Tumor-Derived LIF Promotes GDF15-Driven Cachexia and Adverse Outcomes in Gastric Cancer

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The current clinical study reports on cytokines and metabolites involvement in mediating gastric cancer-induced cachexia and impaired survival. The study is well designed and reports very relevant data for the field.

I have some suggestions in order to improve the quality of the manuscript and to deliver a clearer message.

Major comments:

The main point to address, in my opinion, is the associations vs causative relationship between GDF15 and LIF. Who does come first? The expression of LIF induced by GDF15 was previously reported in glioma cells (PMID: 33431816) and here mentioned, thus not allowing to exclude such dependency even in gastric cancer. The linear regression regarding the correlation among the two factors is not very strong, reducing the robustness of the overall conclusion. A Kaplan-Meier analysis with LIF would be useful to be compared to GDF15, also in light of the inverse association between LIF and low muscularity (Figure 5D). My personal suggestion is to down-tone the statements towards the association, not the causation.

The stratification among high vs low GDF-15 levels should define the threshold and whether this was previously set by others or in the current study. Alongside, the clinical data with the general patient characteristics including demographics should be included in an initial table.

According to Figure 2D and 2H, patients with high GDF-15 survive longer, while the message conceived is the opposite.

Table C in Figure 5: please better clarify why only 19 patients are reported. The individual data are hard to decipher without statistical elaboration and stratification (i.e. according to body weight loss). Please also report in the abstract that only a limited sub-cohort of patients underwent cachexia-related assessments.

The images reported in Figure 6 only minimally support the authors conclusion. The immunofluorescence data (negative controls are missing) should be validated via western blotting, along with a study of cell proliferation / growth upon LIF stimulation. Regarding Figure 7, in case the LIF-stimulated cells grow more, it is likely that the conditioned media will be conditioned by an higher number of cells, posing doubts on the presence of specific mediators rather then being affected by just an higher number of cells during the 3-4 day culture before conditioned media collection. Moreover, alpha-smooth muscle actin should not be expressed by terminally differentiated myotubes…

Minor comments:

• Avoid the use of acronyms in the abstract (e.g. GC for gastric cancer).
• Please thoroughly revise the manuscript for typos or grammatical errors (e.g. line 59, 75)
• Figure 1, in panel C the x axis legend is cut.
• TGCA or TCGA repository? Also ATGC sounds new in figure 2? Please revise.
• Please change surnatant to supernatant.

Author Response

Comments and Suggestions for Authors

The current clinical study reports on cytokines and metabolites involvement in mediating gastric cancer-induced cachexia and impaired survival. The study is well designed and reports very relevant data for the field.

I have some suggestions in order to improve the quality of the manuscript and to deliver a clearer message.

 

Major comments:

Q1: The main point to address, in my opinion, is the associations vs causative relationship between GDF15 and LIF. Who does come first? The expression of LIF induced by GDF15 was previously reported in glioma cells (PMID: 33431816) and here mentioned, thus not allowing to exclude such dependency even in gastric cancer. The linear regression regarding the correlation among the two factors is not very strong, reducing the robustness of the overall conclusion. A Kaplan-Meier analysis with LIF would be useful to be compared to GDF15, also in light of the inverse association between LIF and low muscularity (Figure 5D). My personal suggestion is to down-tone the statements towards the association, not the causation.

R1: We thank the reviewer for this comment. We agree that our data do not allow us to establish a causal relationship between LIF and GDF15 expression. While previous studies have reported GDF15-induced LIF expression in other tumor contexts, such as glioma, our analyses in gastric cancer demonstrate a coordinated upregulation and positive association between the two factors, without defining directionality.

Consistent with this, the correlation between LIF and GDF15, although statistically significant, does not support a strong linear dependency. Kaplan-Meier analyses indicate that GDF15 is robustly associated with overall survival. However, higher LIF expression showed a non-significant trend toward poorer overall survival, as illustrated in Supplementary Figure 1.

 To further explore the functional relationship between LIF signaling and GDF15 expression, we performed additional experimental analyses. LIF stimulation in gastric cancer spheroids induced the transcriptional upregulation of GDF15 together with proliferation- and EMT-associated markers, while pharmacological inhibition of LIFR attenuated these effects (Figure 7A and Supplementary Figure 4). These data support the existence of a functional link between LIF signaling and GDF15 expression in this experimental setting, while not excluding alternative regulatory pathways.

