HMGA1 Has Predictive Value in Response to Chemotherapy in Gastric Cancer

Round 1

Reviewer 1 Report

Pádua et al. have discovered that high‐mobility group AT‐hook 1 (HMGA1) can serve as a prognostic biomarker for gastric cancer (GC) patients receiving an adjuvant chemotherapy. Their study is well conducted, the manuscript is concisely written, and data neatly organized. As such, this inspiring work has the potential to improve the guidance for the clinical diagnosis and treatment of GC. 

1) It is not clear whether for which type of chemotherapy is HMGA1 a better prognostic biomarker? Could the authors further stratify their OS curves based on the exact type of therapeutic regimen used?

2) Although the authors have shown significant correlation between HMGA1 expression and adjuvant chemotherapy in terms of OS, it is not clear whether there is also a significant association in terms of DFS?

3) Please change "i3S-Institute" to "i3S - Institute" (line 6).

4) Please replace "IPATIMUP-Institute" with "IPATIMUP - Institute" (line 7).

5) Please change "CNC-Center" to "CNC - Center" (line 10).

6) Please replace "laboratory" with "Laboratory" (line 11).

7) Please change "5-year" to "5 year" (lines 18, 52, 118).

8) Please replace "over-expressed" with "overexpressed" (line 20).

9) Please change "analyzed, by immunohistochemistry," to "analyzed by immunohistochemistry" (line 23).

10) Please replace "one-hospital" with "single hospital" (line 23).

11) Please change "n=323" to "n = 323" (lines 24, 97).

12) Please replace "(stages I to IV)," with "(stages I to IV)" (line 24).

13) Please change "comparing" to "compared" (line 27).

14) Please replace "GC" with "GCs" (line 36).

15) Please change "Lauren" to "Laurén" (lines 37, 115, 145, 218, 223).

16) Please replace "categorizes" with something like "involves" (line 39).

17) Please change "resection." to "resection" (line 46).

18) Please replace "protein," with "protein" (line 64).

19) Please change "and thereby influencing" to "and thereby influences" or "thereby influencing" (line 66).

20) Please replace "HMGA1 high" with "high HMGA1" (line 75).

21) Please change "aims of this study were" to "aim of this study was" (line 84).

22) Please replace "adenocarcinomas" with "adenocarcinoma" (line 85).

23) Please change "single-hospital" to "single hospital" (line 88).

24) From the "2.1. Patients" Methods section is not clear why the number of patients was reduced to n = 323 from the Mesquita et al. [41] cohort?

25) Please replace "microarrays (TMAs)" with "microarrays" (line 101).

26) Please change "TMAs" to "Tissue microarrays (TMAs)" (line 103).

27) Please replace "immunohistochemical" to "immunohistochemical (IHC)" (line 103).

28) Please replace "that" with "which" (line 104).

29) Please change "ºC" to "oC" (line 107).

30) Please replace "consensus of" with "consensus between" (line 109).

31) Please define abbreviation for "TNM" (line 116).

32) From Figure 1 is not clear what do the IHC images indicate?

33) Please indicate the staining used including explanation of brown and purple staining patterns in the legend to Figure 1. In addition, what are the invaginations that we see in Figure 1A? Why are these absent in Figure 1B? What are the interstitial regions that are completely white devoid of cells? Why some purple cells in Figure 1A appear like undergoing locomotion? Annotating Figure 1 would improve understanding the IHC profile.

34) One data point is missing in the "Chemotherapy (ND)" and "All cases (%)" cell in Table 1.

35) It is not clear why some data points are plotted using bold formatting in Table 1?

36) Please change "Media ± SD" to "Median ± SD", "67.68±11.85" to "67.68 ± 11.85", "68.26±11.61" to "68.26 ± 11.61", "66.39± 12.32" to "66.39 ± 12.32", "32-95" to "32–95", "33-95" to "33–95", "32-87" to "32–87" in Table 1.

37) Please change "threshold<0.05" to "threshold < 0.05" (line 141).

