The 48 Genes that Shape Aging: A Deep Exploration of Horvath’s Universal Mammalian Aging System

I had AI do a quick summary of a very comprehensive deep dive study of Horvath’s 48 aging related genes from the first preprint of his seminal paper Universal DNA methylation age across mammalian tissues -Nature Aging August 2023- The deep dive will be available in my upcoming book on the subject
here’s what it gave us:

What follows is an overview of Stephen Horvath’s Universal Mammalian Epigenetic Aging system. This updated review:

Clarifies that Thymine DNA Glycosylase (TDG), not TET enzymes, is the primary mechanism preventing hypermethylation of these aging-related genes (TDG is α-ketoglutarate dependent).
Explains that the initial 48 genes come from Horvath’s first preprint, and subsequent revisions have added or changed several genes (including transcription factor SP1).
Highlights how SP1 ties together MAO-A/MAO-B, FAD sequestration, WRN protein expression, and a potential impact on aging processes.
Presents a CD38/NAD+ analysis of the 48 genes, discussing how some of them may influence CD38 activity, thereby modulating NAD+ levels.

Throughout, we underscore the interplay of GABA–α-KG–glutamate, the overrepresentation of splicing-related genes, and the newly emphasized roles of SP1 and MAO in driving epigenetic and metabolic shifts that contribute to aging.

1. Horvath’s Universal Mammalian Epigenetic Aging System

Stephen Horvath’s DNA methylation “clock” revolutionized our understanding of biological aging. In his first preprint on the Universal Mammalian Epigenetic Aging system, he highlighted 48 genes whose methylation changes reliably tracked chronological age across mammalian species. These included transcription factors (e.g., REST, NANOG, c-JUN), splicing regulators (e.g., CELF4, CELF6, SON), and developmental patterning genes (e.g., HOXA13, PAX2).

Shifting Gene Sets in Later Revisions

In subsequent revisions of his preprint(s), some genes were replaced, others added, and additional ones were flagged as potentially relevant to aging. By the third revision, Horvath introduced about 10 new genes, including the transcription factor SP1. This evolving list underscores that the search for universal aging markers is ongoing—and that subtle changes in data thresholds can reveal new candidate genes.

SP1 stands out because it appears to:

Bind to the promoter region of MAO-A and MAO-B (both FAD-dependent enzymes).
Influence the production of WRN protein during puberty.
Potentially affect FAD usage in mitochondria and hence ATP production, echoing the way that rising CD38 activity depletes NAD+.

Thus, SP1 emerges as a candidate that links sex-related gene networks (MAO, WRN) to mitochondrial cofactor depletion (FAD, NAD+), reinforcing multiple aging pathways in non-neural tissues.

2. Base Excision Repair (TDG)—Not TET—Drives Gene Demethylation

A central theme in mainstream epigenetics is that TET enzymes (Ten-Eleven Translocation methylcytosine dioxygenases) demethylate DNA. However, Horvath’s 48 aging genes appear to rely more directly on thymine DNA glycosylase (TDG) for demethylation:

TDG functions in the base excision repair (BER) pathway, removing mismatched or modified bases and requiring α-ketoglutarate (α-KG) as a cofactor.
If α-KG levels drop with age (due to declining TCA cycle flux, GABA depletion, etc.), TDG-based demethylation may falter. This can cause hypermethylation and consequent silencing of protective “youth genes.”

Hence, TET enzymes may not be major players in maintaining these 48 genes’ youthful expression states; TDG is the more likely driver. This challenges the mainstream assumption that TETs are the principal route of demethylation in aging.

3. The Original 48 Genes and Their Associations

Below is the full list of Horvath’s original 48 genes from the first preprint, along with key functional or disease-related terms. We will later examine their connection to CD38 and NAD+ metabolism.

less

1.   LHFPL4  — Stroke, GABA #1, Neurogenesis #4

2.CELF6 — Sex M/F, Testosterone, BER/TDG, Splicing, Diabetes, Inflammation, Alzheimer’s, Alopecia

3.LHFPL3 — Male, GABA, Cancer #3, Glaucoma, Aconitase #1

4.EVX2 — Limb Development, Cancer (Zero references: extremely rare hits)

