Applications of Population Genetics to Animal Breeding,from Wright,Fisher and Lush to Genomic Prediction

Although animal breeding was practiced long before the science of genetics and the relevant disciplines of population and quantitative genetics were known, breeding programs have mainly relied on simply selecting and mating the best individuals on their own or relatives’ performance. This is based on sound quantitative genetic principles, developed and expounded by Lush, who attributed much of his understanding to Wright, and formalized in Fisher’s infinitesimal model. Analysis at the level of individual loci and gene frequency distributions has had relatively little impact. Now with access to genomic data, a revolution in which molecular information is being used to enhance response with “genomic selection” is occurring. The predictions of breeding value still utilize multiple loci throughout the genome and, indeed, are largely compatible with additive and specifically infinitesimal model assumptions. I discuss some of the history and genetic issues as applied to the science of livestock improvement, which has had and continues to have major spin-offs into ideas and applications in other areas.THE success of breeders in effecting immense changes in domesticated animals and plants greatly influenced Darwin’s insight into the power of selection and implications to evolution by natural selection. Following the Mendelian rediscovery, attempts were soon made to accommodate within the particulate Mendelian framework the continuous nature of many traits and the observation by Galton (1889) of a linear regression of an individual’s height on that of a relative, with the slope dependent on degree of relationship. A polygenic Mendelian model was first proposed by Yule (1902) (see Provine 1971; Hill 1984). After input from Pearson, Yule again, and Weinberg (who developed the theory a long way but whose work was ignored), its first full exposition in modern terms was by Ronald A. Fisher (1918) (biography by Box 1978). His analysis of variance partitioned the genotypic variance into additive, dominance and epistatic components. Sewall Wright (biography by Provine 1986) had by then developed the path coefficient method and subsequently (Wright 1921) showed how to compute inbreeding and relationship coefficients and their consequent effects on genetic variation of additive traits. His approach to relationship in terms of the correlation of uniting gametes may be less intuitive at the individual locus level than Malécot’s (1948) subsequent treatment in terms of identity by descent, but it transfers directly to the correlation of relatives for quantitative traits with additive effects.From these basic findings, the science of animal breeding was largely developed and expounded by Jay L. Lush (1896–1982) (see also commentaries by Chapman 1987 and Ollivier 2008). He was from a farming family and became interested in genetics as an undergraduate at Kansas State. Although his master’s degree was in genetics, his subsequent Ph.D. at the University of Wisconsin was in animal reproductive physiology. Following 8 years working in animal breeding at the University of Texas he went to Iowa State College (now University) in Ames in 1930. Wright was Lush’s hero: ‘I wish to acknowledge especially my indebtedness to Sewall Wright for many published and unpublished ideas upon which I have drawn, and for his friendly counsel” (Lush 1945, in the preface to his book Animal Breeding Plans). Lush commuted in 1931 to the University of Chicago to audit Sewall Wright’s course in statistical genetics and consult him. Speaking at the Poultry Breeders Roundtable in 1969: he said, “Those were by far the most fruitful 10 weeks I ever had.” (Chapman 1987, quoting A. E. Freeman). Lush was also exposed to and assimilated the work and ideas of R. A. Fisher, who lectured at Iowa State through the summers of 1931 and 1936 at the behest of G. W. Snedecor.Here I review Lush’s contributions and then discuss how animal breeding theory and methods have subsequently evolved. They have been based mainly on statistical methodology, supported to some extent by experiment and population genetic theory. Recently, the development of genomic methods and their integration into classical breeding theory has opened up ways to greatly enhance rates of genetic improvement. Lush focused on livestock improvement and spin-off into other areas was coincidental; but he had contact with corn breeders in Ames and beyond and made contributions to evolutionary biology and human genetics mainly through his developments in theory (e.g., Falconer 1965; Robertson 1966; Lande 1976, 1979; see also Hill and Kirkpatrick 2010). I make no attempt to be comprehensive, not least in choice of citations.     

Louis Reichardt: The long climb to science's summits

From the highest mountains to biology''s own Everest—the brain—Reichardt tackles the biggest challenges of climbing and biology.Louis Reichardt''s scientific career has spanned the simple and the complex. As a graduate student, Reichardt helped to uncover the now renowned DNA regulatory mechanisms that allow one of the simplest life forms, lambda phage, either to hide within a cell or to make its presence known via massive replication (1).Open in a separate windowLouis ReichardtLetting his curiosity for the unknown guide him, Reichardt then forayed into a much more intricate system, the brain. As a postdoc, he showed that growth conditions influenced which neurotransmitters are synthesized by isolated neurons (2). Later, as a professor at UCSF, he discovered synaptotagmin, using the first monoclonal antibody that defined a synaptic vesicle membrane protein (3), showed that expression levels of nerve growth factor in target tissues correlate with the density of innervation (4), and characterized the properties of mice lacking genes encoding the neurotrophins and their Trk receptors (5, 6).Now a professor and director of the Neuroscience Program at UCSF, Reichardt''s laboratory still studies the interface of cell biology and neurobiology, including the involvement of cell adhesion molecules in synaptic development (7). He explains that science is not so different from his favorite hobby, mountain climbing.

“The gist of it was ‘Others more foolish might try this, but it was not for us.’ I thought, ‘We''ll obviously have to do it.’”

     

Flexible open conformation of the AP-3 complex explains its role in cargo recruitment at the Golgi

Vesicle formation at endomembranes requires the selective concentration of cargo by coat proteins. Conserved adapter protein complexes at the Golgi (AP-3), the endosome (AP-1), or the plasma membrane (AP-2) with their conserved core domain and flexible ear domains mediate this function. These complexes also rely on the small GTPase Arf1 and/or specific phosphoinositides for membrane binding. The structural details that influence these processes, however, are still poorly understood. Here we present cryo-EM structures of the full-length stable 300 kDa yeast AP-3 complex. The structures reveal that AP-3 adopts an open conformation in solution, comparable to the membrane-bound conformations of AP-1 or AP-2. This open conformation appears to be far more flexible than AP-1 or AP-2, resulting in compact, intermediate, and stretched subconformations. Mass spectrometrical analysis of the cross-linked AP-3 complex further indicates that the ear domains are flexibly attached to the surface of the complex. Using biochemical reconstitution assays, we also show that efficient AP-3 recruitment to the membrane depends primarily on cargo binding. Once bound to cargo, AP-3 clustered and immobilized cargo molecules, as revealed by single-molecule imaging on polymer-supported membranes. We conclude that its flexible open state may enable AP-3 to bind and collect cargo at the Golgi and could thus allow coordinated vesicle formation at the trans-Golgi upon Arf1 activation.

Eukaryotic cells have membrane-enclosed organelles, which carry out specialized functions, including compartmentalized biochemical reactions, metabolic channeling, and regulated signaling, inside a single cell. The transport of proteins, lipids, and other molecules between these organelles is mediated largely by small vesicular carriers that bud off at a donor compartment and fuse with the target membrane to deliver their cargo. The generation of these vesicles has been subject to extensive studies and has led to the identification of numerous coat proteins that are required for their formation at different sites (1, 2). Coat proteins can be monomers, but in most cases, they consist of several proteins, which form a heteromeric complex.Heterotetrameric adapter protein (AP) complexes are required at several endomembranes for cargo binding. Five well-conserved AP-complexes with differing functions have been identified in mammalian cells, named AP-1–AP-5, of which three (AP-1–AP-3) are conserved from yeast to human (3, 4). The three conserved adapter complexes function at different membranes along the endomembrane system. AP-1 is required for cargo transport between the Golgi and the endosome, AP-2 is required for cargo recognition and transport between the plasma membrane and the early endosome. Finally, AP-3 functions between the trans Golgi and the vacuole in yeast, whereas mammalian AP-3 localizes to a tubular endosomal compartment, in addition to or instead of the TGN (2, 5, 6).Each of the complexes consists of four different subunits: two large adaptins (named α−ζ and β1-5 respectively), a medium-sized subunit (μ1-5), and a small subunit (σ1-5). While μ- and σ-subunits together with the N-termini of the large adaptins build the membrane-binding core of the complex, the C-termini of both adaptins contain the ear domains, which are connected via flexible linkers (2). The recruitment of these complexes to membranes is not entirely conserved. They all require cargo binding, yet AP-1 binds Arf1-GTP with the γ and β1 subunit and phosphatidylinositol-4-phosphate (PI4P) via a proposed conserved site on its γ-subunit (7, 8). AP-2, on the other hand, interacts with PI(4,5)P₂ at the plasma membrane via its α, β2, and μ2 subunits (9, 10, 11).Several studies have uncovered how AP-3 functions in cargo sorting in yeast. AP-3 recognizes cargo at the Golgi via two sorting motifs in the cytosolic segments of membrane proteins: a Yxxφ sorting motif, as found in yeast in the SNARE Nyv1 or the Yck3 casein kinase, which binds to a site in μ3, as shown for mammalian AP-3, which is similar to μ2 in AP-2 (12, 13, 14), and dileucine motifs as found in the yeast SNARE Vam3 or the alkaline phosphatase Pho8, potentially also at a site comparable to AP-1 and AP-2 (15, 16). Unlike AP-1 and AP-2-coated vesicles, which depend on clathrin for their formation (2, 17), AP-3 vesicle formation in yeast does not require clathrin or the HOPS subunit Vps41 (18), yet Vps41 is required at the vacuole to bind AP-3 vesicles prior to fusion (19, 20, 21, 22). Studies in metazoan cells revealed that Vps41 and AP-3 function in regulated secretion (23, 24, 25), and AP-3 is required for biogenesis of lysosome-related organelles (26). This suggests that the AP-3 complex has features that are quite different from AP-1 and AP-2 complexes, which cooperate with clathrin in vesicle formation (2).Among the three conserved AP complexes, the function of the AP-3 complex is the least understood. Arf1 is necessary for efficient AP-3 vesicle generation in mammalian cells and shows a direct interaction with the β3 and δ subunits of AP-3 (27, 28). In addition, in vitro experiments on mammalian AP-3 using liposomes or enriched Golgi membranes suggest Arf1 as an important factor in AP-3 recruitment, whereas acidic lipids do not have a major effect, in contrast to what was found for AP-1 and AP-2 (7, 11, 29, 30). Another study showed that membrane recruitment of AP-3 depends on the recognition of sorting signals in cargo tails and PI3P (31), similar to AP-1 recruitment via cargo tails, Arf1 and PI4P (32).However, since AP-1 and AP-3 are both recruited to the trans-Golgi network (TGN) in yeast (33), the mechanism of their recruitment likely differs. Even though Arf1 is required, yeast AP-3 seems to be present at the TGN before the arrival of the Arf1 guanine nucleotide exchange factor (GEF) Sec7 (33). This implies the necessity for additional factors at the TGN and a distinct mechanism to allow for spatial and temporal separation of AP-1 and AP-3 recruitment to membranes. Structural data on mammalian AP-1 and AP-2 “core” complexes without the hinge and ear domains of their large subunits revealed that both exist in at least two very defined conformational states: a “closed” cytosolic state, where the cargo-binding sites are buried within the complex, and an “open” state, where the same sites are available to bind cargo (7, 8, 10, 34, 35). Binding of Arf1 to AP-1 or PI(4,5)P₂ in case of AP-2 induces a conformational change in the complexes that enables them to bind cargo molecules carrying a conserved acidic di-Leucine or a Tyrosine-based motif, as for all three AP complexes in yeast (8, 34). Additional conformational states and intermediates have been reported for both, mammalian AP-1 and AP-2 complex. AP-1, for example, can be hijacked by the human immunodeficiency virus-1 (HIV-1) proteins viral protein u (Vpu) and negative factor (Nef), resulting in a hyper-open conformation of AP-1 (36, 37).An emerging model over the past years has suggested that APs have several binding sites that allow for the stabilization of membrane binding and the open conformation of the complexes, but there are initial interactions required that dictate their recruitment to the target membrane. Although these interaction sites for mammalian AP-1 and AP-2 have been identified in great detail based on interaction analyses and structural studies (8, 10, 11, 35, 36, 38, 39), structural data for AP-3 is largely missing. The C-terminal part of the μ-subunit of mammalian AP-3 has been crystallized together with a Yxxφ motif-containing a cargo peptide, which revealed a similar fold and cargo-binding site as shown for AP-1 and AP-2 (14). However, positively charged binding surfaces required for PIP-interaction were not well conserved. Although the “trunk” segment of AP-1 and AP-2 is known quite well by now, information on hinge and ear domains in context of these complexes is largely missing. Crystal structures of the isolated ear domains of α-, γ- and β2-adaptin have been published (40, 41, 42), and a study on mammalian AP-3 suggested a direct interaction between δ-ear and δ3 that interfered with Arf1-binding (43). Furthermore, during tethering of AP-3 vesicles with the yeast vacuole, the δ−subunit Apl5 of the yeast AP-3 complex binds to the Vps41 subunit of the HOPS complex as a prerequisite of fusion (18, 19, 21, 22).In this study, we applied single particle electron cryo-microscopy (cryo-EM) to analyze the purified full-length AP-3 complex from yeast and unraveled the factors required for AP-3 recruitment to membranes by biochemical reconstitution. Our data reveal that a surprisingly flexible AP-3 complex requires a combination of cargo, PI4P, and Arf1 for membrane binding, which explains its function in selective cargo sorting at the Golgi.     

