Reshaping the Battlefield: Reprogramming the Melanoma Tumour Microenvironment (TME) by Anti-CTLA-4, Anti-PD-1, and Anti-PD-L1 Monotherapy and Combination Therapy: A Systematic Review and Meta-Analysis of Preclinical and Clinical Evidence

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents a methodologically rigorous dual systematic review and meta-analysis of tumour microenvironment reprogramming by immune checkpoint blockade in melanoma, with prospective PROSPERO registration (CRD420261374242) and full adherence to PRISMA 2020 guidelines. The use of Hedges' g standardized mean difference within a random-effects framework is well justified given the profound measurement heterogeneity described in sections 2.4.1 and 2.4.2. The principal finding of conserved CD8⁺ T-cell infiltration across preclinical and clinical arms (SMD=1.45 and SMD=0.72 respectively) is biologically compelling, and the divergence in PD-L1 dynamics between species, decreased preclinically but increased clinically, is discussed with appropriate mechanistic nuance in section 4.2. The meta-regression identifying superior effector amplification by anti-PD-L1-containing regimens for IFN-γ, CD8/Treg ratio, and apoptosis adds meaningful translational value. Nonetheless, the substantial heterogeneity across outcomes (I²=68–88%) and the limited number of studies contributing to certain pooled estimates, notably four data points for preclinical PD-L1 and a single study for clinical apoptosis, constrain interpretive confidence. Sensitivity analyses partially address this concern. Reporting deficiencies identified through SYRCLE and ARRIVE 2.0 assessments are transparently acknowledged. 

Author Response

We are sincerely grateful to Reviewer 1 for their careful, constructive, and generous assessment of our manuscript. It is deeply rewarding to see our methodological choices, including the prospective PROSPERO registration, the strict adherence to PRISMA 2020, and the use of Hedges' g within a random-effects framework, recognised as appropriate and well justified. We also thank the reviewer for highlighting the conserved CD8⁺ T-cell infiltration, the divergent PD‑L1 dynamics, and the translational value of the anti‑PD‑L1 meta-regression findings as strengths of the work.

The reviewer rightly notes that the substantial heterogeneity (I² = 68–88%) and the very limited number of data points for certain outcomes, specifically the four data points for preclinical PD‑L1 and the single study reporting clinical apoptosis, constrain interpretive confidence. We fully agree with this assessment, and we appreciate the acknowledgement that our sensitivity analyses partially address this concern.

In response, we have added three clarifications to the revised manuscript:

• In Section 4.2 (PD‑L1 Expression; a Critical Divergence), we now explicitly state that the preclinical PD‑L1 decrease, while remarkably homogeneous across the available studies (I² = 0%), should be interpreted as a preliminary signal requiring replication, given the small number of underlying data points.
• In Section 4.2 (Apoptosis and Proliferation), we now underscore that the single clinical apoptosis data point precludes any meta-analytic inference and highlights an important gap in the clinical biomarker literature.
• In Section 4.3 (Interpretation of Heterogeneity), we acknowledge that for outcomes with very few studies, the pooled estimates should be regarded as provisional until further data become available. All additions are highlighted in green.

We also thank the reviewer for recognising the transparency with which we reported the SYRCLE and ARRIVE 2.0 assessments. We believe that honest reporting of methodological limitations is essential, and we are pleased that this was noted positively. Finally, we are grateful that the reviewer rated our research design, results presentation, and conclusions favourably, and that their suggestions regarding the introduction, methods description, and figures/tables were offered in the spirit of improvement. We have carefully addressed these points in the revised manuscript.

We thank Reviewer 1 again for their time and their invaluable contribution to strengthening our manuscript.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript by Karakousis et al. is a comprehensive evaluation of preclinical data in a B16F10 model and clinical data about TME remodeling by immune checkpoint inhibitors. The authors have assembled a large body of literature and the dual preclinical/clinical structure is very ambitious. The focus on TME remodeling rather than clinical response alone is potentially useful, particularly as the field moves toward more rational immunotherapy combinations.

