Well, good afternoon. Thank you very much for coming to Grand Rounds. This is the Nuzhet Atuk distinguished lectureship. 

And before I get started, I wanted to say a few words about kidney diseases. This is the opportunity for nephrology for the next five minutes to talk about the importance of kidney disease, the impact. And then what I'll do is I'll talk about Dr. Atok for a few minutes. And then I'll introduce our speaker, Dr. Matthias Kretzler. 

So as you know, the last two weeks, we've had two outstanding speakers. Last week, Dr. Salant, he represents the present. And the future is what Dr. Kretzler represents. 

So one of the things that's important is to realize that kidney disease has a huge burden across the globe. There are 850 million people with kidney diseases. And that's 20 times more than those with cancer or HIV, and twice as many as patients with diabetes. And you can see there's 13 million with acute kidney injury, 7 million that have reached end stage renal disease. And really only half of them have been treated with RRT. 

Now, the awareness of chronic kidney disease is quite poor. If we look at a survey in the country, 90% are unaware that they have kidney disease. So where are we going from here? Well, we really need to improve our awareness of kidney disease so we can treat patients before they have kidney disease. We need to treat them early to prevent dialysis. But in addition to that, we need to really transform kidney disease research, discovery, and innovation. 

And one area is by Dr. Hu, who is in the biomedical engineering department. We can't work in isolation. I think we need to work in collaboration. And what he's done is he's used this technique called photoacoustic microscopy, where you use pulse laser, and you're able to detect images, and determine functional measures of the kidney. 

So for instance, you can see the brain, vasculature of the retina, and you can see the kidney over here, where you can measure oxygen extraction and bioenergetics. Kidney organoids are being grown in suspension. And they're from pluripotent stem cells. And eventually they can be used, hopefully, for transplantation. 

There's kidney on a chip. These are microfluidics systems which allow for the artificial development of systems where you could determine drug toxicity in a high-throughput manner. There is the ability to measure and image the kidney in 3D. And this is the sympathetic nervous system. And above are the various structures, the collecting duct, proximal tubule, podocytes. And this is alpha smooth muscle actin or sclerosis within the kidney. 

And patients with chronic kidney disease have altered microbiome, which then leads to cardiovascular risk. And so these are all important areas of study. Single-cell sequencing-- instead of bulk sequencing, where you have average gene expression, you have individual identity of different subpopulations of signaling pathways. And then as Dr. Kretzler will probably talk about, is the NIH project in terms of their Kidney Precision Medicine Project, which is a 10 to 15-year project to ethically and safely obtain kidney biopsies to determine subgroups and to stratify patients, determine key pathways, individualize treatments, knowledge base, and improved pipeline of therapies. 

So ordinarily, you have a heterogeneous population. And we treat them. We randomize it into different groups. But with precision medicine, we can define phenotypes and develop specific therapies for patients. Artificial intelligence is on the horizon in order to be able to detect kidney failure in advance. It uses integrated patient information, such as physiological parameters, medications, interventions, and electronic health records. They use deep learning, and they could develop real-time AKI risk score. 

So in this example, we have patients who have creatinine measured on day 1, day 2, and day 3. By day 4, the plasma creatinine rises. But by then, you have acute tubular necrosis. But with artificial intelligence, you have a kidney telemetry in which you can have computers that continuously generate the probability of kidney injury well before there is injury. 

So we need to really innovate. And I'm stealing a quote that Todd Ibrahim used, who is the executive vice president of ASN. "Innovation is a pirate ship that sails into a yacht club." So I think this is where nephrology is headed in terms of innovation. 

So let me tell you about the lectureship. This is a Dr. Atuk. He was born in 1921. He went to the University of Istanbul. And he was an internal medicine resident here at the University of Virginia. In 1951, did a fellowship in hematology and cardiovascular medicine. He joined the faculty in 1956 as an instructor. He was a full professor in 1978. And he was the director of employee health in 1962 and 1988. 

He was a pioneer in providing the extracorporeal therapy in the state of Virginia. He initiated the first hemodialysis session at UVA in 1959. And he subsequently became the director of the dialysis unit. He did some important research in terms of research in catecholamine metabolism, phechromocytoma, and he studied Von Hippel-Lindau disease. 

He's internationally recognized for his work in these areas. He had the largest kindred with phechromocytoma. He was a concerned mentor and a brilliant scientist, and this Lectureship was established by the division to honor his outstanding contributions to science and the practice of medicine, and his extraordinary contributions to the UVA School of Medicine. 

So let me introduce Dr. Kretzler. Dr. Kretzler is the Warner-Lambert Parke-Davis professor of Internal Medicine, Nephrology, and Computational Medicine and Bioinformatics at the University of Michigan Health System. He received his medical degree and PhD from the University of Heidelberg. He was a research fellow at the University of Michigan, did his residency at the University of Munich. He then came to the University of Michigan, where he became the professor of medicine, and his current position with his endowed chairs. 

He is the recipient of many honors. He was a Young Investigator Awardee of the American Society of Nephrology. He was inducted into the American Society of Clinical Investigation and the Association of American Physicians. He was a distinguished fellow of the European Renal Association. He's on numerous editorial boards, steering committees, grant review committees, and he's funded from NIH since 2007. 

His major area of interest is an integrated systems biology analysis of renal disease. He serves as a co-leader for the coordinating centers of the NIH-funded Kidney Precision Medicine program, the Nephrotic Syndrome Study Network, CureGN research network, and JDRF Center of Excellence at the University of Michigan. He has well over 300 publications. We were really pleased to have him here. And please give him a welcome. 

