The Proximity of PD-1⁻CD103⁺ Tissue-Resident CD8⁺ T Cells to Tumor Cells Is Correlated with Improved Clinical Outcomes in Patients with Cholangiocarcinoma

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors made an interesting analysis of the prognostic role of CD8 T cells in cholangiocarcinoma. The paper is well done and the analysis is accurate. English language is proper. Current scientific knowledge has not yet reached a detailed knowledge about neoplastic microenvironment and its role in tumor response to treatments. The authors suggest that CD8+ TRM cells combined with their spatial localization represent a promising frontier in the field of tumor prognosis, leading to possible future tailored therapies. For the future, I encourage the colleagues to further expand their expertise in tumor microenvironment examination, because it seems to represent a promising frontier in oncological treatments.

Author Response

(Reviewer 1):

The authors made an interesting analysis of the prognostic role of CD8 T cells in cholangiocarcinoma. The paper is well done and the analysis is accurate. English language is proper. Current scientific knowledge has not yet reached a detailed knowledge about neoplastic microenvironment and its role in tumor response to treatments. The authors suggest that CD8+ TRM cells combined with their spatial localization represent a promising frontier in the field of tumor prognosis, leading to possible future tailored therapies. For the future, I encourage the colleagues to further expand their expertise in tumor microenvironment examination, because it seems to represent a promising frontier in oncological treatments.

Response to Reviewer 1:

Thank you for the positive assessment of our study and for the thoughtful comment. We fully agree that the tumor microenvironment (TME) remains incompletely understood, particularly regarding how specific immune cell states and their spatial organization influence therapeutic responsiveness. To address this important point and to better contextualize our findings, we have expanded the Discussion to emphasize the emerging role of tissue-resident memory (T_RM) cells and their spatial localization as a promising frontier for prognosis and future tailored therapies.

Specifically, CD103⁺CD8⁺ T_RM cells in lung malignancies and other solid tumors have been repeatedly associated with improved clinical outcomes, yet their mechanistic contribution to anti-tumor immunity, particularly their involvement in responses to PD-1 blockade, remains incompletely defined. In NSCLC, multiplex immunofluorescence analyses indicate that responders to immunotherapy often show a marked increase in intratumoral CD103⁺CD8⁺ T cell density, whereas non-responders do not. These tumor-localized T_RM cells frequently co-express CD49a and CD69 and exhibit elevated PD-1 expression, distinguishing them from CD103^-CD8⁺ TILs. Across solid tumors, T_RM populations also commonly co-express multiple inhibitory checkpoint molecules consistent with functional exhaustion [1]. However, it remains unclear whether PD-1⁺ versus PD-1^- subsets within the CD103⁺CD8⁺ tissue-resident compartment differentially influence responsiveness to anti-PD-1 therapy [2]. Notably, in leukemia patients without anti-PD-1 treatment, better prognosis correlates with a higher proportion of CD103⁺CD8⁺ T cells showing relatively low PD-1 and TIGIT expression [3], which is conceptually consistent with our observation that close spatial proximity between PD-1^-CD103⁺CD8⁺ T_RM cells and tumor cells predicts improved survival in cholangiocarcinoma (CCA). Additional studies report that co-expression of immunosuppressive receptors (e.g., PD-1 and TIGIT) is associated with impaired effector function, and dual blockade of PD-1 and TIGIT can significantly enhance CD8⁺ T cell cytotoxicity [4]. Taken together, these data support the notion that T_RM cells and their spatial positioning relative to tumor cells may provide complementary prognostic information and a rationale for future personalized immunotherapeutic strategies; mechanistically, anti-PD-1 therapy may partially restore effector function in exhausted PD-1⁺ T_RM cells and may be accompanied by dynamic spatial redistribution that could contribute to treatment efficacy.

In line with the reviewer’s encouragement, we will further expand our investigation of the TME in future work (including deeper phenotypic/functional dissection of T_RM subsets and broader spatially resolved profiling), and the above clarifications and perspective have been added to the revised Discussion.

[1]. Gitto, S.; Natalini, A.; Antonangeli, F.; Di Rosa, F. The Emerging Interplay Between Recirculating and Tissue-Resident Memory T Cells in Cancer Immunity: Lessons Learned From PD-1/PD-L1 Blockade Therapy and Remaining Gaps. Front Immunol 2021, 12, 755304, doi:10.3389/fimmu.2021.755304.

