12th Annual Rachmiel Levine Diabetes and Obesity Symposium

February 15-18, 2012; Pasadena, CA Days #1-2 Full Commentary – Draft

Executive Highlights

Greetings from sunny and warm Pasadena and the 12th Annual Rachmiel Levine Diabetes and Obesity Symposium! While very little new data has been presented over the first two days of the conference, we have thoroughly enjoyed hearing some of the latest thoughts on topics ranging from immunotherapy for type 1 diabetes to the pathogenesis of type 2 diabetes from an all-star cast of diabetes researchers and clinicians. We highlight some of our favorite talks from the conference thus far below.

Of perhaps greatest note, Dr. “P.J.” Utz (Stanford University, Palo Alto, CA) presented topline data from a phase 1/2 for Bayhill’s proinsulin DNA vaccine (BHT-3021) for the treatment of type 1 diabetes. Overall, the therapy was found to be safe and well tolerated. Promisingly, although not powered to detect differences in clinical efficacy, the 1 mg dose was also found to statistically significantly preserve C-peptide over placebo at 15 weeks (p=0.026). Dr. Utz indicated that Bayhill is currently planning a phase 2 trial that will examine this dose in a recent-onset type 1 diabetes population. We look forward to hearing more data from this phase 1/2 trial in the coming months.

Dr. Jay Skyler (University of Miami Miller School of Medicine, Miami, FL) delivered an engaging presentation on the need for combination therapy to halt or reverse the progression of type 1 diabetes. While noting that clinical trials examining combination therapies have largely failed to show significant efficacy, he highlighted a number of preclinical studies that have demonstrated initial promise with several different combinations. He closed by presenting what he envisioned would be an effective treatment regimen for the disease.

In a lab to clinic bridging talk, Dr. Kevan Herold (Yale University, New Haven, CT) described the ways in which mice both have been and not been useful for translating of immunotherapies to the clinic. Interestingly, he detailed a new assay that his lab had developed with the help of mouse models that measure beta cell death through quantification of demethylated insulin DNA. His lab has already used the assay in humans. We’ll be excited to see further results using this assay, as we think it could mark a major advance in quantifying beta cell viability, which thus far has only been measured indirectly in humans.

After lunch, Dr. Skyler again took the stage with Dr. Carla Greenbaum (Benaroya Research Institute, Seattle, WA) to debate on whether insufficient biomarkers and inadequate endpoints had caused the downfall of recent phase 3 immunotherapy trials. Dr. Skyler argued that this had indeed been the case and cited decisions and study design shortcomings related to GAD65, teplizumab, and otelixizumab to defend his opinion. Meanwhile, Dr. Greenbaum emphasized that rather than inadequate biomarkers or endpoints, an insufficient understanding of type 1 diabetes pathogenesis and evidentiary conservatism may have doomed these trials. We thought both sides made good points, and we came away with an understanding that much work remains before an immunotherapy for type 1 diabetes might successfully be developed.




Detailed Discussion and Commentary

Controversies in Diabetes Immunology (1)


Pere Santamaria, MD, PhD (University of Calgary, Calgary, Alberta)

In a talk that generated significant audience interest, Dr. Santamaria described a novel method of blunting the diabetes autoimmune response. As background, he detailed how in diabetes, both high avidity and low avidity naïve T cells are activated by antigen presenting cells. The high-avidity naïve T- cells turn into cells that attack islets and subsequently die. On the other hand, the low-avidity naïve T- cells expand into autoregulatory memory T cells, which reengage dendritic cells, suppress these cells’ antigen presentation activities, and thereby blunt the autoimmune response. Excitingly, Dr. Santamaria’s lab has created iron oxide nanoparticles coated with peptide-MHC complexes that expand the autoregulatory memory T cells generated from low avidity T cells. When susceptible animals are treated with these nanoparticles, they don’t develop diabetes. Additionally, when these nanoparticles are coated with disease relevant complexes and injected into newly diabetic mice, they restore normoglycemia. However, because these nanoparticles’ mechanism of action depends on the presence of autoreactive T cells generated by existing autoimmune disease processes, they shouldn’t have an effect in healthy individuals. We’ll be interested to hear further results from animal studies with these nanoparticles and will specifically look forward to learning how efficacious these nanoparticles are at expanding regulatory T-cell pools compared to similarly directed methods.


Q: How late in the disease process can you go in with nanoparticles that have MHC class 4 for the induction of CD4+ T cells? Have you tested whether this approach lends itself to any combination?

A: We have done most of our work in newly diabetic mice, and we haven’t gone to animals that have worse glucose control. Destruction of beta cells in mice is very fast, and when a mouse has been diabetic for a week, they have almost no islets. With a therapy like this, it will be difficult to suppress the autoimmune response enough to make beta cells not go at a late stage. We are however doing imaging studies to find out where the insulin that is made in response to the therapy goes. In regards to the second question, about whether there is a possibility for using with other strategies. I think that any strategy that shows some effect in blunting the autoimmune response, even as part of a combination of other things is good. But I don’t know exactly.

David Harlan, MD (University of Massachusetts, Worcester, MA): Have you adoptively transferred the cells you induced to newly diagnosed diabetic mice? How many do you need to protect newly diagnosed mice?

A: In T cells, you detect 2-4% of cells in the spleen. You have to transfer them. You have to sort them from the mice. You end up with very low numbers. It’s an experiment that’s not practical to do. You have to adoptively transfer to multiple mice. We took an anti-CD4 T cell population and then mixed them with autoreactive T cells and then transferred them to the host to prevent diabetes.

Dr. Harlan: What was the stoichiometry you used?

A: We injected half a million regulators and half a million effectors.

Q: Are there unique surface markers that you can use to identify memory autoregulatory t cells in the mouse?

A: We can use pMHC tetramers.



