When I arrived here we had the essential ideas on how to put everything together in vitro, and how to make all the components homogeneously. That way the histones would not be modified, as they are by making them in E. coli, for example. We knew how to make the DNA extremely pure, and not by attempting to do it via straight synthesis. I started to take advantage of new recombinant DNA technology to piece together homogeneous DNA target sequences.

We had to come up with the right DNA sequence. There are several that work and we chose a particular one. (We have others now and are trying to find many more in the search for very high resolution).

We were able to prepare crystals of the chosen DNA fragment. Nothing quite like it had been done before. By good fortune we had access to the European Synchrotron Research Facility (Grenoble) before commissioning had finished. We needed a lot of beam time to get the phases for our diffraction patterns. This was exciting experimental science, seeing what would work and what would not.

In the end we modified the proteins by expressing variants that would bind heavy atoms, and doing enough of these so we would eventually come up with crystals made from good heavy-atom derivatives. Finally we had two heavy-atom derivatives that worked, and that was enough to crack open the structure at 2.8 Å.

The work in the paper was not only the first third-generation high-intensity synchrotron. We had to develop our technique as we went along. For example, our first data sets revealed that we were causing radiation damage, which we solved by having a larger number of crystals and giving each a shorter exposure. These are the sorts of things to battle with in research!

This is structural biology, a field in which you do what you can do. 90% of the effort in this laboratory at any given time is on biochemistry, crystallizing the target materials. Why do we do this? Let me explain.

It is very difficult to understand a material’s relation to structure without knowing what it looks like, and that is why we do it. It was always obvious that in looking at the nucleosome it was bound to have an impact if we could get the structure. Our atomic structure of the nucleosome core particle of chromatin appears in virtually all molecular biology textbooks published since 1997. For us, the science driver is to understand biological regulation: knowledge of these processes at molecular level will have dramatic consequences for the treatment of human disease.

The potential for DNA structure is still to come. What is most important is to see what it looks like in terms of other DNA binding proteins and regulatory factors, and how these structures differ from the structure in solution. How does it repair itself, for example? This question is at the base of many diseases, including cancer.

In eukaryotic cells the right context is to look at it in chromatin, and that is where the impact of our structure comes: it is the structure of DNA in the context of the way it actually is in the cells. We cannot make progress by looking at interactions between naked DNA and proteins, because that’s not the real world of the life of the cell. Ultimately there is no question in my mind that these structures will be related to biological function in important ways.

In 2002 we presented a paper on the structure of the nucleosome at 1.9 Å resolution. We improved the clarity of the electron density and the accuracy of the atomic coordinates for the histone protein and DNA. The clearer picture has enabled us to do extensive modeling of the water molecules and ions for the first time. The idea was to show why or how the DNA in the nucleosome is different from free DNA.

One of the things the nucleosome is famous for is blocking other proteins from interacting. It occurred to me that the DNA is effectively disguised in very tight bending, and in 2003 we showed how just remarkably curved nucleosomal DNA is.

Currently we are going to factors that interact and looking at them structurally, in combination with the nucleosome. We are trying to look at large complexes that actually move the histone octomer around on the DNA (Chromatin remodeling factors). We are also looking at multiple nucleosomes, and trying to understand the next level of organization of chromatin in its most condensed form, called chromatin fiber. There are well organized but less dense forms which we refer to as nucleosome arrays.

For the last 25 years I have been looking at the interactions between viruses and cells, but not from the point of view of understanding infectious disease. My interest is in seeing how the virus is utilizing the cell machinery. Simply by following what an invading virus does, we can learn more about the cell. We’ve learned many different things over the years.

In the mid-1980s we started on the mechanism of how viral proteins fold as they are assembled and modified by the cell. We learned that viral glycoproteins are synthesized in the ER, just like normal cell glycoproteins, from where they are transported to the surface. As we looked at this we also knew that virus was following other rules using the machinery that cells were using. When you infect a cell with a virus it is just wonderful for looking at protein maturation. We were looking at influenza hemoglutonin synthesis: we found it is synthesized as a glycoprotein with several sugars on it; the sugars are subsequently trimmed prior to transport to the surface.

