Single-cell perspective on the Monocyte-to-HDL cholesterol ratio as a metastasis biomarker in papillary thyroid cancer

Abstract

Background

Papillary thyroid carcinoma (PTC) is a globally widespread inflammation-related cancer, where lymph node metastasis (LNM) poses significant challenges to the prognosis of PTC. The role of the monocyte-to-high-density lipoprotein cholesterol ratio (MHR), a novel inflammation marker attracting increasing attention, in PTC remains unclear.

Methods

Clinical data analysis was adopted to preliminarily explore the relationship between clinical features and LNM of PTC. Single-cell RNA sequencing (scRNA-seq) data from the GSE191288 and GSE193581 datasets were integrated to analyze various single-cell infiltration in PTC. A comprehensive suite of in vitro experiments, encompassing immunohistochemistry, co-culture, and Transwell assays were conducted to elucidate the regulatory role of macrophages and PTC.

Results

Clinical data analysis confirmed MHR could be an independent risk factor for LNM (OR = 1.76, 95%CI: 1.20–2.88, P = 0.011). The single-cell analysis identified cell clusters associated with dysregulated cholesterol homeostasis Q1 and CD68 + C3 macrophage subpopulations, as well as their markers GDF15. Further analysis confirmed that they are closely related to PTC metastasis and malignancy, which also implied a significant correlation between MHR and LNM. The in vitro experiments demonstrated PTC may promote metastasis by mediating inducing macrophage differentiation to the M2 phenotype.

Conclusion

This study revealed the potential role of MHR in PTC from the single-cell perspective for the first time. Combined with the results of clinical studies and basic experiments, it was confirmed that mononuclear/macrophage and cholesterol homeostasis significantly promoted the lymph node metastasis of PTC. Overall, these findings informed robust support for MHR as an emerging marker for preoperative LNM prediction of PTC.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-025-14373-9.

As one of the most common endocrine malignant tumors, the incidence of thyroid carcinoma has shown a rapid upward trend globally in the past 40 years, mainly due to the rapid increase in the incidence of papillary thyroid carcinoma (PTC) [1, 2]. After a complete resection of the primary tumor and metastatic lymph nodes, the prognosis of PTC patients is good, with a 10-year survival rate of 95% [3, 4]. Neck lymph node metastasis (LNM) is the main risk factor for disease recurrence in PTC patients, first metastasizing to the central region (region VI), and then to the lateral neck regions (regions II, III, IV, and V) [5]. Currently, surgical treatment has become the preferred treatment option for PTC, and almost all guidelines from various countries recommend performing a central lymph node dissection (CLND) on the affected side. However, there is still controversy over whether it is necessary to perform a preventive CLND on the contralateral central lymph nodes due to the existence of permanent hypoparathyroidism, nerve injury, and other surgical complications [6]. Postoperative pathological examination showed that about 30–80% of PTC patients had lymph node metastasis in the neck, with the central lymph nodes being the first to metastasize [7]. How to screen out potential LNM patients before surgery is a big challenge that needs to be solved urgently.

Previous studies have indicated that serum inflammatory factor levels are closely related to differentiated thyroid carcinoma [8]. There are also substantial studies supporting that pre-operative detection of these systemic inflammatory response markers can help predict the recurrence of PTC [9] and prognosis [10].

The monocyte-to-high-density lipoprotein cholesterol ratio (MHR) was first proposed in a study on cardiovascular adverse events in patients with chronic kidney disease and is considered a new biomarker that reflects the degree of inflammation and oxidative stress [11]. Currently, much of the research on MHR is focused on the heart, brain, and kidney [12], To date, there has been relatively little research on the relationship between MHR and thyroid disease. Recently, several researchers have indicated that the MHR serves as an independent risk factor for thyroid carcinoma [13], and it is significantly associated with both the presence and size of thyroid nodules, irrespective of gender [14], and the MHR holds promise as a novel biomarker for distinguishing between papillary thyroid carcinoma and benign thyroid nodules [15]. However, the specific impact of MHR in LNM within the thyroid gland remains to be elucidated and the single-cell transcriptomic characteristics, particularly the immune microenvironment in PTC, require further investigation [16, 17].

To address the aforementioned research gap, we conducted a retrospective case–control study and employed propensity score matching to further reduce bias, aiming to identify high-risk factors for LNM in PTC patients and obtain more robust evidence for clinical decision-making. Based on single-cell analysis, public database bioinformatics analysis, and in vitro experiments, we have confirmed from multiple dimensions that monocytes, cholesterol metabolism, and the malignancy and metastasis of thyroid cancer are significantly correlated, and MHR may serve as a novel biomarker for preoperative thyroid LNM.

Results

The relationship between clinical features and the central lymph node metastasis

The results of univariate analysis of clinical characteristics and relevant indicators included in the study showed Gender (P = 0.014), TC(P = 0.025), MHR (P = 0.021), Tumor Size (P = 0.007), FT3 (P = 0.05), and Hashimoto’s thyroiditis (P = 0.038), and ETE (P = 0.038) were associated with CLNM, and the P value of MHR(P = 0.01)and Tumor_Size(P = 0.006) remained statistically significant after propensity score matching (Table 1).

