Diabetes Care Volume 43, January 2020 5

Manuela Battaglia,1 Simi Ahmed,2 Introducing the Endotype Concept Mark S. Anderson,3 Mark A. Atkinson,4 Dorothy Becker,5 Polly J. Bingley,6 to Address the Challenge of Emanuele Bosi,1,7 Todd M. Brusko,4 Linda A. DiMeglio,8 ESETVSI CARE IN PERSPECTIVES Disease Heterogeneity in Carmella Evans-Molina,9 Stephen E. Gitelman,10 Type 1 Diabetes Carla J. Greenbaum,11 Peter A. Gottlieb,12 13 14 Diabetes Care 2020;43:5–12 | https://doi.org/10.2337/dc19-0880 Kevan C. Herold, Martin J. Hessner, Mikael Knip,15 Laura Jacobsen,16 Jeffrey P. Krischer,17 S. Alice Long,11 Markus Lundgren,18 Eoin F. McKinney,19 Noel G. Morgan,20,21 Richard A. Oram,22,23,24 Tomi Pastinen,25 Michael C. Peters,26 Alessandra Petrelli,1 Xiaoning Qian,27 Maria J. Redondo,28 Bart O. Roep,29,30 Desmond Schatz,16 David Skibinski,11 and Mark Peakman31,32 The clinical diagnosis of new-onset type 1 diabetes has, for many years, been considered relatively straightforward. Recently, however, there is increasing awareness that within this single clinical phenotype exists considerable hetero- geneity: disease onset spans the complete age range; genetic susceptibility is complex; rates of progression differ markedly, as does insulin secretory capacity; and complication rates, glycemic control, and therapeutic intervention efficacy vary widely. Mechanistic and immunopathological studies typically show considerable patchiness across subjects, undermining conclusions regarding disease pathways. Without better understanding, type 1 diabetes heterogeneity represents a major barrierbothtodecipheringpathogenesisandtothetranslational effortofdesigning, conducting, and interpreting clinical trials of disease-modifying agents. This re- alization comes during a period of unprecedented change in clinical medicine, with increasing emphasis on greater individualization and precision. For complex disorders such as type 1 diabetes, the option of maintaining the “single disease” approach appears untenable, as does the notion of individualizing each single 1San Raffaele Diabetes Research Institute, IRCCS patient’s care, obliging us to conceptualize type 1 diabetes less in terms of San Raffaele Hospital, Milan, Italy phenotypes (observable characteristics) and more in terms of disease endotypes 2JDRF, New York, NY 3 (underlying biological mechanisms). Here, we provide our view on an approach to Diabetes Center, University of California, San Francisco, San Francisco, CA dissect heterogeneity in type 1 diabetes. Using lessons from other diseases and the 4Department of Pathology, Immunology, and data gathered to date, we aim to delineate a roadmap through which the field can Laboratory Medicine, University of Florida, Gain- incorporate the endotype concept into laboratory and clinical practice. We predict esville, FL 5 that such an effort will accelerate the implementation of precision medicine and has Division of Endocrinology and Diabetes, UPMC Children’s Hospital of Pittsburgh, Pittsburgh, PA the potential for impact on our approach to translational research, trial design, and 6Translational Health Sciences, Bristol Medical clinical management. School, University of Bristol, Bristol, U.K. 7Vita-Salute San Raffaele University, Milan, Italy, and Department of Internal Medicine, IRCCS San Describing aspects of biology as “heterogeneous” often has a negative connotation. It Raffaele Hospital, Milan, Italy is a term that is used when we do not understand a measured or observed aspect of 8Division of Pediatric Endocrinology and disease or when we need to explain data that are not consistent. However, it is evident Diabetology and Wells Center for Pediatric that recognizing that there are “different kinds” of cells, genes, types of response, and Research, Department of Pediatrics, Indiana University School of Medicine, Indianapolis, IN severity of disease could offer a set of opportunities for therapies to work and 9Herman B Wells Center for Pediatric Research, biomarkerstobemeaningful. Thus,itmaybetimetoexploitheterogeneityratherthan Indiana University School of Medicine, Indian- curse it and to use the opportunity to carve out endotypes of type 1 diabetes that apolis, IN 10 have traction both in the clinic and in the laboratory. Division of Pediatric Endocrinology and Diabe- “ ” tes, University of California, San Francisco, San The introduction of the term endotype can largely be attributed to developments Francisco, CA in the field of asthma (1) when it became apparent in the late 1990s that differ- 11Diabetes Program, Benaroya Research Insti- ent pathogenic mechanisms induce a similar symptom cluster and manifest as a tute, Seattle, WA 6 An Approach to Heterogeneity in Type 1 Diabetes Diabetes Care Volume 43, January 2020

phenotype, the implications being 1) that examples are highlighted in Table 1, and typically adopt very basic and standard there are multiple pathways to disease others are expanded below in ENDOTYPE inclusion criteria (e.g., short time from and 2) that pathway-specific therapeutic DEFINITION LED BY OBSERVATIONS AND HYPOTHESES. disease onset, wide age range, single strategies will appear to have limited Examples include continuous as well autoantibody) to assure consistency, success if applied to a population of as qualitative variables and span the enable cross-trial analysis, and, at a subjects in which the pathway is only different stages of disease. Of note, traits practical level, facilitate recruitment. active in a subgroup. From this new are often linked (e.g., age and HLA-specific Yet, one can imagine that factors thinking, the term “endotype” was coined; autoimmunity) in such a way that sug- such as disease