[2]. Corgnac, S.; Malenica, I.; Mezquita, L.; Auclin, E.; Voilin, E.; Kacher, J.; Halse, H.; Grynszpan, L.; Signolle, N.; Dayris, T.; et al. CD103(+)CD8(+) T(RM) Cells Accumulate in Tumors of Anti-PD-1-Responder Lung Cancer Patients and Are Tumor-Reactive Lymphocytes Enriched with Tc17. Cell Rep Med 2020, 1, 100127, doi:10.1016/j.xcrm.2020.100127.

[3]. Liu, L.; Lai, W.; Zhuo, X.; Chen, S.; Luo, X.; Tan, H. Higher frequency of peripheral blood CD103(+)CD8(+) T cells with lower levels of PD-1 and TIGIT expression related to favorable outcomes in leukemia patients. Front Immunol 2024, 15, 1437726, doi:10.3389/fimmu.2024.1437726

[4]. Banta, K.L.; Xu, X.; Chitre, A.S.; Au-Yeung, A.; Takahashi, C.; O'Gorman, W.E.; Wu, T.D.; Mittman, S.; Cubas, R.; Comps-Agrar, L.; et al. Mechanistic convergence of the TIGIT and PD-1 inhibitory pathways necessitates co-blockade to optimize anti-tumor CD8(+) T cell responses. Immunity 2022, 55, 512–526.e519, doi:10.1016/j.immuni.2022.02.005.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

General concept comments

-The article presents many results and has a significant impact. However, it has some lengths that could be reduced without losing the content (the length of the Simple Summary and Materials and Methods section could be reduced).

-The article contains some errors in the figures, legends and text, in particular :

• Figure 4D : Kaplan-Meier survival curves based on spatial proximity of PD-1- naïve T_RM cells to tumor cells (Logrank, p=0.0163). In the graph, wouldn't the blue curve (short-term survival) and the red curve (long-term survival) be reversed?
• Figure 5, Legend on line 391: “The spatial proximity of PD-1+T_RM cells to tumor cells independently predicts favorable clinical outcomes.” PD-1+ T_RM to be changed to PD-1- T_RM
• Discussion section, lines 454-455 “First, the abundance and relative location of PD-1+T_RM cells in relation to tumors could serve as a predictive biomarker for an overall favorable outcome.” PD-1+ T_RM to be changed to PD-1- T_RM
• Discussion section, lines 461 “…similar to the positive spatial phenotype of the PD-1 subset.” Positive to be changed to negative
• Discussion section, lines 476-477 “Proximity of PD-1+CD103+CD8+ T_RM cells to tumor cells is linked to improved clinical outcomes.” PD-1+CD103+CD8+ T_RM to be changed to PD-1-CD103+CD8+ T_RM

Comments for author File: Comments.pdf

Author Response

(Reviewer 2):

-The article presents many results and has a significant impact. However, it has some lengths that could be reduced without losing the content (the length of the Simple Summary and Materials and Methods section could be reduced).

-The article contains some errors in the figures, legends and text, in particular :

• Figure 4D : Kaplan-Meier survival curves based on spatial proximity of PD-1- naïve T_RM cells to tumor cells (Logrank,p=0.0163). In the graph, wouldn't the blue curve (short-term survival) and the red curve (long-term survival) be reversed?
• Figure 5, Legend on line 391: “The spatial proximity of PD-1+T_RMcells to tumor cells independently predicts favorable clinical outcomes.” PD-1+ T_RM to be changed to PD-1- T_RM
• Discussion section, lines 454-455 “First, the abundance and relative location of PD-1+T_RMcells in relation to tumors could serve as a predictive biomarker for an overall favorable outcome.” PD-1+ T_RM to be changed to PD-1- T_RM
• Discussion section, lines 461 “…similar to the positive spatial phenotype of the PD-1 subset.” Positive to be changed to negative
• Discussion section, lines 476-477 “Proximity of PD-1+CD103+CD8+ T_RMcells to tumor cells is linked to improved clinical outcomes.” PD-1+CD103+CD8+ T_RM to be changed to PD-1-CD103+CD8+ T_RM
•  

Response to Reviewer 2:

We sincerely thank the reviewer for the meticulous review and constructive suggestions. Below, we address each point in detail:

- Manuscript length

We have thoroughly revised and condensed the Simple Summary and Materials and Methods sections to clarity and improve conciseness while preserving all essential information. All revisions are reflected in the revised manuscript.

- Errors in figures, legends, and text

• Figure 4D: To eliminate ambiguity between spatial distance and survival duration , we have:

1. Revised all spatial distance descriptors to "proximal" (≤ median distance) and "distal" (> median distance) for the proximity.
2. Survival duration has been consistently defined as "short-term" (<30 months) and "long-term" (≥30 months).