Extracellular DNA Is Required for Root Tip Resistance to Fungal Infection

Plant defense involves a complex array of biochemical interactions, many of which occur in the extracellular environment. The apical 1- to 2-mm root tip housing apical and root cap meristems is resistant to infection by most pathogens, so growth and gravity sensing often proceed normally even when other sites on the root are invaded. The mechanism of this resistance is unknown but appears to involve a mucilaginous matrix or “slime” composed of proteins, polysaccharides, and detached living cells called “border cells.” Here, we report that extracellular DNA (exDNA) is a component of root cap slime and that exDNA degradation during inoculation by a fungal pathogen results in loss of root tip resistance to infection. Most root tips (>95%) escape infection even when immersed in inoculum from the root-rotting pathogen Nectria haematococca. By contrast, 100% of inoculated root tips treated with DNase I developed necrosis. Treatment with BAL31, an exonuclease that digests DNA more slowly than DNase I, also resulted in increased root tip infection, but the onset of infection was delayed. Control root tips or fungal spores treated with nuclease alone exhibited normal morphology and growth. Pea (Pisum sativum) root tips incubated with [³²P]dCTP during a 1-h period when no cell death occurs yielded root cap slime containing ³²P-labeled exDNA. Our results suggest that exDNA is a previously unrecognized component of plant defense, an observation that is in accordance with the recent discovery that exDNA from white blood cells plays a key role in the vertebrate immune response against microbial pathogens.Root diseases caused by soil-borne plant pathogens are a perennial source of crop loss worldwide (Bruehl, 1986; Curl and Truelove, 1986). These diseases are of increasing concern, as pesticides like methyl bromide are removed from the market due to environmental concerns (Gilreath et al., 2005). One possible alternative means of crop protection is to exploit natural mechanisms of root disease resistance (Nelson, 1990; Goswami and Punja, 2008; Shittu et al., 2009). Direct observation of root systems under diverse conditions has revealed that root tips, in general, are resistant to infection even when lesions are initiated elsewhere on the same plant root (Foster et al., 1983; Bruehl, 1986; Curl and Truelove, 1986; Smith et al., 1992; Gunawardena et al., 2005; Wen et al., 2007). This form of disease resistance is important for crop production because root growth and its directional movement in response to gravity, water, and other signals can proceed normally as long as the root tip is not invaded. The 1- to 2-mm apical region of roots houses the root meristems required for root growth and cap development, and when infection does occur, root development ceases irreversibly within a few hours even in the absence of severe necrosis (Gunawardena and Hawes, 2002). Mechanisms underlying root tip resistance to infection are unclear, but the phenomenon appears to involve root cap “slime,” a mucilaginous matrix produced by the root cap (Morré et al., 1967; Rougier et al., 1979; Foster, 1982; Chaboud, 1983; Guinel and McCully, 1986; Moody et al., 1988; Knee et al., 2001; Barlow, 2003; Iijima et al., 2008). Within the root cap slime of cereals, legumes, and most other crop species are specialized populations of living cells called root “border cells” (Supplemental Fig. S1; Hawes et al., 2000). Border cell numbers increase in response to pathogens and toxins such as aluminum, and the cell populations maintain a high rate of metabolic activity even after detachment from the root cap periphery (Brigham et al., 1995; Miyasaka and Hawes, 2000).As border cells detach from roots of cereals and legumes, a complex of more than 100 proteins, termed the root cap secretome, is synthesized and exported from living cells into the matrix ensheathing the root tip (Brigham et al., 1995). The profile of secreted proteins changes in response to challenge with soil-borne bacteria (De-la-Peña et al., 2008). In pea (Pisum sativum), root tip resistance to infection is abolished in response to proteolytic degradation of the root cap secretome (Wen et al., 2007). In addition to an array of antimicrobial enzymes and other proteins known to be components of the extracellular matrix and apoplast of higher plants, the DNA-binding protein histone H4 unexpectedly was found to be present among the secreted proteins (Wen et al., 2007). One explanation for the presence of histone is global leakage of material from disrupted nuclei in dead cells, but no cell death occurs during delivery of the secretome (Brigham et al., 1995; Wen et al., 2007). An alternative explanation for the presence of a secreted DNA-binding protein is that extracellular DNA (exDNA) also is present in root cap slime.exDNA has long been known to be a component of slimy biological matrices ranging from purulent localized human infections to bacterial capsules, biofilms, and snail exudate (Sherry and Goeller, 1950; Leuchtenberger and Schrader, 1952; Braun and Whallon, 1954; Smithies and Gibbons, 1955; Catlin, 1956; Fahy et al., 1993; Allesen-Holm et al., 2006; Spoering and Gilmore, 2006; Qin et al., 2007; Izano et al., 2008). Specialized white blood cells in humans and other species including fish recently have been shown to deploy a complex neutrophil extracellular “trap” (NET), composed of DNA and a collection of enzymes, in response to infection (Brinkmann et al., 2004; Brinkmann and Zychlinsky, 2007; Palić et al., 2007; Wartha et al., 2007; Yousefi et al., 2008). NETs appear to kill bacterial, fungal, and protozoan pathogens by localizing them within a matrix of antimicrobial peptides and proteins (Urban et al., 2006; Wartha et al., 2007; Guimaraes-Costa et al., 2009). Several extracellular peptides and proteins implicated in neutrophil function, including histone, also are present within the pea root cap secretome (Wen et al., 2007). exDNA linked with extracellular histone is a structural component of NETs, and treatment with DNase destroys NET integrity and function (Wartha et al., 2007). Moreover, human pathogens including group A Streptococcus and Streptococcus pneumoniae release extracellular DNase (Sherry and Goeller, 1950). When these activities are eliminated by mutagenesis of the encoding genes, bacteria lose their normal ability to escape the NET and multiply at the site of infection (Sumby et al., 2005; Buchanan et al., 2006). Here, we report that, in addition to histone and other secretome proteins, exDNA also is a component of root cap slime. When this exDNA is digested enzymatically, root tip resistance to infection is abolished.     

The contribution of αβ-tubulin curvature to microtubule dynamics

Microtubules are dynamic polymers of αβ-tubulin that form diverse cellular structures, such as the mitotic spindle for cell division, the backbone of neurons, and axonemes. To control the architecture of microtubule networks, microtubule-associated proteins (MAPs) and motor proteins regulate microtubule growth, shrinkage, and the transitions between these states. Recent evidence shows that many MAPs exert their effects by selectively binding to distinct conformations of polymerized or unpolymerized αβ-tubulin. The ability of αβ-tubulin to adopt distinct conformations contributes to the intrinsic polymerization dynamics of microtubules. αβ-Tubulin conformation is a fundamental property that MAPs monitor and control to build proper microtubule networks.Microtubules are polar polymers formed from αβ-tubulin heterodimers. These tubulin subunits associate head-to-tail to form protofilaments, and typically 13 protofilaments are associated side-by-side to form the hollow cylindrical microtubule. Most microtubules emanate from microtubule organizing centers, in which their minus ends are embedded. GTP-tubulin associates with the fast-growing plus ends as the microtubules radiate to explore the cell interior (see Box).

The cycle of microtubule polymerization.

Fig. 1). The addition of a new subunit completes the active site for GTP hydrolysis, and consequently most of the body of the microtubule contains GDP-bound αβ-tubulin. The GDP lattice is unstable but protected from depolymerization by a stabilizing “GTP cap,” an extended region of newly added GTP- or GDP.Pi-bound αβ-tubulin. The precise nature of the microtubule end structure and the size and composition of the cap are a matter of debate. Loss of the stabilizing cap leads to rapid depolymerization, which is characterized by an apparent peeling of protofilaments. “Catastrophe” denotes the switch from growth to shrinkage, and “rescue” denotes the switch from shrinkage to growth.Open in a separate windowFigure 1.Three structures of GTP-bound αβ-tubulin adopt similar curved conformations. Different αβ-tubulin structures were superimposed using α-tubulin as a reference, and oligomers were generated by assuming that the spatial relationship between α- and β-tubulin within a heterodimer is identical to the relationship between heterodimers. Curvature is calculated from the rotational component of the transformation required to superimpose the α-tubulin chain onto the β-tubulin chain of the same heterodimer. All of the GTP-bound structures (Rb3 complex, Protein Data Bank [PDB] accession no. 3RYH [magenta]; DARPin complex, PDB accession no. 4DRX [green]; TOG1 complex, PDB accession no. 4FFB [blue]) show between 10° and 13° of curvature, which is very similar to the curvature observed in GDP-bound structures (see inset, where the αβ-tubulins from a GDP-bound stathmin complex [PDB accession no. 1SA0] are shown in yellow and orange). A straight protofilament (putty and dark red color, PDB accession no. 1JFF) and a partially straightened assembly (tan) from GMPCPP ribbons are shown for reference.Unlike actin filaments, which grow steadily, microtubules frequently switch between phases of growth and shrinkage. This hallmark property of microtubules, known as “dynamic instability” (Mitchison and Kirschner, 1984), allows the microtubule cytoskeleton to be remodeled rapidly over the course of the cell cycle. “Catastrophes” are GTPase-dependent transitions from growing to shrinking, whereas “rescues” are transitions from shrinking to growing. Numerous microtubule-associated proteins (MAPs) regulate microtubule polymerization dynamics. Discovering how cells regulate and harness dynamic instability is a fundamental challenge in cell biology.A recent accumulation of structural, biochemical, and in vitro reconstitution data has advanced the understanding of dynamic instability and the MAPs that control it. Fresh structural data have provided insight into the process of microtubule assembly and defined how some MAPs recognize αβ-tubulin in and out of the microtubule. In vitro reconstitution experiments are reshaping the understanding of catastrophe and also providing quantitative insight into the mechanism of MAPs. Here, we review this progress, paying special attention to the emerging theme of interactions that are selective for different conformations of αβ-tubulin, both inside and outside the microtubule lattice. We argue for the central importance of recognizing these distinct conformations in the control of microtubule dynamics by MAPs and hence in the construction of a functional microtubule cytoskeleton by cells.

Tubulin dimers and their curvatures

It was clear in early EM studies that αβ-tubulin could form a diversity of polymers (Kirschner et al., 1974). In particular, the first cryo-EM of dynamic microtubules (Mandelkow et al., 1991) revealed significant differences in the appearance of growing and shrinking microtubule ends. Growing microtubule ends had straight protofilaments and were tapered, with uneven protofilament lengths, whereas shrinking microtubule ends had curved protofilaments that peeled outward and lost their lateral contacts. These and other data established the canonical model that GTP-tubulin is “straight” but GDP-tubulin is “curved” (Melki et al., 1989). The idea that GTP binding straightened αβ-tubulin into a microtubule-compatible conformation before polymerization was appealing because it provided a structural rationale for why microtubule assembly required GTP and how GTP hydrolysis could lead to catastrophe. A subsequent cryo-EM study (Chrétien et al., 1995), however, revealed that growing microtubules often tapered and curved gently outward without losing their lateral contacts. These data suggested that GTP-tubulin might not be fully straight at the time of its incorporation into the microtubule lattice, an observation that set the stage for a still-active debate on the structure of GTP-tubulin and of microtubule ends.The atomic details of “straight” and “curved” became apparent when the first structures of αβ-tubulin were solved. The straight conformation of αβ-tubulin was determined from cryo-electron crystallographic studies of Zn-induced αβ-tubulin sheets (Nogales et al., 1998). The structure showed linear head-to-tail stacking of αβ-tubulin along the protofilament, both within and between αβ-tubulin heterodimers. The curved conformation of αβ-tubulin was determined from x-ray crystallographic studies of a complex between αβ-tubulin and Rb3 (Gigant et al., 2000; Ravelli et al., 2004), a microtubule-destabilizing factor in the Op18/stathmin family (Belmont and Mitchison, 1996). In this complex, the individual α- and β-tubulin chains adopted a characteristic conformation distinct from their straight one. Longitudinal interactions also differed from those in the straight conformation (Fig. 1): within and between the heterodimers, successive α- and β-tubulin chains were related by an ∼12° rotation. A chain of these curved αβ-tubulins generates an arc with a radius of curvature resembling that of the peeling protofilaments at shrinking microtubule ends (Gigant et al., 2000; Steinmetz et al., 2000).Straight and curved are not the only two conformations, however. A cryo-EM study of αβ-tubulin helical ribbons trapped using guanylyl 5′-α,β-methylenediphosphonate (GMPCPP), a slowly hydrolyzable analogue of GTP, provided a molecular view of a possible microtubule assembly intermediate (Wang and Nogales, 2005). In these ribbons, GMPCPP-bound αβ-tubulin adopted a conformation roughly halfway (∼5° rotation) between the straight and curved conformations. These partially curved αβ-tubulin heterodimers formed two types of lateral bonds, only one of which resembled those in the microtubule. This structure suggested that at least some αβ-tubulin straightening occurs during polymerization.Until recently, structural information about the conformation of unpolymerized GTP-bound αβ-tubulin was notably lacking. Three recent crystal structures (Nawrotek et al., 2011; Ayaz et al., 2012; Pecqueur et al., 2012) have now provided remarkably similar views of this previously elusive species. In all three structures, GTP-bound αβ-tubulin adopts a fully curved conformation, with its α- and β-tubulin subunits related by ∼12° of rotation (Fig. 1). This curvature is not consistent with models in which GTP binding straightens unpolymerized αβ-tubulin. In each of the structures, αβ-tubulin is bound to another protein, stathmin/Rb3 (Ozon et al., 1997), a designed ankyrin repeat protein (DARPin; Pecqueur et al., 2012), as well as a TOG domain from the Stu2/XMAP215 family of microtubule polymerases (Gard and Kirschner, 1987; Wang and Huffaker, 1997). Biochemical experiments have failed to detect GTP-induced straightening of αβ-tubulin, arguing against the possibility that these unrelated binding partners forced GTP-tubulin to adopt the curved conformation. For example, the affinity of stathmin–tubulin interactions is the same for GTP-tubulin and GDP-tubulin (Honnappa et al., 2003). Similarly, five small molecule ligands that target the colchicine binding site and are predicted to bind only curved αβ-tubulin have equivalent affinity for GTP-tubulin, GDP-tubulin, and αβ-tubulin in the stathmin complex (Barbier et al., 2010). Likewise, a TOG domain from Stu2p binds to GTP- and GDP-tubulin with comparable affinity (Ayaz et al., 2012). Finally, DARPin binds equally well to GTP- and GDP-tubulin even though it contacts a structural element that is positioned differently in the straight and curved conformations (Pecqueur et al., 2012). Taken together with early biochemical experiments (Manuel Andreu et al., 1989; Shearwin et al., 1994), these new data strongly support a model in which unpolymerized αβ-tubulin is curved whether it is bound to GTP or to GDP (Buey et al., 2006; Rice et al., 2008; Nawrotek et al., 2011). According to this model, the curved-to-straight transition occurs during the polymerization process, not before. We discuss some implications of this new view at the end of the following section.

Conformation and dynamic instability

How does GTP hydrolysis destabilize the microtubule lattice and trigger catastrophe? A recent structural study has compared high-resolution cryo-EM reconstructions of GMPCPP microtubules and GDP microtubules to provide some answers to this question (Alushin et al., 2014). The structures show that GTP hydrolysis induces a compaction at the longitudinal interface between dimers, immediately above the exchangeable nucleotide-binding site. This compaction is accompanied by conformational changes in α-tubulin. In contrast, lateral contacts between tubulins were essentially unchanged in the different nucleotide states. These observations suggest that GTP hydrolysis introduces strain into the lattice, but how this strain affects the strength of longitudinal and lateral bonds to destabilize the microtubule remains unknown. The GMPCPP and GDP microtubules also show distinct arrangements of elements that bind to MAPs, which suggests a structural mechanism some MAPs could use to distinguish GTP lattices from GDP lattices (discussed later).In parallel with these structural advances, in vitro reconstitutions (Gardner et al., 2011b) have undermined the textbook view about the kinetics of catastrophe. The seminal measurements of catastrophe frequency (Walker et al., 1988, 1991) assumed that catastrophe occurred with the same probability on newly formed and old microtubules. In other words, the analysis implied that catastrophe was a first-order, single-step process. Although subsequent experiments (e.g., Odde et al., 1995; Janson et al., 2003) indicated that catastrophe involved multiple steps, the first-order view of catastrophe was widely adopted (Howard, 2001; Phillips et al., 2008). Recent experiments using a single-molecule assay for microtubule growth (Gell et al., 2010) have now shown definitively that catastrophe is not a single-step process; rather, newly formed microtubules undergo catastrophe less frequently than older ones (Gardner et al., 2011b). “Age-dependent” catastrophe implies that the stabilizing structure at the end of growing microtubules is evolving to become less effective. The timescale of this evolution is long compared with the kinetics of αβ-tubulin association (Gardner et al., 2011a). Thus, the ageing process probably reports on one or more structural properties of the microtubule end, such as the presence of “defects” in the lattice (Gardner et al., 2011b) or possibly increased tapering of microtubule ends (Coombes et al., 2013).It now seems clear that changes in the curvature of αβ-tubulin during microtubule polymerization are fundamental to microtubule dynamics and the regulatory activities of MAPs. Having straight conformations of αβ-tubulin only occur appreciably in the microtubule lattice provides a simple structural mechanism by which MAPs can discriminate unpolymerized from polymerized αβ-tubulin. Biochemical properties that define microtubule dynamics, like the strength of lateral and longitudinal contacts and the rate of GTP hydrolysis, may differ for curved, straight, and intermediate conformations of αβ-tubulin; e.g., curved forms probably bind microtubule ends less tightly than straight forms. By regulating when and where these different conformations occur, MAPs can tune microtubule dynamics. More speculatively, the complex biochemistry associated with different conformations of αβ-tubulin may contribute to the aging of microtubule ends, which leads to catastrophe. Understanding the connections between αβ-tubulin conformation, biochemistry, and polymerization dynamics is a major challenge for the future. Expanding the current mathematical models (Bowne-Anderson et al., 2013) and computational models (VanBuren et al., 2005; Margolin et al., 2012) of microtubule dynamics to incorporate these new findings about αβ-tubulin structure and age-dependent catastrophe may yield significant insights. In the following sections, we will examine recent studies that demonstrate how MAPs use selective interactions with distinct conformations of αβ-tubulin to control microtubule dynamics and thereby the physiology of the microtubule cytoskeleton.