38) Please replace "P<0.001" with "P < 0.001" (lines 146, 174, 194).

39) Please change "opposition" to "opposition" (line 146).

40) Please replace "GC we" with "GC, we" (line 152).

41) Please change "Kaplan-Meyer" to "Kaplan-Meier" (line 153).

42) Please replace "3A-D" with "3A–D" (line 155).

43) Please change "GC," to "GC" (line 157).

44) Please change "patients.The" to "patients. The" (line 158).

45) Please replace "40, 60" with "40 and 60" (lines 160, 166, 190, 207).

46) Please change "OS, according to HMGA1 expression," to "OS according to HMGA1 expression" (line 163).

47) Please replace "39±2 months versus 24±3" with "39 ± 2 months versus 24 ± 3" (line 173).

48) Please remove italics formatting from "versus" (line 174).

49) Please change "patients non-treated" to "non-treated patients" (line 174).

50) Please replace "chemotherapy we" with "chemotherapy, we" (line 178).

51) Please replace "39±2" with "39 ± 2" (line 179).

52) Please change "30±3" to "30 ± 3" (line 180).

53) Please replace "P=0.051" with "P = 0.051" (line 180).

54) Please change "5A-F" to "5A–F" (line 195).

55) Please replace "40±4" to "40 ± 4" (line 196).

56) Please change "15±4" to "15 ± 4" (line 197).

57) Please replace "26±4 months versus 6±2" with "26 ± 4 months versus 6 ± 2" (line 198).

58) "In the present study, the value of HMGA1 nuclear expression in predicting prognosis and response to treatment, besides its association with clinico-pathological of the tumors, was determined in a series of 323 gastric carcinomas, the largest tumor series used to date with this intent and the more complete regarding clinico-pathological features, survival data and treatment data" (line 210) is too long. Please split into at least two sentences.

59) Please change "HMGA1 nuclear" to "nuclear HMGA1" (line 210).

60) Please replace "cases [[27] – 59%; [42] – 75%; [29] – 74%]" with "cases: 59% [27], 75% [42], 74% [29]" (line 216).

61) Please change "regarding" to "based on" (line 223).

62) Please replace "HMGA1 high nuclear" with "high nuclear HMGA1" (line 227).

63) "Our results unequivocally showed that patients with high levels of HMGA1 expression in the tumor had extraordinarily better OS rates when they are subjected to treatment with adjuvant chemotherapy, implying that they can benefit more from adjuvant treatment than patients with low expression of HMGA1, where no significant difference was observed between patients treated and those not treated" (line 228) is too long. Please split into at least 2 sentences.

64) Please change "like" to "such as" (line 234).

65) Please replace "metastasis, EMT [18,26,27]" with "metastasis and EMT [18,26,27]," (line 235).

66) Please change "considered," to "considered" (line 237).

67) Please replace "management," with "management" (line 238).

68) Please change "65% of the" to "the 65% of" (line 240).

69) Please replace "diagnosed in stages II-IV (those HMGA1 high)" with "with high HMGA1 expression diagnosed in stages II-IV " (line 240).

70) Please change "in what concerns" to "in" (line 243).

71) "In our series, we did not observe a prognostic value for HMGA1" (line 247) contrasts with what is presented in Figures 4 and 5. Do the authors actually mean that there is no "prognostic value for HMGA1" in the absence of chemotherapy?

72) Please replace "HMGA1-high" with "high HMGA1" (line 252).

73) Please change "R.A.," to "R.A." (line 256).

74) Please replace "R.A.," with "R.A." (line 260).

75) Please provide city and state for "Centro Hospitalar de São João (CHSJ)" (line 263) in the Acknowledgements section.

76) Please translate "Norte" into English language (line 273).

77) Please change "Informed patient" to "Informed" (line 279).

78) Please replace "manuscript," with "manuscript" (line 283).