The trimming enzymes are assembled in the ER. It looked to us as if the folding and polymerization happens in the ER, and only when the protein is completely finished does it leave the ER. So, by looking at inhibitors and also at mutants, where something goes wrong with the protein folding, or the trimmer does not form, we realized that only properly formed proteins are transported. Proteins that have something wrong with them are not transported. So that brought up the consumer term "quality control:" inside the cell there is a system that can distinguish different components and decide which to reject. This much had been seen before.

Our contribution was that we were able to generalize: we said there must be a system within the ER which is able to look at each nascent protein one by one. It turns out that the control system is completely generalized: any protein that is synthesized has to pass a quality-control system whose principles we are still trying to sort out. If there is something wrong with the protein it remains in the ER.

There is a whole list of human diseases where faulty proteins are the cause: cystic fibrosis, scurvy, all sorts of diseases where there is something wrong with the protein that’s made, such that it cannot fold into its completely normal form, and therefore it cannot leave the ER, where it is recognized as faulty by the quality-control system. In the case of cystic fibrosis patients the faulty protein is still functional, and if we could make the quality control less stringent these patients might survive.

We are working now on the quality-control mechanism. What happens is this: a faulty protein is made. It is given maybe half an hour to two hours to fold. If the synthesis works the protein immediately leaves, but if it is not folded it is taken out of the ER and degraded. The question is: how are the proteins sorted, given that some of the faults are very slight? We found that, even without any mutations, about 30% of proteins made in the cells are junk: they never fold correctly! Quality control is an on-going process in a healthy cell.

We focused on glycoproteins because there we could see how quality control works. Glycoproteins are proteins which have been modified by carbohydrate molecules, and they are common in the ER. Sugars are added as a polypeptide chain emerges from the ribosome into the ER. An enzyme adds a polysaccharide with 14 different sugars. Immediately after the saccharide is added, other enzymes start to be take sugars out, one by one; this is called trimming, a well-known phenomenon. Every cell biology student has to memories these steps but nobody has been able to supply the logic: this process exists in every organism from yeast to man!

We have shown that each intermediary has meaning for the folding machinery. There are "elected" sugar-binding proteins that grab the new proteins in different ways, and these changes are a record of how the protein is proceeding in the folding pathway, or how it is not proceeding if something is wrong.

We know how unfolded glycoproteins can be recognized and, as it were, recycled for another attempt at folding. Most aspects of the quality control are now understood: the folding promoting chaperones, the adding and subtraction of sugars in the cycle, and a timing sensor (lysosome) for detecting older proteins that won’t fold and must be destroyed.

Understanding these mechanisms is important because two-thirds of the genes in the human body code for glycoproteins. Not all glycoproteins require this system to fold, but nevertheless they all go through it. For individual proteins families, such as, say, collagens, there are additional control systems. Every protein has a specific control system: mistakes are not allowed inside the cell. The system does not test for functionality, but rather the architecture, shape, if you will, of the protein molecule. We are trying to understand the sugar transfer processes. The field is just beginning, really—there must be other sensors for the non-glycoproteins, and we have groups here working on different aspects of this. There is huge interest in the details of the cellular control mechanism, and that’s why our Science paper is so highly regarded.

   Thank you. I would now like to move to a second paper of yours, on the transport pathway to the ER of simian virus 40.

The SV40 virus was discovered as contaminant in polio vaccine grown from monkey cells. Members of this family are oncogenic. It is not pathogenic in humans, but it is in monkeys.

For about 27 years now one of the main lines of work in our laboratory is to do with virus-cell interactions: how do viruses enter cells in order to replicate? 25 years ago we discovered a pathway by which many viruses bind to the surface of the cell, and then the cell makes the mistake of internalizing them. They are transported by clathrin-coated vesicles and they penetrate the cell membrane. The virus with its modified coat proteins fuses with the membrane and gets through to the cell. This pathway explained about the entry of about two-thirds of viruses. Many animal viruses take advantage of this receptor-mediated process to enter the cell.

But there were always some viruses that did not fit the picture. I had worked with an electron microscopist (Jürgen Kartenbeck). By electron microscopy we found that SV40 bypasses most organelles but invades an organelle newly discovered by us, and then ends up in the ER. From here the virus moves to the cytosol and thence to the nucleus. What’s new about this paper is that the study of the virus reveals a completely new and very complicated pathway, involving a new organelle. The pathway bypasses endosomes and the Golgi complex, and is a new infectious route.

