Single Cell Analysis of Human Thyroid Reveals the Transcriptional Signatures of Aging

Abstract

The thyroid gland plays a critical role in the maintenance of whole body metabolism. However, aging frequently impairs homeostatic maintenance by thyroid hormones due to increased prevalence of subclinical hypothyroidism associated with mitochondrial dysfunction, inflammation, and fibrosis. To understand the specific aging-related changes of endocrine function in thyroid epithelial cells, we performed single-cell RNA sequencing of 54,726 cells derived from pathologically normal thyroid tissues from seven patients who underwent thyroidectomy. Thyroid endocrine epithelial cells were clustered into five distinct subpopulations, and a subset of cells were found to be particularly vulnerable with aging, showing functional deterioration associated with the expression of metallothionein (MT) and MHC class II genes. We further validated that increased expression of MT family genes are highly correlated with thyroid gland aging in bulk RNAseq data sets. This study provides evidence that aging induces specific transcriptomic changes across multiple cell populations in the human thyroid gland.

Full text

Endocrinology, 2023, 164, 1–12
https://doi.org/10.1210/endocr/bqad029
Advance access publication 15 February 2023
Research Article
Single Cell Analysis of Human Thyroid Reveals the
Transcriptional Signatures of Aging
Yourae Hong,
1,2,
*Hyun Jung Kim,
3,
*Seongyeol Park,
4
Shinae Yi,
5
Mi Ae Lim,
5
Seong Eun Lee,
5
Jae Won Chang,
6
Ho-Ryun Won,
6
Je-Ryong Kim,
4,7
Hyemi Ko,
7
Seon-Young Kim,
8
Seon-Kyu Kim,
8
Jong-Lyul Park,
8
In-Sun Chu,
9
Jin Man Kim,
10
Kun Ho Kim,
11
Jeong Ho Lee,
3
Young Seok Ju,
3,5
Minho Shong,
4,12
Bon Seok Koo,
6
Woong-Yang Park,
1,2
and Yea Eun Kang
4,12
1
Samsung Genome Institute, Samsung Medical Center, Seoul, Korea
2
Department of Health Sciences and Technology, Samsung Advanced Institute for Health Sciences & Technology, Sungkyunkwan University,
Seoul, Korea
3
Graduate School of Medical Science and Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Korea
4
Genome Insight Technology, Daejeon, Korea
5
Research Institute of Medical Science, Chungnam National University, Daejeon, Korea
6
Department of Otolaryngology–Head and Neck Surgery, College of Medicine, Chungnam National University, Daejeon, Korea
7
Department of Surgery, College of Medicine, Chungnam National University, Daejeon, Korea
8
Personalized Genomic Medicine Research Center, Research Institute of Bioscience and Biotechnology, Daejeon, Korea
9
Genome Editing Research Center, Korea Research Institute of Bioscience and Biotechnology, Daejeon, Korea
10
Department of Pathology, College of Medicine, Chungnam National University, Daejeon, Korea
11
Department of Nuclear Medicine, Chungnam National University Hospital, Daejeon, Korea
12
Division of Endocrinology and Metabolism, Department of Internal Medicine, College of Medicine, Chungnam National University, Daejeon, Korea
Correspondence: Bon Seok Koo, MD, PhD, Department of Otolaryngology–Head and Neck Surgery, College of Medicine College of Medicine, Chungnam
National University, Daejeon 35015, South Korea. Email: bskoo515@cnuh.co.kr; or Woong-Yang Park, MD, PhD, Samsung Genome Institute, Samsung Medical
Center, Seoul, Korea, and Department of Health Sciences and Technology, Samsung Advanced Institute for Health Sciences & Technology, Sungkyunkwan
University, Seoul 06351, Korea. Email: woongyang.park@samsung.com; or Yea Eun Kang, MD, PhD, Division of Endocrinology and Metabolism, Department of
Endocrinology and Metabolism, College of Medicine, Chungnam National University, Daejeon 35015, South Korea. Email: yeeuni220@cnuh.co.kr
*These authors contributed equally to this work.
Abstract
The thyroid gland plays a critical role in the maintenance of whole-body metabolism. However, aging frequently impairs homeostatic maintenance
by thyroid hormones due to increased prevalence of subclinical hypothyroidism associated with mitochondrial dysfunction, inflammation, and
fibrosis. To understand the specific aging-related changes of endocrine function in thyroid epithelial cells, we performed single-cell RNA
sequencing (RNA-seq) of 54 726 cells derived from pathologically normal thyroid tissues from 7 patients who underwent thyroidectomy.
Thyroid endocrine epithelial cells were clustered into 5 distinct subpopulations, and a subset of cells was found to be particularly vulnerable
with aging, showing functional deterioration associated with the expression of metallothionein (MT) and major histocompatibility complex
class II genes. We further validated that increased expression of MT family genes are highly correlated with thyroid gland aging in bulk
RNAseq datasets. This study provides evidence that aging induces specific transcriptomic changes across multiple cell populations in the
human thyroid gland.
Key Words: thyroid, aging, single-cell RNA sequencing, metallothionein, fibrosis
Received: 13 July 2022. Editorial Decision: 9 February 2023. Corrected and Typeset: 2 March 2023
© The Author(s) 2023. Published by Oxford University Press on behalf of the Endocrine Society. All rights reserved. For permissions, please e-mail:
journals.permissions@oup.com
The thyroid gland plays many roles in organ development and
the homeostatic control of fundamental physiological mecha-
nisms, such as body growth and energy metabolism, in all ver-
tebrates (1, 2). Diverse cell types, including epithelial (thyroid
follicular and parafollicular cells), endothelial, and stromal
cells, are necessary to maintain thyroid homeostasis (2).
Thyroid follicular cells, the major cell type of the thyroid
gland, synthesize thyroglobulin (Tg) through iodine uptake
by the sodium iodine symporter and peroxidation by thyroid
peroxidase (TPO) (3). Under normal physiological conditions,
Tg iodination involves a rapid change in iodine levels and local
generation of reactive oxygen species (ROS) in the thyroid
(4, 5). The thyroid hormone synthesis and secretion processes
are regulated by serum thyrotropin (TSH) (3). Thyroid fol-
licles are heterogeneous in size and function, specically in re-
sponse to TSH (6, 7). This interfollicular heterogeneity is
reportedly exacerbated by radiation exposure, iodine excess,
autoimmunity caused by antigen-presenting cells, or micro-
vascular structures (8-11). We hypothesized that the transcrip-
tional heterogeneity within thyroid epithelial cells might
explain the functional aspects of interfollicular heterogeneity,
as a previous study using zebrash showed the heterogeneity
of thyrocytes, including the bimodal expression of the tran-
scription factor pax2a (12).
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023