• Figure 5, Legend (original line 391, revised line 394): "PD-1⁺T_RM" has been corrected to "PD-1^- T_RM".
• Discussion, originallines 454–455, revised line 470–471: "PD-1⁺ T_RM" has been corrected to "PD-1^- T_RM".
• Discussion, originalline 461, revised line 487: "positive spatial phenotype" has been corrected to "PD-1^- T_RM-like state".
• Discussion, originallines 476-477, revised line 504-505: "PD-1⁺CD103⁺CD8⁺ T_RM" has been corrected to "PD-1^-CD103⁺CD8⁺ T_RM".

All corrections have been implemented in the revised manuscript (lines 394, 470-471, 487, 504-505, and Figure 4D). We appreciate the reviewer’s attention to detail, which has significantly improved the manuscript’s clarity and accuracy.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This article is quite innovative, but there are some issues that need to be addressed. First and foremost, it is crucial that the genes related to the regulation of CD8+ T cells have not been clearly characterized. The underlying regulatory mechanisms also require further experimental clarification. The application of PD-1-CD103+CD8+T cells in previous studies has been clarified. However, the innovativeness of this study still needs to be clarified. At the same time, the relationship between PD-1-CD103+CD8+T cells and immunotherapy should be studied. Moreover, if feasible, it is recommended to conduct additional preclinical in vivo experimental studies on the combination of targeted PD-1-CD103+CD8+T cells and immunotherapy.

Author Response

(Reviewer 3):

This article is quite innovative, but there are some issues that need to be addressed. First and foremost, it is crucial that the genes related to the regulation of CD8+ T cells have not been clearly characterized. The underlying regulatory mechanisms also require further experimental clarification. The application of PD-1-CD103+CD8+T cells in previous studies has been clarified. However, the innovativeness of this study still needs to be clarified. At the same time, the relationship between PD-1-CD103+CD8+T cells and immunotherapy should be studied. Moreover, if feasible, it is recommended to conduct additional preclinical in vivo experimental studies on the combination of targeted PD-1-CD103+CD8+T cells and immunotherapy.

Response to Reviewer 3:

Thank you for the constructive comments, we fully agree that the mechanistic and regulatory layers of CD8⁺ T cell biology in the tumor microenvironment remain incompletely resolved and are critical for translating descriptive immune features into actionable therapeutic strategies. In the revised Discussion, we therefore contextualize our findings within the evolving literature on tissue-resident memory (T_RM) cells and immunotherapy, while also clearly delineating the scope and novelty of our work.

Specifically, the presence of CD103⁺CD8⁺ T_RM cells in lung malignancies and other solid tumors has been consistently associated with improved clinical outcomes; however, their precise role in mediating anti-tumor immunity, particularly their involvement in the therapeutic response to PD-1 blockade, remains incompletely understood and is an emerging focus of research. In NSCLC, multiplex immunofluorescence studies suggest that responders to immunotherapy typically show increased intratumoral CD103⁺CD8⁺ T cell density, whereas non-responders do not, and these tumor-localized T_RM cells often co-express CD49a/CD69 and display elevated PD-1 compared with CD103^-CD8⁺ TILs. Across multiple solid tumors, T_RM populations frequently co-express inhibitory checkpoint molecules consistent with functional exhaustion [1], yet it remains unclear whether PD-1⁺ versus PD-1^- subsets within the T_RM compartment differentially influence responsiveness to anti-PD-1 therapy [2]. Importantly, in leukemia patients without anti-PD-1 treatment, better prognosis correlates with a higher proportion of CD103⁺CD8⁺ T cells exhibiting relatively low PD-1 and TIGIT expression [3], which aligns with our observation that close spatial proximity between PD-1^-CD103⁺CD8⁺ T_RM cells and tumor cells predicts improved survival in cholangiocarcinoma (CCA). Additional work has reported that co-expression of immunosuppressive receptors (e.g., PD-1 and TIGIT) on T_RM cells is associated with impaired effector function, and that dual inhibition of PD-1 and TIGIT can significantly enhance CD8⁺ T cell cytotoxicity [4]. Collectively, these data support our central rationale that the PD-1 status of CD103⁺CD8⁺ T_RM cells and, critically, their spatial positioning relative to tumor cells may provide complementary prognostic and translational insight, which extends prior work that largely focused on overall T_RM abundance or checkpoint-positive exhausted subsets, and generates testable hypotheses regarding differential interactions with immunotherapy.

