I was complaining to a colleague, an organic chemist, about the enormous task of keeping up with the current literature on the process of receptor-mediated endocytosis (Box 1). "You cell biologists have it easy," he replied. "All you read is the literature of the last five years; synthetic chemists read the literature of the past 150 years. If someone discovered a synthetic reaction in the nineteenth century, it is still as true today as it was then." I had to agree with my colleague that cell biologists do tend to read, and certainly to cite, only the most recent literature. But is cell biology really different from chemistry? Is something that was discovered in 1970 less true today than it was then?
Cell biology is the science of studying how cells work, usually within the context of the intact cell. Every advance in technology that allows the more precise measurement of a cellular process in time and/or space has been followed by a burst of experimental activity and a new layer of literature. However, often the results of this activity, as spectacular as they seem, are more confirmatory than novel. The reason for this is simple. A well-trained human imagination can sometimes achieve remarkable insight that far outstrips the meagre evidence on which it is based. Later work, with better technology, tests these early insights, substantiates them and adds detail. This has certainly been the case for the study of clathrin-mediated endocytosis. Endocytosis is a highly dynamic process that can now be studied in real time in live cells using fluorescent proteins. However, much of our fundamental understanding of this process came from astute interpretations of morphology that were made 20–40 years ago and current experiments employing light microscopy are invariably interpreted in the context of what was learned from electron microscopy. The brief article that follows describes particularly insightful examples of the earlier work that provide the basis for our current view of the mechanisms for internalizing cell-surface receptors through clathrin-coated pits and recycling them to the plasma membrane.
Interpreting morphology
Coated pits. In 1964, Thomas Roth and Keith Porter described the basic aspects of clathrin-coated-pit formation, relying on the analysis of electron micrographs1. However, the concept of endocytosis predates this study by more than 70 years. At the end of the nineteenth century, Metchnikoff observed that cells ingesting litmus particles turned them from blue to red, indicating that cells internalize particles into an acidic compartment2. By 1931, using cinematographic microscopy, W. H. Lewis photographed macrophages and other cells taking up extracellular fluid into small vesicles, a process he called pinocytosis2. Even with the limited technology that was available to him, Lewis estimated that a macrophage internalized as much as a third of its volume each hour. The concepts of the highly active uptake of extracellular material by cells into acidic compartments and a requirement for membrane recycling were therefore well appreciated3 when Roth and Porter began their studies.
Electron microscopy (EM) poses a challenge for the investigation of dynamic processes such as endocytosis, because it requires samples to be fixed and therefore frozen in time. To surmount this problem, Roth and Porter chose a system in which they could control the timing of cellular events – the deposition of yolk into oocytes that begins when a mosquito has a blood meal. They carried out an anatomical pulse-chase experiment in which mosquitoes were sacrificed at various times after feeding and oocytes were examined by EM. Their beautifully detailed electron micrographs showed that, as the time after feeding increased, small "bristle-coated pits" accumulated at the plasma membrane of the oocyte. The pits contained electron-dense material that Roth and Porter concluded was the yolk proteins, and on the opposite (cytoplasmic) side of the membrane, they were lined by a "bristle" coat.
From these data, Roth and Porter deduced the order of events for the formation and pinching off of a coated vesicle (Fig. 1). They suggested that the coat was responsible for the apparent concentration of yolk cargo in the pits and speculated that the coat might function to curve the membrane, although they acknowledged that, as uncoated endocytic pits had been reported3, 4, the coat might not be required for this function. Today, these conclusions remain the key concepts for clathrin-coat function, and this landmark paper initiated an era of intense examination of the morphology of endocytosis, which resulted in several excellent studies that observed coated pits functioning in the uptake of different types of membrane in many cell types5, 6, 7, 8.
Figure 1 | The sequence of events in the endocytosis of yolk proteins by coated pits of mosquito oocytes.
This schematic drawing depicts the sequence of events in the endocytosis of yolk proteins by coated pits of mosquito oocytes as presented in Roth and Porter's original paper1. The original legend reads in part, "At (1) is shown the first stage of invagination into the oocyte of the protein-coated plasma membrane from the intercellular space. The fully developed pit (2), by pinching off, forms the coated vesicle (3). These vesicles lose their bristles to form dense spheres of similar size (4), which then fuse with other dense spheres (5). Often a flattened empty sac is attached to the droplet (7). This sac may be the membrane remnant of a vesicle or perhaps some element of the Golgi complex that has recently fused with the droplet. The larger droplets (6) coalesce to form the large crystalline proteid yolk bodies (8) of the oocyte." ER, endoplasmic reticulum; L, lysosome. Figure reproduced with permission from Ref. 1 © (1964) the Rockefeller University Press.