Paul “P.J.” Utz, MD (Stanford University, Palo Alto, CA)

Dr. Utz presented topline results for Bayhill Therapeutics’ proinsulin DNA vaccine (BHT-3021) for the treatment of type 1 diabetes. In the trial, 80 adults with type 1 diabetes were randomized to receive once weekly intramuscular injections of BHT-3021 (0.3 mg, 1 mg, 3 mg, or 6 mg) or placebo for 12 weeks. Participants were followed for up to a total of 36 weeks. Overall, the vaccine was found to be safe and tolerable. There was no evidence of an increased immune response toward insulin as a result of treatment. Although not powered to detect differences in efficacy, a U-shaped dose response was observed with BHT-3021 with regards to C-peptide preservation. The 1 mg dose was the most efficacious dose – at both five week and 15 weeks (three weeks following the administration of the last dose), C-peptide levels were elevated above baseline. In contrast, C-peptide levels declined continuously from baseline in the control arm. The difference in C-peptide levels between the 1 mg arm and control arm at 15 weeks was statistically significant (p=0.026). However, C-peptide continued to fall in the 1 mg arm following this time point. According to Dr. Utz, these results suggest that BHT-3021 may act to preserve beta cells in the type 1 diabetes setting; however, the vaccine appears to have the greatest effect during or immediately following the dosing period. Thus, continuous dosing may be required for long- term efficacy. Dr. Utz indicated that Bayhill was currently planning a phase 2 study that will examine the safety and efficacy of the 1 mg dose of BHT-3021 in pediatric or adolescent patients with type 1 diabetes (recently diagnosed individuals). The expected enrollment for the study is 12o. A timeline for initiation was not provided. As a reminder, Roche/Genentech returned the rights to BHT-3021 to Bayhill in 4Q10 as a consequence of Genentech’s Research and Early Stage Development portfolio prioritization process (see the February 2, 2011 Closer Look at http://bit.ly/wPKFLX).

  • BHT-3021 is an injectable formulation of plasmid DNA containing the gene encoding the proinsulin protein. Once injected, the plasmid is taken up, proinsulin is expressed, and the protein is believed to be transported to lymph nodes and presented to T cells by antigen presenting cells. This presentation takes place without the presence of a co-stimulatory factor that is needed to trigger an immune response. Overall, this process is thought to be capable of inducing insulin- and proinsulin-specific immune tolerance, thereby halting or preventing the immune system from attacking its own beta cells.
  • A phase 1/2 trial examined the safety and efficacy of BHT-3021 in people with type 1 diabetes (clinicaltrials.gov identifier: NCT00453375). Participants (n=80) were randomized to receive 0.3 mg (n=14), 1 mg (n=15), 3 mg (n=14), or 6 mg (n=6) of BHT-3021 or placebo (n=23). Treatments were injected once weekly for 12 weeks intramuscularly. The primary analysis was conducted at 15 weeks, and patients were followed for up to 36 months. In the placebo group, 13 participants were crossed over to receive BHT-3021 after 12 months. Enrollment criteria included: type 1 diabetes diagnosis within five years, 18-40 years of age, the presence of antibodies to at least one of insulin, GAD 65, or IA-2, and a stimulated C-peptide response of at least 0.2 pmol/ml. At baseline, Dr. Utz indicated that the groups were mostly balanced except for serum C-peptide levels (338 pg/ml in the 1 mg BHT-3021 group vs 503 pg/ml in the control group). He also noted that Bayhill originally sought to conduct the trial in individuals with recently diagnosed diabetes (due to efficacy considerations), but the FDA requested that the trial be conducted in an older population for safety reasons.
  • BHT-3021 was found to be safe and tolerable. No safety signals were identified, and there was no dose limiting toxicity. No increases in insulin antibodies were reported. Furthermore, initial ELISpot assay analyses suggested that the vaccine altered immune responses toward the insulin antigen but not other antigens unrelated to the pathogenesis of type 1 diabetes.
  • BHT-3021 may preserve beta cell function in people with type 1 diabetes. In the placebo arm, C-peptide levels declined continuously throughout the trial. With BHT-3021, there was a U-shaped dose response with the strongest efficacy observed with the 1.o mg dose. In the1.0 mg arm, C-peptide levels were observed to be higher than baseline levels at both week five andweek 15 (three weeks after dosing in the trial was completed). The change at week 15 was statistically significant vs placebo (p=0.026). C-peptide levels declined following this time point, suggesting that the drug’s effect is greatest during or immediately after the dosing period and that repeat dosing is likely necessary for long-term efficacy.


Harald von Boehmer, MD, PhD (Dana Farber Cancer Institute, Boston, MA)

Dr. von Boehmer described advances his lab had made in the expansion of regulatory T cells. After describing a protocol that generates regulatory T cells with stable FoxP3 expression if high affinity ligands are used, he presented data suggesting that TGF is important in the creation of regulatory T cells in the gut but not the thymus. He then discussed two reasons why T regulatory cell vaccination may have thus far failed to prevent diabetes in preclinical studies. Firstly, he said, T regulatory cell conversion may be inefficient in the NOD mouse. Secondly, the insulin epitope traditionally used in diabetes immunological studies may not be well suited for insulin vaccination. According to Dr. von Boehmer, mTOR inhibitors, rapamycin, and everolimus may help address the first issue. In regards to the second issue, Dr. von Boehmer’s lab has designed an insulin mimetope that binds the MHC complex more effectively than the insulin B:9-23 epitope traditionally used in induction and much more effectively converts naïve T cells to regulatory T cells than insulin B:9-23. When given alone to four- week old NOD mice, this mimetope prevented diabetes; treated mice did not develop diabetes for the rest of their lives. Notably, the mimetope works best in mice with only moderate levels of insulin autoantibodies. It does not prevent diabetes in mice with high autoantibody levels. Dr. von Boehmer’s group is now working to move evaluation of the new mimetope into human studies.

Q: In terms of when you translate this approach to the clinic, what are your thoughts with respect to biomarkers? Are you going to immunize with the peptide in adverse populations? How would you monitor this on an individual basis?

