfinkle (Princeton University, Princeton, NJ) presented her exciting work on the role of C. elegans protein LAF-1 in the phase separation of P-granules. Driven by her observation that purified LAF-1 protein turns cloudy when placed on ice, Elbaum-Garfinkle and colleagues found that LAF-1, a RNA helicase, phase separates into liquid droplet in vitro. Using microrheology (a technique to measure properties of a medium, such as viscosity, based on tracer trajectories in that medium) and quantitative imaging, they have precisely measured the viscoelasticity of LAF-1 droplets and revealed properties consistent with pure viscous materials. The droplet's viscosity and the molecular dynamics within the droplet are highly tunable by modulating environmental conditions, such as salt and RNA concentrations. Additionally, the N-terminal arginine–glycine-rich intrinsically disordered domain of LAF-1 is both necessary and sufficient to drive liquid phase separation (Elbaum-Garfinkle et al., 2015). Collectively, Elbaum-Garfinkle and colleagues have provided mechanistic insights into how molecular-level interactions can give rise to dynamic nonmembrane-bound organelles with tunable mesoscopic properties (properties at intermediate length scale ranging from the size of molecules to microns) that can be functionally regulated by a number of biological parameters during normal development or aging.

Switching from intensive properties to extensive properties, Arne Gennerich (Albert Einstein College of Medicine, New York, NY) presented recent findings from his lab on the single-molecule study of force generation by the cytoplasmic dynein motor (Nicholas et al., 2015). Intrigued by the distinct motility and force production of dynein molecules measured in the yeast Saccharomyces cerevisiae and in mammals, Gennerich and colleagues discovered that the observed differences can be attributed to a previously unexplored C-terminal motor element, the ‘CT-Cap’, which is not present in yeast. Using optical tweezers to measure force and motion, they showed that removal of this element from mammalian dynein leads to an increase in force production and processivity, which resembles that of yeast dynein. At the single-molecule level, this new finding that the CT-Cap has a role in the intra-molecular regulation of dynein motor motility reveals an important molecular distinction between dynein from mammals and that from yeast.

Focusing on how positional information is processed in migrating cells, Gregg Gundersen (Columbia University, New York, NY) spoke about ongoing work from his lab on the role of nuclear–cytoskeletal coupling and mechanics during mammalian cell polarization. In wounded monolayers, fibroblasts start their polarized migration with centrosome reorientation and rearward nuclear movement, which requires a linear array of nuclear membrane proteins associated with dorsal actin cables, the transmembrane actin-associated nuclear (TAN) lines (Luxton et al., 2010). The formation and velocity of a TAN line is highly correlated with nuclear movement, and disrupting TAN line components impairs ordinary nuclear movement and centrosome reorientation (Kutscheidt et al., 2014). Despite the fundamental importance of TAN lines in nuclear movement, the role of nuclear positioning itself remains unclear. To address this question, Ruijun Zhu in the Gundersen lab developed a system that applies quantitative centrifugal force to live fibroblasts on both sides of the wound. With nuclear positions actively altered by centrifugation, the specific questions that now can be asked include: (i) how is nuclear position changed as a function of centrifugal force, (ii) does nuclear positioning on two sides of the wound respond in the same way after centrifugation, (iii) what is the molecular mechanism governing the cell's response to the force-induced change in nuclear positioning and, (iv) what is the role of nuclear positioning during polarized cell migration? Taken together, by combining quantitative imaging and meticulous biochemistry, Zhu and Gundersen have unraveled a novel nuclear-cytoskeletal coupling mechanism essential for nuclear movement during polarization.

Addressing positional properties within the nucleus, R. Ileng Kumaran (Cold Spring Harbor Laboratory, New York, NY) presented his elegantly engineered inducible gene-repositioning system in living mammalian cells, which enables quantitative dissection of fundamental genome functions. Mounting evidence indicates that the 3D-nuclear organization of mammalian cells plays an important role in gene regulation. To directly test the consequence of altering the physical location of a gene inside the nucleus, Kumaran targeted a multi-copy gene expression reporter locus to the nuclear lamina and periphery in living cells. 3D imaging demonstrated the recruitment of gene expression machinery components and the production of nascent transcripts at the repositioned-target locus site. Interestingly, the kinetics of transcription at the nuclear lamina and an internal region were comparable (Kumaran and Spector, 2008). This observation indicates the existence of transcriptionally permissive micro-domains in the nuclear lamina, a region primarily considered as a repressive domain (Kumaran et al., 2008). Moreover, with the same tools, Kumaran and colleagues quantitatively probed the dynamics of DNA double-strand break repair pathways in order to gain mechanistic insights into pathway choice, improving gene editing and genome stability.