Table 1.

Univariate logistic analysis of clinical features of PTC lymph node metastases based on propensity score matching method

		N	Overall population		p-value	q-value	N	Propensity Score Matched population		p-value	q-value
			(-), N = 35	(+), N = 47				(-), N = 35	(+), N = 35
Age, n (%)		82			0.065	0.12	70			0.31	0.52
	≤ 55		21 (60%)	37 (79%)				21 (60%)	25 (71%)
	> 55		14 (40%)	10 (21%)				14 (40%)	10 (29%)
BMI (kg/m2), n (%)		81			0.96	0.96	69			0.73	0.73
	≤ 24		19 (56%)	26 (55%)				19 (56%)	21 (60%)
	> 24		15 (44%)	21 (45%)				15 (44%)	14 (40%)
Gender, n (%)		82			0.014	0.093	70			0.1	0.4
	male		6 (17%)	20 (43%)				6 (17%)	12 (34%)
	female		29 (83%)	27 (57%)				29 (83%)	23 (66%)
MHR, mean (SD)		78	2.78 (1.05)	3.53 (1.73)	0.021	0.093	66	2.78 (1.05)	3.81 (1.92)	0.01	0.073
TC (mmol/L), n (%)		82			0.025	0.093	70			0.11	0.4
	< 5.2		26 (74%)	27 (57%)				26 (74%)	21 (60%)
	< 6.2		9 (26%)	12 (26%)				9 (26%)	10 (29%)
	≥ 6.2		0 (0%)	8 (17%)				0 (0%)	4 (11%)
TG (mmol/L), n (%)		82			0.22	0.28	70			0.29	0.52
	< 1.7		22 (63%)	31 (66%)				22 (63%)	22 (63%)
	< 2.26		9 (26%)	6 (13%)				9 (26%)	5 (14%)
	≥ 2.26		4 (11%)	10 (21%)				4 (11%)	8 (23%)
LDL-C (mmol/L), n (%)		82			0.19	0.26	70			0.36	0.53
	< 3.37		32 (91%)	36 (77%)				32 (91%)	28 (80%)
	< 4.14		3 (8.6%)	9 (19%)				3 (8.6%)	5 (14%)
	≥ 4.14		0 (0%)	2 (4.3%)				0 (0%)	2 (5.7%)
HDL-C (mmol/L), n (%)		82			0.31	0.36	70			0.49	0.62
	< 1		6 (17%)	4 (8.5%)				6 (17%)	4 (11%)
	> 1		29 (83%)	43 (91%)				29 (83%)	31 (89%)
FT3 (pg/ml), mean (SD)		69	3.15 (0.47)	3.39 (0.55)	0.05	0.11	58	3.15 (0.47)	3.26 (0.49)	0.39	0.53
FT4 (ng/dl), mean (SD)		69	1.26 (0.16)	1.26 (0.18)	0.96	0.96	58	1.26 (0.16)	1.23 (0.18)	0.59	0.67
TSH (uIU/ml), mean (SD)		69	2.06 (1.29)	2.58 (1.98)	0.19	0.26	58	2.06 (1.29)	2.64 (2.12)	0.21	0.44
Tumor Size		82			0.007	0.093	70			0.006	0.073
	> 1cm		28 (80%)	24 (51%)				28 (80%)	17 (49%)
	> 2cm		7 (20%)	23 (49%)				7 (20%)	18 (51%)
Number of tumors		82			0.17	0.26	70			0.63	0.67
	solitary		21 (60%)	21 (45%)				21 (60%)	19 (54%)
	multifocality		14 (40%)	26 (55%)				14 (40%)	16 (46%)
Hashimoto's thyroiditis		82			0.038	0.094	70			0.2	0.44
	(-)		14 (40%)	9 (19%)				14 (40%)	9 (26%)
	(+)		21 (60%)	38 (81%)				21 (60%)	26 (74%)
ETE		82			0.038	0.094	70			0.15	0.44
	(-)		25 (71%)	42 (89%)				25 (71%)	30 (86%)
	(+)		10 / 35 (29%)	5 / 47 (11%)				10 / 35 (29%)	5 / 35 (14%)

Mean (SD) for continuous variables: The p value was calculated by Pearson's Chi-squared test, Welch Two Sample t-test or Fisher's exact test.

The q value was calculated by False discovery rate correction for multiple testing

Abbreviations: N numbers of samples, n numbers of subjects, % percentage, SD standard deviation, BMI body mass index, MHR monocyte-to-high-density lipoprotein cholesterol ratio, TC total cholesterol, TG triglycerides, LDL-C Low-density lipoprotein cholesterol, HDL-C high-density lipoprotein cholesterol, TSH thyroid-stimulating hormone, FT4 free thyroxine, FT3 free triiodothyronine, ETE extrathyroidal extension

The multivariate logistic regression analysis showed MHR (OR = 1.65, 95%CI:1.06–2.90, P = 0.048) and Tumor Size > 2cm (OR = 3.98, 95%CI: 1.10–16.5, P = 0.043) were associated with CLNM. After propensity score matching, MHR (OR = 1.76, 95%CI: 1.20–2.88, P = 0.011) and Tumor Size > 2cm (OR = 4.82, 95%CI: 1.56–16.5, P = 0.043) remained independent risk factors for CLNM (Table 2).