That said, the manuscript is very broad and, in several places, the conclusions are stronger than the data appear to support. The manuscript is also extremely long (63 pages!) and may be well beyond the limits for this article type. The article would benefit greatly from a more focused presentation. I would strongly recommend breaking this manuscript into two separate papers, one on B16F10 TME remodeling and another clinical studies because the message is diluted somewhat by the length and lack of clear generalizability between the two sections. The paper would also benefit from clearer framing of the preclinical B16 model earlier to avoid over interpretation of this as a review of all preclinical models. And more cautious interpretation of the anti-PD-L1 findings.

Here are my major comments:

1. The statement that “anti-PD-L1-containing regimens uniquely amplify terminal effector functions” and reveal “superior pharmacological potency” is too strong as written. This may be true within the specific preclinical comparisons included in the analysis, but this has not clearly translated into superior clinical outcomes for anti-PD-L1 therapy in melanoma. The authors should specify that this finding is derived from the preclinical B16F10 literature and should avoid implying that anti-PD-L1 blockade is clinically superior to anti-PD-1 or anti-CTLA-4 therapy in melanoma. This issue is particularly important because the clinical quantitative arm did not include anti-PD-L1 monotherapy or anti-PD-L1-containing combinations. The abstract, highlights, discussion, and conclusion should all be revised to make this distinction clear.
2. Large portions of Section 1.1 could be removed or substantially shortened. The discussion of melanoma etymology, ABCDE criteria, dermoscopy, diagnostic immunohistochemistry, detailed AJCC staging, and broad clinicopathologic classification is outside the main scope of the manuscript. These topics are important clinically, but they distract from the central question of how CTLA-4, PD-1, and PD-L1 blockade alter the melanoma TME. I would recommend opening the introduction with the clinical importance of immune checkpoint blockade in melanoma, the limitations of response and resistance, and the rationale for studying TME remodeling. The purpose of the study is not reached until Section 1.4, and it should appear much earlier.
3. Within the introduction, the current phrasing implies a relatively binary distinction between low-CSD and high-CSD melanomas, with low-CSD tumors characterized by BRAF V600E and high-CSD tumors characterized by NRAS, NF1, or non-V600 BRAF mutations. This should be revised. The WHO pathway-based classification includes multiple melanoma pathways beyond low-CSD and high-CSD melanoma, including desmoplastic melanoma, Spitz melanoma, acral melanoma, mucosal melanoma, melanoma arising in congenital or blue nevi, and uveal melanoma. BRAF V600E is enriched in low-CSD melanoma but is not exclusive to that group; NRAS is also common in low-CSD melanoma. High-CSD melanoma is enriched for NF1, non-V600 BRAF alterations, KIT alterations, and other secondary genomic events. The current wording risks implying a cleaner molecular dichotomy than exists biologically or clinically. This section could be removed to increase focus within the paper on the TME.
4. The decision to focus on the B16F10/C57BL/6 model is reasonable and may improve internal consistency of the preclinical analysis. However, this choice also substantially limits generalizability. I agree with the decision to focus on this but earlier acknowledgement is needed. Throughout the manuscript, statements about “preclinical melanoma” or “preclinical models” should be tempered to specify the model system being analyzed.
5. The preclinical arm evaluates treatment-induced changes in a controlled model system, whereas the clinical arm includes heterogeneous patient cohorts and different types of comparisons. These are complementary but not equivalent. The discussion should avoid treating the two arms as if they provide the same level of mechanistic inference. A clearer separation between “treatment effect in B16F10 models” and “clinical association in human melanoma cohorts” would improve the interpretation.

Minor Comments:

The titles are engaging but perhaps a bit too dramatic such as “unleashing the storm” etc.

Figures 12 and 13 are difficult to read because the text is small and overlapping. The authors could improve legibility by replacing text labels with numbered callouts and a separate legend, or by using leader lines from labels to specific data points.

The manuscript contains too many figures. Several figures appear to restate related concepts and could likely be combined, simplified, or moved to supplementary material.

The manuscript appears to be an enormous amount of work but is too ambitious for one manuscript and as a result is too long and difficult to read. I recommend major revision and deciding which aspect of the manuscript to focus on, with having this broken into two separate manuscripts and removal of the textbook style review at the opening of the paper.

Author Response

Response to Reviewer 2:

We are deeply grateful to Reviewer 2 for the time, expertise, and meticulous care they invested in evaluating our manuscript. The detailed comments have substantially improved the rigour and clarity of our work, and we address each point below. All changes are highlighted within the resubmitted manuscript in red.

