Simulating Crowded Cytoplasm Animating Molecular Biology by bestt571


More Info
                                                      described in the March 2010 issue of            the space of another molecule. The most
                 Simulating                           PLoS Computational Biology.                     complex scenario they ran included
             Crowded Cytoplasm                            “This is an attempt to build a virtu-       excluded volume, electrostatic interac-
             In biology textbooks, the carefully al lab, in which we can study various                tions, and favorable short-range
         rendered cross-section of an E. coli cell biological and biophysical processes as            hydrophobic interactions. The more
         often resembles a well-organized and they might occur inside the cell,” says                 complex simulations performed surpris-
         spacious apartment, with everything in Adrian Elcock, PhD, coauthor and                      ingly well when asked to predict molecu-
         its place and ample room for move- associate professor of biochemistry at                    lar behaviors, such as diffusion and stabil-
         ment. But a recent computational the University of Iowa.                                     ity, in the E. coli cytoplasm.
         recreation of the scene looks more like          The sea of floating proteins inside             The model was able to match exper-
         a Friday night dance floor, with mole- every cell is the background against                  imental observations of how quickly
         cules bumped up against one another in which many cellular reactions take place.             green fluorescent protein diffuses in the
         every direction. In addition to provid- Scientists realized years ago that the cyto-         E. coli cytoplasm. And it was able to
         ing a dramatic, qualitative description plasm is generally not an invisible player           predict the greater stability of the
         of the crowded cytoplasm, this first in those reactions. One of the best-stud-               unfolded state of the protein CRABP,
                                                                        ied examples is macro-        cellular retinoic acid binding protein,
                                                                        molecular crowding            over the folded state inside E. coli.
                                                                        (also called excluded         Although the presence of close neigh-
                                                                        volume effect). Having        bors (crowding) generally stabilizes a
                                                                        large neighbors on            large folded protein, the specific elec-
                                                                        every side changes a          trostatic and hydrophobic interactions
                                                                        protein’s effective con-      of unfolded CRABP with other cyto-
                                                                        centration and influ-         plasmic proteins counteract the crowd-
                                                                        ences its movement            ing effect.
                                                                        and ability to react.             “What this doesn’t mean,” Elcock
                                                                        A biological reaction         emphasizes, “is that crowding effects
                                                                        observed in dilute            are unimportant. It means that crowd-
                                                                        solution can be much          ing is only part of the story.”
                                                                        faster or slower than             A computational box of 1008 pro-
                                                                        the same reaction             teins is still a far stretch from the com-
                                                                        inside a crowded cell.        plex E. coli cytoplasm, says Allen
                                                                            To create the model,      Minton, PhD, a pioneer in the study of
                                                                        Elcock and then gradu-        crowding effects and researcher of
                                                                        ate student Sean              physical biochemistry at the National
                                                                        McGuffee, PhD, start-         Institutes of Health. “But there are a lot
                                                                        ed by gathering known         of questions that only this type of com-
                                                                        structural data for 50 of     putation can answer,” he says. “From a
                                                                        the most common E.            computational point of view, it is a real
                                                                        coli proteins. They then      tour-de-force.”
Combining all available known details about the atomic structures and   combined the detailed         —By Louisa Dalton
concentrations of 50 of the most common proteins within E. coli’s cyto- representations inside a
plasm, Elcock and McGuffee created a model of what it might be like     computer model at
inside the crowded cell. They then simulated 20 microseconds of         known cellular con-                    Animating
jostling with and without various types of molecular interactions,      centrations, creating a             Molecular Biology
including crowding (excluded volume effect) and electrostatic and       strikingly dense model           These days, molecular biologists
hydrophobic interactions. They then compared the results to experi-     of 1008 proteins. The         often gather data over a period of
mental observations. Reprinted from McGuffee SR, Elcock AH, 2010        researchers then set          time—observing shifts as they occur
Diffusion, Crowding & Protein Stability in a Dynamic Molecular Model    that image in motion,         inside groups of cells undergoing natural
of the Bacterial Cytoplasm. PLoS Comput Biol 6(3): e1000694.            running independent           changes. The researchers then face the
doi:10.1371/journal.pcbi.1000694.                                       Brownian dynamics             daunting task of making sense of it all.
                                                                        simulations governed          Now, computational biologists have
         atomically detailed computational by varying energetic descriptions of inter-                devised a software program to easily
         model of E. coli innards is also a tool for molecular interactions. The simplest             visualize and analyze these mountains of
         quantitative predictions of molecular description included only the excluded                 time-series data in animated movie
         conduct within the cell. The model is volume effect: no molecule could take                  form. While these flicks might never

