The microenvironment surrounding thyroid tumors is a highly organized and ever-evolving milieu that integrates neoplastic cells, immune constituents, stromal components, blood vessels, and an array of cytokines (31, 63). The composition and functionality of these immune constituents differ according to the specific tumor histotype and the broader disease trajectory. Tumor-infiltrating lymphocytes (TILs), particularly populations of CD8⁺ cytotoxic and CD4⁺ helper T cells, serve as principal agents of immunological assault against the neoplasm (47, 64); yet, their effective firepower is commonly undermined by a spectrum of immunosuppressive cells, notably regulatory T cells (Tregs), myeloid-derived suppressor cell (MDSC) accumulations, and tumor-associated macrophages (TAMs) (33, 36, 65–67). The more malignant forms, particularly ATC, manifest a striking enrichment of M2-polarized TAMs that, through the secretion of immunomodulatory cytokines and inductive signals for neovascularization, effectively abbreviate tumor immune surveillance (6, 68). Natural killer (NK) cell populations are detectable, yet their cytotoxic functionality frequently diminishes, a setback attributable to the tumor’s ability to enact selective immune evasion (65). As depicted in Figure 1, the transition from DTC to ATC correlates with a shift from a pauci-immune environment to one characterized by heavy macrophage infiltration and complex cytokine networks.

The cytokine and chemokine landscape within the thyroid TMEs critically orchestrates the recruitment and functional modulation of immune cells (70, 71). Increased secretion of immunosuppressive cytokines such as interleukin-10 (IL-10) and transforming growth factor-beta (TGF-β) drives T-cell energy, whereas chemokine interactions along the CXCL12/CXCR4 axis promote tumor cell dispersal and metastatic dissemination (70, 72, 73). Concurrently, inflammatory mediators like IL-6 and TNF-α may paradoxically sustain tumor progression via prolonged inflammatory stimulation (74). Together, these factors craft an immune-privileged microenvironment that supports immune energy and tumor endurance (39).

Thyroid neoplasms adopt multiple immune evasion tactics to escape surveillance and destruction (75). A signature strategy involves selective downregulation of major histocompatibility complex class I (MHC-I) glycoproteins, attenuating the capacity of tumor cells to present neo-antigens to cytolytic T lymphocytes (34). This phenomenon is especially marked in poorly differentiated and anaplastic variants, leading to diminished detection by the adaptive immune compartment (70).

A further important mechanism comprises the increased expression of immune checkpoint ligands, particularly PD-L1, on both malignant thyroid cells and the immune infiltrate (76). The interaction of PD-L1 with PD-1 receptors on effector T cells dampens cytotoxic activity and contributes to T-cell functional exhaustion (70). In ATC, elevated PD-L1 levels are often detected and are associated with a TME that is characterized by an inflammatory signature, both marking a potential target for checkpoint inhibition and reflecting adaptive resistance of the immune compartment (6, 33).

Simultaneously, the macrophage switch toward the M2 phenotype generates a milieu that favors tumor tolerance through the secretion of IL-10, vascular endothelial growth factor (VEGF), and various matrix metalloproteinases (MMPs), all of which promote tumor expansion and secondary spread (71). These pathways further undermine T-cell expansion by degrading tryptophan and generating metabolites that inhibit effector functions (34). Collectively, these signals create a permissive immune environment that reduces the likelihood of immune-mediated tumor rejection and hinders the progression of spontaneous regression (55).

Immune checkpoint molecules represent critical inhibitory pathways that preserve self-tolerance while fine-tuning the intensity of immune reactivity, thereby shielding normal tissues from collateral damage (55). In thyroid malignancies, neoplastic cells exploit these pathways to evade detection and destruction by the immune system (29). A summary of the molecular targets and targeted therapies is described in Figure 3 which summarized the landscape of targeted therapies, highlighting the distinction between established receptor tyrosine kinase inhibitors and the expanding repertoire of immunotherapeutic agents.

To clarify the translational status of these molecular targets, they can be categorized into three stages of clinical development.

Both classes of antigens provide definitional substrates for immunological recognition and therapy, distinguished by their differential expression on neoplastic versus normal tissues (40). Briefly,

Oncofetal antigens, normally confined to fetal development and re-expressed in neoplasia, afford avenues for immunotherapy with narrow off-target toxicity, as summarized.

