Elliot Meyerowitz
Investigator, HHMI-GBMF
George W. Beadle Professor Biology
California Institute of Technology

Research interests. Our laboratory has the goal of understanding the mechanisms of plant development, using both experimental and computational methods to test hypotheses. We concentrate on the shoot apical meristem and its derivative structures (primarily flowers), because this meristem is responsible for the development of the entire above-ground part of the mature plant, and utilizes a number of different pattern-forming processes that are as yet poorly understood. Our experimental organism is predominantly Arabidopsis thaliana, because of the ease with which genetic and molecular biological studies can be done using that model system. We also use other plant species in the laboratory when they offer an experimental advantage.

Peptide signaling in the shoot apical meristem. The Arabidopsis shoot apical meristem (SAM, Fig. 1) is a hemispherical collection of several hundred identical-appearing cells at the tip of each shoot, which serves as the stem cell population for shoot, leaf and flower development. Traditionally, the SAM is considered to be divided into three regions. The central zone is at the tip of the SAM, and comprises the ultimate (pluripotent) stem cells. Surrounding it is the peripheral zone, the collection of multipotent stem cells that derives from the central zone, and from which the primordia of leaves and flowers form. Beneath the central zone and peripheral zone is the rib meristem, from which the cells of the stem, including the vasculature, form. Starting some years ago, our laboratory identified genes that control the cell numbers in the central and rib regions (Clark et al. 1993; 1995; 1997). Molecular cloning of the genes showed that the central zone synthesizes a peptide ligand, CLAVATA3, that is secreted from the central zone cells, and activates a transmembrane receptor kinase, CLAVATA1, in rib meristem cells, eventually causing reduction in the size of the rib meristem (Fletcher et al. 1999; Brand et al. 2000). A signal from the rib meristem back to the central zone regulates central zone size as well, such that more signal gives a larger set of central zone cells. This signaling network controls the size of the meristem. Earlier work has characterized this network in a variety of ways. One notable finding (Reddy and Meyerowitz, 2005) was that live imaging of the central zone while inactivating the CLAVATA3 gene by use of an inducible RNAi construct caused rapid expansion of the central zone by dedifferentiation of peripheral zone cells, normally multipotent stem cells, to central zone cells, which are pluripotent. This unusual control of the stem cell niche explains much about the behavior of SAM cells and the control of their number, and forms the basis for present-day hypotheses of SAM signaling and control. Our current work on the CLAVATA circuit has revealed that the rib meristem receptor, CLAVATA1 (a leucine-rich repeat transmembrane serine-threonine kinase, which sits at the top of the signaling cascade activated by the CLAVATA3 peptide, which is its ligand), is regulated by ligand-induced receptor endocytosis and degradation. A large set of experiments just finished (Nimchuk et al. 2010) and in progress is defining the overall effect of this ligand-controlled degradative mechanism on the dynamics of CLAVATA signaling, and is also defining the cellular components that bind to CLAVATA1 and cause it to be internalized. Future work in this area will include continued analysis of the cell biological components responsible for CLAVATA1 internalization and degradation, and of the downstream components in the CLAVATA signaling pathway. We will also devise detailed computational models of cell-cell interactions in the shoot apical meristem, and iterate between models and experiment, to refine the models to a point where they represent the processes that control this stem cell niche.

Figure 1: Arabidopsis shoot apical meristem. Left, scanning electron micrograph of the shoot apical meristem as it produces flowers on its flank (see Smyth et al. 1990). Center, a frame from a live-imaging movie in which plasma membranes have YFP inserted (Reddy et al., 2004). Right, shoot apical meristem of a triply transgenic plant showing REVOLUTA (red), PINFORMED 1 (blue) and KANADI (green, see Heisler et al., 2005).