[APPLAUSE] 

Thank you, Matthias. 

[INAUDIBLE] 

Mike, thanks a lot for the kind introduction. And it's really a great honor to present the Atuk Lectureship, and to really hopefully get you as excited as the nephrology team here at UVA is about the impact we will have on patients with kidney disease. 

This is an incredibly exciting time. It was at the American Society of Nephrology meeting, which Mitch was a program committee chair, where for the very first time, you had an energy in the room, fueled by many different areas which Mike quickly summarized to you, that kidney is changing, the way we are moving forward with our approaches. 

The reason for that is, as Mike pointed out, that we do have a tremendous problem in front of us. And that problem, as you just have seen, is affecting now 10%, 12% of our population here in North America. It's the same in Europe and in Southeast Asia. And in sub-Saharan Africa, it's about as equivalent there. And yes, diabetic kidney disease as a consequence of the obesity epidemic is the main cause of the disease. But a multitude of other diseases are also causing loss of kidney function. 

And the key part, speaking in the Cancer Center, acutely aware is that the outcome of our patients with kidney disease is worse than many metastatic cancers who are treated in the center, that if you have end stage renal disease on dialysis replacement therapy, 60% of these patients are dead after five years. And this is higher than cancer, cardiovascular disease, heart failure. 

And in parallel to that, the cost-- not only the human costs, but also the cost in health care resources utilized, are staggering. A quarter of the Medicare budget is spent on CKD care. And due to the fact that end stage renal care delivered by the Medicare system is charged back to the federal government. And in absolute numbers, this is 2% of the total federal budget goes to take care of kidney disease patients. 

And as Mike pointed out, we are at the stage now that, through the multitude of effort, including the executive order from the president this year, this status quo has become not acceptable and tolerable. And we are here to change that together due to the fact that we are starting to change the way we define kidney disease. As you have been taught, I have been taught-- and actually that teaching has not changed much in the last 30 years-- we consider kidney disease right now in descriptive disease categorization. 

We are talking about nephrotic syndrome. And that time in itself is a capitulation of the clinician concerning molecular disease definition. Because we don't know them. We have a phenotypic presentation, a gestalt, how a patient with nephrotic syndrome looks like, our pathologists have a similar gestalt of the kidney biopsies. But we have no clue concerning the underlying pathobiology. 

And that led to the frustrating fact, in the kidney clinics, as of recently, we were not able to tell our patients where the disease is coming from, where is it going, and what we can do about it. But it's changing as we speak. 

This year, for the very first time since I became a nephrologist, we had two large phase 3 trials successful in diabetic kidney disease, reducing the risk burden by 30% for our patients. We have now 25 clinical trials activating in glomerular disease, so that suddenly our patients are becoming the rate-limiting step, bringing mechanistic therapies triggered by insight into the disease from cutting-edge genetic and molecular studies. 

What I would like to do over the next 35 minutes, give you an overview, a snippet how nephrology is changing. And together we can make an impact. A key aspect of that approach was that in nephrology we are starting to use prospective cohort studies of our patients which we have followed over time to generate a comprehensive view of what their disease is. 

And yes, phenotype is important. And descriptive histopathology is critical as a starting point. But as you can see, we have key additional aspects, too, of relevance, that we have to go from the phenotypes, to the structure gene and molecular profiling of our patients, towards an understanding of the cost-cutting disease mechanisms. And if we achieve those, we can stratify our patients in sub-population based on the mechanistic disease insight. Then we can actually start to use pathway activity profiling as we have become familiar in oncology in our kidney disease patients, and can stratify our patients in good and bad outcome, and bring targeted therapies to them. 

And I had the pleasure to present, both to the genetic and the renal team over the last couple of days, our strategies how we achieve that in rare diseases, and how we have used, together with a variety of cutting edge bioinformatics collaborators around the world now, these large-scale data sets to define targetable and treatable activities. 

What I would like to do to this audience is to focus my presentation primarily on glomerular disease-- actually more specifically on diabetic kidney disease-- due to the fact of the significant health burden we all experience in our clinical care. 

I will start with a strategy that I will introduce a concept how we can identify prognostic biomarkers in our patients, and rigorously test them, and then bring them forward to help to stratify patients who will do well from those who will need intensive treatment interventions, and also will be most informative participants in clinical trial research. 

In this study we used, initially, renal biopsy findings, generating comprehensive gene expression data sets from these patients, and then mapping them to clinical and histological data with a machine learning approach of [INAUDIBLE] regression to identify the key outcome predictors extracted from the tissue-level organ scan with the transcriptomes measured, and then allow them to identify which non-invasive biofluids are corresponding to the tissue-level information we have available for prognostic biomarker establishment and validation. 

And there has been a study-- this is really a surveillance consisting seen in translational kidney research right now, integrating a multitude of different people from Kerby Shedden, the mathematician behind the machine learning approach, Wenjun Ju, the molecular biologist, and Viji Nair, the bioinformatician behind the transcriptional profiling data, [INAUDIBLE] together around the common problem, how can we extract from these large-scale data sets a key prognostic disease trial? 

And with this approach, we started initially with the European cohort, the European Renal CDNA Bank, which we established now 20 years ago to establish candidate outcome prediction molecule asset, depicted here, with the machine learning approach to identify six molecules in aggregate who were effectively predicting the decline of renal function in the initial training population, and then into two test populations in Europe and North America. 

