Thyroid carcinoma (TC) is one of the most common malignant tumors of the endocrine system, and its biological behavior, therapeutic response, and prognosis are closely associated with the degree of tumor differentiation (1). According to histological differentiation, TC is generally classified into differentiated types—differentiated thyroid carcinoma (DTC, such as papillary thyroid carcinoma (PTC) and follicular thyroid carcinoma (FTC)), poorly differentiated thyroid carcinoma (PDTC) and anaplastic thyroid carcinoma (ATC) (2). Clinically, DTC usually exhibits slow progression and favorable prognosis, whereas PDTC and ATC are characterized by high proliferative activity, extensive genetic abnormalities, and resistance to therapy, resulting in a poor clinical outcome (3, 4).

The prognosis of tumors largely depends on their biological behavior, which is profoundly influenced by the tumor immune microenvironment (TIME) (5). As a dynamic ecosystem composed of immune cells, stromal cells, cytokines, and metabolic products, the TIME plays a pivotal role in regulating the initiation, progression, and therapeutic sensitivity of TC (6–8). The immune composition and functional state of the TIME not only determine the invasiveness and metastatic potential of tumor cells but also affect patients’ clinical outcomes and treatment responses (9). Based on differentiation gradients, single-cell RNA sequencing (scRNA-seq) has systematically delineated distinct immune cell subpopulations and their functional states within the TIME, revealing the immunoregulatory landscape across different differentiation stages of TC (10, 11).

Dissecting this complexity requires advanced transcriptional profiling tools. While bulk RNA sequencing has been instrumental in identifying overall immune signatures, it obscures the cellular heterogeneity central to the function of TIME. In contrast, scRNA-seq resolves this diversity by cataloging immune and stromal subpopulations at unprecedented resolution, directly revealing their functional states and communication networks. Furthermore, spatial transcriptomics is now critical for mapping these cellular interactions within their native tissue architecture, validating the ecological niches predicted by scRNA-seq. A comparative overview of these technologies is provided in Table 1.

Studies have demonstrated that, along with the loss of differentiation in TC, the TIME exhibits a progressive transition from a state of “coexisting activation and suppression” (an inflamed microenvironment with a delicate balance of effector and suppressor forces) to “low infiltration/exclusion” (a paucimmune landscape lacking productive infiltration) and eventually to “high infiltration with exhaustion” (a high-density but dysfunctional state dominated by T-cell exhaustion and myeloid-driven suppression) (12–14). In ATC, such immune exhaustion is frequently characterized by CD8⁺ T cells dysfunction and enrichment of regulatory T cells (Tregs) (15). A summary of representative studies describing the immune composition and major findings across differentiation types is presented in Table 2.

The immune heterogeneity across different differentiation states provides new insights for clinical management. First, immune profiling can serve as an auxiliary reference for pathological classification and risk assessment, identifying specific immune subpopulations may help infer the degree of tumor differentiation and potential aggressiveness (13). Second, immunosuppressive TIME signatures are associated with unfavorable clinical outcomes, including increased risks of recurrence and resistance to radioactive iodine therapy (22, 23). Third, persistent activation of immune checkpoint pathways in patients with PDTC and ATC provides a biological rationale for the application of PD-1/PD-L1 and CTLA-4 inhibitors. Meanwhile, targeting intercellular communication networks represents a promising strategy for future combinational “immunotherapy + targeted therapy” approaches (24, 25).

On this basis, this review centers on the axis of “differentiation gradient–TIME heterogeneity–clinical translation”, systematically summarizing the major immune landscape features of thyroid carcinoma and their potential clinical implications, thereby providing a solid theoretical and practical foundation for precision diagnosis, risk stratification, and personalized immunotherapy in TC.