Accordingly, we have revised the manuscript to down-tone causal interpretations and to consistently frame the LIF–GDF15 relationship as associative rather than causative, while highlighting potentially distinct but converging roles of LIF and GDF15 in the cachexia phenotype.

Q2: The stratification among high vs low GDF-15 levels should define the threshold and whether this was previously set by others or in the current study. Alongside, the clinical data with the general patient characteristics including demographics should be included in an initial table.

R2: The threshold used to stratify patients into high and low GDF15 expression groups has now been explicitly defined in the Methods section, specifying that it was determined within the current study using the auto-selected best cutoff approach. In addition, we have included a new table summarizing the general clinical and demographic characteristics of the patient cohort.

Q3. According to Figure 2D and 2H, patients with high GDF-15 survive longer, while the message conceived is the opposite.

R3. The reviewer is correct; the survival analyses performed on external repositories do not reproduce the prognostic association observed in our internal cohort.

Q4. Table C in Figure 5: please better clarify why only 19 patients are reported. The individual data are hard to decipher without statistical elaboration and stratification (i.e. according to body weight loss). Please also report in the abstract that only a limited sub-cohort of patients underwent cachexia-related assessments.

R4. Cachexia-related analyses were performed in a limited sub-cohort of 19 patients for whom complete CT-based body composition, nutritional, and laboratory data were available. To improve data interpretability, this sub-cohort was stratified according to serum prealbumin levels (<20 vs ≥20 mg/dL), a clinically validated threshold for nutritional risk, as previously reported (doi:10.1007/s10120-024-01472-y). The Results section has been revised accordingly. In addition, the abstract has been updated to explicitly state that cachexia-related assessments were conducted in a limited sub-cohort of patients.

Q5. The images reported in Figure 6 only minimally support the authors conclusion. The immunofluorescence data (negative controls are missing) should be validated via western blotting, along with a study of cell proliferation / growth upon LIF stimulation.

R5. Negative controls for the immunofluorescence experiments are provided in Supplementary Figure 2. In addition, Supplementary Figure 3 includes western blot analyses confirming basal expression of LIF and LIFR in gastric cancer spheroids, together with complementary proliferation assays demonstrating that LIF stimulation enhances cellular proliferation, as assessed by IC-FACS analysis. These findings support the conclusions drawn from Figure 6 by validating the immunofluorescence findings and are consistent with and extend our previous work showing that LIF/LIFR signaling promotes gastric cancer cell proliferation and tumor progression (Di Giorgio et al., Front Oncol 2022; Di Giorgio et al., Biochem Pharmacol 2024). We have revised the manuscript to more clearly reference these supplementary data and to contextualize the observed pro-proliferative effects of LIF within the framework of our previously published studies.

Q6. Regarding Figure 7, in case the LIF-stimulated cells grow more, it is likely that the conditioned media will be conditioned by an higher number of cells, posing doubts on the presence of specific mediators rather then being affected by just an higher number of cells during the 3-4 day culture before conditioned media collection. Moreover, alpha-smooth muscle actin should not be expressed by terminally differentiated myotubes…

R6. We thank the reviewer for this comment. We acknowledge that increased proliferation of LIF-stimulated spheroids may contribute to differences in conditioned media due to variations in cell number. Although spheroids were seeded at identical initial densities and cultured for the same duration, we cannot fully exclude a contribution of cell number to the observed paracrine effects.

To better characterize the conditioned media, LIF concentrations were quantified by ELISA (Figure 7C). Basal LIF levels were detected in conditioned media from untreated spheroids (~17 pg/mL) and were increased in media from LIF-stimulated spheroids (~24 pg/mL), indicating enrichment of LIF-related signaling. These findings support a tumor-driven paracrine effect, while not allowing attribution of the observed biological effects to a single specific mediator.

Regarding αSMA, we agree that it is not a marker of terminal myogenic differentiation. Differentiated C2C12 myotubes exposed to conditioned media from LIF-stimulated spheroids showed reduced myotube diameter and decreased αSMA immunofluorescence (Figure 7D–E), consistent with impaired cytoskeletal organization and compromised myogenic maturation.

Minor comments:

Q1. Avoid the use of acronyms in the abstract (e.g. GC for gastric cancer).