A: In mouse, it was very helpful when we got 2+2 tetramers. We would like to have that in mouse too. We have a good hint from Kappeler how to do this. The plan is to make humanized mice with DQ8 and DQ2 and to repeat the experiments we have done in these mice, because right now what we are doing can’t be very well done in humans. We can then test our markers and reagents in the humanized mice, and then hopefully proceed to human studies.



Matthias von Herrath, MD (La Jolla Institute for Allergy and Immunology, La Jolla, CA), Pere Santamaria, MD, PhD (University of Calgary, Calgary, Alberta, Canada), Paul “P.J.” Utz, MD (Stanford University, Palo Alto, CA), and Harald von Boehmer, MD, PhD (Dana- Farber Cancer Institute, Boston, MA).

Q: Do you have any data in your respective mouse models about what happens to CD8 responses in these mice when you induce tolerance either with a changed peptide or a nanoparticles coated with MHC class II a?

Dr. Santamaria: For us, what we measure is if an individual mouse responds to treatment, and whether they have expansion of cognate T cells in the peripheral blood. This correlates well with reversal of disease and with how long the mouse will remain normoglycemic. We also measure recruitment of autoreactive cells into pancreatic islets and creation of tetramers when islets associate with T-cells. We also do cytokine production assays. We basically see a significant inhibition with treatment of noncognate T cells. We do a number of other experiments in a more reductionist way to show that suppression is antigen specific but is not autoantigen specific at the effector level.

Dr. von Herrath: How is the CD8 response in the BHT-3021 trial correlating with the preservation of C-peptide? Can CD8 cells be used to predict responders?

Dr. Utz: In terms of baseline predictors, we cannot really tell. We aren’t able to tell who is going to respond or not respond yet. The vaccine is hitting the CD8 arm though.

Dr. Harlan: For Dr. Utz, you are looking at autoantibodies sometimes after disease onset, and obviously people with diabetes are not being treated with insulin at this time. Is there any way to go back and identify people that had been autoantibody positive to see if the therapy was effective in this subgroup?

Dr. Utz: The vast majority of patients had not had autoantibody testing done before the trial. I am not able to answer you.

Q: It seems to me that you were targeting primarily participants with late-onset autoimmune diabetes. Was there any association you found between starting C-peptide level and end C-peptide level and age?

Dr. Utz: We have looked at that and the numbers are just too small to look at correlations.

Q: When you showed the patient characteristics, the data was small and it was hard to read the print. The group receiving 0.3 mg had disease of shorter duration than the other subgroups. Did you look at the association between duration of disease and effect of treatment in that group?

Dr. Utz: Yet but the numbers were very small. We did get surprising results, namely that you could see C- peptide at all. However, we designed the study just to assess safety and we expanded the cohorts to add one new dosing group (6 mg) and then later 0.3 mg, further diluting the numbers. The second issue is that it’s a huge challenge to do biomarker studies in these sorts of studies because we have to freeze down the blood, and it all has to be thawed down at one time a year later to show that we see differences.


Controversies in Diabetes Immunology (2)


Jay Skyler, MD (University of Miami Miller School of Medicine, Miami, FL)

Dr. Skyler opened his presentation by noting that there have been several “successful” interventions in the setting of recent onset type 1 diabetes, including with cyclosporine, teplizumab, otelixizumab, rituximab, and abatacept. In clinical trials, each of these therapies has been associated with an initial preservation of or improvement in C-peptide. However, this beneficial effect consistently gives way to progressive C-peptide decline that often mirrors the rate of decline observed in the control group, even with repeat dosing. Dr. Skyler argued that combination therapy might provide a more effective approach for halting or reversing type 1 diabetes progression. To date, however, clinical trials examining combination therapies (MMF plus daclizumab, exenatide plus daclizumab) have largely failed to demonstrate improved efficacy (if any effect at all). The one exception has been a non- controlled trial (n=20) in Brazil that found combination treatment with cyclophosphamide, ATG, and an autologous human bone marrow transplant (AHBMT) provided a 70% likelihood for insulin independence after 40 months. Despite this limited success, Dr. Skyler highlighted several preclinical studies that have provided evidence that combination therapy may ultimately prove effective. These combinations include: 1) anti-CD3 plus exendin-4; 2) GLP-1 plus gastrin; 3) DPP-4 plus a proton pump inhibitor; 4) IL-2 plus rapamycin; 5) anti-CD3 plus nasal insulin; and 5) anti-CD3 plus anti-IL-1. Looking forward, Dr. Skyler proposed that an effective treatment regimen for type 1 diabetes could include: the chronic use of an anti-inflammatory agent (anti-IL1 beta or anti-TNF), the acute use of an immune suppressant (anti-CD3, anti-CD30, or a co-stimulation blocker), the chronic use of an antigen- based therapy (GAD 65, oral insulin, proinsulin peptide), the acute use of a regulatory T cell inducer, and the chronic use of a beta cell survival or regenerative agent (GLP-1, INGAP).

Questions and Answers:

Alexander Fleming, MD (Kinexum LLC, Harper’s Ferry, West Virginia): First of all, let’s keep in mind the complexity of the disease and the need to make the right choices. Ultimately, we need to look at the benefit/risk profile. What concerned me about the South American study was whether it was an ethically permissible study. It was a pretty serious intervention with a substantial risk. I don't know if that can be tested further. You said that the effect could have mostly been from the chemotherapeutic agents. That may be possible. I don’t know if bone marrow transplantations should be considered in the context of type 1 diabetes. We all agree combination approaches are needed. The big problem is going from NOD mice to human studies.



