Understanding, and Reversing, Metabolic Memory Is Within Reach

Abstract

Commentary on: Al-Dabet MM, Shahzad K, Elwakiel A, et al. Reversal of the renal hyperglycemic memory in diabetic kidney disease by targeting sustained tubular p21 expression. Nat Commun. Aug 27 2022;13(1):5062. doi:10.1038/s41467-022-32477-9 Today, it seems self-evident that hyperglycemia is the causative agent for diabetic complications. However, this concept was firmly established only in 1990s with two seminal studies: DCCT (Diabetes Control and Complications Trial)1Diabetes C. Complications Trial Research G. Nathan D.M. et al.The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus.N Engl J Med. Sep 30. 1993; 329: 977-986https://doi.org/10.1056/NEJM199309303291401Crossref PubMed Scopus (22702) Google Scholar in type 1 and UKPDS (UK Prospective Diabetes Study) in type 2 diabetes.2Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). UK Prospective Diabetes Study (UKPDS) Group.Lancet. Sep 12 1998; 352: 837-853Abstract Full Text Full Text PDF PubMed Scopus (18951) Google Scholar The follow-on studies to these trials described the phenomenon of metabolic memory or the legacy effect: that past hyperglycemia leads to sustained increase in the risk of end-organ complications and death long after glucose control has been established.3Nathan D.M. Cleary P.A. Backlund J.Y. et al.Intensive diabetes treatment and cardiovascular disease in patients with type 1 diabetes.N Engl J Med. Dec 22 2005; 353: 2643-2653https://doi.org/10.1056/NEJMoa052187Crossref PubMed Scopus (4135) Google Scholar,4Holman R.R. Paul S.K. Bethel M.A. Matthews D.R. Neil H.A. 10-year follow-up of intensive glucose control in type 2 diabetes.N Engl J Med. Oct 9 2008; 359: 1577-1589https://doi.org/10.1056/NEJMoa0806470Crossref PubMed Scopus (5242) Google Scholar For example, a history of poor glycemic control (6.5-year-long in DCCT/EDIC) increased the risk for development and progression of diabetic kidney disease (DKD), even after 25 years of improved control.5Lachin J.M. Nathan D.M. Group D.E.R. Understanding Metabolic Memory: The Prolonged Influence of Glycemia During the Diabetes Control and Complications Trial (DCCT) on Future Risks of Complications During the Study of the Epidemiology of Diabetes Interventions and Complications (EDIC).Diabetes Care. Sep 21 2021; https://doi.org/10.2337/dc20-3097Crossref Scopus (12) Google Scholar Attention was soon directed to understanding the mechanisms of this memory and whether and how it could be erased. Metabolic memory is an example of environmental variables causing long-lasting changes in gene expression. Therefore, epigenetic alterations have been the focus of intense study as etiologic mechanisms. These modifications are largely environmentally induced changes in structure of DNA or DNA-associated proteins which influence gene expression by keeping the chromosome structure in ‘open’ or ‘closed’ conformation, thus determining whether DNA transcription machinery has access to the gene sequence. Epigenetic modifications include methylation of CpG islands in the DNA, post-translational modification of histones, microRNAs (miRNAs), non-coding RNAs (ncRNAs), long non-coding (lncRNAs) or a cooperative combination of these mechanisms.6Chen Z. Natarajan R. Epigenetic modifications in metabolic memory: What are the memories, and can we erase them?.Am J Physiol Cell Physiol. Aug 1 2022; 323: C570-C582https://doi.org/10.1152/ajpcell.00201.2022Crossref Scopus (3) Google Scholar DNA methylation, the most stable modification, is an interplay between the writer, eraser and reader proteins. Writers, such as DNA methyl transferases (DNMT 1, 3a/3b), are enzymes that methylate CpG dinucleotides; erasers, such as Ten Eleven Translocation (TET-) 1, 2 and 3 proteins oxidize and eventually demethylate cytosines.7Jones P.A. Functions of DNA methylation: islands, start sites, gene bodies and beyond.Nat Rev Genet. May 29 2012; 13: 484-492https://doi.org/10.1038/nrg3230Crossref PubMed Scopus (3803) Google Scholar It is in this context that Al-Dabet et al8Al-Dabet M.M. Shahzad