R1. We removed acronyms from the abstract, replacing them with the full terms where appropriate.

Q2. Please thoroughly revise the manuscript for typos or grammatical errors (e.g. line 59, 75)

R2. The manuscript has been thoroughly revised to correct typographical and grammatical errors throughout the text.

Q3. Figure 1, in panel C the x axis legend is cut.

R3. The x-axis label in Figure 1C has been corrected to ensure full visibility.

Q4. TGCA or TCGA repository? Also ATGC sounds new in figure 2? Please revise.

R4. The correct datasets used in this study are TCGA-STAD and the ACRG cohort. The terms “TGCA” and “ATGC” were typographical errors and have been corrected throughout Figure 2.

Q5. Please change surnatant to supernatant.

R5. The term “surnatant” has been corrected to “supernatant” in Figure 8B.

 

 

Reviewer 2 Report

Comments and Suggestions for Authors

This study presents a comprehensive investigation integrating tumor-intrinsic signaling, bile acid metabolism, and cachectic phenotypes, centered on the LIF–GDF15 axis in gastric cancer. Its strengths include the integration of an in-house cohort with external validation using the TCGA and ACRG datasets, bile acid profiling, and in vitro experiments.

However, the manuscript tends to overextend its conclusions by inferring causal mechanisms from primarily correlation-based observations. In particular, the conclusion that “LIF drives GDF15 and induces cachexia” is not sufficiently supported by the current data.

Major Comments

1. The causal relationship that LIF regulates GDF15 has not been demonstrated.

The title and conclusions of this study strongly suggest a causal relationship in which tumor-derived LIF promotes GDF15 expression and drives cachexia. However, no direct evidence is provided to support this claim. Specifically, data demonstrating that LIF stimulation induces GDF15 expression in tumor cells, that LIF inhibition reduces GDF15 expression, or that LIF–LIFR signaling directly regulates GDF15 transcription are lacking.

At present, the data demonstrate only that LIF and GDF15 are co-expressed and correlated. It cannot be excluded that both molecules are independently induced by a common inflammatory or stress-related pathway. Therefore, the characterization of LIF as an “upstream factor” of GDF15 is not sufficiently justified and requires revision.

2. The association with cachexia is described, but causality is not demonstrated.

The correlation analyses linking LIF/GDF15 expression to cachexia-related indicators such as L3SMI, SATI, weight loss, and serum albumin are intriguing. However, the number of cases included in the CT-based analyses is limited (n = 19), and no multivariate analyses adjusting for disease stage, tumor burden, or inflammatory markers are presented. As a result, it cannot be determined whether LIF and/or GDF15 are independent determinants of cachexia.

Despite this, the Discussion contains several statements implying that LIF/GDF15 “drive” cachexia. It should be clearly stated that this study demonstrates an association with cachexia rather than a causal relationship.

3. The interpretation of bile acid–LIFR antagonism is speculative.

The observation that secondary bile acids, which act as LIFR antagonists, are reduced in gastric cancer is highly novel and potentially important. However, experimental validation is lacking regarding (i) whether bile acid concentrations in gastric cancer tissues are sufficient to meaningfully regulate LIFR signaling, (ii) whether the observed bile acid alterations are a cause or a consequence of tumor progression, and (iii) the extent to which these changes are influenced by cachexia or nutritional status.

At present, the proposed bile acid–LIFR–LIF axis remains a hypothetical model, and its presentation should therefore be accompanied by more cautious wording.

4. Consistency and interpretation of survival analyses.

While high GDF15 expression is associated with poor prognosis in the ACRG cohort, no significant association is observed in the TCGA cohort. The manuscript does not sufficiently address how differences in patient background, data acquisition platforms, or cutoff selection (particularly the use of auto-selected best cutoffs) may contribute to this discrepancy. Strong conclusions should be avoided when prognostic associations are not consistently reproduced across cohorts.

Minor Comments

1. Several grammatical and typographical errors are present (e.g., “drivs,” “musclesm”), and thorough proofreading is required.
2. The limitations associated with the use of “best cutoff” selection in Kaplan–Meier analyses should be explicitly acknowledged in the Methods or Discussion sections.
3. Figures 5 and 7 contain excessive information, particularly in the correlation matrix, which compromises readability.