2 Endocrinology, 2023, Vol. 164, No. 4
Aging deteriorates thyroid function by increasing the preva-
lence of subclinical hypothyroidism, characterized by normal
free thyroxine (T
4
) and elevated TSH levels (13, 14). Under
normal physiological conditions, Tg iodination and coupling
processes occur at the apical pole, on the follicular side of
the cell membrane, in a process requiring a hydrogen
peroxide-generating system, such as the nicotinamide adenine
dinucleotide phosphate oxidase family for hormone synthesis
(4, 5). Accumulation of ROS due to continuous thyroid hor-
mone synthesis is suggested as one of the possible mechanisms
underlying aging-related hypothyroidism (15). This phenom-
enon reects age-dependent histological changes, including in-
creased interfollicular brosis, follicle size, and colloid content
reduction, as well as attening of glandular epithelial cells and
decreased total thyroid gland weight (16). In addition, other
features of aging thyroid include increased susceptibility to
autoimmune reactions and changes in the immune system
caused by failed effector mechanisms of the lymphatic system
(16, 17). Advances in single-cell RNA sequencing
(scRNA-seq) provide a tool for exploring physiological and
pathological transcriptomic changes in the aging thyroid at
single-cell resolution.
In this study, we performed a scRNA-seq of pathologically
normal thyroid tissues from 7 patients who underwent thyroi-
dectomy. The single-cell atlas revealed distinct cell types in the
human thyroid, including epithelial, vascular, and various im-
mune cells, along with broblasts. We focused on the tran-
scriptional heterogeneity of thyroid epithelial cells and
identied a small subset of epithelial cells that is particularly
vulnerable to aging. We also analyzed the cell type-specic
transcriptional changes and ligand-receptor interactions be-
tween cells residing in the human thyroid. Overall, our data-
sets provide a resource for understanding the mechanism
underlying interfollicular heterogeneity and offer insights
into impaired homeostatic maintenance during aging.
Methods
Patient Information and Sample Collection
This study was approved by the Institutional Research and
Ethics Committee at Chungnam National University
Hospital (CNUH- 2020-11-004-001), and informed consent
was obtained from all participants. For scRNA-seq, thyroid
tissues were obtained from 7 patients who underwent thyroi-
dectomy, with conrmed cases of differentiated thyroid can-
cer (Supplementary Table S1) (18). Inclusion criteria were
patients negative for anti-Tg and anti-TPO antibodies with
no structural changes related to autoimmune thyroiditis in
the sonographic examination. Normal thyroid tissues were
sampled at least 3 cm away from the tumor; half of the tissue
samples was examined by pathologists, while the other half
was transported to the laboratory within 30 minutes after re-
section. For bulk RNA-seq, thyroid tissues were obtained
from 65 patients were selected with the same inclusion criteria
(Supplementary Table S6) (18).
Sample Preparation
Thyroid tissues were minced in Hank’s Balanced Salt Solution
for 5 minutes and digested with 400 unit/mL collagenase type
4 (Worthington, Lakewood, NJ, USA) dissolved in RPMI1640
media at 37°C for 60 minutes. The cells were centrifuged at
3100×g or 5 minutes at 4°C, and the pellets were resuspended
in RPMI medium with 10% fetal bovine serum (FBS). After
mechanical dissociation by pipetting, the cell suspension was l-
tered through a 70 μM cell strainer and centrifuged at 3100×g
for 5 minutes at 4°C. The pellets were incubated with red blood
cell lysis buffer (11814389001, Roche Diagnostics, Mannheim,
Germany) for 10 minutes on ice. After centrifugation, the cell
pellets were resuspended in FBS-free RPMI medium and ltered
through a 40 μM cell strainer.
Library Preparation and Preprocessing
The concentration of the cell suspension was measured using
an automated cell counter and adjusted to capture 10 000 cells
per sample. Following the manufacturer’s instructions, cells
from 5 thyroid samples were loaded according to the standard
protocol of the Chromium Single Cell 3′ Reagent Kit (v3
chemistry, 10× Genomics, Pleasanton, CA, USA). Two thy-
roid samples (Young_3, Young_4) were loaded by the
Chromium Single Cell 5′ Reagent Kit (v1.1 chemistry).
Samples were sequenced with HiSeq X (Illumina, San Diego,
CA, USA). scRNA-seq data were mapped, and the number
of molecules per barcode was quantied using the 10× soft-
ware package cellranger (version 3.0.2) and GRCh38 refer-
ence genome supplied by 10× Genomics.
Processing of scRNA-seq Data for Cell Type
Identification and Downstream Analyses
To exclude low-quality cells or doublets, we used the Scrublet
(19) doublet detection algorithm with an expected doublet rate
parameter of 0.1 and ltered cells with fewer than 200 genes or
cells with 25% or higher mitochondrial proportion using the
Seurat R package (v3.2.0) (20). Each gene unique molecular
identier was normalized using total unique molecular identi-
er counts in the corresponding cell, multiplied by a scale fac-
tor of 10,000, and transformed to the natural log scale plus
1. For the clustering of major cell types, the most variable
2000 genes were used to perform principal component ana-
lysis, and the top 30 principal components were converted to
a harmony dimension for correction of sample specicity
(21) and then clustered and projected to the uniform manifold
approximation and projection (UMAP) dimension using
Seurat. For accurate cell type identication, we removed clus-
ters that identied the doublet and repeated these steps. For
subclustering, we performed dimension reduction and cluster-
ing after removing low-quality cells and doublets. To identify
the natural killer (NK) cell subtype, we excluded cells with an
average expression of CD3 (CD3D, CD3E, CD3G) genes
above 1. CellphoneDB (v2.0) (22) was used to calculate trans-
forming growth factor-beta (TGF-β) family interaction pairs
signicantly expressed in broblast and other cell types. For
this, we used default parameters and permuted cell type labels
for each cell 100 times to obtain the signicance of each pair.
Cell cycle phase was predicted using CellCycleScoring
function in Seurat package and identied cell cycle associated
genes (23).
Differentially Expressed Genes Identification and
Enrichment Analyses
To identify differentially expressed genes (DEGs) among clus-
ters, we used the FindAllMarkers and FindMarkers functions
in Seurat and applied the Wilcoxon rank-sum test with
Bonferroni false discovery rate correction. Similarity among
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023