Published by Simbios, the NIH National Center for Physics-Based Simulation of Biological Structures                                             3
       appear at a theater near you, scientists       work, led by Ihor Lemischka, PhD,              the grid, representing molecular shifts
       studying such disparate areas as stem cell     Mount Sinai professor of gene and cell         over the time course of the experimen-
       development and the microbial commu-           medicine, was published in Nature in           tal series. Although GATE was devel-
       nities of the Pacific Ocean will be play-      November 2009. “It was a relatively            oped for stem cell biologists, its poten-
       ing them on their computer screens to          simple approach but it hadn’t been             tial applications are broad, Ma’ayan
       explore how all the genes and proteins         done before,” Ma’ayan says. But                says. Recently, he was contacted by a
                                                                                                       group at the University of British
                                                                                                       Columbia that wants to use the soft-
                                                                                                       ware to analyze changes in marine
                                                                                                       flora and fauna in the Pacific Ocean.
                                                                                                       In this case, the movies will look at
                                                                                                       changes both over time and distance,
                                                                                                       as the researchers sample further from
                                                                                                       the coast.
                                                                                                           Oliver Hofmann, PhD, a computa-
                                                                                                       tional biology research scientist at the
                                                                                                       Harvard School of Public Health, says
                                                                                                       the technology will be very useful for
                                                                                                       the field of molecular biology. “It’s a
                                                                                                       very neat way of visualizing time
                                                                                                       series,” he says. “But it’s not just a pret-
                                                                                                       ty picture you can look at. You can
                                                                                                       explore it interactively too.” It is still
                                                                                                       difficult to coordinate more than two
                                                                                                       types of data timecourses in GATE,
                                                                                                       Hofmann says, and Ma’ayan agrees.
                                                                                                       He says their to-do list includes plans
                                                                                                       to better overlay multiple movies.
                                                                                                       —By Rachel Tompa, PhD

       This screenshot from the GATE software program shows RNA expression levels from experi-
                                                                                                            Capturing Mitosis
       ments on stem cells that were genetically manipulated to differentiate. Each hexagon repre-          Genes in Action
       sents a single gene; red hexagons are genes with increased RNA levels and green are those        During the one-hour drama that is
       with decreased levels. Commonalities among gene annotations are highlighted in blue, and      human cell division, many genes enter
       white lines represent known interactions between proteins. GATE movies animate a series of    and exit the stage. Until now,
       these images to show changes over time. Courtesy of Avi Ma’ayan.                              researchers did not know the identities
                                                                                                     of many of these actors, nor understand
       of a cell type or organism change over         Ma’ayan’s group realized that GATE             their various roles. Now, using a combi-
       the timespan of experiments.                   movies would be even more useful if            nation of high-throughput screening
          “This is a tool that is really useful for   they could incorporate existing biologi-       methods, time-resolved movies and a
       interrogating datasets collected as a time     cal data, such as libraries of protein-pro-    supervised machine-learning algorithm,
       series at multiple layers of regulation,”      tein interactions or annotations of            researchers have identified 572 genes
       says Avi Ma’ayan, PhD, assistant profes-       genes’ functions. The updated software         that are involved in mitosis in human
       sor of pharmacology and systems thera-         was further described in Bioinformatics        cells. The raw data and images are avail-
       peutics at the Mount Sinai School of           in January 2010.                               able to the research community at
       Medicine who spearheads the project.              The movies GATE generates show a  
       “It allows you to form hypotheses for          2-D honeycomb of small hexagons,                  “Researchers can go to the database,
       future experimentation very quickly.”          each representing a single gene or pro-        do a clustering analysis, and extract the
          The software, called GATE (Grid             tein and colored red (for increased            genes that are most interesting from
       Analysis of Time-Series Expression),           expression) or green (for decreased).          their research question point of view,”
       was originally designed to analyze clus-       The hexagons are clustered near other          says Jan Ellenberg, PhD, head of the
       tered gene and protein expression data         genes or proteins with similar behavior        Cell Biology and Biophysics Unit at the
       taken at various time points during stem       patterns in the experiments. When the          European Molecular Biology Laboratory
       cell development, Ma’ayan says. This           movie plays, waves of color shift across       and senior author on the paper pub-