And the team then asked the question, which of these molecules would be most interesting to study. And here, artificial intelligence has to meet human intelligence of the exploratory scientists to identify some molecules who are most impactful to our study. And in this specific instance, for our biomarker work, and in the series of discussions we had here over the last two days-- this is really a strategy also closely pursued here at UVA-- is that if we can identify a molecule with a cell type specific, then this will give us an opportunity to effectively model an organ mechanistic disease progress which is not impacted by parallel activities in other organ systems which so often are seen in our patients with systemic disease. 

And what you can see here is the single-cell RNA-Seq data sets from our code, where we evaluated the candidate biomarkers and could show that this molecule indeed is specific for the ascending Loop of Henle in the kidney compartment. So we have a specific molecule for tubular integrity identified with this approach. And the in situ hybridization could confirm these findings. Or you can see, in the healthy kidney tissue, the signal presiding in the ascending Loop of Henle, and the distal tubules, whereas that signal is lost in the chronic kidney disease state. 

And then Wenjun Ju [INAUDIBLE] very carefully evaluated which of these molecules can be profiled in the urine, and establishing a urinary assay of EGF normalized for serum creatinine. You can see that, with this assay, we can effectively predict the decline of renal function in our patients as with the tissue-level analysis, arguing that indeed this is a good non-invasive surrogate of the intrarenal state. 

Then we moved forward in asking, can we indeed predict long-term outcome of patients with chronic kidney disease in the three cohorts-- North American CKD cohort, rare disease, glomerular nephritis nephrotic syndrome cohort in North America, and then in the PKU IgA cohort in China. And in all three cohorts, the biomarker was able to add significantly on top of the routine outcome predictors used in clinical care, of age, gender, TFR, and albumin to creatinine ratio. [INAUDIBLE] that adding to the current measures of hemodynamic state of the kidney, of integrity of the glomerular filtration barrier, a molecule which is reflective of the tubular health indeed can allow us to get a more comprehensive picture of the kidney in health and in disease. 

And as always, critical for this biomarker studies is to rigorously replicate these findings as we move them into clinical applications. And we're very, very fortunate to have our European sister network from the IMI Summit Consortium actually identifying a parallel strategy using animal models to [INAUDIBLE] molecule and evaluate it, and show the efficacy of that molecule to identify patients at risk of decline of kidney function in type 2 diabetes. 

And Wenjun Ju has a team now, in over 19 cohorts across the world, in more than 20,000 people. It has shown that this indeed is a marker which captures tubular health irrespective of the initiating disease processes, ranging from the general population, to diabetic population, to vasculitis patients, even to children with Alport syndrome. As an additional marker which hopefully will complement our assessment of kidney function using the creatinine and albuminuria-based measures as we are doing here. 

So the main point of this was to introduce you two concepts. One is that, yes, there is a shared mechanism of progression of kidney disease, which can be captured using unbiased profiling strategy if they are coupled with careful and meticulous biologizations based on existing knowledge, and then rigorously replicated in independent cohorts. Because this modeling large-scale data set on cohort of a few hundred patients, you obviously always have the risk of what we statisticians refer to as overfitting, meaning you're identifying spurious associations. And you can clearly evaluate the next clue, then, by going into replication cohorts. 

But this is one aspect which is critical if we want to take care of our patients because kidney disease, that we identify those of highest risk. So next step, obviously, is if we can use the same approach using our comprehensive data sets which we are generating from our cohorts to actually identify the disease mechanisms present in these patients with progressive kidney disease versus those without, and then filter them for those molecules who can be targeted, and in some instances are already targeted by therapeutic strategies. 

And here I will give you three different approaches. One, an identification of the progression pathways of kidney disease, which clinically is a problem. At least we nephrologists are most frequently exposed to the patients come with established disease, and ask if we can slow down the progression of the disease. 

And here I want to give a quick introduction to the understanding currently of the overall dogma of how diabetic necropsy develops. In diabetes, we have systemic hyperglycemia, here depicted as time 0.0, where the disease was establishing itself. And this, it behave similar to type 1 and type 2. And then, driven by the hyperglycemia, we are observing a hyperfiltration, meaning an initial increase of the GFR as the disease presents. And then as the disease progresses, we are losing GFR exponentially until end stage renal disease is reached. 

And in parallel, we see an increase of albuminuria as a measurement of damage to the glomerular filtration barrier failure in the years after diabetes induction. And you can see down here-- and I'm sorry that this pointer makes you probably as dizzy as me-- that you can see here, around 12 to 15 years usually is the time window where, in patients, diabetic kidney disease is developing. And that's roughly happening in a third of the population. 

So the question is, how can we address that further. The problem is that our research, until recently, really has focused on this stage of the disease, where you can see that here we have GFR loss around 60 to 80 millimeter per minute. This becomes clinically visible in the albuminuria when it reaches a rate that it becomes clinically significant with edema formation in our patients. 

But that's clearly not where we would like to be as the disease progression critically takes place leading up to these events. And how can we actually identify what's happening between disease initiation and fully clinically manifested progressive disease. 