severity, age, and un- in this case, the term reflects a subtype gests associations that could be built derlying genetic predisposition could of type 1 diabetes that can be defined into distinct pathobiological entities each influence trial outcomes and treat- by a distinct functional or pathobiological (endotypes). ment responsiveness (“theratypes”)(5). mechanism (that is also tractable ther- These are rarely, if ever, used as strati- apeutically). THE IMPACT OF TYPE 1 DIABETES fiers and when prespecified as covari- Here, we focus on gaining a better HETEROGENEITY ON CLINICAL ates their utility is often limited by understanding of heterogeneity in type TRIALS AND RESEARCH insufficient statistical power. In prac- 1 diabetes and how the endotype con- An overarching goal of type 1 diabetes tice, whether a trial succeeds or fails cept might be introduced to the field in research has been to bring forward in meeting its primary objective(s), order to bring about a sea change in disease-modifying therapies that pre- there are often subgroups of subjects clinical practice and research activity. As serve b-cell function (2). This has who appear to benefit from the therapy. the number of targeted immunotherapy been allied with progress made in One such example is monoclonal anti- treatments under development contin- the design and conduct of interven- CD3 antibody (2). Despite promising ues to grow and associated clinical trial tion (stage 3) and prevention (stages results in phase II trials, a phase III trial activity proceeds unabated, this is a 1 and 2) trials. Despite some successes with this agent did not meet its pri- propitious moment in which to evaluate and considerable knowledge gain, no mary composite outcome of insulin use whether, and how, a strategic approach agent has progressed beyond phase III (,0.5 units/kg per day) and glycated to disease heterogeneity could unlock clinical trials and into clinical practice, hemoglobin A1c (,6.5%) at 1 year. How- the power of disease-modifying drugs and as such, type 1 diabetes remains an ever, some patients appear to have designed to arrest b-cell decline. Since outlier among the autoimmune dis- responded rather well (i.e., younger the existence of heterogeneity in dis- eases. Numerous factors account for subjects with higher C-peptide at study ease traits is a critical component of this, but it is our contention that disease entry and patients from North America the rationale for studying endotypes, heterogeneity contributes to this im- and Europe) (6). More recently, this it is valuable to begin by reflecting on passe in the field (3,4). approach has shown efficacy in preven- the nature of type 1 diabetes diversity. Hitherto, the potential confound- tion of diabetes progression in high-risk Cataloguing all reported aspects of het- ing effect of heterogeneity has been subjects without diabetes; intrigu- erogeneity in detail is beyond the scope insufficiently addressed in the design of ingly, subgroups defined by HLA and of this article, and therefore some key type 1 diabetes clinical trials, which zinc transporter 8 autoantibodies appear

12Barbara Davis Center for Childhood Diabetes, 22Institute of Biomedical and Clinical Science, 31Peter Gorer Department of Immunobiology, University of Colorado School of Medicine, Au- University of Exeter Medical School, Royal Devon Faculty of Life Sciences and Medicine, King’s rora, CO and Exeter Hospital, Exeter, U.K. College London, London, U.K. 13 Department of Immunobiology, Yale Univer- 23NIHR Exeter Clinical Research Facility, Univer- 32King’s Health Partners Institute of Diabetes, sity, New Haven, CT 14 sity of Exeter Medical School, Exeter, U.K. Obesity and Endocrinology, London, U.K. Department of Pediatrics, Medical College of 24 Academic Renal Unit, Royal Devon and Exeter Corresponding authors: Manuela Battaglia, Wisconsin, Milwaukee, WI NHS Foundation Trust, Exeter, U.K. 15 [email protected], and Mark Children’s Hospital, University of Helsinki and 25 Center for Pediatric Genomic Medicine, Child- Peakman, [email protected] Helsinki University Hospital, Clinical and Molec- ren’s Mercy Kansas City, Kansas City, MO ular Metabolism Research Program, University of 26Division of Pulmonary and Critical Care Med- Received 2 May 2019 and accepted 14 October Helsinki, Helsinki, Finland 2019 16 icine, Department of Medicine, and Cardiovas- Department of Pediatrics, University of Florida, cular Research Institute, University of California, S.A., M.S.A., M.A.A., D.B., P.J.B., E.B., T.M.B., Gainesville, FL 17 San Francisco, San Francisco, CA L.A.D.M., C.E.-M., S.E.G., C.J.G., P.A.G., K.C.H., HealthInformaticsInstitute,MorsaniCollegeof 27Department of Electrical and Computer Engi- M.J.H., M.K., L.J., J.P.K., S.A.L., M.L., E.F.M., N.G.M., Medicine, University of South Florida, Tampa, FL 18 neering, TEES-AgriLife Center for Bioinformatics R.A.O., T.P., M.C.P., A.P., X.Q., M.J.R., B.O.R., D.Sc., Department of Clinical Sciences, Clinical Re- and Genomic Systems Engineering, Texas A&M and D.Sk. contributed equally to this work. search Centre, Faculty of Medicine, Lund Univer- University, College Station, TX ̊ M.B. is currently affiliated with the Telethon sity, and Skane University Hospital, Malmö, 28Baylor College of Medicine, Texas Children’s Sweden Foundation, Milan, Italy. 19 Hospital, Houston, TX Department of Medicine, University of 29Department of Diabetes Immunology, Diabetes © 2019 by the American Diabetes Association. Cambridge School of Clinical Medicine, Ad- & Metabolism Research Institute, Beckman Re- Readers may use this article as long as the work is ’ denbrooke s Hospital, Cambridge, U.K. search Institute, National Medical Center, City of properly cited, the use is educational and not for 20 fi Institute of Biomedical and Clinical Science, Hope, Duarte, CA pro t, and the work is not altered. More infor- University of Exeter Medical School, Exeter, U.K. 30 mation is available at http://www.diabetesjournals 21 Department of Immunohaematology and University of Exeter Medical School and Royal Blood Transfusion, Leiden University Medical .org/content/license. Devon and Exeter Hospital, Exeter, U.K. Center, Leiden, the Netherlands Battaglia and Associates 7