Mainly because of the technique we used. We gave the virus a fluorescent coat, but did not inhibit its infectivity, which meant we could then follow it as a visible spot in the microscope. We could then follow with video microscopy, as it first enters the cell. This technique is becoming widely used for following particles through cells.

MLP was discovered as a protein essential for myogenesis of skeletal muscle. It was found as a protein that is highly expressed after denervation of skeletal muscle. Silvia Arber, together with Pico Caroni (co-authors of the paper), found this compound that they could show is one of the essential elements, a transcription factor, that takes skeletal muscle cells towards their final form.

My collaborator Pico Coroni was more interested in the neural connotations of this discovery. He and Silvia went on and did the knock-out of the gene, which, surprisingly, produced a phenotype that was highly specific for the heart. So the defect was much more in heart muscle than skeletal muscle, and this had immediate implications for the study of heart failure.

At this stage we at ETH were yet to be involved. But it was a stunning discovery, and I invited Pico Coroni to our lab, which was already established in the investigation of cardiac cytoarchitecture. Cardiomyocyte development is intimately associated with cardiac differentiation.

One of the major findings was that there was a type of dilated cardiomyopathy in MLP-deficient mice that was similar to the human disease. This also involved other collaborators. I had been at Ken Chien’s (co-author) laboratory in San Diego earlier, and I knew about his superb ability to analyze cardiac phenotypes, as well as the work of John Ross, who was developing these physiological methods. I urged Pico to visit their lab; Ken took their mice and was then extremely helpful in establishing the phenotype. Clearly the finding led to the characterization of the phenotype. There were many facets of dilated cardiomyopathy as similar as a mouse phenotype can ever be to human disease.

The San Diego colleagues very quickly realized the importance of discovering a phenotype with the hallmarks of human dilated cardiomyopathy. Within a year we were getting publicity. Chien’s lab actually monopolized this mouse, and it now sets the gold standard in research on dilated cardiomyopathy. The spread of the literature on MLP deficiency was due to two things: the highly efficient apparatus that Chien ran, plus the ease of keeping the mice (many mutations are lethal), later generations of which have proved to be rather stable.

That’s easy. The MLP-deficient mouse is now the gold standard for models of dilated cardiomyopathy! Heart disease of course is a hugely important area of medical research.

Our job is to identify major genes for the use of linkage maps. We have our own research station with dairy cattle, pigs, and sheep. We are establishing three generation research projects, with some 150-200 markers. Once the map is constructed we run a genome scan to locate the major functional genes relative to the markers.

A lot of money has been spent to help developing countries create a resource population which can be genotyped, and from which we can then find the important functional genes, with an objective to find major genes for important production functions.

I’ll start by explaining what a genetic linkage map does. How can we describe the genome of a living organism? It can be described in terms of the length of the genome, the number of genes, what gene actions are present, and how they are located. The physical map says which genes are present and in which sequence. A linkage map is a map of DNA markers (not necessarily a functional gene). To construct such a map you need polymorphic markers, hopefully equally spaced on the genome.

To construct a linkage map we first need to define a genetically structured population of the organism. The parents and offspring are then genotyped across three generations using 150 or so markers. You can think in terms of "page numbering" the genome. Once this is achieved, highly mathematical mapping functions are used to create the multi-point linkage, which will produce the sequence of marker order and marker distances.

The genetic linkage map can be linked to a physical map. Many institutes now look at comparative genomic mapping, so that if a certain gene is located in one species in a certain location, will that same gene be found in a similar location in another species?

That’s to do with what the map is used for. It is useful in different ways. First of all, we can associate certain chromosomal regions in the organism with how they perform; it’s all about breaking the genome of animals or humans or even plants into molecular genetic components, so that we can quickly say "this particular part of this chromosome has an influence on performance."

Obviously this sort of work has been going on for a century now in selective breeding, but not in terms of the complex molecular structures. Today we are looking at simple characteristics, such as whether certain diseases in humans or certain disorders in animals are related to one gene that is common to more than one species.

In the case of animals we are interested in economic characteristics such as milk yield, meat etc. These properties are not controlled by single genes, but by polygenes. The linkage map helps to find the genes that have a major impact on milk or meat yield. The linkage map ties markers to genes, and the map helps with marker-assisted improvements in conventional breeding programs.