And here we were very fortunate to work with Dr. Rob Nelson from the Intramural NIDDK research program in Phoenix, Arizona, with a unique patient population, the Pima Indian Gila River community, which the NIH team and particularly Rob have studied for over 30 years. This is a Native American population who exhibits significant obesity, diabetes rates, and develop early end organ damage as a consequence of these significant exposures in their 30s and 40s. And these patients show particularly high incidence of kidney disease very early in life. So that compared to the conventional type 2 diabetes population where you have a lot of secondary hypertension and cardiovascular morbidity, the morbidity is driven primarily with kidney alterations. 

And Rob has studied these patients very carefully and meticulously with yearly visits, inpatient to monitor their metabolic and their renal state, and a couple of decades ago tested the hypothesis if blockade of the renin angiotensin system with losartan actually could be preventative strategy to prevent development of early diabetic kidney disease. 

And this study was done on patients who had normal GFR, and had no microalbuminuria, and were pursued for four years. And then patients, at the end of the trial, were offered an exit protocol renal biopsy, which allowed structural assessment of the disease state at a time when we will not clinically obtain tissue. And that really was a critical window into the disease process, and has led to a flurry of additional studies that will give you a few key findings from these efforts over the last 15 years. 

For us-- and certainly for me-- the most striking finding which really has changed the way I teach diabetic kidney disease now is depicted on this picture, that in patients who have diabetes and clinically no indication of disease-- they have normal GFR, their creatinine sometimes is even lower than expected in the hyperfiltration state, and no albuminuria. But if you look here on the morphometric analysis of the kidney, you see these big strands of interstitial fibrosis already being present in patients without clinically obvious disease sign. 

And this has been very carefully and morphometrically analyzed in these protocol biopsies. And in patients without obvious disease signs, we already have a third of the kidney tissue lost, which really tells you how this silent killer is destroying organ function, even without our cool clinical tools being able to detect that. And that really has been the driving force for many in the field now to ask the key question, can we identify what is present here at these early stages in destroying the filter units and the nephrons in these patients early on. 

And here in this specific instance we were fortunate that, early on in the study, we could convince Rob Nelson to actually obtain [?  ?] extra kidney biopsy core for gene expression analysis similar to the one I have shown to you from our biomarker assessments, and then could correlate late these biopsy findings with the measurement of the interstitial fibrosis, which measure GFRs and albuminuria in the gene expression data sets, to then identify what are the molecular [INAUDIBLE] seen in these early structural lesions, and can we link them to the progressive nature of the disease subsequently. 

And I will spare you some of the bioinformatic detail, how in these small sample sizes of 72 patients, we were able to work with genome-scale data sets, these data reduction technologies, like WGCNA analysis. But the bioinformatic team was able to identify, in these patients with early structural lesions, key pathways associated with the fibrosis and subsequent progression of the disease. 

And they captured two key elements of the disease processes. One of them was expected, that yes, we see a lot of metabolic derangement in the tubular compartment, not surprisingly, as tubular epithelial cells get an extra load of glucose transport. And proteinuria, which they are seeing as a consequence of the glomerular filtration barrier failure. 

But certainly, at that time, very surprising to us was a very robust inflammatory signature to the right, to an extent that if you compare that with other glomerular diseases who are driven by a primarily autoimmune disease, that inflammatory signature was equal in strength in the progressive state of the disease. And that really is a key finding in many other organ systems emerging as well now in diabetic and organ damage, that despite the initial metabolic event, the way your organs are losing their function has a significant chronic inflammatory signal associated with it. And I think that's an important feature to consider, because obviously this is an area of intense research and identification of therapeutic modalities could be [INAUDIBLE] our patients. 

And a finding we were asking next is, as we evaluated the early stages of kidney disease in the Pima Indian population, what happens to these pathways as the patients progress to further loss of kidney function, where they are seen in routine clinical care, and undergo, in some instances, clinically-indicated biopsies. 

And here we use our sister cohort in Europe, the European Renal CDNA Bank, we also had obtained renal biopsies from patients with diabetic kidney disease. But these patients had now significantly impaired renal function in the range of CKD stage 3 up to 4. And using these data, we now could ask, can we replicate the finding in early diabetic kidney disease in the Native American population in patients with established kidney disease in a European population, also with very different genetic and environmental exposures. And we indeed could show that from the majority of the transcripts which were detected in the Pima Indian population, 71% actually replicated in a European population. And most convincingly, the directionality of these changes were conserved across the progressive state of the kidney. 

And the question is, how could this be mechanistically explained? And if you look at the molecules who are the key regulatory hub in these regulatory networks, like NF-kappa B, TGF-beta, EMP, retinoids, many of these molecules are either already in clinical trial development, or have been targeted in ongoing therapeutic interventions of progressive kidney disease, and are targeting a mechanism which is active in each individual nephron. And as you lose kidney function, you are losing your individual functional units. But you retain the remaining segment of your nephron population, who, with hypertrophy and hyperfiltration, tries to compensate that, and perpetuates that mechanism, arguing that early therapeutic strategies might be indeed effective throughout the disease process. 

A key finding which we pursue together with Rob Nelson is to ask our gene expression modules, here named in these four different colors, how do they associate with here, this instance, measured GFR, not estimated GFR. And this invasively-measured GFR, as you can see over the first nine years, do not show any association with the transcriptional signatures. But then, over the next 10 years, when renal functional decline becomes visible, this EGFR changes, we start to see very robust associations of the monitored signatures at time of biopsy, arguing that our functional assessment of the disease processes in the kidney tissue are indeed predicting key pathways who drive long-term loss of kidney function, allowing us to use that information for the establishment of diagnostic markers. And you might have seen that epidermal cause factor was indeed one of the molecules depicted by this approach, and can deliver some of these functionalities. 

