To be continued

Beyond Glucose and body Weight: Standards of Care Revaluation and A Citrate-Centred Model of Cellular Metabolic Failure in Type 2 Diabetes

4.5 Branch 1 — lipogenesis, diacylglycerol and the insulin receptor substrate  Cytosolic acetyl-CoA is carboxylated to malonyl-CoA by acetyl-CoA carboxylase. Malonyl-CoA both feeds de novo lipogenesis and inhibits carnitine palmitoyltransferase-1, restricting mitochondrial fatty acid entry — so the same signal that builds lipid also prevents its disposal. The downstream lipid intermediate of pathogenic interest is sn-1,2-diacylglycerol (DAG).  The human evidence for DAG-mediated insulin resistance is among the strongest mechanistic data in the field. Lipid infusion and hyperglycaemic clamp studies in humans show that acute muscle insulin resistance is accompanied by a transient rise in intramyocellular DAG that is temporally coupled to activation of protein kinase C-θ and increased inhibitory serine phosphorylation of insulin receptor substrate-1, with consequent impairment of PI3-kinase activation and GLUT4 translocation [47]. The broader framework — that ectopic lipid species, acting through specific kinases rather than through "lipotoxicity" in the abstract, are the proximate cause of insulin resistance in muscle and liver — has been developed and defended in detail [48,49]. Note that this branch is entirely compatible with the adiposity-centred guideline: weight loss works, on this account, because it relieves DAG accumulation in insulin-responsive tissue. 

 4.6 Branch 2 — histone acetylation and the senescence-associated secretory phenotype  The second branch is the newest and, in our view, the most consequential. Cellular senescence contributes to tissue dysfunction largely through the SASP, the inflammatory secretome of senescent cells. Until recently the dominant model of SASP initiation was cytosolic mitochondrial nucleic acid sensing through cGAS–STING.  In 2026, Martini, Birch, Passos and colleagues reported a second, metabolic layer of control. In senescent cells the mitochondrial pyruvate–citrate–acetyl-CoA axis is upregulated, increasing acetyl-CoA availability to support histone acetylation at SASP genes; while mitochondrial DNA-driven signalling activates inflammatory transcription factors, acetyl-CoA availability is required for robust transcription of SASP genes [50]. Raising acetyl-CoA promoted SASP gene expression, whereas inhibition of SLC25A1 reduced histone acetylation at SASP loci and limited the programme. In vivo, SLC25A1 inhibition reduced chromatin accessibility at SASP loci, dampened inflammation and improved healthspan in aged mice. Aged stromal cells from heart and kidney showed increased p16 and SASP gene expression that was selectively reduced by the SLC25A1 inhibitor CTPI2 without altering p16 expression, and treated aged animals showed reduced frailty on a 31-parameter index [50].  Three features of this result matter for the present hypothesis. First, it identifies precisely the metabolite and transporter that our proposed upstream lesion would overproduce. Second, the intervention suppressed the inflammatory output of senescent cells without abolishing the cell-cycle arrest that constitutes their tumour-suppressive function — a selectivity that senolytic strategies do not offer. Third, the phenotypic benefit was systemic, affecting frailty and muscle function.  The relevance to diabetes is direct. β-cell senescence accumulates with age and insulin resistance and contributes causally to secretory failure; clearance of senescent β cells improves glucose homeostasis in mice [51,52]. Adipose tissue senescence contributes to systemic insulin resistance [53]. β-cell dedifferentiation, rather than death alone, accounts for a substantial share of lost secretory capacity [54]. If a single metabolic node licenses the inflammatory output of senescent cells across islet, adipose, renal and vascular tissue, it is a plausible common denominator for the multi-organ benefit observed when energy surplus is relieved. 

 4.7 The feed-forward loop  The two branches are not parallel. DAG-mediated insulin resistance raises glucose and free fatty acid flux, increasing substrate delivery and further raising NADH/NAD⁺. SASP cytokines — tumour necrosis factor-α, interleukin-6 — activate NF-κB, which in turn modulates expression of both SLC25A1 and ACLY while impairing expression and assembly of respiratory chain components. Released Feα from oxidised clusters propagates local Fenton chemistry. The result, if the hypothesis is correct, is a self-reinforcing cycle with aconitase inactivation at its centre, in which each output increases the input that produced it.  