Immune checkpoint blockade has established itself as a cornerstone of modern cancer immunotherapy, counteracting suppressive receptor-ligand interactions to restore effective T-cell-mediated anti-tumor immunity (97, 98). In thyroid neoplasms, blockade of PD-1/PD-L1 and CTLA-4 pathways has garnered the most investigational traction (50). A summary of the molecular targets and immunotherapy strategies in thyroid cancers is presented in Figure 4.

Therapeutic cancer vaccines are designed to elicit immune responses tailored to tumor-specific antigens (40, 50), such as,

Adoptive cell therapy (ACT) encompasses strategies that expand, engineer, and reinfuse patient-derived or donor-derived immune cells to achieve robust anti-tumor effects (43, 90). For example,

Oncolytic viral therapy (OVT) capitalizes on genetically modified, replication-competent viruses that selectively infect and destroy tumor cells (90). The viral replication cycle releases tumor-derived antigens, thereby stimulating systemic antitumor immunity (41). In thyroid cancer models, oncolytic adenoviruses and vaccinia viruses have shown promising effects, particularly against ATC (110). Further gains in antitumor potency can be achieved by engineering the viruses to produce immune-enhancing cytokines such as granulocyte-macrophage colony-stimulating factor (GM-CSF), which can act synergistically with ICIs (50, 70).

Given the multifactorial underpinnings of immune resistance across thyroid cancers, the development of rationally designed combination strategies is gaining momentum in translational research and clinical practice (58, 108). Briefly,

Multiple Phase I, Phase I/II, and ongoing investigations are systematically evaluating immunotherapy in thyroid malignancies, with a particular focus on advanced disease stages and settings characterized by refractoriness to standard cytotoxic or kinase inhibitor regimens (Table 2) (98, 105, 108, 113).

While the aforementioned trials provide encouraging signals, it is imperative to interpret these results with caution. The majority of data regarding ATC and MTC are derived from Phase I/II single-arm cohorts with small sample sizes, making them susceptible to selection bias. Furthermore, the lack of randomized control arms in many studies complicates the differentiation between the true immunotherapeutic effect and the natural history of the disease in highly selected patient populations. Consequently, while current evidence supports the use of ICIs in advanced ATC (hypothesis-affirming), their role in DTC remains largely hypothesis-generating, necessitating validation through large-scale, randomized controlled trials

The strategic selection of patients most likely to benefit from immunotherapies depends on reliable predictive biomarkers (59), as mentioned.

Despite the relative scarcity of randomized clinical trial data, observational practice has begun to illuminate the practical application of immunotherapy in rare, high-grade thyroid cancers (105, 118). Compassionate use of pembrolizumab in ATC has, when instituted early in the clinical course, been associated with rapid tumor shrinkage, notable extension of survival, and quantifiable gains in quality of life (6, 100). In select thyroid cancer patients, the strategic sequencing of ICIs with targeted kinase therapy has yielded disease stabilization that exceeds the durability associated with monotherapy of the latter (69, 97, 105). Collectively, these real-world data highlight the translational promise of immune-based strategies and encourage systematic exploration in carefully defined, biomarker-guided cohorts (59, 118).

Differentiated thyroid carcinomas, notably papillary and follicular variants, display a low tumor mutational burden, a paucity of tumor-infiltrating lymphocytes, and a limited neoantigen landscape (64, 120). These attributes confer a cold immune phenotype that diminishes the effectiveness of ICIs (69, 98). Addressing this limitation necessitates the incorporation of approaches intended to elevate tumor immunogenicity, including localized radiation, selective kinase inhibitors, and oncolytic viral vectors, either sequentially or on a combinatorial basis (121–124). Addressing this limitation requires strategies to convert “cold” tumors into “hot” environments. For instance, the integration of radiotherapy can induce immunogenic cell death (ICD). This process releases damage-associated molecular patterns (DAMPs) and upregulates MHC-I expression on tumor cells, thereby facilitating the recruitment and infiltration of cytotoxic T lymphocytes. Concurrently, oncolytic viruses are being engineered to secrete cytokines like GM-CSF, further amplifying this systemic immune priming. For the vast majority of PTC and FTC patients, immunotherapy is currently defensible only within the context of clinical trials. The inherently low antigenicity of these tumors means that ICI monotherapy is unlikely to provide clinical benefit. Future success in this subgroup depends entirely on “immune-priming” strategies—such as combinations with TKIs or radiation—that can artificially inflame the microenvironment, rather than relying on pre-existing immunity.