If you look into the underlying molecular networks here depicted for early diabetic kidney disease and progressive diabetic kidney disease-- this is Pima and European population-- tubule interstitial compartment, and intraglomerular compartment, you can see that one of the key pathways differentially regulated in these diseases was inflammatory signaling pathway very well known to the immunologists in the room, the JAK-STAT pathway, where we concordant stage-specific activation in the glomerular compartment early on from receptor ligand kinase to transcription factor. And then as the disease progresses, also into tubular interstitial compartment. 

And that is a compelling finding, because JAK-STAT pathways have been very meticulously started in the immunology and our gene domains, and have been allowing the bioinformatic team in the group to establish kidney-specific JAK-STAT activation networks using public databases like the [INAUDIBLE] regulatory network out of the Princeton group, the promoter modeling activities developed by our group, how we can evaluate how the STAT transcription factors are changing downstream gene expression signatures. And that was benchmark, again, a plethora of experimental data defining how JAK-STAT activation works in different organ systems. 

With that information, the group was able to identify the downstream gene expression signatures which regulates and reflects the intrarenal activation of that pathway with an individual patient-specific signal. And if you look at that, you can see that in this diagram, each off the single bar represents a patient where the individual intrarenal JAK-STAT activity was assessed using the transcription signatures. You can easily see that across these data spaces, from healthy controls to membranous nephropathies in basement membrane disease, minimal change ITA, hypertensive nephropathy, you can see the diabetic kidney disease is actually amongst the diseases with the highest JAK-STAT activity, on par with FSTS and vasculitis. And only lupus patients have a higher JAK-STAT activation, providing independent evidence of the critical role of this inflammatory pathway in progressive loss of kidney function. 

And armed with this information, we approached Eli Lilly, who had developed a JAK-STAT inhibitor, which was selectively regionally enriched and excreted by [INAUDIBLE] if they would be interested to consider repurposing that molecule form the ongoing studies in rheumatoid arthritis and psoriasis into diabetic kidney disease. And the fact that Lilly is a diabetes company, we were able to motivate them to pursue that activity to actually perform a clinical trial with albumin to creatinine ratio after half a year as the endpoint. And the fascinating aspect of this type of repurposing strategies that, from the first discussion to the completion of a phase II trial, we can pursue that in a matter of a few years and not the usual one or two decades. 

And the trial indeed showed a response which was confirming the critical role of the pathway that compared to the placebo arm in the study, where you can see here, in gray, the steady increase of the albumin-to-creatinine ratio. You can see a dose-dependent reduction with a JAK-STAT inhibitor. And that dose-dependent reduction was actually retained even after discontinuation of the trial track, the rush-hour period, arguing that inflammatory pathways indeed are driving significant albumin-to-creatinine ratios in patients with diabetic kidney disease. 

Most convincing for this mechanistic strategy to pursue kidney diseases was the fact that we predicted target engagement biomarkers based on our [INAUDIBLE], meaning molecules who should be regulated by JAK-STAT inhibition in diabetic kidney disease. And Eli Lilly indeed mentioned some of them in the phase II trial. Here are the data for IP-10 CXCL10. You can see that IP-10, again, in a dose-dependent manner, shows significant reduction down to 30% of baseline level under the JAK-STAT inhibitor, and remains repressed throughout. 

The frame here is very different to albuminuria data, where after three months, your target engagement biomarker is responding in urine already after two weeks, allowing, very early on, to screen and stratify patients into respond and non-respond [INAUDIBLE]. 

So this was an example how a molecular profiling approach in renal biopsy tissue allows and novel target to be identified, and then quickly and effectively targeted in a phase II trial, up to the extent to identifying mechanistic biomarkers. 

The second part of the target identification process, I want to give an example how one can pursue disease initiation molecules with this type of strategy. And here I will now focus on the aspect of the early disease manifestation. It's a stage where we see hyperfiltration of the glomerulus without any significant sign of proteinuria yet. 

And here we took advantage of the biopsy probe cohort with the Pima Indian study in Phoenix, where we had the opportunity to measure yearly GFRs in these patients, very nice disease cohorts, over 20 years of time, to see the GFR changes, and then could relate our protocol renal biopsies, how they actually were in relationship to the hyperfiltration state. So if we have patients which we biopsied before they reached hyper filtration at peak of their GFR, or after they are starting to top their GFR down into progressive kidney disease. 

And here you can see the data out of these yearly measurements, where looking at the time scale from minus 15 years, before biopsies, up to minus 15 years after biopsy, one can see an increase of the GFR, and then a subsequent top. And you need this time dynamic to actually learn what a GFR of 120 means. If this is a GFR who is still at a rate of increasing your glomerular filtration rate, or already at some loss. 

And comparing these data sets between peak GFR and GFR after a biopsy, after peak GFR, we were able to identify distinct gene sets who were associated with a variety of clinical parameters, including glycemic control, albuminuria, and structural features. But what we also could do is we could take these molecules now and then ask which key regulatory pathways are found in these molecules activated very early on during the disease process. 

And again, for the immunologists in the room, T cell signaling and exhaustion of T cells as you are starting in lupus nephritis, for example, are clearly seen in these patients. But for the physiologists also, a lot of the elements which are very well-known to be critical for hyperfiltration, like the renin angiotensin system, as a positive control was fine. But an additional element of endothelial cell activation were seen, including the endothelial signaling pathway. 

