本文分析了小鼠不同年龄段的脑、肌肉和脾脏的单细胞基因表达谱，发现了一些与衰老相关的基因表达变化。例如，老年小鼠大脑中炎症相关基因表达上调，而肌肉中则观察到肌肉丢失相关基因表达上调。此外，脾脏中特定免疫细胞群体的积累可能与衰老相关的炎症有关。这些发现为深入理解衰老过程提供了新的视角，并为针对衰老相关疾病的干预措施提供了潜在的研究方向。

🧠 老年小鼠大脑中的炎症相关基因表达上调，例如C4b、H2-K1、LGALS3BP、CTSS、H2D1、LYZ2、GFAP、LGALS3、IFIT3、C3和NEAT1等。这些基因的表达增加表明老年小鼠大脑处于慢性炎症状态，这与阿尔茨海默病等神经退行性疾病的发生密切相关。

💪 老年小鼠肌肉中肌肉丢失相关基因表达上调，例如Alpi和Defa30等，而一些与胶原蛋白相关的基因表达下调，例如Col1a2和Col1a1等。这些基因表达变化可能与老年小鼠肌肉质量下降和功能减退有关。

脾脏中，老年小鼠脾脏中存在特定免疫细胞群体的积累，例如表达Gzmk的细胞毒性T细胞。这些细胞的积累可能与老年小鼠体内炎症的增加有关。

这些基因表达的变化为深入理解衰老过程提供了新的视角，并为针对衰老相关疾病的干预措施提供了潜在的研究方向。例如，刺激神经肽W/B受体以治疗肥胖？靶向防御素以治疗肌少症和其他肌肉疾病？靶向表达Gzmk的T细胞以治疗与年龄相关的炎症性疾病？

Published on September 20, 2024 6:20 PM GMT

A very cool project, sponsored by the Chan Zuckerberg Initiative, was the Tabula Muris and the Tabula Muris Senis — a single-cell gene expression “atlas” of the mouse, and then mice at various ages from 3 months (young adulthood) to 27 months (near the end of their lifespan.)

There’s a handy-dandy differential expression app that lets you see which genes are more expressed, or less expressed, in different tissues with age.

This is great for basic exploration that could give us insights into the mechanisms of aging (and ultimately inform research towards treatments of the diseases of aging.). So let’s play around with it and see if anything interesting is going on!

I’m going to just compare 27-month-old to 3-month-old mice. I’ll refer to genes as “upregulated” (with age) if they’re more abundant in the 27-month-olds, and “downregulated” (with age) if they’re more abundant in the 3-month-olds.

Brain

TL;DR

The aging brain is chronically inflamed, especially in neurodegenerative diseases of aging (Alzheimer’s, dementia, etc). The upregulated genes in old mouse brains are all markers of inflammation.

Interestingly, GPR7, the receptor for neuropeptides B and W, is downregulated with age. This may be a factor in age-related obesity.

Upregulated With Age

C4b

This is the gene for complement factor 4, part of the immune complement system that is active during inflammation and infection.

This is unsurprising; most parts of the body, including the brain, become chronically more inflamed with age.

H2-K1

This is an element of the major histocompatibility complex class 1, a cell surface protein that displays antigen peptides so the immune system can destroy damaged or virus-infected cells.

Again, unsurprising; the aging brain has more damaged cells and more immune activity.

LGALS3BP

this is the gene for binding galectins, a type of protein involved in cell-cell interactions, including immune interactions. In the brain, it seems to be characteristic of activated microglia123 , a type of immune cell involved in brain inflammation and neurodegenerative disease.

As It Is Written: It’s The Microglia, Stupid.

CTSS

This is the gene for cathepsin S, an immune-system gene used for breaking down proteins for antigen presentation.

Once again, unsurprising that it would be upregulated in an aging inflamed brain.

H2D1

This is a component of the MHC class 2 molecules, found on immune cells, including microglia in their activated state.

once again, yep, that’s a sign of neuroinflammation.

LYZ2

This is the gene for lysozyme (LYZ in humans), an antimicrobial enzyme in the innate immune system. It breaks down bacterial cell walls, killing the germs.

You get elevated levels of lysozyme in the brain during brain infections, obviously, but also during brain tumors4, stroke5, and Alzheimer’s6, which are age-related non-infectious diseases associated with an inflammatory response.

Aged mice don’t generally get brain tumors, strokes, or Alzheimer’s per se, but they do develop inflammatory brain damage and impaired cognition.

GFAP

This is the gene for glial fibrilary acidic protein, a major structural protein in astrocytes. In fact, it’s a standard immunostaining marker for distinguishing astrocytes from other cell types.

GFAP expression increases in astrocytes when they become activated (or reactive), part of the neuroinflammation response to damage (from eg brain injury, stroke, tumors, or neurodegenerative disease.)7

Again, no surprises here. We’re seeing more GFAP in the old mice because they have neuroinflammation.

LGALS3

This is the gene for galectin-3, which was discussed above (LGALS3BP is the binding protein for LGALS3).

Again, unsurprising; this is due to activated microglia in neuroinflammation.

IFIT3

This is a gene involved in the innate immune system in response to inflammatory interferon signaling as a response to viral infection.

Again, this looks like a neuroinflammation thing.

C3

This is a component of the complement system, part of the immune response.

Again: more C3 in the aging brain suggests neuroinflammation.

NEAT1

NEAT1 is a non-coding RNA, localized to a structure inside the nucleus known as a paraspeckle. We don’t know exactly what paraspeckles do, but they may be a defense against RNA viruses.