4.8 Supporting pharmacological evidence and counter-evidence  Supporting. Pharmacological inhibition of SLC25A1 with CTPI-2 has been reported to reverse steatosis, glucose intolerance and inflammation in preclinical models of fatty liver disease [55] — the same target, in a metabolic rather than an ageing model, producing the metabolic phenotype the hypothesis predicts. Elevated circulating citrate and altered TCA intermediates have been reported in insulin-resistant states in metabolomic studies, alongside the better-replicated branched-chain amino acid signature [56,57].  Directionality is unproven in humans. Mitochondrial dysfunction in insulin-resistant muscle is well documented [58], but whether it is cause, consequence or parallel phenomenon has been contested for two decades and is not settled by this hypothesis. Aconitase inactivation is reversible. The cluster is repaired by frataxin-dependent mechanisms [40]. A transient, reversible lesion becoming a chronic disease driver requires a reason why repair fails — plausibly chronic substrate surplus and iron dysregulation, but this is assumption, not data. ACLY inhibition in humans has not produced the expected glycaemic signal. Bempedoic acid inhibits ACLY and reduced cardiovascular events in CLEAR Outcomes without deterioration in glycaemic control [59]. If cytosolic acetyl-CoA overflow were the dominant driver of insulin resistance, one might have expected a larger metabolic effect. Tissue-selective activation and the hepatic specificity of the prodrug are possible explanations, but the observation constrains the hypothesis. Antioxidant trials have repeatedly failed. Broad antioxidant supplementation has not reduced cardiovascular or diabetes outcomes [60]. This is an argument against undifferentiated "ROS reduction" as therapy, and an argument for node-specific intervention — but it should temper enthusiasm.

4.9 Falsifiable predictions

Here we list the possible pro predictions as  follows. Reduced ACO2 activity and elevated citrate-to-isocitrate ratio in muscle, adipose or islet tissue from people with insulin resistance, relative to BMI-matched insulin-sensitive controls. Elevated tissue and circulating citrate preceding, not following, the onset of insulin resistance in longitudinal cohorts. Co-variation of SASP markers with citrate/acetyl-CoA flux measures within individuals, independent of BMI. Normalisation of ACO2 activity and citrate ratios after weight loss of a magnitude sufficient to induce remission, with degree of normalisation predicting durability of remission. In humans, an effect of SLC25A1 or ACLY modulation on histone acetylation at SASP loci in accessible cells. Stable-isotope tracer methods, ³¹P and ¹³C magnetic resonance spectroscopy, and single-nucleus multiple profiling of biopsy tissue are all adequate to test these predictions with current technology.  

5. Part III — What "cellular metabolic assessment" would require 

5.1 The case for a stratification layer  Current guideline decisions rest on phenotypic and population-statistical inputs: BMI, HbA1c, estimated glomerular filtration rate, albuminuria, established comorbidity. These are excellent risk descriptors and poor mechanism descriptors. Two patients with identical BMI and HbA1c may differ completely in residual β-cell reserve, oxidative capacity and inflammatory burden — and, plausibly, in what will work for them.  The proposal here is modest in structure and ambitious in content: add a measurement layer that describes the state of the target cells before and during therapy, in the way that lipid subfractionation, cardiac imaging and tumour molecular profiling were each added to their respective fields. The precedent within diabetes already exists — cluster-based subtyping [31], the 1-hour post-load glucose threshold now endorsed as a superior early risk marker [61], and continuous glucose monitoring metrics such as time-in-range that describe dynamics rather than averages. 

 5.2 A candidate five-domain panel, with honest grading    Three caveats apply to the entire panel and should be stated in any clinical protocol. Plasma concentrations of TCA intermediates are determined by renal handling, bone turnover and inter-organ exchange as much as by intracellular flux; a static concentration is a poor surrogate for a rate. Several markers lack accredited clinical assays and reference intervals. And — decisively — no randomised trial has yet shown that treatment guided by any of these markers improves patient outcomes. Until such a trial exists, this panel is a research instrument.

5.3 A stratification proposal

 If the panel were validated, a four-group stratification would follow naturally. We present it as a hypothesis-generating framework, explicitly additive to guideline-directed therapy.    The evidence base for the nutritional components differs by strength. Total energy restriction sufficient for 10–15 kg weight loss has randomised evidence for remission [25]. Mediterranean dietary patterns have randomised evidence for cardiovascular prevention and diabetes incidence [66,67]. Resistance training has consistent evidence for glycaemic control and muscle preservation, and is of particular importance during incretin therapy given documented lean-mass loss [26]. Time-restricted eating shows modest, inconsistent benefits largely attributable to energy reduction [68]. Micronutrient and mitochondrial cofactor strategies — coenzyme Q10, thiamine derivatives, α-lipoic acid, NAD⁺ precursors — remain unproven for metabolic endpoints; nicotinamide riboside, despite strong preclinical rationale, has not improved insulin sensitivity in randomised trials in humans [69,70]. Proprietary dietary systems marketed as "metabolic repair" should be held to the same standard as any drug: randomised trials with hard endpoints, pre-registered.  