Microtubule depolymerases stabilize curved conformations of tubulin

Perhaps the first direct evidence that MAPs might control the conformation of αβ-tubulin came from studies of microtubule depolymerases, which are proteins that promote, accelerate, or induce the depolymerization of microtubules (Howard and Hyman, 2007). Cells use microtubule depolymerases to maintain local control of microtubule catastrophe. Early electron microscopy studies of two unrelated depolymerases, Op18/stathmin and the kinesin-13 Xkcm1, showed that these proteins were able to induce/stabilize the curved conformation of αβ-tubulin and/or curved protofilaments (Desai et al., 1999; Gigant et al., 2000; Steinmetz et al., 2000). Depolymerases are also referred to as “catastrophe factors” because they trigger catastrophes in dynamic microtubules. The localized control of catastrophe is the essential function of depolymerases in cell physiology.The microtubule depolymerase stathmin is inactivated around chromosomes and at the leading edge of migrating cells (Niethammer et al., 2004), creating a gradient of depolymerase activity in these zones. Proteins in the Op18/stathmin family form a tight complex with two curved tubulin dimers (Fig. 2 A). Op18/stathmin proteins have been critical for the crystallization of tubulin (Ravelli et al., 2004; Gigant et al., 2005; Prota et al., 2013) and for biochemical studies of tubulin conformation. Although stathmins are frequently described as tubulin-sequestering proteins, the effect they have on microtubule catastrophe frequencies in vitro is much stronger than would be predicted from the simple sequestration of tubulin (Belmont and Mitchison, 1996). The potency of stathmins suggests that they induce catastrophes through direct interactions with microtubule ends, presumably weakening the bonds of terminal subunits by inducing or stabilizing their curvature (Gupta et al., 2013).Open in a separate windowFigure 2.Proteins that recognize curved αβ-tubulin tend to make long interfaces that span both α- and β-tubulin. (A) A stathmin family protein (blue) forms a long helix that binds two αβ-tubulin heterodimers (pink and green; PDB accession no. 3RYH). (B) The structure of a complex between kinesin-1 and αβ-tubulin (PDB accession no. 4HNA) is shown with the motor in dark green and αβ-tubulin in pink and lime. Depolymerizing kinesins have insertions (red segments modeled based on a crystal structure of MCAK; PDB accession no. 1V8K), such as the KVD finger, that expand the contact region compared with purely motile kinesins. (C) The TOG1 domain (blue) from Stu2, an XMAP215 family polymerase, contacts regions of α- and β-tubulin (pink and green) that move relative to each other in the curved (left, PDB accession no. 4FFB) and straight (right, model substituting straight αβ-tubulin; PDB accession no. 1JFF) conformations of αβ-tubulin. The asterisks show where this relative movement would disrupt the TOG–tubulin interface. Red side chains indicate conserved tubulin-binding residues at the top and bottom of the TOG domain. (D) The TOG2 domain from human CLASP1 (light blue, PDB accession no. 4K92) shows an “arched” interface that in docked models like the ones shown here is not complementary to curved (left) or straight (right) conformations of αβ-tubulin. Curved and straight structures are PDB 4FFB and 1JFF, respectively. Red side chains indicate binding residues similar to those in the polymerase family TOG domains, and asterisks highlight where the arched nature of this TOG prevents a conserved binding residue from contacting its interaction partner on β-tubulin.Kinesin-13s, first identified by their central motor domain (Aizawa et al., 1992; Wordeman and Mitchison, 1995), depolymerize microtubules catalytically using the energy of ATP hydrolysis (Hunter et al., 2003). Kinesin-13s depolymerize microtubules at spindle poles to generate poleward flux (Ganem et al., 2005), at kinetochores to drive anaphase chromosome segregation (Maney et al., 1998; Rogers et al., 2004), and in neuronal processes (Homma et al., 2003). Evidence that kinesin-13s depolymerized microtubules came from the discovery of the Xenopus laevis homologue, Xkcm1, in a screen for kinesin-related proteins involved in spindle assembly (Walczak et al., 1996). Incubation of Xkcm1, also known as MCAK, with GMPCPP microtubules caused peeled protofilaments and significant “ram’s horns” structures to appear at microtubule ends (Desai et al., 1999), which indicates that MCAK binds more tightly to curved structures than to straight ones. As with all kinesins, tight binding of the motor domain is coupled to its ATP hydrolysis cycle. Kinesin-13s first bind the microtubule lattice with an on-rate constant that strongly influences its depolymerase activity (Cooper et al., 2010). Kinesin-13s then target the end of the microtubule via “lattice diffusion,” a random walk mediated by electrostatic interactions that occurs in the ADP state (Helenius et al., 2006). Exchange of ADP to ATP occurs at microtubule ends; in the ATP state, MCAK binds tightly to tubulin dimers and either induces or stabilizes their outward curvature and detachment from the microtubule lattice (Friel and Howard, 2011). The subsequent hydrolysis of ATP causes kinesin-13 to release its tubulin subunit, now detached from the lattice, and begin another cycle of depolymerization (Moores et al., 2002).A distinguishing feature of the kinesin-13 motor domain is an extension of loop L2, known as the KVD finger (Ogawa et al., 2004; Shipley et al., 2004), which protrudes from the motor domain toward the minus end of the microtubule (Fig. 2 B). Alanine substitution of the KVD motif inhibits depolymerase activity in cell-based assays (Ogawa et al., 2004) and in vitro (Shipley et al., 2004). A recent cryo-EM study showed that the kinesin-13 motor domain contacts curved tubulin on three distinct surfaces (Asenjo et al., 2013) that differ from the contact surfaces of kinesin-1 (Sindelar and Downing, 2010; Gigant et al., 2013). The location of the kinesin-13 contact surfaces could allow kinesin-13 to stabilize spontaneous curvature of tubulin dimers at either microtubule end. Alternatively, tight binding of the kinesin-13 motor domain could directly induce curvature in the tubulin dimer. In either case, by promoting curvature at the growing microtubule end, kinesin-13s weaken the association of terminal subunits and induce catastrophes.Kinesin-8s are motile depolymerases (Gupta et al., 2006; Varga et al., 2006) that establish the length of microtubules in the mitotic spindle (Goshima et al., 2005; Rizk et al., 2014), position the spindle (Gupta et al., 2006), and modulate the dynamics of kinetochore microtubules (Stumpff et al., 2008; Du et al., 2010). Unlike the nonmotile kinesin-13s, whose motor domain is fully specialized for depolymerization, kinesin-8 proteins walk to the microtubule end and remove tubulin upon arrival (Gupta et al., 2006; Varga et al., 2006). Although it is unclear if depolymerase activity is fully conserved (Du et al., 2010; Mayr et al., 2011), all kinesin-8s combine motility with a negative effect on microtubule growth. For Saccharomyces cerevisiae Kip3p, the combination of motility and depolymerase activity has a significant functional consequence: Kip3p depolymerizes longer microtubules faster than shorter ones (Varga et al., 2006). This length-dependent depolymerization can be explained by an “antenna model.” In this model, longer microtubules will accumulate more kinesin-8s, which then walk toward the microtubule end, forming length-dependent traffic jams in some cases (Leduc et al., 2012). Because the rate of depolymerization depends on the number of kinesin-8s that arrive at the microtubule end, longer microtubules will be depolymerized more quickly. The “antenna model” depends critically on the high processivity of kinesin-8, which is thought to result from an additional C-terminal microtubule-binding element (Mayr et al., 2011; Stumpff et al., 2011; Su et al., 2011; Weaver et al., 2011); the C terminus may also contribute to a recently described microtubule sliding activity in Kip3p (Su et al., 2013). Intriguingly, a single Kip3p appears to be insufficient to remove a tubulin dimer. Rather, a second Kip3p must arrive at the microtubule end to bump off the first one (Varga et al., 2009).There are less structural and mutagenesis data available to explain the unique ability of kinesin-8s to walk and depolymerize. It is also not clear that all kinesin-8s use the same cooperative mechanism described for Kip3p. Like kinesin-13, the motor domain of kinesin-8 has an extended loop L2. This loop is disordered in the available crystal structure, but has been observed to contact α-tubulin in a cryo-EM reconstruction (Peters et al., 2010). The kinesin-8 loop L2 lacks a KVD sequence, however, and systematic mutations of L2 have not yet determined its role in depolymerase activity. The extent to which kinesin-8s recognize/induce curvature at microtubule ends remains unresolved. Truncated kinesin-8 motor domains can create small peels at the ends of GMPCPP microtubules (Peters et al., 2010), which suggests that kinesin-8 can induce or stabilize curvature. The fact that two kinesin-8s are required to dissociate a tubulin subunit, however, indicates that single motors alone do not substantially weaken the bonds holding the terminal tubulin subunit. Perhaps kinesin-8s do not stabilize curved forms of αβ-tubulin as strongly as kinesin-13s do.Reconstitution of microtubule dynamics in vitro showed that the depolymerizing kinesins affect catastrophe in different ways (Gardner et al., 2011b): kinesin-13s eliminate the aging process described earlier, whereas kinesin-8s accelerate it. Importantly, the local control of catastrophes by depolymerases is accomplished primarily through the local modulation of curvature at microtubule ends.

Growth-promoting MAPs also use conformation-selective interactions with αβ-tubulin

MAPs that accelerate growth or stabilize the microtubule lattice counteract microtubule depolymerases (Tournebize et al., 2000; Kinoshita et al., 2001). XMAP215 was discovered as the major protein in Xenopus extracts that promotes microtubule growth (Gard and Kirschner, 1987). Later, functional homologues were discovered in S. cerevisiae (Stu2p) (Wang and Huffaker, 1997) and other organisms (e.g., Charrasse et al., 1998; Cullen et al., 1999). XMAP215 family proteins localize to kinetochores and microtubule organizing centers, where they contribute to chromosome movements and to spindle assembly and flux (Wang and Huffaker, 1997; Cullen et al., 1999). Loss of XMAP215 family polymerase function leads to shorter, slower-growing microtubules and often gives rise to smaller and/or aberrant spindles (Wang and Huffaker, 1997; Cullen et al., 1999). All family members contain multiple TOG domains that bind αβ-tubulin (Al-Bassam et al., 2006; Slep and Vale, 2007). The molecular mechanisms underlying the activity of these proteins, and the collective action of their arrayed TOG domains, have until recently remained obscure. Recent progress is defining the structure and biochemistry of TOG domains and their interactions with αβ-tubulin. The emerging view is that XMAP215 family polymerases, like the depolymerases, bind to curved αβ-tubulin dimers as an important part of their biochemical cycle. In this section, we will focus on the most recent developments that are shaping the molecular understanding of growth-promoting MAPs, emphasizing the somewhat better studied XMAP215 family.Affinity chromatography using immobilized TOG domains from Stu2p revealed that the TOG1 domain binds directly to unpolymerized αβ-tubulin (Al-Bassam et al., 2006). TOG domains can also bind specifically to one end of the microtubule (Al-Bassam et al., 2006). Crystal structures of TOG domains, sequence conservation, and site-directed mutagenesis defined the αβ-tubulin–interacting surface, which forms a narrow “spine” of the book-shaped domain (Al-Bassam et al., 2007; Slep and Vale, 2007).In early models for XMAP215, the arrayed TOG domains were thought to bind multiple αβ-tubulins (Gard and Kirschner, 1987). Subsequent fluorescence-based reconstitution of XMAP215 activity, however, gave results that were not consistent with this “shuttle” model (Brouhard et al., 2008). The reconstitution assays showed that XMAP215 acted processively, residing at the microtubule end long enough to perform multiple rounds of αβ-tubulin addition. Intriguingly, XMAP215 increased the rate of, but not the apparent equilibrium constant for, microtubule elongation. XMAP215 also stimulated the rate of shrinkage in the absence of unpolymerized αβ-tubulin. Similar observations were made using Alp14 (Al-Bassam et al., 2012), a Schizosaccharomyces pombe XMAP215 homologue. These studies showed that XMAP215 catalyzes polymerization: it promotes microtubule growth by using its TOG domains to repeatedly bind and stabilize an intermediate state that otherwise limits the rate of polymerization.How do TOG domains recognize the microtubule end and promote elongation? Recent structural studies (Ayaz et al., 2012, 2014) suggest that interactions with curved αβ-tubulin play a central role. The crystal structures of complexes between αβ-tubulin and the TOG1 or TOG2 domains from Stu2p revealed that both TOG domains bind to curved αβ-tubulin (Ayaz et al., 2012, 2014; Fig. 2 C). The TOG domains do not interact strongly with microtubules even though the TOG-contacting epitopes are accessible on the microtubule surface (Ayaz et al., 2012). Preferential binding to curved αβ-tubulin (Ayaz et al., 2014) occurs because the arrangement of the TOG-contacting regions of α- and β-tubulin differs between curved and straight conformations (Fig. 2 C). Conformation-selective TOG–αβ-tubulin interactions explain how XMAP215 family proteins discriminate unpolymerized αβ-tubulin from αβ-tubulin in the body of the microtubule. XMAP215 family proteins require a basic region in addition to TOG domains for microtubule plus end association and polymerase activity (Widlund et al., 2011). The polarity of TOG–αβ-tubulin interactions and the ordering of domains in the protein together explain the plus end specificity of these polymerases: only at the plus end can TOGs engage curved αβ-tubulin while the C-terminal basic region contacts surfaces deeper in the microtubule (Ayaz et al., 2012). A recent study proposed that the linked TOG domains catalyze elongation using a tethering mechanism that effectively concentrates unpolymerized αβ-tubulin near curved subunits already bound at the microtubule end (Ayaz et al., 2014). The mechanisms by which these proteins catalyze depolymerization are less understood, although depolymerization can be explained by the catalytic stabilization of an intermediate state (Brouhard et al., 2008). By analogy with the depolymerases described earlier, the stabilization of such a state by arrayed TOG domains seems likely to also depend on the preferential interactions with curved αβ-tubulin.CLASP family proteins (Pasqualone and Huffaker, 1994; Akhmanova et al., 2001) also contain TOG domains, but they are used to different effect: CLASPs do not make microtubules grow faster but instead appear to regulate the frequencies of catastrophe and rescue. For example, in vitro reconstitutions using Cls1p, a CLASP protein from S. pombe, showed that Cls1p promoted rescue (Al-Bassam et al., 2010). CLASP family proteins also localize to kinetochores and contribute to spindle flux (Maiato et al., 2005). Loss of CLASP function affects microtubule stability and causes spindle defects (Akhmanova et al., 2001; Maiato et al., 2005), but does so without significantly affecting microtubule growth rates (Mimori-Kiyosue et al., 2006). CLASPs can also stabilize microtubule bundles/overlaps (Bratman and Chang, 2007). The recently published structure of a CLASP family TOG domain (Leano et al., 2013) provided an unexpected hint about a possible origin of the different activities. Indeed, the structure revealed significant differences with XMAP215 family TOG domains even though the CLASP TOG maintains evolutionarily conserved αβ-tubulin–interacting residues (Fig. 2 D). Whereas the αβ-tubulin binding surface of XMAP215 family TOGs is relatively flat, the equivalent surface of the CLASP TOG is arched in a way that appears to break the geometric match with curved αβ-tubulin (Leano et al., 2013; Fig. 2 D). This suggests that CLASP TOG domains might bind to an even more curved conformation of αβ-tubulin that has not yet been observed, that they do not simultaneously engage α- and β-tubulin, or that they do something else. It is not yet clear how these different possibilities might contribute to the rescue-promoting activity of CLASPs. However, even though the biochemical and structural understanding of how CLASP TOGs interact with αβ-tubulin is less advanced than for XMAP215 family TOGs, the conservation of critical αβ-tubulin–interacting residues makes it seem likely that conformation-selective interactions with αβ-tubulin will play a prominent role.The modulation of microtubule dynamics by XMAP215/CLASP family proteins ensures proper microtubule function in both interphase and dividing cells. As for the depolymerases, specific interactions with curved αβ-tubulin likely underlie the different regulatory activities of XMAP215/CLASP family proteins.