Kevan Herold, MD (Yale University, New Haven, CT)

In this mostly basic science talk that was nevertheless quite clinically relevant, Dr. Herold discussed the ways in which murine type 1 diabetes studies are and are not useful in terms of clinical translation. In some cases, like that of anti-CD20, he said, murine studies have provided good insight for clinical trials, but in others (CTLA4Ig and anti-CD3) they have not properly predicted translatability or therapies’ immune effects. According to Dr. Herold, poor translation of findings in mice to human studies has likely been due to murine immune responses differing from human ones, animal models not accurately recapitulating human type 1 diabetes, the kinetics of diabetes in mice not being the same as those in humans, and mouse beta cells not being biologically the same as human ones. However, mice with humanized immune systems like that which Dr. Herold’s lab recently treated with the anti-CD3 antibody teplizumab and studied (Waldron-Lynch et al., Sci Transl Med 25, 2012) may help bridge some of these murine-human differences and have already provided insight into the effect of anti-CD3 therapy on T cell migration. Murine models have also aided the development of new biomarkers that enhance our as-of-now inadequate understanding of how beta cells die. Moreover, according to Dr. Herold, they will be important for picking out immune mechanisms to target in human type 1 diabetes combination therapy trials. Overall, Dr. Herold said, preclinical murine models are useful for establishing trial rationale and design, elucidating mechanisms, and studying reagents in tissues that can’t be studied in patients.

  • While preclinical mouse studies have provided good rationale for many clinical trials, they have not always been successful at accurately predicting translatability or therapeutic mechanisms. For example, Serreze et al (JEM, 1996) and Wong (Diabetes, 2004) first demonstrated that B cells are important antigen presenting cells in diabetes and Hu et al. (J Immunol, 2007) showed that diabetes could be prevented and treated in NOD mice with the anti-CD20 monoclonal antibody. Building on this preclinical work, Pescovitz et al. tested use of the anti-CD20 antibody rituximab in type 1 diabetes and showed that rituximab treated patients retained significantly higher C-peptide secretion after one year than the placebo group. However, preclinical studies haven’t been as informative with regards to CTLA4Ig or anti-CD3. While treatment of NOD mice with murine CTLA4Ig resulted in exacerbation of type 1 diabetes, Orban et al. showed last year that treatment of patients newly diagnosed with type 1 diabetes with the anti-CTLA4 antibody abatacept modestly improved C-peptide levels. Additionally, although Chatenoud et al. showed that prevention and reversal of diabetes in the NOD mouse after anti- CD3 treatment was associated with increased anti-CD4+CD25+ T cell levels, these cells were not found in the peripheral blood of human subjects treated in clinical trials with the anti-CD3 antibody teplizumab.
  • There are four main reasons why mouse diabetes studies may not translate well to humans. Dr. Herold’s group has begun to address these. According to Dr. Herold, poor translation of findings in mice to human studies may be due to murine immune responses differing from human ones, animal models not accurately recapitulating human type 1 diabetes(because NOD mice are inbred, don’t live in the same environment as humans, and have several immune deficiencies not found in humans), the kinetics of diabetes in mice not being the same as those in humans, and mouse beta cells not being biologically the same as human ones. To address this, Dr. Herold’s lab has created mice with humanized immune systems. Using this model, they have shown that anti-CD3 treatment causes both CD4+ and CD8+ cell levels to go down in the spleen, lung, and bone, but that these cells migrate to the gut and produce IL-10 that is detectable in the peripheral circulation.
  • Mouse models have helped identify biomarkers that enhance our understanding of when and how beta cells die. A major problem with current trials in type 1 diabetes, Dr. Herold emphasized, is that we don’t know how or if beta cells die because we traditionally could not measure beta cell death, but rather only beta cell “mass.” To address this, Dr. Herold’s group has developed a nested PCR assay that measures beta cell death via quantification of demethylated insulin DNA (according to Dr. Herold, only dead beta cells should have demethylated insulin DNA). In mice, this assay has not surprisingly demonstrated that beta cell death increases significantly prior to development of diabetes, even before changes in fasting blood glucose are seen. Dr. Herold and colleagues have also used this assay to show that people with new onset diabetes have much higher levels of demethylated insulin DNA than healthy controls and that there is an inverse relationship between demethylated insulin DNA levels and C- peptide.
  • Mouse studies will be needed to identify the optimal immune therapies to use together in combination trials. Since these trials will be so large, costly, and complicated, Dr. Herold argued, mice will be needed to understand on a mechanistic level which combination of therapies will have synergistic or complementary effects. Notably, results from early mouse studies with combination therapies have been positive and have shown, for example, that joint treatment of mice with anti-CD3 and anti-IL-1 significantly decreases levels of antigen-specific T cells.

Questions and Answers

Q: Have you had a chance to apply the demethylation assessment to samples in the prevention trial?

A: I didn’t show the data because we are in the process of doing that. It turns out to be more complicated than we are showing here. There is some suggestion that there are changes. What we haven’t done is look at later timepoints. We should look at changes at six months and I think that would predict outcome at 12 months. I’m not sure about the kinetics. We are doing this right now.

Dr. Harlan: About the CTLA4Ig example that you gave: the rule has so far been efficacy in NOD mice or BB rat is a go signal. The rule has said that we shouldn’t use this. This makes me wonder what might have been ruled out that could have been effective in people. I wanted to reflect on that philosophical question.

A: I always thought about it the other way. I thought of it as being the gatekeeper. That is if studies made diabetes worse in the NOD mouse then we shouldn't proceed. But your point is well taken. Things that failed in the NOD mouse might be useful in humans. It’s a hard question. I'm sure there are many for example that would raise as a point of discussion later on that lots of beta cell specific agents might have been discarded because the NOD is largely autoimmune driven in the mouse. Those types of agents might potentially have been useful but we ignored them.



Bart Roep, MD, PhD (Leiden University Medical Center, Leiden, The Netherlands), Jay Skyler, MD (University of Miami Miller School of Medicine, Miami, FL), Kevan Herold, MD (Yale University, New Haven, CT), Francesco Dotta, MD (University of Sienna, Sienna, Italy), Decio L. Eizirik, MD, PhD (Free University of Brussels, Brussels, Belgium)

Dr. Skyler: In response to the question Kevan raised about why did abatcept go forward when it was controversial in animals. For a TrialNet for a therapy to be tested it either has to have a beneficial effect in animals or be a therapy that is used in other autoimmune diseases successfully. It [abatacept] has been approved for two other indications previously.