Finally, we acknowledge that deeper characterization of gene regulatory programs and mechanistic validation (including the relationship between PD-1^-CD103⁺CD8⁺ T_RM cells and immunotherapy as well as, if feasible, preclinical in vivo testing of strategies targeting these cells in combination with checkpoint blockade) are important next steps; we have explicitly added these clarifications and future directions in the revised Discussion.

[1]. Gitto, S.; Natalini, A.; Antonangeli, F.; Di Rosa, F. The Emerging Interplay Between Recirculating and Tissue-Resident Memory T Cells in Cancer Immunity: Lessons Learned From PD-1/PD-L1 Blockade Therapy and Remaining Gaps. Front Immunol 2021, 12, 755304, doi:10.3389/fimmu.2021.755304.

[2]. Corgnac, S.; Malenica, I.; Mezquita, L.; Auclin, E.; Voilin, E.; Kacher, J.; Halse, H.; Grynszpan, L.; Signolle, N.; Dayris, T.; et al. CD103(+)CD8(+) T(RM) Cells Accumulate in Tumors of Anti-PD-1-Responder Lung Cancer Patients and Are Tumor-Reactive Lymphocytes Enriched with Tc17. Cell Rep Med 2020, 1, 100127, doi:10.1016/j.xcrm.2020.100127.

[3]. Liu, L.; Lai, W.; Zhuo, X.; Chen, S.; Luo, X.; Tan, H. Higher frequency of peripheral blood CD103(+)CD8(+) T cells with lower levels of PD-1 and TIGIT expression related to favorable outcomes in leukemia patients. Front Immunol 2024, 15, 1437726, doi:10.3389/fimmu.2024.1437726

[4]. Banta, K.L.; Xu, X.; Chitre, A.S.; Au-Yeung, A.; Takahashi, C.; O'Gorman, W.E.; Wu, T.D.; Mittman, S.; Cubas, R.; Comps-Agrar, L.; et al. Mechanistic convergence of the TIGIT and PD-1 inhibitory pathways necessitates co-blockade to optimize anti-tumor CD8(+) T cell responses. Immunity 2022, 55, 512–526.e519, doi:10.1016/j.immuni.2022.02.005.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The authors aimed to identify the functional significance of tissue-resident memory T cells, particularly how their spatial positioning relative to tumor cells influences anti-tumor immunity to support development of immunotherapies. Through integration of mIHC and scRNA-seq data, they found intra-tumor PD-1⁺CD8⁺ T cells correlate with poorer survival. The authors further compare the tissue-resident memory T cells which make up a larger percentage of total CD8⁺ T cells in tumor than in adjacent normal tissue, dividing them into those that are naïve (PD-1^-) and exhausted (PD-1⁺).  They determine that PD-1^- CD8⁺ Trm cells’ proximity to tumor cells is a prognostic factor for survival in cholangiocarcinoma, whereas PD-1⁺ cells of the same subtype are associated with worse survival and are farther away from tumor cells. They suggest this indicates effective immune surveillance requires not only the presence of specific T cell subsets, but also their location relative to tumor cells. Finally, scRNAseq analysis delineates some of the gene signatures associated with the two Trm subsets, with the PD-1- Trm T cells showing enrichment in effector genes and a cytotoxic signature, while the PD-1+ Trm T cells display checkpoint gene enrichment and an exhaustion signature.. Overall, this paper is a useful entry in the literature, but requires extensive revision to be considered for publication.

Major Comments:

1. A major issue is that the language throughout is inappropriately strong, making statements that imply causality when only correlation has been shown. All causality language should be replaced with neutral language. This includes but is not limited to:
   1. The title “yields improved outcomes”
   2. Line 228 “outcomes were primarily influenced by”
   3. Line 235 “had a strong negative impact”
   4. Line 248 “crucial in determining”
   5. Line 332 “should be considered as their role in killing cancer cells”
   6. Line 384 “precisely delineates” (this is more just overly strong)
   7. Line 442 “this functional capacity in only achieved…”
2. Most importantly, there is little use of multiple testing or false discovery rate correction throughout the manuscript or discussion of why these were not used. As it stands, the p-values listed are mostly borderline significant and likely will not hold under FDR correction.
3. The number of patients and samples and the platforms they were run on should be clearly described at the beginning of the Results instead of only in the Methods. Also, there should be a description of sample overlap between the mIHC and scRNAseq. Finally, there needs to be a description of CCA subtypes included in Supp Table 1, iCCA vs eCCA. The two subtypes have different survival outcomes and should not be mixed together without clear delineation.
4. Lines 227-230, it is not very clear what is meant by survival being correlated with “relative abundance within the tumor nests” but not with “intra-tumor density”. Relatedly, “tumor nests” should be more clearly defined with examples shown.
5. Figure 4 is unclear as currently written.
   1. It is not clear what constitutes short- vs long-term survival and how the cutoff was determined. The same goes for short vs long proximity.
   2. 4D is very confusing as “short” and “long” are used both to refer to survival and to proximity… please make this more clear.
   3. 4D does not contain any correlation metrics between distance and survival, yet a log-rank =0.0163 is linked to “proximity” in the text, this is very difficult to understand as there is only a p=0.06 between average distance and short vs long-term groups. The overall significance is unclear.
   4. It is not clear why two different proximity measurements are used between Fig. 4 and Fig. 5, this needs to be clarified.
6. 6 text describes several specific activation and exhaustion genes, but these are mostly not individually labeled in either the volcano plot (Fig. 6A) or the heatmap (Fig. 6C), preventing clear assessment by the reader of these important markers. This is especially glaring given the GSEA p-value of 0.168 (Figure 6E) was not significant.
7. The manuscript would benefit from clarification on the role of HRA000863 data in this manuscript vs. the TMA described in Line 109, especially as the following section 2.2 seemingly indicates the scRNA-seq processing was performed on the TMA.
8. Further detail on cell type annotation beyond “canonical lineage markers” in Line 135 is necessary.
9. Effect sizes need to be reported throughout when reporting findings, not only p-values.
10. The manuscript suggests improved clinical outcomes, though the only metric provided is survival. Language change reflecting the correlation with increased survival would more accurately present their data.
11. There is little detail on whether the samples processed are treatment naïve or following some form of therapy.

Minor Comments:

1. Lines 235-241 seem to have a lot of redundancy, saying the same thing multiple times, please correct this.
2. Be sure to clearly say at each new section which dataset is being described, for example line 270 should clearly say scRNAseq data.
3. Line 111 describes 67 of the 70 patients underwent curative resection, though this seems abnormally high given line 117 states 15 of these 67 patients were Stage IV.
4. Information on the total number of cells could be better described.
5. Clarification that the TMA processed is a single TMA would help as there is no detail provided on integration when discussing the single-cell processing.
6. Line 382 is an incomplete sentence.
7. Minor language change to reflect “predictive” value as “prognostic” would more accurately present findings.
8. Clarification on why the representative mIHC chosen as representative would be beneficial.
9. Figure legends throughout would benefit from specification of number of patients, samples, cells, etc.
10. Figure 4 highlights that PD-1^- T_RM confers a survival advantage, while PD-1⁺ T_RM correlates with poor prognosis (Line 342-343), as does Section 3.5 where Lines 357-358 states that high proximity of the naïve T_RM subset served as a predictor of a favorable prognosis. However, Figure 5 states that proximity PD-1⁺ T_RM cells predicts favorable outcomes (Lines 391-392). Presumably, this is a simple typo since the rest of the figure references PD-1^- T_RM.

Author Response

(Reviewer 4):

The authors aimed to identify the functional significance of tissue-resident memory T cells, particularly how their spatial positioning relative to tumor cells influences anti-tumor immunity to support development of immunotherapies. Through integration of mIHC and scRNA-seq data, they found intra-tumor PD-1⁺CD8⁺ T cells correlate with poorer survival. The authors further compare the tissue-resident memory T cells which make up a larger percentage of total CD8⁺ T cells in tumor than in adjacent normal tissue, dividing them into those that are naïve (PD-1^-) and exhausted (PD-1⁺).  They determine that PD-1^- CD8⁺ Trm cells’ proximity to tumor cells is a prognostic factor for survival in cholangiocarcinoma, whereas PD-1⁺ cells of the same subtype are associated with worse survival and are farther away from tumor cells. They suggest this indicates effective immune surveillance requires not only the presence of specific T cell subsets, but also their location relative to tumor cells. Finally, scRNAseq analysis delineates some of the gene signatures associated with the two Trm subsets, with the PD-1- Trm T cells showing enrichment in effector genes and a cytotoxic signature, while the PD-1+ Trm T cells display checkpoint gene enrichment and an exhaustion signature.. Overall, this paper is a useful entry in the literature, but requires extensive revision to be considered for publication.