5.   FOXB1   — Glutamate

6.   PRDM13  — Puberty, Alzheimer’s, Hypertension

7.   ZIC5    — Glycolysis

8.   CELF4   — Sex M/F, Splicing, Glutamate, Diabetes, Obesity, Hypertension

9.   LARP1   — Female, mTOR, Glycolysis, Cancer #6

10.  ZIC4    — Stroke, GABA, Glutamate, Aconitase #8

11.  DLX6-AS1— Female, Apoptosis, Cancer #1, mTOR, Glycolysis

12.  OTP     — Autism, Obesity, LH, Estradiol, DHT, Testosterone, Parkinson’s, Female, mTOR, Calcification, ATM, Necrosis, Cancer #7, Diabetes, Inflammation

13.  DBX1    — Female, LH, GABA, Glutamate, Neurogenesis #5, Ca2+ #1

14.  SON (aliases) — Vitamin D, FSH, Menopause, Splicing, Aconitase #3

15.  IRX1    — Scoliosis, Neural Development, Lung & Kidney, Cancer, Retina, Rheumatoid Arthritis, Heart Disease

16.  NRN1    — Male, Alzheimer’s, mTOR, Stroke, ATM, AKT, ERK, Glutamate, Necrosis, Apoptosis, Diabetes #2, Glaucoma, Ca2+ #2

17.  GRIK2   — GluK2 (Kainate Receptor), Rare/Autism, Sex M/F, ATM, GABA, Glutamate, Obesity, Alzheimer’s

18.  POU3F2  — (?) Splicing?, Female, SIRT1, Neural Development

19.  SNX1    — Parkinson’s, Diabetes #4, Hypertension, Aconitase #2

20.  NR2E1   — (?) Splicing?, SIRT1, Neurogenesis #2, Diabetes, Inflammation, Aconitase #9

21.  TLX3    — Estradiol, Cancer #8

22.  VSX2 (?)— Female, Glutamate, Neural Development, Glaucoma

23.  TBX18   — Pacemaker Cells, Heart Development, Ca2+ #5

24.  OTX1    — FSH, GABA

25.  TCF12   — T-cells, Cancer, Cranial Structure, Muscle

26.  REST    — Obesity, Alopecia #2, Ca2+ #8, Aconitase #5

27.  SALL1   — FSH, Menopause

28.  ZIC1    — Female, FSH, Progeria, Mitochondria, SIRT1, Neural Dev, Obesity, Osteoporosis

29.  ZIC2    — Female, Ca2+

30.  SIX2    — Female, Hypertension, Krebs Cycle #1

31.  HOXA13  — Female, Menopause, Osteoporosis #1

32.  NEUROG2 — Female, Alzheimer’s, Neurogenesis #1, Neural Development #2

33.  EGR3 (?)— Estradiol, ERK, SIRT1, GABA, ESCs, Necrosis, Inflammation, Succinate #1, Ca2+ #4

34.  FOXD3   — FSH, Mitochondria, WRN, ESCs, Necrosis, Apoptosis, Pyruvate #5, Aconitase #6

35.  OBI1-AS1— CCR4–NOT, Glaucoma, Hypertension, Mitochondria, SIRT1, Necrosis, Cancer, Apoptosis, Diabetes, Inflammation, Succinate #4, Ca2+

36.  PHOX2B  — pH #1, Hypoventilation, Neuroblastoma, Parkinson’s/Dopamine, Crohn’s, Neuron Differentiation, Depression, Islet B-cells, Pyruvate #8

37.  NEUROD1 — Splicing Defects → Melatonin, FSH #1, Diabetes, Mitochondria, WRN, Krebs, Neurogenesis, Alzheimer’s, Pyruvate #2

38.  PAX2    — Female, Progesterone, Estradiol, Pyruvate #9

39.  PAX5 (?)— CD38, ATM, WRN, Cancer #5, Osteoporosis, Alopecia, Succinate

40.  HDAC2   — Inflammation, Progeria, ATM, Mitochondria, WRN, SIRT1, Krebs, Necrosis, Apoptosis, Alzheimer’s, Alopecia, SIRT2

41.  TWIST1  — Female, Cancer #2, DHEA, Calcification, AKT, Osteoporosis, Alopecia, SIRT3, Sclerostin, Succinate

42.  CTCF    — Vitamin D, α-KG, Diabetes #3, Succinate #6, Aconitase #10

43.  TF (Transferrin)— Macrophages, FSH, Mitochondria, Krebs, Diabetes, Succinate #2, Pyruvate #3

44.  NKX2    — Atrial Fibrillation, WRN, ESCs, Necrosis, Diabetes #5, SIRT2, Succinate #3, Ca2+ #6

45.  PRC2    — SIRT1, Cancer #4, Osteoporosis, Alopecia, SIRT2, SIRT3, Succinate

46.  NANOG   — (a surrogate Yamanaka Factor) Female, Progeria, FSH, Mitochondria, WRN, ESCs, Alopecia, SIRT2, SIRT3, Pyruvate #4