Author Response

Reviewer 2:

This study presents a comprehensive investigation integrating tumor-intrinsic signaling, bile acid metabolism, and cachectic phenotypes, centered on the LIF–GDF15 axis in gastric cancer. Its strengths include the integration of an in-house cohort with external validation using the TCGA and ACRG datasets, bile acid profiling, and in vitro experiments.

However, the manuscript tends to overextend its conclusions by inferring causal mechanisms from primarily correlation-based observations. In particular, the conclusion that “LIF drives GDF15 and induces cachexia” is not sufficiently supported by the current data.

 

Major Comments

The causal relationship that LIF regulates GDF15 has not been demonstrated.

Q1. The title and conclusions of this study strongly suggest a causal relationship in which tumor-derived LIF promotes GDF15 expression and drives cachexia. However, no direct evidence is provided to support this claim. Specifically, data demonstrating that LIF stimulation induces GDF15 expression in tumor cells, that LIF inhibition reduces GDF15 expression, or that LIF–LIFR signaling directly regulates GDF15 transcription are lacking.

R1. We thank the reviewer for this important comment. We have now provided experimental evidence showing that stimulation of gastric cancer spheroids with recombinant human LIF induced a significant transcriptional upregulation of GDF15, together with proliferation- and EMT-associated genes (c-MYC and SNAIL1), supporting activation of tumor programs linked to aggressiveness and cachexia-related signaling (Figure 7A).

Moreover, pharmacological inhibition of LIFR signaling reduced the expression of GDF15 as well as KI-67 and SNAIL1 (Supplementary Figure 4), indicating that GDF15 expression in this experimental context is, at least in part, dependent on LIF-LIFR signaling. While these findings do not establish a direct transcriptional mechanism or exclude additional regulatory inputs, they provide functional evidence that LIF signaling contributes to the regulation of GDF15 expression in gastric cancer cells.

Accordingly, we have revised the manuscript to clarify that our data support a functional association between tumor-derived LIF signaling and GDF15 expression, without claiming a unidirectional or exclusive causal pathway, and we have tempered all statements implying that LIF “drives” cachexia.

Q2. At present, the data demonstrate only that LIF and GDF15 are co-expressed and correlated. It cannot be excluded that both molecules are independently induced by a common inflammatory or stress-related pathway. Therefore, the characterization of LIF as an “upstream factor” of GDF15 is not sufficiently justified and requires revision.

The association with cachexia is described, but causality is not demonstrated.

The correlation analyses linking LIF/GDF15 expression to cachexia-related indicators such as L3SMI, SATI, weight loss, and serum albumin are intriguing. However, the number of cases included in the CT-based analyses is limited (n = 19), and no multivariate analyses adjusting for disease stage, tumor burden, or inflammatory markers are presented. As a result, it cannot be determined whether LIF and/or GDF15 are independent determinants of cachexia.

Despite this, the Discussion contains several statements implying that LIF/GDF15 “drive” cachexia. It should be clearly stated that this study demonstrates an association with cachexia rather than a causal relationship.

R2. We agree that our data demonstrate co-expression and correlation between LIF and GDF15, without establishing a causal or upstream-downstream relationship. We have therefore revised the manuscript to remove statements implying that LIF acts as an upstream driver of GDF15 and to consistently frame their relationship as associative.

In addition, we acknowledge that the correlations between LIF/GDF15 expression and cachexia-related parameters are based on a limited CT-assessed sub-cohort (n = 19) and univariate analyses. Accordingly, we have revised the Discussion to explicitly state that these findings demonstrate an association with cachexia rather than a causal relationship, and to recognize the lack of multivariate adjustment as a limitation.

Q3. The interpretation of bile acid–LIFR antagonism is speculative.

The observation that secondary bile acids, which act as LIFR antagonists, are reduced in gastric cancer is highly novel and potentially important. However, experimental validation is lacking regarding (i) whether bile acid concentrations in gastric cancer tissues are sufficient to meaningfully regulate LIFR signaling, (ii) whether the observed bile acid alterations are a cause or a consequence of tumor progression, and (iii) the extent to which these changes are influenced by cachexia or nutritional status.

At present, the proposed bile acid–LIFR–LIF axis remains a hypothetical model, and its presentation should therefore be accompanied by more cautious wording.

Consistency and interpretation of survival analyses.