Cytokinin control of gene expression in the shoot apical meristem. One major question raised by the CLAVATA signaling system is how the gene expression domains are defined and maintained - that is, how it is that CLAVATA3 is expressed in (and defines) the cells of the central zone, and CLAVATA1 in the cells of the rib meristem. One recent discovery is that the plant hormone cytokinin plays a role in this (Gordon et al. 2009), such that increased cytokinin concentration in the SAM leads to incorporation of peripheral zone cells, ordinarily surrounding the rib meristem, into the rib meristem. This shows that another boundary (like that between central and peripheral zones) is dynamic, and that cells at the border can differentiate to the fate characteristic of either cell type, with the balance between cell types regulated by a diffusible factor. It also opens a set of new questions regarding which of the cytokinin receptors is responsible (we have recently found that mutations in the three receptor genes have different effects on gene expression in the SAM), and regarding the regulation of cytokinin concentration and gradients within the SAM. Future work will be directed to discovering the characteristics of this system, and to developing computational models for the translation of a dynamic cytokinin gradient and a dynamic pattern of hormone receptor activities into SAM gene regulation domains.

The origin and control of phyllotactic pattern, and physical signaling in the SAM. Another set of problems posed by the actions of the SAM is its formation of leaf and flower primordia. In addition to describing in mechanistic detail the formation of new floral primordia, which has led to a series of surprising observations and testable hypotheses for the sequential activation of the genes that define floral anlage and their adaxial and abaxial patterns of gene expression (Heisler et al. 2005), we have analyzed the pattern in which new leaf and floral primordia arise in the meristematic peripheral zone. This origin of this pattern, the phyllotactic pattern, is a question of long standing in plant biology, with the first mathematical approaches being taken almost 150 years ago (and recognition of the regular phyllotactic pattern over 2,000 years ago). In Arabidopsis and in most other plants, the phyllotactic pattern is a spiral with successive primordia appearing approximately 140 degrees radially from each other. By a combination of live imaging (Heisler et al. 2005, Michniewicz et al. 2007) and computational modeling (Jönsson et al. 2006) we have been able to provide a detailed mechanistic and causal hypothesis for the phyllotactic pattern, based on feedback in the transport of the plant hormone auxin with its efflux carrier, the PIN1 protein (Fig. 2). Each premise of the model is explicit and tested, and together they provide a solution to the old problem of the formation of this characteristic plant pattern. The solution is in fact a new class of developmental model - a regulated transport model, which adds to the two earlier-known classes of models for the generation of pattern by groups of cells (reaction-diffusion models, and mutual inhibition models).

Figure 2: Phyllotaxis in Arabidopsis. Left: confocal microscope image showing the pattern of floral primordia surrounding the shoot meristem (green, plasma membrane, red, chloroplast autofluorescence, Roeder et al., 2010). Middle: level and position of the PINFORMED 1 auxin transporter protein, whose gene is auxin-induced, and whose position allows inference of the direction of auxin transport (arrows, see Heisler et al., 2005). Right: from a computational model of auxin concentration dynamics in the meristem (see Jönsson et al., 2006).

One premise of the phyllotaxis model raised a new set of possibilities for plant development - we (and others) have been able to demonstrate that the PIN1 protein at the plasma membrane of each epidermal cell of the SAM is found adjacent to the neighboring cell with the highest auxin concentration (it is this feedback that gives the meristem its dynamic pattern-forming capabilities). One possible mechanism for a cell to sense the auxin concentration of its neighbor is physical - auxin causes meristem cells to expand by relaxing their cell walls, and the stress that such expansion creates on their neighbors could be a signal for cellular polarization, and for PIN1 relocalization. We recently tested this possibility and found that SAM cells can indeed sense stress, and respond in a vectorial fashion to the principle direction of stress by reorganizing their microtubule cytoskeletons (Hamant et al. 2008). Confirmation of the hypothesis that physical stress is responsible for PIN1 location, and therefore plays a central role in the signaling and feedback system that creates auxin flow across the meristem, was just published as Heisler et al., 2010 (Fig. 3). The future possibilities for this work are extensive, both to learn the mechanism by which stress is directionally sensed, and to follow up on the implications of the cytoskeletal rearrangement motivated by changing stress patterns. As microtubules define the pattern of cellulose deposition in cell walls, and therefore the reinforcement of plant cells against stress, and also are thought to play a central role in establishing the plane of plant cell division, we appear to be on the brink of an explanatory model for cell wall deposition, organization of cell division, and their feedbacks, which could provide a predictive model for meristem cell expansion, division, and differentiation. The computational as well as experimental challenges (the need for finite element model descriptions of dynamic tissues, as well as novel methods for live imaging of multiple dynamic cell parts) are likely to occupy much of our laboratory-F¢s effort, and much effort-A on the part of our collaborators, in the near future.