Comment 1: This manuscript by Karakousis et al. is a comprehensive evaluation of preclinical data in a B16F10 model and clinical data about TME remodeling by immune checkpoint inhibitors. The authors have assembled a large body of literature and the dual preclinical/clinical structure is very ambitious. The focus on TME remodeling rather than clinical response alone is potentially useful, particularly as the field moves toward more rational immunotherapy combinations. The manuscript appears to be an enormous amount of work.

Response 1: We sincerely thank the reviewer for these generous remarks. We are grateful for the recognition of the manuscript as a comprehensive synthesis, for the appreciation of the large body of assembled evidence, and for the characterisation of the dual-arm framework as ambitious. The reviewer's appraisal that our focus on TME remodelling reflects an important direction for the field is particularly encouraging, and we deeply appreciate the acknowledgement of the extensive work involved. These positive observations have been a powerful motivation during the revision process, and we have endeavoured to address every subsequent concern with the same care that the reviewer brought to their reading.

Comment 2: The titles are engaging but perhaps a bit too dramatic such as “unleashing the storm” etc.

Response 2: We fully agree with the reviewer that certain original titles employed a more poetic style than is conventional for scientific reporting, and we appreciate this constructive observation. In the revised manuscript, we have replaced the three affected titles with descriptive alternatives:

• Section 1.1: The title “From Melanocyte to Malignancy: The Rising Burden and Biology of Melanoma”has been changed to “From Melanocyte to Malignancy: Epidemiology and Molecular Pathogenesis of Melanoma”.
• Section 1.2: The title “Releasing the Brakes, Unleashing the Storm: Molecular Cascades of CTLA -4, PD -1, and PD- L1 Blockade”has been changed to “Molecular Mechanisms of CTLA‑4, PD‑1, and PD‑L1 Immune Checkpoint Blockade”.
• Section 1.4: The title “Mapping the Immune Battlefield: A Dual Systematic and Meta analytic Cartography of TME Reprogramming”has been changed to “A Dual Systematic and Meta-Analytic Cartography of TME Reprogramming”.

We thank the reviewer for prompting us to adopt a more restrained tone throughout the Introduction.

Comment 3: Large portions of Section 1.1 could be removed or substantially shortened. The discussion of melanoma etymology, ABCDE criteria, dermoscopy, diagnostic immunohistochemistry, detailed AJCC staging, and broad clinicopathologic classification is outside the main scope of the manuscript. These topics are important clinically, but they distract from the central question of how CTLA-4, PD-1, and PD-L1 blockade alter the melanoma TME. I would recommend opening the introduction with the clinical importance of immune checkpoint blockade in melanoma, the limitations of response and resistance, and the rationale for studying TME remodeling. The purpose of the study is not reached until Section 1.4, and it should appear much earlier. [I recommend] removal of the textbook style review at the opening of the paper.

Response 3: We are grateful to the reviewer for this constructive guidance, which has prompted a meaningful improvement in the manuscript’s structure and readability. We have substantially condensed Section 1.1 by approximately 50%, retaining a concise summary of the essential epidemiological, pathogenetic, diagnostic, and staging concepts that are necessary to convey the clinical significance of melanoma and to ground the subsequent discussion of TME-directed immunotherapy in the reality of clinical practice. The shortened text now offers a much quicker transition to the clinical importance of immune checkpoint blockade, the limitations of response and resistance, and the unmet need that motivated this synthesis. We also fully agree with the reviewer’s observation that the study rationale was presented too late in the original manuscript. In response, we have reordered the Introduction so that the study rationale, objectives, hypotheses, and PROSPERO registration details (originally Section 1.4) now appear immediately after the condensed clinical background, as the new Section 1.2. This, places the purpose of the review much earlier in the flow, exactly as the reviewer suggested.

Comment 4: Within the introduction, the current phrasing implies a relatively binary distinction between low-CSD and high-CSD melanomas, with low-CSD tumors characterized by BRAF V600E and high-CSD tumors characterized by NRAS, NF1, or non-V600 BRAF mutations. This should be revised. The WHO pathway-based classification includes multiple melanoma pathways beyond low-CSD and high-CSD melanoma, including desmoplastic melanoma, Spitz melanoma, acral melanoma, mucosal melanoma, melanoma arising in congenital or blue nevi, and uveal melanoma. BRAF V600E is enriched in low-CSD melanoma but is not exclusive to that group; NRAS is also common in low-CSD melanoma. High-CSD melanoma is enriched for NF1, non-V600 BRAF alterations, KIT alterations, and other secondary genomic events. The current wording risks implying a cleaner molecular dichotomy than exists biologically or clinically. This section could be removed to increase focus within the paper on the TME.