Endocrinology, 2023, Vol. 164, No. 4 3
cell types or subclusters was determined using Pearson’s cor-
relation coefcient with the mean expression of cell types
with the top 2000 variable genes.
To identify broblast functionality using signicant genes,
we used ingenuity pathway analysis with signicant genes
and mapped them in the Ingenuity Pathway Knowledge
Base. Scoring for gene ontology terms was used to identify
cell functionality depending on age or subcluster. The mean
expression of each gene was calculated using GO Biological
Process Genesets from MsigDB (v7.1). Moreover, the proin-
ammatory and anti-inammatory score was calculated using
a signature geneset (24).
Bulk RNA-seq and Processing
Total RNA was extracted from 65 tumor-adjacent normal
thyroid tissues using TRIzol reagent (15596026, Invitrogen,
Waltham, MA, USA). Complementary DNA was prepared
from the total RNA using Moloney Murine Leukemia Virus
reverse transcriptase and oligo-dT primers (28025013,
Invitrogen). Libraries were generated using the TruSeq
Stranded mRNA LT Sample Prep Kit (Illumina) according
to the manufacturer’s protocol. The NovaSeq 6000 system
(Illumina) was used to generate 101 bp paired reads. The
adapter content in the reads was trimmed using Cutadapt
(v3.1), and STAR (v2.7.3a) was used to align the reads.
Gene expression levels were determined using RSEM (version
1.3.1). The read counts were normalized by library size using
the DESeq2 algorithm and converted to transcript per million
values using expected gene lengths from RSEM. The correl-
ation coefcient between the transcript per million values
and age was calculated for each gene.
Deconvolution of Bulk RNA-seq Data
Cell fractions from the bulk RNA-seq data for normal thyroid
samples were enumerated using the CIBERSORTx online
platform (https://cibersortx.stanford.edu/) with our single-cell
reference data (25). The reference dataset was constructed
based on the expression of 2000 highly variable genes to ex-
clude noise by low-quality genes. Using the single-cell refer-
ence le, we followed the instructions of CIBERSORTx with
default parameters.
Analysis of the Genotype-Tissue Expression and
The Cancer Genome Atlas Datasets
Correlation analyses were conducted by R package corrplot. The
Genotype-Tissue Expression (GTEx) dataset was downloaded
from the GTEx portal (http://gtexportal.org/). The thyroid car-
cinoma dataset of The Cancer Genome Atlas database was ob-
tained from University of California Santa Cruz database
(https://xenabrowser.net/). In the thyroid carcinoma and GTEx
datasets, we selected RNA-seq results of normal tissues. In the
GTEX dataset, we performed correlation analysis after batch cor-
rection about death type. To analyze correlation among age, MT
genes, and thyroid function related genes, we tested with Pearson
and Spearman methods in datasets. Correlation matrices were
performed using the “corrplot” package in R studio.
Cell Culture and Quantitative Reverse Transcription
Polymerase Chain Reaction
We used Nthy-ori3-1, which was purchased from Sigma-
Aldrich. Cell lines cultured using RPMI 1640 medium
(Welgene, Daejeon, Korea) supplemented with 10% FBS
(Gibco, CA, USA), 100 U/mL penicillin, and 100 ug/mL strep-
tomycin at 37°C in a humidied atmosphere of 5% CO
2
.
Total RNAs of cell lines were isolated using TRIzol reagent
(Life Technologies, Eugene, OR, USA) according to the man-
ufacturer’s protocol. For synthesizing cDNA, reverse tran-
scription was performed with total RNA, Moloney Murine
Leukemia Virus reverse transcriptase (Invitrogen, Carlsbad,
CA, USA), and oligo-dT primers (Promega). Quantitative re-
verse transcription polymerase chain reaction was performed
using the synthesized cDNA with SFCgreen-Cyan qPCR
Master Mix (2X)-Low ROX (SFC Probes, Korea) by a 7500
Real-Time PCR System. Relative expressed genes were nor-
malized by 18S ribosomal RNA as a housekeeping gene
(Supplementary Table S7) (18).
Serum Measurement
Concentrations of triiodothyronine (T3), free thyroxine (free
T4), and TSH were measured in the Chungnam National
University Hospital using Gamma pro (Kaien, Korea) and
Dream Gamma 10 (Shin Jin Medics, Korea).
Statistical Analysis
Statistical signicance between multiple parameters was de-
termined using a one-way analysis of variance, and a 2-sided
Wilcoxon rank-sum test was used to compare the 2 parame-
ters. For the similarity of 2 variables, we used Pearson’s
correlation coefcient.
Results
A Single-Cell Atlas of Human Thyroid Tissue in
Young and Old Adults
To characterize age-dependent physiological changes in the
human thyroid, we selected 7 patients (four 20- to 30-year-old
and three 50- to 60-year-old patients) who underwent thyroi-
dectomy for differentiated thyroid cancer. We excluded pa-
tients with Hashimoto’s thyroiditis or other destructive
thyroiditis conditions to ensure that we could evaluate the ef-
fects of physiological aging on the thyroid. The normal tissues
adjacent to the tumor were used in the study. The old group
showed signicantly increased TSH levels (Young; 0.04 ±
1.03, Old; 1.40 ± 0.21, P value = .007) and decreased free
T4 (Young; 1.40 ± 0.21, Old; 1.03 ± 0.03, P value = .042)
(Fig. 1A and 1B; Supplementary Table S1) (18). The variations
between the 2 groups were within the normal range and did
not show signs of overt or subclinical hypothyroidism. We
rst analyzed the cell types residing in the thyroid tissue,
examining a total of 54 726 cells (Fig. 1C; Supplementary
Table S2) (18). Unsupervised clustering analysis revealed 16
distinct clusters corresponding to 7 cell categories, including
epithelial cells, broblasts, smooth muscle cells (SMCs)/peri-
cytes, endothelial cells (ECs), myeloid cells, B cells, T cells,
and NK cells. Every cluster was observed in both age groups
and each patient, and the overall cell type categories persisted
during aging despite a high degree of individual variation
(Fig. 1D; Supplementary Fig. S1) (18). We used canonical
markers to annotate cell types in the thyroid tissue, and each
cell type consisted of a variable number of cells from each pa-
tient (Fig. 1E; Supplementary Fig. S2) (18). We assessed the
annotation accuracy using hierarchical clustering of the
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023

4 Endocrinology, 2023, Vol. 164, No. 4
pairwise correlation among cell types using 2000 highly vari-
able genes (Fig. 1F).
Heterogeneous Subpopulations of Thyroid
Epithelial Cells
We analyzed a total of 19 012 epithelial cells and identied 5 dis-
tinct subpopulations based on Louvain clustering (Fig. 2A). All
clusters were found in both young and old groups, with different
proportions in each patient (Fig. 2B). Although the number of
genes varied among subclusters, none of them represent a popu-
lation of dead or dying cells according to mitochondrial gene
percentages (Fig. 2C and 2D). All of the epithelial subpopula-
tions exhibited comparable level of genes involved in thyroid
hormone synthesis and iodine uptake, including thyrotropin re-
ceptor (TSHR), TPO, and thyroglobulin (TG) (Fig. 2E). As
shown in Supplementary Fig. S2C, SLC26A4 and DUOX1
were found to be expressed in the epithelial cell clusters (18).
DEGs for each subcluster represented the specialization of the
thyroid epithelial cells (Supplementary Fig. S3; Supplementary
Table S3) (18). Cluster 0 showed increased expression of
DUSP1, JUN, and FOS, which mediates cell proliferation asso-
ciated with thyroid receptor β (26, 27). Similarly, cluster 1 ex-
pressed higher levels of G2/M checkpoint genes, such as
GADD45B and CDKN1A (28-30). Cell cycle scoring also con-
rmed an increased proportion of cells in the G2/M phase in C1
compared with that in C0, corresponding to the expression of
cell cycle-related genes (31-34) (Fig. 2F). The messenger RNA
expression of MT genes such as MT1G, MT1F, and MT2A
was increased in cluster 2 (Fig. 2G). Moreover, MT1G expres-
sion was negatively correlated with TSHR and TPO expression,
suggesting that increased MT expression is associated with the
expression of genes related to thyroid function
(Supplementary Fig. S3E) (18). Cluster 4 showed higher
Figure 1. Single-cell atlas of human thyroid tissue in young and old adults. (A) Experimental scheme of the study and number of cells analyzed in each age
group. (B) Serum levels of TSH, free T4, and T3 in each age group. (C) UMAP plot illustrating various cell types in the human thyroid tissue. (D) UMAP plots
separated by age group. I Dot plot shows the expression of marker genes for each cell type. The dot size indicates the fraction of expressing cells, and the
color indicates the expression level. Bar plots illustrate the proportion of patient-derived cells for each cell type. (F) Hierarchical clustering of pairwise
correlation between annotated cell type clusters. Pearson’s correlation values were calculated from the mean expression of each cluster.
Abbreviations: T3, Triiodothyronine; T4, thyroxine; TSH, serum thyrotropin; UMAP, uniform manifold approximation and projection.
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023