Major Comments:

1. A major issue is that the language throughout is inappropriately strong, making statements that imply causality when only correlation has been shown. All causality language should be replaced with neutral language. This includes but is not limited to:

1. The title “yields improved outcomes”
2. Line 228 “outcomes were primarily influenced by”
3. Line 235 “had a strong negative impact”
4. Line 248 “crucial in determining”
5. Line 332 “should be considered as their role in killing cancer cells”
6. Line 384 “precisely delineates” (this is more just overly strong)
7. Line 442 “this functional capacity in only achieved…”
   2. Most importantly, there is little use of multiple testing or false discovery rate correctionthroughout the manuscript or discussion of why these were not used. As it stands, the p-values listed are mostly borderline significant and likely will not hold under FDR correction.
   3. The number of patients and samples and the platforms they were run on should be clearly described at the beginning of the Results instead of only in the Methods. Also, there should be a description of sample overlap between the mIHC and scRNAseq. Finally, there needs to be a description of CCA subtypes included in Supp Table 1, iCCA vs eCCA. The two subtypes have different survival outcomes and should not be mixed together without clear delineation.
   4. Lines 227-230, it is not very clear what is meant by survival being correlated with “relative abundance within the tumor nests” but not with “intra-tumor density”. Relatedly, “tumor nests” should be more clearly defined with examples shown.
   5. Figure 4 is unclear as currently written.
8. It is not clear what constitutes short- vs long-term survival and how the cutoff was determined. The same goes for short vs long proximity.
9. 4D is very confusing as “short” and “long” are used both to refer to survival and to proximity… please make this more clear.
10. 4D does not contain any correlation metrics between distance and survival, yet a log-rank =0.0163 is linked to “proximity” in the text, this is very difficult to understand as there is only a p=0.06 between average distance and short vs long-term groups. The overall significance is unclear.
11. It is not clear why two different proximity measurements are used between Fig. 4 and Fig. 5, this needs to be clarified.
    6. 6 text describes several specific activation and exhaustion genes, but these are mostly not individually labeled in either the volcano plot (Fig. 6A) or the heatmap (Fig. 6C), preventing clear assessment by the reader of these important markers. This is especially glaring given the GSEA p-value of 0.168 (Figure 6E) was not significant.
    7. The manuscript would benefit from clarification on the role of HRA000863 data in this manuscript vs. the TMA described in Line 109, especially as the following section 2.2 seemingly indicates the scRNA-seq processing was performed on the TMA.
    8. Further detail on cell type annotation beyond “canonical lineage markers” in Line 135 is necessary.
    9. Effect sizes need to be reported throughout when reporting findings, not only p-values.
    10. The manuscript suggests improved clinical outcomes, though the only metric provided is survival. Language change reflecting the correlation with increased survival would more accurately present their data.
    11. There is little detail on whether the samples processed are treatment naïve or following some form of therapy.

Minor Comments:

1. Lines 235-241 seem to have a lot of redundancy, saying the same thing multiple times, please correct this.
2. Be sure to clearly say at each new section which dataset is being described, for example line 270 should clearly say scRNAseq data.
3. Line 111 describes 67 of the 70 patients underwent curative resection, though this seems abnormally high given line 117 states 15 of these 67 patients were Stage IV.
4. Information on the total number of cells could be better described.
5. Clarification that the TMA processed is a single TMA would help as there is no detail provided on integration when discussing the single-cell processing.
6. Line 382 is an incomplete sentence.
7. Minor language change to reflect “predictive” value as “prognostic” would more accurately present findings.
8. Clarification on why the representative mIHC chosen as representative would be beneficial.
9. Figure legends throughout would benefit from specification of number of patients, samples, cells, etc.
10. Figure 4 highlights that PD-1^-T_RM confers a survival advantage, while PD-1⁺ T_RM correlates with poor prognosis (Line 342-343), as does Section 3.5 where Lines 357-358 states that high proximity of the naïve T_RM subset served as a predictor of a favorable prognosis. However, Figure 5 states that proximity PD-1⁺ T_RM cells predicts favorable outcomes (Lines 391-392). Presumably, this is a simple typo since the rest of the figure references PD-1^- T_RM. 

Response to Reviewer 4:

We sincerely thank the reviewer for thorough evaluation and insightful suggestions. All comments have been meticulously addressed below, with revisions implemented throughout the manuscript.

