High lymphocyte signature genes expression in parathyroid endocrine cells and its downregulation linked to tumorigenesis

Summary

Background

To date, because of the difficulty in obtaining normal parathyroid gland samples in human or in animal models, our understanding of this last-discovered organ remains limited.

Methods

In the present study, we performed a single-cell transcriptome analysis of six normal parathyroid and eight parathyroid adenoma samples using 10 × Genomics platform.

Findings

We have provided a detailed expression atlas of parathyroid endocrine cells. Interestingly, we found an exceptional high expression levels of CD4 and CD226 in parathyroid endocrine cells, which were even higher than those in lymphocytes. This unusual expression of lymphocyte markers in parathyroid endocrine cells was associated with the depletion of CD4 T cells in normal parathyroid glands. Moreover, CD4 and CD226 expression in endocrine cells was significantly decreased in parathyroid adenomas, which was associated with a significant increase in Treg counts. Finally, along the developmental trajectory, we discovered the loss of POMC, ART5, and CES1 expression as the earliest signature of parathyroid hyperplasia.

Interpretation

We propose that the loss of CD4 and CD226 expression in parathyroid endocrine cells, coupled with an elevated number of Treg cells, could be linked to the pathogenesis of parathyroid adenoma. Our data also offer valuable information for understanding the noncanonical function of CD4 molecule.

Funding

This work was supported by the National Key R&D Program of China (2022YFA0806100), National Natural Science Foundation of China (82130025, 82270922, 31970636, 32211530422), Shandong Provincial Natural Science Foundation of China (ZR2020ZD14), Innovation Team of Jinan (2021GXRC048) and the Outstanding University Driven by Talents Program and Academic Promotion Program of Shandong First Medical University (2019LJ007).

Research in context.

Evidence before this study

The parathyroid gland, one of the last organs to be discovered in the body, remains poorly understood due to difficulties in obtaining living samples. Consequently, previous single-cell studies were based exclusively on tumour samples. There is still no single-cell data available on the normal parathyroid gland. Our knowledge of this tiny yet crucial organ’s physiology and pathology is limited, with minimal information available on its cellular composition, specific molecular markers, receptors, and secretion spectrum, aside from its role in monitoring serum calcium levels and releasing parathyroid hormone (PTH) to regulate calcium and phosphorus metabolism. Furthermore, the pathogenesis of parathyroid adenomas remains unclear due to the challenges in procuring normal or para-tumour samples, despite extensive research on the subject.

Added value of this study

This study performed a single-cell RNA sequencing analysis on upper and lower pairs of the parathyroid gland and parathyroid adenomas. Our findings suggest a different function between the upper and lower parathyroid glands. Also, we discovered an elevated expression of CD4 and CD226 in parathyroid endocrine cells. Furthermore, this atypical expression of CD4 and CD226 were significantly decreased in tumoral cells, suggesting their involvement in the pathogenesis of parathyroid adenomas. Additionally, our data also offer valuable information for understanding the noncanonical function of CD4 molecule.

Implications of all the available evidence

This study provided a detailed single-cell transcriptomic atlas of both normal and tumoral parathyroid glands and confirmed many important observations made previously. Furthermore, the elevated expression levels of lymphocyte markers in parathyroid endocrine cells encourage further exploration into the role of these markers in the physiology of endocrine cells and pathogenesis of endocrine tumours.

Introduction

The parathyroid gland is the last major organ to be recognized in the human body. Two pairs of parathyroid glands are usually located behind the thyroid, each about the size of a soybean. Despite its vital role in the regulation of serum calcium levels by secreting parathyroid hormone (PTH), little is known about this tiny gland because of the difficulty in obtaining fresh specimens from humans and rodent models because of its tiny size (about 500 μm for rats and 50 μm for mice).

According to available literature, the two pairs of parathyroids in humans originate from different branchial pouches, with the lower pair developing with the thymus. It is generally believed that they are equivalent, but no relevant study has been conducted so far. The parathyroid parenchyma consists mainly of two types of cells according to haematoxylin and eosin staining: chief and oxyphil cells. Chief cells are responsible for the secretion of PTH and are characterized by high levels of chromogranin A (CHGA), calcium-sensing receptor (CaSR), and PTH. Oxyphil cells also express PTH, CaSR and CHGA, but with larger cytoplasm and more eosinophilic staining.¹ Ritter et al. suggest that oxyphil cells produce more PTH-like hormone (PTHLH) and calcitriol but less PTH,¹ while others suggest that they express higher levels of CaSR than chief cells² and are responsible for secondary hyperparathyroidism.^3,4 Whether functional differences exist between chief and oxyphil cells remains to be investigated.

Some bulk transcriptomic and proteomic studies have suggested expression signatures of parathyroid chief and oxyphil cells,5, 6, 7 but these studies were based on chief and oxyphil adenoma samples using whole tissue lysates. More recently, a single-cell RNA sequencing (scRNA-seq) study provided an expression atlas of the cell heterogeneity and tumour environment of parathyroid adenomas,⁸ but there is still no data available on the normal parathyroid gland.

Furthermore, it is not known whether there exist other specialized endocrine cell subpopulations within the parathyroid gland, as endocrine glands are usually multifunctional. Studies suggest that the parathyroid gland may also synthesize PTHLH, CHGA, leptin and calcitriol using immunohistochemical approaches.^1,9,10 However, the significance of these molecules in the parathyroid gland has not been fully elucidated, and our understanding of its cell composition, specific molecular markers, receptors, and secretion spectrum remains limited. Additionally, always due to the difficulty in obtaining normal fresh parathyroid samples, the pathogenesis of parathyroid adenomas is also unclear, little is known about the biological marker and the expression signature of the parathyroid adenoma types.

In the present study, we carried out a single-cell transcriptomic study of parathyroid glands obtained from six healthy samples and eight parathyroid adenoma samples. Our data highlights the expression atlas of normal parathyroid at single cell levels and the different expression patterns observed between the upper-lower pairs of parathyroid glands. More importantly, we found an elevated expression of CD4 and CD226 in parathyroid endocrine cells and a lack of CD4 positive lymphocytes in normal parathyroid glands. Furthermore, we observed a loss of CD4 and CD226 expression in tumoral parathyroid endocrine cells, which may be linked to pathogenesis of parathyroid adenomas.

Methods

Ethics approval and consent to participate

All procedures were reviewed and approved by the Ethics Committee for the Investigation of Human Subjects in the Shandong provincial hospital SWYX No.2022-048, No.2022-049 and No.2022-1018. Written informed consent was obtained from each patient prior to sample collection.

Human specimens

A total of six normal parathyroid samples (three normal upper glands, three lower glands from four healthy donors) and eight parathyroid adenoma samples collected at Shandong Provincial Hospital were included in our study. Sex and age were confirmed by reviewing the donors’ and patients’ medical records.

The overall health status of the brain-dead donors was confirmed for liver, kidney and heart transplantation. We examined their blood calcium and phosphorus levels from previous medical records. The status of the parathyroid glands was also examined by the surgeons upon removal to confirm the absence of aberrancies.