Although durable remissions can accompany checkpoint blockade, the therapy is not devoid of risks (44). Immune-related adverse events may affect virtually any organ, yet endocrinopathies such as thyroiditis, subsequent hypothyroidism, adrenal insufficiency, and hypophysitis assume particular significance in the context of thyroid malignancy (100). These events can complicate the clinical course by destabilizing pre-existing hormonal dysregulation (102, 108). Effective mitigation hinges on prompt recognition, the integration of endocrinologists with oncologists, and the provision of comprehensive patient education that includes symptoms to be monitored (125). While manageable, these events are significant. Clinical observations indicate that immune-related thyroid dysfunction occurs in approximately 15–30% of patients receiving checkpoint inhibitors. Effective management requires a tiered protocol: asymptomatic or mild cases (Grade 1–2) typically allow for the continuation of immunotherapy with appropriate hormone replacement (e.g., levothyroxine for hypothyroidism), whereas severe inflammatory reactions (Grade 3–4) necessitate the temporary suspension of therapy and the administration of high-dose corticosteroids.

Immunotherapy for aggressive thyroid cancers faces both intrinsic and adaptive resistance (33, 67). Intrinsic or primary resistance often stems from barriers such as poor infiltration of activated T cells, defective MHC molecule expression (89), and a predominance of immunosuppressive elements, including regulatory T cells, myeloid-derived suppressor cells, and alternatively activated macrophages (126). Conversely, adaptive or acquired resistance evolves in response to pressure from therapy and may manifest as compensatory upregulation of inhibitory checkpoint receptors (127); most notably, T cell immunoglobulin and mucin-domain-containing molecule-3 and lymphocyte-activation gene 3, loss of the target antigen because of mutations or selective pressure, or a reorganized, immunosuppressive TME (31, 44). Elucidation of these escape pathways is essential to inform next-generation clinical strategies that incorporate rationally designed combination therapies (82, 128). Adaptive resistance often evolves under therapeutic pressure. A key mechanism involves the compensatory upregulation of alternative immune checkpoints, such as TIM-3 and LAG-3, following initial PD-1 blockade. This “checkpoint switching” limits T-cell effector function. Consequently, future clinical designs are increasingly focusing on dual-blockade strategies (e.g., anti-PD-1 plus anti-TIM-3) or the addition of MEK inhibitors to sensitize the tumor to immune recognition.

The infrequency of aggressive thyroid neoplasms, specifically ATC and advanced MTC, complicates the speedy and sufficient recruitment of patients for investigational studies (105, 129). Geographic dispersion of a small cohort of patients, the absence of trial sites in resource-limited environments, and narrow eligibility definitions combine to constrain enrollment, which in turn delays the accrual of statistically powerful, high-quality clinical evidence (97, 108).

Predictive biomarkers for immune checkpoint blockade in thyroid cancers have included tumor cell-surface PD-L1 expression, TMB, and select immune transcriptomic signatures (59, 79, 129). However, their performance for accurate patient stratification remains inadequate (130). The lack of standardized, validated biomarkers restricts the ability to individualize immunotherapy interventions (50). Furthermore, preclinical validation often utilizes immunocompromised murine models (67), which inadequately mirror the complexity of human tumor-immune system interactions, thereby limiting the translational value of observed therapeutic responses and contributing to clinical trial attrition (50, 88).

The extensive biological heterogeneity of thyroid cancers requires an immunotherapy platform based on precision medicine principles (31, 131). Multimodal profiling, spanning whole-exome sequencing (132), RNA sequencing (73, 133, 134), and mass spectrometry-based proteomics (127, 135), can map patient-specific neoepitopes, the full spectrum of somatic mutations (136), and pre-treatment immune microenvironments (137, 138). When assimilated into clinical workflows, these layers of information can guide the judicious selection of immunotherapeutic regimens, be it checkpoint inhibitors, engineered T cell therapies, or peptide-based vaccines, so that each intervention is calibrated to the distinct immunogenic and oncogenic profile of the tumor at hand (88, 94, 108, 139).

