observation9, especially of systems where the companion is somewhat less massive than the primary disk, and is in or near a gravita-tionally bound orbit. Recent extensions have taken into account observations of several extreme processes, among them high-velocity collisions between galaxies that fall together from distant points, but that are both bound to a greater structure such as a galaxy group or cluster. The interacting system NGC 5291 (ref. 10) is a prominent example of the hyper-extended rings that can result (Fig. 2).In large rings, the gas density is likely to be low, so the star formation rate will also be low. Mapelli and colleagues’ central idea is that such a weak ring will generally be so dim as to be barely visible, and that the huge disk that it leaves behind it as it propagates outwards will look remarkably like a GLSB. Moreover, in the time (a billion years or so) that it takes the material to travel out to such large dis-tances, the high-speed companion will in many cases have moved off the immediate scene. It is therefore hardly surprising if we see no obvious evidence of the past fracas. The authors4 use extensive data on four GLSBs to select, from a grid of computer simulations with a range of initial collision parameters, those models, and the times during their evolution, that are most like the observed systems. In all four cases, a model is found in which the surface-brightness profiles, optical colours, and the gas distributions and kin-ematics of the simulated galaxy all agree well with observations. In three out of four cases, possible companions are also in view.The obvious consequence of the hypoth-esis is that GLSBs are disturbed bodies, and not the stable, quiescent galaxy disks that Figure 2 | Collision remnant. A multi-wavelength image of the hyper-extended ring galaxy NGC 5291. they had been assumed to be. This means that hier archical galaxy-formation theory within a CDM model is not required to yield GLSBs as final products: the challenge they represent to the theory vanishes. In fact, the best-fit models of Mapelli et al. all include ‘haloes’ of CDM surrounding the colliding galaxies. Sound as it might like a fractured fairy tale or a Wagner opera gone wrong, the evidence seems to suggest that cosmic gentle giants spring from bejewelled rings. But mysteries remain, such as how smaller, gas-rich low-surface-brightness galaxies form, and how they persist so quietly until they are lit up by a collision. With rapid advances in our under-standing of galaxy formation, the prospects are bright for a speedy resolution. ■Curtis Struck is in the Department of Physics and Astronomy, Iowa State University, Ames, Iowa 50011, USA. e-mail: [email protected]. Pickering, T. E., Impey, C. D., van Gorkom, J. H. & Bothun, G. D. Astron. J. 114, 1858–1882 (1997). 2. Impey, C. & Bothun, G. Astrophys. J. 341, 89–104 (1989).3. Edmunds, M. G. Nature 341, 105–106 (1989).4. Mapelli, M. et al. Mon. Not. R. Astron. Soc. (in the press); preprint at www.arxiv.org/abs/0710.5354 (2007).5. Springel, V., Frenk, C. S. & White, S. D. M. Nature 440, 1137–1144 (2006).6. Springel, V. et al. Nature 435, 629–636 (2005).7. Governato, F. et al. Mon. Not. R. Astron. Soc. 374, 1479–1494 (2007).8. Lynds, R. & Toomre, A. Astrophys. J. 209, 382–388 (1976).9. Appleton, P. N. & Struck-Marcell, C. Fund. Cosmic Phys. 16, 111–220 (1996).10. Bournaud, F. et al. Science 316, 1166–1169 (2007).STRUCTURAL BIOLOGYIon pumps made crystal clearDavid C. GadsbyThe function of every cell in our bodies depends on the work of proteins known as ion pumps. Several new crystal structures cast fresh light on how three different pumps deal with their distinct cargoes of ions.Ion pumps toil tirelessly in cells throughout all kingdoms of life, transporting ions across membranes. To investigate the workings of these microscopic machines, X-ray crystal structures of a calcium ion pump known as SERCA have been determined1–5. But although those structures depict SERCA in several con-formations, none of them caught the pump in the act of releasing its cargo of ions. More-over, nagging questions remained about how much SERCA might differ from other, geneti-cally related ion pumps — such as those that transport ions of different sizes and charges from calcium, or that require additional pro-tein subunits. In this issue, three papers6–8 from the same group go a long way towards addressing those concerns by describing the first atomic structures of a SERCA pump with its ion pathway open6 and of two related pro-teins — a sodium–potassium pump7 and a proton pump8.The three pumps described in these papers belong to a family known as phosphory-lated-type (P-type) pumps, named after the phosphate group whose addition and removal controls their activity. P-type pumps inhabit all our cells and are essential for life. Without sodium–potassium pumps, many vital func-tions would fail. For example, there would be no electrical signals in our brains or hearts, or in any nerves or muscles; and without SERCA pumps, there would be no muscle contraction. Not surprisingly, P-type pumps are hot tar-gets for therapeutics — for example, digoxin (a treatment for heart problems) targets sodium–potassium pumps, and the latest ant acids act on the proton–potassium pumps in our stom-achs. So the stakes are high — determining the structure and mechanism of each P-type pump is crucial for further drug discovery.P-type pumps reside either in the surface membranes of cells or in the membranes of intracellular organelles such as the endoplasmic or sarcoplasmic reticulum. In all cases, one end of the pump opens to the cytoplasm and the other end opens either to the outside of the cell or to the interior (lumen) of the organelle. The pumps adopt two main conformations9, known as E1 and E2 (Fig. 1, overleaf). The ion-binding sites are found deep inside the region of the pump that crosses the membrane; in E1, these sites are accessible to ions in the cytoplasm. Ion binding promotes the phosphorylation of the pump, in which a phosphate group is added to a single amino-acid residue. The source of the phosphate is an ATP molecule; a side product (ADP) is formed that briefly remains associated with the pump. In the resulting E1P state, the bound ions are occluded — they are inaccessible from either side of the membrane. The pump then releases the ADP and relaxes to the E2P conformation, whereupon a