Table 2.

Multivariate logistic regression analysis based on the results of univariate analysis

Characteristic	Overall population		p-value	Propensity Score Matched population		p-value
	OR	95% CI		OR	95% CI
Gender
Male	—	—
Female	0.35	0.08, 1.45	0.2
TC (mmol/L)
< 5.2mmol/L	—	—
< 6.2mmol/L	2	0.52, 8.25	0.3
≥ 6.2mmol/L	129,717,512	0.00, NA	>0.9
MHR	1.65	1.06, 2.90	0.048	1.76	1.20, 2.88	0.011
Tumor Size
> 1cm	—	—		—	—
> 2cm	3.98	1.10, 16.5	0.043	4.82	1.56, 16.5	0.008
FT3	1.86	0.52, 7.23	0.3
Hashimoto's thyroiditis
(-)	—	—
(+)	1.74	0.41, 7.82	0.5
ETE
(-)	—	—
(+)	0.74	0.13, 3.79	0.7

Abbreviations: OR Odds Ratio, CI Confidence Interval, TC total cholesterol, MHR monocyte-to-high-density lipoprotein cholesterol ratio, ETE extrathyroidal extension

Overall single-cell atlas of carcinoma and para-carcinoma tissues in PTC

Seventeen samples from two databases were subjected to single-cell integrated analysis (Fig. S1A–B). Following rigorous quality control, the analysis yielded 34,191 high-quality cells, which were subsequently analyzed in further detail. We subsequently adopted the UMAP technique to visualize the high-dimensional scRNA-seq data and successfully classified cells into six clusters, Furthermore, these cell subclasses were verified with well-acknowledged cell markers (Fig. 1A, S1C).

Fig. 1.

PTC cell clustering based on cholesterol homeostasis and cell malignancy level. A UMAP plot of PTC cell basic clustering, and violin plot of the corresponding expression of signature genes. B InferCNV analysis of thyroid cancer cells based on single-cell sequencing data. C Violin diagram of cholesterol homeostasis difference analysis between the normal group, paratumor group, and tumor group. D The heat map of the expression levels of multiple well-acknowledged cell markers of homeostasis and thyroid differentiation, and the immune checkpoints in the normal group, paratumor group, and tumor group. E Further clustering based on cholesterol homeostasis differences in inferCNV cell classification. F The scatter diagrams of correlation analysis between cholesterol homeostasis and EMT, cell migration, and cell proliferation. G The heat map of the expression levels of multiple well-acknowledged cell markers of endoplasmic reticulum stress, Exosome, EMT, and angiogenesis in the Q1-Q4 cell clusters. H The GO biological process analysis of Q1-Q4 cell clusters. I The proportion of different malignant degree cells in the Q1-Q4 cell cluster

Dysregulation of cholesterol homeostasis is associated with the progression and metastasis of thyroid cancer

To further investigate the potential correlation between MHR and the progression and metastasis of PTC, as well as the possible heterogeneity among PTC cells, we acquired comprehensive whole-genome CNV profiles for tumor cell through inferCNV analysis (Fig. S2A). We reduced the dimension and clustered PTC cells into distinct CNV subgroups (normal group, paratumor group, and tumor group). The findings indicated that PTC cells displayed tumor-extrinsic heterogeneity at the CNV level (Fig. 1B). Subsequently, building on the aforementioned results, we investigated the differences in cholesterol homeostasis among cell clusters exhibiting varying degrees of malignancy. The findings revealed significant disparities in cholesterol homeostasis levels between the normal group, paratumor group, and tumor group. There was a positive correlation between cholesterol homeostasis scores and the degree of malignancy. Specifically, the more unstable the cholesterol homeostasis, the higher the malignancy of the tumor might be (Fig. 1C). We also investigated the expression levels of multiple well-acknowledged cell markers of homeostasis and thyroid differentiation, as well as the immune checkpoints, in 3 groups. The results support our previous findings in the present study, most of the cholesterol homeostasis markers and immune checkpoints are highly expressed in tumor group cell clusters. Interestingly, we found that the normal group exhibited a highly differentiated state and the tumor group showed a low differentiated state, while the paratumor group was observed “inverse” results (low differentiated state) in thyroid differentiation (Fig. 1D). Next, based on differences in cholesterol homeostasis, we reduced the dimensions of PTC cells, clustered them into 18 diverse cell clusters, and further clustered them into Q1-Q4 cell clusters (Fig. 1E). Correlation analysis showed that cholesterol homeostasis score was highly correlated with epithelial-mesenchymal transition (EMT) (r = 0.458, p < 0.001), cell migration (r = 0.545, p < 0.001), and cell proliferation (r = 0.424, p < 0.001)(Fig. 1F). Meanwhile, we investigated the expression levels of multiple well-acknowledged cell markers of endoplasmic reticulum stress (ER stress) and exosome, EMT, and angiogenesis. The results indicated markers associated with cell proliferation-related functions were prominently highly expressed in the Q1 Cell cluster, which had a high cholesterol homeostasis score and was similar to the more malignant cell clusters (Fig. 1G). Based on the results of Fig. 1B, we further analyzed the differences in CNV levels (cell malignancy degree) in Q1-Q4 cell clusters, and the results showed that cells with higher malignancy degree in Q1 cell clusters accounted for the highest proportion (Fig. 1H). Additionally, we explore the potential biological procession these clusters might be correlated with, and the results showed that biological processions including “response to lipid”, “tissue migration”, “regulation of myofibroblast”, “differentiation positive”, and “regulation of macrophage” were highly correlated with Q1 cell cluster (Fig. 1I).