4 BIOMEDICAL COMPUTATION REVIEW        Summer 2010                                             
lished in Nature on                                                                       and transcription factors to describe
April 1, 2010.                                                                            the information flow within cells.
    The research                                                                              The work serves as part of a larger
addressed an age-old                                                                      effort within Gerstein’s group to develop
problem in the study                                                                      real-world analogies to explain how bio-
of     cell     division,                                                                 logical systems use and process informa-
Ellenberg says. “We                                                                       tion. Previously, the group had shown
didn’t know all the                                                                       that hierarchies in biological regulatory
genes or the proteins                                                                     systems resemble directed social struc-
involved,” he says.                                                                       tures such as governments and corpora-
“So we decided that                                                                       tions. That study, published in PNAS in
we had to do this gene                                                                    2006, found that “middle managers
discovery ourselves.”                                                                     rule,” Gerstein says. Transcription fac-
    First, Ellenberg and This microscopy image captures the mitotic spindle (green)       tors in the middle layers of the networks
his colleagues in the and the chromosomes (red) of a dividing cell. EMBL                  have the most regulatory interactions
Mitocheck consor- researchers videotaped mitosis for 22,000 different gene                with other genes. “The genes in the
tium developed the knockouts. Videos and data for all 22,000 genes are available          middle are much more essential. If you
technology to do sys- at Courtesy of Thomas Walter & Jutta             knock them out, the organism is much
tematic high through- Bulkescher / EMBL.                                                  more likely to die.”
put screens of multiple                                                                       In a paper published online in PNAS
samples of all 22,000 human genes and at the Broad Institute, who was not                 in March 2010, Gerstein and his col-
then visually match each knockout to a involved in the research. “[The insights           leagues took that work a step farther,
phenotype. They relied on RNA inter- into mitosis are] just the tip of the ice-           seeking to understand how cells avoid
ference to knock out each of the berg of the knowledge that will be                       failure at the sites of middle manager
approximately 22,000 individual genes. extracted from this single experiment,”            bottlenecks in five species ranging from
They then printed more than 384 of Carpenter says.                                        E. coli to humans. First, they identified
these samples at a time on microarray        The researchers’ next project, called        which genes are regulated by other
chips. Because mitosis occurs transient- Mitosys, will explore the molecular              genes in each organism. They then
ly (approximately once every 24 activity of the 572 mitosis-related genes.                stacked the levels of regulators in hierar-
hours), the researchers developed —By Sarah A. Webb, PhD                                  chies and placed them between two
microscopes to capture movies of each                                                     extreme types of social hierarchies, auto-
sample from four such microarrays in                                                      cratic and democratic, and showed cel-
parallel over the course of 48 hours.
                                               Cells’ Collaborative                       lular regulatory hierarchies have "inter-
    Analysis of so much visual data—          Middle Management                           mediate" structures. They found that, in
nearly 200,000 movies—required               Like corporate and governmental              all five organisms, coregulation happens
supervised machine learning. First, a organizations, cells rely on middle man-            most at the middle level and least at the
human expert annotated examples of agers to keep things running smoothly.                 bottom. And more complex organisms
different morphologies observed within These “middle managers” function as a              exhibit more collaborative, “democrat-
the movies. A computer then extracted critical bridge that controls the flow of           ic” regulatory structures with more
a numerical signature with 200 differ- information traffic. According to recent           interconnections. For example, yeast
ent parameters that it correlated with research, however, the middle managers             has about one regulator for every 25 tar-
those characteristics. After iterative often partner with one another, ensur-             gets whereas in humans the ratio is
training with movies of just 3000 dif- ing that the failure of one manager does-          much smaller, about one to 10.
ferent individual cells, the computer n’t bring down the entire organization.                 “The parallels between government
analyzed additional movies and identi- Moreover, this partnering becomes more             structure and regulatory network struc-
fied phenotypes with 90 percent accu- extensive in more complex organisms.                ture are provocative,” says Trey Ideker,
racy. The researchers also developed         “Understanding the system isn’t              PhD, associate professor of medicine and
new distance measures for clustering about the function of the individual                 bioengineering at the University of
algorithms to categorize the differences parts. [It’s about] understanding the            California, San Diego, who was not
in cell division behavior.                importance of these information flow            involved with the study. One question,
    The scale of these experiments and bottlenecks and how natural systems                says M. Madan Babu, PhD, an investiga-
the use of time-lapse imaging over two get around them,” says Mark Gerstein,              tor in the MRC-Laboratory of Molecular
days are “unparalleled and nothing short PhD, professor of bioinformatics at              Biology at the University of Cambridge,
of phenomenal,” says Anne Carpenter, Yale University. He and his colleagues               is the function of these hierarchies with-
PhD, director of the Imaging Platform have been studying networks of genes                in a cell. “Are they really important? Or