Patients with parathyroid adenoma were selected according to the diagnostic criteria of parathyroid adenoma¹¹ (hypercalcemia, high PTH levels and Tc-99m scan). The specimens were also confirmed through immunohistochemical analysis after removal. We have excluded multiglandular hyperplasia, atypical parathyroid tumour and parathyroid carcinoma. To ensure the representativeness of samples, we proceeded to digest half of the gland.

Single-cell RNA-sequencing

Normal and tumour parathyroid samples were dissociated to single-cell suspension using Liberase TM (Roche, China). The cell suspension was subjected to the Gel Bead Kit V3 (10× Genomics, Pleasanton, CA) for library preparation and sequenced on Illumina NovaSeq 6000 Systems using paired-end sequencing (150 nt), and aligned against GRCh38 human reference genome using Cellranger V6.

Quality control of single-cell RNA-seq data

Seurat (v4.1.0) was applied for quality control procedures. Cells that fit any of the following criteria were considered as low-quality cells: the proportion of mitochondrial gene counts (>20%), Unique Molecular Identifiers (UMIs) < 500 or UMIs > 20,000, expressed genes < 500 or expressed genes > 20,000. After that, the batch effect correction of scRNA-seq data were applied as indicated in each analysis as appropriate.

Unsupervised clustering and identification of signature genes for cell clusters

Filtered single-cell sequencing data were Log transformed using the NormalizeData() function in Seurat. Harmony was used to correct the batch effect of scRNA-seq data. The FindNeighbors() and FindClusters() functions in Seurat were both applied for cell clustering. To find the optimal cluster resolution, RunTSNE() and RunUMAP() function were both performed for visualization when appropriate. And the differentially expressed genes (DEGs) in each subcluster were identified through the built-in function of Seurat FindAllMarkers().

Pseudo-time trajectory analysis

The trajectory analysis was performed using the Monocle3 package (v3.1.0) to reveal the cell evolution trajectory. The batch effect correction was performed by built-in align algorithm as recommended by Monocle3. The Uniform Manifold Approximation and Projection (UMAP) method in Monocle3 was applied to reduce the dimensions with default settings. As the normal parathyroid cells have been clearly separated and localized in the right side of UMAP projection plot in our study, we set these cells as the root state.

Cell–cell interaction analysis

To investigate the potential interactions between different cell types, we performed cell–cell interaction analysis using CellChat, Interaction networks between all these cell clusters were investigated. The significant cell type-specific interactions between L-R pairs (p < 0.05) were selected for evaluation and visualization.

SCENIC analysis

Single-Cell rEgulatory Network Inference and Clustering (SCENIC) analysis was performed to reveal the gene regulatory network in PTH⁺ cell clusters. The gene-motif rankings (500 bp upstream or 100 bp downstream of the transcription start site) were used to determine the search space around the transcription start site.

Immunohistochemical staining

Fresh samples were fixed in paraformaldehyde and embedded in paraffin. 5-μm sections of parathyroid adenomas were mounted on glass slides. After deparaffinization and hydration, slides were boiled in Tris–EDTA buffer at 95 °C for 15 min and treated with 3% H₂O₂ solution for 10 min to block endogenous peroxidase activities. The slides were then incubated with 5% normal goat serum for 1 h at room temperature and incubated overnight at 4 °C respectively with anti-PTH (Abcam, ab154792, RRID: AB_3094549), anti-CD4 (Abcam, ab133616, RRID: AB_2750883) or anti-CD226 (Abcam, ab212076, RRID: AB_3094550) primary antibody at 1:150 in TBST buffer. After washing, the slides were incubated with HRP-conjugated goat anti-rabbit secondary antibody (Abcam, ab6721, RRID: AB_955447) for 1 h. Finally, the slides were revealed by DAB and stained with haematoxylin following the manufacturer’s instructions.

Statistical analysis

All of the statistical analyses in the histograms were performed using GraphPad prism. Results are expressed as means ± S.E.M. Significance was assessed using unpaired Student’s t test or by Mann–Whitney U test when data did not meet normal distribution. Statistical significance was defined as p value < 0.05.

Role of funders

The funders had no role in study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication.

Results

Single-cell transcriptomic atlas of human parathyroid glands

To obtain a comprehensive understanding of the parathyroid glands, we performed single-cell RNA-seq (10X Genomics) to profile six normal parathyroid glands samples from four healthy brain-dead donors and eight adenoma samples from patients with parathyroid adenoma(Fig. 1a) according to the parathyroid adenoma diagnostic criteria¹¹ and histochemical diagnosis following parathyroidectomy (Supplementary Fig. S1a). We excluded multiglandular hyperplasia, atypical parathyroid tumour. In addition to the pathological classification, we also indicated its anatomical location (Fig. 1a) and clinical data (Supplementary Tables S1 and S2).

Fig. 1.

The expression atlas of the normal parathyroid gland. (a) Schematic representation of the healthy donors and patients included in analysis. (b) The expression of marker genes for the major clusters. (c) UMAP plot showing the major clusters of cells in normal parathyroid gland. (d) UMAP plot showing the major clusters of PTH⁺ cells. (e) The top expressed genes for each cluster of PTH⁺ cells from healthy donors. (f) UMAP plot showing the expression profile of key genes in PTH⁺ cells from healthy donors. (g) Heatmap showing the DEGs in PTH⁺ cells from healthy donors between the upper and lower gland. (h) Expression levels of CD4, CD226 and their related genes in different cell population within normal parathyroid gland. (i) Immunohistochemical analysis of CD4 and CD226 in paraffin-embedded normal parathyroid gland sections. Top right panel is the no primary antibody negative control. (j) Splicing junction analysis of parathyroid endocrine cells derived CD4 transcripts. (k) UMAP plot showing the major clusters of lymphoid cells in normal parathyroid gland. (l) The top expressed genes for each cluster of lymphoid cells in normal parathyroid gland. (m) UMAP plot showing the expression of lymphoid cell signature genes.

We first focused on the normal parathyroid gland samples. A total of 60,325 single cells passed the quality filtering. After correcting the batch effect using Seurat V4¹² and Harmony,¹³ we primarily applied uniform manifold approximation and projection (UMAP) clustering and identified 7 main cell populations, which were labelled as parathyroid endocrine cells, (PTH, CaSR, GCM2, VDR), lymphoid cells (CD3E, CXCR4, NKG7), neutrophils (FCGR3B, CXCR2, CMTM2), myeloid cells (LYZ, S100A8, S100A9, CD14), smooth muscle cells (ACTA2, TAGLN, CNN1), fibroblasts (COL1A1, COL1A2, TAGLN, ACTA2) and endothelial cells (PECAM1, CD34, VWF) (Fig. 1b and c). In addition, we explored the expression profiles in each cluster to confirm whether there are unreported cell populations with endocrine functions in parathyroid glands. However, no other cell population was found.