Table 1—Examples of heterogeneity in type 1 diabetes–associated traits Early disease (stages 0–2) Clinical disease onset (stage 3) Established disease (stage 4) Phenotypic c BMI: Children ,12 years old have higher rate of type 1 diabetes c Age of diagnosis: Typically younger age associates with lower number of c Emergence of other autoimmune progression if overweight/obese (25); disease course in overweight/obese is insulin-containing islets, serum C-peptide concentrations, duration of diseases (reviewed in 30). modified by presence of polymorphism in TCF7L2 (26). remission period; higher frequency of diabetic ketoacidosis; hyper- hi c Ethnicity: Risk of type 1 diabetes development differs according to ethnic immune CD20 insulitis; HLA-DR3/DR4 haplotypes; gene polymorphisms group (27). in PTPRK, THEMIS, and IAA; and higher overall number of AAbs, excess mortality, and frequency of cardiovascular disease in stage 4 (reviewed in 28). Adult onset more frequently GADA only, in association with other autoimmune diseases; gene PFKFB3 (29). c Coexistence of other autoimmune diseases (reviewed in 30). Genetic c HLA: Typically at age ,2 years, IAA emerges first and associates with c CTSH polymorphism associates with higher daily insulin c Chromosome 1 gene variants HLA-DRB1*0401/DQA1*0301/DQB1*0302 genotype; at age .6 years, GADA dose and faster disease progression (61). associate with severity of b-cell emerges first and associates with HLA-DRB1*0301/DQA1*0501/DQB1*0201 c TCF7L2 polymorphism associates with milder immunologic and loss (60). genotype (34). Specific haplotypes with high-risk DQ genes associate with rapid metabolic phenotype (62). progression to stage 3 (35,36); low-risk DQ genes (e.g., DQB1*0602) reduce type 1 diabetes development (37). c Genetic risk score: Specific constellations of gene polymorphisms associate with faster progression to stage 3 (38–40). Immune c Autoantibodies: type, affinity, titer, spreading, and tendency to revert (43–46). c Autoantigen-specific reactivities (41,42). c IL-2 signaling defect (55–57), Treg c Type I interferon signature: detected in peripheral whole blood activity (reviewed in 58), CD8 antigen and purified neutrophils (47–49). experience, and exhaustion (59) c Activation of innate immunity: detected in circulation (22,50). c T-cell signatures: CD4 proinflammatory (IFN-g), regulated (IL-10) (51,52), T follicular helper cells (53,54). Metabolic c Proinsulin/insulin processing: dysregulated, residual c Insulin secretory pattern: early insulin response associates c Insulin secretion: long-term insulin staining, proinsulin staining pattern (reviewed in 33). with faster rate of loss of insulin secretion (32). sustainers/nonsustainers (31) c Pathology: cellular constituents and extent of insulitis (17,18); evidence of viruses. c Metabolic: glucose and Index60 in relation to age and C-peptide (31,32). AAbs, autoantibodies; Treg, regulatory T cell. care.diabetesjournals.org 8 An Approach to Heterogeneity in Type 1 Diabetes Diabetes Care Volume 43, January 2020

to be differentially responsive to the over to type 1 diabetes?” This is not so subjects with high risk of type 1 diabetes drug (7). Similarly, while oral insulin easy, as type 1 diabetes has several that examine the timing of emergence of did not realize its primary objective complicating features. As examples: the specific autoantibodies indicate an early to delay progression from stage 1 to target organ is inaccessible for scien- peak of incidence of insulin autoantibody stage 3 type 1 diabetes, an indepen- tific interrogation in living individuals, (IAA)asthe first marker of autoimmunity, dently randomized subgroup (distin- there is uncertainty about the canonical strongly linked to the HLA-DR4 haplo- guished by having first-phase insulin immunological pathways that are re- type; in contrast, GAD autoantibodies release lower than a specified thresh- sponsible for b-cell death (and probably (GADA) emerge as the sole marker of old) had a 31-month delay in disease there is heterogeneity), and the disease autoimmunity later, and with a strong progression (8). These observations burden is greatest in children and ado- link to the HLA-DR3 haplotype (Table 1). should be an incentive for stratification lescents, restricting some types of ex- This example raises an important ques- to be built into trial design and for the perimentation. Indeed, age is clearly tion, namely, whether a specific endo- field to contemplate development of such a major driver of heterogeneity type represents a discrete, etiological drugs that work for the few rather in type 1 diabetes (Table 1) and other event and pathway or whether it is a than the many. Moreover, and impor- diseases that it merits further discussion distinct outcome and pathological track tantly for this discussion, this kind of (see below). Taken together, these dis- that arises on the background of causal observation is a major learning oppor- ease features militate against easy sol- mechanisms that are the same for all tunity (9) and could emerge as being a utions to the endotype question. disease cases. It is probably too soon to critical path to