While high GDF15 expression is associated with poor prognosis in the ACRG cohort, no significant association is observed in the TCGA cohort. The manuscript does not sufficiently address how differences in patient background, data acquisition platforms, or cutoff selection (particularly the use of auto-selected best cutoffs) may contribute to this discrepancy. Strong conclusions should be avoided when prognostic associations are not consistently reproduced across cohorts.

We thank the reviewer for these comments. With respect to the bile acid-LIFR axis, we would like to clarify that the antagonistic activity of secondary bile acids toward LIFR has been previously demonstrated by our group through biochemical, structural, and functional approaches (Di Giorgio et al., Biochem Pharmacol 2024). Therefore, the concept of bile acids acting as endogenous LIFR antagonists is not speculative per se. In the present study, we extend these findings by showing that gastric cancer tissues display a selective depletion of secondary bile acids with LIFR antagonist activity, thereby supporting a loss of endogenous restraint on LIF signaling in the tumor microenvironment.

We agree, however, that the current study does not directly establish whether intratumoral bile acid concentrations are sufficient to quantitatively modulate LIFR signaling in vivo, nor does it determine whether the observed bile acid alterations represent a cause or a consequence of tumor progression or are influenced by cachexia or nutritional status. Accordingly, we have revised the manuscript to present the bile acid-LIFR-LIF axis as a working model and have adopted more cautious wording throughout the Results and Discussion sections.

Regarding survival analyses, we acknowledge that the association between high GDF15 expression and poor prognosis was observed in the ACRG cohort but not consistently reproduced in the TCGA-STAD dataset. We have expanded the Discussion to address potential sources of discrepancy between cohorts, including differences in patient populations, data acquisition platforms, and cutoff selection strategies, particularly the use of auto-selected best cutoffs. Consequently, we have moderated our conclusions to avoid overinterpretation of prognostic associations that are not uniformly replicated across datasets.

Minor Comments

Q1.Several grammatical and typographical errors are present (e.g., “drivs,” “musclesm”), and thorough proofreading is required.

R1.The manuscript has been carefully proofread and all grammatical and typographical errors, including those highlighted by the reviewer, have been corrected.

Q2.The limitations associated with the use of “best cutoff” selection in Kaplan–Meier analyses should be explicitly acknowledged in the Methods or Discussion sections.

R2. We have revised the Methods section to explicitly clarify the use of a data-driven “Auto-select best cutoff” approach for Kaplan–Meier survival analyses, as implemented in the Kaplan–Meier Plotter. We now acknowledge that this procedure involves testing multiple expression thresholds with false discovery rate correction and that, despite this correction, the use of an optimized cutoff may introduce bias and potentially overestimate prognostic effects.

Q3.Figures 5 and 7 contain excessive information, particularly in the correlation matrix, which compromises readability.

R3.While Figures 5 and 7 include multiple parameters, this integrated presentation was intended to capture the multifactorial nature of the analyzed phenotypes. The main findings are highlighted and explained in the Results section to facilitate interpretation.

Reviewer 3 Report

Comments and Suggestions for Authors

This study identifies the LIF/GDF15 axis as a central driver of cancer-associated cachexia in gastric cancer, demonstrating that tumor-derived LIF promotes GDF15 expression and directly impairs skeletal muscle differentiation, while gastric tumors selectively deplete secondary bile acids with LIFR antagonist activity, creating a microenvironment that sustains LIF signaling and correlates with reduced muscle mass, impaired nutritional status, and poor clinical outcomes. The manuscript is well-writen. The methodology is clear and appropriate.  However, some issues should be adressed.
1.  The link between tumor cytokine expression and cachexia-related features is shown through the correlation matrix and patient-level data. CT-based measurements of lumbar skeletal muscle and subcutaneous fat are presented, and the main correlations are summarized in the heatmap. These results give an initial view of how tumor LIF and GDF15 relate to nutritional and inflammatory markers. The negative correlations with albumin and muscle index shown in Figure 5D are in line with a cachexia profile. However, the study does not include multivariable analyses to account for possible confounding factors. As a result, the correlations in the heatmap (Figure 5D) may be influenced by other variables such as tumor stage, BMI, age, sex, comorbidities, or overall inflammatory status.
2. Systemic cachexia features are assessed using CT-based muscle and fat measurements (Figure 5A–B) and serum nutritional markers and blood counts shown in the patient summary (Figure 5C), with correlations presented in the matrix (Figure 5D). These analyses help link tumor gene expression with clinically relevant outcomes. The conclusion that higher tumor LIF/GDF15 is associated with systemic inflammation and nutritional decline is based on these cross-sectional correlations. The study relates tumor mRNA expression to systemic cachexia features but does not measure circulating LIF or GDF15 protein levels, nor key inflammatory markers such as C-reactive protein or IL-6.