Author Response

Reviewer 1

Pádua et al. have discovered that high‐mobility group AT‐hook 1 (HMGA1) can serve as a prognostic biomarker for gastric cancer (GC) patients receiving an adjuvant chemotherapy. Their study is well conducted, the manuscript is concisely written, and data neatly organized. As such, this inspiring work has the potential to improve the guidance for the clinical diagnosis and treatment of GC. 

1) It is not clear whether for which type of chemotherapy is HMGA1 a better prognostic biomarker? Could the authors further stratify their OS curves based on the exact type of therapeutic regimen used?

For clarification of this point, we have now included a Supplementary table with the treatment regimens. As can be seen on Supplementary table 1 (Table S1), there are 16 treatment regimens used, which hampers a statistical analysis with all of them, due to the reduced number of cases in each group.

2) Although the authors have shown significant correlation between HMGA1 expression and adjuvant chemotherapy in terms of OS, it is not clear whether there is also a significant association in terms of DFS?

We apologize for not mentioning these results in the original manuscript. In fact, we had already analysed the correlation between HMGA1 expression and adjuvant chemotherapy regarding DFS but we have not observed differences. We have now included this information in the results part of the revised version of the Ms, as Supplementary Figure 1 (Figure S1).

All the requests numbered by the Reviewer 1 as 3) to 23) as well as 25) to 31), 36) to 70) and 72) to 78) were accordingly modified and they can be tracked in the revised manuscript.

24) From the "2.1. Patients" Methods section is not clear why the number of patients was reduced to n = 323 from the Mesquita et al. [41] cohort?

The TMAs we have used in the current study, were also used in previous studies ([42], doi.org/10.1186/s12964-019-0465-9; doi.org/10.1007/s10120-020-01064-6; doi.org/10.3390/biomedicines9091249). When we tested HMGA1, we observed that for some cases (more precisely 10), we had exhausted the tumor material and for that reason it was impossible to perform the IHC analysis of HMGA1 in all the cases we had initially.

32) From Figure 1 is not clear what do the IHC images indicate? 33) Please indicate the staining used including explanation of brown and purple staining patterns in the legend to Figure 1. In addition, what are the invaginations that we see in Figure 1A? Why are these absent in Figure 1B? What are the interstitial regions that are completely white devoid of cells? Why some purple cells in Figure 1A appear like undergoing locomotion? Annotating Figure 1 would improve understanding the IHC profile.

We have now included the requested information in the legend to Figure 1. Figure 1 A and 1 B refer to two different gastric cancer cases, that are morphologically different since gastric tumors are very heterogeneous. The invaginations seen in Figure 1 A are glands that appeared to have in its interior some cellular content. This is most likely due to the orientation of the tissue block, and is an artifact. For this reason we chose another representative gastric cancer case, with high HMGA1 expression. We thank the Reviewer for pointing this out and hope it is more clear now.  

34) One data point is missing in the "Chemotherapy (ND)" and "All cases (%)" cell in Table 1.

We apologize for the missing value in Table 1. We have included it (0.6) in the revised version of the Ms.

35) It is not clear why some data points are plotted using bold formatting in Table 1?

Thank you for pointing this out. The data points formatted in bold refer to n-values, while those not bolded refer to the corresponding percentages. We did that in order to facilitate the reading of the table.

36) Please change "Media ± SD" to "Median ± SD

Thank you for pointing this out. In fact, it is "Mean ± SD" and we have changed accordingly in the revised version of the Ms.

71) "In our series, we did not observe a prognostic value for HMGA1" (line 247) contrasts with what is presented in Figures 4 and 5. Do the authors actually mean that there is no "prognostic value for HMGA1" in the absence of chemotherapy?

In our series, we observe that HMGA1 does not have prognostic value in all cases (Figure 2). Even when we stratify by chemotherapy treatment there is no prognostic value (Figure 4C and D). In Figure 4C it is almost significant but we cannot consider significant. What we see is that when we stratify patients based on HMGA1 expression, those with High expression have better overall survival when treated with chemotherapy, that is benefit more from chemotherapy. We consider this result is predictive of response to chemotherapy.