Response 4: We fully agree with the reviewer’s observation that the original wording risked implying a cleaner molecular dichotomy than exists biologically or clinically, and that this discussion diverted focus from the central TME question. We have therefore entirely removed the CSD classification section from the Introduction, as suggested.

Comment 5: The paper would also benefit from clearer framing of the preclinical B16 model earlier to avoid over interpretation of this as a review of all preclinical models. The decision to focus on the B16F10/C57BL/6 model is reasonable and may improve internal consistency of the preclinical analysis. However, this choice also substantially limits generalizability. I agree with the decision to focus on this but earlier acknowledgement is needed. Throughout the manuscript, statements about “preclinical melanoma” or “preclinical models” should be tempered to specify the model system being analyzed.

Response 5: We thank the reviewer for this thoughtful and constructive observation. We wish to respectfully note that the stringent restriction to the B16F10/C57BL/6 model is clearly stated from the very outset of the manuscript; the Abstract explicitly specifies that the preclinical arm comprises "58 preclinical (B16F10/C57BL/6; 46 quantitative) studies." Furthermore, an extensive, dedicated justification of our stringent preclinical model standardisation and inclusion criteria is provided early in the Materials and Methods (Section 2.1.1.1, "Rationale for Stringent Preclinical Model Standardization and Inclusion Criteria"); precisely to make it clear that the preclinical population under study is a highly specific one. We are grateful for the reviewer's acknowledgement that this decision is reasonable and may improve the internal consistency of the preclinical analysis.

The limitation of generalisability is also explicitly and transparently acknowledged in the Limitations section (Section 4.6).  In direct response to the reviewer’s comment, we have added a clarifying sentence to the Limitations section to further underscore this point. The full passage now reads: "First, the preclinical synthesis was deliberately restricted to the unmodified B16 F10 melanoma model in immunocompetent C57BL/6 mice. While this stringent standardization enhanced internal validity and reduced experimental variance, it necessarily limits the generalizability of the findings to other syngeneic models (e.g., YUMM, RET) and to the broader landscape of melanoma heterogeneity. Consequently, the preclinical findings should be interpreted as specific to this model system and extrapolated to other contexts with due circumspection."

We fully agree with the reviewer's recommendation regarding terminology. For this exact reason, throughout the revised manuscript, we have carefully tempered statements about "preclinical melanoma" or "preclinical models" to specify "in the B16F10 preclinical melanoma model" or "in B16F10 preclinical models". The specific changes made are as follows:

• Section 3.6.1.1 (Spatial and Stromal Architecture): "the preclinical corpus"→ "the B16F10 preclinical corpus"
• Section 3.6.1.8 (Systemic/Peripheral Biomarkers): "immunocompetent murine models"→ "the immunocompetent B16F10 murine model"
• Section 4.2 (CD8⁺ T‑cell Infiltration): "the preclinical corpus"→ "the B16F10 preclinical corpus"
• Section 4.2 (Apoptosis and Proliferation): "the preclinical corpus"→ "the B16F10 preclinical corpus"
• Section 4.2 (Spatial and Stromal Architecture): "the preclinical corpus"→ "the B16F10 preclinical literature"
• Section 4.2 (Alternative Checkpoint Upregulation): "Preclinically"→ "In the B16F10 model"
• Section 4.4 (Pharmacological Implications): "superior preclinical potency"→ "superior potency of PD‑L1 blockade in the B16F10 model"
• Section 5 (Conclusion): "The preclinical corpus"→ "The B16F10 preclinical evidence base"

We trust that these systematic revisions fully address the reviewer's concern and ensure that the reader is consistently reminded of the specific model system being analysed.

