This morning, I had a phone conversation with a scientist abroad, talking about a joint paper project. At the end, we agreed on who was going to write which part of the manuscript. My colleague, let’s call him Peter, said something along the lines of “I will write the introduction and will create figure 1, and you should do figures 2 and 3″. I wasn’t too enthusiastic because figure 2 looked like a lot of work, but I agreed.

Only after spending several hours on this blasted figure 2, and another phone call, it turned out that Peter had not said “and you should do figures 2 and 3″, but rather “and Yu should do figures 2 and 3″.

Now we have two versions of figure 2, one by me and one by Yu.

Sigh.

Reading about this story made me think about how acceptable it is to publish stuff in a journal that you have founded yourself and where you are acting as editor-in-chief. Before knowing about the El Naschie affair, I would not have seen problems here. Sure, you could probably select your own reviewers or bypass peer review altogether. But editors-in-chief are typically prominent scientists (right?) and prominent scientists don’t behave that way. Or do they? Maybe I have been too naive. Obviously, I don’t have any fist-hand experience as I never founded my own journal and no publisher in their right mind would ask me to act as editor-in-chief of anything.  On the other hand, I am on the editorial board of a few journals, and in this role you are typically asked by the publisher to ‘submit your best manuscripts’ to the journal in order to make the world a better place. And for increasing the impact factor.

Another reason for my somewhat careless attitude to ‘self-publishing’ is that I have seen several examples of excellent papers published in self-edited journals. Just by looking at the numbers alone, some journals might appear to be in a situation similar to  Chaos Solitons & Fractals. Maybe not quite the same, if the amazing number of 321 El Nashie papers in ‘CSF’ mentioned in Derek’s post is true.

There is one journal in my area, which is dominated by papers from the chief editor:  Biology Direct. Just look at the numbers: Since its inception in 2006, Biology Direct has published 24 papers from Eugene Koonin. This might appear small compared to 321,  but if you consider that the journal has published only 131 papers altogether,  Eugene Koonin has authored more than 18% of all the published manuscripts. Before you jump to wrong conclusions, there are several big differences to the El Naschie case: Eugene Koonin also publishes scores of papers in inconspicuous journals, including Nature and their ilk, he really works at the NCBI (I have proof of that), and his papers in Biology Direct do make sense (to me, at least). Most importantly, there is undeniable evidence that Eugene did not use his position to bypass peer review: Biology Direct has an open peer review policy, the reviewers’ comments are published with the papers!

Post scriptum: After checking out the web page of El Naschie, I have the strong  feeling that there is some problem with this guy. Even without understanding a single equation of his E-learning E-Infinity  theory, someone who sees himself as a  ‘central figure in the field of fractal cosmology’ and at the same time publishes on how to slow the aging process cannot expect to be taken seriously.

I have just been informed that there will be 2nd such conference in September 2009, at the same venue in Riva del Garda. The 2009 title will be “Ubiquitin and Ubiquitin-like modifiers in health and disease“.

Again, the organizers (mainly the Rubicon folks) have assembled a very promising line-up of invited speakers,  including Aaron Ciechanover, Allan D’Andrea, Allan Weissman, Brenda Schulman, Chris Lima, Dan Finley, Frauke Melchior, Helle Ulrich, Heran Darwin, Hermann Schindelin, Ivan Dikic, Jeff Brodsky, Keiji Tanaka, Maria Masucci, Mark Hochstrasser, Matthias Peter, Michael Rape, Michele Pagano, MikeTyers, Paolo di Fiore, Ray Deshaies, Ron Kopito, Simona Polo, Stefan Jentsch, Sylvie Urbe, Wade Harper and Yosef Yarden. This is clearly going to be exciting.

Judging by the presence of Heran Darwin and Hermann Schindelin, the ubiquitin-like modifiers will this time also include their prokaryotic counterparts, including the mysterious Pup pathway in mycobacteria. Compared to 2007, there is another advantage for the prospective 2009 audience: this time you won’t have to sit through 30 min of ubiquitin bioinformatics. It looks like the organizers have learned their lessons and didn’t invite me any more. It is possible, though, that Matthias Peter is going to talk about our most recent collaborative project, which is going to hit the (scientific) tabloid press in January 2009. This one has bacteria in it, too.