Major Comments

1. We sincerely thank the reviewer for the thoughtful and constructive comment. In response, we have carefully revised the manuscript to replace the language implying causality with neutral, correlational phrasing.
2. The title “yields improved outcomes” has been changed to “is correlatedwith improved outcomes”.
3. Original line 228, revised line 227“outcomes were primarily influenced by” now reads “outcomes were primarily associated with”.
4. Original line 235, revised line235 “had a strong negative impact” has been replaced with “were associated with reduced”.
5. Original line 248, revised line243 “crucial in determining” has been reworded to “CCA patient survival is associated with”.
6. Original line 332, revised line332 “should be considered as their role in killing cancer cells” has been revised to “may reflect their potential involvement in anti-tumor immune responses”.
7. Original line 384, revised line388 “precisely delineates” (this is more just overly strong) has been softened to “identifies”.
8. Original line 442, revised line458 “this functional capacity in only achieved…” has been changed to “the association with a functional phenotype occurs predominantly”.

In addition, we have reviewed the entire text for similar instances of imprecise or overly assertive phrasing and revised them accordingly to maintain a neutral language. All these changes have been clearly highlighted in the marked version of the revised manuscript for ease of review.

2. Multiple Testing Correction

In the original manuscript, most analyses were performed using conventional statistical tests (e.g., t-tests and Logrank tests), while FDR correction was applied to scRNA-seq analyses (Figure 6 and supplementary materials). We acknowledge that some reported p-values are marginally significant, which is likely due to the limited clinical sample size (67 cases in the TMA cohort and 14 cases in the scRNA-seq cohort), resulting in reduced statistical power.

To address concern, we have now applied FDR correction to the survival and boxplot-based analyses and included the adjusted results in Supplementary Table S5. Although statistical significance was modestly attenuated after correction, the overall trends remained unchanged and did not alter our main conclusion that PD-1^- T_RM cells are associated with favorable prognosis in cholangiocarcinoma.

3. Sample Size, Platforms, and Cohort Details

1. Added to Results section 3.1:“About these two independent cohorts, mIHC was performed on 67 primary iCCA samples; scRNA-seq was derived from 14 iCCA patients with paired tumor and adjacent normal tissues.”
2. The TMA and scRNA-seq data were derived from two independent cohorts: the mIHC dataset originates from clinical tissue sections collected in our study, while the scRNA-seq dataset was downloaded from a public database. Additionally, both cohorts exclusively included iCCA (not eCCA), and this clarification has been added to the Methods, Results and table S1.
   4. Terminology for Tumor Nests and Density

Original lines 227-230, revised line 228-230, the terms "tumor nest" and "intra-tumor" were used interchangeably, which caused confusion. In fact, both refer to tumor compartments identified by software-based histological segmentation. We have now standardized the terminology to consistently use "intra-tumor".

5. Survival Periods and Spatial Proximity Definitions

The definitions and descriptions of survival duration and spatial proximity have been revised as follows:

1. Regarding tumor distance terminology, the original terms "long" and "short" were prone to ambiguity with survival. We have now adopted the anatomical terms "proximal" and "distal" based on median cutoff values.
2. The statistical significance in survival differences between proximity groups versus the non-significant spatial distribution between survival groups stems from our initial median-based grouping, which ignored clinical timelines and yielded irrelevant categories. Given the 5-year (60-month) follow-up, we have now adopted a clinically grounded 30-month cutoff to stratify patients into distinct survival groups, complementing Kaplan–Meier survival analyses.
3. 4 and Fig. 5 employ complementary spatial metrics: spatial distance quantifies CD8⁺T_RM–tumor proximity (Section 3.4), while G-cross analysis characterizes spatial distribution and density gradients (Section 3.5), yielding distinct biological insights. These complementary approaches have been specifically elaborated in Results sections 3.4 and 3.5 respectively, with their biological implications clearly differentiated.

     6 Gene Annotations and GSEA statistics

1. Key genes (e.g., KLRC1, CTLA4) are now labeled in Fig6A (volcano plot), for easy reference by readers.
2. While our primary focus was the significant upregulation of the cytotoxic signature in PD-1^-T_RM cells, the initial presentation of the exhaustion signature trend in PD-1⁺ T_RM cells required rigorous validation. Re-examination of the code and raw data identified a technical error: due to an R script parsing issue, the 167-gene exhaustion signature geneset from Zheng et al. (Nat Commun 2020, 11:6268, PMID: 33293583), was inadvertently truncated to the first 27 genes during input. This severely compromised statistical power, yielding a non-significant p-value. After correcting the script to import the full gene set and re-running GSEA, the exhaustion signature in PD-1⁺ TRM cells achieved strong statistical significance (p = 0.001), robustly supporting the biological interpretation of an exhausted phenotype in this subset.

7. Data Source Clarification

The scRNA-seq data (HRA000863) and TMA originate from two distinct cohorts, and this distinction has been explicitly clarified in the text.