So this is therefore a very rich data set, A, confirming existing therapeutic avenues we have in place with the renin angiotensin system, giving a framework how endothelin inhibition might work in diabetic kidney disease, but also pulling in additional novel candidates for early intervention in our patients. 

And with this approach, the next question was, can we identify these cellular compartments responsible for these activation signatures. And as Mike mentioned, we are very fortunate in the framework of the Kidney Precision Medicine Project to really have established, now, technologies that we can take our studies, which up to now have used microdissected kidney biopsy compartments, glomerular, and tubular compartments, into the single-cell level. And that's an effort led by three faculties in the team, [INAUDIBLE], and Celine Berthier, who have established the technology to use a single kidney biopsy to generate a comprehensive gene expression signatures of individual cells. 

And for example, here, you can see out of 23,000 individual kidney cells extracted from tubule nephrectomy on affected parts of the living donor biopsies, that we are able to identify what an individual cell is expressing in the range of 1,000 transcripts per cell, and then are able to group these cells together according to the similarity of their expression signatures. It's a truly agnostic approach. We only know which genes a cell produces. And then you sort the cells together according to what transcripts we have detected in them. 

When you get unique populations of cells, these clusters, blobs, you see the onset [INAUDIBLE] depicted. But the power of the technology is that then we can look what these cells are in these different clusters. And we can search for molecules which we a priori know are cell-type-specific, and therefore add a tentative label to these cells, saying that, for example, the cell population down here expresses transcripts which we know are exclusively found in proximal tubule epithelial cells that we can assign that over all cell cluster to be derived from proximal tubule epithelial cells. 

The exciting part is, as you can see, there are distinct subpopulation visible which we can again [INAUDIBLE] with the underlying signatures who are different between these subpopulation to see where they relate to into the functional space. 

And we have done that in the interim also now, in patients with diabetic kidney disease, from the Pima Indian protocol biopsies over the last five years. We [INAUDIBLE] a small segment of the biopsy in a preservative which we were hoping to work for single-cell analysis, and indeed it did. And here you can see an analysis where we used 90,000 cells from 45 patients, and can show that indeed we are detecting a distinct cellular population along the nephron segment. And very importantly, also, robust immune cell population, including some of the cells like Tregs of specific interest to the research teams. 

We now are able to ask, if we take our hyperfiltration gene expression signatures which we have extracted from patients where we had 15 years of follow-up after biopsy, using technologies from the early 2000s, can we take that signature and overlay it with the signatures we have on the single-cell level to try to define which cell types are driving these hyperfiltration gene responses, and should be in the crosshairs of our intervention strategy. 

And indeed you can see here, in dark, are the cells who are expressing the hyperfiltration genes, that these are the endothelial cells, the vascular smooth muscle genes, and the immune cell population, arguing that vascular immune cell interactions are critical driver of this early disease stages, harnessing the approaches toward these compartments to be pursued in therapeutic interventions. 

And here we are we are very fortunate that, based on independent research over the last 15 years, where we participated early on as well with our targeted studies, actually the endothelin pathway was already identified as a key driver of early vascular stress in hypertension and in kidney disease. And the endothelin pathway is a key regulator of many compartments of renal dysfunction, not only in the vasculature, but also in the immune cell compartments, and then subsequently for the epithelial cellular compartment in the kidney. 

And depending if you read The Lancet, you might have noticed that actually there was a phase III trial reported this year where the SONAR trial with AtRasentan showed a significant effect of endothelin receptor inhibition towards progression to end stage renal disease, or 40% loss of GFR. 

And this was one of the exciting events in the nephrology of 2019, that this strategy where careful mechanistic discovery [INAUDIBLE] in basic science, in kidney disease, led to the development of a therapeutic agent. 

The challenge is, if you look here at the survival curves, that we have a segment of our patients responding to the treatment. But we have a large number of patients still not responding to the intervention. And that results in an intention-to-treat ratio which is high, and poses, then, a challenge to bring these approaches towards care delivery in the T3, T4 framework. 

So a key challenge we are pursuing together with a European network, the PDKD activities, in a public-private partnership supported by the EU, is the question, can we identify what are the key drivers of the endothelin-signaling response in patients with diabetic kidney disease to identify predictive biomarkers which can be used to stratify patients to be targeted by that therapy. 

And with this network, we used a multi-scalar approach, where we evaluated the effect of AtRasentan in the relevant human cell population. And based on our single-cell data and published data, we mesangial cells to be the target for the intervention, and then perform genome-wide expression profiling to identify AtRasentan-responsive genes in mesangial cells. 

With Charlie Alpers from Seattle, we performed a study to block endothelin receptors with AtRasentan in probably the best progressive mouse models of diabetic kidneys disease in BTBR, Ob/Ob mouse. And then we use the phase II trial, who provided the rationale for the phase III endothelin receptor trials, the RADAR trial, to evaluate bio samples available from these patients, with urine, serum, and plasma, for metabolites, targeted proteomes and microRNA assays to see what are the responses in humans with type 2 diabetes are to AtRasentan and the associated reduction in proteinuria. 