Endocrinology, 2023, Vol. 164, No. 4 5
expression of autophagy regulation-relate genes such as XIST,
NEAT1, and VMP1 (35-37), but overall transcriptome was
similar to C1. Interestingly, genes involved in major histocom-
patibility complex (MHC) class II formation and transport,
namely HLA-DRB1 and CD74, were dominantly expressed in
cluster 3, a minor subset of epithelial cells (38, 39). Although
this cluster expressed both MHC class I- and II-related DEGs,
the expression of MHC class II molecule such as HLA-DRA
was predominantly specic to cluster 3 (Fig. 2H).
Age-Dependent Deterioration on Thyroid Epithelial
Cell Subpopulations
Next we examined whether thyroid function was maintained dur-
ing the aging process. The expression of canonical epithelial
markers such as TPO, TSHR, and SLC5A5 signicantly decreased
during aging, except for TG and PAX8 (Fig. 3A; Supplementary
Fig. S4A) (18). The thyroid function is decreased in old patients,
whereas increased TG expression is possibly in response to in-
creased TSH levels in older patients. Age-dependent changes in
gene expression of thyroid epithelial cells were enriched in EIF2 sig-
naling, oxidative phosphorylation, and mitochondrial dysfunction
using ingenuity pathway analysis (Fig. 3B and 3C). As gene expres-
sions involved in thyroid function were separated by age group and
subcluster, cluster 2 (C2), which expressed higher levels of MT
genes, showed the most signicant functional deterioration among
other epithelial cell populations (Fig. 3D and 3E; Supplementary
Fig. S4B) (18). Although MT gene family serves a protective role
for oxidative stress in the thyroid (40-42), increased expression
of MT genes in C2 could not improve thyroid function possibly
due to ROS accumulation and mitochondrial dysfunction
(Fig. 3F; Supplementary Fig. S4C) (18). Additionally, MHC class
II molecules were exclusively expressed in cluster 3, and the gene
expressions were signicantly upregulated in the elderly group
(Fig. 3G). The genes enriched in older patients include MHC class
II and other immune molecules such as ANXA1 and MAFF, show-
ing specic enrichment in the antigen presentation pathway
(Supplementary Fig. S4D) (18).
Upregulated Signatures Related to Fibrosis in the
Aged Thyroid
Fibrotic changes are known to occur in the human thyroid
during aging (43, 44). We analyzed a total of 9105 broblasts
Figure 2. Single-cell RNA sequencing of thyroid epithelial cells reveals distinct subpopulations. (A) UMAP plot of thyroid epithelial cells. (B)
Composition of epithelial subclusters among individuals. (C and D) Violin plots illustrate the number of genes (C) and mitochondrial percentage (D) of
each subclusterI(E) Violin plots show gene expression related to thyroid function for each subcluster. (F) UMAP plot of cell cycle scores and bar plots of
cell cycle proportion for each cluster. (G) Bar graphs illustrate the mean of scaled expression of MT family genes for each subcluster. (H) UMAP plot
indicates cells expressing MHC class II or I molecules with TG and TPO. Black dots indicate coexpressing cells. The MHC class II coexpression is mainly
localized in subcluster 3.
Abbreviations: MHC, major histocompatibility complex; MT, metallothionein; TG, thyroglobulin; TPO, thyroid peroxidase; UMAP, uniform manifold
approximation and projection.
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023

6 Endocrinology, 2023, Vol. 164, No. 4
and other stromal cell types, including SMCs/pericytes and
vascular or lymphatic ECs, to accurately determine the com-
position of the stromal cell population (Fig. 4A;
Supplementary Fig. S5) (18). The heatmap of cell type-specic
DEGs also revealed strong expression of cell type markers,
such as DCN, FBLN1/2, and COL1A2 for broblasts;
ACTA2 and TAGLN for SMCs/pericytes; VWF for vascular
ECs; and LYVE1 for lymphatic ECs (Supplementary
Fig. S5B; Supplementary Table S4) (18). Although the overall
proportion of each cell type was sustained during aging, we
observed a slightly higher proportion of broblasts in older
patients with a high degree of variation (Supplementary
Fig. S5D and 5E) (18). When we further analyzed the age-
dependent changes in broblasts, genes enriched in older pa-
tients included those related to brosis, such as COL14A1
and FBN1, immune response genes such as CD81, and com-
ponents of the immunoglobulin complex (Fig. 4B). The ca-
nonical pathway analysis of old-enriched DEGs indicated
that inammatory and brotic signaling are activated during
the aging process (Fig. 4C). Since TGF-β is the primary factor
mediating brosis (45), we examined brosis-related
ligand-receptor interactions among different cell types using
CellPhoneDB (Fig. 4D). Fibroblasts interacted with various
immune, vascular, and epithelial cells, and TGF-β interactions
were strengthened in older patients. Therefore, increased ex-
pression of TGF-β receptors TGFBR1 and TGFBR2 was con-
sistently observed in broblasts (Fig. 4E).
We next analyzed 23 758 immune cells to observe cytokine-
producing activated immune cells involved in brosis. Based
on DEG analysis of each subcluster, we identied the immune
Figure 3. Age-dependent changes in thyroid epithelial cells. (A) Violin plots show gene expression related to thyroid function between the 2 age groups.
(B) Volcano plot indicates age-enriched genes in epithelial cells, and yellow- or green-colored dot indicates the age-enriched genes with a minimum
fraction of 0.25 for each group and old-enriched top10 genes are shown. (C) Top 10 enriched ingenuity canonical pathways for age-enriched DEGs in
epithelial cells. (D and E) Violin plots show expression of thyroid function related genes (D) and MT family genes (E) for each subcluster. Red-colored box
indicates log2 fold change of more than 0.5 or less than −0.5 between 2 age groups. (F) Age-dependent changes to cell death in response to oxidative
stress GO term are shown for each subcluster. (G) Violin plots show MHC class II gene expressions for each subcluster. Significance between young
and aged samples was determined using two-sided Wilcoxon test. * P value <.05, ** P value <.01, *** P value <.001, **** P value <.0001.
Abbreviations: DEG, differentially expressed genes; GO, gene ontology; MHC, major histocompatibility complex; MT, metallothionein.
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023