8. Cell Annotation Details

Original lines 135, revised line 130: The signature markers used for clustering in the cell type annotations, such as T cells (CD3E, CD4, CD8A) and B cells (CD19, MS4A1, CD79A), have been supplemented.

9. Effect Sizes Reported

All statistical measures actually included in each chart have been labeled.

10. Clinical Outcome Metrics

While incorporating additional clinical indicators would more accurately reflect the prognostic impact of CD8⁺ T_RM cells. However, the current cohort lacks sufficient information to support this, and we will address this limitation by refining our study design in future research cohorts.

11. Treatment-naïve status

Clinical cohorts included in this study consist of primary, treatment-naive patients, as detailed in Supplementary Material 1

Minor Comments

1. Redundancy (Original lines 235–241, revised line 235-237): has been revised and polished..
2. The datasets used at the beginning of each section have been explicitly stated.
3. Among the patients, 15 indeed reached Stage IV, which may be attributed to the challenge of early detection and diagnosis of iCCA; these patients presented at advanced stages upon hospital admission.
4. Total number of cells in the scRNA-seq dataset (n = 343,170) added to Methods.
5. The tissue microarray (TMA) and single-cell data belong to two independent cohorts, as clarified in the Methods section.
6. Incomplete Sentence (Original lines 382, revised line387): has been rewritten.
7. The entire manuscript has been reviewed, and instances of "predictive" have been corrected to "prognostic" where appropriate.
8. The current marker panel comprehensively covers tumor cells and CD8⁺T cell effector subsets, effectively reflecting the spatial characteristics of the tumor microenvironment. Its representativeness has been elaborated in Results 3.1.
9. Annotations for sample size and cell count have been added to each figure.
10. The notation of PD-1⁺T_RM in Fig5 was indeed a typographical error, which has now been corrected to PD-1^- T_RM.

We deeply appreciate the reviewer’s rigorous feedback, which has significantly strengthened the manuscript‘s clarity, accuracy, and scientific rigor. All revisions are marked in the revised manuscript.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript has undergone comprehensive revisions, significantly enhancing its clarity and formatting. Key improvements include updates to the Simple Summary and Materials and Methods sections, as highlighted in the comments, as well as the Title, Introduction, Results, Discussion, and Conclusions sections. These changes have enhanced the overall quality of the manuscript.

The Discussion section has been expanded to provide a deeper analysis, with four added references that enhance the context and reinforce the findings. All corrections have been implemented in the revised manuscript (lines 394, 470-471, 487, 504-505, and Figure 4D).

Additionally, p-values in several figures (2B, 4B, 4D, 6D, and 6E) have been revised, with corresponding updated figure legends and text to ensure consistency. A typographical error on line 340 ("confered") has also been identified and needs to be corrected to "conferred."

Overall, these revisions have substantially improved the manuscript, ensuring greater clarity and accuracy.

Author Response

Review 2#

The manuscript has undergone comprehensive revisions, significantly enhancing its clarity and formatting. Key improvements include updates to the Simple Summary and Materials and Methods sections, as highlighted in the comments, as well as the Title, Introduction, Results, Discussion, and Conclusions sections. These changes have enhanced the overall quality of the manuscript.

The Discussion section has been expanded to provide a deeper analysis, with four added references that enhance the context and reinforce the findings. All corrections have been implemented in the revised manuscript (lines 394, 470-471, 487, 504-505, and Figure 4D).

Additionally, p-values in several figures (2B, 4B, 4D, 6D, and 6E) have been revised, with corresponding updated figure legends and text to ensure consistency. A typographical error on line 340 ("confered") has also been identified and needs to be corrected to "conferred."

Overall, these revisions have substantially improved the manuscript, ensuring greater clarity and accuracy.

Response to Reviewer 2

Thank you for your positive assessment of our revised manuscript and for acknowledging the improvements made in the first-round revision, including section-wide updates, expansion of the Discussion with additional references, and the revision of statistics with corresponding updates to the figure legends and text.

In this second-round revision, we have addressed the typographical issue: “confered” has been corrected to “conferred” (line 340).

We appreciate your careful check in this round, and we have conducted another thorough review of the manuscript to further ensure clarity, consistency, and overall quality.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

Authors have done a thorough job in correcting the manuscript

Author Response

Review 4#

Authors have done a thorough job in correcting the manuscript

Response to Reviewer 4

Thank you for your positive comment and the recognition that we have addressed the concerns raised in the previous round and improved the manuscript accordingly. In this second-round revision, we have also carefully proofread the entire manuscript, implementing several minor typographical refinements, to further ensure clarity, consistency, and overall quality.

Author Response File: Author Response.pdf