Author Response

This study identifies the LIF/GDF15 axis as a central driver of cancer-associated cachexia in gastric cancer, demonstrating that tumor-derived LIF promotes GDF15 expression and directly impairs skeletal muscle differentiation, while gastric tumors selectively deplete secondary bile acids with LIFR antagonist activity, creating a microenvironment that sustains LIF signaling and correlates with reduced muscle mass, impaired nutritional status, and poor clinical outcomes. The manuscript is well-writen. The methodology is clear and appropriate.  However, some issues should be adressed.

Q1. The link between tumor cytokine expression and cachexia-related features is shown through the correlation matrix and patient-level data. CT-based measurements of lumbar skeletal muscle and subcutaneous fat are presented, and the main correlations are summarized in the heatmap. These results give an initial view of how tumor LIF and GDF15 relate to nutritional and inflammatory markers. The negative correlations with albumin and muscle index shown in Figure 5D are in line with a cachexia profile. However, the study does not include multivariable analyses to account for possible confounding factors. As a result, the correlations in the heatmap (Figure 5D) may be influenced by other variables such as tumor stage, BMI, age, sex, comorbidities, or overall inflammatory status.

R1. We thank the reviewer for this comment. We agree that the correlations presented in Figure 5D are based on univariate analyses and do not account for potential confounding factors such as tumor stage, BMI, age, sex, comorbidities, or global inflammatory status. The aim of this analysis was to provide an exploratory overview of the relationships between tumor LIF/GDF15 expression and cachexia-related clinical, nutritional, and body composition parameters in a well-characterized sub-cohort of patients. We acknowledge that these associations do not establish independence or causality. Accordingly, we have revised the manuscript to explicitly recognize the lack of multivariable adjustment as a limitation and to frame these findings as hypothesis-generating rather than definitive.

Q2. Systemic cachexia features are assessed using CT-based muscle and fat measurements (Figure 5A–B) and serum nutritional markers and blood counts shown in the patient summary (Figure 5C), with correlations presented in the matrix (Figure 5D). These analyses help link tumor gene expression with clinically relevant outcomes. The conclusion that higher tumor LIF/GDF15 is associated with systemic inflammation and nutritional decline is based on these cross-sectional correlations. The study relates tumor mRNA expression to systemic cachexia features but does not measure circulating LIF or GDF15 protein levels, nor key inflammatory markers such as C-reactive protein or IL-6.

R2. We thank the reviewer for this observation. We agree that the associations between tumor LIF/GDF15 expression and systemic cachexia-related features are based on cross-sectional correlations. In the revised manuscript, we have expanded the analyses by including C-reactive protein measurements and by stratifying patients according to nutritional status using serum prealbumin levels (doi:10.1007/s10120-024-01472-y). As correctly noted, circulating LIF or GDF15 protein levels were not assessed. These limitations are now explicitly acknowledged, and the Results and Discussion have been revised to frame the findings as associative rather than causal.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors revised the manuscript in a satisfatorily manner, implementing most of my suggestions.

I just noticed the persistence of typos (e.g. 'hepatocarcinoman' in line 220 or 'Weigt' in Figure 5c), suggesting that a thorough revision is still needed. 

Author Response

R1. We thank the Reviewer for pointing this out. We apologize for the remaining typographical errors and confirm that the manuscript has now undergone a thorough proofreading. All identified typos, including those mentioned (e.g., “hepatocarcinoman” and “Weigt”), have been corrected in the revised version.

Reviewer 2 Report

Comments and Suggestions for Authors

This study demonstrates a high level of originality by comprehensively analyzing tumor-intrinsic signaling, bile acid metabolism, and cachectic phenotypes in gastric cancer, centered on the LIF–GDF15 axis. Its major strengths lie in the multi-layered approach, which integrates analysis of an institutional cohort with TCGA and ACRG datasets, bile acid profiling, and in vitro experiments.