K. Elwakiel A. et al.Reversal of the renal hyperglycemic memory in diabetic kidney disease by targeting sustained tubular p21 expression.Nat Commun. Aug 27 2022; 13: 5062https://doi.org/10.1038/s41467-022-32477-9Crossref Scopus (2) Google Scholar present their study in Nature Communications. Epigenetic modifications underpinning metabolic memory have been studied in kidney tissue from DKD animal models or cultured or urine-derived human kidney cells. Human peripheral blood mononuclear cells have also been used for these studies, with some of the results replicated in micro-dissected tubules from human kidney biopsies (previously reviewed6Chen Z. Natarajan R. Epigenetic modifications in metabolic memory: What are the memories, and can we erase them?.Am J Physiol Cell Physiol. Aug 1 2022; 323: C570-C582https://doi.org/10.1152/ajpcell.00201.2022Crossref Scopus (3) Google Scholar,9Aranyi T. Susztak K. Cytosine Methylation Studies in Patients with Diabetic Kidney Disease.Curr Diab Rep. Aug 30 2019; 19: 91https://doi.org/10.1007/s11892-019-1214-6Crossref Scopus (4) Google Scholar). Several studies from one group utilized kidney tissues from patients with or without DKD (reviewed in9Aranyi T. Susztak K. Cytosine Methylation Studies in Patients with Diabetic Kidney Disease.Curr Diab Rep. Aug 30 2019; 19: 91https://doi.org/10.1007/s11892-019-1214-6Crossref Scopus (4) Google Scholar). As a whole, these studies have identified a number of epigenetic modifications in genes involved in renal fibrosis, inflammation, oxidative stress, apoptosis, mitochondrial and endothelial dysfunction and other pathways implicated in DKD pathogenesis.6Chen Z. Natarajan R. Epigenetic modifications in metabolic memory: What are the memories, and can we erase them?.Am J Physiol Cell Physiol. Aug 1 2022; 323: C570-C582https://doi.org/10.1152/ajpcell.00201.2022Crossref Scopus (3) Google Scholar To translate these mechanisms to human DKD, several pieces of information are needed (Table 1). While no single manuscript can cover all these elements, successful translation to human DKD does require all of them.Table 1Data elements required for translation to Human DKDCausalitySatisfies Hill’s criteria for causality:•Temporality: modification occurs before DKD•Strength of association•Replication: association is found in multiple studies/groups•Biologic gradient: if feasible•Biologic plausibility, e.g., localization to kidney-specific enhancers•Consistency with other knowledge: correlation with gene expression•Robust study design: e.g., ascertainment of cases and controls•ReversibilitySpecies specificity9Aranyi T. Susztak K. Cytosine Methylation Studies in Patients with Diabetic Kidney Disease.Curr Diab Rep. Aug 30 2019; 19: 91https://doi.org/10.1007/s11892-019-1214-6Crossref Scopus (4) Google ScholarIs identified or replicated in human kidney tissueCell type specificityThe kidney cell type(s) in which the modification occurs is determined.Condition specificityThe modification is shown to occur in DKD Open table in a new tab The authors8Al-Dabet M.M. Shahzad K. Elwakiel A. et al.Reversal of the renal hyperglycemic memory in diabetic kidney disease by targeting sustained tubular p21 expression.Nat Commun. Aug 27 2022; 13: 5062https://doi.org/10.1038/s41467-022-32477-9Crossref Scopus (2) Google Scholar use a combination of tools (cell lines, animal models, human tissue and biofluid) to make several important contributions: •They begin by characterizing histologic and molecular features of memory-associated injury patterns in experimental DKD, defining them as features that are induced by hyperglycemia and not reversed by subsequent glucose control. For example, they show that fibrosis displays a memory-associated injury pattern, as does the compilation of histologic and protein expression features identified as senescence in proximal tubules. •The authors then devise a simple and clever study design to identify cyclin-dependent kinase inhibitor, p21 (Cdkn1a), as a top contender for association with memory phenotype: in a streptozotocin-induced diabetic mouse model, they look for all genes