Dr. Fleming: I just want to point out that these are different conditions and that there are different therapeutic options for these conditions. We will agree to disagree about whether it is ok to go with that approach for type 1 diabetes.

Dr. Roep: In NOD mice, there is evidence suggesting that B cells can be pathogenic. Is there any evidence that this is true in humans? I am not denying the suppression of B cells can help delay the progression of diabetes?

Dr. Herold: Are they directly pathogenic I don't know of any evidence suggesting that they are directly pathogenic. But as an antigen presenting cell, they may be very important. One of the lessons we have learned from the rituximab trial and preclinical trials is that the kinetics of the disease, the immunological kinetics, may be different in mice and humans. If B cells acted as antigen presenting cells, by the time you get to diagnosis, you would think that they would no longer have an important role. They would no longer be needed whatsoever. However, based on human data, it seems that at the time of diagnosis, B cells still play some important role in modulating the immune response. I think that the abatacept data helps prove that.

Q: As a clinician, from these very important talks this morning, I have learned that type 1 diabetes can be prevented and cured if you’re a mouse. And I was particularly fascinated with Dr. Dotta’s talk that considered some LADA patients anecdotally. I wonder if things are just going slower in these patients and if some immune processes are the same in LADA patients. Are there are any trials going on to see what is happening in these people without having them donate their pancreas? Anything that is seeing how this goes along across a period of years? What is keeping diabetes from occurring earlier in life?

Dr. Dotta: I don’t know, this is a key question. The answer to this question will shed light on the pathogenesis of type 1 diabetes. It might be a genetic disposition. LADA patients are for sure very interesting. This will be useful to test.

Dr. Skyler: It is very hard to design longer term studies with LADA. The natural history of C-peptide decline is not well defined in LADA. And secondly the rate of decline is so slow that to find a difference between intervention and placebo would require either a very large sample size or a very long duration of study. It is very difficult to map such a study.

Dr. Herold: Another aspect that maybe Carla wants to comment on is the effect of age. I think some of us have thought a long time about whether there is an effect of age on the decline of C-peptide and I think that work Carla put together and my own work on the effect of age on anti-CD3 shows that there is an effect. The effect is on untreated groups. The rate of decline is slower in older people. I am not answer the question with regards to difference but it does seem to be different.

Carla Greenbaum, MD (Benaroya Research Institute, Seattle, WA): We looked at a batch of data from clinical trials – placebo and controlled groups together – to get a better understanding of what happens under clinical trials conditions. In people over the age of 21, the rate of fall was different than people younger than that. What was interesting was that for the cohorts under age 21, you had a similar rate of fall. It wasn’t any different than teenagers. It was also interesting that in the very youngest cohort, they started with less C-peptide. We have little information about normal C-peptide levels in children. Jerry Palmer and others have done studies in order to be able to design definitive trials in that group.

Q: Are there any data to suggest obesity-induced insulitis can accelerate beta cell apoptosis in type 1 diabetes?

Dr. Dotta: That is tough to say. When we look at apoptosis with insulitis, we never find it. It is a process that lasts for just a few hours or a few minutes or is just not frequent. So, we don’t have a good way to quantify it. What you propose is possibly suggestive and interesting though. In type 2 diabetes, fatty pancreases are more prone to losing beta cells regardless of BMI. Therefore, visceral fat and obesity in the pancreas may contribute to beta cell death in type 2 diabetes.

Dr. Eizirik: I would also caution that it is not clear that there is insulitis in type 2 diabetes. Most available data suggest that a mild innate immune response may occur in type 2 diabetes, but it never reaches the level observed in type 1 diabetes. There is no data to suggest that you have an adaptive and targeted immune assault on pancreatic beta cells in type 2 diabetes. I wouldn't use the word insulitis in type 2 diabetes.

Q: There seems to be a huge gap between the advances in diabetes research at the cellular level and the clinical application of this knowledge toward achieving insulin independence. Which of the approaches do you foresee as the most promising for allowing individuals with type 1 diabetes to achieve insulin independence?

Dr. Skyler: I think it will be a combination of things and not just one approach.

Dr. Herold: I agree with Dr. Skyler. It’s not going to be simple. I wouldn’t take this as discouraging news. We don't treat cancer with a single drug. We don't treat chronic viral infections with a single drug. Dr. Skyler is right. We need a combination of agents that will target both the innate and adaptive immune response to beta cells.

Dr. Roep: I think it also speaks to the need for better biomarkers to assess beta cell mass and function.

Q: I was struck by that example of the patient with persistent autoimmunity. That is the bad news and it’s not surprising. The good news is that there is persistent beta cell mass in patients with established type 2 diabetes and we know that from other evidence. How do we explain that we’re not seeing some degree of spontaneous recovery in situations where patients are immunoprotected or immunomodulated? There is no proliferation with KI-67 staining, so we conclude that there is no regeneration going on. Doesn’t this suggest that there must be some low-level islet regeneration that could allow people with established type 1 diabetes to respond to immunomodulatory therapy?

Dr. Herold: I guess the question I would have would be what you mean by spontaneous. NOD data would suggest that rather than a significant increase in beta cell mass, there is functional recovery. I don’t think it’s spontaneous because it doesn’t happen on its own. At least the way we’ve been thinking about it is that beta cells, when they get near the time of diagnosis, are undergoing significant metabolic stress and are becoming very poor at making insulin. Then they’ll stop making insulin and that’s when you present. It doesn't mean they are dead. If you let the immune response continue they will be dead. If you can alleviate the immune response there is some chance of recovery.

Dr. Roep: Could it be that diagnosis is the point of no return, but you get nonspontaneous recovery before that? Could it be that the majority of people losing tolerance to islets remain healthy in the end? It could be that there is a lot of recovery going on without us knowing.

Dr. von Herrath: In the trials in individuals with recently established diabetes, there appears to be a pattern by which we see some C-peptide preservation or even some improvements in secretion. Then we see C-peptide continue to decline, even if another dose is given. What is happening here? There seems to be a regain of beta cell function. Why do these drugs stop working?