Potential association of cholesterol dysregulation with thyroid cancer-associated fibroblasts

We have noticed that the Q1 cell subpopulation, which is strongly associated with the progression and metastasis of thyroid cancer, is also related to regulating fibroblasts. Previous studies have indicated that active fibroblasts are not present in the normal thyroid gland under physiological conditions [18, 19] and tumor-associated fibroblasts can maintain the growth and progression of thyroid cancer by producing soluble factors and extracellular matrix proteins, thus deeply affecting the behavior and invasiveness of thyroid cancer cells [20, 21].

To explore this potential association, we further cluster fibroblast epithelial cell clusters into iCAF and myoCAF and investigate the expression levels of their respective marker genes, the results were visualized by UMAP (Fig. 2A). Subsequent results of GO/ KEGG analysis suggesting CAF and myoCAF are not only associated with the intrinsic functions of fibroblasts, but they also exert a significant role in cell proliferation, differentiation, and migration (Fig. 2B). Notably, they are also involved in the activation and regulation of macrophages. The results of the global cellular communication mode analysis showed that Q1 was strongly correlated with multiple signaling pathways in both input and output signaling modes, with much higher correlation strength than Q2-Q4, iCAF, and myoCAF (Fig. 2C). We then carried out a cellular communication role analysis and found that Q1 cell clusters perform the function of transmitters in both TGF-β and GDF signaling pathways with high importance, and fibroblasts also play an important role in many signaling pathways responsible for cell growth and proliferation, including TGF-β and GDF signaling pathways (Fig. 2D).

Fig. 2.

Thyroid cancer-associated fibroblasts cell clustering, cell communication analysis, and immunohistochemistry. A The UMAP plot of fibroblast clusters in PTC and corresponding markers. B The GO and KEGG analysis of fibroblast clusters in PTC. C Global cell communication analysis between fibroblast clusters and Q1-Q4 cell clusters and well-acknowledged signaling pathways in Outgoing and Incoming signaling patterns. D Cell communication analysis to determine roles and contributions of fibroblast clusters and Q1-Q4 cell clusters in TGF-β, GDF, and PDGF signaling pathway. E IHC staining for GDF15 and myofibroblast marker α-SMA

Previous studies have found that an increase in GDF15 levels is positively correlated with the depth of tumor invasion, lymphatic vessel invasion, and tumor size in various types of cancer, in which the Smad and other signaling pathways may involved [22, 23]. In patients with thyroid cancer, high expression of GDF15 has been shown to be associated with an increased frequency of lymph node metastasis [24]. As a member of the TGF-β superfamily, GDF15 has received increasing attention in the field of cancer metastasis [25]. Hence, we selected GDF15 as the representative marker for subsequent analysis.

Herein, the pan-cancer analysis showed that GDF15 was significantly overexpressed in thyroid cancer samples compared to the control group (P < 0.001, Fig. S2B). Human PTC samples were collected to perform IHC staining for GDF15 and myofibroblast marker α-SMA. Notably, stronger signals for GDF15 and α-SMA were observed in cancer regions compared with the weaker and fewer positive signals in the adjacent normal regions, suggesting the potential correlation between malignancy of tissue and the activation of fibroblasts and their associated pathways, as well as TGF-β and GDF signaling pathways (Fig. 2E, S2C).

Cell lineage and potential role of mononuclear/macrophage cells in thyroid carcinoma

Macrophages were the mature forms of monocytes within tissues [26]. Specifically, monocytes circulate in the bloodstream and possess the capacity to respond promptly, being capable of reacting promptly to infection or injury, while macrophages residing in tissues and are accountable for long-term immune surveillance and defense [27, 28]. During tissue injury or infection, monocytes are rapidly recruited to the damaged tissue, a process typically mediated by cytokines and chemokines, such as the chemokine CCL2 [29]. Once they arrive at the target tissue, monocytes differentiate under the influence of the local microenvironment and transform into macrophages [30]. The differentiated macrophages can perform various functions, including phagocytosis, antigen presentation, cytokine secretion, etc., to participate in the body’s immune response and tissue repair [31].