Anyway, this is going to be an excellent opportunitly to learn what is going on in the ubiquitin field – especially if you are based in Europe.

I am the last person to claim that microarrays are a perfect tool for tackling all questions conceivable . They are not.  DNA microarrays can be seen as some kind of hammer that is being (rightfully) applied to a few nails, but unfortunately also to lots of objects with no nail-like properties whatsoever. Microarray data are problematic in many different ways. However, we should be careful not to throw out the baby with the bath water.

Here are the main points of criticism that have been raised in the recent posts, along with my comment. I might exaggerate to some extent, but this just serves to make my point more clear.

1) Microarrays are useless, because it has been shown that protein levels correlate poorly with mRNA levels. You hear this argument a lot, especially from the mass-spec people, who want to convince you that only they have a handle on the truth. I admit freely: microarray are mostly useless if you want to learn about protein levels. This is not a nail, go use another tool. You should use microarrays mainly if you are interested in mRNA levels. There are lots of interesting applications for that, e.g. learning which transcription factors are activated. Several stress responses, including those to toxic substances, lead to a dramatic and very specific induction of certain mRNAs. You don’t have to know if the corresponding proteins are really being made, the transcriptional response is the earliest and most specific indicator of many stress conditions. This knowledge can be very useful in its own right. Just don’t try to predict whether there is more protein A in the cell than protein B, just by looking at their microarray signals.

By the way, microarrays are somewhat better in judging changes of protein levels, rather than the protein levels themselves. But still, if protein levels are what you are after, your should turn to another tool.

2) Microrray experiments cannot be trusted because the statistical significance values are wrong. This argument is reiterated here, and the author certainly has a point. Somewhat surprisingly, the examples used in the blog post talk about genetic associations studies rather than the common gene-expression  microarrays. There also seems to be some confusion about the numbers of SNPs vs the number of genes. Nevertheless, the main problem is shared between GWAS and transcriptomics studies: a microarray gives you tons of data and chances that one of the genes appears as strongly regulated just by chance alone is substantial. On the other hand, this ‘multiple testing’ problem is well known in the microarray field and is routinely taken into account. There are methods to correct for the bias in p-value (best known is the ‘Bonferroni correction‘),  Thus, a situation similar to the one described in the blog post would certainly not reach a p-value of 0.05, at least not  in a responsible microarray analysis.

3) Batch effects play a major role and often conceal the real regulation. Admitted, there are batch effects. However, with modern microarray platforms and hybridization methods they can be be safely neglected – at least in comparison to other common noise sources. Obviously, batch effects depend on the technology used. I have experience with three different microarray platforms (two major vendors and and one type developed by the company I work for), and for each of them the batch effects were typically much smaller than the noise from sample preparation or inter-individual differences.

While we are talking about noise sources, here are what I consider the main offenders:

1) Sampling. Particularly problematic when dealing with surgical or biopsy samples. Are you sure that each of your biopsies samples exactly the same tissue structure? With the same relative proportions of cells?  Same amount of blood in the tissue samples? Least problematic when comparing things like treated and untreated cell lines.

2) Inter-individual differences. This problem is ofter under-appreciated but is slowly gaining publicity. Most problematic when dealing with human samples or other outbred (animal-)populations. The differences between ‘healthy’ tissue of two donors are often much more pronounced than between ‘healthy’ and ‘diseased’ tissue of the same donor. Less problematic when dealing with imbred strains or cell culture. Even then, there still might be inter-individual differences related to e.g. nutrition status, circadian effects, etc.

3) Extreme amplification protocols. For many microarray studies, the available material is severely limited. Compliance of tissue donors is often inversely correlated with the size of the biopsy needle. There are several protocols for getting sufficient cDNA for microarray analysis out of very small samples, and some of them are clearly better than others. However, all of them share one common problem: less starting material means more dramatic amplification, which in turn means more noise.