In the 1970s, the investigation of nutrient and hormone binding to cells led to the conclusion that cells have specific receptors on their surfaces for the uptake of extracellular molecules. In a remarkably prescient review article that was published in Nature in 1979, Goldstein, Anderson and Brown interpreted this binding data in light of their own studies of low-density lipoprotein (LDL)-receptor endocytosis9, and outlined the basic tenets that constitute the modern view of receptor-mediated endocytosis10. Using light microscopy and EM, they had observed that when LDL particles bound to cells they clustered over coated pits and were internalized by them11. Biochemical experiments that measured LDL uptake had indicated that the receptors must recycle after internalization and be used repeatedly. Most importantly, Goldstein et al. had identified a patient suffering from familial hypercholesterolaemia, whose fibroblasts bound LDL but were incapable of delivering it to coated pits. Therefore, several years before the LDL receptor was partially purified and used to generate antibodies, these researchers predicted that cell-surface receptors for LDL and other ligands would be transmembrane proteins with a cytoplasmic binding site for an element of the endocytic coat (Fig. 2). In addition, they proposed that the receptors would be recognized and concentrated through crosslinking by cytoplasmic coat proteins. This concept of the clathrin-coated pits as a common entry point for cell-surface receptors was firmly established by work showing that the receptors for epidermal growth factor (EGF), insulin and 
2-macroglobulin clustered together on the plasma membrane and were internalized together12. Quantitative electron microscopy also showed that receptors for LDL and EGF shared the same coated pits13. In the decades that followed, more advanced techniques were applied to the study of clathrin-mediated endocytosis and a great deal was learned (TIMELINE). However, our basic understanding of this process was established through the interpretation of simple observations that were made before molecular tools were available.
Figure 2 | An early proposal for a mechanism of receptor-mediated endocytosis.
This simple line drawing is taken from a 1979 review article by Goldstein, Anderson and Brown, who proposed, on the basis of little physical evidence, that endocytic receptors are transmembrane proteins that have cytosolic signals for clathrin coats and that these receptors recycle. LDL, low-density lipoprotein. Figure reproduced with permission from Ref. 10 © (1979) Macmillan Magazines Ltd.
Sorting endosomes. Roth and Porter admitted that the fate of coated vesicles after they had pinched off from the membrane was not clear. They saw small vesicles containing yolk and some larger ones, and they suggested that the small ones fused together to form larger vesicles. However, the complexity of the internal membranes prevented further conclusions. Early progress in understanding the sequence of events after internalization was provided by morphological studies of the endocytosis of viruses through coated pits (Ref. 14 and references therein). Viruses that were bound to cells at low temperatures, which prevented internalization, could be located inside the cells after brief intervals of warming. Under these conditions, single virions were first seen in coated vesicles, and minutes later groups of virions were observed in larger uncoated vesicles, the early endosomes, and later still in multivesicular endosomes and lysosomes. These studies revealed the fate of internalized ligands, but did not report on the fate of the ligand receptors.
Although Goldstein and colleagues knew that the LDL receptor must recycle to the plasma membrane after separating from its LDL ligand, they did not speculate on the pathway that the receptors follow or on how the receptors could be separated from LDL. An important conceptual advance was made by Geuze and colleagues15. They investigated the pathway that is followed by the asialoglycoprotein receptor and its ligand after clathrin-mediated internalization by labelling them on frozen thin sections using antibodies and colloidal gold. Although others had observed that endocytosed tracers entered a linked network of tubules and vesicles16, 17, the labelling technique used by Geuze and colleagues allowed them to observe that the receptors were concentrated in tubules leading from central vacuoles that were enriched in ligand. Geuze et al. called this compartment CURL (compartment of uncoupling ligand from receptor; Fig. 3), but it is now known as the sorting endosome. Interestingly, Roth and Porter described something that looked like CURL in their 1964 paper (Fig. 1).
Figure 3 | An early proposal for a sorting endosome.
This line drawing, which is taken from a 1983 article by Geuze et al., shows the morphology of sorting endosomes (labelled CURL (compartment of uncoupling ligand from receptor)), postulates that tubules form on the recycling pathway, and shows the progression from sorting endosomes to multivesicular bodies to lysosomes. CV, coated vesicle; G, Golgi; L, lysosome; MVB, multivesicular body; PM, plasma membrane. Figure reproduced with permission from Ref. 15 © (1983) Elsevier Science.
The geometry of the sorting endosome membrane system was indicative of a mechanism for partitioning ligands into a pathway that leads to lysosomes, and membrane-bound receptors into one leading back to the cell surface, simply on the basis of differences in the ratio of membrane surface to internal volume in the tubules and the central vacuole18. Ligands released from receptors would diffuse into the volume of the central vacuole and the receptors would partition with most of the membrane into the tubules. Experimental evidence supporting this idea was provided by Maxfield and colleagues19, 20. In experiments that were the precursors of current work with intrinsically fluorescent proteins, they used separate fluorescent tracers for ligands and receptors, and were able to study the kinetics of ligand and receptor sorting quantitatively in individual endosomes in live cells. They concluded that endosomal sorting was likely to be a process in which the sorting endosome fused repeatedly, but for a limited time, with newly formed endocytic vesicles. The receptors would be rapidly sorted away through the tubular extensions that were observed by Geuze et al. and these tubules would form repeatedly from the sorting endosome. In only a few rounds of sorting receptors away from ligand, such partitioning would achieve the fidelity of sorting that is observed experimentally20.