endotype definition (see be definitive on this aspect of type 1 below). FROM PHENOTYPES TO diabetes endotypes, and this important In sum, there are opportunities to ENDOTYPES IN TYPE 1 DIABETES: concept will require careful teasing apart conduct smarter clinical and laboratory A ROADMAP TO MAXIMIZING using cohort studies and a better knowl- studies. The risk of continuing with cur- OPPORTUNITIES edge of causality. For example, one can rent trends is an excess of nonproductive Several approaches are beginning to hypothesize a pathway in which toler- science,poor utilization of resources, and emerge that could assist in defining en- ance to (pro)insulin is breached early disillusion among our patients and con- dotypes, including greater accessibility to following presentation of (pro)insulin stituencies. tools for sophisticated human immuno- or related peptides by class II HLA mol- phenotyping; specific, targeted immune ecules on the HLA-DR4 haplotype, lead- LEARNING FROM OTHER DISEASES therapies; and the application of systems ing to T- and B-cell activation and One way to start integrating endotypes immunology and new statistical tools. autoantibody production; and tolerance into type 1 diabetes studies and appre- These offer opportunities that are based to GAD is similarly brokendperhaps at ciate their potential impact is to draw on 1) observation/hypothesis-driven ap- a slower pace or following different upon experiences in other complex dis- proaches, as well as 2) unsupervised/ precipitantsdby presentation of GAD eases in which they have been an im- data-driven methodologies and 3) re- peptides by HLA-DR3 haplotype–linked portant part of the development of sponse to therapy. In each case these molecules. In both situations, the un- precision medicine approaches. Asthma approaches will benefit from the oppor- derlying causative event could be shared is the prototypical example, in which the tunities that arise to study natural history (e.g., virus-mediated damage to islets) or delineation of a subset of patients whose cohorts such as TrialNet (11), The Envi- distinct (e.g., molecular mimicry for pro- airway disease is driven by type 2 cyto- ronmental Determinants of Diabetes in insulin or GAD), and in both there is a kines (the T2-high/low paradigm) led to the Young (TEDDY) (12), and INNODIA common pathogenesis involving pro- use of the term “endotyping” to describe (13) as well as responder/nonresponder gression to multiple autoantibodies, sig- subpopulations in which the underlying subgroups in clinical trials (8). nifying increased risk of disease, as well disease is caused by a uniform pathobio- as progression to diabetes (14). These logic or molecular mechanism (1). The Endotype Definition Led by processes could be termed the “proin- endotyping paradigm proved critical for Observations and Hypotheses sulin autoimmune-DR4” (PADR4) and advancing a new age of asthma medi- A clear recognition of the heterogeneous “GAD autoimmune-DR3” (GADR3) endo- cations and has invoked the use of a more traits present in type 1 diabetes has given types. Going forward, the field could stringent definition of the term “endo- rise to numerous examples of possible focus on defining the related but distinct type” that incorporates successful dis- pathophysiological processes that could pathophysiological pathways more pre- ease modification by a therapeutic agent be considered to be compatible with the cisely, as well as using these two markers that targets the putative pathobiological definition of an endotype, and focused (autoantibody and HLA) as stratifiers for mechanism. The essential requirements study addressing one or two of these any therapeutic that emerges as being for success of the endotype model in seems a reasonable place to start. With particularly efficacious in limiting or re- asthma were a robust understanding of this in mind, one of the more obvious versing specific autoantigen presenta- at least one pathobiological pathway, a examples relates to a phenotype (e.g., tion and loss of tolerance (15). One of therapeutic intervention that interdicts development of a specific islet cell au- the challenges in this context is that this, and a robust biomarker to identify toantibody) and its link to a genotype the reliable identification of these two the disease subtype (10). (e.g., HLA) that would strongly infer that endotypes currently requires sampling An obvious question, therefore, is, a distinct pathophysiological process is close to first seroconversion. Indeed, “Can we simply transfer these principles in operation. Birth cohort studies of findings from the Type 1 Diabetes Prediction care.diabetesjournals.org Battaglia and Associates 9

and Prevention (DIPP) study suggest that of an age-related physiological differ- assigned as present/absent across a scale perhaps only half of those testing pos- ence in immune responsiveness (e.g., (i.e., color shades) for each given subject. itive for GADA at diagnosis had GADA as B-lymphocyte number and percentage Given a sufficient number of subjects, the first detectable autoantibody, mak- are higher in the blood in young children there would be the potential to identify ing the case that better biomarkers of compared with later childhood) (19). subgroups of subjects whose disease is PADR4andGADR3willberequired(16). Age thus functions as a proxy for changes reflected in palettes with a similar color/ Asecondexampleisgivenbythedem- in immune and metabolic function. A shade composition. In addition, the pal- onstration that in the pancreas, two dis- much better understanding of the mat- ette colors could go beyond the mea- tinct types of insulitic lesions are present in