Published by Simbios, the NIH National Center for Physics-Based Simulation of Biological Structures                                     5
       are they something that is emergent             have a better chance of eventually
       because of the complexity of the system         landing in our lungs.
       and has no consequence whatsoever?”                 “The conventional wisdom is that
          Regulatory networks are definitely           the thermal plume from your body pro-
       important for organism function,                tects you from particles falling from
       Gerstein notes. So the question of              above,” says John B. McLaughlin,
       whether the networks emerged in                 PhD, professor of chemical and bio-
       response to complex roles or the sys-           molecular engineering at Clarkson          The positions of 2-micrometer particles
                                                       University and coauthor of the study.      inside a 20-degree-Celsius room with a man-
                                                       “We found that, in our small room at       nequin heated to 25 degrees Celsius, three
                                                       least, that is not true.” Such findings    minutes after particles were released
                                                       can help engineers design better venti-    through a floor event. In this simulation, 31
                                                       lation systems, McLaughlin says.           out of 1000 particles fell directly onto the
                                                       “Studies have shown that schoolchild-      mannequin’s warm body; none managed to
                                                       ren learn more and office workers are      leave the room through the ceiling vent. Yet
                                                       more productive in environments            when the mannequin was the same temper-
                                                       where the concentration of particles in    ature as the room, no particles fell onto the
                                                       the air is very low.”                      body, and 160 out of 1000 particles escaped.
                                                           Airflow dynamics are notoriously       Results were similar for simulations with 10-
                                                       tough to model computationally, large-     micrometer-diameter particles.
                                                       ly because of the huge range of physi-
                                                       cal scales in equations for turbulent      meeting in Portland, Oregon.
                                                       fluids. McLaughlin and his colleagues         “The computational and the experi-
                                                       used a direct numerical simulation         mental go hand in hand when studying
                                                       approach that offers accuracy but          complex turbulent flows such as those
                                                       requires intensive computational           around human beings,” says Mark N.
                                                       resources. Their computational mod-        Glauser, PhD, professor of mechanical
                                                       els of airflow and particle paths were     and aerospace engineering at Syracuse
                                                       built in a 4.8-square-meter virtual        University, whose empirical results
       Diagrams of hierarchical networks: In an        room at two-centimeter resolution          helped guide McLaughlin’s modeling.
       autocratic network, such as the military,       using three-millisecond time steps over    Fundamentally, experiments can help
       there is a clear chain of command. In a dem-    about three minutes of total simulated     validate computational models and
       ocratic network, many members interact and      time. In each simulation, a mannequin      give physical insights that spur new
       regulate each other. And in an intermediate     sits motionless in the middle of the       simulations. “Then the simulation
       network, such as exists within a law firm and   room. A stream of air suffused with        tools can be used to probe a broader
       many cells, the hierarchy shares features of    particles—each with the density of         range of parameter space ‘virtually,’ as
       both types. As biological organisms become      sand and about the size of a grain of      well as look in more detail at flow
       increasingly complex, their organization        pollen—shoots up through a floor vent      physics,” Glauser says. For example, the
       becomes more democratic.                        in front of the chair. Particles fan out   models from McLaughlin’s team can
                                                       throughout the room, with a ceiling        track individual particles in a turbulent
       tem’s complexity allows organisms to            vent as the only exit.                     flow—a feat that’s nearly impossible in
       carry on these complex interaction is a             In simulations where the man-          real-life experiments.
       “chicken and egg type of issue.”                nequin was bestowed with realistic         —By Regina Nuzzo, PhD
       —By Sarah A. Webb, PhD                          body heat, researchers could see the
                                                       hot air surging off the body and inter-
             Hot Bodies a Lure                         acting with particulates. This thermal               Brain Folding
                                                       plume pulled rising particles directly         In the four months before birth, a
             for Unseen Specks                         into the mannequin’s breathing zone.       fetus’s brain grows from a smooth tube
          We can’t see them, but tiny parti-           At the same time, the plume blocked        of neurons into a highly crinkled, con-
       cles—dust, pollen, microbes, and the            the path of particles traveling near the   volved mass of tissue. Because the cere-
       like—swirl around us in complicated,            ceiling, forcing them to fall down into    bral cortex has a surface area nearly
       turbulent pathways. New numerical               the mannequin’s personal space, dou-       three times as big as that of its skull cav-
       simulations suggest that, at least in           bling the trapping effect of the plume.    ity, scientists have proposed that this
       tiny indoor spaces, our body heat may           The work was presented in March 2010       real-estate-space squeeze is what drives
       pull them even closer, where they               at the American Physical Society           the brain’s folding process. Now results