To ensure a comprehensive and unbiased retrieval of relevant literature, a systematic search was conducted across four major electronic databases: PubMed, Web of Science, Embase and Cochrane Library. The search strategy employed the following keywords and their combinations: “Thyroid Neoplasms”, “Thyroid Cancer, Papillary”, “Adenocarcinoma, Follicular”, “Thyroid Carcinoma, Anaplastic”, “Single-Cell Gene Expression Analysis”, “Tumor Microenvironment”, “Tumor-Associated Macrophages”, “Dendritic Cells”, “Killer Cells, Natural” to retrieve English articles published from January 2015 to October 2025. The final literatures included for in-depth analysis were selected based on strict criteria: (1) high relevance to the research theme; (2) rigorous experimental design; (3) complete and reliable data. To ensure objectivity, a double-blind screening process was adopted, where two researchers independently evaluated literature quality, with discrepancies resolved through expert group discussions. In the comprehensive analysis, we paid special attention to the mutual verification between different studies, incorporating both supporting evidence and opposing viewpoints to ensure the scientificity and reliability of conclusions. The overall literature screening and inclusion process is summarized in Figure 1. 940 records were retrieved from PubMed, Embase, Web of Science, and Cochrane Library based on predefined MeSH.

DTC primarily includes PTC and FTC, accounts for more than 90% of all malignant thyroid tumors (1), with a subset of patients progressing to radioiodine-refractory disease (26). In recent years, beyond the classical driver mutations in BRAF, RAS, and TERT (27), increasing attention has been directed toward the pivotal role of TIME in the initiation, progression, and therapeutic response of PTC (28, 29).

Given that PTC represents the predominant subtype of DTC, this review focuses on the immunological landscape of DTC using PTC as a representative model. Histologically, PTC tissues are characterized by extensive infiltration of macrophages, T cells, natural killer (NK) cells, neutrophils, and dendritic cells (DCs) (11, 29). ScRNA-seq studies have revealed that the TIME of PTC exhibits a state of “coexisting immune activation and suppression”: on one hand, active antitumor immune cell populations are present, while on the other hand, phenomena such as T cells exhaustion, enrichment of immunosuppressive macrophages, and impaired antigen presentation are frequently observed (11, 18). The complex cellular interactions within the TIME of PTC are illustrated in Figure 2, highlighting the interplay among T cells, B cells, NK cells, DCs, and macrophage subpopulations.

PDTC, positioned between PTC/FTC and ATC, represents an intermediately aggressive subtype. It is characterized by loss of differentiation markers, increased proliferative and necrotic tendencies, reduced radioiodine uptake, and markedly elevated therapeutic difficulty (50). The 5 year survival rate lies between those of differentiated and undifferentiated carcinomas but remains substantially lower than that of typical PTC (2, 51). Molecularly, PDTC features driver gene alterations such as TERT promoter and TP53 mutations, as well as aberrations in the PI3K/AKT signaling pathway, accompanied by upregulation of dedifferentiation markers in its transcriptional profile (52). Within the immune microenvironment, PDTC displays markedly reduced tumor infiltrating immune cells, including sparse T cells and TAMs, consistent with a “cold tumor/immune exclusion” phenotype (53). Expression of inhibitory immune checkpoints—PD-L1, PD-L2, PD-1, LAG-3, and TIM-3—are generally limited (54). In certain cases, mild immune activation signals can be detected, suggesting low immunogenicity and a propensity to develop an immune evasive microenvironment (55). Moreover, BRAF-associated signaling may further contribute to immune suppression and reduced tumor immunogenicity (18, 56). Collectively, PDTC exhibits an “immune inert and suppressive” signature, representing a refractory subtype that poses a major challenge for immunotherapeutic strategies in thyroid cancer. Representative immune interactions within the TIME of PDTC are shown in Figure 3.

ATC represents one of the most aggressive forms of thyroid malignancy, characterized by rapid clinical progression, high resistance to conventional radioiodine therapy and most systemic treatments, and an extremely high mortality rate (19, 34, 77). ATC exhibits extensive immune cells infiltration accompanied by potent immunosuppressive signaling, which may partially explain its relatively higher clinical responsiveness to immune checkpoint blockade therapy (12, 34, 78). Representative immune dysfunctions within the TIME of ATC are shown in Figure 4.