uration of the key physiological systems surement of known traits but also include subjects with recent-onset type 1 diabetes, in childhood would undoubtedly help system approaches (such as immunom- distinguishable by the degree of cellular here, and perhaps endotypes in which ics by mass cytometry, whole blood infiltrate and presence of CD201 Bcells the pathobiology diverges from the phys- transcriptomics, metabolomics, and (termed hyperimmune CD20hi and pauci- iology would be of particular interest. proteomics) as in the design of the immune CD20lo) (17,18). This phenotype At the very least, mechanistic and dis- INNODIA consortium studies (13). There is carries important implications for endo- covery science studies aiming to uncover insufficient space here to dojustice to the type definition and treatment strategies. endotypes should be careful to select par- many studies that have defined po- For example, the hyperimmune CD20hi ticipants a priori (for example, according to tentially important immune and meta- status, which appears to be most overt age,sex,autoantibody status,andHLA)oras bolic phenotypes that could contribute in the younger age-group, could be re- bins according to these features post hoc. to complex endotype definitions in sponsive to B-cell depletion therapies. An emphasis on age relatedness is thus type 1 diabetes; therefore, as a means In both of these examples, age could an important part of the roadmap and to illustrate the palette as a potential be a confounding influence. For the will undoubtedly help yield clearer data part of the roadmap, several of the more putative PADR4 and GADR3 endotypes, and uncover the nature of physiology/ prominent examples are shown in Fig. 1, it will be important to examine the role pathobiology relationships and their proxies. along with a strategy for discovering how of age and whether this is a proxy for they could be used going forward. For different gene-environment interactions Data-Driven Endotype Discovery example, subjects with multiple dom- (e.g., diet, infection) or immunological Beyond these clear and somewhat binary inant phenotypes indicative of im- maturity. It is plausible that what ap- examples, one innovative approach that mune hyperresponsiveness (e.g., multiple pears as a pathobiological phenomenon could be adopted in type 1 diabetes for autoantibodies, high antigen-specific (e.g., a greater preponderance of B defining more complex endotypes is the T-cell proliferation, activated CD8 T cells) lymphocytes in islet immune infiltrates palette model, proposed by McCarthy will cluster together (Fig. 1). McCarthy in young children with type 1 diabetes) (20). The principle is that several selected describes several advantageous features is actually a reflection, at least in part, major traits (i.e., palette colors) are of this model, including: implicit acknowl- edgment of the multifactorial nature of type 1 diabetes, potential to reflect pro- gression rates and response to therapy, enablement of targeted therapies (e.g., for T cell, B cell, interferon), focusing of re- search efforts onto therapeutics and en- couraging identification of the extremes (“archetypes”), and the potential to iden- tify surrogates that are more facile to measure than multiple different pheno- types. Developing this model could be envisaged as a collaborative effort across the key type 1 diabetes research networks to achieve sufficient data points for clusters to appear.

Endotype Definition Led by Responders Versus Nonresponders Analysis Further insights into disease pathways that could lead us to endotypes follow a reversed discovery track; these are learn- ings from the study of clinical responses in the setting of intervention trials, in Figure 1—The palette model for defining endotypes. A series of characteristics are defined and which a therapeutic agent appears to be graded using immunoassays and the data analyzed for evidence of clustering to reveal complex most effective in a subgroup of patients. endotypes. Examplesforanti-CD3andoralinsulinare 10 An Approach to Heterogeneity in Type 1 Diabetes Diabetes Care Volume 43, January 2020

given above; further indications of such are pursued. A greater impact might be erogeneity and defining endotypes in type 1 theratypes include the analysis of the seen in the design of immunotherapy diabetes. The authors thank the following for effects of costimulation blockade on im- trials in the short-term and adoption of helpful discussion of the topics aired in this manuscript: Tee Bahnson (Benaroya Research mune compartments in the setting of disease-modifying therapies into clinical Institute), Susan Geyer (TrialNet Coordinating the TrialNet intervention study with the practice in the longer term. In trials, the Center [University of South Florida]), Georg costimulation blocking agent CTLA4-Ig clear definition of type 1 diabetes endo- Hollander (University of Oxford), and Edward (Abatacept) (21). A plasma-induced tran- types that associate with responsiveness Wakeland (University of Texas Southwestern Medical Center). scription assay showed that the patients to specific therapies could provide suffi- Funding. The Leona M. and Harry B. Helmsley exhibiting high innate inflammatory bias at ciently compelling early-phase outcome Charitable Trust sponsored the 2-day meeting baseline exhibited more rapid disease pro- data so that drugs make a faster transition that was held on this topic in January 2018. gression as well as a greater therapeutic to market and are explicitly earmarked for Duality of Interest. No potential conflicts of response to CTLA4-Ig (22). In another use in a disease subset. To arrive at these interest relevant to this article were reported. Author Contributions. The following authors study, a treatment-induced change in advances will take sustained, high-quality contributed to the idea conceptualization and the configuration of memory/na¨ıve com- research that must be conducted with 2-day meeting preparation, attended the meet- 1 partments of CD4 T lymphocytes was cognizance of the potential positive/ ing in person, and delineated the manuscript reported (23). These findings further sup- negative impact of heterogeneous traits backbone: M.B., S.A., M.A.A., T.M.B., C.E.