SYNDROMES OF KETOSIS-PRONE DIABETES

Abstract

Ketosis-prone diabetes (KPD) is a heterogeneous condition characterized by patients who present with diabetic ketoacidosis but lack the phenotype of autoimmune type 1 diabetes. Here I review progress in our understanding of KPD and its place in the expanding universe of “atypical diabetes.” I focus on investigations of our collaborative research group at Baylor College of Medicine and the University of Washington using a longitudinally followed, heterogeneous, multiethnic cohort of KPD patients. We have identified clinically and pathophysiologically distinct KPD subgroups, separable by the presence or absence of islet autoimmunity and the presence or absence of beta cell functional reserve. The resulting “Aß” classification of KPD accurately predicts long-term glycemic control and insulin dependence. I describe key characteristics of the KPD subgroups, their natural histories, and our investigations into their immunologic, genetic, and metabolic etiologies. These studies serve as a paradigm for the investigation of atypical forms of diabetes.

INTRODUCTION

Traditionally, patients with diabetes are classified as having either autoimmune type 1 diabetes (T1D) or obesity-associated type 2 diabetes (T2D). In recent decades, the worldwide epidemic of diabetes has revealed a range of distinct phenotypes that defy assignment to one of these binary types of diabetes. One strikingly variant form of diabetes that has emerged is manifested by presentation with diabetic ketoacidosis (DKA) of patients lacking the classic phenotype of autoimmune T1D. This has led to the recognition of an array of heterogenous syndromes of ketosis-prone diabetes (KPD) (1).

DKA IN ATYPICAL PATIENTS

Beginning in the 1960s, reports from West Africa described small numbers of patients with reversible forms of diabetes (2,3) who appeared to shift from a T1D phenotype (insulin-dependent) to a T2D (non-insulin–dependent) phenotype. These included patients who presented with ketosis and initially required insulin therapy but were later found to have a complete remission of diabetes (4). Through the 1980s and 1990s, larger and more heterogeneous groups of atypical patients presenting with DKA were reported, including obese African-American children lacking islet cell autoantibodies and high-risk human leukocyte antigen (HLA) alleles (5), overweight adult Afro-Caribbean patients who had clinical characteristics of T2D (termed “Flatbush diabetes”) (6), and obese African-Americans with late-onset T2D (7,8). Consistent features of the patients in these reports included a significant improvement in endogenous insulin secretion (i.e., recovery of beta cell function) over time, and the ability to dispense with insulin therapy weeks to months after the episode of DKA.

Since the turn of the millennium — especially following the establishment of a standardized classification scheme (see below) — the number of such reports has multiplied and expanded the incidence of KPD and KPD-like syndromes worldwide to include patients from a wide range of geographic regions and ethnic backgrounds, including: Japanese (9); US Hispanics, Caucasians, and Native Americans (10-13); Peruvians (14); European Caucasians (15); Pakistanis (16); Asian Indians (17); Chinese (18); Koreans (19); and Thais (20).

In the early 2000s, three investigative groups tracked longitudinal cohorts of patients with KPD and attempted to classify them into distinct subgroups based on clinical characteristics and presumed pathophysiological differences. In addition to the taxonomic and heuristic usefulness of these efforts, a major motivation was to predict long-term clinical behavior, in particular the need (or lack thereof) for lifelong insulin treatment. Our collaborative group at Baylor College of Medicine and the University of Washington (UW) initiated a longitudinal, prospective study of multiethnic patients with a range of phenotypes of KPD in Houston, Texas, and introduced a classification scheme based on two criteria: islet autoantibodies and beta cell functional reserve (21). Mauvais-Jarvis et al. (22) published a retrospective 10-year follow-up study of KPD in immigrants from sub-Saharan Africa living in Paris, and classified autoantibody-negative KPD patients according to insulin requirement. Ramos-Roman et al. (23) reported outcomes in DKA patients and atypical clinical courses compared with typical patients with T1D and T2D in Dallas, Texas.