To explore the lineage status of monocytes in PTC tissue, we next performed unsupervised dimensionality reduction of monocytes. First, PTC cells were further clustered into 7 types of myeloid cell clusters (Fig. 3A). Figure 3B more directly shows the distribution of various myeloid cell clusters. Based on the findings above, macrophage clusters were characterized by different marker genes and further clustered into C1-C8 cluster macrophages, proliferating macrophages, and low-quality clusters (Fig. 3C). Meanwhile, by comparing the cell distribution density of macrophage clusters of PTC patients with and without LNM, we found significant differences in C3-macrophage clusters, implying that the C3 cluster might exert an essential role in PTC metastasis, and the results were visualized with a thermal density plot (Fig. 3D). Notably, the GO/KEGG analysis of C3-cluster macrophages showed significant correlation with cholesterol metabolism, including “Cholesterol metabolism” pathway and “Negative Regulation Of Cholesterol Storage” biological procession (Fig. 3E). Furthermore, we investigated the expression levels of markers of foam cells, the Homeostatic foamy gene, the pathogenic foamy gene, and the immune checkpoints in these types of macrophage clusters. Following our previous findings, almost all of these markers exhibited high expression in C3 clusters (Fig. 3H). We also investigated cellular cholesterol homeostasis and expression of GDF15 and SPP1(Tumor macrophages with high expression of SPP1 tend to have a poor prognosis [32]) of the macrophages clustering, and results showed that C3 clusters exhibited high GDF15 and SPP1 expression and Dysregulated cholesterol homeostasis (Fig. S3A). Cell communication analysis revealed a significant association between macrophage CD68 + C3 clusters and fibroblasts and T cells (Fig. S5C-D).

Fig. 3.

Single-Cell perspective on macrophages in PTC and corresponding cell clustering. A-B Myeloid cell clusters in PTC. C Further macrophage clustering in PTC based on the results of myeloid cell clusters. D Thermal density maps of macrophage density distribution were compared between PTC with lymph node metastasis and macrophage without PTC metastasis. E The GO and KEGG analysis of CD68 + Macro C3 cell cluster. F The detection of TREM2 expression via qRT-PCR and the detection of co-culture migration ability of TPC-1 cells and M2 macrophages. G The heat map of the expression levels of multiple well-acknowledged cell markers of foam cells, the Homeostatic foamy gene, the pathogenic foamy gene, and the immune checkpoints in the 10 macrophage clusterings

PTC may promote metastasis by mediating inducing macrophage differentiation to the M2 phenotype

Macrophages can be mainly classified into two subgroups, namely M1 and M2. M1 phenotype macrophages are tumor-resistant due to their intrinsic phagocytic ability and enhanced antitumor inflammatory responses [33, 34]. In contrast, M2 possesses a set of tumor-promoting abilities, including immuno-suppression, angiogenesis, and neovascularization, as well as stromal activation and remodeling [35].

To further explore the role of macrophages in PTC, we then investigated the differences in regulon activity scores (RSS) of different subtypes of macrophages (C1-C8 and Proliferating macrophage), which showed that LXRα was highly expressed in CD68 + C3 clusters (Fig. S3B). In the TPC-1 thyroid cancer cell line, the expression of LXRα was knocked down by siRNA interference technology, and then Transwell experiments were conducted on the knockdown and control groups. The results showed that the cell migration ability of the LXRα-si group was significantly reduced, proving that LXRα expression is important for the migration of thyroid cancer (Fig. S3C-D).

THP-1 lymphocytes were induced to generate M0 macrophages by PMA treatment. M0 macrophages were further induced by sTPC-1, the conditioned medium of TPC-1 cells, to promote their transformation into M2 macrophages. In the co-culture experiment between the M2 macrophages induced by the aforesaid steps and TPC-1 thyroid cancer cells, it was found that M2 macrophages could promote the migration of TPC-1 cells, while the LXRα-si M2 macrophages demonstrated a slightly diminished capacity to promote TPC-1 cell migration (Fig. 3F). Furthermore, we investigated the expression levels of markers of foam cells, the Homeostatic foamy gene, the pathogenic foamy gene, and the immune checkpoints in these types of macrophage clusters. Following our previous findings, almost all of these markers exhibited high expression in C3 clusters (Fig. 3G).

Pseudotime analysis of macrophage clustering in tumor tissue was performed, and results indicated that with the progression of PTC, macrophages gradually tend to differentiate into CD68 + C3 macrophages, with increased dysregulated cholesterol homeostasis and GDF15 expression (Fig. 5A-B).

Fig. 5.

Further exploration of macrophages and cholesterol homeostasis in PTC. A Cell differential abundance analysis of CD68 + Macro_C3 cell cluster in primary solid tumor and solid tissue normal. B Cell differential abundance analysis of CD68 + Macro_C3 cell cluster in PTC with/without lymph node metastasis. C-D Survival analysis of high and low CD68 + Macro_C3 cell abundance in PTC. E Pearson correlation analysis between GDF15 and fibroblasts, macrophages, myeloid, and Q1-Q4 cell clustering in the present research, as well as tumor immune markers

Single-cell characteristics of thyroid cancer microenvironment and further exploration

Next, our study focused on T/NK cells, 11 cell clusters were obtained after re-clustering, including 3 CD4 + T cell clusters, 4 CD8 + T cell clusters, 2 NK cell clusters, 1 native T cell cluster, and a proliferating cell cluster (Fig. 4A-B).