Single-cell transcriptomic atlas of parathyroid endocrine cells under physiological condition

As previously reported, the parathyroid endocrine cells also showed high expression levels of other endocrine molecules such as CHGA, PTHLH, POMC and leptin.^1,9,10 However, we did not find leptin expression in the parathyroid endocrine cells. Additionally, we found that the parathyroid endocrine cells also showed high expression levels of secretory molecules such as pleiotrophin (PTN) and chemerin (RARRES2). The full list of parathyroid endocrine cell specific expression genes was shown in Supplementary Table S3.

We next re-clustered the PTH-expressing cells (Fig. 1d). To avoid the influence of sex difference, the parathyroid cells from four female donors were selected. Despite the significant difference in morphologies between chief and oxyphil cells, we did not find two distinct groups of cells in the DEGs plot (Fig. 1d). Cluster 5 was found to be localized separately from the main clusters, but it did not exhibit any unique gene expression (Fig. 1e). Cluster 2 and 3 may represent the oxyphil cell population because they showed a higher expression level of the previously reported oxyphil cell marker PTHLH (Fig. 1f). However, CaSR and VDR, two previously reported differentially expressed genes between chief and oxyphil cells were found homogeneous across clusters (Fig. 1f). This observation indicates that there may not be a fundamental difference between these two types of cells at the transcriptional level. Additionally, we noticed a small subpopulation of cells (cluster 11) that express both endothelial cell and parathyroid cell markers (CD34⁺ VIM⁺ CaSR⁺ GCM2⁺) (Fig. 1e and f).

We next investigated the difference between upper and lower gland. In addition to the embryogenic marker homeobox genes gene (HOXB-AS3, HOXB4-9), we also observed a differential expression pattern in PTH-expressing cells between upper and lower glands (Fig. 1g and Supplementary Fig. S1b). Clusters expressing LPL, FABP5 (C2 and C7) and metallothionein (e.g., MT1X) were mainly from lower parathyroid gland (Fig. 1f and Supplementary Fig. S1b). GSEA enrichment analysis¹⁴ showed that the metallothionines bind metals, retinoid metabolism and chloride transmembrane transport are more enriched in lower parathyroid glands (Supplementary Fig. S1c), indicating a different function in zinc homeostasis and lipid metabolism between upper and lower glands.

Finally, we investigated the sex difference of parathyroid glands. However, we did not find much difference between sex, except for some Y- or X-chromosome associated genes (Supplementary Fig. S1b).

Single-cell transcriptome reveal a lack of CD4 T cell in the normal parathyroid gland but a high expression of CD4 in PTH secreting cells.

Interestingly, we found that the parathyroid endocrine cells showed a high expression level of CD4 and CD226, which normally should not be expressed in non-immune cells (Fig. 1h and Supplementary Fig. S1d). We found their expression levels were even higher than in lymphocytes (Fig. 1h and Supplementary Fig. S1d). We have confirmed the localization of CD4 and CD226 protein in PTH-expressing cells using immunohistochemical approach (Fig. 1i). This anti-CD4 and anti-CD226 antibody have also been assayed on the human lymph node section to confirm its specificity (Supplementary Fig. S1e). Additionally, we have detected all the splicing junctions of full-length CD4 transcript in endocrine cells using STAR-Solo analysis, suggesting the presence the full-length CD4 molecule (Fig. 1j).

We next analysed a set of lymphocyte genes that related to CD4 and CD226 molecules such as CD4 upstream transcription factors and CD4 downstream signalling molecule in all types of cells in the parathyroid. We found endocrine cells showed a stronger expression of CD4 T cell master regulators GATA3 and ThPOK (ZBTB7B) than lymphocytes, which was associated with a decreased CD4 silencers expression such as RUNX3 (Fig. 1h). Although PTH-secreting cells are CD4 positive, these cells are CD3 and LCK negative (Fig. 1h). Instead, we found a strong expression of a newly discovered intracellular partner of CD4, SPG21, in the endocrine cells.

Additionally, we have remarked an exceptionally low count of CD4⁺ T cells in the normal parathyroid tissues (Supplementary Fig. S1d). We subsequently investigated lymphoid cells within normal parathyroid tissues. The re-clustering of lymphoid cells showed ten subpopulations (Fig. 1k). The expression profile of each cluster was presented in Fig. 1l. We have identified two natural killer (NK) cell clusters by the high expression levels of NKG and GNLY. In CD8⁺ clusters, we have identified the CD8-RGS1, CD8-LMNA, CD8-DUSP1 and CD8-GZMK clusters using known marker genes (Fig. 1k and l). However, we did not find canonical CD4⁺ cell populations (i.e., CD4 CCR7, CD4 IL10, CD4 FoxP3) as previously reported in single-cell analysis. Instead, we identified a large proportion of CD4 and CD8 double negative population expressing CCR7 or CTLA4 (Fig. 1m). Due to the lack of CD4 expression, the remaining T clusters were marked by their high expression levels of CCR7(Fig. 1k). Overall, our data of lymphoid cells showed a low proportion of CD4 T cell in the parathyroid gland.

The different expression atlas in parathyroid endocrine cells between normal and adenoma tissues

We next assayed the parathyroid adenomas, we selected all parathyroid hormone secreting cells (identified by high expression levels of PTH, CaSR, CHGA and GCM2) from healthy donors and adenomas. Re-clustering revealed 7 distinct clusters (Fig. 2a). We have noted that cells from 6 different normal samples were homogeneously projected in the same cluster, indicating a good batch effect correction of our data. However, each adenoma sample was projected into a unique cluster, indicating a high degree of individual heterogeneity in adenomas.

Fig. 2.

The differential gene expression between normal and tumoral parathyroid endocrine cells. (a) UMAP plot showing the clusters of PTH⁺ cells in normal and tumoral parathyroid gland. (b) Heatmap showing the CD4, CD226 and the classical parathyroid tumour markers expression in normal and tumoral parathyroid gland. (c) Representative immunohistochemical image of tumoral parathyroid gland sample and the possible para-tumoral rim is indicated by black arrow. (d) Heatmap showing the top 50 DEGs in PTH⁺ cells between normal and tumoral parathyroid endocrine cells. (e) Enrichment analysis of DEGs between normal and tumoral parathyroid endocrine cells. (f) UMAP plot showing the possible para-adenomas (remnant) clusters of PTH⁺ cells. (g) Heatmap showing the DEGs in parathyroid endocrine cells between normal, remnant and tumoral parathyroid clusters. (h) SCENIC analysis showing the regulon activity in parathyroid endocrine cells from normal, remnant and adenoma clusters.

We firstly assayed the previously reported parathyroid adenomas markers including MEN1, parafibromin (CDC73), CASR and APC.¹¹ However, we did not find significant differential expression of these genes between normal and tumoral samples due to the heterogeneity within the adenoma samples (Fig. 2b). In contrast, our results demonstrated a significant decrease in the expression of CD4 and CD226 genes in tumoral endocrine cells, with almost complete loss of expression in certain adenoma samples (Fig. 2b). Also, we confirmed the decreased CD4 and CD 226 expression in adenomas tissues compared to the rim of normal tissue using immunohistochemical approach (Fig. 2c).