Needless to say that most of these problems can be overcome by using really large sample numbers. Unfortunately, this if often impossible due to limited availability of samples or money. As a consequence, we have to live with the shortcomings mentioned above. I usually recommend that microarray results should not be considered the final outcome of an experiment, but rather as a method for identifiying candidate genes that can be used for a more detailed follow-up study.

For the sake of full disclosure: if you haven’t noticed already, I am working for a company that sells microarrays, microarray services and microarray data analysis. Obviously, this affiliation might bias my view of things. Nevertheless, I speak only for myself and not for my employer. I have tried my best to keep this brief discussion as unbiased as possible, it is just meant to reflect my personal experiences from about 10 years of microarray data analysis.

Despite these gross exaggerations, I do not regret having read the paper, as it has introduced me to a potentially interesting concept called ‘chaperone-mediated autophagy’ (CMA). Judging by the name, this is something I should have known about before – I must admit that this is not the case. One of the reasons is that there appears to be only one group (and their offshoots) working on CMA, They seem to have discovered it, too. If you want to know the details, have a look at this open-access review by J. Fred Dice.

In brief, CMA is a weird way of degrading cytoplasmic proteins. No, not by the proteasome, by the lysosome! After reading the CMA review, there is one thing for sure: had it been my task to intelligently design nature, CMA is not a route I would have taken into consideration. Ok, maybe after a few bottles of a good Chianti. But then, there is ERAD, where lumenal proteins are squeezed back into the cytoplasm by some obscure mechanism, with the idea to degrade them by the proteasome. Why not use a similarly obscure mechanism to squeeze cytoplasmic proteins into the lysosomal lumen for degradation? Apparently, selected cytoplasmic proteins that harbour a certain motif (the KFERQ-related peptide) are recognized by cytoplasmic Hsc70 chaperones together with their usual set of co-chaperones (Hip, Hsp40, Hsp90, Bag-1). Next, this substrate/chaperone assembly is recognized by a receptor consisting of the lysosome-associated membrane protein LAMP-2a, a splice form of the LAMP2 gene. LAMP2 forms a multimeric complex in the lysosomal membrane, which is thought to act as the translocation pore. Apparently, lysosomal chaperones are also required for pulling the unfolded protein through the pore.

This intriguing mechanism has a fundamental flaw, though: it does not involve ubiquitin. As everybody knows, ubiquitin is required for every interesting biological process. Thus, there is clearly some more work to do. It would also help, if a completely unrelated group would contribute to the CMA field. As I have talked about before, I am always skeptic about new proteins/genes/pathways/processes that have only been observed by a single group.

What is the connection between CMA and aging? It had been previously observed (by the Cuervo group) that CMA is abolished if LAMP-2a is knocked down, and that both CMA and lysosomal LAMP-2a levels decline with age. In the new Nature Medicine paper, Zhang and Cuervo have generated a mouse model where the expression level of LAMP-2a in the liver can be modulated. When analyzing aged mice with a continuously high LAMP-2a expression in the liver, the authors found that CMA efficiency did not decline with age and that  the aged liver contained much less intracellular accumulation of damaged proteins. Also, liver function of aged mice with high LAMP-2a level was comparable to that of young mice and better than that of untreated mice at the same age. No matter what you think about CMA, this is a very interesting result.

Update: After doing some more searches, I noticed that there is much more literature on CMA than I previously thought, with different groups involved as well. The “KFERQ” motif has even been reviewed in TiBS as early as 1990. I really wonder how this could have eluded me.

1. Search for the term “ubiquitin OR proteasome”. Check how many entries are found, and read all of them.
2. Search for my own name. Vanity rules. High-scoring matches are typically 15-year old Usenet postings of mine, asking silly questions that nobody cares to answer. Runners-up are caused by other people appropriating my name, including one European scientist accused of scientific misconduct. Yuck. That wasn’t me, I promise!
3. Search for the term “ataxin-3″ or “Rpn13″ (according to taste). If the highest scoring matches start like “comparison shop for ataxin-3 at cooldealz.com” or similar, the search engine gets bonus points.