-M., porttheexistenceofdiscreteendotypesof and phenotypes. Performing experiments C.J.G., P.A.G., M.J.H., J.P.K., S.A.L., E.F.M., N.G.M., R.A.O., T.P., M.C.P., X.Q., M.J.R., type 1 diabetes that exhibit distinct im- with human samples, and taking into B.O.R., D.Sc., D.Sk., and M.P. In addition, the munoregulatory profiles at clinical onset consideration the possibility that, for ex- following authors (being part of the TrialNet and that these may be useful for design ample, males and females have different Steering Committee) significantly contributed and analysis of clinical trials. immunological behavior depending on to the discussion and brainstorming that followed These examples, in addition to many age, hormonal status, BMI, and other the abovementioned meeting: M.S.A., D.B., P.J.B., E.B., L.A.D.M., S.E.G., K.C.H., M.K., L.J., M.L., A.P., others (24), provide support for a strat- factors, is likely to yield data of higher D.Sc., and D.Sk. Thus, all authors made substantial egy that is being increasingly adopted to quality, with lower variance, and thus contributions to conception and design of the understand drug mechanisms of action, make a more incisive contribution to manuscript, participated in drafting the manu- human physiology, and disease, namely, knowledge and understanding. If these script or revising it critically for important in- fi “ ” tellectual content, and gave nal approval of the use of experimental medicine studies codes of practice are widely adopted the version to be submitted. M.B. and M.P. wrote (for example, a drug or intervention is used and studies and clinics are conducted the manuscript. to examine hypothetical changes in the against a background of wide awareness immune system as the primary end point) of the endotype concept, then there is the References rather than clinical trials (efficacy or safety definite potential for significant advances 1. Lotvall¨ J, Akdis CA, Bacharier LB, et al. Asthma is the outcome). One could also contem- in practice to be made. endotypes: a new approach to classification of plate the use of combinations of therapies During an era that is unprecedented in disease entities within the asthma syndrome. J – (eachwithdistinctmechansimsofactions) the application of immune and biologic Allergy Clin Immunol 2011;127:355 360 2. GreenbaumC, VanBueckenD, LordS. Disease- across a diverse population to highlight therapies to disorders as diverse as can- modifying therapies in type 1 diabetes: a look drugs with distinct, subgroup-dependent cer, hypercholesterolemia, and psoriasis, into the future of diabetes practice. Drugs 2019; effects. type 1 diabetes remains an outlier in 79:43–61 terms of not having a disease-modifying 3. Ludvigsson J. Time to leave rigid traditions in CONCLUSIONS: MOVING FROM type 1 diabetes research. Immunotherapy 2017; therapy beyond single hormone re- 9:619–621 PHENOTYPES TO ENDOTYPES IN placement. This means that despite 4. Woittiez NJC, Roep BO. Impact of disease TYPE 1 DIABETES representing a major unmet need, it stands heterogeneity on treatment efficacy of immu- Ultimately, the considerable effort re- to miss out on the benefits of precision notherapy in type 1 diabetes: different shades of – quired to establish robust endotypes of medicine. One of the barriers to overcome gray. Immunotherapy 2015;7:163 174 fi 5. Agache I, Akdis CA. Precision medicine and type 1 diabetes must be justi ed in terms in order to address this current, parlous phenotypes, endotypes, genotypes, regiotypes, of its importance for, and impact upon, status is the impact of disease heteroge- and theratypes of allergic diseases. J Clin Invest clinical management, clinical trial design, neity. We propose that defining, under- 2019;130:1493–1503 and research studies on disease patho- standing, and applying disease endotypes 6. Sherry N, Hagopian W, Ludvigsson J, et al.; genesis. Examples of the bearing this in type 1 diabetes are steps that warrant Proteg´ e´ Trial Investigators. Teplizumab for treat- ment of type 1 diabetes (Proteg´ e´ study): 1-year might have are therefore worth consid- keen attention as we design new labora- results from a randomised, placebo-controlled ering. In relation to new patients being tory and clinical studies. A revised model trial. Lancet 2011;378:487–497 seen for the first time in the type 1 for disease investigation and management, 7. Herold KC, Bundy BN, Long SA, et al.; Type diabetes clinic, for example, the identi- entailing categorization of patients by bi- 1 Diabetes TrialNet Study Group. An anti-CD3 fication of endotypes with rapid and ology, should replace the “one size fits all” antibody, teplizumab, in relatives at risk for type 1 diabetes. N Engl J Med 2019;381:603–613 unrelenting progression to a state of approach and would be transformational. 8. Writing Committee for the Type 1 Diabetes minimal C-peptide secretion, as opposed TrialNet Oral Insulin Study Group, Krischer JP, Schatz to prolonged honeymooning with limited DA, Bundy B, Skyler JS, Greenbaum CJ. Effect of oral insulin requirement, could guide manage- insulinonpreventionofdiabetesinrelativesof Acknowledgments. This manuscript is the re- patients with type 1 diabetes: a randomized ment decisions such as pump adoption or sult of a 2-day meeting, held with the generous clinical trial. JAMA 2017;318:1891–1902 other advanced technologies and the in- support of the Leona M. and Harry B. Helmsley 9. Atkinson MA, Roep BO, Posgai A, Wheeler tensity with which education programs Charitable Trust, focused on understanding het- DCS, Peakman M. The challenge of modulating care.diabetesjournals.org Battaglia and Associates 11