Fig. 4.

Single-Cell perspective on T/NK cell in PTC and corresponding cell clustering. A-B T/NK cell clusters in PTC. C Global cell communication analysis between CD4 + T regulation cell, CD8 + exhaustion cell, and Q1-Q4 cell clusters and well-acknowledged signaling pathways in Outgoing and Incoming signaling patterns. D Cell communication analysis to determine roles and contributions of CD4 + T regulation cell, CD8 + exhaustion cell, and Q1-Q4 cell clusters in GDF and CXCL signaling pathway. E Survival analysis of high and low GDF expression PTC patients with CD8 + T cells and CD4 + Treg cells enriched/ decreased. F Survival analysis of high and low CXCL16 expression PTC patients with CD8 + T cells and CD4 + Treg cells enriched/ decreased

Global cellular communication analysis showed that the correlation strength between CD4 + Treg cells and CD8 + exhausted T cells and various signaling pathways was much lower than that of the Q1 cell cluster (Fig. 4C). CD8 + exhausted T cells showed a certain correlation with GDF and CXCL signaling pathways, while CD4 + Treg cells were unrelated to GDF signaling pathway and weakly related to CXCL signaling pathway (Fig. 4D).

The high GDF15 expression group exhibited worse Relapse Free Survival (RFS) versus the low GDF15 expression group when CD8 + T-cells (HR = 2.34, 95%CI (0.98—5.58), P = 0.048) and CD4 + Treg cells (HR = 2.45, 95%CI (0.96—6.21), P = 0.052) enriched (Fig. 4E), and results showed no statistical difference when CD8 + T-cells and CD4 + Treg cells decreased (Fig. S4A). Analogously, the high CLCX16 expression group exhibited worse RFS versus the low GDF15 expression group when CD8 + T-cells (HR = 2.77, 95%CI (1.2—6.4), P = 0.013) and CD4 + Treg cells (HR = 3.8, 95%CI (1.5—9.66), P = 0.0025) enriched (Fig. 4F), and no statistical difference results were also observed when CD8 + T-cells and CD4 + Treg cells decreased(Fig. S4B).

Additionally, when macrophages were enriched, the high GDF15 expression group exhibited worse RFS, and reversed results were obtained when macrophages were decreased (Fig. S4C). The above results implied that Q1 clusters and CD68 + C3 macrophage may be mainly associated with promoting tumor metastasis.

Next, we further elucidated the relationship between CD68 + Macro_C3 cell clusters and thyroid cancer LNM through some supplementary analyses. The differential analysis showed that C3 subtype macrophages expressed higher levels in PTC tissues compared to normal tissues (P < 0.001)(Fig. 5A). It is also worth noting that C3 subtype macrophages were higher in PTC tissues with lymph node metastasis versus those without lymph node metastasis (P < 0.05)( Fig. 5B). Patients with high expression of C3 macrophages in their thyroid cancer may have a worse Progression-Free Interval (PFI) (P = 0.02, HR = 1.91, 95%CI:1.1–3.35) and Disease-Free Interval (DFI) (P = 0.036, HR = 2.32, 95%CI:1.03–5.2) (Fig. 5C-D). The Pearson correlation heatmap shows the association between GDF15 and Q1-Q4 cell clusters, CD68 + C3 macrophages, other cells, and various markers. It is worth noting that GDF15 shows a strong positive correlation with the Q1 cluster and CD68 + C3 macrophages (Fig. 5E).

Materials and methods

Clinical data

A retrospective study was conducted to collect the clinical data of 200 patients with papillary thyroid carcinoma continuously treated at the First Affiliated Hospital of Nanchang University from February 2022 to November 2023, and 82 patients’ data were finally included in the study. Inclusion criteria: 1) Pathological examination after surgery confirmed PTC. 2) Multiple lesions were classified according to the largest diameter of the tumor in the pathological examination results. 3) Thyroidectomy with central lymph node dissection ± lateral lymph node dissection. Exclusion criteria: 1) Other types of thyroid tumors (follicular thyroid carcinoma, medullary carcinoma, poorly differentiated carcinoma, etc.) except PTC. 2) Thyroid Papillary Microcarcinoma. 3) History of thyroid surgery or head and neck radiation therapy. 4) Distant metastasis. 5) Incomplete data and lost follow-up. All research content and surgical plans were approved by the patient. The methods for handling clinical data and propensity score matching are presented in Supplementary Methodology.

Single-cell RNA-Sequencing and in vitro experiments

We downloaded publicly available scRNA-seq data of PTC from the GSE191288 [17] and GSE193581 [16] in the Gene Expression Omnibus (GEO) database, which included 13 primary PTC and 4 adjacent normal tissues, shown in supplemental tables (Table S1). The Specific procession of Single-cell RNA-Sequencing analysis (Data screening, Normalization, dimensionality reduction, clustering, cell cluster annotation, subset cell characterization, Pseudotime analysis, Transcription factor module analysis, Cell–cell communication analysis, and enrichment analysis) and in vitro experiments (Immunofluorescence staining, Transwell assay, etc.) were available in Supplementary Methodology.