Here are a few observations with notable search engines.

Google: Finds the largest number of entries, some of them sensible. Lists a certain amount of shopping sites, typically offering antibodies to everything biological that I throw at it (except for my name). For a while, the top entry on Rpn13 was a blog post here at my site, but this time is long gone and my blog post has disappeared in the abyss (i.e.  after position 10). What I hate about google is the microsoftesk attempt to outsmart the user by also showing results with ‘spelling corrections’ applied to, say, my name. Unfortunately, there are lots of people with a name similar to mine, and I am tired to see their pages on top when all I want to see is me, me, me.

Cuil: used to find a lot of entries, most of them with funny pictures of no obvious connection to the query. By now, the pictures seem to have gone, but so do a lot of the hits, too. As of today, the top hit on ‘ataxin-3′ points to a paper of mine (10 bonus points!), but strangely not to the paper itself but rather to a Nodalpoint page linking to the paper.

Gigablast: Not an NCBI product, despite the name. Ugly color scheme. There are two useful features: The categories on top are simple and intuitive and even seem to work (unlike some other search engines with built-in categorization). Even better: There is an easy way to filter for new entries (“freshness=2″ restricts search to entries from the last two days). This is really useful, and I haven’t figured out how to do it in google (although I guess that it is possible somehow)

Vadlo: I learned about this just today from GTO. Supposed to be dedicated to life science content, e.g. protocols and the like. Fails miserably on my benchmarks. I don’t want go buy ataxin-2 antibodies when searching for ataxin-3. And no, I didn’t mean ‘Rp13′ when I typed ‘Rpn13′, thank you very much. The main feature of this site seems to be a daily science cartoon. I had to gasp when looking at todays cartoon (see below, the original is here: Life_in_Research_Cartoon_1138.html). Are they really making fun about a guy who talks about ubiquitination, shows a structure slide that looks a lot like mine, and is supposedly unable to run a decent agarose gel? How could they possible know? But wait, I am much fatter than the guy in the cartoon, I don’t usually wear a tie, and there are more than 3 people in the audience. This cannot be me, what a relief.

. . . . . .

What caught my eye, though, was the statment on author contributions (isn’t this the part of a paper that everybody reads first?). Anyway, this is what they say:

Author contributions: J.G.E., P.R., M.K., and K.O.S. designed research; J.G.E., M.P., B.P.H., A.L., E.G., K.B., and G.W. performed research; J.G.E., M.P., D.E.G., K.S.M., Y.W., L.R., C.B.-A.,V.K., I.A., E.V.K., P.H., N.K., and K.O.S. analyzed data; and J.G.E., D.E.G., E.V.K., and K.O.S. wrote the paper.

You can probably guess what I am hinting at. Seven people “performed research”, while 12 people “analyzed data”, with only a small overlap. It is no surprise that a genome sequencing paper needs a lot of people doing data analysis – this is what genome papers are all about. So far I had assumed that the analysis of genome data should be considered as ‘research’, probably more so than cutting chromosomes into pieces, or operating lab robots and sequencing machines. Apparently, some people see this differently.

In hindsight, I should have known better. At some point in the past, I used to work in a research institute (a very nice place, by the way) with a revealing organization of their phone directory. If memory serves me, they had individual pages entitled “Research group X” and “Research group Y”, listing the PI, postdocs, and students. At the end, they had a page with no particular heading, which held entries for 1) Janitor, 2) Workshop 3) Graphics studio, 4) Bioinformatics. You see, we are just one of these facilities …

Before describing the new model, let me briefly recount the conventional wisdom on proteasome architecture. In the center, there is the 20S core particle, whose structure is very well understood. It consists of four stacked rings of seven different subunits each. The two middle rings contain the beta subunits – three of them with catalytic activity, the other ones are closely related but inactive. The two outer rings contain the alpha subunits, which are distantly related to the beta subunits; however, none of the alphas is catalytically active.