whose mRNA expression in the kidneys is altered by hyperglycemia AND whose alteration is not reversed by subsequent glucose control. They identify p21 as one of the most highly altered of these mRNAs which follow this memory expression pattern. They then confirm the memory-pattern expression of p21 in several other DKD animal models and ascertain the tubular epithelium as the cell-type where this mechanism operates. •They proceed to flesh out the pathway from hyperglycemia to p21 induction to kidney injury, showing that: •Hyperglycemia reduces DNA methyl transferase 1 expression, thereby hypomethylating p21 promoter and increasing its expression. •Increase in p21 expression releases senescence-associated secretory proteins, promoting tubular injury and fibrosis. Suppression of p21 expression rescues this phenotype. •The authors also show increased expression of p21 as well as proteins associated with senescence in kidney tubules from people with diabetic kidney disease. They further find increased p21 protein in cross-sectional urine samples from people with DKD, as well as those with non-diabetic CKD. In addition, they find that urine p21 is higher in the diabetic participants of the LIFE-Adult cohort (a population-based cohort study in Germany),10Loeffler M. Engel C. Ahnert P. et al.The LIFE-Adult-Study: objectives and design of a population-based cohort study with 10,000 deeply phenotyped adults in Germany.BMC Public Health. Jul 22 2015; 15: 691https://doi.org/10.1186/s12889-015-1983-zCrossref PubMed Scopus (215) Google Scholar and it is roughly proportional to their KDIGO CKD progression risk. •Of special interest to translation, they fully examine pathway reversibility: activated Protein C (aPC), together with glucose control, increases DNA methyl transferase 1 expression, which re-methylates, and suppresses, p21 expression, reversing tubular injury, senescence, fibrosis, and albuminuria: •aPC augments DNA methyl transferase 1 expression and rescues kidney injury in wild type mice, made diabetic with streptozotocin; •aPC-deficient (Thrombomodulin pro/pro mutant) mice, made diabetic with streptozotocin, have higher p21 and worse kidney injury and •knocking out p21 in the above aPC-deficient mice reverses kidney injury.8Al-Dabet M.M. Shahzad K. Elwakiel A. et al.Reversal of the renal hyperglycemic memory in diabetic kidney disease by targeting sustained tubular p21 expression.Nat Commun. Aug 27 2022; 13: 5062https://doi.org/10.1038/s41467-022-32477-9Crossref Scopus (2) Google Scholar Given the depth and breadth of this work,8Al-Dabet M.M. Shahzad K. Elwakiel A. et al.Reversal of the renal hyperglycemic memory in diabetic kidney disease by targeting sustained tubular p21 expression.Nat Commun. Aug 27 2022; 13: 5062https://doi.org/10.1038/s41467-022-32477-9Crossref Scopus (2) Google Scholar it seems more appropriate to discuss remaining questions, rather than limitations: •In DKD animal models, diabetes alone induces p21 expression and tubular injury; in human kidneys, p21 expression is increased in DKD, but not in diabetes without kidney disease.8 In addition, CKD increases p21 in human kidneys, in absence of diabetes. Therefore, diabetes is neither necessary, nor sufficient for p21 upregulation in human kidneys.8 How, then, is this pathway regulated in human DKD or CKD? Given these differences between mice and humans, translation to human DKD requires elucidation of how the pathway operates in human kidneys. •How does a history of systemic hyperglycemia specifically suppress DNA methyl transferase 1 expression in a tissue- (e.g., kidney) and cell-type (tubular)-restricted manner? How does the effect of DNA methyl transferase 1 suppression remain confined to specific DNA sequences (e.g., p21 promoter)? •Another interesting finding is the presence of p21 in human urine samples, its correlation with GFR and albuminuria and its association with the risk of DKD progression. Urine p21 correlates with eGFR when latter is <55-60 mL/min/1.73m2, but not higher. This pattern is interestingly reminiscent of p21 increase in people with reduced GFR (DKD