Figure 3: Mechanical stress controls microtubule orientation and PIN1 auxin transporter localization. Left, finite element model of principle stress directions in shoot meristem after 1-cell laser ablation (Hamant et al., 2008). Center, experiment confirming model prediction in left panel (see Hamant et al., 2008). Right, PINFORMED 1 auxin transporter aligns parallel to microtubules in meristem epidermal cells (Heisler et al., 2010).

Flower development and gene regulatory networks. Once new floral primordia are made in the meristematic peripheral zone, processes of pattern formation define the positions of the early bract (it is typically prevented from later development, and therefore absent in mature flowers) and of the flower itself, then define the radial domains that later will be occupied by the different types of floral organs - sepals, petals, stamens, and carpels. The study of these processes were among the problems for which our laboratory originally developed the molecular genetics of Arabidopsis in the early 1980s, and this study has occupied a part of the laboratory ever since. Earlier work included the derivation of the ABC model of flower development, which specifies how a set of transcription factors acts in specification of floral organ identity (Meyerowitz et al., 1991; Bowman et al., 1991, Coen and Meyerowitz, 1991), and more recently we have gained extensive information on the genes regulated by the ABC transcription factors, first by microarray analysis, and then by next-generation RNA sequencing and ChIP-seq (e.g. Kaufmann et al., 2010, Jiao and Meyerowitz, 2010). We have recently adapted methods for cell-type specific RNA profiling, and have used them to assess the staged regulation of genes downstream of the ABC genes in different cell types, which is leading toward a gene regulatory network description of the events of floral development (e.g. Ito et al., 2007; Jiao and Meyerowitz, 2010). Future work in this area includes continued development of new methods for cell type-specific analysis of transcriptional and DNA-binding events in developing flowers and in shoot apical meristems, to advance the work described above, as well as continued genetic analysis of the genes that occupy key regulatory nodes in the network. Cell-type specific RNA-seq will also be used on mutant, as well as wild-type plants (as now), to establish the changes in gene regulatory networks caused by alterations in the level of expression of key regulatory genes. We also plan new live-imaging studies with mutant and transgenic plants to develop models for the specification of patterns of cell types in developing sepals (e.g. Roeder et al., 2010; 2012), which will relate the work on cell division pattern control in the shoot apical meristem (above) to the patterned cell divisions that give floral organs their shape and size.

Plant regeneration. One additional property of shoot apical meristems is their ability to form de novo, both during embryogenesis and leaf growth, but even more remarkably, during regeneration in tissue culture. How a mass of cultured cells, callus, can give rise to highly organized and functioning shoot (and root) meristems is a mystery. One set of experiments in the laboratory is directed toward a mechanistic understanding of the processes of pattern formation in shoot regeneration. These studies are taking two directions. The first is to understand the nature of callus tissue, which according to the literature of plant tissue culture is an undifferentiated set of cells that forms and grows in response to hormonal cues, with the relevant hormones being auxin and cytokinins. We have recently shown that the literature is incorrect, and that Arabidopsis callus tissue, whether formed from roots, cotyledons, or petals, resembles lateral root primordia in its spatial patterns of gene expression, and in the genetic requirements for its formation (Sugimoto et al. 2010). We have also started the analysis of the earliest events in SAM regeneration from growing callus tissue - while we no longer consider this to be pattern formation from undifferentiated or dedifferentiated cells, it is still, and remarkably, an example of transdifferentiation of root into shoot cells - and we are observing the changes in hormone response, and gene activity as new meristems form, by live imaging, using laser scanning confocal microscopy (Gordon et al., 2007; 2009). These observations will lead in the future to new experiments in which new reporters are observed, and in which a variety of different genotypes of tissue are used, to learn the contribution of each of the key regulatory genes discovered. Future work will also include development of computational models of the cell-cell interactions and gene regulatory networks responsible for the formation of callus, and its transformation into shoot apical meristems.