Endocrinology, 2023, Vol. 164, No. 4 7
landscape of normal human thyroid: (1) M1- and M2-like
myeloid cells; (2) CD4 and CD8 T cells; (3) naïve, prolifera-
tive, and memory B cells; (4) NK cells; and (5) dendritic cells
(Fig. 4F; Supplementary Fig. S5; Supplementary Table S5)
(18). Although the proportion of immune cell types did not
change in the aging thyroid, cytokine TGFB1 expression
was predominantly increased in myeloid cell populations
(Fig. 4G). When we further analyzed the age-dependent
changes in macrophages, several mitochondrial genes were
upregulated in older patients, which is possibly related to
mitochondrial dysfunction and inammatory signatures
(Fig. 4H and 4I).
Figure 4. Fibrotic changes and increased TGF-β signaling in the aged thyroid. (A) UMAP plot of thyroid stromal cell populations. (B) Volcano plot
indicates old-enriched genes in fibroblasts, and the top 10 old-enriched genes are shown. (C) Top 10 enriched canonical pathways for age-enriched
DEGs in fibroblasts. (D) Fibrosis-related ligand-receptor interactions were examined among different cell types in the aging thyroid. The analysis was
performed by CellPhoneDB (v2.0) and only indicated significant pairs from the young or old group. (E) Violin plots show gene expression related to TGF-β
for each stromal cell type. (F) UMAP plot of thyroid immune cell populations. (G) Violin plots show gene expression related to TGF-β for each stromal cell
type. (H) Volcano plot indicates old-enriched genes in myeloid cells, and the top 20 old-enriched genes are shown. (I) Top 10 enriched canonical
pathways for age-enriched DEGs in myeloid cells. (E, G) A 2-sided Wilcoxon test was performed. * P value <.05, ** P value <.01, *** P value <.001,
**** P value <.0001.
Abbreviations: DEG, differentially expressed gene; TGF-β, transforming growth factor-beta; UMAP, uniform manifold approximation and projection.
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023

8 Endocrinology, 2023, Vol. 164, No. 4
Age-associated Gene Expression and Cell Type
Composition Determined Using Deconvolution of
Bulk RNA-seq Data
To complement our ndings, we additionally performed de-
convolution analysis using the bulk RNA-seq dataset of 65
normal thyroid tissues in the Chungnam National University
Hospital database (Supplementary Fig. S6A; Supplementary
Table S6) (18). Correlation analysis showed a tendency for
free T4 to decrease with aging, but there was no difference
in the TSH levels (Supplementary Fig. S6A) (18). The cellular
composition of 65 human thyroid samples was estimated us-
ing CIBERSORTx (25), and the overall cellular proportion
of each cell type was sustained during aging, even between
age groups (Fig. 5A and 5B). However, we observed slightly
higher proportions of broblasts and SMC/pericytes in older
patients. The correlations between age and gene expression
of TG, PAX8, TSHR, TPO, and SLC5A5 in epithelial cells
demonstrated that thyroid function is well preserved with
aging (Fig. 5C; Supplementary Fig. S6B) (18). We identied
age-correlated genes based on the bulk RNA-seq data
Figure 5. Age-related gene expression and composition change determined using bulk RNA-seq deconvolution data. (A) Cellular composition of thyroid
cell types determined using bulk RNA-seq deconvolution with CIBERSORTx. Each individual is listed according to age. (B) Boxplots show estimated
proportions of each cell type between 3 age groups from CIBERSORTx results. A 2-sided Wilcoxon test was performed for the comparison of bulk
RNA-seq data for young (<40) and old (>60) groups. (C) Correlation graph between age and expression of thyroid function-related genes (TG, PAX8). R
and P values were calculated using Pearson’s correlation. (D) Heat map shows the expression of 65 age-correlated genes derived from bulk RNA-seq
data. (E) Correlation graph between age and MT gene family (MT1F, MT1G) expression. R and P values were calculated using Pearson’s correlation. (F)
Correlation graph between MT and thyroid function-related genes. R and P values were calculated using Pearson’s correlation. (G) The expression of
MT genes after iodine treatment of Nthy-ori 3-1 cells. (H) The measurement of thyroxine in the cultured medium after iodine treatment of Nthy-ori 3-1
cells. (I) Dot plot of correlation among age, MT genes, and thyroid function-related genes in the TCGA and GTEx datasets. (G, H) One-way analysis of
variance was used for the statistical analysis. * P value <.05, ** P value <.01, *** P value <.001.
Abbreviations: GTEx, Genotype-Tissue Expression; MT, metallothionein; RNA-seq, RNA sequencing; TCGA, The Cancer Genome Atlas.
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023