To translate these molecular insights into clinical practice, we propose a hierarchical biomarker framework:

Recent improvements in next-generation sequencing and computational immunology have streamlined the characterization of neoantigens arising from non-synonymous mutations (40, 140). Machine learning (ML) algorithms can now evaluate peptide immunogenicity, enabling the selection of neoepitopes most likely to elicit durable T cell responses (141, 142). In the context of thyroid cancers, especially ATC, where a pronounced mutational burden is often present, such neoantigen-informed vaccines can function synergistically with ICIs by diversifying and sustaining the pool of tumor-specific cytotoxic T lymphocytes that can drive long-term disease control (6, 69, 95, 143, 144).

Reprograming the TME to counteract immunosuppressive barriers remains a highly promising therapeutic strategy (31). Approaches currently under investigation include selective depletion or re-education of tumor-associated macrophages (67), neutralization of myeloid-derived suppressor cell activity (58, 66), antagonization of immunosuppressive cytokines such as TGF-β and IL-10, and augmentation of dendritic cell maturation (33, 145). Such interventions, when sequenced or co-administered with ICIs or ACT, have the potential to convert immunologically inert, well-differentiated thyroid cancers into immunologically active and responsive tumors (31, 63, 69, 146).

Recent studies suggest the gut microbiome can influence both systemic immune homeostasis and the therapeutic response to immunotherapy (117, 147). Modulatory strategies, including dietary interventions, probiotics, and fecal microbiota transplantation, have been proposed to tilt systemic immune polarization toward a more pro-inflammatory profile, thereby augmenting therapeutic efficacy (148). Longitudinal studies of microbiome composition in thyroid cancer patients receiving immunotherapy are warranted to discern achievable microbial signatures that could serve as biomarkers or therapeutic targets (59, 149).

Advances in cellular engineering are poised to overcome the limitations imposed by the TME (68). Next-generation CAR-NK cells (150), T-cell receptor–engineered T cells, and cytokine-secreting armored CAR-T cells can be tailored for enhanced specificity (56, 86, 151), sustained persistence, and durable resistance to immunosuppressive conditions (31, 59). The application of “safety switch” technologies in these constructs can constrain on-target, off-tumor toxicities, thereby safeguarding normal thyroid tissue and adjacent organs from collateral damage while retaining therapeutic efficacy against tumor cells (126, 131).

The promise of future iodine-resistant thyroid cancer management hinges on rational, multimodal constructs that integrate surgery (17, 152–154), small-molecule targeted inhibitors, radiotherapy, and immunotherapy into a unified therapeutic roadmap (49, 155, 156). For instance, a neoadjuvant immunotherapy phase could induce a downward stage migration in locally advanced disease, thereby enabling a more complete surgical resection while simultaneously entrenching durable immunological memory that guards against later relapse (101, 157). Subsequent radiotherapy and targeted kinase blockade might be temporally interleaved to unleash tumor-associated antigen exposure and amplifying immune activation, after which a prolonged immunotherapy maintenance phase would consolidate the antitumor effect (49, 98, 113, 131, 158).

Despite the promising horizon of personalized immunotherapy, significant technological and logistical hurdles impede its widespread clinical adoption. Firstly, the “vein-to-vein” time interval remains a critical bottleneck. For patients with rapidly progressing malignancies like ATC, the weeks required to manufacture autologous CAR-T cells or personalized neoantigen vaccines may exceed their life expectancy. Consequently, a key transformation pathway lies in the development of “off-the-shelf” (allogeneic) cellular products and rapid-manufacturing vaccine platforms (e.g., mRNA technology) to ensure timely therapeutic accessibility. Secondly, bioinformatics standardization is urgently needed. While machine learning accelerates neoantigen discovery (as discussed in section 7.2), the lack of consensus on prediction algorithms leads to inter-institutional variability, hindering the validation of biomarkers across multi-center trials. Establishing harmonized computational pipelines will be a prerequisite for moving these precision tools from the laboratory to standard clinical practice.