The molecular era: the clathrin coat
Our understanding of the molecular basis of clathrin-mediated endocytosis actually began in 1969. Kanaseki and Kadota examined coated vesicles in membrane fractions and observed that the endocytic coat has a regular cage-like structure that is composed of hexagons and pentagons8. This indicated how the coat might generate curvature and how it can assume different dimensions, by altering the ratio of hexagons to pentagons. This idea was substantiated by a careful EM study by Heuser21. Another important advance came from work by Pearse. She discovered how to purify clathrin-coated vesicles and found that the coat was composed primarily of one protein, which she named clathrin22, 23.
These discoveries initiated a period of intense biochemical examination of the properties of the coat and of the requirements for its assembly in vitro. This led to the discovery that the basic unit of the coat is a trimer of clathrin (named a triskelion)24, 25, 26, that triskelions require other proteins to bind to membranes27, and that proteins that are found in substoichiometric amounts in coat preparations can facilitate the reassembly of clathrin coats in vitro28, 29, 30. In a morphological study of clathrin-coated vesicles that were prepared for cryoEM, Vigers, Crowther and Pearse showed that the 'assembly' or 'accessory' proteins (APs) were positioned under clathrin next to the membrane31. This suggested that these were the components postulated by Goldstein, Anderson and Brown to contain the binding site for endocytic receptors. The cDNAs encoding clathrin32, APs33, 34 and many endocytic receptors were subsequently cloned (TIMELINE). With the availability of cDNAs, researchers could use the strategies of site-directed and random mutagenesis to identify functional domains in proteins, and the internalization signals predicted by Goldstein and colleagues were located and characterized35, 36, 37, 38, 39. Researchers could also express and purify proteins for reconstitution experiments. In practice, this was not easy, and although many laboratories worked on the problem, nine years passed between the observations of Vigers and colleagues and the convincing work by Ohno et al., which showed that AP50 binds to internalization signals40.
None of the early investigators said much about how coated pits pinch off from the membrane. In fact, this is still the least understood part of coated-pit function. However, an important insight into this process was supplied by a mutant Drosophila melanogaster, known as Shibire, which became paralysed at high temperatures41. In an EM study of Shibire flies, Kosaka and Ikeda discovered that, at high temperatures, deep coated pits with long 'collared' necks formed on presynaptic membranes42. Apparently Shibire cells had a defect in the fission of coated vesicles from synaptic membranes. Subsequent studies showed that the tubular–vesicular endosomes (sorting endosomes) disappeared from Shibire cells 7–10 minutes after a shift to higher temperatures43, 44, which is consistent with a 'maturation' model for endocytosis. When the mutated gene in Shibire was finally identified and cloned45, it was found to encode dynamin, a large GTPase that was originally identified as a microtubule-binding protein46. We still do not know why dynamin is required for the fission of coated vesicles from the membrane, and research on dynamin function is now a large and active field47.
In the past decade, our understanding of the molecular details of clathrin-mediated endocytosis has increased enormously. Many new proteins have been discovered that function as cargo-specific adaptors to bring receptors into the coat48, 49. The introduction of the technology of fusing intrinsically fluorescent proteins to proteins that are involved in endocytosis has allowed endocytosis to be observed in live cells with unparalleled resolution in both space and time50, 51, 52, 53, 54. The three-dimensional structures of some clathrin adaptor proteins have been solved55 and, in a tour de force, the crystal structures of individual segments of clathrin have been combined with electron-diffraction data to give us a detailed model of the clathrin coat56. Current work is not only supplying an increased understanding of the details of mechanisms that were first described decades ago, but occasionally also uncovers important aspects that were not anticipated by earlier work – for example, the fact that constitutive clathrin-mediated endocytosis is controlled by signal transduction to a greater extent than was formerly believed57, 58 and that lipid-modifying enzymes are important regulators of endocytic traffic59.
Future directions and conclusion
We currently do not know enough about how endocytic pathways are regulated by cell signalling, and how membrane proteins are sorted in the endocytic pathway to many different destinations. The ability to image endocytosis in real time will doubtlessly be useful for studying these issues. The sorting of membrane into the inner vesicles of multivesicular bodies – a fascinating puzzle both topologically and mechanistically – is currently an area of intense investigation, particularly following the discovery that HIV hijacks the machinery of the multivesicular body for its own maturation60, 61. Non-clathrin-mediated endocytosis, which has a history as old as clathrin-mediated endocytosis and is a story that deserves telling on its own, now seems to be increasingly important, and we do not know how many non-clathrin pathways there are, nor do we understand how they are connected to each other. Of considerable medical importance, we still do not know how cholesterol moves from LDL particles in lysosomes to become part of membranes throughout the cell.
Returning to my original question (is cell biology really different from chemistry?), do cell biologists differ from chemists in the way they use the literature? Cell biology is unlike chemistry in that the older literature is generally not useful for designing today's experiments. New technology in cell biology is almost always superior to the old. However, the danger of not knowing the older literature is that old ideas can be mistaken for new ones.