5.4 An organisational proposal, and its problems  The manuscript from which this review derives proposes that large hospitals establish a department of cellular metabolic medicine, functioning as an upstream assessment layer for endocrinology, cardiology and nephrology rather than as a competitor to them. The intuition is defensible: specialties are organised by organ, while the lesion under discussion is organised by cell type and is shared across organs. Existing mitochondrial medicine clinics, lipid clinics and cardiometabolic clinics are partial precedents.  The obstacles are equally real and should be named: no reimbursement pathway exists for most of the proposed assays; the workforce would need to combine clinical biochemistry, endocrinology and geriatrics; and, most importantly, a service should not be built before the tests it would deliver have demonstrated clinical utility. The defensible near-term version is not a department but a research programme embedded in existing diabetes centres — one that collects the panel prospectively alongside standard care and tests whether it predicts anything that BMI, HbA1c and C-peptide do not.  

6. Incentives, evidence generation and who pays for repair  Any proposal that substitutes durable repair for continuous pharmacotherapy encounters a structural problem that is economic rather than scientific, and it is worth stating without conspiracy.  Therapies that must be taken indefinitely generate recurring revenue; therapies that cure shrink their own market. The canonical illustration is hepatitis C: direct-acting antivirals achieved cure rates above 95% and produced peak annual revenues near USD 19 billion for their developer in 2015, after which sales declined steeply as the prevalent pool was treated — a commercial outcome that is the direct consequence of therapeutic success [71]. The contrast with the incretin class is stark: the global GLP-1 market was estimated at roughly USD 70 billion in 2025 and is projected to exceed USD 85 billion in 2026, sustained by indefinite administration in a growing population [72].  It is a description of where capital flows under standard return-on-investment logic, and it has a specific and important consequence for the questions raised in this paper: the trials that would test whether cellular repair reduces or eliminates the need for chronic pharmacotherapy are precisely the trials that no commercial sponsor has an incentive to fund. That gap is a legitimate target for public and philanthropic research funding, for value-based and outcomes-linked contracting, and for the kind of investigator-initiated trial that produced DiRECT and Look AHEAD.  It also cuts the other way, and the review would be dishonest if it did not say so. Incretin therapies have demonstrated mortality, cardiovascular, renal and hepatic benefit in large randomised trials [8,9,19–24]. Nutritional and "repair" strategies have demonstrated remission in one well-conducted trial [25] and remain otherwise unproven at the level of hard outcomes. Economic critique of the incumbent does not confer evidence on the challenger.  

7. Conclusion  The history of the ADA Standards of Care is a history of the field correcting its own model when trials contradicted it: glucose alone proved insufficient, then outcomes reorganised the algorithm, then adiposity became the master lever. Each correction improved care substantially and left a residue of patients whom the new model does not describe.  We have proposed that a single node — oxidative inactivation of mitochondrial aconitase, with consequent citrate export and cytosolic acetyl-CoA overflow — could link substrate surplus to both insulin resistance and inflammaging, and that the 2026 demonstration of acetyl-CoA-licensed SASP transcription gives this proposal an experimental anchor it previously lacked. If the model is correct, the next stratification layer in diabetes care will be cellular rather than anthropometric, and it will sit upstream of pharmacotherapy rather than replacing it.  If the model is wrong, the panel proposed here will fail to add information beyond BMI, HbA1c and C-peptide, and the predictions in section 4.9 will not be observed. Both outcomes are informative, and both are testable now. The purpose of medicine is not to abolish disease but to treat each patient more precisely than we could yesterday; the question worth asking of any new framework is simply whether it makes that possible.  

Statements

Conflicts of interest. The author declares no financial conflicts of interest relating to this work. Author's contributions. The authors BX and HJ both conceived the hypothesis, reviewed the literature and wrote the manuscript. Disclaimer. This article is a narrative review and hypothesis. It does not constitute clinical guidance. None of the assessments or interventions discussed in Part III is validated for clinical decision-making, and none should be used to defer, modify or withhold guideline-directed diabetes therapy.  

支持单位：