January 31-February 3, 2018; Pasadena, CA; Days #3-4 Highlights – Draft

Executive Highlights

We may have made the trip back north from Pasadena, but we still have one more installment of highlights from Levine-Riggs. In the meeting’s final days, TrialNet chair Dr. Carla Greenbaum gave an update on the consortium’s efforts, and we were particularly struck by her discussion of possible type 1 diabetes endotypes – that is, different forms with different phenotypic expression, suggesting distinct pathobiological mechanisms and distinct approaches to treatment. Beta cell mass data indicate that type 1 diabetes does not manifest the same in all patients, meaning all patients probably should not receive the same immunotherapy.

On the type 2 diabetes/obesity therapy front, we also gleaned fascinating insights into the mechanisms underlying the mechanism of action for much-hyped obesity candidate FGF21 as well as evidence for future research into therapies that target AMPK and peripheral serotonin, two signals mediating the cellular energy gauge.

Below, we bring you our top 5 highlights from Levine-Riggs days #3-4, encompassing all this and more.

In case you missed it, take a look at our previous coverage of day #1 and day #2. We’re already looking forward to next year’s meeting!

Top 5 Highlights

1. TrialNet’s Dr. Carla Greenbaum on Type 1 Diabetes Endotypes, Targeting Immunotherapy to Ideal Patients

During TrialNet’s update on disease-modifying therapy for type 1 diabetes, Dr. Carla Greenbaum presented evidence of type 1 diabetes endotypes, or subtypes of the condition. She argued for the need to better understand the etiology of type 1 in order to target immunotherapy to the right patients. Dr. Greenbaum showed how disease-modifying immunotherapy has worked in type 1 diabetes trials, but went on to explain that current research is trying to unlock answers on when immunotherapy will work optimally. This requires identifying when someone with antibodies will progress to type 1 diabetes, the mechanisms underlying that progression, and ways to prolong or increase the effect of immunotherapy. Dr. Greenbaum reviewed an emerging model of the natural history of type 1 diabetes: The first-phase insulin response (FPIR) changes, on average, ~1.5 years before diagnosis, accompanied by non-linear drops in insulin secretion after diagnosis. The drop-in insulin secretion is steeper in the first year after diagnosis, but then levels out: This is the target area for TrialNet’s beta cell preservation trials. Those studies, along with prevention trials, are accumulating data on the progression to type 1 diabetes. Dr. Greenbaum described the LIFT (Long-Term Investigative Follow-Up) study, which has enrolled 65 people to-date from among those in prevention studies who develop type 1 diabetes. LIFT aims to tease apart the biology of this phenotype – is it immune-mediated or is something else happening with the beta cells? Another TrialNet project on functional beta cell mass in 38 adults, all with 2+ antibodies but without having developed type 1 diabetes, has revealed a wide range of function, reproducible within individuals. Researchers believe this could represent endotypes of type 1 diabetes ­– that is, different forms of the disease with distinct pathobiological mechanisms that should be targeted in treatment. Dr. Greenbaum acknowledged the possibility that heterogeneity in beta cell mass reflects time of observation, but noted that it’s also possible that these truly are endotypes of (i) intact functional mass but impaired insulin secretion; and (ii) impaired functional mass + impaired insulin secretion – these distinctions could have important implications for how and who to treat with immunotherapy.

