CD8⁺ T cells are a crucial subset of T cells, showing the most significant difference after neoadjuvant chemoimmunotherapy (nCIT) (29). CD8⁺ T cells are T cells that have undergone differentiation and maturation to express CD8 molecules. They can secrete large amounts of IFN-γ and granzyme B to synergize against cancer cells (30). Numerous studies have found that nCIT increases CD8⁺ T cell infiltration more significantly than neoadjuvant chemotherapy (28), prolonging overall patient survival (31). And the same trend can be observed before and after nCIT, whereas regulatory T cells (Treg) show the opposite trend. One of the studies has indicated that the density of CD8⁺ tumor-infiltrating lymphocytes (TILs) in the pCR and major pathological response (MPR) groups is higher than in the pathological partial response (pPR) and pathological non-response (pNR) groups. Additionally, the ratio of CD8⁺ to FoxP3⁺ TILs in the pCR/MPR group is significantly greater than in the pPR/pNR group, suggesting that this ratio could emerge as a novel biomarker after nCIT in ESCC. Interestingly, in the patients not achieving pCR, CD8⁺ TILs were observed to be mostly distributed in the periphery of the tumor, correlating with a worse prognosis, while the same cellular distribution was found in the lymph nodes (32).

Based on the expression of antigenic clusters, CD8⁺ T cells are classified into different T cell subpopulations. In addition to the regulatory CD8⁺ T cell subset (33), there are Tc1 cells with the most typical cytotoxic function (34). In the tumor microenvironment, CD8⁺ T cells generally resist cancer cells by differentiating into cytotoxic T cells (29). However, T cells chronically exposed to persistent antigens or chronic inflammatory environments are prevented from acquiring memory cell homeostatic responsiveness (34–36). The immediate protection and recall response to antigens conferred by memory lymphocytes constitutes a critical component of the organism’s immune defense (37). Nowadays, studies have shown a decrease in the ratio of central memory cells (TCM) to effector memory T cells (TEM) in samples of patients who have achieved pCR/MPR, reversing the high pre-treatment TCM/TEM ratio (28). The process of TCM-to-TEM, which occurs on antigenic restimulation, produces greater anti-tumor capacity (38, 39).

T cells, during their process of progressive exhaustion, upregulate the expression of inhibitory pathway molecules, including PD-1 and CTLA-4 (40). Thus, ICB facilitates the reactivation of exhausted CD8⁺ T cells by blocking the inhibitory molecules. However, with an in-depth study, it was found that even with ICB, a large number of exhausted T cells (Tex) still existed in the TME (41) and it could be divided into ICI-permissive and ICI-resistant subpopulations (42, 43). However, two clusters of cells, CD8⁺ Tex-SPRY1 and CD8⁺ Tex-XAF1, were found to respond significantly to ICI treatment. Furthermore, the anti-tumor signature genes ENTPD1, CXCL13, and HLA-DR genes were enriched in these two CD8⁺ Tex subsets. The CD8⁺ Tex-SPRY1 cells were in the early stage of exhausted T cells and showed high similarity to progenitor CD8⁺ T cells. Moreover, these samples from patients who achieved pCR had higher expression of SPRY1 and CD8 in the tumor stroma and intra-tumoral regions before and after nCIT, which was not observed in patients who did not achieve pCR ( Figure 2 ). It is suggested that patients with high CD8⁺ Tex-SPRY1 cell infiltration have a higher sensitivity to nCIT and that CD8⁺ Tex-SPRY1 may be a potential biomarker for nCIT in ESCC.

CD8⁺ T cells showed a significant correlation with nCIT, but the heterogeneity of the TME also limits its stability as a biomarker. CD4⁺ T cells, unlike CD8⁺ T cells that can directly kill tumor cells, function mainly by releasing different cytokines to regulate other immune cells and play an equally important role as CD8⁺ T cells. CD4⁺ T cells represent a heterogeneous subset of T cells, and different levels of CD4⁺ T cell numbers are associated with different outcomes. Some studies have noted that a significant increase in CD4⁺ T cells is observed after nCIT compared to neoadjuvant chemotherapy alone and that higher CD4⁺/CD8⁺ ratios are associated with better pathological response (44). However, it has also been noted that the high level of CD4⁺ T cell distribution within a distance of 0-10 μm from the tumor cells predicts a low survival rate (45). Tumor-infiltrating CD4⁺ T cells exert their function by undergoing differentiation into various subsets directed by functional polarization. Moreover, the differentiation of these subsets is significantly influenced by the characteristics of the TME in which they exist. For example, Th1 cells exert antitumor effects, whereas Th2 cells counteract the effects of Th1 cells, and the Th9 is unchanged after nCIT (28).