The outer faces of the core particle, formed by the alpha subunits, can associate with at least two different regulatory particles. The more familiar (and probably also more important) one is the ’19S regulatory particle’. The 19S particle itself consists of a ‘base’ structure proximal to the 20S core proteasome, and a ‘lid’ sitting on top of the base. By the way, it was also Michael Glickman (while working with Dan Finley), who showed that base and lid are separate particles held together by the Rpn10/S5a subunit. The lid will probably be a topic for another post; here, I will only talk about the ‘base’ and how it attaches to the 20S core.

The ‘base complex’ consists of at least 10 proteins: 6 homologous AAA-ATPases forming a hexameric ring like most other AAA ATPases, 2 large proteins called Rpn1 and Rpn2 in yeast, and finally Rpn10/S5a, which is the classical ubiquitin receptor of the proteasome and is also required for holding the lid in place. A very recent addition is Rpn13, which is probably also a component of the base.

Archea have a proteasome very much like the eukaryotic one, the same is true for a group of eubacteria, the actinomycetes (which include the notorious Mycobacterium tuberculosis). These prokaryotic proteasomes are somewhat simpler: they contain only the 20S core (made from homomeric rings of only one alpha and beta subunit each) and a hexameric ring of ATPases very similar to the eukaryotic base. In prokaryotes, there is no doubt that the ATPase ring is stacked on top of the alpha ring of the 20S core, and the same architecture was assumed for the eukaryotic base. The big Rpn1 and Rpn2 proteins are usually depicted as blobs, sitting somewhere on top of the ATPase ring, together with Rpn10 forming a ‘hinge structure’ for attaching the lid.

This architecture has now been challenged by the new paper from Michael Glickman’s group in Haifa. He describes a number of results from very different approaches, which all converge on an alternative model in which Rpn1 and Rpn2 are more proximal to the 20S core than the AAA ATPases are. Apparently, the whole thing got started by a bioinformatical paper by Andrey Kajava, coincidentally a former office-mate of mine during my Lausanne years. Andrey had predicted that the repeat motifs of Rpn1 and Rpn2 would give the proteins a toroid structure. When this first paper came out, it was greeted by some skepticism, in particular from the cryo-EM people who did not see a convincing toroid electron density on top of the ATPase ring.

Rina Rosenzweig in Michael Glickman group took this idea one step further; they collaborated with Pawel Osmulski and Maria Gaczynska, who use atomic force microscopy (AFM) for studying macromolecular assemblies. According to the published AFM data, both Rpn1 and Rpn2 form individual rings. When analyzed together, a ring with the same diameter was observed, which was twice as high, suggesting that Rpn1 and Rpn2 stack on top of each other. Crosslinking experiments showed that Rpn2, but not Rpn1, appeared to be a direct neighbor of the 20S core particle: according to MS analysis, Rpn2 was crosslinked to five different alpha subunits.

Finally, a series of atomic force micrographs of 20S core particles reconstituted with various factors, had been obtained. I think the figure to the right looks quite convincing. The rightmost panel shows the 20S core with a relatively flat surface. In the 2nd panel from the right, Rpn2 appears to form a circular structure on top of the 20S core. In the 3rd panel from the right, Rpn2 and Rpn1 combined result in a circular structure of higher profile (bottom row), while the addition of the full base particle, including the ATPases, does broaden but not further heighten the profile.

The obvious model for explaining these results looks like the one shown on the left. The two rings of Rpn2 (yellow) and Rpn1(red) form a funnel-like structure on top of the ‘entrance’ of the 20S core particle (green). The six AAA ATPases (orange) form a ring encircling the funnel.

The two most important questions are i) is this model correct? and ii) will it have a strong impact on the future of proteasome research?