2. FGF21: A Powerful Insulin Sensitizer and Early-Stage Obesity Candidate

University of Iowa’s Dr. Matthew Potthoff characterized FGF21 as “one of, if not the most potent insulin sensitizer ever identified.” Past studies provide compelling evidence that FGF receptor activation increases energy expenditure via adipose tissue, the hypothalamus, and the hindbrain. Dr. Potthoff’s research has elaborated on this, describing the unique role FGF21 plays in each of these tissues. He discussed how FGF21 activity is responsible for increasing energy expenditure via a neural mechanism, but added that FGF21 activity in adipose tissue also has an acute effect on insulin sensitization. In preclinical studies, a single injection of FGF21 increased plasma insulin levels by >50% in a mouse model of type 2 diabetes. While this is extremely early-stage evidence, it suggests that FGF21 agents in development could have profound beneficial effects in type 2 diabetes. That said, we note that Pfizer and Lilly both discontinued FGF21 candidates for diabetes after underwhelming glucose-lowering efficacy. Many FGF21 candidates in the pipeline currently are being investigated for obesity, and this also holds exciting promise, since the drugs could stimulate meaningful weight loss via increased energy expenditure. Novo Nordisk and Genentech both have FGF-targeting therapies in phase 1 for obesity. See our obesity drug competitive landscape for more on this.

• Dr. Potthoff called attention to findings from 23andMe’s recent Genetic Weight Report. Highlighted earlier this year at JPM 2018, the project used machine learning techniques on genetic data from >one million 23andMe participants to identify genetic loci that predispose a person to weigh more or less than average. Variants of the FGF21 gene were associated with dietary macronutrient intake and sweet preference – this further suggests the potential for FGF21 analogs to contribute to the next generation of obesity therapy. 23andMe’s new Weight Loss Intervention Study will go further to identify genetic variants associated with weight loss and the effectiveness of different diet/exercise methods. In our view, it’s noteworthy that 23andMe’s first foray into producing original research addresses obesity – a scientifically challenging area made even more difficult by extreme stigma and low public awareness of obesity as a disease. If there was ever an area where innovative companies like 23andMe could make an impact by identifying genes of interest and potential new drug targets, this is it. 

3. TrialNet Study Reveals Possible microRNA Biomarkers for Type 1 Diabetes; Improving Predictive Power of Autoantibodies

University of Miami Diabetes Research Institute’s Dr. Alberto Pugliese discussed the possibility of using microRNA, non-coding RNA molecules that regulate gene expression but also circulate through the bloodstream, as a potential biomarker for type 1 diabetes risk and progression. Using samples from TrialNet participants (n=300), which includes relatives of patients with type 1 diabetes, Dr. Pugliese’s research group uncovered that elevated levels of seven distinct microRNAs were associated with increased risk of developing type 1 diabetes among relatives who were already at high risk due to multiple islet autoantibodies. Moreover, the levels of these microRNAs were correlated with measures of metabolic impairment and indices of disease progression from an oral glucose tolerance test. Dr. Pugliese’s microRNA signature could help boost predictive power on the autoantibody front – autoantibodies by themselves are imperfect predictors of disease progression – by identifying which autoantibody-positive people are at highest risk of progression to diabetes. Next up, Dr. Pugliese hopes to conduct longitudinal studies to further define changes in microRNA profiles during disease progression and identify microRNA signatures that can aid in predicting disease progression. We would be particularly interested in further studies of autoantibody-negative relatives, which may aid in identifying those at risk of developing autoantibodies, potentially leading to the discovery of early biomarkers.

4. Scripps’ Dr. Nicholas Schork on Regulatory Challenges Surrounding Algorithms, AI for Personalized Medicine, and Doing Away with Phase 3 Trials

Dr. Nicholas Schork (Scripps, San Diego, CA) explored potential problems with the algorithms underlying new biomedical technology (they’re ever-changing, they depend on the quality of input data, and outcomes can’t easily be attributed to drugs vs. the algorithm) while also advocating for their further development and incorporation into care models. He argued that we need better methods for vetting and deploying these technologies, suggesting that step one is answering this question: What makes a technology successful? Is it individual outcomes? Improvements in quality of life? Cost-savings? Only then can we move onto verifying and evaluating new tech-based strategies for treating disease, Dr. Schork explained. He discussed “basket” trials of cancer drugs, where enrolled patients have their mutations profiled and then, on the basis of a drug’s mechanism of action and how it might impact the pathology induced by that mutation, a patient is steered toward a particular “basket” of treatment. The mutation, rather than the type of cancer, is treated. This introduces some uncertainty, because if a trial has poor outcomes, is it the drug’s fault or the algorithm’s fault? Moreover, Dr. Schork emphasized how everything depends on the biology linking mutations to treatments – and that science is changing all the time. Can you revise the rules after a trial has begun? Dr. Schork led us to think not: Doing so would be like changing the chemical composition of a pill halfway through. In this sense, algorithms, much easier to change than a drug’s composition, present unique, novel challenges to regulatory agencies.