In addition to quantitative experimental measurements, quantitative modeling of complex biological systems is a powerful tool to bridge our knowledge of microscopic molecular processes with macroscopic observations. Three speakers of this session were interested in using a combination of experiments and theory to study how the cytoskeleton and molecular motors self-organize into active structures that are dynamic and motile. Alex Mogilner (New York University, New York, NY) talked about fascinating work from his lab on the design principles of actin treadmilling. As a major driver of cell migration, the turnover (treadmilling) of actin has been studied in great biochemical detail, but a quantitative understanding of this phenomenon remains incomplete. With a simple geometry, the fast-crawling fish epithelial keratocyte and its lamellipodial fragments provide an elegant biophysical system that is amenable for both microscopic measurement and mathematical modeling. To test the hypothesis that disassembled G-actin undergoes recycling in order to be reassembled by global diffusion in the fragments, a simple 1D model was built with the distribution of different forms of actin being measured in lamellipodial fragments using quantitative imaging (Ofer et al., 2011). Remarkably, Mogilner and colleagues have discovered that more than half of actin in the fragments does not participate in the filamentous actin (F-actin) network that undergoes rapid turnover, but instead constitutes a diffusing fraction comprising monomers and oligomers, most of which is not available for polymerization. Moreover, numerical solutions to their model, which assumes global globular actin (G-actin) recycling, predict that diffusible actin density is almost uniform over the fragment, in excellent agreement with experimental measurements. Collectively, modeling suggests that the ‘global’ turnover and treadmilling mechanism allows for diffusion in order to recycle actin effectively, resulting in steady-state cell migration that is poised for rapid, focused acceleration.

Besides the ability to self-assemble from subunits into polymers, another key feature of the cytoskeleton is the interplay between polymer filaments and motor proteins. This liaison and its dynamic regulation shape the higher-order behaviors of a given biological system – a recurring topic throughout the symposium (Fig. 2). Changing gears from actin, Sebastian Fürthauer (New York University, New York, NY, and Harvard University, Cambridge, MA) described a novel phenomenon of the microtubule networks. Using Xenopus egg extracts, Fürthauer and colleagues have discovered that stable microtubule networks can undergo spontaneous rapid bulk contraction (Foster et al., 2015). They proposed that dynein motors drive the self-organization of microtubules into asters by clustering microtubule minus-ends and that the motors generate contractile stresses. Their experimental data fit well with an active fluid theory in many aspects, including a strong dependence of the time scale of contraction on the initial geometry of the network and on dynein inhibition. Their results have demonstrated that motor-driven clustering of filament ends could provide a generic mechanism for contraction, which can be important to understand the self-organization of meiotic spindles in vivo.

In his keynote address, Jonathon Howard (Yale University, New Haven, CT) emphasized the importance of combining theory with experiments in quantitative biology, sometimes by reducing the problem to simpler systems (Howard, 2014). Cytoskeleton systems that include motors cause deformations in surrounding structures (e.g. in the extracellular matrix surrounding a metastasizing cancer cell) when they generate force; these deformations can provide feedback on motor and cytoskeletal activity. Despite many measurements, an integrated quantitative understanding remains challenging owing to the intrinsic complexity of the mechanical properties of cells and tissues. Using a much simpler system – the photosynthetic algae Chlamydomonas (Geyer et al., 2013) – Howard showed how mechanical feedback coordinates thousands of dynein motors, which act collectively to shape the beat of axoneme, the conserved mechanical structure inside cilia and flagella (Riedel-Kruse et al., 2007). Dyneins produce forces that deform the axoneme; they then sense the stresses and strains, resulting in spatial and temporal synchronization. Together with his colleagues, Howard demonstrated that the waveform of the Chlamydomonas axonemes can be decomposed into a static circular shape and a dynamic traveling wave that, when superimposed, creates the breaststroke-like beat. He further demonstrated that the dynamic beat is consistent with curvature-controlled dynein activity that argues against sliding and normal force control, thereby addressing a longstanding debate in the field.