Deconvolution of TCGA RNA-seq data

The expression data of TCGA and the clinical information of patients were downloaded from the UCSC database (https://xenabrowser.net/). The BayesPrism R package (version 2.1.2) [36] model with default parameters was used to deconvolve immune cell abundances in the TCGA bulk-seq expression profile datasets by constructing a matrix from the scRNA-seq data as a reference. The abundances of the macrophage subtypes of each TCGA patient were classified into a high and a low group on the basis of the optimal cut point returned by the “surv_cutpoint” function.

Statistical analysis

Python (version 3.11) and R (version 4.3.1) were used for statistical analysis. Quantitative data was described by (‾x ± s), and Welch’s two-sample t-test was used to compare quantitative data. Count data was described by n (%), and Pearson’s chi-square test and Fisher’s exact test were used to compare count data. The variables with significant differences in univariate analysis were included in the multivariate logistic analysis. The significance was tested by the F test, and the odds ratio (OR) and 95% confidence interval (CI) of the research factors were estimated. P < 0.05 indicates statistically significant differences.

Discussion

Many studies suggest that thyroid cancer is considered a major immune "hot" malignant tumor, meaning that there are more immune cells in its tumor microenvironment [37, 38]. Thyroid cancer cells can suppress the immune response in various ways, including secreting immunosuppressive factors and expressing immune checkpoint molecules (such as PD-L1), which can inhibit the activity of T cells and help tumor cells evade immune surveillance [39, 40]. Different types of immune cells could have distinct roles in the development and progression of thyroid cancer. For instance, regulatory T cells (Tregs) and tumor-associated macrophages (TAMs) are regarded as key cells that facilitate tumor growth, which enable tumor cells to survive and spread by suppressing the function of effector T cells [41, 42].

Consistent with previous studies [43, 44], In the present research, the in vitro experiments found that thyroid cancer may promote metastasis by mediating inducing macrophage differentiation to the M2 phenotype. CXCL16, mainly produced by PTC-stimulated macrophages and able to promote PTC cell migration and invasion [45, 46], was confirmed by survival analysis to be associated with lower RFS and OS with high expression. Meanwhile, based on single-cell sequencing data, we identified the macrophage clusters (CD68 + _C3) most related to LNM in thyroid cancer. Further differential analysis showed that it was upregulated in the tumor tissue and significantly upregulated in PTC with combined LNM compared to non-metastatic PTC. Survival analysis also confirmed that patients with high expression of CD68 + _C3 macrophage clusters often had worse PFI and DFI. In summary, our results confirm from multiple angles that monocytes/macrophages are closely related to LNM of PTC and have the potential to be predictive indicators.

Previous studies have pointed out that thyroid cancer cells recruit immunosuppressive cells such as Tregs and myeloid-derived suppressor cells (MDSCs) by upregulating the expression of CXCLs to suppress anti-tumor immune responses and help tumor cells evade host immune surveillance [47–49]. In this study, the results of T/NK cell single-cell clustering show that exhausted T cells are distributed roughly the same as CD4 + T cells, while cytotoxic cells are distributed roughly the same as NK cells and CD8 + cells, which confirms the widespread existence of immune suppression in thyroid cancer cells and mainly related to CD4 + T cells. Notably, the results of cell communication analysis show that CD8 + T exhausted cells with CXCL signaling pathway show a moderate correlation, with CD4 + Tregs showing a weaker correlation, which is consistent with the findings of other scholars and simultaneously validates our previous results. We also speculate the immune heterogeneity that we found may also be part of the reason for the mixed efficacy of PTC in immunotherapy, further research may inform new progress in the field of PTC immunoresistance.

Cholesterol metabolism plays an important role in the development, invasion, and metastasis of tumors [50, 51]. In particular, the ratio of low-density lipoprotein cholesterol (LDL-C) to high-density lipoprotein cholesterol (HDL-C) is believed to be related to the metastasis of cancer [52]. Specifically, an increase in cholesterol levels, especially an increase in LDL-C [53], may be associated with a higher rate of lymph node metastasis [54]. In this study, we identified the Q1 subpopulation with the highest (most unstable) cholesterol homeostasis score and found that it was significantly associated with the malignant degree and multiple functional pathways of cell proliferation and metastasis, including TGF-β and GDF. GDF15, a marker of these pathways, was observed by immunohistochemistry to be significantly overexpressed in PTC tissues compared to normal tissues. Moreover, GDF15 was found to be significantly positively correlated with both the Q1 subpopulation and the CD68 + C3 subpopulation. We have thus uncovered an intriguing association between GDF15 and cholesterol metabolism, which is more importantly, these results are consistent with our previous findings (Fig. 1), indicating a significant association between cholesterol homeostasis disorder and the occurrence of PTC.