• Dr. Schork provided an interesting conceptual schematic of the progression from traditional, to stratified, to precision, to individualized/personalized medicine. Do outcomes improve with n=1 care? Dr. Schork outlined a method to investigate this question: (i) Collect data from N patients on factors impacting drug response (history, genomic profile, biomarkers, pathology analysis). (ii) Form an N x N similarity matrix from patients’ drug response profiles. (iii) Lastly, cluster patients using that matrix to find the rules that impact treatment response. Repeating this over and over pushes predictive capabilities toward greater and greater specificity – but the question remains of when enough is enough. Citing three recent studies of AI in identifying diabetic retinopathy, making treatment decisions for acute myeloid leukemia, and classifying skin cancer, Dr. Schork suggested that AI can perform at least as well as a panel of physicians; however, he said he views these tools as providing decision support to physicians, rather than replacing them altogether.
• A murmur went through the room when Dr. Schork broached the idea of eliminating phase 3 trials. While he didn’t advocate explicitly for this movement, he shared that many people think it would save time, money, and energy to do so. The idea is that, if safety/efficacy have been shown in phases 1 and 2, an infrastructure could be built to objectively determine who responds best in a real-world setting and under what circumstances. We find this last point compelling, though we also note that phase 2 results can be particularly misleading, that bringing a drug to market bears its own significant costs (and manufacturers want to know what they have on their hands), and that flooding the market with new drugs that have less clinical data backing them might cause more confusion for real-world practice decisions. Nevertheless, we absolutely support this form of real-world data collection, aimed at vetting drugs and expediting adoption.

5. Harnessing the “Cellular Fuel Gauge” (AMPK vs. Serotonin) for Diabetes/Obesity Drug Development

Dr. Gregory Steinberg (McMaster University, Hamilton, Canada) provided a deep dive into the molecular pathways governing the metabolic sensing of nutrient availability, an area with important implications for future therapy development since type 2 diabetes involves a mismatch in sensing of cellular energy demand and nutrient availability. According to Dr. Steinberg, cells sense energy using two gauges: AMP-activated protein kinase (AMPK) to signal that energy is running low, and serotonin to signal that energy is running high. Dr. Steinberg explained that activation of AMPK suppresses lipid synthesis and inflammation, while simultaneously increasing glucose uptake, fatty acid oxidation, and mitochondrial function. Accordingly, AMPK levels are elevated in response to exercise and metformin therapy. On the other hand, serotonin suppresses this process, so peripheral serotonin levels are elevated in people with type 2 diabetes and in animal models with diet-induced obesity. Given this framework, Dr. Steinberg noted the data suggest that identifying new ways to manipulate these “ancient fuel gauges” – that is, activating AMPK and/or inhibiting peripheral serotonin – could be a promising avenue for future type 2 diabetes therapies. We’re aware of one obesity therapy, Arena/Eisai’s Belviq (lorcaserin), that targets serotonergic signaling, but this drug acts primarily in the brain’s serotonin signaling pathways rather than the peripheral serotonin system where the cellular energy gauge lies. Dr. Steinberg noted, however, that it’s important to remember serotonin has opposite effects in the brain versus the periphery. With the acknowledgement that this research is very early stage, we are certainly intrigued by the AMPK/serotonin cellular energy gauge story – this is a preclinical area we’ll be watching closely.

-- Ann Carracher, Abigail Dove, Payal Marathe, and Kelly Close