Slide 1

Many thanks for the invitation to beautiful Barcelona in early summer. I’m very excited to present a progress report on our recent developments in the transcriptional analysis of human renal biopsies. We began with the same dilemma everybody faces in nephrology, namely that we classify our patients according to descriptive disease categories, which results in having multiple different disease mechanisms for the inflammatory renal disease present in the same diagnostic category of patients.

Slide 2

This has significant implications, because we end up with what I call the ‘mixed bag’ problem: we lump our patients together with the diagnosis of i.e. IgA, but we have different mechanisms active inside their disease category. This results in our not being able to give good prognostic information to our patients and not knowing which subgroup of our patients will respond to the available therapies for this disease. The result is that treatment becomes an ‘art of trial and error’ in nephrology as we try all the therapies available in the hope that at least one will finally work.

Slide 3

As you will see in this slide which is 250 years old, some things are not new. Here, an early nephrologist is looking at the urine of this nephrotic and anemic lady – which we still do – and coming up with the same descriptive disease categorization we do. He told her that the phlegma is not leaving her body, and he will probably give medications as toxic as the ones we prescribe, received with same doubts as today, at least if you look at the family, and alternative modes of intervention which are also already consulted.
(At the time my father gave me the above picture – he recommended that I should in contrast to this gentlemen at least wear a tie….)

Slide 4

So - what is the way out of this dilemma in nephrology? How do we eliminate the ‘mixed bag’ approach? I think we should consider a shift in how we categorize and define our patients and their diseases and move on to what we now call mechanism- based personalized patient management. In other words, we try to define the disease process active in each individual patient and then base the prognostic information we give that patient, on that mechanistic insight into the disease process. We discover which stage of inflammation activation they’re in and also attempt to target our therapy to interfere with these specific mechanisms actively destroying nephrons. This is already becoming the standard in oncology. I think in nephrology we have a unique chance to implement that as well because we are a tissue-based speciality of internal medicine so we should be able to harness the power of molecular medicine available nowadays for our patient management.

Slide 5

To do that, however, we needed to envision developing a very structured approach. That’s what we did nine years ago with the leadership of Detlef Schlondorff in Munich Germany to develop a tissue procurement protocol which applies a molecular medicine technique to renal biopsies to generate an expression map for this new landscape, an gene expression atlas of renal disease, to then hopefully extract relevant disease markers and pathways, one that is finally designed intrinsically into the study to develop strategies for clinical implementation.

Slide 6

This protocol that we designed is now known as the European Renal Cell DNA Bank (ERCB) protocol. Part of a renal biopsy is taken, and then transferred into biopsies within an mRNA stabilizing solution that goes to the core facility where we micro-dissect into nephron segments. These undergo focused analysis or genome-wide expression profiling, then we con-currently file them in a databank system that captures the clinical data and histology parameters of our patients.

Slide 7

That project has been ongoing for the last eight years. Through an EU funding framework, we have been able to accumulate more than 2,000 biopsies capturing the standard categories of renal diseases. Many of you in the audience are very active collaborators in this exciting project.

Slide 8

We are extremely grateful for that, because together we’ve been able to begin studies where we can now map what genes are regulated in human renal disease in a nephron segment-specific manner. After some optimization of the technologies we profile in a genome-wide manner capturing more or less all mRNAs analyzed in this tissue with several control populations and a variety of primarily glomerular-based diseases.

Slide 9

With this currently available information, we then had to develop a highly standardized reproducible approach to process this data so that it is comparable between different disease categories and, further, apply it to other published literatures. In that endeavor, We’ve been very fortunate to work with the National Centre of Integrative Biomedical Informatics at the University of Michigan. It allows us to have an expertise core facility available to develop bioinformatic tools together with leading bioinformaticians and then transfer the use of those tools into data analysis systems and pipelines to exploit and extract the most useful information from these large databases.