COMPREHENSIVE CHARACTERIZATION OF PATIENTS PRESENTING WITH DKA: THE Aß SUBGROUPS

Our collaborative Baylor-UW research team has identified, classified, and tracked an ever-expanding cohort of KPD patients since 1999. This cohort, located in Houston, Texas, now numbers close to 900 patients who are identified at the time of presentation to the hospital with DKA, have consented to be followed in a dedicated KPD clinic and to have their data and samples reposited in a secure, longitudinal database and registry (1,11). Enrolled patients have tests of islet autoimmunity (autoantibodies directed against the 65 kDa glutamic acid decarboxylase [GAD65Ab], zinc transporter T8 [ZnT8Ab], or islet antigen-2 [IA2Ab]; and T1D HLA class II T1D susceptibility alleles in a subset) and beta cell function (fasting and glucagon-stimulated C-peptide levels) at the time of the initial clinic visit following hospitalization for the index episode of DKA); the latter are repeated at 6- to 12-month intervals. A protocol based on the results of these two tests and longitudinal clinical behavior is followed to determine if, how and when insulin therapy may be withdrawn. Based upon an unbiased multivariate analysis of prospective data thus collected over 12 months in the first ∼100 patients in the registry, rigorous cutoffs for autoantibody positivity or negativity, and of beta cell functional reserve (present or absent) were established, and the Aß classification scheme was developed (11). This 2 × 2 factorial construct defines four KPD subgroups:

• 1)A+ß- KPD: defined by presence of autoantibodies and absence of beta cell function;
• 2)A+ß+ KPD: defined by presence of autoantibodies with preserved beta cell functional reserve;
• 3)A-ß- KPD: defined by absence of autoantibodies with absence of beta cell function; and
• 4)A-ß+ KPD: defined by absence of autoantibodies with preserved beta cell functional reserve.

In the diverse ethnic population of our hospital district (∼ 40% Hispanic, 30% African-American, 20% Caucasian, and 10% Asian and other), the approximate frequencies of these subgroups are: A-ß+ KPD (50%), A-ß- KPD (20%), A+ß- KPD (20%), and A+ß+ KPD (10%) (11).

Patients with the β- forms of KPD are diagnosed with diabetes at a young age, lean, and completely insulin dependent (11). Less than 1% of patients classified initially as β- manifest any improvement in beta cell function over time. A+β- KPD is synonymous with autoimmune T1D. The demographic and clinical features of A-β- KPD are similar to those of A+β- KPD, but these patients lack islet autoantibodies and have a lower frequency of HLA T1D risk alleles and a strong family history of diabetes (11).

A-β+ KPD patients resemble T2D patients in most clinical characteristics, but present with DKA. They have substantial beta cell functional reserve following recovery from the index DKA episode, and this reserve improves progressively over time (11). Insulin therapy can be withdrawn for the majority of these patients within 4 to 8 months, and they remain insulin-independent with excellent glycemic control on oral medications for a median of 4 years of follow-up. The natural history of A-β+ KPD–patients reveals two clinical trajectories: approximately half have sustained excellent beta cell function for many years with good glycemic control, and the other half, after a brief period of improvement and insulin independence, show progressive decline in beta cell function, worsening glycemic control, and relapse to insulin requirement (24). The critical baseline characteristic that predicts these divergent outcomes is the association of a clinically significant stressful event (e.g., infection, myocardial infarction, and trauma) with the index episode of DKA. Those with an identifiable stressor (denoted “provoked” A-β+ KPD) lose beta cell function over time, whereas those who developed DKA without an associated stressful event (denoted “unprovoked” A-β+ KPD) preserve beta cell function and remain insulin-independent for years. The distinctive syndrome of unprovoked A-β+ KPD is characterized by late onset, male predominance, obesity, DKA at initial diagnosis of diabetes, and lack of HLA class II T1D susceptibility alleles or T cell reactivity to islet autoantigens (24,25). In contrast, provoked A-β+ KPD has no sex predilection and a history of T2D; many patients in this subgroup have “occult” islet autoimmunity manifested as T cell reactivity to islet autoantigens (24,26).

A+β+ KPD patients resemble T2D patients clinically, but have evidence of islet autoimmunity and usually present with DKA at diagnosis of diabetes. They have a variant autoimmune process compared to patients with A+β- KPD or autoimmune T1D, with a predilection for epitope-specific GAD65-Ab (27) and co-express both high-risk and protective T1D HLA class II alleles (11). Approximately half of this subgroup exhibit progressive worsening of beta cell function following the index DKA and need insulin therapy, whereas the remainder can discontinue insulin for prolonged periods of time after the index DKA episode (1,11).

VALIDITY OF THE Aß CLASSIFICATION OF KPD

To standardize the taxonomy of KPD, we compared the four extant KPD classification systems, using as an index of accuracy the ability to predict long-term beta cell functional reserve, essential for the clinically important outcomes of glycemic control and insulin dependence, in a multiethnic cohort of 294 KPD patients with long-term follow-up (28). Multiple statistical analyses revealed that the Aß system is the most accurate, providing an area under the receiver operator characteristic curve of 0.972 (28). The classification of diabetes and nomenclature recommended by the American Diabetes Association yielded the lowest accuracy in predicting clinical outcomes among KPD patients. We and a growing number of investigators worldwide use Aß classification for its value in predicting glycemic outcomes and insulin requirement shortly after the index episode of DKA, and as a reliable foundation for genetic and mechanistic studies of the etiology and pathophysiology of KPD.