Sensing conformation at lattice contacts

Thus far, we have described how microtubule polymerases and depolymerases bind selectively to curved conformations of the αβ-tubulin dimer. These interactions play a significant role in the movement of tubulin dimers into and out of the microtubule polymer. Once in the polymer, αβ-tubulin dimers make contacts with neighboring tubulins. Recently, three MAPs were shown to bind microtubules at lattice contacts: (1) the Ndc80 complex, a core kinetochore protein; (2) doublecortin (DCX), a neuronal MAP; and (3) EB1, the canonical end-binding protein. Here we will summarize recent progress demonstrating how these proteins recognize distinctive features of lattice contacts.The Ndc80 complex is a core component of the kinetochore–microtubule interface (Janke et al., 2001; Wigge and Kilmartin, 2001; McCleland et al., 2003), forming a “sleeve” that connects the outer kinetochore to microtubules of the mitotic spindle (Cheeseman et al., 2006; DeLuca et al., 2006). Loss of Ndc80 function leads to chromosome segregation errors in mitosis (McCleland et al., 2004; DeLuca et al., 2005). Ndc80 binds to microtubules at the longitudinal interface between α- and β-tubulin and extends outward toward the plus end at an ∼60° angle (Cheeseman et al., 2006; Wilson-Kubalek et al., 2008). Ndc80 binds to both the intradimer and interdimer interface and forms oligomeric arrays (Alushin et al., 2010). The binding of Ndc80 to this longitudinal lattice contact may confer a preference for straight rather than curved microtubule lattices, because the shape of the Ndc80 binding site is expected to change as a protofilament bends (Alushin et al., 2010; Fig. 3 A). Preferential binding to straight protofilaments might allow the Ndc80 complex to remain attached to the end of a shrinking microtubule. Indeed, reconstitutions of the Ndc80 complex interacting with dynamic microtubules show that the curved shrinking end acts as a “reflecting wall,” giving rise to “biased diffusion” (Powers et al., 2009). Interestingly, the Ndc80 complex also promotes rescue (Umbreit et al., 2012), and selective binding to straight lattice contacts may contribute to this rescue activity.Open in a separate windowFigure 3.Proteins that bind microtubules can distinguish unique configurations at lattice contacts. (A) Ndc80 (light and dark blue) binds the contact within (dark blue) and between (light blue) αβ-tubulin heterodimers (pink and green). The left shows part of an Ndc80 array on straight protofilaments (PDB accession no. 3IZ0). The right shows that neighboring Ndc80 molecules clash when modeled onto a curved protofilament. Individual Ndc80s may read the conformation at a single joint, or the change in conformation may disrupt cooperative interactions between adjacent Ndc80s. (B) Two views of DCX (blue) binding a lattice contact at the vertex of four αβ-tubulins, PDB accession no. 4ATU. Cooperative interactions on the microtubule allow DCX to discriminate between the subtle changes that accompany different protofilament numbers (11: orange, EMDataBank [EMD] accession no. 5191; 13: red, EMD accession no. 5193; 15: yellow, EMD accession no. 5195). (C) EB1 (left, dark blue) binds at the same vertex as DCX (PDB accession no. 4AB0), but EB1 binds preferentially to GTP vertices over GDP vertices, and is not sensitive to protofilament number. The same section of microtubule with EB1 removed (right) shows the location of nucleotide-dependent changes at the four-way vertex: helix H3 of β-tubulin (red patch at the lower right of the four-way junction), and the intermediate (Int.) domain of α-tubulin (yellow patch at the top left of the four-way junction). pfs, protofilaments.DCX, a MAP expressed in developing neurons (Francis et al., 1999; Gleeson et al., 1999) and mutated in cases of subcortical band heterotopia (des Portes et al., 1998; Gleeson et al., 1998), is unique in its ability to bind specifically to 13-protofilament microtubules over other protofilament numbers (Moores et al., 2004; Fig. 3 B). DCX contains two nonidentical, microtubule-binding “DC” domains (Taylor et al., 2000) that share a ubiquitin-like fold (Kim et al., 2003). A cryo-EM reconstruction showed that a single DC domain binds to microtubules at the vertex of four tubulin dimers in the so-called “B” lattice configuration (Fourniol et al., 2010). The DCX binding site is ideally situated to detect the subtle changes at lattice contacts that result from different protofilament numbers, which range from 11 to 16 for mammalian microtubules (Sui and Downing, 2010). Despite their ideal location, protofilament preference is not a property of single DCX molecules. Rather, it is cooperative interactions between neighboring DCX molecules that are sensitive to the spacing between protofilaments (Bechstedt and Brouhard, 2012). In vitro, this selectivity enables DCX to nucleate homogeneous, 13-protofilament microtubules (Moores et al., 2004). The function of DCX in developing neurons remains unclear, with models ranging from microtubule stabilization (Gleeson et al., 1999) to regulation of kinesin traffic (Liu et al., 2012).EB1, the canonical end-binding protein (Morrison et al., 1998), uses its calponin homology (CH) domain (Hayashi and Ikura, 2003) to bind the same lattice contact as DCX (Maurer et al., 2012). EB1 forms “comets” by binding rapidly and tightly to a distinct feature at the growing microtubule end but only weakly to the “mature” lattice (Bieling et al., 2007). Recent work has defined this distinctive feature as the nucleotide state. EB1 binds preferentially to microtubules built from GTP analogues (Zanic et al., 2009; Maurer et al., 2011). Combined with careful analysis of the size, shape, and dynamics of EB1 comets (Bieling et al., 2007), these results established that EB1 recognizes microtubule ends by binding specifically to the “GTP cap,” which is an extended region of the microtubule end that is enriched with GTP- and GDP-Pi-tubulin dimers. A recent cryo-EM reconstruction of the CH domain of Mal3 (the S. pombe EB1) bound to GTPγS microtubules provided a possible structural mechanism for how EB1 might differentiate GTP from GDP lattices (Maurer et al., 2012; Fig. 3 C). Mal3 was observed to contact helix H3 of β-tubulin, which connects directly to the exchangeable nucleotide-binding site. EB1 also contacts the regions of α-tubulin that move during the compaction of the lattice that follows GTP hydrolysis (Alushin et al., 2014). Mutation of conserved EB1 residues that contact either helix H3 or the compacting region of α-tubulin disrupts the end-tracking behavior of EB1 (Slep and Vale, 2007; Maurer et al., 2012). Interactions with helix H3 and the compacting region of α-tubulin also enable EB1 to accelerate the transitions of tubulin from the GTP state to the GDP state; in other words, EB1 acts as a “maturation factor” for the microtubule end (Maurer et al., 2014). EB1 recruits a large network of plus-end-tracking proteins (Akhmanova and Steinmetz, 2008) through interactions with the EB1 C terminus (Hayashi et al., 2005; Honnappa et al., 2006) and EB1 homology domain (Honnappa et al., 2009). This diverse and complex protein network is essential for the regulation of microtubule dynamics, the capture of microtubule ends by the cell cortex (Kodama et al., 2003) and endoplasmic reticulum (Grigoriev et al., 2008), and the positioning of the mitotic spindle (Liakopoulos et al., 2003).As mentioned earlier, microtubule ends also show unique structural configurations, namely tapered, outwardly flared, and flattened structures collectively described as “sheets” (Chrétien et al., 1995). The sheets contain distinctive lattice contacts, and recent work shows that the microtubule-binding activities of DCX and EB1 are sensitive to these structural features. DCX, for example, binds specifically to the outwardly flared sheets (Bechstedt et al., 2014), which enables DCX to track microtubule ends. Evidence for the ability of EB1 to recognize or control a distinct lattice configuration comes from the reconstitutions showing that EB1 promotes elongation synergistically with XMAP215 (Zanic et al., 2013): lack of a detectable direct EB1–XMAP215 interaction suggested that the observed synergy was mediated through alterations of the microtubule end structure itself. Further evidence that EB1 can affect the structure of the microtubule lattice comes from data showing that EB1 can nucleate “A” lattice microtubules in vitro (des Georges et al., 2008) and influence protofilament number distributions (Vitre et al., 2008; Maurer et al., 2012). The connection between the structure of microtubule ends, their nucleotide state, and microtubule dynamics is an important open question.

Conclusions and outlook

The αβ-tubulin dimer adopts a range of conformations as it moves in and out of the microtubule polymer, including changes to its intrinsic curvature and changes to its lattice contacts. These different conformations affect microtubule dynamics by altering the strength of lattice association and the rate of GTP hydrolysis. The work we discussed here has revealed an intimate linkage between these different conformations and the activities of key proteins that regulate microtubule dynamics. It is now clear that selective interactions with distinct conformations of unpolymerized and polymerized αβ-tubulin define the cell physiology of the microtubule cytoskeleton. Recently developed methods for purifying or overexpressing αβ-tubulin (des Georges et al., 2008; Johnson et al., 2011; Widlund et al., 2012; Minoura et al., 2013) are facilitating structural studies and allowing the biochemistry of αβ-tubulin polymerization to be dissected in unprecedented detail. Microtubule structural biology is entering a golden age, where the pace of new structural information is accelerating. We anticipate that future crystallographic and high-resolution cryo-EM studies will define the strategies used by other MAPs to recognize and control the conformation of αβ-tubulin, and may reveal new conformations of αβ-tubulin inside and outside of the microtubule. Reconstitutions of microtubule dynamics are rapidly increasing in complexity and are beginning to reveal how the activities of multiple MAPs can reinforce or antagonize each other (Zanic et al., 2013). More complex reconstitutions are also defining the minimal requirements for creating cellular-scale structures like the mitotic spindle (Bieling et al., 2010; Subramanian et al., 2013). Reconstitutions will also greatly advance the understanding of the dynamics and regulation of microtubule minus ends. As the ever-advancing structural data are integrated with reconstitution data, incorporated into computational models, and correlated with cell biology experiments, a robust, multiscale understanding of microtubule biology will come within reach.     

The Modest Beginnings of One Genome Project

One of the top things on a geneticist’s wish list has to be a set of mutants for every gene in their particular organism. Such a set was produced for the yeast, Saccharomyces cerevisiae near the end of the 20th century by a consortium of yeast geneticists. However, the functional genomic analysis of one chromosome, its smallest, had already begun more than 25 years earlier as a project that was designed to define most or all of that chromosome’s essential genes by temperature-sensitive lethal mutations. When far fewer than expected genes were uncovered, the relatively new field of molecular cloning enabled us and indeed, the entire community of yeast researchers to approach this problem more definitively. These studies ultimately led to cloning, genomic sequencing, and the production and phenotypic analysis of the entire set of knockout mutations for this model organism as well as a better concept of what defines an essential function, a wish fulfilled that enables this model eukaryote to continue at the forefront of research in modern biology.THE yeast Saccharomyces cerevisiae genome project culminated with the first sequenced eukaryotic genome (Goffeau et al. 1996) and was followed by the functional analysis of almost all of its ∼6000 genes. This project produced a near complete collection of deletion mutants that among other things attempted to define the number of genes that were essential for growth of this organism in the laboratory (Winzeler et al. 1999; Giaever et al. 2002). These projects had their roots in earlier studies that began with Carl Lindegren’s first genetic map (Lindegren 1949; Lindegren et al. 1959), continued with the herculean efforts of Robert Mortimer and colleagues who compiled data from hundreds of investigators who had mapped genes (Mortimer and Hawthorne 1975; Mortimer and Schild 1980, 1985; Mortimer et al. 1989, 1992), and continues today through the Saccharomyces Genome Database led by Mike Cherry (Cherry et al. 1997, 2012) that curates genomic and related information.This is the story of the first functional genomic analysis of yeast, which began with our investigation of chromosome I, the organism’s smallest. It is worth telling because it traces its routes to the “phage school” and shows how what was deemed non-hypothesis–driven research and data collection were actually tied to a fundamental question and indeed to several hypotheses. It is also about how sharing information, unselfish donation of materials, and true collaboration within the yeast community enabled this organism to rise to the forefront of molecular biology. It is told with the idea that it really does take a village to decipher a genome.     