Dr. Herold: In the ABATE trial, we gave anti-CD3 twice. 20-40% developed anti-drug antibodies. There was a higher proportion of these individuals in the non-respondent group than the respondent group. So, there may be some effect of neutralizing antibodies. There was also a group of individuals, a not unsubstantial group, that were stable with respect to C-peptide over two years. However, when you look at the average of the entire study group, you don’t see that there were some people that were quite significant responders. For the question you raised, why did this happen? I think we have to turn to our beta cell colleagues. We may have in front of us breakthroughs in immune responses that we don’t quite realize yet. There may be something about the programming of beta cells at the time of diagnosis or even before that may lead to their eventual demise.

Dr. Harlan: Annette Ziegler has also reported that children born via C-section are at significantly greater risk for diabetes than those born vaginally. Had you seen this?

Dr. Herold: No, but blue eyes is also a risk.

Dr. Harlan: Blue eyes isn’t something we do. C-section is something we do.

Dr. Eizirik: One thing that is interesting in that context is that there are major changes in methylation of human islets at that time. We published a study where we looked at global methylation patterns from type 2 diabetes individuals. There are 271 genes with major changes. It is not to be excluded that epigenetic changes early in life may have an early impact in the ways cells are going to function in the coming years. This is a pure speculation, but is possible that neonatal phenomena will have a long term impact on cells.

Q: What are you thoughts about the wisdom of using cellular approaches to treat type 1 diabetes such as mesenchymal stromal cells?

Dr. Skyler: Mesenchymal stromal cells have been effective at treating some diseases, such as CV diseases. They are probably anti-inflammatory agents, so this is the same thing we talked about today and yesterday.


Debate: Insufficient Biomarkers and Endpoints Doomed Recent Phase 3 Clinical Trials


Jay Skyler

In the opening presentation of the debate, Dr. Skyler argued that insufficient biomarkers and endpoints did indeed doom recent phase 3 diabetes immunotherapy clinical trial results. Dr. Skyler started off by reviewing GAD-alum vaccination, which in phase 2 trials was shown to be effective at preserving stimulated in C-peptide secretion in people diagnosed with type 1 diabetes in the prior six months, but in a series of recent trials (a US Diamyd-sponsored study, a European Diamyd-sponsored study, and a TrialNet-sponsored study), failed to show a benefit versus placebo. He emphasized in his discussion that GAD65’s phase 2 trial never met its primary endpoint and the compound was only shown to be efficacious in a small subgroup. It may have been imprudent to move forward into large scale clinical trials based on such limited efficacy data, he suggested. Subsequently, Dr. Skyler discussed how teplizumab’s phase 3 PROTÉGÉ trial failed to show efficacy after promising phase 2 results due to incorrect choice of endpoint and shortcomings in patient selection in phase 3. Finally, Dr. Skyler described what might have gone wrong with otexliziumab, whose phase 3 DEFEND-1 trial also failed to meet its primary endpoint in a phase 3 trial. According to Dr. Skyler, far too low of an otelixizumab dose was used in DEFEND-1, which sacrificed efficacy for the sake of safety concerns that were not even particularly concerning to the study’s data and safety monitoring board. Generally, from the GAD65, teplizumab, and otelixizumab results, Dr. Skyler said, we see that small pilot studies need to be viewed with caution and that better biomarkers are needed to serve as endpoints for trial design.

  • Efficacy signals from early GAD65 studies were based on small numbers, possibly explaining recent phase 3 letdowns. As a reminder, a 2008 NEJM study by Ludvigsson et al. suggested that GAD treatment was associated with significantly less decline in stimulated C- peptide levels than alum treatment in patients treated within zero and six months of diagnosis. However, it is often overlooked that this trial failed to meet its primary endpoint (change in fasting C-peptide after 15 months). Significant results were only obtained from it when examining a very small subgroup of patients (n=11 and 14 GAD and placebo patients, respectively) who had been diagnosed with type 1 diabetes within the last six months. Despite incorrectly choosing an endpoint for its early trials, Diamyd went ahead with large-scale clinical trials based on the positive findings in this small subroup, as did TrialNet. Not surprisingly, three recent sets of results (from a European Diamyd-sponsored study presented at ADA 2011 and recently published in NEJM, from a US-Diamyd sponsored study whose full results haven’t yet been presented, and from a TrialNet sponsored study also presented at ADA 2011) have suggested that GAD does not confer a benefit in terms of C-peptide decline versus placebo. Notably, in recent trials, GAD treatment was associated with increased GAD antibody levels, suggesting that GAD alum is capable of eliciting an immune response, but not necessarily a protective one. Dr. Skyler emphasized that this is not the first antigen-specific type 1 diabetes therapy to fail. He also detailed how hard antigen specific studies are to conduct because of how challenging it is to translate dosing from animal models to humans, to know whether the compound is being dosed via the right route, whether adjuvant should be given at all, and when the antigen should be given, and to assess whether the antigen is indeed causing an immune response. Antigen-specific therapies might thus best be used as a component of a combination, he suggested.
  • Incorrect choice of endpoint and shortcomings in patient selection impeded proof of teplizumab’s efficacy in the phase 3 PROTÉGÉ trial. The composite endpoint used in this trial (A1c <6.5% and insulin dose of < 0.5 U/kg/day), Dr. Skyler argued, was arbitrarily selected, not statistically robust, and puzzlingly, different than that used in successful phase 2 trials. It was used so that people would recognize the drug’s market potential, but eventually failed to help the drug in this way. If the study had used an A1c of < 7.0% and an insulin dose of < 0.25U/kg/day as the primary composite endpoint, Dr. Skyler noted, the 14-day low dose teplizumab endpoint would have been shown to be effective. Additionally, this study had shortcomings related to demographics and age groups included. Although what we know about type 1 diabetes mostly applies to Europoid Caucasians, 28% of PROTÉGÉ’s participants were from India. Notably, this means that results could have been swayed by the study’s Indian population having different genetic markers and different responses to immunotherapy than other participants. Additionally, enrollment of a range of age groups in the study may have been problematic. Interestingly, secondary analyses demonstrated that children 8-11 had a dramaticmaintenance of C-peptide post treatment in PROTÉGÉ even while other groups didn’t. Since C- peptide levels are known to decline much more slowly after disease onset in people over age 21, enrollment of older subjects in the study may have confounded its results.
  • Use of a too low of an otelixizumab dose in late stage clinical trials led to this drug’s disappointing phase 3 results. Although it was known that use of the otelixizumab dose that proved effective in phase 2 trials was associated with cytokine release syndrome and reactivation of mononucleosis, according to Dr. Skyler, the timing and severity of these side effects was known and they weren’t concerning to the trial’s data safety monitoring board. Nevertheless, those who developed the drug decided to reduce the dose used in phase 3 trials by 1/16th versus that used in phase 2 trials (3.1 mg versus 48 mg in phases 3 and 2, respectively). Dr. Skyler argued that it was this reduction that caused a lack of efficacy.