I mentioned above that the reactions to the new model and the underlying data have been mixed. During the last year, I have discussed it with several people from the field and heard comments ranging from ‘certainly correct, is in perfect agreement with my new data’ to ‘complete rubbish, my new data prove that the model is wrong’. I for myself find that the data look convincing, but I am not much of an expert in atomic force microscopy. I am aware of the specificity problems with photochemical crosslinking, but as far as I can tell this part of the paper doesn’t look too bad, either. So, judging by the data, I don’t think that the new model can be easily dismissed. What I don’t like about the model are two things. First, I find it hard to believe that the arrangement of core particle and ATPases should be different in eukaryotes and prokaryotes (which lack Rpn1/2). Second, I see a problem with Rpn2 being completely buried inside the proteasome structure. After all, Rpn13 has been found to attach to the complex via binding to Rpn2, which is not really compatible with its purely internal position. Michael (sort of) addressed this problem by admitting that some part of Rpn2 might stick out somewhere. I don’t know what to think, probably time will tell.

With regard to the relevance of the new model, it is interesting to note the complete absence of both associated ‘news & views’ material and press releases claiming new avenues to treat cancer. I find this somewhat surprising, as this new finding is much closer to a ‘gatekeeping’ role than Rpn13 is. It should be kept in mind that the entrance pore of isolated 20S proteasome core partciles is closed. For opening this pore, an interaction with the base complex is required. According to the new model, Rpn2 would be in the best position to do this job. I am convinced that – as soon as more data supporting the new model becomes available – a lot of the accepted truths about ubiquitin recognition by the proteasome and shuttling substrates into the 20S catalytic chamber will have to be reconsidered.

Addition: After discussing this post with Michael Glickman, I should point out that even in his new model there are contacts between the ATPases and the 20S core. Thus, the new model is not that much different from the archaeal proteasome. The major difference is that in the revised eukaryotic model, the opening in the ATPase ring is large enough to accomodate the toroidal Rpn2/Rpn1 subcomplex.

This line reminds me to some degree of scientists doing social networking. I am not so much thinking of Facebook and the like, but rather of their scientific siblings like Nature Network an SciLink, as they seem to gain popularity in the scientific blogosphere (see here, here, here, here, and here). All of these services ask you for your affiliation, workplace and several other obvious and semi-obvious data. The scientifically inclined services also ask you for your publication list. This can be a bit tedious, depending on how prolific you are. I imagine that Eugene Koonin would have to hire a summer student for entering his publication list.

When all of these questions have been answered, you are asked to build a network by connecting your entry to that of ‘friends’ who are also represented in the system. This is where – at least for me – the problems begin. As everybody knows (or at least has always assumed), real scientists® don’t have friends. Just like the cool guy in the movie.

What we do have are colleagues, competitors, co-workers (and maybe some more c-words). In terms of scientific social networking, they would pass for friends,  I guess. The major problem here is that they are typically not present on the same services. And if they are, it can be very hard to find them. I tried it myself,  typing in just about everybody I know to be active in my area of work. The result – nada. Things look different with science bloggers – they all seem to be present. Almost everybody, almost everywhere.

I am obviously not in the networking business (and happy about it, too) but I have invested some thoughts on how to automatically find (or guess) related souls present in the social network. The only possibility I can think of is the one large, untapped resource of science networks: the publication list. It should be possible to construct complete co-authorship networks, based on who authored a paper together which whom. Obviously, this network would contain a number of highly connected nodes, either due to very prolific and collaborative people, or due to people with names like ‘Smith, J’. Nevertheless, this would be an interesting resource that I have not seen implemented so far.

Even if you are not going for the full network, this approach might help the networking sites to find other network members has been your co-author (or more indirectly, a co-author of your co-author). This candidate list could be presented to you and might help you to identify which other network members might be of interest for you. Maybe this approach has been tried before, but everything I have seen so far is either based on geographical proximity of your working place, or on matches between tags and keywords, which the network users might have assigned to themselves.