Endocrinology, 2023, Vol. 164, No. 4 9
(Fig. 5D). Interestingly, among the 65 genes, the expression of
MT family genes positively correlated with age, primarily due
to the aging of thyroid epithelial cells (Fig. 5E). There were
strong correlations between MT and thyroid function-related
genes such as TG and PAX8 (Fig. 5F; Supplementary Fig. S6C
and 6D) (18). We analyzed associations between free T4 and
thyroid hormone production-related genes as well as MT
genes in patients who underwent bulk RNA seq, but no signi-
cant association was found (Supplementary Fig. S6E). This
suggests that various changes at the gene level occur in the nor-
mal thyroid gland rather than changes in thyroid hormone
caused by physiologic aging. In addition, the expressions of
MT genes signicantly upregulated in Nthy-ori 3-1 thyroid
follicular epithelial cells after the iodine treatments, along
with the increases of T3 and T4 in cell lysates (Fig. 5G and
5H). Increased expression of the MT genes and elevated levels
of T3 were observed in conditions with excess iodine, suggest-
ing that MT gene expression is increased to compensate for
the increased thyroid function (iodine uptake and hormone
synthesis) and that MT acts as an antioxidant, which corre-
sponds with previous ndings (46). The correlations of MT
and thyroid function-related genes were also validated in nor-
mal thyroid tissues from 2 databases, The Cancer Genome
Atlas and the GTEx (Fig. 5I). Collectively, based on bulk
RNA-seq and public databases, our results identied the asso-
ciation of MT family and aging, consistent ndings with
single-cell RNA-seq data.
Discussion
In this study, we compiled and analyzed a single-cell transcrip-
tomic atlas of human thyroid tissues to understand the mech-
anisms underlying aging. The atlas revealed that cell
type-specic features of aged thyroid were correlated with
known aging mechanisms, including oxidative stress response
and mitochondrial dysfunction. Under physiological condi-
tions in the thyroid, hydrogen peroxide, a highly reactive oxi-
dant, which together with oxidized iodine is used in the
peroxidation reaction, is crucial for thyroid hormone synthet-
ic process, playing multiple roles in thyrocyte biology (4, 5);
however, under pathological conditions, high ROS levels
may overcome the protective capacity of the aforementioned
enzymes leading to oxidative stress, resulting in thyrocyte
damage (4, 47, 48). Additionally, rapid changes in iodine in-
take may chemically stress thyrocytes and trigger an auto-
immune response (49). This inammatory reaction also
induces oxidative stress (4), consistent with increased mito-
chondrial dysfunction and oxidative phosphorylation signa-
tures in epithelial cells, broblasts, and myeloid cells in the
aged thyroid. Using single-cell RNA-seq, we identied key
mechanisms in the aging thyroid without autoimmune dis-
ease, consistent with previous ndings connecting aging and
mitochondria (50, 51).
Our data indicated the presence of an aging epithelial cell
subpopulation (C2), in which genes related to thyroid hor-
mone synthesis were downregulated, and the expression of
MT family genes was upregulated. The functions of MT genes
include detoxication of metals, including copper and zinc;
ROS scavenging; cell survival; angiogenesis; and proliferation
(52-54). Although MT genes are reportedly expressed in nor-
mal thyroid tissues (55-57), the role of MT in the human thy-
roid has not been fully understood. A previous study identied
the increase of MT1 expression in the liver of old mice
compared to young mice, with a reduction of T3, T4, and de-
creased secretion of zinc in the blood, suggesting increased
MT did not serve to maintain zinc homeostasis in aging
mice (58). Another study also showed an increased MT of
lymphocytes in peripheral blood mononuclear cells from eld-
erly patients, compared to young patients, with decreased bio-
availability of zinc ion and decreased NK cell activity despite
the increase of interleukin 6 and cortisol, which regulate MT
expression (59); however, no study reported on the role of MT
in the aged human thyroid. Our data revealed that expression
of the MT gene family in epithelial cells was correlated with
age, even in bulk RNA-seq. Upregulated MT expression
showed positive correlations with thyroid function-related
genes, such as TG and PAX8, suggesting that the MT gene ex-
pression is vital for thyroid homeostasis during physiological
aging. We also validated that an increase in MT gene expres-
sion protects thyrocytes from iodine stress in vitro. Despite the
protective effects of MT as antioxidants, a subset of epithelial
cells failed to improve thyroid function, making this subpopu-
lation vulnerable to aging. We have dened this subpopula-
tion as the major driving force of functional deterioration,
possibly leading to hypothyroidism in the elderly.
Although several studies have reported the presence of
MHC-positive thyroid epithelial cells, most studies focused
on their roles in autoimmune thyroid diseases such as
Graves’ disease and Hashimoto’s thyroiditis (60-62).
Thyroid-specic overexpression of allogeneic MHC class I in
mice resulted in thyroid atrophy, growth delay, and hypothy-
roidism, along with distended and irregular rough endoplas-
mic reticulum (63). In addition, MHC class II-expressing
thyroid epithelial cells act as autoimmune triggers for the de-
struction of thyroid follicles post-immunization (64, 65). In
this study, we rst identied the presence of MHC class I-
and II-expressing epithelial cells in normal thyroid. The subset
of epithelial cells (cluster 4) predominantly expressed MHC
class II molecules, including HLA-DRA, DRB1, DRB5,
DQB1, and CD74, in association with TPO and TG.
Interestingly, increased expression of the MHC classes was ob-
served in the aged thyroid. Although proportions of diverse im-
mune cell types were maintained during aging, the increased
expression of MHC-I/II molecules in the epithelial cells may in-
duce inammatory responses in the aged thyroid.
TGF-β is a physiological regulatory factor for thyroid cell
differentiation and proliferation (66, 67). TGF-β is normally
expressed and secreted by thyrocytes, acting as a potent in-
hibitor of thyroid cell growth by inhibiting iodide uptake
and metabolism (67). TGF-β is also known to reduce sodium
iodine symporter gene expression in thyroid follicular cells via
decreased PAX8 DNA binding activity (68) and TG biosyn-
thesis inhibition in rat thyroid cell lines FRTL-5 (69). In the
human thyroid, aging activated TGF-β related ligand-receptor
interactions among epithelial, stromal, and immune cells.
Interestingly, most of the subpopulations of stromal and im-
mune cells in aged thyroid represented signicantly increased
TGF-β signaling-related genes. A previous study identied the
importance of TGF-β signaling in inltrative papillary thyroid
carcinoma, representing tumor invasion edge together with
lymphocytes or tumor-associated macrophages (70), high-
lighting the importance of the epithelial-stromal connections
in the thyroid. Since aged human thyroid exhibited brotic
changes with inltrated stromal cells, our results suggest
that TGF-β signaling may be involved in signicant extracellu-
lar signaling of tissue brosis.
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023

10 Endocrinology, 2023, Vol. 164, No. 4
In the absence of an evaluation of individual thyroid tissue
from healthy controls, we have analyzed gene expression pro-
les from the “normal” tumor-adjacent (3 cm) tissue in pa-
tients. Although we perform single-cell dissociation within
30 minutes after tissue sampling, the cell number and cell
type composition varied depending on patient and tissue qual-
ity, and parafollicular cells were not detected in our single-cell
dataset. Additionally, we could not identify the cell clusters
expressing SLC5A5 or CALCA due to single-cell dissociation
procedure or 3′ or 5′ messenger RNA method. Despite the
study’s limitations, our single-cell data identied aging mech-
anisms in the human thyroid gland, including TGF-β signaling
and mitochondrial dysfunction, and discovered increased ex-
pressions of MT family genes and MHC class II genes in aged
epithelial cells. In conclusion, we characterized the human
aged thyroid at the single-cell level, offering potential insights
into the stress response mechanisms in physiological aging.
Acknowledgments
This research was supported by a grant from the Korea Health
Technology R&D Project through the Korea Health Industry
Development Institute, funded by the Ministry of Health &
Welfare, Republic of Korea (Grant No. HR20C0025),
by the Korea Science and Engineering Foundation grant
funded by the Korean government (MOST) (Grant No.
2019M3E5D1A02068560), and a National Research
Foundation of Korea grant funded by the Korean government
(MEST) (Grant No. 2021R1C1C1011183) and (MSIT)
(Grant No. 2021R1C1C2010490). This research was also -
nancially supported by the Ministry of Trade, Industry, and
Energy, Korea, under the System Industry Infrastructure and
R&D Support Program (reference number P0009796) super-
vised by the Korea Institute for Advancement of Technology.
This work was supported by a research fund of Chungnam
National University in 2020. This work was supported by
The Health Fellowship Foundation.
Author Contributions
Y. E. K., H. J. K., B. S. K., M. S., and W.-Y. P. conceived the
study. Y. H., H. J. K., Y. E. K., S. Y., and S. E. L. participated
in data acquisition. J. W. C., H.-R. W., J.-R. K., and
H. K. were responsible for on-site specimen collection and la-
boratory testing quality evaluation. Y. H., H. J. K., Y. E. K.,
S. P., S.-K. K., J.-L. P., and I.-S. C. analyzed the data. S.-Y. K.,
J. H. L., S. P., K. H. K., and Y. S. J. provided advice on the
methodology. Y. H., H. J. K., and Y. E. K. drafted the
manuscript. B. S. K., M. S., J. M. K., and W.-Y. P. revised
the manuscript. All authors read and approved the nal
manuscript.
Disclosures
The authors have nothing to disclose.
Data Availability
Original data generated and analyzed during this study are in-
cluded in this published article or in the data repositories [The
accession number for the single cell RNA sequencing datasets
reported in this paper is GSE182416. The bulk RNA is in
GSE223765 (token: unqlmqyqbfsnhid)].
References
1. Maenhaut C, Christophe D, Vassart G, et al. Ontogeny, anatomy,
metabolism and physiology of the thyroid. In: Endotext.
MDText.com, Inc.; 2015:1-4.
2. Muller S, Liu SJ, Di Lullo E, et al. Single-cell sequencing maps gene
expression to mutational phylogenies in PDGF- and EGF-driven
gliomas. Mol Syst Biol. 2016;12(11):889.
3. Coscia F, Taler-Verc
ic
 A, Chang VT, et al. The structure of human
thyroglobulin. Nature. 2020;578(7796):627-630.
4. Žarkovic M. The role of oxidative stress on the pathogenesis of
graves’ disease. J Thyroid Res. 2012;2012:302537.
5. Karbownik M, Lewinski A. The role of oxidative stress in physio-
logical and pathological processes in the thyroid gland; possible in-
volvement in pineal-thyroid interactions. Neuro Endocrinol Lett.
2003;24(5):293-303.
6. Studer H, Peter HJ, Gerber H. Natural heterogeneity of thyroid
cells: the basis for understanding thyroid function and nodular goi-
ter growth. Endocr Rev. 1989;10(2):125-135.
7. Sellitti DF, Suzuki K. Intrinsic regulation of thyroid function by
thyroglobulin. Thyroid. 2014;24(4):625-638.
8. Boltze C, Brabant G, Dralle H, Gerlach R, Roessner A, Hoang-Vu
C. Radiation-induced thyroid carcinogenesis as a function of time
and dietary iodine supply: an in vivo model of tumorigenesis in
the rat. Endocrinology. 2002;143(7):2584-2592.
9. Šošic-Jurjevic B, Filipovic B, Miloševic V, Nestorovic N, Negic N,
Sekulic M. Effects of ovariectomy and chronic estradiol administra-
tion on pituitary–thyroid axis in adult rats. Life Sci. 2006;79(9):
890-897.
10. LiVolsi VA. The pathology of autoimmune thyroid disease: a re-
view. Thyroid. 1994;4(3):333-339.
11. Dayan CM, Daniels GH. Chronic autoimmune thyroiditis. N Engl J
Med. 1996;335(2):99-107.
12. Gillotay P, Shankar M, Haerlingen B, et al. Single-cell transcrip-
tome analysis reveals thyrocyte diversity in the zebrash thyroid
gland. EMBO Rep. 2020;21(12):e50612.
13. Biondi B, Cooper DS. The clinical signicance of subclinical thyroid
dysfunction. Endocr Rev. 2008;29(1):76-131.
14. Gesing A, Lewinski A, Karbownik-Lewinska M. The thyroid gland
and the process of aging; what is new? Thyroid Res. 2012;5(1):16.
15. Bowers J, Terrien J, Clerget-Froidevaux MS, et al. Thyroid hor-
mone signaling and homeostasis during aging. Endocr Rev.
2013;34(4):556-589.
16. Mariotti S, Franceschi C, Cossarizza A, Pinchera A. The aging thy-
roid. Endocr Rev. 1995;16(6):686-715.
17. Rosato E, Salsano F. Immunity, autoimmunity and autoimmune
diseases in older people. J Biol Regul Homeost. 2008;22(4):217.
18. Hong Y, Kim H, Park S, et al. Supplementary data for: Single cell ana-
lysis of human thyroid reveals the transcriptional signatures of aging
information. Uploaded February 20, 2022. Figshare. Doi:106084/
m9gshare21316995
19. Wolock SL, Lopez R, Klein AM. Scrublet: computational identica-
tion of cell doublets in single-cell transcriptomic data. Cell Syst.
2019;8(4):281-291.e9.
20. Stuart T, Butler A, Hoffman P, et al. Comprehensive integration of
single-cell data. Cell. 2019;177(7):1888-1902.e21.
21. Korsunsky I, Millard N, Fan J, et al. Fast, sensitive and accurate in-
tegration of single-cell data with harmony. Nat Methods. 2019;16-
(12):1289-1296.
22. Efremova M, Vento-Tormo M, Teichmann SA, Vento-Tormo R.
CellPhoneDB: inferring cell-cell communication from combined ex-
pression of multi-subunit ligand-receptor complexes. Nat Protoc.
2020;15(4):1484-1506.
23. Tirosh I, Izar B, Prakadan SM, et al. Dissecting the multicellular
ecosystem of metastatic melanoma by single-cell RNA-seq.
Science. 2016;352(6282):189-196.
24. Martinez FO, Gordon S, Locati M, Mantovani A. Transcriptional
proling of the human monocyte-to-macrophage differentiation
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023