47.  BDNF    — Male, Depression w/ SIRT2 & SIRT3, Succinate #7, Ca2+ #3, Pyruvate #7, Aconitase #7

48.  c-JUN   — Male, Apoptosis, Inflammation, Macrophage, Atherosclerosis, Mitochondria, AKT, WRN, ERK, Necrosis, Cancer, Diabetes, Osteoporosis, Alopecia, SIRT2, SIRT3, Succinate #5, Ca2+ #7, Pyruvate #6, Aconitase #4

(Note: Some genes are marked “?” for potential involvement in RNA splicing or discovered references.)

4. SP1, MAO Enzymes, and WRN: A New Pivot in the 3rd Revision

Among the 10 newly introduced genes in Horvath’s 3rd revision, SP1 has garnered attention. Research suggests SP1:

Activates the promoter of MAO-A and MAO-B.
Influences WRN expression during puberty—a key life phase when robust DNA repair is crucial.
MAO-A/MAO-B are FAD-dependent enzymes that rise in many tissues with aging. Excess MAO can sequester FAD, reducing the pool available for complex II and other mitochondrial enzymes. This scenario parallels how CD38 can sequester NAD+, contributing to energy deficits.
WRN protein is essential for genome stability; its deficiency leads to Werner syndrome (a segmental progeria).

Thus, SP1 may create a cross-talk between sex hormones, mitochondrial cofactors (FAD, NAD+), and progeroid gene networks (WRN). Overactivity of MAO in aging tissues further drains FAD, akin to how elevated CD38 drains NAD+—both diminishing ATP production in mitochondria and aggravating aging.

5. GABA–Glutamate–α-KG Axis and Splicing Overrepresentation (Brief Recap)

From prior discussions:

Low GABA + high glutamate fosters excitotoxicity and dedifferentiation, partly by silencing REST in non-neural tissues.
α-KG is crucial for TDG-based demethylation. When α-KG drops, you get hypermethylation and gene silencing.
A remarkable ~31% of these 48 genes relate to mRNA splicing, whereas genome-wide, only 1.6% of genes are splicing-related. This 19.5-fold overrepresentation hints that splicing errors could be central in epigenetic aging.

6. Detailed Analysis: How These 48 Genes May Influence CD38 and NAD+ Levels

CD38 is a multifunctional enzyme that hydrolyzes NAD+ into nicotinamide and ADP-ribose, thereby depleting NAD+ pools critical for sirtuins (SIRT1, SIRT2, SIRT3) and other vital pathways. Sustained overexpression of CD38 in aging tissues correlates with NAD+ decline, mitochondrial dysfunction, and systemic inflammation.

We now examine how the 48 genes might intersect with CD38 function or NAD+ homeostasis:

PAX5 (Gene #39): Notably lists CD38 among its associations. PAX5 is critical in B-cell lineage commitment; B-cells are major CD38-expressing immune cells. Dysregulation can upregulate CD38, accelerating NAD+ consumption and fueling immunosenescence or chronic inflammation.
c-JUN (#48): c-JUN modulates inflammatory cytokines and can promote a feed-forward loop of immune activation. Chronic inflammation can upregulate CD38 in macrophages and other immune cells.
SIRT1, SIRT2, SIRT3 references: Many genes (e.g., POU3F2, NR2E1, HDAC2, PRC2, NANOG, BDNF) are connected to or regulated by sirtuins. When CD38 is high, NAD+ depletion reduces sirtuin activity, indirectly affecting the expression or stability of these genes.
MAO-A/MAO-B (via new gene SP1): Although not in the original 48, SP1’s role in MAO induction can parallel CD38 in draining an essential cofactor (FAD vs. NAD+). Double cofactor depletion (FAD and NAD+) drastically compromises mitochondria.
WRN (e.g., FOXD3 #34, NKX2 #44, c-JUN #48 all mention WRN): Proper WRN function depends partly on robust DNA repair capacity, which can be hampered by low NAD+ (sirtuin inactivation) and inadequate TDG activity. The synergy of high CD38 and high MAO could push cells toward a progeroid state.

While not every one of these 48 genes directly targets CD38, many influence or respond to pathways that modulate NAD+ usage, from immune activation (PAX5, c-JUN) to splicing (HDAC2 can repress beneficial NAD+-protective genes, etc.).