b-cell autoimmunity in type 1 diabetes. Lancet Prevention Trial-Type 1 Study Group. A new 39. Bonifacio E, Beyerlein A, Hippich M, et al.; Diabetes Endocrinol 2019;7:52–64 approach for diagnosing type 1 diabetes in TEDDY Study Group. Genetic scores to stratify 10. Kuruvilla ME, Lee FE, Lee GB. Understanding autoantibody-positive individuals based on pre- risk of developing multiple islet autoantibodies asthma phenotypes, endotypes, and mecha- diction and natural history. Diabetes Care 2015; and type 1 diabetes: a prospective study in nisms of disease. Clin Rev Allergy Immunol 38:271–276 children. PLoS Med 2018;15:e1002548 2019;56:219–233 26. Redondo MJ, Steck AK, Sosenko J, et al.; Type 40. Redondo MJ, Geyer S, Steck AK, et al.; Type 11. Battaglia M, Anderson MS, Buckner JH, et al. 1 Diabetes TrialNet Study Group. Transcription 1 Diabetes TrialNet Study Group. A type 1 di- Understanding and preventing type 1 diabetes factor 7-like 2 (TCF7L2) gene polymorphism and abetes genetic risk score predicts progression of through the unique working model of TrialNet. progression from single to multiple autoantibody islet autoimmunity and development of type 1 Diabetologia 2017;60:2139–2147 positivity in individuals at risk for type 1 diabetes. diabetes in individuals at risk. Diabetes Care 12. Rewers M, Hyoty¨ H, Lernmark A,˚ et al.; Diabetes Care 2018;41:2480–2486 2018;41:1887–1894 TEDDY Study Group. The Environmental Deter- 27. Tosur M, Geyer SM, Rodriguez H, Libman I, 41. So M, Elso CM, Tresoldi E, et al. Proinsulin minants of Diabetes in the Young (TEDDY) study: Baidal DA, Redondo MJ; Type 1 Diabetes TrialNet C-peptide is an autoantigen in people with type 1 2018 update. Curr Diab Rep 2018;18:136 Study Group. Ethnic differences in progression diabetes. Proc Natl Acad Sci U S A 2018;115: 13. Mathieu C, Lahesmaa R, Bonifacio E, Achenbach of islet autoimmunity and type 1 diabetes in 10732–10737 P, Tree T. Immunological biomarkers for the de- relatives at risk. Diabetologia 2018;61:2043– 42. Arif S, Moore F, Marks K, et al. Peripheral and velopment and progression of type 1 diabetes. 2053 islet interleukin-17 pathway activation charac- Diabetologia 2018;61:2252–2258 28. Leete P, Mallone R, Richardson SJ, Sosenko terizes human autoimmune diabetes and pro- 14. Ziegler R, Alper CA, Awdeh ZL, et al. Specific JM, Redondo MJ, Evans-Molina C. The effect of motes cytokine-mediated b-cell death. Diabetes association of HLA-DR4 with increased preva- age on the progression and severity of type 1 2011;60:2112–2119 lence and level of insulin autoantibodies in first- diabetes: potential effects on disease mecha- 43. Ziegler AG, Rewers M, Simell O, et al. Se- degree relatives of patients with type I diabetes. nisms. Curr Diab Rep 2018;18:115 roconversion to multiple islet autoantibodies Diabetes 1991;40:709–714 29. Cousminer DL, Ahlqvist E, Mishra R, et al.; and risk of progression to diabetes in children. 15. Roep BO, Wheeler DCS, Peakman M. Anti- Bone Mineral Density in Childhood Study. First JAMA 2013;309:2473–2479 gen-based immune modulation therapy for genome-wide association study of latent auto- 44. Achenbach P, Lampasona V, Landherr U, type 1 diabetes: the era of precision medicine. immune diabetes in adults reveals novel insights et al. Autoantibodies to zinc transporter 8 and Lancet Diabetes Endocrinol 2019;7:65–74 linking immune and metabolic diabetes. Diabe- SLC30A8 genotype stratify type 1 diabetes risk. 16. Ilonen J, Lempainen J, Hammais A, et al.; tes Care 2018;41:2396–2403 Diabetologia 2009;52:1881–1888 Finnish Pediatric Diabetes Register. Primary islet 30. Kahaly GJ, Hansen MP. Type 1 diabetes 45. Achenbach P, Koczwara K, Knopff A, Naserke autoantibody at initial seroconversion and auto- associated autoimmunity. Autoimmun Rev 2016; H, Ziegler A-G, Bonifacio E. Mature high-affinity antibodies at diagnosis of type 1 diabetes as 15:644–648 immune responses to (pro)insulin anticipate the markers of disease heterogeneity. Pediatr Di- 31. Oram RA, Jones AG, Besser REJ, et al. The autoimmune cascade that leads to type 1 di- abetes 2018;19:284–292 majority of patients with long-duration type 1 abetes. J Clin Invest 2004;114:589–597 17. Arif S, Leete P, Nguyen V, et al. Blood and islet diabetes are insulin microsecretors and have 46. Ziegler A-G, Nepom GT. Prediction and path- phenotypes indicate immunological heterogeneity functioning beta cells. Diabetologia 2014;57: ogenesis in type 1 diabetes. Immunity 2010;32: in type 1 diabetes. Diabetes 2014;63:3835–3845 187–191 468–478 18. Leete P, Willcox A, Krogvold L, et al. Differ- 32. Greenbaum CJ, Beam CA, Boulware D, et al.; 47. Vecchio F, Lo Buono N, Stabilini A, et al.; ential insulitic profiles determine the extent of Type 1 Diabetes TrialNet Study Group. Fall in DRI_Biorepository Group; Type 1 Diabetes Trial- b-cell destruction and the age at onset of type 1 C-peptide during first 2 years from diagnosis: Net Study Group. Abnormal neutrophil signature diabetes. Diabetes 2016;65:1362–1369 evidence of at least two distinct phases from in the blood and pancreas of presymptomatic and 19. Tosato F, Bucciol G, Pantano G, et al. Lym- composite type 1 Diabetes