To our knowledge, we were the first to report a strong correlation between MHR and lymph node metastasis in thyroid cancer, which could potentially serve as a preoperative predictor. Compared with many practical radiomics studies on preoperative prediction of LNM in PTC [55] (such as ultrasound-base radiomics [56, 57] and CT Radiomics [58], MHR as a blood marker, although less precise, was relatively easier to obtain and convenient, and may have a higher judgment efficiency, which could obtain higher patient acceptance and did not depend on the technology of operation personnel. Other inflammation indicators (such as systemic immune inflammatory response index (SII) [59], neutrophil-to-lymphocyte ratio (NLR) [60], platelet-to-lymphocyte ratio (PLR) [61], etc.) are mostly shallow and one-sided cross-sectional studies. This study explores the relationship between MHR and LNM in PTC from multiple dimensions, including clinical data analysis and single-cell sequencing analysis, and verified this correlation to some extent via in vitro experiments, which partially fills in the gaps of cross-sectional studies and elevates the accuracy of our study.

Monocytes, a major component of the human immune system, contribute to the production of pro-inflammatory cytokines and pro-oxidants [62]. HDL exhibits antioxidant and anti-inflammatory properties by reducing the release of pro-inflammatory cytokines from monocytes and macrophages. Additionally, HDL suppresses the expression of cytokine-activated adhesion molecules and facilitates cholesterol efflux from peripheral tissues, thereby attenuating inflammatory progression [63]. Previous studies have demonstrated that elevating HDL levels via overexpression of human ApoA-I not only reduces lesion size of atherosclerosis and lipid content but also modifies the genetic profile of lesional macrophages. HDL prevents the differentiation of human monocytes into the pro-inflammatory M1 phenotype [64, 65]. However, its role in promoting M2 macrophage polarization remains controversial. Some studies indicate that, unlike in murine macrophages, HDL does not significantly influence M2 polarization in human macrophages [66].

MHR has emerged as a novel indicator of inflammation, oxidative stress, and metabolic syndrome, reflecting the balance between monocyte-driven inflammation/oxidative stress and HDL-mediated protection. Recent studies [67, 68] have proposed innovative therapeutic strategies, such as HDL-mimetic nanotherapeutics, which modulate monocyte recruitment and promote macrophage polarization toward an anti-inflammatory phenotype, showing significant clinical potential. MHR demonstrates superior predictive value for clinical outcomes compared to separate monocyte counts or HDL concentrations [12]. However, the underlying molecular mechanisms and signaling pathways remain poorly understood. Further investigation into the relationship between MHR, cholesterol dysregulation, and macrophage polarization is essential to advance novel therapeutic approaches. Our study provides critical insights that help address this research gap.

Several limitations should be noticed. Firstly, due to the constraints of experimental conditions and other factors, this study is a single-center retrospective study with a relatively small sample size, which may cause some bias in the results. Secondly, the precise upstream and downstream regulatory mechanisms by which cholesterol homeostasis and monocytes facilitate thyroid cancer metastasis remain to be further investigated.

In summary, in this study, we revealed the potential role of MHR in PTC lymph node metastasis from a single-cell perspective, and the CD68 + C3 cell cluster of macrophages, and cholesterol dysregulation was significantly associated with LNM. Our study confirms that MHR can serve as a promising predictive marker and provides a feasible solution for predicting the risk of LNM in PTC patients before surgery.

Authors’ contributions

Z.K.-N. conceptualized the study design and interpreted the results. T.T.-Y. provided methodological suggestions and managed the data of the public databases. H.Y. was in charge of analyzing the data and drafting the manuscript. J.L. and H.Y.-H. were very helpful in completing the experimental content. S.S.-S and H.T. provided many suggestions for the article's writing. M.H.-X, H.-L., and Y.T. played a great help in visualization of results. X.P.-Z. and X.-M. were of immense help in the manuscript process, and are responsible for ensuring that the descriptions and findings are accurate. All authors read and approved the final manuscript.

Funding

The science and technology plan of Jiangxi Provincial Administration of Traditional Chinese Medicine,2022B1001,2023B1221,the Jiangxi Provincial Natural Science Foundation,20232BAB206020,20224BAB216022,Science and Technology Plan of Jiangxi Provincial Health Commission,202130202

Data availability

The datasets and computer code produced in this study are available in the following databases: 1) scRNA-seq data:GSE191288 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE191288), GSE193581 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE193581). 2) RNA-Seq data:The expression data of TCGA and the clinical information of patients were downloaded from the UCSC database (https://xenabrowser.net/). Other data that support the findings of this study are available from the corresponding author upon request.

Declarations

Ethics approval and consent to participate

The studies involving human participants were reviewed and approved by the Medical Ethics Committee of the First Affiliated Hospital of Nanchang University, and this study followed the ethical standards laid down in the 1964 Declaration of Helsinki. This ethics committee strictly follows China’s Good Clinical Practice (CCP) and related regulations. Written informed consent was obtained from participants to collect and analyze specimens. This research was approved by the Ethics Committee of the First Affiliated Hospital of Nanchang University (project approval number: ITS2024783).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.