KPD CLASSIFICATION AS A CATALYST FOR ETIOLOGIC DISCOVERY

I present here a summary of studies by the Baylor-UW group to discover etiologic and pathophysiologic mechanisms underlying the KPD syndromes. An important theme common to all these investigations, is the ability of the Aß classification scheme to define stable phenotypes, and a disease registry, with long-term follow-up to permit comparative studies of large numbers of patients belonging to different KPD subgroups.

A-ß- KPD: Discovery of Novel Monogenic and Occult Autoimmune Forms of Diabetes

The strong family history (usually in multiple generations) of relatively early onset diabetes suggested that the disease in many A-ß- KPD patients might be due to a familial trait. Following the examples of monogenic diabetes characterized by the Maturity-Onset Diabetes of Youth syndromes, we hypothesized that variants in genes required for beta cell development, regeneration, or function could constitute an important etiologic basis for this subgroup. A pilot investigation showed that 26% of A-β- KPD patients have an increased frequency (compared with ethnic-specific population controls) of potentially pathogenic variants in HNF-1α and PDX-1, encoding critical transcription factors (hepatocyte nuclear factor-1α and pancreas-duodenum homeobox-1) for beta cell development (29). This has led to detailed investigations of novel mechanisms leading to a developmental block in beta cell development and a defective insulin secretion using induced pluripotential stem cell technology on skin fibroblasts obtained from these patients (Yang, Borowiak, Patel, Balasubramanyam, October 2018). Thus, the ability to accurately circumscribe the A-ß- KPD phenotype greatly enhances the success rate of identifying monogenic or oligogenic causes of beta cell dysfunction in diabetes. The possibility also exists that A-ß- KPD patients may be erroneously labeled “A-” because of a hitherto unknown or untested islet autoantibody or because they have cellular autoimmunity directed at the pancreatic islets without a corresponding circulating marker of humoral immunity. Applying a test of cellular autoimmunity (T cell reactivity directed against islet autoantigens) to a subset of A-ß- KPD patients has shown that some have an “occult” form of islet autoimmunity mediated by cellular rather than humoral mechanisms (26).

A+ß+ KPD: Discovery of Variant Forms of Autoimmunity in Diabetes

The clinical phenotypes and natural histories of the two A+ KPD subgroups differ substantially as described above. Our comparative investigations of autoimmune markers have revealed immune network differences that underlie the later-onset and more slowly progressive nature of beta cell dysfunction in A+ß+ compared to A+ß- KPD. We first evaluated the autoimmune response expressed by autoantibodies directed toward specific epitopes of the GAD65 and discovered significant associations between GAD65Ab specific for the epitope DPD, presence of beta cell functional reserve, and the A+ß+ phenotype (27). Subsequently, we evaluated KPD patients belonging to all four subgroups for anti-idiotypic (i.e., “masked”) GAD65Ab(DPD). We discovered that KPD patients with preserved beta cell functional reserve (the ß+ subgroups) had a significantly higher frequency of masked GAD65Ab(DPD) than KPD patients without beta cell functional reserve (ß^- subgroups) (30). Masked GAD65Ab(DPD) were also more frequent among A+ß+ KPD than among A+ß- KPD. Patients bearing HLA haplotypes associated with a high risk of islet autoimmunity were very likely to lack both overt and masked GAD65Ab(DPD) (30). These findings indicate a role for GAD65Ab epitope-specificity, and for anti-idiotypic antibodies to GAD65 in defining the clinical phenotypes of autoantibody-positive KPD. A+ß+ KPD patients also have elevated levels of circulating unmethylated and methylated insulin DNA, markers respectively of ongoing beta cell destruction and inflammation, compared to all other KPD subgroups (31). This strengthens the concept that the pathophysiology of islet dysfunction in A+ß+ KPD involves a slow, prolonged decline in mass or function, rather than a rapid destruction as in autoimmune T1D.

A-ß+ KPD: Distinctive Subtypes and Pathophysiologic Mechanisms

The natural history of the syndrome of provoked A-ß+ KPD, characterized by initial improvement in beta cell function following the index episode of DKA and later, inexorable decline to insulin dependence, raised the possibility that these patients might have a progressive form of occult islet autoimmunity in a manner similar to some patients with A-ß- KPD. Approximately one-third of provoked A-ß+ KPD patients display T cell reactivity to islet antigens despite lacking the classic islet autoantibodies of autoimmune diabetes (26).