We next investigated the transcriptional changes between cells from normal donors and adenomas. Differentially gene expression analysis showed an upregulation of LPL, CCND1, FABP5, HMOX1, and a downregulation of CD226, PTH, ALAS1 and CD4 in tumoral parathyroid endocrine cells. We also observed an upregulation of amyloidogenic peptide transthyretin (TTR) and amyloid precursor (APP) in endocrine cells of adenoma samples, suggesting a possible amyloid deposition in the tumorigenesis of parathyroid adenomas (Fig. 2d).

Metascape enrichment analysis (metascape.org) showed that the reactive oxygen species (ROS), phosphorus metabolic process and VEGF signalling pathways were highly enriched in adenoma clusters. We also noticed that the gonadotropin-releasing hormone receptor pathway was enriched, suggesting a possible role of such pathways in the pathogenesis of parathyroid adenomas (Fig. 2e). We have also remarked an upregulated PPAR signalling pathway using Gene Set Enrichment Analysis (GSEA) analysis (Supplementary Fig. S2a).

Normally, to avoid the bias introduced by individual differences, differentially gene expression analysis should be performed between adenoma and para-adenoma tissues. Although a rim of normal tissue in parathyroid adenomas can occasionally be seen in immunohistochemical images, the technical limitations of parathyroid adenoma surgery make it impossible to collect this tiny layer of para-adenomas tissues. Here, we noticed that a fraction of cells from adenoma samples were projected within the normal cell clusters in the t-Distributed Stochastic Neighbor Embedding (t-SNE) or UMAP plot, we call this fraction of cells “remnant”, which may represent the remaining normal endocrine cells (Fig. 2f).

We next performed a differential gene expression analysis between the normal cell cluster, the remnant cluster and the adenoma cluster. We found that the remnant cluster showed an increase in transthyretin (TTR), transcobalamin (TCN1), and Cyclin D1 (CCND1) but a decreased in the rate-limiting enzyme in heme biosynthesis ALAS, DNA damage stress sensor GADD45A and tumour suppressor DNAJB4 compared to normal cluster (Fig. 2g). Enrichment analysis showed increased interferon signalling, copper homeostasis and steroid metabolic pathways in remnant cluster (Supplementary Fig. S2b). Between the remnant and adenoma clusters, we found that the adenoma clusters showed an increased levels of S100A1, HPGD, LPL and decreased levels of CD226 (Fig. 2g).

To gain insight into the mechanisms underlying cellular heterogeneity between normal cells and adenomas or between different types of adenomas, we performed a single-cell regulatory network inference and clustering (SCENIC) analysis¹⁵ to predict potential transcription factors driving transitions between normal and tumour states. We found increased activity of the early growth response (EGR 1–4), vitamin D binding protein (DBP), and SOX2, SOX4 regulons in adenoma cells (Fig. 2h and Supplementary Fig. S2c), as well as inhibition of the tumour suppressor regulon interferon regulatory factor 1 (IRF1).

Cell communication analysis reveals a significant alteration of CD4 and CD226 signalling between normal and tumoral parathyroid endocrine cells

The cell–cell signalling network is crucial for maintaining proper tissue function. To further investigate this, we conducted a comprehensive analysis of cell communication networks in normal parathyroid and adenoma tissue, using the CellChat package¹⁶ for quantification and comparison.

In normal parathyroid tissue, CellChat analysis on these cells identified parathyroid endocrine cells as the dominant communication cells in normal parathyroid tissues, which secrete and receive the greatest number of signal molecules (Fig. 3a). In adenoma tissue, a dominant communication was found between the myeloid cells and parathyroid endocrine cells, suggesting a macrophage and monocyte mediated immunoreaction in parathyroid adenomas (Fig. 3b and c). We also found an increased communication between myeloid cells and endothelial cells, suggesting an increased recruitment of myeloid cells from blood vessels. No remarkable increase in T/NK cell communication was found (Fig. 3b and c). These data may suggest that the innate immune system, but not the adaptive immune system, plays a major role in the immune response of parathyroid adenoma.

Fig. 3.

Cell communication network in normal and tumoral parathyroid glands. (a) Heatmap showing the relative importance of each cell cluster based on the computed signalling network. (b) Weight of significant ligand–receptor pairs between any pair of two cell populations in normal and adenoma samples. (c) The differential ligand–receptor pairs between any pair of two cell populations in normal and adenoma samples. (d) The significant ligand–receptor pairs that contribute to the signalling sending from endocrine cells to other clusters or vice versa in normal parathyroid samples. (e) Relative differences of overall information flow of significant signalling pathways in normal and adenoma samples. (f) The differential ligand–receptor pairs that contribute to the signalling sending from endocrine cells to other clusters or vice versa between normal parathyroid samples.

We have discovered several auto-signalling ligand–receptor pairs that are specific for normal parathyroid endocrine cells, in particular, the CD226-nectin pair (Fig. 3d left), which cannot be detected in adenoma samples (Fig. 3d right), indicating that it appears to be critical for maintaining normal parathyroid cell function. Furthermore, we have found several other auto-signalling pairs in normal parathyroid endocrine cells, including the pleiotropin-nucleolin (PTN-NCL), CADM1-CADM1, and E-cadherin (CDH–CDH) ligand–receptor pairs (Fig. 3d). We also detected a possible HLA-CD4 signalling that flowed from myeloid cells to parathyroid endocrine cells (Supplementary Fig. S3a). In contrast, our analysis did not reveal significant CD4 signalling within T/NK cell populations, but we did detect CD8 signalling (Supplementary Fig. S3b and c).

We next conducted a detailed analysis of context-specific signalling pathways. The significant signalling pathways were selected based on differences in the overall information flow between normal parathyroid tissues and adenomas. Overall, our analysis uncovered that normal parathyroid tissue enriched the nucleolin, CD226, NEGR, SELL, CDH1, TENASCIN, and IFN signalling pathways, while adenomas tissues enriched the VISFTIN, THY, SELE, TNF, and TGFβ signalling pathways (Fig. 3e). More precisely, we identified significant upregulation of macrophage migration inhibitory factor (MIF) signalling from parathyroid cells to myeloid cells. MIF is a pleiotropic cytokine that arrests ‘random’ immune cell movement and promotes ‘directed’ cell migration, suggesting the recruitment of immunocytes in adenoma tissues. Also, VEGF signalling exhibited significant enhancement from parathyroid cells to endothelial cells in adenoma samples in adenoma samples, indicating enhanced angiogenesis (Fig. 3f).

Moreover, we found that the endocrine cell specific CD226-nectin, PTN-NCL, CADM1-CADM1, and E-cadherin auto-signalling ligand–receptor pairs were dispersed in tumoral endocrine cells (Fig. 3f). All of them were reported to play a critical role in cell adhesion, tumour development and angiogenesis, suggesting a potential role in the pathogenesis of this disease. We observed a noteworthy upregulation of CD8 signalling in T/NK cells, indicating an activation of CD8 immunity (Supplementary Fig. S3d and e). We also found a downregulation of HLA-CD4 signalling that flowed from myeloid cells to parathyroid endocrine cells and a shift from integrin α2/β1 to α5/β8 signalling in adenoma samples (Fig. 3f). That may be due to the decrease in CD4 and integrin α2 in tumoral parathyroid endocrine cells (Fig. 2d).