TrialNet data. Di- symptomatic type 1 diabetes. JCI Insight 2018;3: phocytes subsets reference values in childhood. abetes 2012;61:2066–2073 122146 Cytometry A 2015;87:81–85 33. Rodriguez-Calvo T, Richardson SJ, Pugliese A. 48. Ferreira RC, Guo H, Coulson RMR, et al. A 20. McCarthy MI. Painting a new picture of Pancreas pathology during the natural history of type I interferon transcriptional signature pre- personalised medicine for diabetes. Diabetologia type 1 diabetes. Curr Diab Rep 2018;18:124 cedes autoimmunity in children genetically at risk 2017;60:793–799 34. Simell S, Hoppu S, Simell T, et al. Age at for type 1 diabetes. Diabetes 2014;63:2538– 21. Orban T, Bundy B, Becker DJ, et al.; Type development of type 1 diabetes– and celiac 2550 1 Diabetes TrialNet Abatacept Study Group. disease–associated antibodies and clinical dis- 49. Kallionpa¨a¨ H, Elo LL, Laajala E, et al. Innate Co-stimulation modulation with abatacept in ease in genetically susceptible children observed immune activity is detected prior to serocon- patients with recent-onset type 1 diabetes: from birth. Diabetes Care 2010;33:774–779 version in children with HLA-conferred type 1 a randomised, double-blind, placebo-controlled 35. Beyerlein A, Bonifacio E, Vehik K, et al.; diabetes susceptibility. Diabetes 2014;63:2402– trial. Lancet 2011;378:412–419 TEDDY Study Group. Progression from islet au- 2414 22. Cabrera SM, Engle S, Kaldunski M, et al.; Type toimmunity to clinical type 1 diabetes is influ- 50. Cabrera SM, Chen Y-G, Hagopian WA, 1 Diabetes TrialNet CTLA4-Ig (Abatacept) Study enced by genetic factors: results from the Hessner MJ. Blood-based signatures in type 1 Group. Innate immune activity as a predictor of prospective TEDDY study. J Med Genet 2019; diabetes. Diabetologia 2016;59:414–425 persistent insulin secretion and association with re- 56:602–605 51. Coppieters KT, Dotta F, Amirian N, et al. sponsiveness to CTLA4-Ig treatment in recent-onset 36. Krischer JP, Liu X, Lernmark A,˚ et al.; TEDDY Demonstration of islet-autoreactive CD8 T cells in type 1 diabetes. Diabetologia 2018;61:2356–2370 Study Group. The influence of type 1 diabetes insulitic lesions from recent onset and long-term 23. Orban T, Beam CA, Xu P, et al.; Type 1 Di- genetic susceptibility regions, age, sex, and fam- type 1 diabetes patients. J Exp Med 2012;209: abetes TrialNet Abatacept Study Group. Reduc- ily history on the progression from multiple 51–60 tion in CD4 central memory T-cell subset in autoantibodies to type 1 diabetes: a TEDDY study 52. Arif S, Tree TI, Astill TP, et al. Autoreactive costimulation modulator abatacept-treated pa- report. Diabetes 2017;66:3122–3129 T cell responses show proinflammatory polari- tients with recent-onset type 1 diabetes is as- 37. Pugliese A, Boulware D, Yu L, et al.; Type zation in diabetes but a regulatory phenotype in sociated with slower C-peptide decline. Diabetes 1 Diabetes TrialNet Study Group. HLA-DRB1*15: health. J Clin Invest 2004;113:451–463 2014;63:3449–3457 01-DQA1*01:02-DQB1*06:02 haplotype protects 53. Kenefeck R, Wang CJ, Kapadi T, et al. Fol- 24. Malmegrim KCR, de Azevedo JTC, Arruda autoantibody-positive relatives from type 1 dia- licular helper T cell signature in type 1 diabetes. J LCM, et al. Immunological balance is associated betes throughout the stages of disease progres- Clin Invest 2015;125:292–303 with clinical outcome after autologous hemato- sion. Diabetes 2016;65:1109–1119 54. Viisanen T, Ihantola E-L, Nant¨ o-Salonen¨ K, poietic stem cell transplantation in type 1 di- 38. Steck AK, Dong F, Waugh K, et al. Predictors et al. Circulating CXCR51PD-11ICOS1 follicular T abetes. Front Immunol 2017;8:167 of slow progression to diabetes in children with helper cells are increased close to the diagnosis of 25. Sosenko JM, Skyler JS, DiMeglio LA, et al.; multiple islet autoantibodies. J Autoimmun type 1 diabetes in children with multiple auto- Type 1 Diabetes TrialNet Study Group; Diabetes 2016;72:113–117 antibodies. Diabetes 2017;66:437–447 12 An Approach to Heterogeneity in Type 1 Diabetes Diabetes Care Volume 43, January 2020

55. Long SA, Cerosaletti K, Bollyky PL, et al. in individuals with long-standing type 1 diabetes. locus on chromosome 1 and multiple variants in Defects in IL-2R signaling contribute to dimin- Diabetes 2015;64:3891–3902 the MHC region for serum C-peptide in type 1 ished maintenance of FOXP3 expression in 58. Hull CM, Peakman M, Tree TIM. Regulatory diabetes. Diabetologia 2018;61:1098–1111 CD41CD251 regulatory T-cells of type 1 diabetic T cell dysfunction in type 1 diabetes: what’s 61. Fløyel T, Brorsson C, Nielsen LB, et al. CTSH subjects. Diabetes 2010;59:407–415 broken and how can we fix it? Diabetologia 2017; regulates b-cell function and disease progr- 56. Schwedhelm K, Thorpe J, Murray SA, et al. 60:1839–1850 ession in newly diagnosed type 1 diabetes Attenuated IL-2R signaling in CD4 memory 59. Yeo L, Woodwyk A, Sood S, et al. Autoreactive patients. Proc Natl Acad Sci U S A 2014;111: T cells of T1D subjects is intrinsic and dependent T effector memory differentiation mirrors b cell 10305–10310 on activation state. Clin Immunol 2017;181: function in type 1 diabetes. J Clin Invest 2018;128: 62. Redondo MJ, Geyer S, Steck AK, et al.; Type 67–74 3460–3474 1 Diabetes TrialNet Study Group. TCF7L2 ge- 57. Yang JHM, Cutler AJ, Ferreira RC, et al. 60. Roshandel D, Gubitosi-Klug R, Bull SB, netic variants contribute to phenotypic hetero- Natural variation in interleukin-2 sensitivity in- et al.; DCCT/EDIC Research Group. Meta- geneity of type 1 diabetes. Diabetes Care 2018; fluences regulatory T-cell frequency and function genome-wide association studies identify a 41:311–317