Recently, a single-cell sequencing study classified CD4⁺ T cells into six major subsets based on specific markers. Among them, CD4-C1-CCR7 carries naïve signature genes, including TCF7, CCR7, LEF1, and SELL. CD4-C6-FOXP3 has the most distinctive Treg profile, expressing high levels of the Treg signature genes FOXP3, IKZF2, IL2RA, and CTLA-4. The study noted that Tregs are heavily enriched in ESCC and that Tregs accounted for more than 50% of the total CD4⁺ T cells in the tumors, compared to only 25% in the adjacent normal tissues (46). At the same time, numerous studies have indicated that Treg cells can significantly inhibit T cell-mediated antitumor effects (47). The density of Treg cells in ESCC tissues was higher than that in normal tissues. However, after nCIT, the number of Treg decreased while CD8⁺ T cells increased (28). FoxP3⁺ TILs were found in higher quantities in tumor samples from patients who achieved pNR compared to those who achieved pCR (32). It was also found that there exists a subset of T cells expressing CD4⁺CD25^-CD69⁺Foxp3^-LAP⁺, which inhibits T cell proliferation and promotes tumor immune escape by secreting TGF-β. This subset was positively correlated with the pathologic response (48). However, only high Treg density in ESCC tissues is not predictive of patient survival (45). The calculation of the ratio between Treg and CD4⁺ T cells is required to observe a positive correlation with better pathological response (49). Thus, the ratio of FoxP3⁺ T cells to CD4⁺ T cells is expected to be a biomarker for predicting the response to nCIT, as mentioned previously for CD8⁺ TILs/FoxP3⁺ TILs.

Macrophages are known as the “scavengers” of the human organism. They can not only phagocytose pathogens but also exert the function of presenting antigens by processing antigens to form MHC complexes. Macrophages migrate into the TME and become tumor-associated macrophages (TAMs), which play an important role in the initiation and regulation of antitumor immune responses (50). Recently, it was found that dendritic cells and macrophages were mostly distributed within 30 μm from tumor cells and that patients with closer spatial distance of TAM’s distribution tend to have better immune responses and longer survival (45). This is because chemotherapy causes tumor cells to release more tumor-associated antigens, and the closer the distance between TAMs and tumor cells, the better for TAMs to be exposed to these more new antigens, which provide immunomodulatory signals to the T cells to promote anti-tumor immune responses (51). This suggests that spatial distances between TAMs and tumor cells are potential predictive biomarkers (52).

TAMs can be activated by M1 and M2 types. M1-TAMs exert proinflammatory and antitumor effects (53), enhancing the antigen-presenting capacity of DC and augmenting tumoricidal effects of infiltrating T cells (54). Tumor tissues in the pCR group were observed to have a significant M1 polarization level before treatment and a significantly higher M1/M2 ratio than in the non-pCR group after treatment (29, 55). Among M1-type macrophages, Macro-MMP9 and Macro-FOLR2 subset produced large amounts of cytokines such as IL-15, CXCL1, and CXCL9, which induced the activation of CD8⁺ Tex-SPRY1 subset. Thus Macro-MMP9 and CD8⁺ Tex-SPRY1 form a positive feedforward cycle and anti-tumor ability (29). Whereas M2-TAMs could inhibit M1-TAMs effects to promote immunesuppression (56), the number of PD-L1⁺CD163⁺ cells infiltrated was significantly increased in the non-pCR group (57). Recently, it has been shown that macrophage migration inhibitory factor (MIF) promotes the conversion of M1-type TAMs to M2-type, and patients with high baseline MIF levels are significantly associated with adverse pathologic responses.