Maturation Stress Generation in Poplar Tension Wood Studied by Synchrotron Radiation Microdiffraction

Tension wood is widespread in the organs of woody plants. During its formation, it generates a large tensile mechanical stress, called maturation stress. Maturation stress performs essential biomechanical functions such as optimizing the mechanical resistance of the stem, performing adaptive movements, and ensuring long-term stability of growing plants. Although various hypotheses have recently been proposed, the mechanism generating maturation stress is not yet fully understood. In order to discriminate between these hypotheses, we investigated structural changes in cellulose microfibrils along sequences of xylem cell differentiation in tension and normal wood of poplar (Populus deltoides × Populus trichocarpa ‘I45-51’). Synchrotron radiation microdiffraction was used to measure the evolution of the angle and lattice spacing of crystalline cellulose associated with the deposition of successive cell wall layers. Profiles of normal and tension wood were very similar in early development stages corresponding to the formation of the S1 and the outer part of the S2 layer. The microfibril angle in the S2 layer was found to be lower in its inner part than in its outer part, especially in tension wood. In tension wood only, this decrease occurred together with an increase in cellulose lattice spacing, and this happened before the G-layer was visible. The relative increase in lattice spacing was found close to the usual value of maturation strains, strongly suggesting that microfibrils of this layer are put into tension and contribute to the generation of maturation stress.Wood cells are produced in the cambium at the periphery of the stem. The formation of the secondary wall occurs at the end of cell elongation by the deposition of successive layers made of cellulose microfibrils bounded by an amorphous polymeric matrix. Each layer has a specific chemical composition and is characterized by a particular orientation of the microfibrils relative to the cell axis (Mellerowicz and Sundberg, 2008). Microfibrils are made of crystalline cellulose and are by far the stiffest constituent of the cell wall. The microfibril angle (MFA) in each layer is determinant for cell wall architecture and wood mechanical properties.During the formation of wood cells, a mechanical stress of a large magnitude, known as “maturation stress” or “growth stress” (Archer, 1986; Fournier et al., 1991), occurs in the cell walls. This stress fulfills essential biomechanical functions for the tree. It compensates for the comparatively low compressive strength of wood and thus improves the stem resistance against bending loads. It also provides the tree with a motor system (Moulia et al., 2006), necessary to maintain the stem at a constant angle during growth (Alméras and Fournier, 2009) or to achieve adaptive reorientations. In angiosperms, a large tensile maturation stress is generated by a specialized tissue called “tension wood.” In poplar (Populus deltoides × Populus trichocarpa), as in most temperate tree species, tension wood fibers are characterized by the presence of a specific layer, called the G-layer (Jourez et al., 2001; Fang et al., 2008), where the matrix is almost devoid of lignin (Pilate et al., 2004) and the microfibrils are oriented parallel to the fiber axis (Fujita et al., 1974). This type of reaction cell is common in plant organs whose function involves the bending or contraction of axes, such as tendrils, twining vines (Bowling and Vaughn, 2009), or roots (Fisher, 2008).The mechanism at the origin of tensile maturation stress has been the subject of a lot of controversy and is still not fully understood. However, several recent publications have greatly improved our knowledge about the ultrastructure, chemical composition, molecular activity, mechanical state, and behavior of tension wood. Different models have been proposed and discussed to explain the origin of maturation stress (Boyd, 1972; Bamber, 1987, 2001; Okuyama et al., 1994, 1995; Yamamoto, 1998, 2004; Alméras et al., 2005, 2006; Bowling and Vaughn, 2008; Goswami et al., 2008; Mellerowicz et al., 2008). The specific organization of the G-layer suggests a tensile force induced in the microfibrils during the maturation process. Different hypotheses have been proposed to explain this mechanism, such as the contraction of amorphous zones within the cellulose microfibrils (Yamamoto, 2004), the action of xyloglucans during the formation of microfibril aggregates (Nishikubo et al., 2007; Mellerowicz et al., 2008), and the effect of changes in moisture content stimulated by pectin-like substances (Bowling and Vaughn, 2008). A recent work (Goswami et al., 2008) argued an alternative model, initially proposed by Münch (1938), which proposed that the maturation stress originates in the swelling of the G-layer during cell maturation and is transmitted to the adjacent secondary layers, where the larger MFAs allow an efficient conversion of lateral stress into axial tensile stress. Although the proposed mechanism is not consistent with the known hygroscopic behavior of tension wood, which shrinks when it dries and not when it takes up water (Clair and Thibaut, 2001; Fang et al., 2007; Clair et al., 2008), this hypothesis focused attention on the possible role of cell wall layers other than the G-layer. As a matter of fact, many types of wood fibers lacking a G-layer are known to produce axial tensile stress, such as normal wood of angiosperms and conifers (Archer, 1986) and the tension wood of many tropical species (Onaka, 1949; Clair et al., 2006b; Ruelle et al., 2007), so that mechanisms strictly based on an action of the G-layer cannot provide a general explanation for the origin of tensile maturation stress in wood.In order to further understanding, direct observations of the mechanical state of the different cell wall layers and their evolution during the formation of the tension wood fibers are needed. X-ray diffraction can be used to investigate the orientation of microfibrils (Cave, 1966, 1997a, 1997b; Peura et al., 2007, 2008a, 2008b) and the lattice spacing of crystalline cellulose. The axial lattice spacing d₀₀₄ is the distance between successive monomers along a cellulose microfibril and reflects its state of mechanical stress (Clair et al., 2006a; Peura et al., 2007). If cellulose microfibrils indeed support a tensile stress, they should be found in an extended state of deformation. Under this assumption, the progressive development of maturation stress during the cell wall formation should be accompanied by an increase in cellulose lattice spacing. Synchrotron radiation allows a reduction in the size of the x-ray beam to some micrometers while retaining a strong signal, whereby diffraction analysis can be performed at a very local scale (Riekel, 2000). This technique has been used to study sequences of wood cell development (Hori et al., 2000; Müller et al., 2002). In this study, we report an experiment where a microbeam was used to analyze the structural changes of cellulose in the cell wall layers of tension wood and normal wood fibers along the sequence of xylem cell differentiation extending from the cambium to mature wood (Fig. 1). The experiment was designed to make this measurement in planta, in order to minimize sources of mechanical disturbance and be as close as possible to the native mechanical state (Clair et al., 2006a). The 200 and 004 diffraction patterns of cellulose were analyzed to investigate the process of maturation stress generation in tension wood.Open in a separate windowFigure 1.Schematic of the experimental setup, showing the x-ray beam passing perpendicular to the longitudinal-radial plane of wood and the contribution of the 004 and 200 crystal planes to the diffraction pattern recorded by the camera. [See online article for color version of this figure.]     

Permeation Redux: Thermodynamics and Kinetics of Ion Movement through Potassium Channels

The fundamental biophysics underlying the selective movement of ions through ion channels was launched by George Eisenman in the 1960s, using glass electrodes. This minireview examines the insights from these early studies and the explosive progress made since then.The recent passing of George Eisenman (December 18, 2013) inspired us to revisit the topic most associated with his passionate input, namely how the membrane proteins known as ion channels control passive movements of ions across biological membranes. Ion permeation has captivated biophysicists for more than half a century, and only now, with the combined advent of atomic-level structures and sophisticated computational wizardry, are the secrets of this amazing process beginning to be revealed. Why “amazing”? For example, because K⁺-selective ion channels can discriminate between K⁺ and Na⁺ ions, which differ in radius by a mere 0.38 Ångstrom, and do so with 1000:1 reliability and at lightning speed near the diffusion limit, the dwell time of an ion in the pore of a channel is as fleeting as ∼10⁻⁸ s. Understanding this remarkably-tuned process in K⁺ channels requires attention to two perspectives: the ability of specific channels to discriminate between the ions they might encounter (i.e., selectivity); and the kinetics of ion movement across the channel pore (i.e., conduction).The classical thermodynamic explanation of ion selectivity is that the relative free energy difference of ions in the pore relative to the bulk solution is the critical quantity to consider (1–4). Some of the earliest insights into thermodynamic selectivity derive from studies of ion binding to aluminosilicate glass electrodes (5,6). Depending on the composition of the glass, these electrodes, originally developed for their proton sensitivity, can exhibit a dramatic range of selectivities among the five alkali metal cations. In rank order, one might expect as many as 5 × 4 × 3 × 2 × 1 = 120 different sequences of selectivities among these five cations. Remarkably, however, in the vast literature of selectivity in biological membranes, typically only 11 sequences are observed (with some exceptions). These became known as the “Eisenman sequences”. The exact same selectivity sequences are observed in glass electrodes of various compositions.Why are the free energy differences the way they are for a given system? To answer this question, one needs a physical mechanism. For Eisenman, numerical calculations stood as a critical component of the process of better understanding Nature. In other words, proposing a physical mechanism that is qualitatively reasonable is not enough—one must also test it by constructing atomic models leading to actual quantitative predictions (Fig. 1). In the early days, the concept of the anionic field strength of a binding site was formulated and tested with direct calculations based on exceedingly simple atomic hard-sphere models of ions, water molecules, and coordinating ligands such as shown in Fig. 1 A (2,5). Remarkably, these simple calculations led to the Eisenman selectivity sequences. Eisenman was able to account for the limited class of sequences by considering the equilibrium binding of cations to the glass, and the energetic competition between water and glass for the ions. The critical factor that determines the selectivity sequence of a given glass is the anionic field strength of the binding site on the glass. Briefly, the smallest group Ia cation, Li⁺, holds water most tenaciously, so it will only dehydrate and bind in the presence of a strongly negative electrostatic potential.Open in a separate windowFigure 1Structural models used in theoretical studies of ion selectivity. (A) Simple model used to introduce the concept of field strength leading to 11 cationic selectivity sequences (2,5,6). Ions, water, and ligands are represented by simple hard-spheres with embedded point charges. Selectivity arises from the difference in the interaction energy of the cation with a water molecule (top) and an anionic coordinating ligand (bottom). (B) Ion-selective transfer process is depicted with atomic models incorporating all molecular details in the case of solvation in liquid water (top) and binding to the K⁺-selective ionophore valinomycin (bottom). Such atomic models were used to carry out some of the earliest MD free energy simulations on ion binding selectivity (12,13,15).By contrast, the largest cation, Cs⁺, holds water least tenaciously. It cannot bind readily to a strongly negative site because the site itself greedily clings to water molecules, and thus prevents Cs⁺ binding. However, Cs⁺ is more willing, relative to the smaller cations, to dehydrate and bind in the presence of a weakly negative electrostatic potential. At the extremes, the highest anionic field strength glass shows a selectivity sequence ofLi⁺ > Na⁺ > K⁺ > Rb⁺ > Cs⁺(sequence XI), and the lowest anionic field strength glass shows a selectivity sequence ofCs⁺ > Rb⁺ > K⁺ > Na⁺ > Li⁺(sequence I).A very simple model, based on the relative Gibbs’ free energies of binding and hydration, explains why there are only 11 sequences (5–7). The critical factor underlying the pattern of these selectivity sequences is that the “ion-site interaction energies fall off as a function of cation size as a lower power of the cation radius than do ion-water interaction energies” (5,6). The icing on the cake is that ion selectivity of channels in membranes appears to follow similar principles (7). The thermodynamic principles are evidently analogous. Moreover, Eisenman’s contributions went far beyond the monovalent cation selectivity of potassium channels. His theoretical approach was seminal in understanding both cation and anion selectivity in a diverse range of physical and biological systems (8,9).The advent of molecular dynamics (MD) simulations around this period was of critical importance to the field. This made it possible to construct increasingly realistic models of proteins (10), including ion channels (11), and examine the ion selectivity of carriers using the alchemical free energy perturbation (FEP) technique (12,13). With no experimental structures yet available for the ion-selective regions of biological K⁺ channels, an important step forward was Eisenman’s realization that other ion-selective systems could be used to computationally test the structural basis of his selectivity theory. Both peptidelike small ionophores, such as valinomycin and nonactin, and the ion-coordinating fivefold symmetry sites in icosahedral virus structures, thus caught his attention (13). As it turned out, these types of structures were indeed very relevant for the selectivity problem, because K⁺-channel filters were eventually shown to be lined likewise by carbonyl groups (14). With the crystallographically determined valinomycin structure at hand, its selectivity could be energetically analyzed by atomistic computer simulations, as illustrated in Fig. 1 B (12,15). The anionic field strength (represented by the carbonyl ligand dipole moment) could then be varied artificially, and the successive progression through the different selectivity sequences, as a function of field strength, directly observed. Likewise, Eisenman and Alvarez (13) made computational predictions for the binding energetics and selectivity of the Ca²⁺ binding site at the fivefold symmetry axis of satellite tobacco necrosis virus, and they subsequently showed experimentally that this binding site had a marked rare-earth ion size selectivity (16). To this day, the general computational FEP/MD framework based on equilibrium thermodynamics used in these studies continues to be a critical tool to understand ion channels (17), transporters (18), and pumps (19).Despite these early insights, it was always clear to Eisenman that explanations of selectivity solely based on thermodynamic equilibrium were too simple to account for the detailed properties observed in biological systems. Since the halcyon days of equilibrium binding studies on glass electrodes, the permeation landscape presented by the pores of ion channels has emerged as richer than anticipated. One important realization is that binding and conduction of ions through a channel may act as contradictory processes, because although an ion has to leave the comfort of its hydration shells to selectively enter the mouth of a channel pore, if it binds the channel too tightly, it cannot move rapidly through it. This mini-conundrum is most apparent, perhaps, for K⁺-channels, which attract K⁺ ions much more forcefully than Na⁺ ions, yet conduct K⁺ ions much faster than Na⁺ ions.Another factor evident in early studies of permeation is that ions encounter a series of obstacles (i.e., energy barriers) and binding sites (i.e., energy wells) as they wend their way through the pore. One approach to understanding permeation is to consider that ions hopscotch from one well to the next over a series of barriers. When the number of barriers is rather limited, say <5, one can use so-called “rate theory” (20) to analyze and formulate the free energy profile experienced by an ion crossing the membrane. Hille (21) proposed that selectivity derives largely from the selectivities of the barriers, not the wells. Eisenman and Horn (7) later considered the possibility that binding sites and barriers within a particular channel might have different selectivity sequences. For example, if a channel presents two barriers, one of which has selectivity sequence I and the other has selectivity sequence XI, the channel as a whole will have an intermediate selectivity sequence that is not an Eisenman sequence at all. Rather, it is a so-called “polarizability sequence” (7). Interestingly, contemporary studies indicate that successive binding sites along K⁺-selective channels display different selectivities (22). Another concept based on Eyring barrier models is that the energy levels for wells and barriers may not be static, and may therefore fluctuate on a timescale relevant to ion permeation (23). Finally, the biophysics of ion permeation and later structural studies show that multiple ions may cohabit the same channel simultaneously, and the interactions among these ions have profound consequences for ion conduction and selectivity.Fast forward to the 21st century: atomic-level structures and all-atom simulations seem to have blown the permeation field wide open, as suggested by recent reviews (24–27). Once the KcsA channel structure was solved (14), the structural origin of K⁺-ion permeation could finally be addressed by computer simulations of the “real structure” and a number MD simulation studies provided novel insight (22,28–31). Needless to say, George Eisenman took great interest in these simulations even though he had by then retired. Also, in the case of KcsA, the initial work largely revolved around calculations of equilibrium ion binding and selectivity, barrier heights, and energy landscape mapping (22,31), because direct all-atom simulations of spontaneous permeation were not possible. However, the general type of knock-on mechanism with multiion occupancy of the channel selectivity filter, involving key distinct states (22,31), and a surprisingly flat energy landscape (22), appear to be robust features of these channels.Even with the advent of MD simulations, the concept of field strength has kept its relevance. For example, the selectivity filter in MD simulations of the KcsA channel displayed a range of atomic flexibility that seemed somewhat shocking at the time because a traditional host-guest mechanism of selectivity would require a fairly rigid cavity-size. Yet, free energy computations indicated that this was not strictly necessary to establish the thermodynamic free energy differences needed to support ion selectivity (32). The resilience of Eisenman’s ideas is not entirely surprising because, as foreseen early on by Bertil Hille (21), the concept of field strength remains “useful if the dipoles of the channel are free to move and can be pulled in by small ions and pushed back by large ones”.Nevertheless, despite the exciting progress, the chapter on ion selectivity in K⁺ channels is far from closed. Very recently, a number of studies have revealed some extremely intriguing multiion aspects of selectivity in K⁺ channels that appear to stand squarely outside the realm of equilibrium thermodynamics. By examining the properties of MthK (33) and NaK (34) mutants, Liu and Lockless (35) and Sauer et al. (36) showed that the channel becomes K⁺-selective only if there are four consecutive binding sites along the filter. This has culminated more recently with studies of two engineered mutants of the NaK channel, referred to as “NaK2K” and “NaK2CNG”. According to reversal potential measurements from single-channel electrophysiology, the NaK2K construct is K⁺-selective and the NaK2CNG construct is nonselective. Remarkably, despite being nonselective in ion permeation, the NaK2CNG filter displays an equilibrium preference for binding K⁺ over Na⁺, as indicated by measurements with isothermal titration calorimetry and concentration-dependent ion replacement within the filter observed through crystallographic titration experiments.K⁺-selective channels bind two or more K⁺ ions in the narrow filter, whereas the nonselective channels bind fewer ions. Based on the crystallographic titration experiments, the NaK2K construct has two high-affinity K⁺ sites whereas the NaK2CNG construct has only one K⁺-selective site. These experiments show that both K⁺-selective and nonselective channels select K⁺ over Na⁺ ions at equilibrium, implying that equilibrium selectivity is insufficient to determine the selectivity of ion permeation (35,36). The data indicate that having multiple K⁺ ions bound simultaneously is required for selective K⁺ conduction, and that a reduction in the number of bound K⁺ ions destroys the multiion selectivity mechanism utilized by K⁺ channels. Although these experimental results are intriguing, the underlying microscopic mechanisms remain unclear. The implication is that the multiion character of the permeation process must, somehow, be a critical element for establishing selective ion conduction through K⁺ channels.The progress made, and the challenges that remain, are perhaps best illustrated by returning to computational studies of the simplest membrane spanning structure known, namely the gramicidin A channel. Before detailed studies of selectivity and conductance of K⁺-channels were launched, computational work on ion conduction through membrane channels was largely focused on this simple channel (37–41). In this case the permeation selectivity was monotonically size-dependent (Eisenman sequence I) and, in this respect, less interesting than K⁺-selective channels. However, from an energetic point of view it was puzzling how this single helical structure could yield free energy barriers low enough to permit high conductivity (7,42). Computer simulations of increasing complexity in this case established that the combined effect of several contributions to ion stabilization along the pore (from the protein, membrane, single-file waters, and bulk solution) indeed results in low barriers to permeation (11,39,40). Furthermore, the most realistic model comes in close agreement with experimental measurements (11,43), although it is clear that work is still needed.     