Additionally, we detected a significant increase in amyloid precursor protein signalling from parathyroid cells to both immunocytes and endothelial cells in adenoma samples (Fig. 3f), which is coherent with the increased amyloidogenic genes expression in tumoral endocrine cells (Fig. 2d).

scRNA-seq analysis showed a significant immunosuppression environment in adenoma tissues

Next, we aimed to depict a more detailed immune cell atlas of normal and tumoral parathyroid glands. In contrast to the PTH-secreting cells, immune cells showed less individual heterogeneity. Cells of all samples were projected into a single cluster, no patient-specific cluster was found. The re-clustering of lymphoid cells of all samples revealed 8 clusters (Fig. 4a), including two NK cell clusters (cluster 3, 4), two clusters of CD4⁺ T cells (cluster 5, 6), two clusters of CD8⁺ T cells (0, 2), one double negative cluster (cluster 1) and one cycling cell cluster (cluster 7) (Fig. 4b and Supplementary Fig. S4a). Within these clusters, we found that the adenoma samples showed significantly higher proportions of cells in cluster 5 and 7, which correspond to Treg and cycling T cells, respectively (Fig. 4c).

Fig. 4.

Immune environment between normal and tumoral parathyroid. (a) UMAP plot showing the clusters of lymphoid cells in normal and tumoral parathyroid samples. (b) The top expressed genes for each cluster of lymphoid cells. (c) The normalized proportion of cells from normal and tumoral parathyroid samples contributing to each cluster. (d) Heatmap showing the top 50 DEGs in lymphoid cells between normal and tumoral parathyroid samples. (e) The enriched pathways of upregulated genes in lymphoid cells from tumoral parathyroid samples. (f) The normalized proportion of cells that expressing classical T cell subset markers from normal and tumoral parathyroid samples. (g) UMAP plot showing the clusters of myeloid cells in normal and tumoral parathyroid samples. (h) The top expressed genes for each cluster of myeloid cells. (i) Heatmap showing the top 50 DEGs in myeloid cells between normal and tumoral parathyroid samples. (j) The enriched pathways of upregulated genes in tumoral parathyroid samples. (k) The normalized proportion of cells from normal and tumoral parathyroid samples contributing to each cluster. (l) The enriched pathway in each myeloid cell clusters in normal and tumoral parathyroid samples. Histograms are expressed as mean ± S.E.M. (n = 6 normal versus 8 tumour samples), ∗p < 0.05, ∗∗p < 0.01 by Mann–Whitney U test.

Differential gene expression analysis revealed an upregulation of lymphotoxin β (LTB), CD74, GZMK, CD52 and a downregulation in heat shock proteins in total lymphoid cells from adenoma samples compared to normal samples (Fig. 4d). Enrichment analysis showed that the regulation of T cell activation, and the oxidative phosphorylation pathway were highly enriched (Fig. 4e). Robust rank aggregation (RRA) of enriched pathway of each cluster also reveals an upregulation of oxidative interferon response and oxidative phosphorylation pathway in adenomas samples (Supplementary Fig. S4b). We subsequently counted the number of cells expressing the classical T cell subgroup markers. In line with the above-mentioned observations, we found a significant increase in CD4⁺ helper cell and decreased CD8 cytotoxic T cell counts in the adenomas. Moreover, these observations were associated with an increased count of cells expressing hallmarkers characteristic Treg and naïve T cells (Fig. 4f).

scRNA-seq analysis of normal and tumoral parathyroid glands revealed a significant downregulation of CD163 and increased counts of dendritic cells

We next focused on myeloid cells between normal and tumoral samples. A total of 12 clusters were found, comprising 8 clusters for macrophages, two for monocytes, one for dendritic cells, and one for cycling cells (Fig. 4g). Within the macrophage populations, we observed a gradient in the expression of CCLs. Macrophage clusters expressing high levels of CCL2-4 termed Macro CCL⁺ clusters (clusters 0, 1, 6). In contrast, the Macro CCL⁻ clusters (cluster 2, 3) were characterized by notably decreased expression levels of CCL genes (Fig. 4h). We identified one population of macrophages (clusters 4) that expressed high levels of LYVE1 and MRC1 (CD206), which may represent tissues-resident macrophages. This cluster also showed higher expression levels of M2 marker CD163 (Fig. 4h). Additionally, we remarked a lack of MACRO and ARG1 expressing macrophages in all parathyroid samples (Supplementary Fig. S4c). For the dendritic cell (DC) cluster, we identified DC2 cells marked by high CD1C expression. However, we were unable to detect the frequently mentioned DC subtype in the literature that expresses high levels of XCR1 or LAMP3 (Supplementary Fig. S4c). The two monocytes were termed as classical monocytes and FCGR3A monocytes by high levels of S100A8, FCGR3A expression, respectively (Fig. 4g and Supplementary Fig. S4c).

Differential gene expression analysis revealed an upregulation of HLA-DRB5, P2RY13, APOC1, CD14, APOE and a downregulation of M2 marker CD163 (Fig. 4i). Enrichment analysis showed that myeloid cells from adenoma samples were characterized by an increase in antigen processing and presentation (Fig. 4j). In addition, we noticed a significant increase in cell proportions of cluster 1 (macrophage) and 5 (DC) in adenoma samples (Fig. 4k). RRA analysis of each cluster indicated that the cluster 1 of tumoral samples showed increased IL6 and TGFβ signalling but absence of interferon response (Fig. 4l). The DC cluster in both normal and tumoral samples showed decreased interferon response IL6 and TGFβ signalling (Fig. 4l).

Developmental trajectory defines distinct states of adenomas

To dynamically dissect the evolution and molecular signatures during the normal-adenoma sequence, we applied Monocle3 and its built-in batch effect correction algorithm to perform a single-cell trajectory analysis. Our trajectory analysis showed three major developmental trajectories, with the normal cell cluster on the left side and adenoma clusters on the right side. We therefore started the trajectory on the left side of the normal cell cluster and then bifurcated into the main branches on the right (Fig. 5a). Of these three branches, the chief cell adenomas were predominantly distributed at the end of the lower branch and the oxyphil adenomas were predominantly distributed in the upper branch (Supplementary Fig. S5a). Consistent with this distribution, cells from patient 7 (mixed adenomas) were distributed in both branches (Supplementary Fig. S5a). The middle branch, shared by normal and adenoma samples, appeared to be in an apoptotic state, reflected by high proportion of mitochondrial genes (Supplementary Fig. S5b and c). Moreover, the presence of lipomas in patient 5 did not significantly alter its evolutionary trajectory, cells from this patient were mainly located in the chief cell adenoma trajectory (Supplementary Fig. S5a).

Fig. 5.