Macrophages exhibit high plasticity and heterogeneity (58). CCR4⁺CCR6⁺ TAM were found to be significantly reduced after nCIT, while the percentage of total macrophages in the tumor microenvironment was not changed. The CCR4/CCR6 chemokine system score was lower in tumor regression grading (TRG) of grade 0/1 compared to patients with TRG2/3, suggesting that the CCR4/CCR6 chemokine system would be a potential biomarker (28). Another subset of TAM, TAM-TREM2, has an immunosuppressive phenotype and is functionally close to the M2 subtype (59). Their high level of infiltration is associated with worse OS (60). Further studies showed that TAM-TREM2 expressing key genes of the complement system, C1Q, APOE, and SPP1, promote infiltration of Tex and Tregs as well as tumor proliferation (61) ( Figure 3 ). Meanwhile, the TIDE algorithm predicted that signature genes of TAM-TREM2 are associated with immune resistance. It is proposed that this subset is expected to be a predictive biomarker for nCIT (60).

In TME, dendritic cells (DCs), similar to macrophages, exert antigen-presenting function and regulate the immune response, thereby influencing tumor immune surveillance and escape (62). Among long non-coding RNAs, lncRNA ENST00000560647 has been shown to inhibit dendritic cell-mediated antigen presentation, leading to the inactivation of CD8⁺ T cells and the facilitation of immune evasion (63). DCs can differentiate into classic dendritic cell cDC and plasma cell-like dendritic cells pDC in TME (64). The current study found that the density of pDC in tumor tissues of patients not treated with nCIT is higher than in adjacent normal tissues, and then significantly decreased after receiving nCIT. This is consistent with the fact that pDCs produce IFN-α which leads to differentiation of Tregs and immune escape (65). The TCGA database survival analysis also illustrated that the characterized genes of pDC can differentiate patient survival well. Another DC subset, cDC, whose quantitative changes are mainly influenced by classical cDC type 1. cDC1 can activate CD4⁺ and CD8⁺ effector T cells, which are potentially responsive to immunotherapy (66). Therefore, the regulation of DC to exert its antitumor effects has become one of the hotspots in immunotherapy, including the use of tumor vaccines. Tumor vaccines utilize anti-CLEC9A antibodies to specifically deliver the immunogenic tumor antigen NY-ESO-1 to human CD141⁺ DC cells for anticancer therapy (67). Meanwhile, overexpression of MAGE-A3 and CALR on DCs to study their potential for generating anti-tumor immune responses is also one of them (68).

NK cells are the core cells of the organism’s innate immune cells. NK cells have a broader-spectrum immune role that not only is limited to fending off viruses but also has a crucial role in killing tumor cells (69). CD56⁺ is the characteristic gene of NK cells, and NK cells can be further categorized into two major subsets based on CD56 expression: CD56^bright and CD56^dim NK cells (70). CD56^bright NK is the precursor cell of CD56^dim NK (71), but CD56^dim NK expresses higher levels of FCγ receptor III (CD16) and possesses stronger natural cytotoxicity and antitumor effects (72, 73). The present study found that CD16⁺ cells were less in tumor tissues than in normal tissues and were significantly higher after nCIT, while CD16^- cells exhibited higher exhaustion scores (28). Meanwhile, CD56^dim NK cells were significantly reduced in the responders compared to non-responders within the nCIT group, whereas no significant change was observed in the neoadjuvant chemotherapy group. And, the pCR group had more abundant CD56^dim NK cells in the stromal area than the non-pCR group. These results show that CD16⁺CD56^dim NK cells are associated with a better pathological response (55). ICB could activate numerous PD-1-expressing NK cells (74, 75), and CD16 could strongly self-activate NK cells (76). These all support an immunotherapeutic approach using modified NK cells to resist tumors, known as chimeric antigen receptor-NK (CAR-NK) cell therapy.

CAR-NK cell therapy, from adoptive cell therapy (ACTs), involves the transduction of a specially designed chimeric antigen receptor into NK cells. This enables CAR-NK cells to specifically recognize antigens on the surface of tumor cells and be activated, subsequently releasing cytotoxic molecules to eliminate tumor cells (77). Research on adoptive cell therapy, which began with tumor-infiltrating T cells, has demonstrated promising efficacy in hematological malignancies. However, the development of CAR-T therapy is constrained by issues related to infiltration into tumor tissues, as well as resistance to the tumor microenvironment (TME). Compared to CAR-T therapy, CAR-NK therapy can reduce the cytokine release syndrome (CRS) and neurological toxicities associated with CAR-T treatment (78). These advantages position CAR-NK as a potential “universal” product (79). CD19, CD20, and CD33 are the main targets for CAR-NK cells, which are frequently expressed in hematological malignancies (80). Additionally, CAR-NK cells have been shown to be effective against solid tumor targets such as Her2, EpCAM, and EGFR, which are common in colorectal and ovarian cancers (81–83). This broadens the applicability of CAR-NK cell therapy. Currently, CAR-NK cell therapy faces challenges such as viral transduction difficulties. Researchers are developing CAR-NK cells capable of recognizing multiple antigens to reduce the risk of tumor escape due to the loss of a single antigen (84). Moreover, CAR-NK cell therapy is being explored in combination with other treatment modalities to expand the applicability of immunotherapy. It is anticipated that ACTs will also become a milestone in immunological treatment.