and CKD), but not those with diabetes and intact GFR. Similarly, urine p21 correlation with the urine-albumin-to-creatinine ratio (UACR) appears bimodal; when UACR is <500 mg/g, p21 has little-no correlation with it; with UACR>500 mg/g, urine p21 does rise to ∼200-600 pg/mL range, but still has no correlation with UACR within that range. Because of the strong correlations between eGFR and UACR, and the correlation of most individual urine proteins and the total urine protein, regressing p21 on UACR and eGFR may provide a clearer picture of their associations. A fully adjusted regression may also explain why p21 is not linearly rising from low to very high KDIGO risk subgroups in the LIFE-adult cohort. It is also worth asking whether urine p21 correlates with kidney p21 and what urine p21 levels are in people with diabetes but not DKD. The study by Al-Dabet et al8Al-Dabet M.M. Shahzad K. Elwakiel A. et al.Reversal of the renal hyperglycemic memory in diabetic kidney disease by targeting sustained tubular p21 expression.Nat Commun. Aug 27 2022; 13: 5062https://doi.org/10.1038/s41467-022-32477-9Crossref Scopus (2) Google Scholar goes beyond identification of the epigenetic modification to meticulously, and comprehensively, elucidate the pathways, from a history of hyperglycemia to tubular injury. Additionally, as outlined above, this study takes a significant, and unique, step towards translation: they identify a mechanism for reversing injurious effects of metabolic memory by showing that addition of aPC to glucose control enables a reversal of injury via suppression of p21 expression. Scoring another point for clinical translation, the authors further dissect the precise mechanisms by which aPC suppresses p21 and the injury caused by metabolic memory. aPC sits at the helm of a major two-pronged anticoagulant and cytoprotective system. The anticoagulant arm begins by Protein C (PC) binding to thrombomodulin and thrombin and generation of activated Protein C (aPC), which restrains coagulation by removing thrombin and inactivating factors Va and VIIIa. On the other hand, the aPC-initiated, cytoprotective arm controls inflammatory response and endothelial cell apoptosis, stabilizes endothelial and epithelial barriers and promotes regenerative capacity in a cell- and context-specific manner.11Griffin J.H. Zlokovic B.V. Mosnier L.O. Activated protein C: biased for translation.Blood. May 7 2015; 125: 2898-2907https://doi.org/10.1182/blood-2015-02-355974Crossref PubMed Scopus (169) Google Scholar This arm involves aPC binding to endothelial protein C receptor (EPCR) and signaling through and/or activating several receptors (e.g., Protrase-Activated Receptor [PAR]1, PAR2, PAR3, etc.) along the way.11Griffin J.H. Zlokovic B.V. Mosnier L.O. Activated protein C: biased for translation.Blood. May 7 2015; 125: 2898-2907https://doi.org/10.1182/blood-2015-02-355974Crossref PubMed Scopus (169) Google Scholar The reality is likely even more complex, with extensive cross-talk between these two signaling arms. For example, PC activation depends on, and is likely regulated by, binding to thrombomodulin and thrombin, i.e., components of the anticoagulant signaling arm. On the flip side, binding to EPCR augments aPC anticoagulant activity;11Griffin J.H. Zlokovic B.V. Mosnier L.O. Activated protein C: biased for translation.Blood. May 7 2015; 125: 2898-2907https://doi.org/10.1182/blood-2015-02-355974Crossref PubMed Scopus (169) Google Scholar also PAR-1 does double duty in coagulation (in myocardial infarction) and cytoprotective signaling (in sepsis), depending on the tissue and clinical condition.12Aisiku O. Peters C.G. De Ceunynck K. et al.Parmodulins inhibit thrombus formation without inducing endothelial injury caused by vorapaxar.Blood. Mar 19. 2015; 125: 1976-1985https://doi.org/10.1182/blood-2014-09-599910Crossref PubMed Scopus (62) Google Scholar Using an aPC mutant with significantly diminished anticoagulant signaling (3K3A-APC) and parmodulin-2, a biased PAR-1 signal modulator,12Aisiku O. Peters C.G. De Ceunynck K. et al.Parmodulins inhibit thrombus formation without inducing endothelial injury caused by vorapaxar.Blood. Mar 19. 