Developmental trajectory defines distinct states of adenomas. (a) UMAP pseudotime-ordered analysis of PTH⁺ cells from normal and adenoma samples, insert in the lower-left is the UMAP plot of PTH expression. (b) Enrichment analysis of DEGs between the upper branch of adenoma cell and normal parathyroid endocrine cell cluster in UMAP pseudotime plot. (c) Enrichment analysis of DEGs between the lower branch of adenoma cell and normal parathyroid endocrine cell cluster in UMAP pseudotime plot. (d) UMAP pseudotime-ordered plots showing the expression profile of key genes in PTH⁺ cells. (e) Representative UMAP pseudotime-ordered plots and immunohistochemical image of patient 1. (f) Representative UMAP pseudotime-ordered plots and immunohistochemical image of patient 6. (g) Representative UMAP pseudotime-ordered plots and immunohistochemical image of patient 4. (h) Schematic presentation of the development of parathyroid adenomas.

We also performed the differential gene expression and enrichment analysis on the cells between the different branches. We identified an upregulation of SNAP25, INA, CHGB, and “don’t eat me” signal CD24 as a signature of chief cell adenoma end stage markers while S100A1, HPGD, LPL as oxyphil cell adenoma end stage markers (Fig. 5b and Supplementary Fig. S5c). Enrichment analysis suggests that the cholesterol metabolism and adipogenesis pathway genes were enriched in the upper branch (Fig. 5c), while vitamin D receptor pathway genes were enriched in the lower branch (Fig. 5d). Inflammation response pathways were found in the both branches. It should be noted that despite dramatic changes in gene expression along the tumour evolutionary trajectory, adenoma cells still maintained an overall high expression level of parathyroid cell markers. Furthermore, we did not detect the expression of known EMT-driven transcription factors (Snail, Twist1/2, Zeb1/2) or proliferation markers (MKI67) in adenoma clusters, suggesting a non-malignant nature of parathyroid adenoma (Fig. 5b).

Next, we characterized specific marker genes to define adenoma precursor cell population. Consistent with cell communication analysis results, we found a gradual loss of CD226 expression during developmental trajectory from normal cell to adenoma (Fig. 5d). More interestingly, we found the loss of expression of POMC, ADP-ribosyltransferase 5 (ART5), carboxylesterase 1 (CES1) and RvD1 receptor GPR32 was the earliest gene expression signature of adenoma precursor cell (Fig. 5b and Supplementary Fig. S5c).

In addition to the overall trajectory map, we then separately depicted the evolutionary trajectory of individual patients. To avoid sex bias and embryonic origin bias, we matched patients to normal donor samples based on sex and the upper or lower gland of tumour origin. Cells from all patients with adenoma, except patient 4, show two evolutionary trajectories: the adenoma trajectory and the apoptotic trajectory (Fig. 5e and f). In addition, the endothelial-like cluster was found in the trajectory between branch 1 and 2, mainly from patient 4. Correspondingly, we detected an endothelial or follicular-like structure using immunohistochemistry in patient 4, a secondary evolutionary route to the endothelium-like cell route emerged in patient 4 at the end of adenoma trajectory (Fig. 5g).

In conclusion, along the developmental trajectory, we discover the loss of POMC, ART5, CES1 and GPR32 expression as the earliest signature of parathyroid adenomas. We also identified two sets of signature genes corresponding to the trajectory to chief and oxyphil parathyroid adenomas (Fig. 5h).

Discussion

The parathyroid gland is one of the last organs to be discovered in our body. Up to now, our understanding of gene expression, cell composition and the function of this organ is still very limited. Therefore, the first question we want to explore in the present study is the cell composition and expression atlas of this gland. We want to confirm whether there are any yet undiscovered new types of endocrine cells within the parathyroid gland, or whether these PTH-secreting endocrine cells have functions that we have not yet discovered. Here, we provide a comprehensive single-cell transcriptomic atlas to characterize normal and tumoral parathyroid glands. In the present study we found that the parenchymal cells of the parathyroid gland were mainly composed of PTH⁺ cells (parathyroid endocrine cells), endothelial cells, fibroblasts and smooth muscle cells. No other clusters of cells with a specific expression profile were found, suggesting that the parathyroid may not contain other types of cells that have not been identified. Although adipocyte is often seen in parathyroid sections under microscope, we did not find adipocyte cluster in our data, which may be due to their loss during digestion and centrifugation processes. Moreover, we found that the PTH⁺ cells also expressed high levels of pleiotrophin, chemerin and POMC, a subpopulation of cells expressed PPY suggesting that the PTH⁺ cells have the potential to secrete related hormones.

We have detected some cells with “hybrid” expression features, especially the cluster of CD34⁺ PTH⁺ cells showed a mixed type of endothelial and parathyroid cells. The presence of these hybrid cells may not be attributed to the encapsulation of two cells into one sequencing reaction volume, commonly known as “doublets”. Because this type of cell has been previously reported by Corbetta et al. by immunohistochemical imaging.¹⁷ Additionally, apart from this cluster of CD34⁺ PTH⁺ double positive cells, the number of other types of “hybrid” cells was significantly lower than CD34⁺ PTH⁺ double positive cells.

CD34 is first identified on hematopoietic stem and progenitor cells.¹⁸ However, strong evidence demonstrates CD34 is expressed by a number of other nonhematopoietic progenitors cell including muscle satellite cells, epithelial progenitors, and vascular endothelial progenitors.¹⁸ Corbetta et al. found that the CD34⁺ PTH⁺ double positive cells proliferated, though at a low rate, and may go through endothelial lineage differentiation,¹⁷ suggesting that these cells may not represent parathyroid progenitor cells but may be involved in its spontaneous angiogenesis of parathyroid. Therefore, in our trajectory analysis, we did not start the trajectory from this population of cells.

Despite the significant difference in morphology between chief cells and oxyphil cells, in the normal PTH⁺ cell population, no significantly different clusters were found either in tSNE or in UMAP plot. This observation suggests that the chief and oxyphil cells may represent two different statuses of the same types of cells. In contrast, we found significant differences in the expression profiles between the upper and lower pairs of glands, particularly in the expression of genes related to lipid metabolism and metallothionine metabolism, which were significantly higher in the lower gland than in the upper gland. This is also reflected in the anatomical features of the lower glands, for example, the lower gland has a higher fat content, is lighter in color and is often enveloped by a fat capsule.¹⁹ In addition, our recent ultrasound imaging study has shown that the lower glands are hyperechoic in ultrasonography whereas the upper glands are hypoechoic, which also suggests a higher fat content in the lower glands.¹⁹ Whether the lower glands have additional functions in regulating lipid and metallothionein metabolism needs further investigation.