Endocrinology, 2023, Vol. 164, No. 4 11
and polarization: new molecules and patterns of gene expression. J
Immunol. 2006;177(10):7303-7311.
25. Newman AM, Steen CB, Liu CL, et al. Determining cell type abun-
dance and expression from bulk tissues with digital cytometry. Nat
Biotechnol. 2019;37(7):773-782.
26. De Bosscher K, Beck IM, Dejager L, et al. Selective modulation of
the glucocorticoid receptor can distinguish between transrepression
of NF-κB and AP-1. Cell Mol Life Sci. 2014;71(1):143-163.
27. Porlan E, Vidaurre OG, Rodríguez-Peña A. Thyroid hormone
receptor-beta (TR beta 1) impairs cell proliferation by the transcrip-
tional inhibition of cyclins D1, E and A2. Oncogene. 2008;27(19):
2795-2800.
28. Chen Y, Ma X, Zhang M, et al. Gadd45a regulates hematopoietic
stem cell stress responses in mice. Blood. 2014;123(6):851-862.
29. Johnson D, Walker C. Cyclins and cell cycle checkpoints. Annu Rev
Pharmacol Toxicol. 1999;39(1):294-312.
30. Magimaidas A, Madireddi P, Maifrede S, Mukherjee K, Hoffman
B, Liebermann DA. Gadd45b deciency promotes premature senes-
cence and skin aging. Oncotarget. 2016;7(19):26935-26948.
31. Mern DS, Hoppe-Seyler K, Hoppe-Seyler F, Hasskarl J, Burwinkel
B. Targeting Id1 and Id3 by a specic peptide aptamer induces
E-box promoter activity, cell cycle arrest, and apoptosis in breast
cancer cells. Breast Cancer Res Treat. 2010;124(3):623-633.
32. Alani RM, Young AZ, Shifett CB. Id1 regulation of cellular senes-
cence through transcriptional repression of p16/Ink4a. Proc Natl
Acad Sci U S A. 2001;98(14):7812-7816.
33. Wong Y-C, Wang X, Ling M-T. Id-1 expression and cell survival.
Apoptosis. 2004;9(3):279-289.
34. Yamaguchi F, Takata M, Kamitori K, et al. Rare sugar D-allose in-
duces specic up-regulation of TXNIP and subsequent G1 cell cycle
arrest in hepatocellular carcinoma cells by stabilization of p27kip1.
Int J Oncol. 2008;32(2):377-385.
35. Chen W, Li S, Zhang F. Role of lncRNA XIST/microRNA-19/
PTEN network in autophagy of nucleus pulposus cells in interver-
tebral disc degeneration via the PI3K/Akt signaling pathway. Cell
Cycle. 2021;20(17):1629-1641.
36. Kong Y, Huang T, Zhang H, et al. The lncRNA NEAT1/miR-29b/
Atg9a axis regulates IGFBPrP1-induced autophagy and activation
of mouse hepatic stellate cells. Life Sci. 2019;237:116902.
37. Wang P, Chen X, Wang Y, et al. Essential role for autophagy pro-
tein VMP1 in maintaining neuronal homeostasis and preventing
axonal degeneration. Cell Death Dis. 2021;12(1):1-14.
38. Rock KL, Reits E, Neefjes J. Present yourself! By MHC class I and
MHC class II molecules. Trends Immunol. 2016;37(11):724-737.
39. Neefjes J, Jongsma ML, Paul P, Bakke O. Towards a systems under-
standing of MHC class I and MHC class II antigen presentation.
Nat Rev Immunol. 2011;11(12):823-836.
40. Schmid KW, Greeff M, Hittmair A, et al. Metallothionein expres-
sion in normal, hyperplastic, and neoplastic thyroid follicular and
parafollicular C cells using monoclonal antimetallothionein anti-
body E9. Endocr Pathol. 1994;5(2):114-122.
41. Mori K, Yoshida K, Hoshikawa S, et al. Effects of perinatal expos-
ure to low doses of cadmium or methylmercury on thyroid hormone
metabolism in metallothionein-decient mouse neonates.
Toxicology. 2006;228(1):77-84.
42. Zhang N, Wang L, Duan Q, et al. Metallothionein-I/II knockout
mice aggravate mitochondrial superoxide production and peroxir-
edoxin 3 expression in thyroid after excessive iodide exposure.
Oxid Med Cell Longev. 2015;2015:267027.
43. Denham M, Wills E. A clinico-pathological survey of thyroid glands
in old age. Gerontology. 1980;26(3):160-166.
44. Gordon MB, Klein I, Dekker A, Rodnan GP, Medsger TA Jr.
Thyroid disease in progressive systemic sclerosis: increased fre-
quency of glandular brosis and hypothyroidism. Ann Intern
Med. 1981;95(4):431-435.
45. Davidson S, Coles M, Thomas T, et al. Fibroblasts as immune reg-
ulators in infection, inammation and cancer. Nat Rev Immunol.
2021;21(11):704-717.
46. Houessinon A, François C, Sauzay C, et al. Metallothionein-1 as a
biomarker of altered redox metabolism in hepatocellular carcin-
oma cells exposed to sorafenib. Mol Cancer. 2016;15(1):1-10.
47. Eng PH, Cardona GR, Fang S-L, et al. Escape from the acute
Wolff-Chaikoff effect is associated with a decrease in thyroid so-
dium/iodide symporter messenger ribonucleic acid and protein.
Endocrinology. 1999;140(8):3404-3410.
48. Sies H. Biological redox systems and oxidative stress. Cell Mol Life
Sci. 2007;64(17):2181-2188.
49. Ruiz-Riol M, Martinez-Arconada M, Alonso N, et al.
Overexpression of metallothionein I/II: a new feature of thyroid fol-
licular cells in graves’ disease. J Clin Endocrinol Metab. 2012;97(2):
446-454.
50. Asa SL, Mete O. Oncocytic change in thyroid pathology. Front
Endocrinol. 2021;12:437.
51. Cho BA, Yoo S-K, Song YS, et al. Transcriptome network analysis
reveals aging-related mitochondrial and proteasomal dysfunction
and immune activation in human thyroid. Thyroid. 2018;28(5):
656-666.
52. Swindell WR. Metallothionein and the biology of aging. Ageing Res
Rev. 2011;10(1):132-145.
53. Pitt BR, Schwarz M, Woo ES, et al. Overexpression of metallothio-
nein decreases sensitivity of pulmonary endothelial cells to oxidant
injury. Am J Physiol. 1997;273(4):L856-L865.
54. Andrews GK. Regulation of metallothionein gene expression by
oxidative stress and metal ions. Biochem Pharmacol. 2000;59(1):
95-104.
55. Nartey N, Cherian MG, Banerjee D. Immunohistochemical local-
ization of metallothionein in human thyroid tumors. Am J
Pathol. 1987;129(1):177.
56. Liu Z-M, van Hasselt CA, Song F-Z, et al. Expression of functional
metallothionein isoforms in papillary thyroid cancer. Mol Cell
Endocrinol. 2009;302(1):92-98.
57. Ferrario C, Lavagni P, Gariboldi M, et al. Metallothionein 1G acts
as an oncosupressor in papillary thyroid carcinoma. Lab Investig.
2008;88(5):474-481.
58. Mocchegiani E, Giacconi R, Cipriano C, et al. Metallothioneins (I
+II) and thyroid–thymus axis efciency in old mice: role of cortico-
sterone and zinc supply. Mech Ageing Dev. 2002;123(6):675-694.
59. Mocchegiani E, Giacconi R, Cipriano C, et al. MtmRNA gene ex-
pression, via IL-6 and glucocorticoids, as potential genetic marker
of immunosenescence: lessons from very old mice and humans.
Exp Gerontol. 2002;37(2-3):349-357.
60. Migita K, Eguchi K, Otsubo T, et al. Cytokine regulation of HLA on
thyroid epithelial cells. Clin Exp Immunol. 1990;82(3):548-552.
61. Kasuga Y, Matsubayashi S, Akasu F, Miller N, Jamieson C, Volpé
R. Effects of recombinant human interleukin-2 and tumor necrosis
factor-α with or without interferon-γ on human thyroid tissues
from patients with graves’ disease and from normal subjects xeno-
grafted into nude mice. J Clin Endocrinol Metab. 1991;72(6):
1296-1301.
62. Menconi F, Monti MC, Greenberg DA, et al. Molecular amino acid
signatures in the MHC class II peptide-binding pocket predispose to
autoimmune thyroiditis in humans and in mice. Proc Natl Acad Sci
U S A. 2008;105(37):14034-14039.
63. Frauman AG, Chu P, Harrison LC. Nonimmune thyroid destruc-
tion results from transgenic overexpression of an allogeneic major
histocompatibility complex class I protein. Mol Cell Biol.
1993;13(3):1554-1564.
64. Li Y-S, Kanamoto N, Hataya Y, et al. Transgenic mice producing
major histocompatibility complex class II molecules on thyroid cells
do not develop apparent autoimmune thyroid diseases.
Endocrinology. 2004;145(5):2524-2530.
65. Kimura H, Kimura M, Tzou S-C, et al. Expression of class II major
histocompatibility complex molecules on thyrocytes does not cause
spontaneous thyroiditis but mildly increases its severity after im-
munization. Endocrinology. 2005;146(3):1154-1162.
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023