2015; 125: 1976-1985https://doi.org/10.1182/blood-2014-09-599910Crossref PubMed Scopus (62) Google Scholar the authors conclude that the aPC effect in kidney tubules is largely mediated via cytoprotective mechanisms (e.g., through PAR-1). Identifying which aPC-originated pathway leads to inhibition of metabolic memory is important because on the face of it, aPC seems a superb candidate for erasing metabolic memory, especially as it has been previously developed as a pharmaceutical. However, its broad pleiotropic effects on coagulation and cytoprotection mean that targeting it is fraught with the risk of ‘off-target’ adverse events. In fact, aPC has been evaluated in prior clinical trials in sepsis, as well as ischemic stroke. However, it was withdrawn from the market for safety concerns and lack of significant clinical effect. One reason may be that its beneficial cytoprotective effects were derailed by aPC-induced off-target increases in serious bleeding.13Bernard G.R. Vincent J.L. Laterre P.F. et al.Efficacy and safety of recombinant human activated protein C for severe sepsis.N Engl J Med. Mar 8. 2001; 344: 699-709https://doi.org/10.1056/NEJM200103083441001Crossref PubMed Scopus (5055) Google Scholar, 14Kalil A.C. LaRosa S.P. Effectiveness and safety of drotrecogin alfa (activated) for severe sepsis: a meta-analysis and metaregression.Lancet Infect Dis. Sep. 2012; 12: 678-686https://doi.org/10.1016/S1473-3099(12)70157-3Abstract Full Text Full Text PDF PubMed Scopus (68) Google Scholar, 15Christiaans S.C. Wagener B.M. Esmon C.T. Pittet J.F. Protein C and acute inflammation: a clinical and biological perspective.Am J Physiol Lung Cell Mol Physiol. Oct 1 2013; 305: L455-L466https://doi.org/10.1152/ajplung.00093.2013Crossref PubMed Scopus (48) Google Scholar As a result, knowing that cytoprotective aPC effects are largely responsible for its opposition to metabolic memory allows us to focus therapeutic strategies on aPC mutants with preferential cytoprotective effects. Metabolic memory adds a significant, and often unseen, risk factor to those typically managed by nephrologists, one that impacts both risk assessment and medical management. As regards risk factors, nephrologists are well acquainted with hyperglycemia as a risk factor for DKD development and progression (though more data exists on the former than latter). Equally important is the fact that past hyperglycemia is an independent risk factor (and more serious for being hidden) for DKD progression, the magnitude and duration of which must be estimated and considered in DKD management. This is not a trivial task given the discontinuous healthcare structure in the US and the widespread healthcare disparities, especially in populations with high prevalence of diabetes and DKD. Also relevant to risk stratification, glycemic control early in the diabetes course has a more profound impact on metabolic memory than later in the disease, additionally highlighting the importance of early detection of diabetes and institution of intensive glucose control.16Lind M. Imberg H. Coleman R.L. Nerman O. Holman R.R. Historical HbA1c Values May Explain the Type 2 Diabetes Legacy Effect: UKPDS 88.Diabetes Care. Jul 7 2021; https://doi.org/10.2337/dc20-2439Crossref Scopus (22) Google Scholar On the management front, as we learn more about the mechanisms of metabolic memory and develop specific therapies, a new class of drugs will emerge. To correctly position, and use, these agents within the DKD toolset, we need to recognize that management of the risk posed by metabolic memory requires both the control of current hyperglycemia as well as use of specific interventions to reverse the damages caused by the history of hyperglycemia. Support: None. Financial Disclosure: The author declares that she has no relevant financial interests. Peer Review: Received October 30, 2022 in response to an invitation from the journal. Direct editorial input from an Associate Editor and a Deputy Editor. Accepted in revised form December 21, 2022.