Computational modeling not only advances the understanding of intra-cellular processes, but also extends the ability to comprehend complex multicellular systems. Bacteria usually live in a multicellular community and engage in many cooperative behaviors. One outstanding question is how to maintain consistent cooperativity in the face of non-cooperators. In a population of Pseudomonas aeruginosa bacteria, each individual can produce rhamnolipids that are shared by the whole population to facilitate growth. Using the expression of the rhamnolipid-synthesis operon rhlAB in P. aeruginosa as a measure of cooperativity between cells, Kerry Boyle (Sloan Kettering Institute, New York, NY) showed that when the environment is nutrient rich, cooperativity is largely determined by population density. However, when nutrients are limited, cooperativity is highly nutrient dependent (Boyle et al., 2015). These findings are quantitatively characterized in a 1D model that relates population density with varying nutrient levels, the predictions of which are in close agreement with further experimental measurements. Because of the swarming nature of P. aeruginosa populations, future work involving 2D modeling will bring additional insights into the regulatory dynamics observed on a petri dish. Focused on another aspect of multicellular populations, Yevgeniy Plavskin (New York University, New York, NY) spoke about the statistical properties of spontaneous mutations that constitute the raw material for evolution. Because most current methods used to study the genetic basis of trait variation can only reveal the effects of polymorphisms that have already been filtered by selection, Plavskin and colleagues took a bottom-up approach using the well-sequenced mutation-accumulation strains of Saccharomyces cerevisiae in order to ask how many mutations affect phenotype and how strong these effects are. Enabled by a high-throughput imaging platform that automatically segments and tracks individual yeast colonies, Plavskin and colleagues have been able to precisely measure growth rates in tens of thousands of yeast microcolonies. Based on these rates, they constructed a statistical model to infer the distribution of the effect size of mutations. Modeling and maximum likelihood estimation revealed the mean effect size, the shape of mutational effect distribution and the percentages of mutation with neutral, positive or negative effects. Collectively, these findings have provided a new perspective to investigate the effects of nonfiltered spontaneous mutations that contribute to the process of evolution.

Recent years have witnessed profound progress in the transition of cell biology toward a more quantitative science. As revealed during the symposium, cell biologists have generated fresh insights into challenging and longstanding questions, and have formulated new trailblazing ideas when equipped with quantitative data from elegantly designed experiments. An outstanding question in biology remains as to how the dynamic structures within and between cells are organized; in particular, how the size, shape and behaviors of these structures can be ordered at micron-length scales by using nanometer-scaled building blocks with either ordered structure or of intrinsically disordered nature. Undoubtedly, cutting-edge engineering and imaging approaches will continue to elucidate how organization principles contribute to emergent material and mechanical properties while giving rise to cellular functions. Of equal importance is the understanding of how molecular interactions of proteins and nucleic acids can ensure robust yet adaptive information propagation over time (e.g. long-term genome stability, evolution) and space (e.g. anterior–posterior polarity of single-cell zygotes, head-to-tail axis in Drosophila embryo or quorum sensing in bacterial community). Lastly, NYQBoC highlighted the unprecedented role of computational tools in handling ‘big data’ today – ranging from automated high-throughput screens to machine-learning-based extraction of morphology from complex imaging dataset. Because data alone does not speak for itself, the symposium has underscored the essential role of theories and models in quantitative cell biology. Building mathematical models, both deterministic and probabilistic, is not only a key to recapitulate and explain experimental findings, but also a crucial step in guiding future experiments by generating testable predictions. Confronted with complex processes, computational simulation and modeling help us to develop better intuition, extending our ability to build an integrative and quantitative comprehension of the systems behavior at molecular, cellular or multicellular scales.

We are grateful to all speakers and poster presenters for sharing their inspiring work. We apologize for not including highlights from the poster session owing to space limitations. NYQBoC was generously supported by ASCB, the Office of Graduate Affairs and the Department of Pathology and Cell Biology at Columbia University Medical Center, The Rockefeller University and EMD Millipore. The authors thank Julien Berro (Yale University) for critically reviewing the manuscript.