Interactions Among Flower-Size QTL of Mimulus guttatus Are Abundant but Highly Variable in Nature

The frequency and character of interactions among genes influencing complex traits remain unknown. Our ignorance is most acute for segregating variation within natural populations, the epistasis most relevant for quantitative trait evolution. Here, we report a comprehensive survey of interactions among a defined set of flower-size QTL: loci polymorphic within a single natural population of yellow monkeyflower (Mimulus guttatus). We find that epistasis is typical. Observed phenotypes routinely differ from those predicted on the basis of direct allelic affects in the isogenic background, although the direction of deviations is highly variable. Across QTL pairs, there are significantly positive and negative interactions for every trait. Across traits, specific locus pairs routinely exhibit both positive and negative interactions. There was a tendency for negative epistasis to accompany positive direct effects and vice versa for the trait of corolla width, which may be due, at least in part, to the fact that QTL were identified from their direct effects on this trait.EPISTASIS contributes significantly to intrapopulation variation in floral morphology, development time, and male fitness components of Mimulus guttatus (Kelly 2005). The aggregate effect of interactions among QTL substantially alters the resemblance of relatives and phenotypic response to inbreeding. However, previous experiments did not identify the specific character of interactions between QTL. For example, it is not clear whether interactions change the rank order of QTL genotypes. If the direction of allelic effect changes with genetic background, so-called “sign epistasis” (Weinreich et al. 2005), the same selection pressure may favor different alleles in different genomic contexts, e.g., different subpopulations of a species (De Brito et al. 2005). When the trait is fitness, these kinds of interactions naturally generate peaks and valleys in genotypic fitness landscapes (Wright 1932; Burch and Chao 2000).Sign epistasis can involve a reversal of allelic effect at one or both loci of an interacting pair. Poelwijk et al. (2007) define the double reversal as “reciprocal sign epistasis” and contrast sign epistasis generally to “magnitude epistasis” where the magnitude but not the direction of allelic effects changes with genetic background. Magnitude epistasis includes synergistic interactions (alleles have greater effect in combination than individually) and less-than-additive or diminishing returns interactions (alleles have lesser effect in combination) (see Crow and Kimura 1970). Alternatively, one can classify interactions as positive or negative (Phillips et al. 2000)—positive if the observed phenotype of an allelic combination exceeds that predicted from direct effects at each locus, and negative if the phenotype of the combination is less than the additive prediction. Unfortunately, there is no simple logical mapping from the magnitude/sign epistasis classification to the positive/negative classification, nor to epistasis in the classical sense (Bateson 1909), wherein one locus masks the effect of another. The taxonomy of epistasis is further complicated by dominance (Routman and Cheverud 1997), higher-order interactions (Templeton 2000), and environmental dependencies (Brock et al. 2010).Molecular genetic studies provide clear examples of sign epistasis. Here, the interacting polymorphisms are often within the same gene. For example, the stability of RNA secondary structures requires matching of nucleotides at different positions. Whether a particular nucleotide change increases or reduces stability depends entirely on the identity of the nucleotide at its paired site (Chen et al. 1999). With “compensatory evolution” (Moore et al. 2000), a mutation that is neutral or detrimental in the original genetic background becomes advantageous by compensating for some other mutation that has recently fixed or at least become prevalent within the population. Mutations conferring antibiotic resistance often have deleterious side effects that reduce bacterial fitness in the absence of the drug. These side effects are attenuated by secondary mutations that are often detrimental in the original genotype (Levin et al. 1997; Schrag et al. 1997).Despite the progress in research on microbes (Weinreich et al. 2005; Elena et al. 2010), we currently know little about the prevalence or nature of epistasis for quantitative traits (Carlborg and Haley 2004) and particularly its impact on standing (segregating) variation within natural populations. Eshed and Zamir (1996) found extensive epistasis among QTL in Lycopersicon (tomato), but interactions primarily influenced the magnitude of single-locus effects and not their direction. In contrast, Kroymann and Mitchell-Olds (2005) documented a case of QTL effect reversal in Arabidopsis thaliana. The high allele for biomass accumulation in the Ler-0 accession becomes the low allele when introgressed into another line (the Col-0 accession). Patterns of gene sequence variation suggest that this polymorphism is maintained by balancing selection. In Avena barbata, two loci with negligible average effects exhibit sign epistasis for fitness in a cross between mesic and xeric genotypes (Latta et al. 2010).Even with these examples, it is difficult to evaluate the quantitative frequency of epistasis or regularities in its nature from the current literature. In part, this is because discovery of QTL×QTL interactions is typically idiosyncratic. In a segregating mapping population such as F₂’s or recombinant inbred lines, there is limited replication of particular multi-locus genotypes. As a consequence, it is difficult to accurately estimate the mean phenotype of any particular multi-locus genotype. Also, there are an enormous number of pairwise tests in a full simultaneous scan, which leads to very stringent significance levels (see box 2 of Carlborg and Haley 2004). Thus, while genomic scans have successfully identified interactions (e.g., Li et al. 1997; Cheverud 2000; Montooth et al. 2003), it is hard to know if the many nonsignificant interactions are due to absence of effect or absence of power.Our intention in this study was to conduct a comprehensive survey of interactions among a specific set of monkeyflower QTL. These loci are polymorphic within a single contiguous natural population. We first mapped flower-size QTL within nearly isogenic lines (NILs). Measurements on the NILs isolate the “direct effect” of each QTL, i.e., how the polymorphism affects phenotype in a single uniform genetic background. We then intercrossed the various single-QTL NILs in all combinations to generate the four double homozygotes for each pair of QTL: aabb, AAbb, aaBB, and AABB. This design allows high replication of each multi-locus genotype and hence reasonable power to detect even moderate epistasis (e.g., Moyle and Nakazato 2009). Applying this methodology, we find that epistasis is more the rule than the exception, although the form of interaction is highly variable among QTL pairs.     

MRAP2 inhibits β-arrestin recruitment to the ghrelin receptor by preventing GHSR1a phosphorylation

The melanocortin receptor accessory protein 2 (MRAP2) is essential for several physiological functions of the ghrelin receptor growth hormone secretagogue receptor 1a (GHSR1a), including increasing appetite and suppressing insulin secretion. In the absence of MRAP2, GHSR1a displays high constitutive activity and a weak G-protein–mediated response to ghrelin and readily recruits β-arrestin. In the presence of MRAP2, however, G-protein–mediated signaling via GHSR1a is strongly dependent on ghrelin stimulation and the recruitment of β-arrestin is significantly diminished. To better understand how MRAP2 modifies GHSR1a signaling, here we investigated the role of several phosphorylation sites within the C-terminal tail and third intracellular loop of GHSR1a, as well as the mechanism behind MRAP2-mediated inhibition of β-arrestin recruitment. We show that Ser²⁵² and Thr²⁶¹ in the third intracellular loop of GHSR1a contribute to β-arrestin recruitment, whereas the C-terminal region is not essential for β-arrestin interaction. Additionally, we found that MRAP2 inhibits GHSR1a phosphorylation by blocking the interaction of GRK2 and PKC with the receptor. Taken together, these data suggest that MRAP2 alters GHSR1a signaling by directly impacting the phosphorylation state of the receptor and that the C-terminal tail of GHSR1a prevents rather than contribute to β-arrestin recruitment.

The “hunger hormone” ghrelin is secreted by X/A cells of the oxyntic mucosa of the stomach in response to a low energetic state, which leads to an increase in appetite (1, 2) and prevents hypoglycemia (3, 4). Ghrelin is the agonist of the growth hormone secretagogue receptor 1a (GHSR1a), a G-protein–coupled receptor (GPCR) expressed in the brain and in multiple peripheral organs including the heart and the endocrine pancreas. Activation of GHSR1a by ghrelin in hypothalamic agouti-related protein (AgRP) neurons potently stimulates feeding (5, 6, 7). In pituitary somatotrophs, GHSR1a stimulation promotes growth hormone release (8, 9, 10). Finally, in cardiomyocytes, ghrelin increases cell survival and contractility (11, 12) while in the endocrine pancreas the hormone inhibits insulin secretion (13, 14).GHSR1a primarily couples to Gα_q/11, thus stimulating the production of intracellular inositol triphosphate (IP) 3. Like other GPCRs, agonist stimulation results in phosphorylation of GHSR1a by kinases, including GPCR kinase 2 (GRK2) and PKC (15), and β-arrestin recruitment. Notably, GHSR1a contains several phosphorylation sites within the C-terminal tail, some of which have been shown to be important for β-arrestin recruitment (16). However, although other putative phosphorylation sites are present in the third intracellular loop (ICL3) of GHSR1a, their role in β-arrestin recruitment has not yet been described.When expressed in heterologous cells, GHSR1a displays a high constitutive activity and a limited ghrelin-stimulated responses (17).Both constitutive- and agonist-stimulated GHSR1a signaling are regulated by the single transmembrane melanocortin receptor accessory protein 2 (MRAP2), which functions to drastically reduce GHSR1a constitutive activity and increase ghrelin-stimulated responses (17). Additionally, MRAP2 significantly inhibits ghrelin-induced β-arrestin recruitment to GHSR1a (17). As such, MRAP2 is essential for several physiological functions of ghrelin including its orexigenic activity (18) and its insulinostatic actions (14). Global or AGRP neuron–targeted deletion of MRAP2 abrogates the effect of ghrelin on food intake (18) and global or pancreatic δ-cell-targeted deletion of MRAP2 prevents ghrelin-mediated inhibition of insulin secretion (14).Although expressed in AGRP neurons and pancreatic δ-cells (thus promoting G-protein coupling and inhibiting β-arrestin-dependent signaling), MRAP2 is not present in every GHSR1a-expressing tissue. Consequently, it is possible that β-arrestin signaling plays an important role in the physiological function of ghrelin in tissues where MRAP2 is absent. Whereas, the inhibition of β-arrestin recruitment to GHSR1a by MRAP2 is well established and the domains of MRAP2 required for this function have been identified (17), the molecular mechanism by which MRAP2 alters GHSR1a signaling is not yet understood. In this study, we investigated the importance of GHSR1a phosphorylation for β-arrestin recruitment and the mechanism involved in MRAP2-mediated inhibition of β-arrestin recruitment.     

Relationships between Growth,Growth Response to Nutrient Supply,and Ion Content Using a Recombinant Inbred Line Population in Arabidopsis