Just for fun, I have tried to (manually!) analyse my co-authorship relations to a select groups of people: bloggers found in my blogroll. I haven’t tried all of them  – this work can become quite tedious if more than one intermediate co-author is involved. However, for all science bloggers I have tried so far, I was able to find a connection with two or less intermediate authors. Here are some examples:

Direct co-authorship

I found only one example blog called ‘Research Highlights from the Aravind group‘ In the past, I have co-published occasionally together with L.Aravind and people in his group.

One intermediate author

Paweł Szczęsny (Freelancing Science) ↔ Andrei Lupas ↔ me
Roland Krause (nftb) ↔ Peer Bork ↔me
Jonathan Eisen (Tree of life) ↔ Eugene Koonin ↔ me
Lars Juhl Jensen (Buried treasure) ↔ Peer Bork ↔ me

Two intermediate authors

Ian York (Mystery rays) ↔ Alfred Goldberg ↔ Daniel Finley ↔ me
Pedro Beltrao (Public rambling) ↔ Luis Serrano ↔ Peer Bork ↔ me
Neil Saunders (what you’re doing..) ↔ Bostjan Kobe ↔ Andrei Kajava ↔ me
Jason Stajich (fungal genomes) ↔ Ewan Birney ↔ Philipp Bucher ↔ me
Keith Robison (omics omics) ↔ Emad Alnemri ↔ Vishva Dixit ↔ me

Sometimes, it was easier than expected to find a link to somebody working in a very different area. In other cases, I found a very strong first-degree link to somebody working over years on the same subjects as we were, but always as competitors – no common publication.

If there is anything the non-expert knows about ubiquitin, it is the fact that ubiquitin becomes attached to other proteins and earmarks them for destruction by the proteasome. In order to do this job, the proteasome has to recognize if a protein carries a ubiquitin degradation signal. The complete 26S proteasome consists of at least 32 different stoichiometric subunits and some of them are thought to act as ubiquitin recognition components, or ‘ubiquitin receptors’. There is no shortage of contenders for this job, though. At least three different ubiquitin recognition systems had been described previously:

Rpn10 or S5a, depending on if you are a yeast or mammalian person. Or PSMD4, if you are a nomenclature person.  The ubiquitin recognition properties of this 19S subunit, positioned at the interface between base and lid of the 19S cap, have been known for more than 10 years. The ubiquitin binding is done by a short helical motif, which was later shown to occur in lots of other ubiquitin binding proteins and has been called ubiquitin interaction motif or UIM.

Rpn1 or S2 (PSMD2) is another subunit of the 19S base complex involved in the recognition of ubiquitinated targets. This recognition, however, is not direct. Instead of recognizing ubiquitin itself, Rpn1 recognizes ubiquitin-like domains that are present in a class of ‘adaptor proteins’. These adaptors contain  a ubiquitin-like domain at their N-terminus, while the C-terminus carries a UBA domain, another widespread class of ubiquitin recognizing modules. It is this UBA domain (in proteins like Rad23, Dsk1, Ddi1) that is responsible for the actual recognition of the degradation target.

Rpt5 or S8 (PSMC3) is one of the six AAA ATPases that form the base complex of the 19S proteasome cap. All six ATPases are quite similar to each other, but Rpt5 has ben singled out as ‘the’ receptor for polyubiquitin chains, which constitute the degradation signal. I don’t have an opinion on this myself, but I have noticed that most people in the area don’t believe in Rpt5 being a real ubiquitin receptor under physiological conditions.

This was the situation before the Rpn13 paper appeared. I will summarize the contents of the paper further down, but you can also read the associated news & views article to get a better idea. Let me briefly recount the story behind this paper and my (admittedly very minor) contribution. In 2000, Rati Verma in Ray Deshaies group performed a proteomics screen in yeast, looking for new proteasome-associated proteins. One of the proteins they focused on was Rpn13, which they described to be a new stoichiometic proteasome subunit. When I heard Rati talk about their results on one of the early Zomes conferences, we began looking for a human version of Rpn13. To our surprise, we came up with a protein known as ADRM1, described as a cell adhesion protein. Not what you would expect for a proteasome subunit. Nevertheless, I was sure that we were right, and expected the cell adhesion phenotype to be an epiphenomenon, if not an artifact.