One of the key findings of our present study is the atypical expression of lymphocyte markers such as CD4 in parathyroid endocrine cells. CD4 is a classical surface marker of T helper cells, and its expression was traditionally thought to be exclusively confined to immune cells, particularly lymphocytes and certain macrophage populations. The extracellular domains of CD4 synergize with the T-cell receptor (TCR) to engage with the major histocompatibility complex class II (MHC II) molecules, whereas its tiny transmembrane and intracellular signalling domains are situated at the C-terminal region. To evaluate the presence of functional CD4 molecule in endocrine cells, we firstly confirmed the presence of full-length CD4 transcript by using STAR-solo splicing analysis. However, due to the technique limit of 10X genomic sequencing, we cannot really quantify the proportion of full-length CD4 transcript. Nevertheless, our immunohistochemical photo confirmed a membrane localization of CD4 and CD 226 staining in the PTH expressing cells, indicating the parathyroid endocrine cells may equipped functional CD4 and CD226 molecules. Hellman et al. in 90s have reported that parathyroid tissue express a “CD4-like” protein using immunohistochemical approaches²⁰ and they also reported a “CD3-like” protein in parathyroid tissue. Here, we confirmed that parathyroid endocrine cells express the same CD4 as T cells (at least at transcriptional levels). However, we did not detect any CD3 subunit in parathyroid endocrine cells. Also, we did not detect the expression of canonical signal transduction molecule LCK tyrosine kinase. Instead, we found a high expression of SPG21. These observations suggest that the CD4 protein in parathyroid endocrine cells may initial a noncanonical intracellular signalling cascade, or alternatively, the primary role of CD4 in parathyroid endocrine cell may reside in mediating extracellular interactions with other cells, particularly the immune cells.

Recently, Young et al. reported that activated mouse Th2 cells may also produce PTH, which acted as a cytokine for local cell communication.²¹ The expression of CD4 and PTH in these two cell types may potentially be associated with the expression of GATA3 in these two cell types. Evidence suggests that the transcription factor GATA3 plays a crucial role in the differentiation of Th2 cells, and mutations in the GATA3 gene have been found to cause hypoparathyroidism and abnormalities in parathyroid gland development.²²

Interestingly, we also searched the public single-cell sequencing database to investigate the expression pattern of CD4 and CD226 in other types of endocrine cells. Although the expression of lymphocyte signature has not been reported in other cells, we found a CD4 and CD226 expression in some other cells according to the database of Human Protein Atlas (proteinatlas.org ²³), such as certain hepatocytes, Muller cells, pancreatic somatostatin cells, testis Leydig and Sertoli cells, enteroendocrine cells (parathyroid showed the highest expression levels). Therefore, this atypical expression of T cell markers in parathyroid endocrine may not simply due to the fact that the parathyroid and thymus shared a common embryonal progenitor. We can hypotheses that CD4 expression may have an undiscovered role.

In the present study, we did not further investigate the role of CD4 and CD226 expression in parathyroid endocrine cells. To this objective, tissues specific CD4 or CD226 knockout mice are needed. Although there are global CD4 or CD226 knockout mice, to the best of our knowledge, no study has investigated the parathyroid glands in these mice, because the main objective of using these mice is to study transplantation or the immune system. It is also very challenging to obtain samples of such small size glands (about 50 μm) from mice without transgenic fluorescent protein labelling.²⁴ At least until now, there have been no reports mentioning the presence of impaired calcium-phosphorus metabolism in these genetically defective mice, suggesting that CD4 or CD226 deficiency is unlikely to affect PTH secretion function.

In our study, another unusual observation associated with the high expression levels of CD4 and CD226 in parathyroid endocrine cells is the almost absence of CD4 T helper cells in the normal parathyroid glands. Lymphocytes have been documented to express the type 1 PTH receptor (PTH1R) and patients suffering from primary or secondary hyperparathyroidism have elevated levels of PTH and immune alterations,²⁵ suggesting physiological levels of PTH may be necessary for a proper immune response. In vitro studies with high concentrations of PTH have shown a dose-dependent decreases in helper/suppressor ratio.²⁵ Within the parathyroid glands, the presence of locally extremely high concentration of PTH may strongly affect immune cell functions. Whether the presence of CD4 and CD226 in the can prevent the infiltration of CD4 positive T cells in parathyroid glands, or the parathyroid environment favour the maturation of double negative T cells, needs further investigation.

The second question we want to investigate in the present study is the difference between normal and tumoral parathyroid endocrine cells. Previous genomic studies suggest that parathyroid adenoma is associated with mutations in MEN1, a potent tumour suppressor.^26,27 However, this form of hereditary parathyroid adenoma accounts for only 5–10% of patients and typically occurs at a younger age. Furthermore, carriers of mutations in MEN1 are usually associated with the development of other neuroendocrine tumours such as pancreatic and duodenal, pituitary, adrenal and gastric cancers.²⁸ Our understanding of the pathogenesis of the different subtypes of parathyroid adenoma is still limited due to the heterogeneity of this adenoma. Previous immunohistochemical and transcriptomic studies have also identified some key genes involved in PTH regulation, cell cycle regulation, mitochondrial function or apoptotic pathways, such as CDC73, APC, EZH2, CASR, EZH1, ZFX, mTOR, and cyclin D1, which may play important roles in the pathogenesis of parathyroid adenomas.^1,26,29, 30, 31 However, these classical techniques cannot balance the specificity and high throughput with comprehensive analysis of all types of cells in normal parathyroid glands and tumours. So far, the progression of a parathyroid adenoma from a normal gland to an adenoma remains unclear, and atypical parathyroid adenomas may progress to carcinoma without appreciable differences from typical adenomas.³² In addition, the limited availability of living samples from this tiny organ has hampered our knowledge of the normal parathyroid gland itself. Therefore, it is critical to improve our understanding of the genetic and transcriptomic signatures that differ between normal parathyroid glands and adenomas.

The second key finding of our study is that the CD4 and CD226 expression significantly decreased in tumoral endocrine cells. CD226 is an adhesion molecule belonging to the immunoglobulin superfamily and is known to be expressed on the majority of immune cells. It competes with TIGIT for activating anti-tumour responses. Recently, CD226 has been identified as a novel inducible adhesion molecule on human endothelial cells.³³ The loss of cell surface CD226 has been attributed to TGF-β1 signalling and hypoxia. Knockdown of CD226 not only increases HIF-1α translation but also promotes the phosphorylation of ERK and AKT levels.³⁴ Both AKT and ERK signalling are aberrantly activated in a wide range of human cancers.³⁵

Furthermore, our single-cell analysis revealed that parathyroid adenomas showed an immunosuppressive environment, particularly the increase in Treg count, despite the heterogeneity within the different adenoma samples. T_reg cells are crucial for initiating tumour immune evasion. The increased Treg may be due to the uncontrolled secretion of PTH in parathyroid adenomas, as exposed to high levels of PTH stimulates the Treg development,³⁶ and surgical hypoparathyroidism showed decreased CD4⁺ T, CD4 T reg and CD4⁺ naïve T cell.³⁷ Consist with this, we found an increased total CD4⁺ T cells and CD4⁺ naïve T cells in parathyroid adenoma samples. These observations indicated that the supra-physiological concentrations of PTH within parathyroid tissues significantly affect immune cell homeostasis, and over-secretion of PTH may also played a crucial role in the pathogenesis of parathyroid adenomas.