Slide 10

For the remainder of my talk, I’d like to focus mainly on the molecular definition of renal disease. This is work in progress, and we’re clearly very early in the process with a lot to learn, but I’ll outline our current approach how we could use that information. Then I’ll give you a couple of slides on how we can actually use that information to learn about disease process just beyond the marker identification.

Slide 11

The main challenge in our studies is to breaking a large signature down on a small set of molecules which might be of direct clinical application in our data set. With these genome-wide approaches, we have focussed over the last years our studies and they have been published over the last 4-5 years where we have sets for very defined differential diagnostic questions, for example, steroid responsiveness of a nephrotic syndrome, CD20 activation in membranous and others.

Slide 12

And they can be combined on a molecular diagnostic marker platform to evaluate them in independent patient cohorts, if they are truly able to predict outcome of our patients prospectively.

Slide 13

I want to give you a quick overview about our systems biology approaches of renal disease and how we can identify molecular pathways. In this, we are now obviously exposed to the tremendous challenge of having more than 18 million expression data points from 10 renal diseases. We therefore need to develop an effective strategy to extract regulated molecules and pathways relevant overall in an unbiased manner and also relevant for the respective research flows. This morning I presented our studies and how they were integrated in the long standing effort of our collaborators  in Madrid on apoptosis and diabetes. That is a prime example of how these tools can actually facilitate research and expose the relevance of animal models and in vitro systems to human pathology.

Slide 14

I would like to use one of our studies recently published in Diabetes where we used this comprehensive data set and used the sequential filtering step and integrating different systems biology tools to come up with a testable hypothesis and define using it.

Slide 15

So we started out with genome-wide expression profiling, then used a categorization tool which assigns a specific function to each gene to define in which molecule or category these genes were relevant, then imported them in pathway maps where defined regulatory cascades are displayed.  

Slide 16

And, in this instance, we use them to define that inflammation is driven by NF-kB in renal tissue in diabetes, that we could use this NF-kB dependent signature to group our patients to progressors or non-progressors and then, using transcriptional promoter modeling tools to actually define a novel specific module NF-kB IRF1 to be activated in diabetes, predict gene expression signatures. We would then use these predictions to verify processes actively going on with progressive loss of renal function in diabetes. That has been published. If you’re more interested in these type of approaches, please feel free to have a look at the manuscript and contact me for further questions.

Slide 17

One clear aspect which is emerging and will become more pertinent in years to come is that these data sets are potentially tremendously useful for everybody doing renal research. So a key challenge is how to make these data sets, if they published, available in a manner that they can be integrated in ongoing research projects.
One aspect which we used in our ERCB study was that we had direct supervised mining in specific research contexts using individual application of software tools. This is obviously effective, but very labor- and cost- intensive and not practical as a long-term approach.

Slide 18

There are publicly available depositories, and most journals nowadays require expression data to be deposited in these public blood forms. Our diabetes data, for example, are deposited online  and in theory extraction of expression levels of defined molecules can be done, however that clearly requires expert knowledge which is very limited in linking with relevant clinical information. So one option would be to develop a kidney-specific search tool which allows disease-focused data base mining and integration of these data with one’s ongoing research. The first tools have actually been implemented in oncology. They do that very effectively by the Oncomine team from the University of Michigan with system tools used by more than 8,000 users a month as it allows datasets from the public domain to be imported. It has a very intuitive user-friendly mining tool.

Slide 19

We are currently actively working together with the Oncomine team to develop a similar approach for a renal data set and I’d like to give you a quick overview here where we use publicly available data pertinent to the kidney and kidney disease. We have extensive data annotation tools where comparability between the different blood forms is assured, then via a web- based server these data sets can be queried for expression of specific molecules. One can even see if there are subgroups of differentially-regulated genes present in datasets. Poor expression of molecules can be displayed, enrichment for defined functional categories can be seen and even complex pathway mapping tools. That’s a flexible system where novel emerging databases relevant for renal diseases can be easily integrated, however it is really dependent upon community involvement and submission.
We currently have the hardware in place and are starting to upload publicly available datasets of renal diseases. Please feel free to visit the website. If you want to take a look at www.nephromine.org that clearly is an early better version. We hope to have the relevant data uploaded by fall 2007 and to have it operational in a manner that I truly can offer that as a service to the communities.