In the brain, NEAT1 also promotes inflammatory responses to traumatic brain injury8 and is upregulated in a variety of human neurodegenerative diseases (Alzheimer’s, Parkinson’s, Huntington’s, ALS)9 so it may also be a neuroinflammation thing.

H2-Q6

This is an MHC2 gene

again, that’s inflammation

H2-Q7

This is an MHC2 gene

again, that’s inflammation

Downregulated with Age

GPR7

This is a binding protein for neuropeptide B and neuropeptide W, proteins particularly expressed in parts of the brain like the amygdala that regulate stress and fear.

Mice lacking this gene approach strange mice, rather than hesitating cautiously as mice usually do.10

Mice lacking this gene also develop adult-onset obesity.11

Progression

The earliest of these genes to be upregulated with age is C4b, which shows up at 9 months. By 21 months, most of these genes are upregulated. GPR7 downregulation doesn’t show up until 27 months.

Muscle

TL;DR

Mice, like humans, lose muscle and strength with age; mice begin to show elevated levels of muscle-loss-associated proteins quite young, before “middle age”. At least one of them, a defensin, belongs to a class of proteins known to cause muscle loss when administered experimentally. Perhaps anti-defensin strategies might be a good research direction for preventing sarcopenia?

Also, collagen genes are downregulated with age; this might be a side effect of collagen accumulation in muscles (a known age-related phenomenon.)

Upregulated With Age

H1fnt

This is a histone, one of the “beads” around which DNA is normally coiled.

Why is its expression upregulated in aged muscle? I’m generally stumped. Overall, aging is associated with loss of histones.

Alpi

This is the gene for alkaline phosphatase, mostly used for regulating homeostasis in the intestine. In skeletal muscle, you see alkaline phosphatase mostly in muscle wasting diseases12 and high serum alkaline phosphatase is a biomarker for sarcopenia (loss of muscle mass with age).13

I don’t think it’s known, causally, why alkaline phosphatase seems to be associated with muscle loss.

Defa30

This is a gene for defensin (human ortholog DEFA3), an antimicrobial peptide.

Defensins cause muscle loss!1415 And they’re associated with muscle degeneration in inflammatory myopathies.16

Downregulated with Age

Cilp2

This is a cartilage protein; usually its expression increases with age in the presence of osteoarthritis.

Why is it downregulated with age in muscle? Beats me!

Col1a2

That’s a collagen gene.

In general, collagen accumulates in aged muscles, causing stiffness; but this is not a result of increased gene expression but rather lack of ability to clear it out.17

It’s not unusual for production of a substance in the body to go down in response to that substance being abundant, in a standard negative feedback mechanism; that could (speculatively) be what’s going on here.

Col1a1

Also a collagen gene; also might be downregulated as a result of collagen buildup.

Progression

Alpi begins to be upregulated super early, at 9 months. H1fnt and Defa30 are also upregulated by 15 months. Downregulated Cilp, Col1a2, and Col1a1 show up later, at 18-21 months.

Spleen

TL;DR

The aging spleen is interesting because removing it altogether seems to make animals live longer. What is the aged spleen doing wrong, at a gene expression level?

Looks like, granzyme K-positive cytotoxic T cells. They accumulate with age and promote inflammation throughout the body. (They aren’t a unique “culprit” for age-related inflammation, though; inject them into a young mouse and they don’t promote the same phenotypes.)

Upregulated with Age

C130026I21Rik

This is a nuclear body protein like gene; it’s homologous to the transcriptional regulator SP140, which is found in white blood cells and downregulates genes involved in autoimmune diseases.18

Snhg11

This is a long non-coding RNA that promotes growth in a wide variety of cancers.192021

Gzmk

This is the gene for granzyme K, found in the granules of immune cells (NK cells and cytotoxic T lymphocytes).

In humans, higher granzyme K levels are associated with older age and CMV infection.22 In fact, granzyme K-positive cytotoxic T cells are a highly clonal population23 that grow more common with age across organs in mice and are associated both with age and with age-related inflammatory markers in humans.

Progression

Snhg11 and Gzmk start to get upregulated at middle age (12 months); C130026I21Rik pops up only slightly later, at 18 months.

Conclusions

This is just a starting point (I picked a couple tissues to play with but there are lots more.)

But already it’s easy to observe patterns that make sense in context (old brains are inflamed; old muscles have weird collagen; old spleens are enriched for particular populations of immune cells) and to generate ideas for interventional research directions.

Speculative Research Directions:

stimulate the neuropeptide W/B receptor to treat obesity?

target defensins to treat sarcopenia and other muscle diseases?

target granzyme K-positive T cells to treat age-related inflammatory conditions?

The majority of drug discovery programs, remarkably enough, do not seem to have come from any kind of genetic analysis of this kind, where you notice that the relevant target is overexpressed (or mutated, or whatever) in disease.

Obviously, the “sit on your butt and read/play with omics data” approach to idea generation offers only a slim chance of actually coming up with something successful; but I keep seeing data points suggesting that, despite how easy it is to do, it’s underrated and under-practiced in biomedical research.

Maybe Research Idea-Generation At Scale Is Valuable?

Someone (maybe me!) should start compiling a database of “someone should try this” experimental approaches suggested by systematic “omics” datasets (like the Tabula Muris Senis and CellxGene, and GWAS studies, and much much more.)

This requires a common-sense filtering step based on context from the research literature — I don’t think writing a simple script to automate it would be very useful — but it is of course accelerated by LLM tools like Perplexity that help you search for research papers that answer your questions.

Once fully populated, this sort of resource might be valuable to researchers; it might even be a higher-quality version of some of the functions that fully automated “AI for target discovery” tools aim to provide biotech companies.

Discuss