Moreover, programmed death-ligand 1 (PD-L1) is considered a crucial immune checkpoint protein. Previous studies have suggested that decreased PD-L1 expression plays a key role in the development of parathyroid adenomas.³⁸ However, our study revealed that PD-L1 expression was very low in both normal and tumour al parathyroid, indicating that PD-L1 may not be involved in the pathogenesis of parathyroid adenomas. In contrary, anther immune checkpoint protein CD24 was found to be highly enriched in chief cell adenomas, suggesting a potential immune evasion mechanism.

Although the parathyroid adenomas showed a high degree of individual variation, each patient’s cells had its own unique expression profile. For example: high expression of the prostaglandin synthesis gene (HPGD), ectopic expression of the upper and lower gland decision gene HOXB, high expression of lipid metabolism genes or endothelial cell markers. Nevertheless, we found several common markers across tumour types, all of which showed loss of expression of POMC, CD4 and CD226. Consistent with our findings, a previous study has reported the down-regulation of POMC in parathyroid adenomas samples.³⁹ Our single-cell study provides further evidence that the downregulation of POMC expression is located in PTH⁺ cells. Our data also highlighted CD4, CD226, ART5 and GPR32, which have not been reported, as suggestive biomarkers that were potentially involved in adenoma initiation and progression.

The results from our pseudo-time analysis revealed that the loss of expression of the POMC, CES1 and APOD genes were prior to that of CD4 and CD226. It is highly possible that these genes function as decisive elements in the determination of cellular fate at the onset of tumour growth. However, the regulation of the expression of these three genes belong to three different regulons, we do not yet have direct experimental evidence for their upstream signalling. Remarkably, the hallmark of genes involved in cell proliferation, migration, vessel formation and EMT-driven transcription factors were expressed at very low levels in adenoma samples, suggesting a non-malignant nature of parathyroid adenoma. However, we observed a potential endothelial evolutionary trajectory of adenoma cells. This may be linked to the inherent spontaneous angiogenesis characteristic of parathyroid cells, especially linked to the small proportion of parathyroid-derived CD34⁺ cells, which co-expressed both endothelial progenitors and parathyroid specific genes.

In the present study, we provided a single-cell transcriptomic atlas of normal parathyroid from brain-dead donors and parathyroid adenomas. We highlighted those parathyroid endocrine cells showed high expression levels of CD4 and CD226. We also discovered that this atypical expression of lymphocyte markers can be observed in diverse types of endocrine cells. Furthermore, loss of CD4 and CD226 expression in parathyroid endocrine cells is potentially associated with immune dysfunction and tumorigenesis. This phenomenon serves as a common hallmark among diverse types of endocrine cells. Additionally, loss of POMC, APOD, ART5 and CES1 expression as the earliest signature of parathyroid adenomas. These observations encourage further exploration into the role of these markers in the development and progression of endocrine tumours.

Our study has several limitations. Firstly, the sample size in our study is limited and the sex is not balanced due to the parathyroid adenomas are more prevalent in females. Secondly, apart from healthy donor samples, adenoma samples were collected over a period of two years. Though we applied the Harmony, the built-in algorithm of Monocle3 or Cellranger V6 for batch effect correction, as appropriate. It is still a challenge to fully correct the batch effect and artifacts may be introduced. Thirdly, due to the limitations of the indications for surgical treatment of parathyroid adenomas, only those that affect calcium and phosphorus homeostasis are surgically removed. Therefore, our adenoma samples only contain those functioning tumours that cause hyperparathyroidism. Further in vivo and in vitro experiments are required to validate these findings in the future.

Contributors

Conceptualization, J.Z. and Y.S.; Resources, C.G. and K.L.; Methodology, L.S., J.L.; Investigation, B.G., P.G., B.W., C.G. and J.L.; Formal Analysis, C.G. and J.L.; Validation, L.S., Q.W., J.Z. and Y.S.; Visualization, B.G., P.G., C.G. and J.L.; Writing-Original Draft, C.G. and J.L.; Writing-Review & Editing, L.S., Q.W., J.Z. and Y.S.; Supervision, L.S., Q.W., J.Z. and Y.S. All authors read and approved the final version of the manuscript. In addition, C.G., and J.L. have verified the underlying data, Y.S. was responsible for the decision to submit the manuscript.

Data sharing statement

The data reported in this paper have been deposited in the OMIX, China National Center for Bioinformation (https://ngdc.cncb.ac.cn/omix: accession no. OMIX005492).

Declaration of interests

The authors declare no competing interests.

Appendix ASupplementary data

Supplementary Fig. S1.

Fig. S1. The expression atlas of the normal parathyroid gland related toFig. 1. (a) Representative immunohistochemical images of parathyroid adenoma samples included in this study, part of the same image of patient 6 is used in figure 5f. (b) Heatmap showing DEGs in PTH⁺ cells from healthy donors. (c) Enrichment analysis of DEGs between upper and lower parathyroid endocrine cell clusters. (d) UMAP plot showing the expression of T cell signature genes in the major population of normal parathyroid gland cells. (e) Immunohistochemical analysis of CD4 and CD226 in paraffin-embedded human lymph node sections.

Supplementary Fig. S2.

Fig. S2. The differential gene expression between normal and tumoral parathyroid endocrine cells related toFig. 2. (a) Enrichment analysis of DEGs between normal and tumoral parathyroid endocrine cell clusters. (b) Enrichment analysis of DEGs between normal and “remnant” parathyroid endocrine cell clusters. (c) SCENIC analysis showing the regulon activity in PTH⁺ cells from each parathyroid gland samples.

Supplementary Fig. S3.

Fig. S3. Cell communication network in normal and tumoral parathyroid glands related toFig. 3. (a) Signalling sending from other clusters to endocrine cells in normal parathyroid samples. (b) Signalling sending from T/NK cluster to other clusters in normal parathyroid samples. (c) Signalling sending from other clusters to T/NK cluster in normal parathyroid samples. (d) Differential signalling sending from T/NK cluster to other clusters between normal parathyroid samples and adenomas samples. (e) Differential signalling sending from other clusters to T/NK cluster between normal parathyroid samples and adenomas samples.

Supplementary Fig. S4.

Fig. S4. Immune environment between normal and tumoral parathyroid related toFig. 4. (a) UMAP plots showing the expression profile of key genes in lymphocyte in normal and tumoral parathyroid glands. (b) The enriched pathway in each lymphoid cell clusters in normal and tumoral parathyroid samples. (c) UMAP plots showing the expression profile of key genes in myeloid cells in normal and tumoral parathyroid glands.

Supplementary Fig. S5.

Fig. S5. Developmental trajectory defines distinct states of adenomas related toFig. 5. (a) UMAP pseudotime-ordered analysis of PTH⁺ cells from normal and adenoma samples coloured by donor and patient ID. (b) UMAP pseudotime-ordered analysis of PTH⁺ cells from normal and adenoma samples coloured by cells states. (c) Heatmap showing the DEGs in different cell states, the different branches in trajectory maps as indicated in (B).