Author Response File: Author Response.docx

Reviewer 2 Report

This paper is a well-written original article assessing prognostic impact of HMGA1 expression in patients who had adjuvant chemotherapy for gastric cancer (GC). In this article, the authors clarified that high levels of HMGA1 expression were associated with good response in adjuvant chemotherapy. This finding may have significance in clinical setting, but there are some points to be commented or revised.

Major Comments;

1. In IHC analysis, the authors focused on the proportion of stained cells to classified GC patients into two groups. To assess the protein expression of HMGA1 more accurately, not only proportion of stained cells but also intensity of staining should be focused on. Was intensity of staining associated with long-term outcomes?
2. In table1, HMGA1 expression was significantly associated with intestinal Lauren-type. Is HMGA1 associated with differentiation of GC cells? Is there any report about impact of HMGA1 on differentiation of GC cells?
3. There are some articles reporting that diffuse-type GC is more resistant for chemotherapy compared to intestinal-type GC. From this standpoint, the date in this article is not efficient to prove the prognostic impact of HMGA1 in patients with adjuvant chemotherapy. The authors should add the multivariate analyses into Table1 to show that HMGA1 itself is independently associated with chemo sensitivity.
4. As the authors mentioned, MHMGA1 has been reported to have oncogenic functions in various cancer, so the finding in this article seems to contradict these previous reports. There is unfortunately no data and reference which support their idea form the point of molecular biology in this article. The authors should clarify the molecular functions of HMGA1 which causes chemo sensitivity in GC.

Author Response

Reviewer 2

This paper is a well-written original article assessing prognostic impact of HMGA1 expression in patients who had adjuvant chemotherapy for gastric cancer (GC). In this article, the authors clarified that high levels of HMGA1 expression were associated with good response in adjuvant chemotherapy. This finding may have significance in clinical setting, but there are some points to be commented or revised.

Major Comments;

In IHC analysis, the authors focused on the proportion of stained cells to classified GC patients into two groups. To assess the protein expression of HMGA1 more accurately, not only proportion of stained cells but also intensity of staining should be focused on. Was intensity of staining associated with long-term outcomes?

- We have not classified cases according to the intensity of staining but only according to the percentage of positive tumor cells. In our experience the intensity of the staining is more subject to technical variations and therefore more difficult to compare between cases, hence we opted to stratify the cases according to the percentage of stained cells.

In table1, HMGA1 expression was significantly associated with intestinal Lauren-type. Is HMGA1 associated with differentiation of GC cells? Is there any report about impact of HMGA1 on differentiation of GC cells?

- To our knowledge there is only one study referring the association of HMGA1 with poor differentiation in gastric cancer. We cite this study in the discussion part of the manuscript (line 294).

There are some articles reporting that diffuse-type GC is more resistant for chemotherapy compared to intestinal-type GC. From this standpoint, the date in this article is not efficient to prove the prognostic impact of HMGA1 in patients with adjuvant chemotherapy. The authors should add the multivariate analyses into Table1 to show that HMGA1 itself is independently associated with chemo sensitivity.

- Thank you for suggesting the multivariate analysis. We have added it as Table 2 to the Results part of the manuscript and concluded that (line 262) “Furthermore, HMGA1 expression was associated with a lower risk of death in a Cox multivariable analysis restricted to patients diagnosed in stages II-IV treated with chemotherapy and that includes relevant clinico-pathological factors like Laurén classification, growth pattern, TNM staging, vascular and perineural invasion as covariates (Table 2; HR=0.53; 95% CI 0.30-0.95; P=0.03)”. We included the respective information in the Methods and Discussion parts of the manuscript.

As the authors mentioned, HMGA1 has been reported to have oncogenic functions in various cancer, so the finding in this article seems to contradict these previous reports. There is unfortunately no data and reference which support their idea form the point of molecular biology in this article. The authors should clarify the molecular functions of HMGA1 which causes chemo sensitivity in GC.

