WillGaoLab · scientific reading notes
Read Paper with Will · 001
Reading A Hormone Cell Atlas: The Human Endocrine System at Cellular Resolution
Paper: Fei L, Huang-Doran I, Lawler K, et al. Science 393, no. 6806 (2026): eaeb2672. DOI: 10.1126/science.aeb2672
One-sentence summary
A valuable resource that constructs the human Hormone Cell Atlas and maps predicted hormone production and reception, adipocyte endocrine remodeling, and monogenic endocrine disease genes at cellular resolution.
The Question
We usually learn the endocrine system organ by organ, but hormones operate through communication across the whole body. Hormones coordinate essential biological processes, including metabolism, growth, and reproduction. However, hormone biology has traditionally been studied through individual hormones, endocrine organs, or specific signaling pathways.
This paper asks:
How can we understand the human endocrine system at cellular resolution and connect hormone signaling with physiology and disease?
01 · Build the atlas
Figure 1: Build the human Hormone Cell Atlas by integrating sc/snRNA-seq data and hormone-related gene information
The authors built a cross-tissue human endocrine atlas by integrating sc/snRNA-seq datasets with curated hormone–receptor information (Figure 1A). This provides a cellular framework to predict hormone-producing and hormone-responsive cell types across human tissues. The major contribution of this figure is to establish the foundation for a system-level endocrine map rather than studying individual hormones or organs in isolation.
With the map established, the next step is to predict which cells produce the signals.
02 · Map hormone producers
Figure 2: Map predicted hormone-producing cell types
The atlas identifies predicted hormone-producing cell types across tissues and shows that hormone-related expression patterns are strongly associated with cell identity. Beyond classical endocrine organs, some unexpected cell types were predicted to produce hormones.
Key Example
- SCT/secretin expression in plasmacytoid dendritic cells (pDCs), supported by transcript- and protein-level evidence (Figures 2E and 2G–J)
This expands the traditional view that hormone production is restricted to classical endocrine organs and suggests a broader endocrine network across tissues.
Mapping predicted hormone-producing cells defines the potential signal sources; the next step is to predict which cells receive them.
03 · Map hormone receivers
Figure 3: Map predicted hormone-responsive cell types and infer endocrine communication networks
By mapping hormone receptor expression, the atlas predicts hormone-responsive cell populations and infers endocrine communication networks. The study recovers classical endocrine axes and predicts cross-tissue hormone interactions.
Key Example
- GLP1R and GIPR co-expression in cardiomyocytes and cardiac pacemaker cells (Figure 3H)
This demonstrates that endocrine regulation is not a simple one-organ-to-one-organ pathway but a distributed cellular communication network.
The authors then applied this framework to adipose tissue, an important metabolic and endocrine organ.
04 · Explore adipocyte remodeling
Figure 4: Explore adipocyte endocrine remodeling during differentiation, across depots and metabolic states
The atlas reveals that predicted adipocyte endocrine profiles change dynamically during differentiation and differ across adipose depots and metabolic states.
Adipose progenitor cell (APC) → committed preadipocyte (CPA) → mature adipocyte
Maturation is associated with changes in:
- hormone production
- hormone reception
Key Example
- regulatory programs, including SCENIC-predicted regulon activity across APCs, CPAs, and mature adipocytes (Figure 4I)
Adipocytes should be considered dynamic endocrine cells rather than passive energy storage cells. This raises questions about how endocrine programs contribute to metabolic protection or dysfunction.
The atlas was then used to connect monogenic disease genes with specific cellular and anatomical contexts.
05 · Connect genes with disease
Figure 5: Map monogenic endocrine disease genes to specific cell types and anatomical zones
The atlas links monogenic endocrine disease genes with their cell-type-specific and anatomical expression patterns.
Key Example
- Expression of KCNQ1 and KCNJ5 in adrenal cortical cells, together with protein-level evidence of KCNQ1 expression in human adrenal tissue, suggests a potential role for potassium channels in adrenal function and aldosterone regulation (Figure 5E–F).
By linking disease-associated genes to the specific cellular and anatomical contexts where dysfunction may arise, the atlas turns molecular alterations into testable, mechanistic hypotheses for endocrine disorders.
My Assessment
My Thoughts
My overall impression is that the major strength of this paper lies in the development of a comprehensive endocrine cell atlas rather than in uncovering new biological mechanisms. Much of the validation demonstrates that the atlas recapitulates established endocrine biology, which is appropriate for demonstrating the quality of the resource. The study proposes many intriguing hypotheses, but only a limited number are followed up with functional experiments, leaving substantial opportunities for future mechanistic work.
A Note on Scope
Reading depot and sex comparisons in an integrated atlas
One thing worth holding in mind when reading the adipose comparisons: this atlas is assembled from many independent studies, and the adipose depots sampled are unlikely to be evenly represented across them. A difference that appears to distinguish one depot from another may therefore be partly a difference between the studies that contributed those depots. The same caution applies to the sex comparison, which is collapsed across depots — if depot representation is not balanced between sexes, an apparent sex effect could partly reflect depot composition. This does not undercut the atlas; disentangling biological from technical variation is the central difficulty of any cross-study integration, and the authors are explicit that these are predictions. It does mean that the depot- and sex-resolved panels are best read as a starting point for targeted comparison rather than as settled effect estimates.