Some years later, in 2004, Grzegorz Zapart in the lab of Ivan Dikic in Frankfurt performed a Y2H screen for novel ubiquitin interactors. Among the long list of putative interactors, many of them very interesting proteins, was ADRM1 – the adhesion molecule. Now, this is my only real contribution to the paper: when I saw the list of interacting proteins, I remembered our identification of ADRM1 as the human Rpn13 ortholog and thus a likely proteasome subunit. I felt that a new proteasome subunit binding to ubiquitin would be something to write home about, and recommended to Ivan that he should focus on ADRM1 rather than the other interactors. It (kind of) worked, and Koraljka Husnjak did the further characterization of the protein, showed that it really binds to ubiquitin, mapped the interaction region to the N-terminus, and found lots of other interesting details about the protein. Next thing, the collaboration was joined by the group of Dan Finley (Harvard), who had shown that a simultaneous knockout of all known proteasomal ubiquitin receptors in yeast still supported ubiquitin-dependent protein degradation. Thus, it was an obvious idea to test if Rpn13 is responsible for this residual activity.

In the meantime, several other groups (1,2,3,4) demonstrated that ADRM1 really is a proteasome subunit. With these publications, my own contribution to the current project became virtually obsolete, except maybe giving Ivan’s and Dan’s groups two years head start. It was also shown that one important function of ADRM1 is to anchor the deubiquitinating enzyme Uch37 to the proteasome. This could not be the whole story, though, as yeast has an Rpn13 but no Uch37. Later on, the collaboration was further extended by adding two structure groups: Kylie Walters (Minneapolis) solved the NMR structure and Michael Groll (Munich) did the X-ray structure of ADRM1 (or hRpn13, as it is called now) in isolation and bound to ubiquitin. The majority of the structure work is published in a second paper, although some of the NMR work features also in the main article.

The ubiquitin binding mode of Rpn13 is somewhat unusual. Most other ubiquitin binding domains are shared by many proteins, use a contiguous binding surface and use one or more alpha-helices for contacting the Ile-44 patch of ubiquitin. The N-terminal ubiquitin binding domain of Rpn13 has a beta-sheet fold, very similar to the PH domain despite the absence of sequence relationship. Contact to ubiquitin is made through multiple non-contiguous loop regions, and no homolog outside of the Rpn13 family could be detected. The above figure shows yeast Rpn13 (botton) in contact with ubiquitin (top). Highlighted in red are three Rpn13 residues, which – when mutated – no longer support ubiquitin binding. Two other noteworthy features of ubiquitin recognition by Rpn13 are the unusually high affinity (Kd of 300 nM for monoubiquitin, 90 nM for diubiquitin, compared to the uM Kd valued measured for other ubiquitin receptors) and the ability to also recognize ubiquitin-like domains besides ubiquitin itself.

What is the significance of the finding? Rpn13 is certainly one of the more interesting ubiquitin recognition modes by the proteasome. Is it a ‘New Target for Cancer Drugs’, as the press releases ask (see e.g. here) ? Not very likely. Is it true that ‘A discovery of this kind happens only once in a researcher’s lifetime’, as Ivan Dikic is cited in the release? I hope not. As usual, the press releases on this publication (there are several ones) consist mainly of hot air. At least they agree with me on the insignificance of my contribution, illustrated by dropping me from the list of groups involved. That’s fine by me, though.

One final question: It this the end of the line for proteasome ubiquitin receptors? Chances are that there is still more to come. Apparently, yeast cells deleted for Rpn10, all UBL-UBA adaptors, and Rpn13 are still viable (while deletions of the catalytic proteasome subunits are lethal). Either there is still one more ubiquitin receptor around that would support some basal level of recognition and degradation, or ubiquitin independet proteasomal degradation is more important than expected.

P.S. As usual, other bloggers have beaten me in reporting on Rpn13. See e.g. Think Gene.