- In our view, a better response to chemotherapy, in a patient, does not necessarily mean increased chemosensitivity in vitro. Indeed, this is a hypothesis that does not find support in the molecular data obtained so far for HMGA1. However, it is not the only hypothesis. Another hypothesis is that tumors with high expression of HMGA1 are more aggressive (as demonstrated in vitro), therefore if these tumors are not treated more aggressively also (for instance with the inclusion of chemotherapy) their course will be worse. This is our hypothesis, which is supported by published data showing that gastric tumors with HMGA1 high expression have been associated with features of higher malignancy, such as increased proliferation, migration, metastasis, EMT in vitro. In fact there are examples of this mechanism in the literature. For instance, CDX2 negative colorectal tumors are more aggressive, due to loss of differentiation. Nevertheless patients with this type of tumor benefit more from chemotherapy. This is shown in the following highly cited paper (DOI: 10.1056/NEJMoa1506597).

Author Response File: Author Response.docx

Reviewer 3 Report

Comments:

1. In addition to 323 patients, do authors have any samples with HMGA1 level < 80% but >20%? If yes, please add it to the manuscript.
2. The representative protein levels of high and low HMGA1 by western blot should be added to the manuscript.
3. HMGA1 is closely related to C-Myc, EMT, SUZ12 and CCDC43. Any data from the patients with the levels of C-Myc, EMT events, SUZ12 and CCDC43?

Author Response

Reviewer 3

1) In addition to 323 patients, do authors have any samples with HMGA1 level < 80% but >20%? If yes, please add it to the manuscript.

- We thank the Reviewer for pointing this out and apologize for the mistake when describing the cutoff we have used to distinguish between cases with high expression and those with low expression of HMGA1. We have corrected this, by replacing the sentence in line115 with the following: “Samples were considered HMGA1-low when they were negative for HMGA1 or it was present in less than 20% of malignant cells and HMGA1-high otherwise, in consensus of 2 observers”.

2) The representative protein levels of high and low HMGA1 by western blot should be added to the manuscript.

- At this point, it is impossible to perform this experiment as we do not have fresh material from gastric tumors that could allow the extraction of protein with enough quality to evaluate HMGA1 expression by Western blot. In our view, this experiment is not very informative for the following reasons: the classification into high and low expression is based on the extension and not intensity of HMGA1 expression; tumor samples have other cell types that could bias the quantification of HMGA1 by western blot and, finally, the idea is to eventually apply the assessment of HMGA1 expression in the clinic, which is much more viable with a simple method like immunohistochemistry than with western blot.

3) HMGA1 is closely related to C-Myc, EMT, SUZ12 and CCDC43. Any data from the patients with the levels of C-Myc, EMT events, SUZ12 and CCDC43?

- We do not have any data in this series regarding the expression levels of c-myc, EMT-related genes, SUZ12 or CCDC43.

Author Response File: Author Response.docx

Reviewer 4 Report

In the present manuscript "HMGA1 has Predictive Value in Response to Chemotherapy in Gastric Cancer", the authors try to evaluate HMGA-1 as a potential biomarker in predicting response of adjuvant chemotherapy in gastric cancer patients. The authors try to distinguish their study from previous studies recognizing HMGA-1 as a biomarker for worse prognosis in patients with gastric cancer. The study is scientifically sound and well written. There are following questions for the author:

1. Why do you think high HMGA-1 expression correlates well with response to adjuvant chemotherapy? Is there any mechanistic precedence in the literature for this?
2. What sort of experiments do the authors propose to identify this mechanistic link between HMGA-1 and chemotherapy?
3. Since HMGA-1 impacts cell metabolism (Introduction and Ref. 17), do the authors think that intercepting this pathway may be useful as combination therapy in gastric cancer patients?
4. What was the effect of adjuvant chemotherapy on patient survival between patients according to Laurén classification, that is between intestinal, mixed and diffuse subtypes?