B cells are the major humoral immune cells, exerting antibody-dependent cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP) (85). There is growing evidence that tumor-infiltrating B cells (TIL-B) and plasma cells have a crucial role in immunomodulation (86). One study showed a significant infiltration of B cells in a subgroup of “immune activation-dominant” samples. COL19A1, as a receptor for TIL-B, positively correlates with pathological responses and is expected to be a biomarker predicting the prognosis of nCIT (87). The present study classified TIL-B into five major subsets: naïve B cells (NBCs); activated B cells (ABCs); memory B cells (MBCs); germinal center B cells (GCBs), and the ASC (88). Two of these B cell subsets are sensitive to immunotherapy, GCB-LMO2 and CD55⁺ Bm. First, patients with a pathological response showed higher GCB cell infiltration than pathological non-response patients, and a significant increase in the interaction between the GCB-LMO2 and the CD4⁺ T-CXCL13 subsets was observed. GCB-LMO2 expresses the conventional GCB cell markers GL-7 and CD23, which are distributed within TLS and promote an antitumor immune response. CD55⁺ Bm belongs to the group of memory B cells that express relevant genes such as CD20, which are distributed outside the TLS. The expression of its signature gene, CD55, is down regulated with increased expression of LMO2. CD55⁺ Bm expresses ADGRE5, which inhibits CD8⁺ T cells and releases cytokines that interact with Treg cells to inhibit responsiveness to immunotherapy ( Figure 4 ). It is illustrated that comprehensive B cell profile characteristics can predict nCIT responsiveness (89).

Tertiary lymphoid structure (TLS) refers to the organization of immune cell aggregates that resemble lymphoid tissues, formed within non-lymphoid tissues. In the physiological state, TLS is not present in normal tissues and is mainly found in the infiltrative margins of tumors (90). In recent years, TLS has been found to have a positive effect on the prognosis of many cancers, including lung cancer (91) and ovarian cancer (92). However, both RFS and OS were higher in the TLS-positive group than in the TLS-negative group, and the mature TLS group containing germinal centers (GCs) had better long-term efficacy and prognosis (93, 94). After nCIT, TME with abundant TLS was the most common type among the patients who achieved pCR or MPR (95, 96). Moreover, a study has developed a pathology image model using deep learning to quantitatively calculate the TLS ratio for predicting the pathological response to nCIT in ESCC. In TME, TLS is constructed by GCBs, T cells, and B cells (7). And it can only be activated in the presence of GCB-LMO2 cells. Meanwhile, patients with TLS high in CD55⁺ Bm cells and low in GCB-LMO2 cells were stratified as having the worst clinical outcome (89). This indicates that the antigen-presenting role of GCB cells contributes to the maturation of cytotoxic T cell (97). Follicular B cells (Bfo-NEIL1), which originate from GCB and express NEIL1, co-localize with CD8⁺ Tex-SPRY1 within TLS and exhibit better pathological responses. This co-localization demonstrates a synergistic interaction that significantly improves RFS. Bfo-NEIL1 cells express TNF and IL-23 to activate CD8⁺ Tex-SPRY1 and promote the positive feedforward cycle between Macro-MMP9 cells and CD8⁺ Tex-SPRY1. Additionally, CD4⁺ T-CXCL13 cells were induced by B-T interaction to activate the Tfh phenotype, which in turn activated CD8⁺ Tex-SPRY1 again. ( Figure 5 ) Meanwhile, CD8⁺ Tex-SPRY1 induced the activity of germinal centers and the TLS, coinciding with the recent studies that the CD8⁺ response is present in the TLS (90, 98).