Introduction

A decline in nicotinamide adenine dinucleotide (NAD+) metabolism is widely discussed as one of the biochemical features of aging. NAD+ is required for redox reactions, but it also serves as a substrate for enzymes involved in stress responses, deoxyribonucleic acid (DNA) repair, mitochondrial homeostasis, inflammation, and chromatin regulation. This has made NAD+ biology a major focus in aging research. Nicotinamide mononucleotide (NMN) and nicotinamide riboside (NR) are often described as NAD+ precursors with the potential to improve healthspan-related phenotypes in preclinical models. The present study does not provide new animal or wet-laboratory data; it is a secondary reanalysis of an existing public microarray dataset. Reviews by Imai et al., Verdin, and Yaku et al. frame NAD+ decline as a broad aging mechanism, but they also make clear that NAD+ biology is distributed across many pathways rather than confined to a single molecular axis [1-5].

The study by Mills et al. [6] remains one of the central preclinical datasets for long-term NMN administration during normal aging. In that study, wild-type C57BL/6N mice received oral NMN at 300 mg/kg/day during aging. The authors reported suppression of age-associated body weight gain, improved energy metabolism, enhanced insulin sensitivity, improved plasma lipid profiles, better eye function, and attenuation of age-associated gene expression changes in metabolic tissues, with particular emphasis on skeletal-muscle mitochondrial oxidative metabolism. The public microarray dataset GSE85718 captures this design across skeletal muscle, liver, and white adipose tissue, with age, treatment, and tissue all represented.

That distinction matters. A simple comparison between old control mice and old NMN-treated mice can identify genes that differ at old age, but it does not establish that NMN altered the age-related trajectory. A young-age NMN effect can also be misleading, because it may reflect treatment biology that has little to do with aging. The appropriate statistical question is whether the age-associated slope differs by treatment. Thus, the design tests whether the rate of age-related transcript change differs under NMN rather than merely testing expression differences at a single time point.

If a gene rises with age in control mice and the interaction term is negative, or if a gene falls with age in control mice and the interaction term is positive, the result is consistent with NMN shifting old-age expression toward the young-control state. In this article, an "NMN-rescue candidate" refers to a gene meeting this sign-reversal pattern together with nominal evidence for both the control aging effect and the age-by-treatment interaction. For example, a gene that falls with age in control mice but has a positive interaction coefficient would be considered a candidate for NMN-associated attenuation of that age-related decline. The broader concept of age-associated transcriptional drift has also been described in other biological systems, although its magnitude and direction can vary by tissue and cell type [7].

This approach also addresses several problems that limit translation in the NAD+ precursor field. First, mechanistic models are often too simple. A common narrative is that NAD+ increases sirtuin activity, activates peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α)-like programs, and thereby improves mitochondria. That model is useful, but it is unlikely to explain all tissue-specific effects of NMN. Second, tissue heterogeneity remains poorly resolved. Skeletal muscle, liver, and adipose tissue differ sharply in function, metabolic load, cell composition, and transcriptional response to aging. A single universal transcriptional rescue program would therefore be biologically surprising. Third, the field lacks reliable molecular biomarkers of NMN response. If NAD+ precursors are to move toward precision use, pharmacodynamic markers will be needed to show that the intervention engages relevant biology in a given tissue or patient group. Finally, it remains difficult to link transcriptional changes to functional outcomes such as insulin sensitivity, lipid handling, mitochondrial oxidation, and cellular homeostasis.

The present analysis was designed with those limitations in mind. It uses GSE85718 to test whether NMN modifies age-associated gene expression changes in metabolic tissues. It then asks whether any rescue candidates are robust across tissues and whether those candidates point to plausible mechanisms. The working hypothesis was that NMN would not reverse a single shared transcriptional aging program in all tissues. Instead, it was expected to produce tissue-dependent rescue signatures that converge on metabolic function. Based on the downstream results, the strongest mechanistic leads were RAB11A-linked vesicular trafficking and CPT2-linked fatty acid oxidation.