As the degree of differentiation decreases, the TIME of TC exhibits a progressive transition from “coexisting immune activation and suppression” → to a “cold/immune excluded” state → and finally to a “highly infiltrated but exhausted” phenotype (9, 12, 34, 79). This gradient is accompanied by increasing heterogeneity, dedifferentiation, and therapeutic refractoriness. The dynamic remodeling of the TIME across thyroid cancer differentiation is illustrated in Figure 5.

Characterized by a “coexistence of immune activation and suppression” state, with moderate immune cells infiltration and partial preservation of cytotoxic CD8⁺ T cells function (32). Immune evasion primarily depends on CD8⁺ T cells exhaustion and the immunosuppressive network orchestrated by M2-polarized TAMs and Tregs (86, 116). This immune landscape correlates with increased tumor aggressiveness and radioiodine (RAI) refractoriness, suggesting the potential of strategies that enhance immunogenicity and promote antigen presentation or redifferentiation reactivation.

Displays a “cold tumor/immune excluded” phenotype, featuring impaired CD8⁺ T cells maturation, TAMs mediated immunosuppressive remodeling, reduced and dysfunctional NK cells activity, attenuated Tfh–B cells cooperation, increased PD-L1 expression, decreased cDC1 frequency, and defective antigen presentation (56, 59, 61, 74, 75, 117). This subtype represents a major immunotherapeutic challenge, strongly associated with RAI resistance and treatment failure. Therapeutic strategies targeting this immune profile may include myeloid reprogramming combined with dual/multiple checkpoint blockade and restoration of antigen presentation capacity.

Exhibits a “highly infiltrated yet exhausted” immune state, characterized by extensive TIME reprogramming, high level expression of multiple checkpoint pathways (PD-1/PD-L1, LAG-3, TIM-3, etc.), and massive enrichment of myeloid derived suppressor cells (118, 119). Inhibitory NK cells dominate, while TLSs formation and upregulation of IL-10 and PD-L1 reflect the functional duality of B cells. Absolute or relative loss of cDC1 and impaired maturation of cDC2 lead to disrupted antigen presentation, spatially forming an “immunosuppressive loop” and a functionally compromised tumor core (9, 18, 78, 79). Although ATC retains partial responsiveness to immune checkpoint inhibitors, its therapeutic benefit is highly heterogeneous. Optimal strategies may require reversal of T cells exhaustion, remodeling of myeloid suppression, and restoration of cDC1 functionality to maximize clinical efficacy.

The profound immune dysfunction and heterogeneity observed in ATC may be rooted in its cellular origins. The prevailing model posits that ATC arises through the dedifferentiation of pre-existing differentiated carcinomas. Genomic analyses delineated this trajectory, identifying key co-mutations (e.g., TERT and TP53) that drive progression (119). Crucially, single-cell transcriptomics visually captured this process, revealing ‘anaplastic-like’ cells residing within BRAF-mutant papillary thyroid carcinomas, thereby providing direct evidence of clonal evolution along a dedifferentiation continuum (34).

While scRNA-seq excels at cataloging cellular diversity, it lacks spatial context. Spatial transcriptomics directly addresses this by mapping gene expression within intact tissue architecture, providing critical insights into the thyroid cancer ecosystem. This technology validates cell-cell interactions predicted by scRNA-seq and reveals key spatial patterns of immune evasion. For instance, it has visually confirmed the presence of tertiary lymphoid structures in PTC, localizing active immune collaborations. In contrast, it delineates the “immune-excluded” phenotype in PDTC and ATC, demonstrating how cytotoxic CD8⁺ T cells are restricted to the tumor stroma or invasive margin, while the core is dominated by immunosuppressive cells like M2-polarized TAMs. The co-localization of exhausted T cells with PD-L1⁺ myeloid cells provides a spatial rationale for immunotherapy responses. Ultimately, the synergy between scRNA-seq (identifying the “players”) and spatial transcriptomics (defining the “stage”) is forging a spatially-resolved understanding of the TIME, which holds promise for guiding spatially-informed precision immunotherapy in the future.