Comment 6: In several places, the conclusions are stronger than the data appear to support. And more cautious interpretation of the anti-PD-L1 findings. The statement that “anti-PD-L1-containing regimens uniquely amplify terminal effector functions” and reveal “superior pharmacological potency” is too strong as written. This may be true within the specific preclinical comparisons included in the analysis, but this has not clearly translated into superior clinical outcomes for anti-PD-L1 therapy in melanoma. The authors should specify that this finding is derived from the preclinical B16F10 literature and should avoid implying that anti-PD-L1 blockade is clinically superior to anti-PD-1 or anti-CTLA-4 therapy in melanoma. This issue is particularly important because the clinical quantitative arm did not include anti-PD-L1 monotherapy or anti-PD-L1-containing combinations. The abstract, highlights, discussion, and conclusion should all be revised to make this distinction clear.

Response 6: We are grateful to the reviewer for this critical observation. We fully agree that the original wording risked overinterpreting the preclinical meta-regression findings and could inadvertently imply a clinical superiority that is not supported by the evidence, particularly given the absence of anti‑PD‑L1‑containing regimens in the clinical quantitative arm. In response, we have carefully tempered the relevant statements across all four sections identified by the reviewer, anchoring the findings explicitly to the B16F10 model and removing any language that could be construed as a claim of clinical superiority. The specific revisions are as follows:

• Highlights (second bullet): The original claim of "superior pharmacological potency" has been removed. The revised bullet now reads:
  "In the preclinical B16F10 literature, anti‑PD‑L1‑containing regimens amplified terminal effector functions to a greater degree than anti‑CTLA‑4 monotherapy, producing larger effect sizes for IFN‑γ, the CD8/Treg ratio, and apoptosis, a differential not observed with anti‑PD‑1 or anti‑CTLA‑4 alone."
• Abstract (final sentence): The original phrasing "define the superior effector potency of PD‑L1 blockade" has been replaced. The revised sentence now reads:

"These findings establish the TME as a critical determinant of ICI outcomes, indicate that PD‑L1 amplifies effector functions in B16F10 model, and highlight translational gaps in TME reprogramming."

• Discussion (Section 4.4): The sentence describing the largest effect sizes has been revised to specify the preclinical context:
  "Notably, in the preclinical B16F10 analysis, anti‑PD‑L1‑containing regimens consistently produced the largest effect sizes, both as monotherapy and in combination with anti‑CTLA‑4."
• Conclusion: A new sentence has been added to explicitly state the preclinical origin of the finding and its untested status in the clinical arm:

"The superior effector amplification by anti‑PD‑L1‑containing regimens was observed exclusively in the preclinical B16 F10 model and has not been evaluated in the clinical quantitative arm; its translational applicability therefore awaits prospective validation."

We trust that these revisions, which carefully circumscribe the anti‑PD‑L1 findings to the preclinical domain, fully address the reviewer's concern and ensure that the manuscript does not imply clinical superiority where none has been demonstrated.

Comment 7: The preclinical arm evaluates treatment-induced changes in a controlled model system, whereas the clinical arm includes heterogeneous patient cohorts and different types of comparisons. These are complementary but not equivalent. The discussion should avoid treating the two arms as if they provide the same level of mechanistic inference. A clearer separation between “treatment effect in B16F10 models” and “clinical association in human melanoma cohorts” would improve the interpretation.

Response 7: We completely agree with the reviewer that the preclinical and clinical arms provide complementary but not equivalent levels of mechanistic inference. This distinction is inherent in the study design and is explicitly described in the Methods, where the clinical comparator categories are collectively considered under the umbrella term “favorable vs non-favorable outcome” (Section 2.7.2.1, Comparators and Subgroup Analysis), making it clear that the clinical analysis captures associations rather than controlled treatment effects. In direct response to this comment, we have added a clarifying sentence at the very beginning of Section 4.2 to ensure that the reader is explicitly reminded of this distinction throughout the comparative Discussion. The added sentence reads:

“Throughout this comparison, it is important to distinguish between the treatment effects observed in the controlled B16F10 preclinical model and the clinical associations derived from heterogeneous patient cohorts; the two arms provide complementary but not equivalent levels of mechanistic inference.”

We trust that this addition fully addresses the reviewer’s concern.

Comment 8: The manuscript contains too many figures. Several figures appear to restate related concepts and could likely be combined, simplified, or moved to supplementary material. Figures 12 and 13 are difficult to read because the text is small and overlapping. The authors could improve legibility by replacing text labels with numbered callouts and a separate legend, or by using leader lines from labels to specific data points.