So as you have seen with the single-cell data, indeed mesangial cells are the target of endothelin signaling, and expressing the respective receptors. The BTBP Ob/Ob mouse model of progressive kidney disease is a robust mouse model with significant glomerular sclerosis [INAUDIBLE] releasing pattern, and actually some interstitial damage already. And it therefore was selected as a model system to ask, can we see the effect of the treatment. And here, for example, you see two principle patterns that some molecules are increased in the Ob/Ob mouse model, and then repressed by AtRasentan. Our molecule is lost in the mouse model, and then recovered by the intervention. And if you do that genome-scale-wide, you can select out of these two trends the response parameters which you would like to bring back to the humans. 

And that's what we did together in the network across the Atlantic, that we asked, what does the Ob/Ob mouse model do? And you can see here the endothelin pathway, with activation in red, of the molecules of the endothelin pathway throughout the regulatory cascade. And if you overlap the mouse model with a human disease, you can see that indeed the same molecules are regulated between mouse and man. So the mouse is a good model for this pathway at least. 

And then the exciting part is that, in the mouse, indeed AtRasentan reversed the key elements who are activated in mouse and men, arguing that even the intervention arm of the study is meaningful, and allows us to now use the molecules who are responding to AtRasentan in the mouse tissue as a starting point to filter out human data sets. 

And that was the second part of the study, where we now go into metabolite and proteomic data sets to ask which of these molecules are responsive to AtRasentan in the human, filtering them against the mechanistic biomarkers we have established in tissue and mouse models. And with this pathway, we identified in the targeted protomic analysis in urine, 160 pathways to be baseline predictive, and 200 of dynamic biomarkers, and subsequently filter them down to a set of molecules depicted here, from the mechanistic animal studies and tissue culture studies, to predictive biomarkers from the human biofluids and the dynamic target engagement biomarkers. 

And you can see, similar to what I have shown you with the single-cell data, that we identify, up here, pathways who are imminently targetable. And NRF2 mediated oxidative stress pathway, for example, is a pathway currently targeted by the Bardoxolone trials. And we see several inflammatory mediators, and down here, endothelial cell activation mediators as well. 

And with this approach, we then used the pathway selection strategy to limit our analysis of filtering these data down to the key outcome predictors of molecules who were supported by the multiple evidence of research in this context. And using machine learner, we pruned the 57 molecules down to a small set of eight molecules. Here was able to effectively predict response to AtRasentan in 80% of the patients. And it's a list here contains several known inflammatory mediator and stress factors and outcome predictors of chronic kidney disease, but now anchored back against the AtRasentan response signals. 

And these data are currently validated in the urine sample available from the phase III trial from SONAR, in the cohort of 180 patients, to confirm these findings. And then will be utilized, actually, above and beyond the AtRasentan study, as we have several other clinical trials, particularly the Sparsentan molecule currently being tested in other proteinuric diseases like FSGS, minimal change disease, and IgA, to see if this enrichment strategy can be utilized beyond diabetic kidney disease. 

[INAUDIBLE] I would like to wrap up with a final outlook how we actually envision to bring these precision medicine concepts forward to our patients in ongoing clinical trials. Where as I mentioned earlier, we are in a situation that we changed from a clinical trial or two in a five-year funding cycle of research activities to now nearly 20 trials being active in this year or next year. And so a key feature is how can we bring the right patients to the right trials at the right time. 

And here, we are working together with a patient interest group, NephCure Kidney International, and an NIH-funded research network, to establish that principle in the nephrotic syndrome framework, and with the hope that the same philosophy can be translated to other common diseases, including diabetic kidney disease. 

NEPTUNE captures patients with a very comprehensive information data set at time of presentation, again using kidney biopsies as one of the key entry points, and we capture comprehensive clinical phenotypes, and then molecular profiles. And we have done that over the last 10 years to define disease subtypes along the molecular, clinical, and histological phenotypes which we discussed yesterday, adrenal gland rounds at length, and now are transitioning to use that information to initiate therapeutic trials in functional defined patient cohorts. 

And the strategy in that context is that we take our patients with tissue, biofluids, clinical information, and histological assessments. We bring these data sets together in a molecular nephrology board, who over the last two years has learned how to use that information to identify in each patient along molecular pathways who are targeted by ongoing or future clinical trials, the same pathway activation scores I have shown you for JAK-STAT or for the endothelin pathway. 

And with this instrument in place, we are currently initiating a clinical trial framework that we will profile our patients, we will communicate to our patients those who are at high or low risk of disease progression, and then among those at risk of disease progression, where their kidney tissue, blood, or urine indicates activity states who are available for therapeutic intervention into clinical trials. 

It's a hope that we transition from the challenge you saw with the AtRasentan trial, that only a small proportion of the patients respond to treatment, requiring the several-thousand sample sizes needed right now in phase III trials, to a targeted approach with a majority of the patients have your pathway of interest active, and therefore will allow you to cut down the sample size in your clinical trials significantly. Currently the power calculation can argue, if you reach enrichment of 80%, we can do these trials in up to 68 patients, which clearly changes the field significantly. 

And in parallel, at least for glomerular diseases, we have, together with the FDA, developed a strategy that proteinuria can be, at least often, indication and initial registration endpoint, really turbo-charging nephrology towards a field where very effective clinical trials can test innovative molecular compounds in a functional context. 

So that's where I would like to wrap up, to-- hopefully I've given you an example of how precision medicine can actually be implemented outside of oncology, and that kidney nephrology is an ideal place to do that, as we have multiple avenues towards assessing disease activities in the end organ manifesting the disease, ranging from biopsies of the kidney down to the liquid biopsy of the urine, leading to [INAUDIBLE] disease classification, integrating that information in a molecular nephrology board to bring the right patients to the right trial at the right time. And this is happening. This is happening here, and this is happening across the globe as these research efforts are integrated with efforts in Europe, Southeast Asia, and now also sub-Saharan Africa. 