Carla Greenbaum, MD (Benaroya Research Institute at Virginia Mason, Seattle, WA)

Taking the con position, Dr. Greenbaum argued that the failures of the recent phase 3 trials for Diamyd’s GAD65 vaccine, Tolerx’s anti-CD3 antibody otelixizumab, and Macrogenic’s anti-CD3 antibody teplizumab were not caused by the endpoints and biomarkers used; rather, she contended that the trials failed because the drugs “just didn’t work.” Addressing the endpoints used first, she highlighted data from the European C-peptide Trial Study Group and TrialNet that demonstrated that the mixed meal tolerance test (when used to measure C-peptide response) is a robust, highly reproducible, and sensitive measure of beta cell function. Thus, according to Dr. Greenbaum, the endpoints used in these phase 3 trials (meal-stimulated C-peptide [Gad 65; otelixizumab]; a composite of total daily insulin usage and A1c [teplizumab]) were not the problem. Similarly, she argued that the biomarkers used in the trials were also adequate. For the GAD 65 vaccine, GAD antibody levels increased with treatment. With the anti-CD3 therapies, effects on CD8+ T cells and regulatory T cells were observed. Instead, taking a big picture view, she stated that it was not actually surprising that the trials failed since over 50% of all phase 3 drug trials fail due to poor efficacy. Additionally, she suggested that evidentiary conservatism – the tendency to base clinical information on a narrow class of evidence and to make go/no go decisions based on subjective views – may have played a role, causing the companies to launch phase 3 trials without strong evidence for clinical efficacy. Another potential cause of failure may have been an inaccurate understanding of the pathology of type 1 diabetes. She noted that studies have showed that a significant percentage of individuals (60%) with type 1 diabetes treated with placebo still posses clinically significantly levels of C-peptide after two years. Thus, when taking examining the effect of an immune therapy on C-peptide levels, it is important to look at the effect of placebo on matched controls. Finally, Dr. Greenbaum reviewed a series of data suggesting that the rates of beta cell destruction alter during the course of disease progression, indicating that the underling immune processes may also change with time. Therefore, it will be critical to assess the time points at which different therapies may have greater or lesser efficacy.

Questions and Answers

Dr. Fleming: Carla, you argued well partly because you were responsible for putting together the joint NIH-JDRF workshop that recommended C-peptide as regulatory endpoint for pipeline studies. I’m sure you haven’t backed away from it. Jay, I think you made valid criticisms of other parts of trials. I would have to agree with Carla though that C-peptide is still best way to measure insulin secretion.

Dr. Skyler: But that’s not what was measured as primary endpoint for PROTÉGÉ. 

Dr. Greenbaum: And the data was negative in the primary analysis.

Dr. Herold: I have a question for Dr. Skyler. The way that the endpoint ended up being changed was not actually from a desire to change the endpoint, but actually from pressure from regulatory authorities. Dr. Fleming may want to comment on that. When we had the meeting with the FDA, the agency was unwilling to accept C-peptide as an endpoint. In some ways, we were backed into that situation. However, in the original anti-CD3 trial, that endpoint was missed. I agree with the point that C-peptide is a biological thing that we are detecting. But we need to look at the bigger picture. When do you say something isn’t working? Take GAD 65 for example. In the initial trial, there were only a small number of individuals and insulin usage never changed. The problem becomes complicated because glucose control is typically pretty good, taking changes in A1c off the table. We need to see the bigger picture. We need to see improvements in C-peptide and insulin use to see a biological effect.

Dr. Skyler: I don’t disagree with that. My biggest argument about the endpoint used in that trial is that you needed to reach two arbitrary thresholds to meet the endpoint. If you look at something on a continuous basis, you will have greater power to see if there is an effect. This is a subtle criticism. But the trial basically combined endpoints with two arbitrary numbers. It makes it difficult to fit it into that box. If we looked at C-peptide and A1c on a continuous basis, we might have found statistical significance with each as separate components. Those things come into play too.

Dr. Greenbaum: I would also add that we want to see have how everything is aligned. We want to see how homologous and large the study population is. We need to look at the whole disease and realized that we are getting data from real subject. There many be a number of different factors that could influence the results from a particular person. C-peptide is the most robust measure we have at the moment. We would have more consistency in trial results if the way the trials were conducted were more consistent. I agree that we want to see improvements in both C-peptide and insulin use.

Q: I have two comments. I’m afraid that if we have a short follow-up, we won’t be able to tell properly if it works. The windows we use may not even be the appropriate ones. I’m also afraid we’re not using all the appropriate endpoints. We never include biomarkers, with the exception being oral insulin, which we would deliver if we had stuck to the original rules. Since this is an autoimmune disease and the endpoint is not immune but metabolic, should we add immune markers to cope with disease heterogeneity?

Dr. Skyler: You need something robust enough so that you can use biomarkers. Carla showed that biomarkers were adequate with anti-CD20 because it lowers B-lymphocytes. I want to know what you use for an altered peptide ligand where you don’t have that. How can you look for its effect reliably? I want something that can reflect disease processes, not just reduce B-lymphocytes because it’s a B lymphocyte drug, because that doesn’t necessarily correlate with disease state.