Regarding the primary concern raised in the initial peer review—namely, the overinterpretation of causality from correlational data—the revised manuscript has been appropriately modified overall. In particular, retracting assertive statements such as “LIF drives GDF15 or cachexia” and repositioning these claims as functional associations has significantly improved the scientific validity of the study.

Below, the manuscript is re-evaluated according to the major points of concern.

1. Claims of causality in the LIF–GDF15 relationship

The newly added LIF stimulation and LIFR inhibition experiments in human gastric cancer spheroids can be regarded as supportive evidence demonstrating that LIF signaling is functionally associated with GDF15 expression. Notably, the concurrent modulation of tumor progression-related genes such as c-MYC and SNAIL1 alongside GDF15 provides useful context for understanding the biological consequences of LIF signal activation.

However, as the authors themselves explicitly acknowledge, evidence for a direct transcriptional regulatory mechanism or a unidirectional and exclusive control relationship between LIF and GDF15 remains lacking.

Nevertheless, the revised manuscript avoids causal assertions in the title, abstract, and main text, consistently limiting interpretations to functional associations. On this basis, the concern regarding causal overstatement in this section is considered adequately addressed.

2. Association with cachexia and statistical limitations

The associations observed between LIF/GDF15 expression and cachexia-related indicators—including L3SMI, SATI, body weight loss, and serum albumin—are intriguing and represent results of substantial hypothesis-generating value. However, as noted in the initial review, the limited number of cases available for CT-based analysis (n = 19), as well as the absence of multivariate adjustment for disease stage, tumor burden, and inflammatory markers, remain important methodological constraints.

In the revised manuscript, these findings are clearly framed as associations rather than evidence of causation, and the statistical limitations are explicitly acknowledged. Furthermore, the discussion has been appropriately tempered. Taken together, these revisions largely resolve the issue of overinterpretation in this section.

3. Interpretation of the bile acid–LIFR–LIF axis

The concept that secondary bile acids function as endogenous antagonists of LIFR is supported by prior literature (Di Giorgio et al., Biochemical Pharmacology, 2024). Therefore, this aspect of the study does not warrant being characterized as entirely speculative.

That said, several key questions remain unresolved, including the extent to which tumor-specific bile acid concentrations can regulate LIFR signaling in vivo, whether observed bile acid alterations are a cause or a consequence of tumor progression, and how cachexia or nutritional status may influence these processes. In this context, the authors’ decision to reposition this framework as a “working model” is appropriate. The revised model figure and accompanying text clearly indicate its hypothetical nature, which improves conceptual clarity.

4. Consistency of prognostic analyses

The discrepancy whereby high GDF15 expression is associated with poor prognosis in the ACRG cohort but not replicated in TCGA-STAD is appropriately discussed. Explicit consideration of cohort characteristics, platform differences, and cutoff determination (auto-selected best cutoff) strengthens the interpretative framework.

Moreover, clearly acknowledging the limitations and potential biases inherent to the best-cutoff approach in the Methods and Discussion sections is commendable from the standpoint of analytical transparency. In the revised version, prognostic conclusions are no longer overstated, resulting in a more balanced and cautious interpretation.

5. Minor points

Grammatical and typographical errors have been adequately corrected.

Figures 5 and 7 remain information-dense, and further improvement in readability may be possible; however, the intent to present the results comprehensively is understandable and does not constitute a major concern.

 

The reviewer sincerely appreciate the authors’ careful and constructive efforts in addressing our concerns, and thank them for their hard work in revising the manuscript.

Author Response

R1. We thank the Reviewer for the careful re-evaluation of our revised manuscript and for the positive and constructive feedback. We are pleased that the revisions have adequately addressed the major concerns raised in the initial review, particularly regarding the interpretation of causality, statistical limitations, and the framing of the bile acid–LIFR–LIF axis as a working model.

As noted by the Reviewer, we have revised the manuscript to consistently avoid causal overstatements, to explicitly acknowledge methodological constraints, and to provide a more balanced interpretation of prognostic analyses across independent cohorts. In addition, Figures 5 and 7 have been refined to improve readability while preserving the comprehensive presentation of the results. We appreciate the Reviewer’s recognition of these efforts and thank them for their valuable guidance, which has substantially improved the clarity and rigor of the study.