In summary, CD38 emerges as a major NAD+ consumer in aging, and multiple genes in this set either feed into or are downstream of the pathways that upregulate CD38 or worsen NAD+ depletion.

7. LARP1, Lamin A Truncation, and the Evolutionary Perspective

A significant subplot involves short LARP1 driving splicing errors in ATM, which then leads to truncated lamin A (progerin). Evolution appears to use this splicing-based “cell sacrifice” mechanism for normal fetal processes (e.g., ductus arteriosus closure). In advanced age, the same mechanism reappears pathologically.

The GABA–α-KG–splicing synergy, combined with CD38-driven NAD+ depletion, all converge on a scenario where cells lose their capacity to maintain correct gene expression, ironically replaying a developmental cell-death program that now triggers tissue dysfunction.

8. Conclusions and Future Directions

TDG (Not TET): Contrary to the mainstream focus on TET enzymes, these 48 genes likely rely on TDG (an α-KG-dependent base excision repair enzyme) to maintain their youthful, demethylated states.
SP1 and MAO: The third revision of Horvath’s preprint reveals new players like SP1, bridging sex-hormone pathways, MAO-driven FAD depletion, and WRN expression. This broadens the mechanistic tapestry of how epigenetic aging can be accelerated by cofactor exhaustion (FAD and NAD+).
CD38 and NAD+: Several genes (especially PAX5 and c-JUN), or their pathways, can upregulate CD38 in immune cells, leading to chronic NAD+ depletion. This further undermines sirtuin function, fosters inflammation, and amplifies splicing/repair failures.
Overlapping Mechanisms: GABA deficiency, α-KG shortage, CD38-mediated NAD+ depletion, and splicing errors appear interlinked. Correcting one might not suffice unless the entire metabolic-epigenetic network is balanced.

Potential Interventions

Restoring α-KG: Could rejuvenate TDG-based demethylation, preventing aberrant silencing of protective genes.
CD38 Inhibitors: May preserve NAD+ pools, supporting sirtuin activity and normal splicing.
MAO Blockade: Prevents excessive FAD depletion, potentially synergizing with NAD+ restoration.
Short LARP1 Antagonists: Could reduce truncated lamin A/ATM, slowing progeroid phenotypes.

Next Steps: The field should closely watch how Horvath’s gene list evolves. As new regulators like SP1 come to light, bridging FAD and NAD+ depletion, the potential for combination therapies—targeting splicing fidelity, GABA restoration, α-KG supplementation, and enzyme inhibition—becomes a promising frontier in slowing or reversing aspects of the aging clock.

9. Final Thought

Stephen Horvath’s evolving Universal Mammalian Epigenetic Aging system underscores the complex synergy among epigenetics (DNA methylation), metabolism (α-KG, NAD+, FAD), and transcription/splicing networks. With each new revision—introducing genes like SP1—we see deeper connections forming between energy cofactors, progeroid genes (WRN, lamin A), and master transcription factors. As research continues, TDG-mediated demethylation, CD38/MAO-related cofactor draining, and splicing misregulation appear to be the crux of how mammals age. Unlocking these interconnected pathways promises a multi-pronged approach to restoring metabolic and epigenetic homeostasis in old age.

My non-AI note-Let me add one more amazing insight- if you study all the abstracts related to DLX6-AS1 and DLX6 you will see that it mostly involves cancer, but there are a few abstracts showing this gene is involved in diabetes, schizophrenia, autism and Alzheimer’s. When you try to figure out what all these diseases have in common it turns out to be deficiencies or dysregulation of GABA. It turns out that declining SERUM GABA is associated with increased incidence of cancer and diabetes. Dysregulation of GABA is also a major factor in schizophrenia , autism, and Alzheimer’s. One hypothesis I am mulling over is that declining serum GABA might lead to the inactivation of the REST gene which could cause reactivation of neural development genes in somatic cells which might be the main cause of somatic cancers with aging. Also in my upcoming book I do a detailed analysis of what serum amino acid changes occur with aging. There are dramatic increases and dramatic declines in many of the amino acids in serum. Notably while inhibitory GABA declines its excitatory partner glutamate increases dramatically. This can explain why there appearance of GABA and glutamate related genes in Horvath’s aging gene set. If you would be interested in an advanced pdf copy of my very detail book which is a super deep dive into Horvath’s aging genes and the newer ones added and removed form his list please send an email request to [email protected].

Blog