Response 8: We thank the reviewer for this practical suggestion. The manuscript does contain a substantial number of figures, but each serves a distinct and necessary purpose in telling the full story of this dual systematic review and meta-analysis: the Graphical Abstract (recommended by the journal), the mechanistic scheme of the checkpoint inhibitors (Figure 1, condensing complex molecular pathways), the PRISMA flow diagrams (Figures 2 and 3, required for transparent reporting), the risk-of-bias traffic light and summary plots (Figures 4–7, necessary to visualise the SYRCLE, ARRIVE 2.0, RoB 2, and ROBINS‑I assessments), the forest, bubble, and funnel plots (Figures 8–13, the standard graphical output of the meta-analytic workflow), and the recapitulating schematic of the main findings (Figure 14). In response to the reviewer’s concern, however, we have taken two concrete steps to reduce visual load and improve legibility. First, we have moved the bubble plots for both preclinical and clinical meta-regression (originally Figures 10 and 11) to the Supplementary Materials. Second, for the funnel plots (Figures 12 and 13), which by their nature contain many overlapping study names and years as directly generated by the statistical software, we have retained data point labels only for the outlying studies, those falling outside the pseudo-confidence contours (the “Outside H₀” option within JASP), and removed the labels from the remaining points. We note that in some funnels, a minor degree of label overlap persists among the outliers due to the inherent density of points; however, the plots are now substantially more legible than in the original submission, and the visual assessment of asymmetry is fully preserved.  We trust that these measures address the reviewer’s concern without compromising the completeness of the reported evidence.

Comment 9: That said, the manuscript is very broad. The manuscript is also extremely long (63 pages!) and may be well beyond the limits for this article type. The article would benefit greatly from a more focused presentation. I would strongly recommend breaking this manuscript into two separate papers, one on B16F10 TME remodeling and another clinical studies because the message is diluted somewhat by the length and lack of clear generalizability between the two sections. [The work] is too ambitious for one manuscript and as a result is too long and difficult to read. I recommend major revision and deciding which aspect of the manuscript to focus on, with having this broken into two separate manuscripts.

Response 9: We sincerely thank the reviewer for this constructive assessment of the manuscript's scope and structure. We have reflected on this recommendation with great care, and we respectfully but firmly believe that the unified, dual-arm format is not an overambitious flaw but rather the defining methodological and conceptual innovation of this work. We wish to explain our reasoning in detail, drawing on both regulatory considerations and scientific rationale.

First, we wish to clarify the nature of the work under review and the composition of the page count to which the reviewer refers. This is not a single systematic review or meta-analysis; it is a dual systematic review and meta-analysis (SR/MA); two complete, prospectively registered syntheses conducted in parallel and then compared within a unified framework. The 63 pages include approximately 10 pages of abbreviations and references and 12 pages occupied by the figures that are essential for the manuscript’s narrative (PRISMA flow diagrams, forest plots, funnel plots, risk-of-bias visualizations, and the mechanistic and recapitulating figures). The main text itself therefore comprises roughly 41 pages, which is consistent with the scope of synthesising 102 studies across two species with independent search strategies, risk-of-bias assessments, meta-analyses, sensitivity analyses, meta-regressions, GRADE assessments, and a comparative narrative synthesis for each arm. The manuscript’s length is a direct consequence of its registered scope: two complete SR/MAs, each of which individually would justify a full-length publication, presented together to enable the direct comparison that is the study’s central objective.

Second, the dual-arm systematic review was prospectively registered as a single, unified protocol on PROSPERO (CRD420261374242). The registered protocol specifies a single review encompassing both preclinical and clinical evidence, with a comparative synthesis as the primary objective. Splitting the manuscript into two separate publications after the results are known would constitute a deviation from the registered protocol and would raise a significant methodological concern: it would create two publications from one registration, with the second publication lacking any prospective registration. Such an act could be viewed as salami-slicing of a prospectively registered systematic review, which is strongly discouraged by both PRISMA 2020 and Cochrane guidance. The integrity of our PROSPERO registration compels us to maintain the unified structure.