12 Endocrinology, 2023, Vol. 164, No. 4
66. Roberts A, Sporn M. The transforming growth factor-βs. In:
Peptide Growth Factors and Their Receptors I. Springer;
1991:419-472.
67. Pang X-P, Park M, Hershman JM. Transforming growth factor-
beta blocks protein kinase-A-mediated iodide transport and protein
kinase-C-mediated DNA synthesis in FRTL-5 rat thyroid cells.
Endocrinology. 1992;131(1):45-50.
68. Costamagna E, García B, Santisteban P. The functional inter-
action between the paired domain transcription factor Pax8
and Smad3 is involved in transforming growth factor-β
repression of the sodium/iodide symporter gene. J Biol Chem.
2004;279(5):3439-3446.
69. Kang H-C, Ohmori M, Harii N, Endo T, Onaya T. Pax-8 is essen-
tial for regulation of the thyroglobulin gene by transforming growth
factor-β1. Endocrinology. 2001;142(1):267-275.
70. Eloy C, Santos J, Cameselle-Teijeiro J, Soares P, Sobrinho-Simões
M. TGF-beta/Smad pathway and BRAF mutation play different
roles in circumscribed and inltrative papillary thyroid carcinoma.
Virchows Arch. 2012;460(6):587-600.
Downloaded from https://academic.oup.com/endo/article/164/4/bqad029/7040488 by KRIBB(Kor Res Inst of Biosci & Biotech) user on 15 March 2023