Author Response

Reviewer 4

In the present manuscript "HMGA1 has Predictive Value in Response to Chemotherapy in Gastric Cancer", the authors try to evaluate HMGA-1 as a potential biomarker in predicting response of adjuvant chemotherapy in gastric cancer patients. The authors try to distinguish their study from previous studies recognizing HMGA-1 as a biomarker for worse prognosis in patients with gastric cancer. The study is scientifically sound and well written. There are following questions for the author:

1) Why do you think high HMGA-1 expression correlates well with response to adjuvant chemotherapy? Is there any mechanistic precedence in the literature for this?

- In our view, a better response to chemotherapy, in a patient, does not necessarily mean increased chemosensitivity in vitro. Indeed, this is a hypothesis that does not find support in the molecular data obtained so far for HMGA1. However, it is not the only hypothesis. Another hypothesis is that tumors with high expression of HMGA1 are more aggressive (as demonstrated in vitro), therefore if these tumors are not treated more aggressively also (for instance with the inclusion of chemotherapy) their course will be worse. This is our hypothesis, which is supported by published data showing that gastric tumors with HMGA1 high expression have been associated with features of higher malignancy, such as increased proliferation, migration, metastasis, EMT in vitro. In fact there are examples of this mechanism in the literature. For instance, CDX2 negative colorectal tumors are more aggressive, due to loss of differentiation. Nevertheless patients with this type of tumor benefit more from chemotherapy. This is shown in the following highly cited paper (DOI: 10.1056/NEJMoa1506597).

2) What sort of experiments do the authors propose to identify this mechanistic link between HMGA-1 and chemotherapy?

In light of what we have discussed above, we propose to use a panel of gastric cancer cell lines with and without HMGA1, treated with a platinum- and/or a fluoropyrimidine-based chemotherapy agent and evaluate the response to chemotherapy. Nevertheless, as we mentioned in the previous response this has already been done (doi.org/10.1186/1471-2407-14-851; doi.org/10.1038/s41388-018-0394-x; doi.org/10.1007/s00109-017-1520-x; doi.org/10.26355/eurrev_201905_17803) and we discuss these results and put forward another hypothesis.

3) Since HMGA-1 impacts cell metabolism (Introduction and Ref. 17), do the authors think that intercepting this pathway may be useful as combination therapy in gastric cancer patients?

We agree with the Reviewer’s considerations. Indeed, Cao’s data show that HMGA1 seems to be a glycolysis regulator in gastric cancer and suggest that the HMGA1-c-Myc-glycolysis axis may be used as a new target in cancer therapy, in combination with commonly used chemotherapeutic agents. We have now clearly included the point in the Discussion of the revised version of the Ms.

4) What was the effect of adjuvant chemotherapy on patient survival between patients according to Laurén classification, that is between intestinal, mixed and diffuse subtypes?

Chemotherapy significantly improved the OS of patients having tumors classified as mixed according to the Laurén classification (mean OS of 35 ± 4 months versus 18 ± 4 months in cases not treated; P = 0.005). In patients with tumors of the intestinal and diffuse-type there is a trend to an increase in OS when patients are treated, but there is no statistical significance. We show the results obtained in the file attached below for the Reviewer assessment, but we did not include them in the manuscript as we think that they do not add to the main message of our study.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

This paper has been improved by author’s adequate revision and has become better than the original version.

However, authors do not meet the reviewer’s demand for clarifying the molecular functions of HMGA1 which causes chemo sensitivity and supports author’s hypothesis in GC. MHMGA1 has been reported to have oncogenic functions in various cancer, so the finding in this article seems to contradict these previous reports. Though authors took CDX2 as a molecule supporting their hypothesis, CDX2 has not reported to be directly associated with HMGA1 expression. Therefore, previous literature about CDX2 is not able to efficiently support and reinforce authors hypothesis.

From this standpoint, investigating and clarifying the molecular functions of HMGA1 which supports their clinical date is critically important. Authors should plan and add the experiment in vitro or in vivo to clarify the molecular functions of HMGA1 associating outcomes of patients with chemotherapy.