We have used a range of investigative methods to understand the basis of the most unique and complex syndrome of KPD, that of unprovoked A-ß+ KPD. Despite the striking clinical phenotype, these patients offer no tantalizing autoimmune or genetic clues to their pathophysiology; hence, we used unbiased comparative metabolomics analyses of stored serum samples to discover leads. We found evidence for potential defects in two metabolite families — both involved defects in amino acid metabolism, a surprising finding for a disease conventionally associated mainly with defects in carbohydrate or lipid metabolism. One pathway implicated branch chain amino acid metabolism, and the other implicated the arginine/citrulline/glutamine metabolic cycle (32). We coupled these static metabolite measurements with kinetic validation, using stable isotope infusions and mass spectrometric analysis. We identified a distinctive pathogenic sequence of impaired ketone oxidation and fatty acid utilization for energy, together with accelerated catabolism of the ketogenic amino acid leucine and associated impairment of tricarboxylic acid cycle anaplerosis (32) — these might collectively explain the predilection to develop ketosis in these otherwise T2D-like patients. In regard to the finding of defects in the processing of arginine (an insulin secretagogue), we tested the hypothesis that during sustained hyperglycemia, KPD patients experience diminished intracellular arginine availability due to increased arginine hydrolysis, and this deficit is associated with impaired insulin secretion. We found that these KPD patients actually have increased intracellular arginine availability in the euglycemic state, but this reserve is severely compromised during hyperglycemia, resulting in an inadequate arginine “kick” to sustain insulin secretion during hyperglycemic crises (33). A clinically notable corollary to this finding is that exogenous arginine administration restores a virtually normal insulin response (33).

KPD IN THE SPECTRUM OF ATYPICAL DIABETES

It is useful to conclude this review by placing these advances related to syndromes of KPD in the larger, emerging concept of atypical diabetes. This concept enfolds not only rare monogenic diabetes syndromes but also clusters of phenotypically distinct forms of diabetes that span a spectrum between two poorly defined poles of T1D and T2D. Because of a lack of etiological information, some of these unique and variant forms of diabetes are currently subsumed under the broad category of T1D by virtue of certain characteristics (e.g., early onset, insulin requirement, proneness to ketosis, and lean body mass index) while some are grouped under the broad category of T2D by virtue of other characteristics (e.g., later onset, response to oral agents, metabolic syndrome, high body mass index). It is essential to develop processes to distinguish the full range of atypical forms from the current, indistinct polar categories, and thereby uncover their natural histories and pathogenic mechanisms. A tremendous gap exists in our ability to categorize these atypical characteristics into specific diabetes endotypes and to develop scientifically and clinically relevant classification schemes for them. A comprehensive approach to define clusters and syndromes of atypical diabetes within the diabetes spectrum will facilitate discoveries of novel etiologic pathways and uncover mechanisms and therapeutic targets that can improve our understanding of the traditional poles of T1D and T2D. Lessons learned from our identification, classification, and investigations of syndromes of KPD should provide a paradigm and a valuable catalyst in this important effort.

DISCUSSION

Reiser, Chicago: Diabetes type 1.5, where would that fit in…?

Balasubramanyam, Houston: So, we try to avoid terms like that. The point is that you can come up with any kind of terminology for it, but the question is what does it mean? People have used terms like 1.5 or type-3, but I think that what we really need is an etiologically accurate classification of diabetes. I think that's what we're trying to get at.

Reiser, Chicago: On your magnetic spectrum slide — the last one you showed — if you look into let's say development of complications like CKD, for example, where do you see spikes in your band?

Balasubramanyam, Houston: We haven't tracked people sufficiently long to find that out, but that's really the point. The other way to try to approach this — which has been very effective — is to use machine learning and clustering analysis. In just the last 3 years there have been three reports — actually two reports: one from our group about to come out — using a whole lot of data including complications, biochemical data, and clinical data, to use algorithms to see how they cluster in terms of phenotype and behavior. The group from Scandinavia showed that one of the very distinct clusters of type-2 diabetes actually has the hallmark of rapid progression to kidney disease. The group from the Broad and the MGH has overlaid that with genomic analysis as well, and found yet other clusters. We have done that in an atypical group and also found a number of clusters. Some of them have to do with long term behavior. So, I think there's exciting times ahead for that.