Q: And if you have the right biomarkers, you would know who you should never give islet transplantation to or who you should not give autologous human hematopoietic stem cell therapy to, not based on anything but autoimmune responses. There are a number of ways. Even in pilot trials with DiaPep277 there are immune phenotypes that will predict who will not respond but they are not randomized among the right arms.

Dr. Greenbaum: This is a classic issue if you are a company in drug development or a regulatory agency because you want something that will be widely used. So do you want to do the trial in the general population and go back and say who did a better job or only do it in the people you are predicting you will respond. I agree we are not using biomarkers well enough. Oftentimes we don’t know what to do. It’s a challenge when you’re working with funders and regulatory agencies.

Q: Someone said that don't treat all patients in same way in cancer. I think we have to prepare to do that in the same way in diabetes.


Therapeutic Beta Cells


Kevin D’Amour, PhD (ViaCyte, San Diego, CA)

Dr. D’Amour provided an overview of ViaCyte’s preclinical development of an encapsulated cell therapy for insulin-dependent diabetes. As a reminder, the therapy is comprised of human embryonic stem cell (hESC)-derived pancreatic progenitor cells (Pro-Islet) packaged into an implantable, retrievable, non- biodegradable, vascularizing, and selectively permeable macroencapsulation device. Following transplantation, the device acts to protect and promote the maturation of the progenitor cells into pancreatic hormone producing cells, including insulin-producing cells. In vitro studies have demonstrated the ability of Pro-Islet pancreatic progenitor cells to form islets that resemble human islets (based on structure and hormone staining) at 377 days. With regards to manufacturing, Dr. D’Amour noted that the company is capable of producing 20 billion pancreatic cells in approximately four weeks. Studies using hESC and endoderm markers have suggested that the process is highly reproducible, with no hESCs detected at five days into differentiation and with few mesoderm and no ectoderm cells present. With the macroencapsulation device, Dr. D’Amour stressed that it would protect enclosed cells from both alloimmune and autoimmune rejection. While the device is permeable to antibodies and cytokines, he believed that the device would pose enough of a barrier to avoid most immune destruction (suggesting that occasional Pro-Islet replacement may be necessary for long-term use). In immunocompromised, streptozotocin (STZ)-induced diabetic mice, the combined product was found to be safe (including no teratoma development) and effective in protecting against hyperglycemia for up to 53 weeks. Addressing human use, Dr. D’Amour revealed that the product would be implanted subcutaneously (at least in initial trials), that multiple capsules would likely be needed, and that probable locations for implantation include the trunk and the lower back. A timeline for the initiation of clinical trials was not provided. As a reminder, JDRF announced in December that it would help fund, alongside the California Institute for Regenerative Medicine (CIRM), a three-year series of preclinical studies for ViaCyte’s product (see the December 28, 2011 Closer Look at http://bit.ly/ztBR3O).

Questions and Answers

Q: Is the device permeable to cytokines?

A: Yes, the device is permeable to antibodies and cytokines. However, we think the device will be a barrier enough in itself to avoid most immune destruction. We are not going to avoid the immunological mechanism; it just won’t be as accessible.

Q: Can you talk about fibroblast overgrowth and the durability of the product?

A: Because we have such a highly enriched endoderm population, we see very little mesoderm and zero ectoderm cells. Outside, the device becomes very well vascularized and very stable. We have had the device in animals for up to 53 weeks. At 53 weeks, mice are extremely old. We will only learn about durability most likely once we start testing in humans.

Q: Is oxygen tension normal in the device?

A: I think it is somewhat self-limited. Because the blood vessels can’t get super close to the cells, some of the cells in the middle of the device won’t receive as much oxygen. So, expansion becomes somewhat limited, etc.

Q: I assume you will be testing the device subcutaneously in humans. Have you though about looking at portal transplantation?

A: For our first human studies, we will be sticking with subcutaneous insertion for safety advantage reasons. However, if this is not shown to be efficacious enough, we will examine other options.

Q: Have you attempted to retrieve the device? How long have these animals survived post- removal?

A: Yes, we have. We have followed these mice for about one to two weeks. We have been quite pleased with how easily it slides out. For the mice, the device is almost like what the size of a cafeteria tray would be like to us. The device will be much smaller, in a relative sense, in humans. Still, the mice did well and it was not very difficult.

Q: How many capsules will be needed to achieve insulin independence?

A: There are two aspects for dosing. The first is how many cells are captured in each device and what level are you trying to achieve at the time of implantation in humans? Ultimately, this could be something like 200,000 islet equivalents ultimately on board. That is different from figuring out what level of cells must remain functional for patients to remain insulin independent. The islet transplantation literature is fairly unhelpful in this regard. We know what is delivered, but we have little knowledge of what remains there at the time of insulin independence. I can tell you that even with the largest device we are examining, it will be a multiple devices in humans. We will only find out how many will really be needed once we begin examining the product in humans. We also have a development program under way that is seeking to make a product that maximizes cell density.

Q: During the in vivo differentiation period, is there any variability depending on the host environment?

A: We have looked into this this. No matter what strain you use, the product is pretty adaptive. There may be some kinetic impact depending on the particular strain. However, the device dose cap ultimate output, regardless of how many cells you start out with. It just gets there at different times. We have examined implantation at numerous different sites in rats and mice, and the product has worked in each of these. We’ve used six different strains of mice, and they’ve all worked. It is very generalizable to the in vivo environment.

Q: In humans, where do you intend to transplant the device?

A: We are in discussions with endocrinologists, islet transplant experts, and plastic surgeons. Most likely, it will be on the trunk and lower back. We don’t wan the patients to have great access to it. Still, we have not fully decided yet.

Q: What materials do you use in the device?

A: It is comprised of an inert polymer that has been used for transplantations in humans for decades.

-- by Ben Kozak, Lisa Rotenstein, and Kelly Close