Growth is an integrative trait that responds to environmental factors and is crucial for plant fitness. A major environmental factor influencing plant growth is nutrient supply. In order to explore this relationship further, we quantified growth-related traits, ion content, and other biochemical traits (protein, hexose, and chlorophyll contents) of a recombinant inbred line population of Arabidopsis (Arabidopsis thaliana) grown on different levels of potassium and phosphate. Performing an all subsets multiple regression analyses revealed a link between growth-related traits and mineral nutrient content. Based on our results, up to 85% of growth variation can be explained by variation in ion content, highlighting the importance of ionomics for a broader understanding of plant growth. In addition, quantitative trait loci (QTLs) were detected for growth-related traits, ion content, further biochemical traits, and their responses to reduced supplies of potassium or phosphate. Colocalization of these QTLs is explored, and candidate genes are discussed. A QTL for rosette weight response to reduced potassium supply was identified on the bottom of chromosome 5, and its effects were validated using selected near isogenic lines. These lines retained over 20% more rosette weight in reduced potassium supply, accompanied by an increase in potassium content in their leaves.Plants in natural environments face abiotic constraints limiting growth and ultimately affecting their fitness. In response to such constraints, flowering time (Korves et al., 2007) and seed dormancy (Donohue et al., 2005) as well as vegetative growth (Barto and Cipollini, 2005; Milla et al., 2009) are the main traits controlling fitness (for review, see Alonso-Blanco et al., 2009). These traits are under the control of complex networks integrating genetic (G) and environmental (E) factors as well as their interaction (G × E). Due to the implications for food and renewable energy sources, dissecting the genetic architecture that underlies plant growth is becoming a priority for plant science (Rengel and Damon, 2008; Carroll and Somerville, 2009; Gilbert, 2009).Plant growth is highly dependent on mineral nutrient uptake (Clarkson, 1980; Sinclair, 1992). Minerals can be distinguished into two categories based on the amount required by plants: micronutrients, which are found in relatively small amounts in the plant (such as copper and iron), and macronutrients, which constitute between 1,000 and 15,000 μg g⁻¹ plant dry weight (such as potassium and phosphate; Marschner, 1995, Buchanan et al., 2002). Phosphate is an important structural and signaling molecule with an essential role in photosynthesis, energy conservation, and carbon metabolism. Its deficiency leads to a reduction of growth and an increase of pathogen susceptibility (Marschner, 1995; Williamson et al., 2001; Abel et al., 2002; López-Bucio et al., 2005; Poirier and Bucher, 2008; Vijayraghavan and Soole, 2010). Potassium is not incorporated into any organic substances but acts as the major osmoticum of the cell, controlling cell expansion, plasma membrane potential and transport, pH value, and many other catalytic processes (Maathuis and Sanders, 1996; Armengaud et al., 2004; Christian et al., 2006; Di Cera, 2006). Potassium deficiency leads to reduced plant growth, a loss of turgor, increased susceptibility to cold stress and pathogens, and the development of chlorosis and necrosis (Marschner, 1995; Véry and Sentenac, 2003; Ashley et al., 2006; Amtmann et al., 2008). To cope with changes in nutrient availability, plants have evolved different mechanisms of adaptation, such as changes in ion transporter expression and activity (Ashley et al., 2006; Jung et al., 2009), morphological changes, such as an increase in root growth to explore more soil volume (Marschner, 1995; Shirvani et al., 2001; Jiang et al., 2007; Jordan-Meille and Pellerin, 2008), or acidification of the surrounding soil in order to mobilize more mineral nutrients (for review, see Ryan et al., 2001). Although these adaptations are well known, the mechanisms involved in sensing and signaling low mineral nutrient status are less well understood, despite significant progress in this area being made (Doerner, 2008; Jung et al., 2009; Luan et al., 2009; Wang and Wu, 2010).One approach to identify genes that are involved in plant responses to environmental factors is to perform a quantitative trait locus (QTL) analysis on a mapping population grown in contrasting environments, allowing the identification of QTL-environment (QTL × E) interactions. Some QTLs for growth-related traits in response to environmental changes were cloned already. For example, the differential response of root growth of some Arabidopsis (Arabidopsis thaliana) accessions to phosphate starvation led to the identification of allelic differences responsible for this phenotype (Reymond et al., 2006; Svistoonoff et al., 2007). Other studies have identified QTLs for shoot dry matter under changing nitrogen supply (Rauh et al., 2002; Loudet et al., 2003). In parallel to natural variation for growth, natural variation for ion content has also been reported. In Arabidopsis, considerable variation in the content of mineral nutrients exists both in seeds (Vreugdenhil et al., 2004; Waters and Grusak, 2008) and in leaves (Harada and Leigh, 2006; Rus et al., 2006; Baxter et al., 2008a; Morrissey et al., 2009). Furthermore, changes in mineral nutrient homeostasis have also been reported to be associated with characteristic multivariate changes in the leaf ionome, the mineral nutrient and trace element composition of an organism or an organ (Baxter et al., 2008b). Due to higher throughput and lower costs, such “omics” analyses examining alterations of large numbers of certain molecules at once have recently become available for mapping purposes. Some QTL studies have linked the variations of these omics data to variation of growth or other physiological traits. For instance, Meyer et al. (2007) and Schauer et al. (2008) linked plant growth or morphological traits to a synergistic network of metabolomic compounds in Arabidopsis and tomato (Solanum lycopersicum), respectively. In addition, Sulpice et al. (2009) associated differences in growth with starch content using a set of Arabidopsis accessions. Compiling the importance of ions in the process of cell division (Lai et al., 2007; Sano et al., 2007) or cell expansion (Philippar et al., 1999; Elumalai et al., 2002), ionomics appears to be a major unexplored field for understanding growth.In this study, we focus on variation in plant growth, the root and leaf ionomes, and their response to varying supplies of potassium and phosphate. Studying variations for these traits among recombinant inbred lines (RILs) in Arabidopsis enabled us to detect QTL and QTL × E interactions for all of these traits. To understand the observed variation in plant growth, predictors that explained a high percentage of variation of growth-related traits have been selected especially among the root and leaf ionomes. The colocalization between growth-related trait QTLs and QTLs for their predictors allowed us to point out genetic regions of possible causality. In addition, the effect of a growth-response QTL on reduced potassium supply was validated with selected near isogenic lines (NILs) that maintained a higher rosette weight when grown in reduced potassium supply. This growth advantage went along with significant changes in ion contents that further emphasize the impact of the ionome in plant growth variations.     

In-cell infection: bringing uninvited guests

Cell-in-cell structures resulting from live cell engulfment were identified more than 100 years ago, but their physiological significance has remained largely obscure. Now Ni et al. identify a new role for cell-in-cell structure formation, called “in-cell infection” that spreads Epstein-Barr virus from infected B cells to epithelial cells, an activity that may predispose to cancer.Epstein-Barr virus (EBV) is a common herpesvirus infecting up to 90% or more of the human population that causes mononucleosis, is associated with autoimmune conditions, and predisposes to cancer¹. EBV persists as a latent infection within B cells and predisposes to cancers of B cell origin, including Hodgkin''s and Burkett''s lymphoma, due to expression of latency genes, which leads to B cell transformation. Infected individuals are also predisposed to developing nasopharyngeal and gastric carcinoma, as epithelial cells also harbor latent EBV. However, while the mechanism of EBV entry into B cells is well characterized, how EBV infects epithelium has remained obscure. In a recent paper published in Cell Research, Ni et al.² identify a novel mechanism for EBV infection of epithelial cells, which they term “in-cell infection”, an insidious mode of viral entry that takes advantage of whole cell ingestion.Viral infection is generally mediated by viral envelope glycoproteins that bind to specific receptors on target cells, leading to membrane fusion and viral entry. To infect B cells, the EBV envelope protein gp350 binds to the complement receptor 2 (CR2) on target cells, followed by interaction of gp42 with MHC class II molecules, and virus-to-target cell fusion is mediated by gp42, gH and gL proteins³. Unlike B cells, epithelial cells do not normally express complement receptors or MHC class II molecules, and are generally not infected by purified EBV. Previously described alternative modes of EBV infection of epithelial cells include “cell-to-cell” and “transfer” infection, where B cells have been found to act as carriers to mediate infection through cell adhesion protein-dependent conjugation^3,4,5,6.Ni et al.² now describe a different mode of epithelial cell infection, called “in-cell infection”, that occurs by ingestion of whole EBV-infected B cells, leading to the formation of “cell-in-cell” structures. B cell ingestion in this context resembles “entosis”, a mechanism previously found to mediate cell-in-cell structure formation in epithelial cultures and human tumors⁷. Entosis also promotes the uptake of hematopoietic cells into epithelial cells or cancer cells of various types^8,9. Incredibly, the authors find that entosis-like internalization of latent EBV-infected B cells (Akata) into cultured nasopharyngeal carcinoma cells (CNE-2) leads to the activation of EBV and the transfer of virus to host (CNE-2) cells. Cells infected in this manner express viral gene products, and produce virions upon stimulation that can infect naïve cells of either B cell or epithelial cell origin, indicating potent infection ability and altered tropism of EBV produced by this mechanism.Frequent cell-in-cell structure formation involving EBV-infected B cells is shown by the authors to occur in clinical nasopharyngeal carcinoma samples, suggesting that the in-cell infection mechanism is a likely contributor to viral spread in vivo, and may be linked to carcinoma development. Intriguingly, entosis itself may participate in tumorigenesis by promoting aneuploidy¹⁰, and by supplying cancer cells with nutrients¹¹. As the authors found that EBV infection promoted entosis-like cell uptake, this mode of viral spread could affect tumorigenesis by multiple mechanisms. For in-cell infection, it seems that the nutrients taken in upon the death of internalized cells come mixed with virus that is insidiously transferred to hosts, in a manner perhaps like a Trojan horse enterring with a hidden viral payload (Figure 1).Open in a separate windowFigure 1In-cell infection delivers virus to insusceptible host cells. The B cell infected by EBV resembles a Trojan horse that delivers a hidden viral payload to host epithelial cells.The identification of in-cell infection by Ni et al.² makes a significant contribution to cell-in-cell research by identifying a new pathophysiological role for an entosis-like process. Cell-in-cell structures were first reported over 100 years ago, but the mechanisms that control the formation of such structures and their significance are only now starting to emerge¹². As is often the case with groundbreaking research, the discovery of in-cell infection² raises many new interesting questions. What is the mechanism of B cell internalization into epithelial cells? Entosis is previously described to involve cell adhesion receptors, such as E-cadherin, and Rho-kinase that promotes the actomyosin contraction that drives cell uptake⁷. The molecular mechanism controlling the entry of EBV-infected B cells into epithelial cells will be important to uncover, as other mechanisms in addition to entosis can also mediate the uptake of live cells¹². How is EBV activated by the formation of cell-in-cell structures? How is EBV transferred from internalized cells to hosts? And importantly, can other viruses, such as HIV, spread by in-cell infection? The answers to these questions await further research.     

An emerging consensus on voltage-dependent gating from computational modeling and molecular dynamics simulations

Developing an understanding of the mechanism of voltage-gated ion channels in molecular terms requires knowledge of the structure of the active and resting conformations. Although the active-state conformation is known from x-ray structures, an atomic resolution structure of a voltage-dependent ion channel in the resting state is not currently available. This has motivated various efforts at using computational modeling methods and molecular dynamics (MD) simulations to provide the missing information. A comparison of recent computational results reveals an emerging consensus on voltage-dependent gating from computational modeling and MD simulations. This progress is highlighted in the broad context of preexisting work about voltage-gated channels.Voltage-gated K⁺ (K_V) channels and prokaryotic voltage-gated Na⁺ (Na_V) channels are formed by four subunits surrounding a central aqueous pore that allows ion permeation. Each subunit consists of six transmembrane α-helical segments called S1 to S6; the first four of these, S1–S4, constitute the voltage-sensor domain (VSD), whereas the S5–S6 segments assemble to form an ion-selective pore domain (see Fig. 1). The VSDs respond to changes in the potential difference across the cell membrane. When the membrane is depolarized, the VSD in each subunit undergoes a conformational transition from a resting to an activated state, and this information is communicated to the ion-conducting pore to promote its opening (Bezanilla et al., 1994; Zagotta et al., 1994). The activation of the VSD and opening of the pore are associated with the transfer of an electric charge ΔQ across the membrane, called the “gating charge” (Sigworth, 1994). Opening of the voltage-gated K⁺ channel Shaker corresponds to the outward translocation of a large positive charge on the order of 12–14 elementary charges (Schoppa et al., 1992). Four highly conserved arginines along S4 (R1, R2, R3, and R4) underlie the dominant contributions to the total gating charge of Shaker and appear to be mainly responsible for the coupling to the membrane voltage (Papazian et al., 1991; Aggarwal and MacKinnon, 1996; Seoh et al., 1996). The overall structure of eukaryotic voltage-gated Na⁺ channels, which are composed of four analogous subunits covalently linked in a single polypeptide, appears to be similar (Catterall, 2012).Open in a separate windowFigure 1.Overall view of the voltage-activated Kv1.2 K⁺ channel (Protein Data Bank accession no. 3LUT). (A) Two of the four subunits of the channel are displayed from a side view. The VSD comprises the transmembrane segments S1–S4, and the pore domain comprises the transmembrane segments S5–S6. (B) The tetramer is displayed from the extracellular side (each subunit is a different color). The two views are related by a 90° rotation.The nature of the conformational change within the VSD, and how it is communicated to the pore domain, is the key question that must be answered to explain voltage-dependent gating. Ultimately, we need to know the 3-D structure of the multiple resting and activated states of the VSDs and their relationship to the closed and open conformations of the pore at atomic resolution to understand the voltage-dependent gating mechanism in molecular terms. However, although x-ray crystallographic structures of the Kv1.2 channel, Kv1.2/Kv2.1 chimera, and bacterial Na_VAb channels have provided information on the conformation of the active state (Long et al., 2005, 2007; Payandeh et al., 2011), no atomic resolution structure of a K_V or Na_V channel in the resting state is currently available. This has motivated the use of computations to provide the missing information about channel gating (Yarov-Yarovoy et al., 2006, 2012; Pathak et al., 2007; Bjelkmar et al., 2009; Delemotte et al., 2010, 2011; Khalili-Araghi et al., 2010, 2012; Schwaiger et al., 2011; Vargas et al., 2011; Jensen et al., 2012). These computational studies have relied on different approaches, including Rosetta modeling, a protein-folding method using knowledge-based potentials, and molecular dynamics (MD) simulations, consisting of propagating Newton’s classical equation of motion as a function of time using an all-atom force field. Remarkably, despite the considerable variations in computational methodologies and in template x-ray structures used, a highly consistent picture is emerging from these studies. Here we briefly review the most recent results in the broad context of preexisting work about voltage-gated channels.