6 BIOMEDICAL COMPUTATION REVIEW        Summer 2010                                          
from a computational three-dimension-              geometry of the cortex.                            ers, this imbalance subtly shapes what
al geometric model agree that the skull                The team simulated how the cortex              kinds of folds become most prominent.
does help guide the wrinkling—but                  grows under various conditions: with-                 Computational models can help
they also suggest that a growing brain             out a skull, with a skull of fixed size,           explain normal brain development as
folds up regardless of its container.              and with a skull that grows at the same            well as what happens when things go
   “Mechanical constraints imposed                 time as the brain does. As expected,               wrong, says Bernard S. Chang, MD,
by the skull are important regulators,”            brains grown in a skull were more con-             assistant professor of neurology at
says Tianming Liu, PhD, assistant pro-             voluted than those allowed to develop              Harvard Medical School. For example,
fessor of computer science at                      unfettered. But even without a skull to            in some forms of microcephaly, the
University of Georgia and lead author              confine it, a cortex will still fold in on         brain surface is almost completely
on the study, which was published in               itself, results showed. This happens as            smooth with no folds; in others, the
May 2010 in the Journal of Theoretical             a natural response in a fast-growing               folding is normal. “A model that pre-
Biology. “But our simulations indicate             cortex, as the tissue attempts to reduce           dicts how folding is affected by the
that skull constraint is not necessarily           the increasing mechanical tension                  skull’s physical constraints might help
the dominant mechanism.”                           among axons, dendrites, and neu-                   us to understand why some patients
   The computational model under-                  roglia, Liu says.                                  have one form and not another,” he

Growth of the cortex under different assumptions. From left to right,        a fetus' developing skull). Major cortical folds developed much earli-
the images show simulated development of the cortex over time. The           er and faster during simulations with skull constraints. But the cortex
cortex grows (a) within a skull of fixed size, (b) without a skull, or (c)   increases its surface area and convolutes itself to reduce the fast-
within a skull that also grows at the same time (which corresponds to        growing internal tension, with or without skull constraint.

neath the simulations had two main                    Tweaking other parameters in the                says. Since animal models don’t cap-
features: geometric constraints of the             model revealed how cellular growth                 ture the complexity of the human
skull, and partial differential equations          affects these folding patterns. When               brain, and doing repeated MRIs of
that model biological processes driving            neurons themselves grow rapidly —                  developing fetuses for research isn’t fea-
the growth of neurons. To start off the            during synapse development and neu-                sible, Chang says, “we need to rely on
simulation, researchers used MRI data              ron dendritic projection, for example—             these theoretical models as tools to
from the brains of two human fetuses;              the cortical folding increases dramati-            help us understand what we’re observ-
then solutions to the differential equa-           cally too. And when certain areas of               ing clinically.”
tions guided the changing surface                  the cortex grow more quickly than oth-             —By Regina Nuzzo, PhD ■

Published by Simbios, the NIH National Center for Physics-Based Simulation of Biological Structures                                                    7

To top