Advances in Single-Cell and Spatial Omics Are Driving a Paradigm Shift in TIME Research—from Quantitative Assessment to Qualitative Understanding. The evolution of analytical approaches—from quantifying immune cells infiltration density and immune checkpoint expression (e.g., PD-1, PD-L1) (18, 120), to single-cell and spatial transcriptomic mapping that delineates lineage trajectories, functional reprogramming, and intricate cell–cell communication networks (9, 78, 118)—has elevated the focus from “immune cells quantity” to the broader concept of “immune ecology.” These advances have revealed a strong correlation between tumor differentiation status and the dynamic evolution of TIME in thyroid carcinoma (86, 116). The dedifferentiation continuum from PTC → PDTC → ATC is mirrored by TIME transitioning from “coexistent activation and suppression” to “cold/immune excluded” and ultimately to a “highly infiltrated but exhausted” state. This progressive pattern provides a biological explanation for clinical heterogeneity and reflects the bidirectional co-evolution between tumor cells and their immune microenvironment.

For studying dedifferentiation and TIME, future directions evolution include (1): multi-omics integration and spatial reconstruction (121), enabling dynamic tracking of key cells populations—such as pre-exhausted CD8⁺ T cells, TREM2⁺ TAMs, and cDC1s—before and after therapy (2). Single-cell trajectory analyses to monitor lineage transitions and functional state shifts during tumor progression (122) (3). Spatial transcriptomics to identify the precise localization of critical immune populations, constructing a three-dimensional immune ecosystem and providing spatially resolved targets for disrupting immunosuppressive networks (123, 124). From a therapeutic perspective, mechanism driven combination strategies should be tailored according to differentiation hierarchy: DTC: immune activation approaches based on antigen presentation and T cells restoration (such as cancer vaccines or local radiotherapy combined with PD-1 blockade) (125, 126); PDTC: myeloid reprogramming combined with multi-checkpoint inhibition (53); ATC: highest potential for combinatorial immunotherapy, emphasizing anti-exhaustion strategies, STING pathway activation, and stromal unlocking NK cells reactivation (127–129). Recently, targeted plus immunotherapy neoadjuvant regimens—for example, BRAF^V600E driven combinations—have entered clinical trials and yielded promising early results, demonstrating the feasibility of achieving both rapid tumor regression and durable immune responses within a short therapeutic window (130). Key clinical studies targeting the tumor immune microenvironment are summarized in Table 3.

The single-cell architecture of the ATC TIME elucidates the limited efficacy of PD-1 blockade, revealing that PD-L1 upregulation occurs within a profoundly immunosuppressive ecosystem. This resistance is multifaceted: CD8⁺ T cells are locked in pre-exhausted or terminally exhausted states; M2-polarized and TREM2⁺ TAMs dominate, secreting IL-10, TGF-β, and adenosine; and loss of cDC1 with impaired cDC2 maturation creates an antigen presentation deficit (88, 95). Consequently, PD-1 monotherapy often fails to reverse this entrenched suppression. In contrast, the success of dabrafenib plus trametinib in BRAF-mutant ATC demonstrates that targeted therapy can achieve profound tumor regression (136), while emerging evidence suggests MAPK inhibition may also favorably remodel the TIME. This synergy is exemplified by lenvatinib plus pembrolizumab, pointing toward a future paradigm of rational combinations—co-targeting tumor cells, specific immune resistance hubs (e.g., TREM2⁺ TAMs), and checkpoint pathways—to overcome resistance.

This review primarily reflects the immune landscape of PTC, PDTC, and ATC. ScRNA-seq data for FTC and MTC remain limited, largely due to their lower incidence and the challenges in procuring fresh surgical specimens. Existing studies are often constrained by small sample sizes and potential selection biases. Future multi-center collaborations are essential to define the TIME of these understudied subtypes.

In summary, high resolution temporal and spatial immune mapping has delineated an evolutionary TIME atlas from differentiated to undifferentiated thyroid carcinoma, laying a conceptual and technological foundation for precision immunotherapy. This paradigm shift—from static immune cells distribution to dynamic immune system reconstruction—highlights a state stratified, mechanism guided framework for adaptive clinical trial design and marks the beginning of a mechanistically driven era in thyroid cancer immunotherapy.