Computational Morphodynamics. The sum of the experimental approaches we emphasize in our work - live imaging of gene function and of protein levels and subcellular locations, forming hypotheses that encompass the dynamic information gained by real-time imaging, and testing the hypotheses both by experiment, and by developing computational models where the same parameters tested in vivo can also be tested in silico - we call by the name Computational Morphodynamics (see Figs. 2, 3). In addition to our direct studies of plant development using the methods of Computational Morphodynamics, we also have written reviews (such as Chickarmane et al., 2010; Roeder et al. 2011; Cunha et al. 2012) and held workshops (such as the Computational Morphodynamics workshop at the Kavli Institute for Theoretical Physics at the University of California, Santa Barbara from August 23-September 23, 2009) to inspire similar approaches in plant science, and in other areas of developmental biology, such as animal development. Our lab with others have co-founded Caltech's Biological Network Modeling Center (BNMC, bnmc.caltech.edu), which provides assistance and advice on image processing to automatically or semiautomatically segment images such as laser scanning confocal microscope images of gene expression domains and cell division patterns; and to develop algorithms for tracking cells in time series images. The BNMC is also engaged in developing ways of facilitating computational modeling in biology, in particular by development of software such as Cellerator (www.cellerator.org), which generates ordinary differential equation models from arrow-based reaction networks, and then simulates and analyzes the function of the networks by numerical solutions to the equations. This type of software allows biologists without expertise in mathematical modeling to develop and test computational models, and therefore to refine their hypotheses and thinking about biological processes. Continued development of this software and its extension to two- and three-dimensional tissues (as in Cellzilla, an early-stage expansion of Cellerator) will be an important part of future work.

One other venture in computational morphodynamics is what we call the Computable Plant group (www.computableplant.org), an international collaboration of biologists, engineers, computer scientists and mathematicians that collaborates in many combinations of personnel to develop and apply methods of image processing and computational modeling to the understanding of development; many publications have resulted from this ongoing collaboration. Computable Plant group meetings are by videoconference every Wednesday morning (in California, Wednesday evening in Europe), and include groups in England (Henrik Jönsson, Sainsbury Laboratory, Cambridge), Sweden (Pawel Krupinski, University of Lund), Germany (Marcus Heisler, EMBL Heidelberg), and Switzerland (Gerarado Tauriello, ETH Zürich) in additional to our computation group at Caltech, the BNMC, and our close and long-time collaborators at the University of California, Irvine, led by Eric Mjolsness. It is through our collaboration with these computational scientists that we have been able to use mathematical models to advance our experimental work.

A final point. One could ask about the practical applications of the work that we have done, and plan to do. While the largest reason for performing the studies described is to gain a fundamental understanding of developmental mechanisms, the extraordinary importance of shoot apical meristems for human welfare is an additional motivation. At least 80% of human nutrition derives directly from shoot apical meristems - all of our grains, other fruits, and leaf crops are SAM products. A sad statistic from the World Health Organization is that more people die of starvation each year than of cancer - and furthermore, the victims of starvation, unlike those of cancer, are generally children. SAMs are also the source of much of the natural world - almost anywhere on land, the primary aspect of visible nature is the plants, and the visible parts are direct SAM derivatives. Our crops, building materials, many of our medicines, fibers, fabrics - the material basis of human life - come from the immediate developmental functions of the SAM. Thus understanding its function and development in mechanistic detail is an important practical, as well as theoretical, endeavor, and an endeavor with some urgency, as human welfare depends so directly on a fundamental understanding of plant growth.