With this, I would like to wrap up, to thank the team in the networks and in Michigan who make these activities possible through funding support from NIH, EU, foundations, and private entities, my team, most of all our patients and their families, due to their dedication-- in some instances over a couple of decades-- who helped us to hopefully give some venues to make their kidney filters a little bit more happy than they sometimes are. Thanks a lot for your attention. 

[APPLAUSE] 

Any questions? 

I'm sorry. 

No, you're fine. Just, yeah, if you have any questions, we have a couple more minutes. Just raise your hand for the microphone. 

Does the glomerulus start in [INAUDIBLE] diabetic [INAUDIBLE]? Is it though to be, like, passive accumulation of glycosolated proteins? Or is it though to be an active process whereby the epithelial cells [INAUDIBLE] muscle cells are secreting the scarring material? 

Yeah, it is an active and highly-regulated process. And you saw some of the molecular pathways who came out of the earlier study where the hyperfiltration is providing the first stress stimulus on the glomerulus reflective of that. 

And there's also nice genetic evidence in support of that, which has published the largest [INAUDIBLE] study in type 1 diabetes. And the key hit out of that study identifying risk loci for diabetic kidney disease was actually in collagen IV alpha 3, which is a specific molecule of the glomerular basement membrane critical for cell membrane interactions. And the variance was actually protective who avoided the widening and the activation of the GBM in diabetic kidney disease, really stimulating a lot of discussions now how matrix 3 modulation can be targeted. 

And this is where, like I showed yesterday, actually, some of our key progression predictors are also involved with tissue matrix metalloproteinases. [INAUDIBLE] that's a very actively highly-regulated process. The challenge will be to find innovative way how to target these mechanisms effectively. A TGF-beta inhibitor trial by Eli Lilly did not show the effect that-- the expected effect. But TGF-beta is a complex molecule. And using a sledgehammer approach might not be the definite answer in that context. 

Thank you for a wonderful lecture-- couple lectures. If, in the Pima Indians, the use of losartan and the reduction in proteinuria was insufficient to avoid the scarring, how do you feel about proteinuria as a progressor of disease? 

This is very interesting. The Pima trial was negative for the pre-specified endpoint, meaning measured GFR after five years. What was interesting is the secondary endpoint changes in morphometric alterations was actually significant, but wasn't sufficient for the FDA at that time point. After 12 years [INAUDIBLE] of follow-up with measured GFRs, now the GFRs are different, and really arguing that these studies are so critical that you have an opportunity to link early findings with long-term outcomes so that you know which endpoints you can model in this context. 

And so BDKD's a European public-private partnership network. The main purpose of that network is to identify these outcome predictors for clinical trial endpoints. And the challenge with albuminuria is that it is not stringently associated. We do see patients in contrast to the dogma. I showed with initial loss of GFR and an increase of albuminuria. There are patient with progressive diabetic kidney disease without albuminuria. We see, probably, a dominant tubule interstitial pathophysiology driving for or arguing that inflammatory mechanisms, particularly in the interstitial, are fully capable of advanced GFR reductions without primarily glomerular filtration variant damage. 

Great talk. You have a wonderful slide, the image here. Do you see the value in medical imaging-- Of course X-ray contrast and MRI contrast lately is a problem for kidney disease-- but ultrasound contrast, do see a value in that, with perfusion imaging, and also targeted for endothelial biomarkers of inflammation? 

Yeah. There's a lot of effort underway, including, as you saw, here, at UVA. In the Pima Indian cohort, we have functional MRI done at time of biopsy so that we can correlate a functional MRI with the histology and the molecular signatures present at the same time. And the BDKD network in Europe is doing a comprehensive study across six sites, where they use a wide variety of different imaging modalities and protocol renal biopsies to see how the imaging modalities relate to the kidney function at time of biopsy. And then with progressive future decline of kidney function, which will need to be ascertained. 

I think I have time for one more question. Thank you, Dr. Kretzler. 

Hi. I really enjoyed your talk. One quick thing-- this new genome and microarray will help us better understand how to manage the inflammatory process. But I was just wondering, we know, for example, in diabetics, about 30% or 40% end up with kidney disease. But there's a large group that does not. And I was wondering if you can use this technology to examine the people who do not so that we can prevent the inflammatory process. 

Yeah. This is absolutely critical, because as always, listen to your patients, and they will tell you a lot of things. And the key part is for these studies to define the protective mechanisms is a proper adjudication of the disease course over time. And as you have seen, many patients show a progressive decline of kidney function only after 15, 20 years. A lot of studies following their principal use, follow-up time periods of two to five years, you might have adjudicating patients into non-progessive states, and they actually just have a delay in their progression. 

So these long-term follow-up data are critical. And the data is actually embedded in what I presented to you. Because we compare not against healthy living donor biopsies, but in progressive versus non-progressive diabetic kidney disease in this Pima population. So the data is already present. We just would need to display it differently than I did today. 

All right, great. Thank you so much. I think we have one more announcement. 

Yes, so on behalf of the Department of Medicine and Division of Nephrology, we would like to thank you for being our Atuk Lecturer this year. And we'd like to present you with a picture of [INAUDIBLE]. 

Aw, that is great. Thank you so much. 

[APPLAUSE] 

Thank you so much.