Computational models of the resting state

Early models of the resting-state conformation of the VSD were obtained using the Rosetta method (Yarov-Yarovoy et al., 2006; Pathak et al., 2007); these initial models were subsequently refined with all-atom MD simulations (Khalili-Araghi et al., 2010) and with high-resolution Rosetta algorithms (Yarov-Yarovoy et al., 2012). Independent studies obtained very similar conformations using a combination of experimentally derived constraints based on engineered cross-links and metal bridges during MD simulations (Delemotte et al., 2010, 2011; Henrion et al., 2012). Moreover, these earliest models (Yarov-Yarovoy et al., 2006; Pathak et al., 2007) predicted pairs of neighboring residues before they were identified experimentally (Campos et al., 2007). Subsequent refinement of structural models of the resting state made it possible to demonstrate the existence of a consensus 3-D conformation of the VSD that satisfied a wide range of experimental data (Vargas et al., 2011). Rosetta models for the bacterial sodium channel NaChBac are very similar to those of K_V channels and have been extensively tested by disulfide cross-linking studies (DeCaen et al., 2008, 2009, 2011; Yarov-Yarovoy et al., 2012).In practice, the atomic models of the resting state have either been refined by imposing inter-residue distances that are consistent with the experimentally derived constraints (Yarov-Yarovoy et al., 2006, 2012; Delemotte et al., 2010, 2011), or by explicitly modeling the side chains involved in the various cross-links or metal bridges themselves (Vargas et al., 2011; Henrion et al., 2012). Although these various models display high similarity, they all relied to varying degrees on computational “shortcuts” to obtain meaningful results about the VSD conformations within a reasonable computational time. Ultimately, the dream would be to “visualize,” atom-by-atom, how the channel moves as a function of time in response to a realistic membrane potential.This has now become possible in part by relying on the virtual reality provided by computer simulations. Computer trajectories “simulating” the effect of membrane hyperpolarization on a voltage-gated ion channel were generated by several research groups, with the goal of triggering deactivation to directly observe the conformational response and reorganization of the VSD (Treptow et al., 2004, 2009; Nishizawa and Nishizawa, 2008, 2009; Bjelkmar et al., 2009; Denning et al., 2009; Delemotte et al., 2011; Freites et al., 2012). More recently, Jensen et al. (2012) used long (hundreds of microseconds) MD simulations to visualize a complete spontaneous conformational transition of a voltage-gated K⁺ channel upon changes of the membrane potential. As in the previous simulations, large negative hyperpolarizing membrane potentials were applied (−750 mV) to shorten the time for the voltage-dependent transition toward the resting state within accessible computing time (although some simulations were also generated at −375 mV). The long MD simulations performed by Jensen et al. (2012) led to a computationally derived model of the resting-state conformation of the channel very similar to those deduced previously with different methods (Yarov-Yarovoy et al., 2006, 2012; Pathak et al., 2007; Bjelkmar et al., 2009; Delemotte et al., 2010, 2011; Khalili-Araghi et al., 2010; Schwaiger et al., 2011; Henrion et al., 2012) (Fig. 2). The results confirm and substantially strengthen the consensus from previous computational studies.Open in a separate windowFigure 2.Main elements of secondary structure of the VSD in the active and resting state. (A) The VSD in the active conformation (taken from the x-ray structure; PDB accession no. 3LUT). (B) A superposition of different models of the resting-state configuration of the VSD obtained by independent research teams (Delemotte et al., 2010; Schwaiger et al., 2011; Vargas et al., 2011; Jensen et al., 2012) using different constraints and methodologies. The four helices, S1 (gray), S2 (yellow), S3 (red), and S4 (blue), are displayed. The spheres correspond to the Cα atoms of E1 (Glu226) and E2 (Glu236) along S2, and R1 (Arg294) along S4.By several quantitative measures, all the mentioned atomic models of the resting state display a high degree of similarity, indicating that a consensus on the structure of the resting state and the mechanism of VSD function has emerged from the independent computational studies. The backbone Cα carbons of all the models lie within 3–4 Å root mean square deviation (RMSD) of one another (Table S1). This is comparable to the RMSD among the four VSDs of tetrameric structures obtained in various models. The RMSD among the four VSDs of the tetrameric structure of Delemotte et al. (2010, 2011) vary between 2.7 and 3.9 Å, and the RMSD between the four VSDs of the tetrameric structure of Jensen et al. (2012) vary between 1.3 and 3.4 Å. For comparison, the VSDs of the x-ray structures of the Kv1.2 and Kv1.2/Kv2.1 chimera display RMSDs of 1.5 Å for the same region. The largest deviations of any of the resting-state models from a hypothetical average configuration are <3 Å; the two main outliers are one subunit from Delemotte et al. (2010, 2011) and one subunit from Jensen et al. (2012) (Fig. S1).In all of the resting-state models, the S4 helix is predominantly rotated and translated inward along its main axis relative to the x-ray structures of the activated conformation, whereas the S1 and S2 helices retain their configuration. Averaging over all models, the absolute vertical displacement of S4 at the level of the Cα of the R1 position is ∼10 Å, with a spread of 3–4 Å (Table S1). There is some uncertainty in estimating the vertical translation of S4 because of the large structural fluctuations exhibited by the flexible VSDs. For example, the vertical position of R1 in the two x-ray structures of the active state of the K_V1.2 VSD differs by 2.6 Å (Long et al., 2005, 2007), and the net vertical displacement in the VSDs of the four subunits from Jensen et al. (2012) is 13.2, 16.2, 14.9, and 12.0 Å. All the resting-state models place the Cα of the R1 between the two acidic side chains E1 and E2 along the S2 helix, and R1 is also located above the highly conserved Phe located in the middle of S2. Most importantly, all of the resting-state models show the positive gating charge residues along S4 in position to either form salt bridges with acidic residues in the S1–S3 helices, or interact with the aqueous regions or the polar head groups. Similar conclusions have been obtained from the experiments on Na_V channels that combined structural modeling with disulfide cross-linking experiments (DeCaen et al., 2008, 2009, 2011; Yarov-Yarovoy et al., 2012).The term “resting state” in the above discussion is used to loosely describe the conformation in which S4 inhabits its most inward position. Upon careful consideration, however, “the” resting state is probably an oversimplified concept because it is likely that S4 does not withdraw to the same position in all channels as result of sequence variations, regardless of the applied membrane potential. Furthermore, several of these resting states of the VSD may favor the nonconducting closed state of the pore. Consistent with this idea, hyperpolarization of the membrane potential slows the activation of voltage-gated channels, a behavior known as the “Cole–Moore effect” (Cole and Moore, 1960), because the VSDs must move through more resting states before activation when they start from a more negative membrane potential. To highlight the structural differences among the various proposed resting-state models shown in Fig. 2 (right), it is useful to realign the different models with respect to the S1–S3 helices, which are the most stable structural elements. The set of models, displayed in Fig. S2, places the S4 helix at various depths relative to S1–S3, with a slight spread in the tilt angle of its main axis. The initial resting-state model of K_V1.2 (Yarov-Yarovoy et al., 2006) captured a resting state in which the S4 segment is not drawn as far inward as the models of Fig. 2 and which, therefore, likely represents an initial step toward activation. Models of a full range of resting states of the VSD extending to quite negative voltages have been developed for both K_V and Na_V channels (DeCaen et al., 2009, 2011; Delemotte et al., 2012; Henrion et al., 2012; Yarov-Yarovoy et al., 2012).The existence of multiple resting states is supported by a wide range of experiments, including analysis of multiple engineered metal bridges tracking successive states of the VSD (Henrion et al., 2012), noncovalent interactions between R2 and a tryptophan residue inserted in S1 (Lacroix et al., 2012), and disulfide cross-linking results of S4 gating charges with ion pair partners in S2 and S3 (DeCaen et al., 2009, 2011; Yarov-Yarovoy et al., 2012). It is expected that the multiple resting states of the VSDs of K_V and Na_V channels all stabilize the pore in its nonconducting closed conformation over a specific range of negative membrane potentials. Interestingly, a series of such resting states is observed in the long MD simulations of Jensen et al. (2012) as pauses in the inward movement of the VSD after the pore has closed. Thus, the conformations obtained from extremely long MD simulations are consistent with the resting state predicted by several independent studies using various computational approaches including knowledge-based structure prediction algorithms (Yarov-Yarovoy et al., 2006, 2012; Pathak et al., 2007) and MD simulations with experimental constraints (Delemotte et al., 2010; Vargas et al., 2011; Henrion et al., 2012). The final resting state from long MD simulations of Jensen et al. (2012) appears closest to the state reported in Delemotte et al. (2012), with the most inward position of S4. It is very satisfying that the conformational states are visited as the result of spontaneous transitions during long unbiased MD trajectories of the protein submitted to a negative membrane potential. The ability to simulate the spontaneous conformational transitions strengthens confidence in our current understanding of the physical forces and molecular interactions governing the voltage-gating process at the atomic level.

Mechanism of voltage-dependent activation

A consistent mechanistic perspective of voltage gating has emerged from these computational studies. The scenario that most accurately conveys the conformational change occurring within the VSD during activation as observed in these computations is the classical helical screw–sliding helix mechanism in which the S4 segment retains its helical conformation as it moves principally along its long axis (Catterall, 1986a,b; Guy and Seetharamulu, 1986). The gating charges are not directly exposed to the lipid hydrocarbon, and the S3–S4 helix-turn-helix does not move as a highly concerted structural motif across the membrane during voltage gating as proposed in the more recent paddle model (Jiang et al., 2003). Rather, sequential formation of salt bridges involving the gating residues plays an important role as proposed by Clay Armstrong (1981). Lastly, the concept of the “focused electric field” (Islas and Sigworth, 2001; Asamoah et al., 2003; Starace and Bezanilla, 2004), in which the spatial variation in the transmembrane potential affecting the gating charges of the VSD is concentrated over a narrow region that is considerably thinner than the full bilayer membrane, has been clarified by explicit calculations of the gating charge contributions based on all-atom MD simulations following two different approaches (Khalili-Araghi et al., 2010; Delemotte et al., 2011) and further supported by structural modeling studies (Yarov-Yarovoy et al., 2012). The gating charge calculations of Jensen et al. (2012) following the methodology of Khalili-Araghi et al. (2010) provided an additional confirmation of the concept of a focused electric field. Nonetheless, the S4 segment moves outward through the focused field, and the mechanism of voltage gating is not primarily a rearrangement in the transmembrane field as proposed in the transporter model (Chanda et al., 2005).The MD simulations of Jensen et al. (2012) showed that the ion-conducting pore closes before any of the four VSDs have undergone a transition to the most stable resting-state conformation. This sequence of events is consistent with kinetic models with discrete states developed long ago to describe voltage gating in the Shaker K⁺ channel (Bezanilla et al., 1994; Zagotta et al., 1994; Schoppa and Sigworth, 1998). According to these kinetic models, the first step involved in closing an activated channel is the closing of the pore domain, followed by the independent transitions of the four VSDs toward the resting state. Recent x-ray structures of bacterial Na_V channels provide examples of this intermediate state, showing a closed pore domain associated with VSDs in their activated conformation (Payandeh et al., 2011, 2012; Zhang et al., 2012).In summary, the major advances are that the resting-state conformation of the VSD reached by the long MD simulations is consistent with the results of numerous previous studies using different computational methods (Yarov-Yarovoy et al., 2006, 2012; Pathak et al., 2007; Bjelkmar et al., 2009; Delemotte et al., 2010, 2011; Khalili-Araghi et al., 2010; Schwaiger et al., 2011; Henrion et al., 2012), and that the sequence of events seen in the long simulations appears to be in qualitative accord with classical kinetic models of the voltage-gating process (Bezanilla et al., 1994; Zagotta et al., 1994; Schoppa and Sigworth, 1998). However, there is no experimental data at the large negative voltages used in the long simulations, and one must be cautious in trying to extrapolate the experimental time constants determined around −100 mV for ionic currents (Rodríguez and Bezanilla, 1996) and gating currents (Rodríguez et al., 1998).

Novel mechanistic hypotheses from MD simulations

Some novel ideas about the mechanism of voltage-dependent gating are suggested by the computational studies but do not yet have direct structural or experimental support. For example, sections of the S4 segment are observed in 3₁₀ helical conformation in x-ray crystal structures (Long et al., 2007; Clayton et al., 2008; Vieira-Pires and Morais-Cabral, 2010; Payandeh et al., 2011, 2012; Zhang et al., 2012). An intriguing suggestion from several simulation studies is the concept of a sequential dynamical transition to a 3₁₀ helical conformation for all or part of the S4 segment as it moves through the most hydrophobic region of the VSD (DeCaen et al., 2009, 2011; Khalili-Araghi et al., 2010; Schwaiger et al., 2011; Yarov-Yarovoy et al., 2012). Although the presence of some amount of 3₁₀ helical conformation is supported by available data (Villalba-Galea et al., 2008), the concept of a dynamic 3₁₀ transition of S4 during the voltage-gating process will require further experimental validation.Another suggestion from the long MD simulations is that the closure of the pore domain is driven by a rapid de-wetting transition taking place in the intracellular vestibule. A similar de-wetting process was previously found in voltage-driven simulations of an isolated pore domain, in the absence of the VSDs (Jensen et al., 2010). The de-wetting process results in a closed pore with a nearly dry central cavity. However, evidence that ions may be captured in the cavity of a closing Shaker K⁺ channel argues against a complete de-wetting (Baukrowitz and Yellen, 1996a,b; Ray and Deutsch, 2006). We note here also that the structures of the preopen and inactivated states of prokaryotic Na_V channels have closed, water-filled pores (Payandeh et al., 2011, 2012; Zhang et al., 2012). More experimental work is required to determine whether de-wetting drives pore closure or arises only at very negative membrane potentials.Lastly, the relationship of the voltage-gating transition displayed by the long MD simulations to the three major conformations of a VSD (resting, activated, and relaxed) observed in most of the S4-based VSDs (Villalba-Galea et al., 2008; Lacroix et al., 2011) remains uncertain. Moreover, the transitions observed in long MD simulations do not appear to reproduce all of the early components of the gating current (time constant of ∼10 µs) observed in Shaker K⁺ channels (Sigg et al., 2003). New experiments in which long MD simulations and gating current measurements are made in parallel on the same channel will give more insight into these issues.

Confidence in the computational results

Several models derived from a combination of experimental data and computations, produced from different approaches, have converged to yield a low resolution picture of the resting-state conformation, defined within ∼3–4 Å RMSD (Fig. 2, right). Upon a closer look, fine differences can be noted among the various models (Fig. S2), which points to the concept of multiple resting states in which the segment S4 is drawn to different depths toward the intracellular side. The broad agreement among the various computational methods, most likely, is not fortuitous, and the picture emerging represents a genuine advance in understanding voltage-gated channels. The implication, if the computational results are to be trusted, is that many of the apparently conflicting measures about voltage gating can be resolved. Nevertheless, some might argue that the controversy about the resting-state conformation of the VSD will remain until an experimental x-ray structure becomes available. In this context, it is important to note that all structures are models, even x-ray crystal structures. However, structural models from x-ray crystallography rely on a huge amount of experimental data and are derived from rigorously established procedures that have been extensively tested and cross-validated. It is expected that experimental structure determination will increasingly rely on sophisticated computational modeling to complement low resolution data (Chen et al., 2007; Trabuco et al., 2008; Brunger et al., 2012). In the early 1980s, NMR structures were considered tentative models until it was demonstrated that the results were consistent with x-ray crystallography (Billeter et al., 1989). What ultimately matters is the quantifiable level of confidence that can be attributed to a structural model. As the methodologies become more and more reliable and consistent, computational modeling will play an increasingly important role in structural biology (DiMaio et al., 2011; Lange et al., 2012). The present situation, in which the proposed models of the VSD (Fig. 2) are supported by such a wide range of computational approaches applied by different investigators, is unprecedented to our knowledge. For this reason, our view is that one may be (cautiously) optimistic that the resting-state structures of Fig. 2 and the related structures in other computational papers cited here are close (within ∼3–4 Å RMSD) to reality. Notably, this accuracy can be predicted from the distribution of models themselves (even the early ones).At this point it is prudent to sound a note of caution. The progress documented in this Viewpoint on understanding structural aspects of a membrane-bound channel protein has been made possible by using novel computational methodologies and an empirical potential energy (force field) that subsumes polarization and nuclear quantum effects in an average fashion. Although MD simulation articles often imply that every in silico detail from the trajectory is real, it is useful to remind ourselves that this is not necessarily the case. Experience indicates that current approximations are more successful in predicting conformational states than transition rates. The reason is that the overall topology of the potential energy surface, with its wells and barriers, is more or less correct, even though the relative depths of the wells and heights of the barriers may be imperfect. As a consequence, the order in which the chain of events takes place in a simulation during a complex conformational transition may not reflect reality, as some parts may undergo transitions that are too slow, whereas other parts undergo transitions that are too fast. Therefore, although a strong consensus is emerging on the nature of the conformation of the resting state, the dynamic properties of the gating process require more scrutiny. This will be challenging notwithstanding the massively increased length of recent MD trajectories.

Conclusion

A clear consensus on the mechanism of voltage-dependent gating is emerging from various studies based on a wide range of computational and experimental methods. This consensus, which is to be celebrated, highlights the increasingly important role of computational modeling in linking molecular structure to biological function by supplementing missing information. It is important, however, to remain prudent in assessing the significance of details and features of the computationally derived models that have not yet been experimentally validated. Even if the resting-state conformation of the VSD reached by the simulations is correct, and the sequence of events is in accord with classical kinetic models of the voltage-gating process, it is possible that the rates of the individual processes is differentially affected by insufficient sampling, force field inaccuracies, and the large membrane potential typically applied so far. Nevertheless, it is encouraging to note that, despite their inherently approximate nature, current computational models can provide meaningful answers to important questions about complex biomolecular systems. Further studies using computational methods in concert with structure–function experiments seem likely to soon reveal the missing details of VSD function.

Online supplemental material

Table S1 provides all the data about minimum global RMSD of the VSDs using best pairwise alignments for all models. Fig. S1 shows the deviation of backbone atom of each model relative to the average. Fig. S2 shows a superposition of all the VSD models in the resting-state configuration aligned with respect to S1–S3 helices. Fig. S3 presents a quantitative structural comparison of all the VSD models with the crystal structures. Video 1 includes an animation showing all the available VSD models rotating in superposition. The online supplemental material is available at http://www.jgp.org/cgi/content/full/jgp.201210873/DC1.