Third, we respectfully note that neither the journal's instructions nor the major international reporting guidelines that explicitly govern SR/MAs impose a page or word limit that this manuscript exceeds. Cells MDPI explicitly states that "Cells has no restrictions on the length of research manuscripts, provided that the text is concise and comprehensive." The PRISMA 2020 statement, the internationally recognized standard for systematic review reporting, imposes no page limit; it only observes that individual journals may set limits, and Cells has chosen not to do so. The Cochrane Handbook for Systematic Reviews of Interventions, which provides guidance for the conduct and reporting of SR/MAs, states that "there is no formal word limit for Cochrane reviews" and that the suggested guideline of 10,000 words may be exceeded "when the question is unusually broad or complex." A dual SR/MA spanning preclinical and clinical evidence, with two independent meta-analyses, is, by any reasonable standard, a broad and complex question, and its length is a direct reflection of its registered scope rather than a lack of focus. Our manuscript, at approximately 20,000 words of main text excluding references, reflects the fact that it contains two complete systematic reviews, a preclinical meta-analysis of 58 studies and a clinical meta-analysis of 44 studies, each with independent searches, risk-of-bias assessments, meta-analyses, sensitivity analyses, meta-regressions, and GRADE assessments. We therefore believe the manuscript length is not a sign of inadequate focus but a direct consequence of the review's registered scope.

Fourth, we note that neither of the other two reviewers raised any concern about the manuscript's length or its unified structure. We therefore believe that the consensus of the other two expert reviewers, who evaluated the same manuscript and did not identify length or focus as a problem, supports the unified presentation.

Fifth, and most importantly, the dual-arm, directly comparative framework is not merely a formatting choice, it is the intellectual core of the study. Our central research question was whether preclinical TME reprogramming translates to clinical observations, and whether the two species show conserved or divergent patterns of response. This question can only be answered by placing both arms in direct, quantitative comparison within a single analytical framework. Separating them into two manuscripts would destroy the very comparison the PROSPERO protocol was designed to perform. The complementary nature of the two arms is evident not only in the quantitative meta-analysis, where the strengths and limitations of each dataset become directly comparable, but also in the comparative narrative synthesis, which systematically juxtaposes preclinical and clinical observations for each TME parameter, thereby highlighting species-specific divergences and pinpointing the most pressing gaps that future research should address. The preclinical arm alone would lose its translational anchor; the clinical arm alone would lose its mechanistic depth. The whole is greater than the sum of its parts, and the whole is what we registered, designed, and executed. The dual-arm format is precisely what distinguishes this synthesis from existing reviews and provides its unique value to the field.

We have, however, taken the reviewer's concerns about focus and readability very seriously. As detailed in our responses to Comments 1‑8, we have substantially shortened Section 1.1, moved the study rationale earlier in the Introduction, tempered the anti‑PD‑L1 language throughout, added earlier caveats about the B16F10 model, clarified the distinction between preclinical treatment effects and clinical associations, moved the bubble plots to the Supplementary Materials, and improved the readability of the funnel plots. We believe these revisions collectively address the reviewer's concern about focus without sacrificing the structural integrity that makes this work distinctive.

We hope the reviewer will understand that our decision to preserve the unified format is grounded not in obstinacy but in fidelity to the registered protocol, compliance with journal and international reporting standards, and a sincere conviction that the dual-arm comparison is the defining contribution of this review. We are deeply grateful for the reviewer's time and insights, which have substantially improved the manuscript.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

It was an honor to review this paper on tumor environment and changes caused by the therapy with checkpoint inhibitors. I believe it is an important topic and the readers will find it interesting and educational. It is well written. The section on the methods of this analysis is comprehensive and appropriate. The results are exciting. The references are appropriate. I do not have any concerns or comments about this research

Author Response

We are sincerely grateful to Reviewer 3 for their generous and encouraging words, and for the time they invested in reviewing our manuscript. We are delighted that the reviewer found the topic important, interesting, and educational, and we deeply appreciate the kind recognition of the manuscript's clarity of expression. We thank the reviewer for their positive appraisal of the Methods section, which we designed with rigour as the guiding principle, and we are truly gratified that our results were described as "exciting." We also appreciate the reviewer's confirmation of the appropriateness of our references. To receive this review is an honour, and we thank Reviewer 3 once again for their time, expertise, and generous endorsement.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have adequately addressed my concerns. The revisions are appropriate and while we disagree on the dual arm approach of the paper, their argument is valid. I commend the authors for their efforts and recommend accepting the manuscript.