Slide 20

So, to sum up I have presented how one can envision using the emerging tools of systems biology for a personalized molecular nephrology approach. Also how identification of candidate molecular markers out of renal biopsies can be envisioned. This is clearly only a starting point. We’re far from complete, and it will take many more years to come up with defined markers as we have on immunofluorescence levels. How transcription pathways can be defined and I’ve not gone at all into how promoter modeling tools can be used to define comparative genomics for tissue and functional regulation of genes. The next steps are required to evaluate the diagnostic power of these tools, to modulate identified pathways and functional studies and to develop web-based search tools to identify them.

Slide 21

In the end, and most importantly, I would like to thank all the members of the ERCB for their participation. Clemens Cohen, who is coordinating the ERCB study out of Munich (now moved to Zurich) and his team and the collaborators in Michigan who are helping us one way or the other in our data analysis, and the reference pathologists of this study, Maria Pia Rastaldi in Milan, Herman Joseph Grone in Heidelberg and Paco Mampaso from Madrid who tragically died last year. I would like to devote this talk in his honor. Thank you very much for your attention.

Slide 22

Chairman: Thank you very much Matthias for sharing with us this innovative approach, as well as new data. Because of lack of time we have time for only one or two questions and short answers.

Question: Congratulations this is an enormous amount of work and so exquisitely analyzed. I have a simple question that I’m confronting doing some proteomic analysis and it is what does one do considering that the disease process perhaps changing constantly depending on the staging constantly changes its makeup? Do you have to develop special sets for different stages of the same disease process or are there clever tricks to overcome it?

Prof Kretzler: Michael thanks a lot for the comment. That really is one of the areas where everybody in the field is struggling right now. The renal biopsy is a snap shot picture and we would like to have the movie and we don’t have a movie -  at least in humans. On a tissue basis we do not have this opportunity. I think there are two obvious ways to help out somewhat. First, to use sequential studies in animal models and compare the data generated in animal models with humans. There are a growing number of genome-wide expression profiling data available on animal systems. Second, it would be great if some of these molecules would be amendable to non-invasive screening. Proteomic tools are clearly leading the way in that corner. Studying urine with proteomic tools will be one way to see how we can define the continuity and the gaps and the chums in a disease process and maybe combining transcriptional analysis kind of upstairs inside the kidney with proteomic analysis in the urine might be one way to address that issue.

Question: Great lecture thank you. A question, have you ever compared the gene expression profile in renal tissue with those that you can obtain in peripheral leukocytes from the same patients? Because there are some studies in other diseases and in other system organs indicating that leucocytes are basically an expression of mRNA can relatively decently reflect what’s going on in other system organs? I don’t know whether that has been ever done for the kidney and done specifically for proteinuric kidney disease.

Prof Kretzler: Actually it has been published by several groups. There is a beautiful study relating expression in leucocytes and kidney biopsies in IgA nephritis from Ron Falks group and there are several studies out in the transplant community where leucocytes and gene expression-based profiles from tissues are available. We have just uploaded one of these studies in nephromine so that it can be shared, searched and compared between those compartments. It’s not trivial because obviously the blood leucocytes will be influenced by many things and there’s a lot of noise in these analyses to begin with so a lot of these genome-wide expression profiling studies in leucocytes and have drowned in the background noise picked up by everything influencing it. So there again I think the point you are making relating leucocytes with specific intrarenal signatures might help to fill out these large data sets containing the most informative diagnostic parameters and we are currently pursuing a study, most particularly in lupus nephritis.

Chairman: Thank you very much Matthias for a beautiful lecture.