Selective inhibition of TGF-β1 produced by GARP-expressing Tregs overcomes resistance to PD-1/PD-L1 blockade in cancer

TGF-β1, β2 and β3 bind a common receptor to exert vastly diverse effects in cancer, supporting either tumor progression by favoring metastases and inhibiting anti-tumor immunity, or tumor suppression by inhibiting malignant cell proliferation. Global TGF-β inhibition thus bears the risk of undesired tumor-promoting effects. We show that selective blockade of TGF-β1 production by Tregs with antibodies against GARP:TGF-β1 complexes induces regressions of mouse tumors otherwise resistant to anti-PD-1 immunotherapy. Effects of combined GARP:TGF-β1/PD-1 blockade are immune-mediated, do not require FcγR-dependent functions and increase effector functions of anti-tumor CD8+ T cells without augmenting immune cell infiltration or depleting Tregs within tumors. We find GARP-expressing Tregs and evidence that they produce TGF-β1 in one third of human melanoma metastases. Our results suggest that anti-GARP:TGF-β1 mAbs, by selectively blocking a single TGF-β isoform emanating from a restricted cellular source exerting tumor-promoting activity, may overcome resistance to PD-1/PD-L1 blockade in patients with cancer.

The authors make the claim that anti-GARP:TGFB1 acts via effects in blunting the immunosuppressive activity of Tregs, however they do not provide any mechanistic data proving this to be the case. IN fact GARP is also clearly expressed at high levels in endothelial cells. The authors argument for not investigating drug responses in Treg-specific Garp-KO mice is not solid. They responded that, others have previously reported that Treg-specific Garp-KO mice do not show reduced growth of GL261 or MC38 tumors by comparison to WT. However, the correct experiment to undertake to address this point would be to treat tumor-bearing Foxp3-Garp-KO mice with anti-PD-1 in order to determine whether loss of Treg-associated GARP can synergize with anti-PD-1 in promoting tumor regression. I agree that it would take too long now to undertake the genetic KOs experiments for the purpose of this manuscript. An alternative strategy to address whether GARP expression on other cell types, such as endothelial cells, contribute to tumor regression, would be to immunologically deplete Tregs by, for example, using anti-CD25 depleting antibodies, and then determine whether anti-GARP-TGFB1 has any additional effect on tumor growth (see ).
Figure 2: what is the X axis of the middle column of graphs shown on the far left of the page? This is unnecessary, and makes it appear as though there is missing data.
Another unusual way that the authors have presented data is seen Figure 4. Here they present the number of cells per mm3, whereas it is more usual to present flow cytometry data as a percentage of live cells, or as a percentage of another cellular population. How do they calculate this parameter?
This issue also highlights the fact that the authors do not show immunohistochemical analysis/confirmation of their flow cytometry findings of no change in T cell numbers after therapy. Simple IHC for CD3+ or CD8+ T cells would support the flow cytometric findings. Moreover, it would also be informative to see the distribution of T cells across the tumor, especially in view of publications by Mariathasan et al., Nature 2018 and Dodagatta-Marri et al 2019 JITC, that show a redistribution of immune cells within the tumor after anti-TGFβ therapy. This is pertinent to the current manuscript, since Dodagatta-Marri et al suggest that a considerable portion of the antitumor effect of anti-TGFβ1/2/3 therapy is due to drug-action on Treg cells.
On page 10, the authors highlight the difference between their Treg-mediated mechanism of blocking GARP:TGFB1 with that of an anti-TGFβ study in a mammary tumor model that exhibits spatial T cell exclusion from the tumor parenchyma, due to physical blockade by TGFB-responsive CAFs (Mariathasan et al. 2018). But their proposed mechanism of action of anti-GARP may be more similar to that described by Dodagatta-Marri et al 2019 for anti-TGFβ1-3 antibodies, i.e. blocking active TGFβ on Tregs. Despite high GARP on endothelial cells, the author did not address the effect of anti-GARP:TGFβ1 on the vascular system. Are their morphological or functional changes in the vascular or lymphatic networks in vivo? This could be addressed for example by IHC. They mention unpublished negative data on in vitro analysis of GARP function on primary ECs, but this might not reflect the in vivo situation.
It is a shame that for immune profiling analysis of tumors, the authors only included 5 tumors per arm, and used the parameter of cells per mm2 (Fig 4). It seems that there could a 5 or even 10 fold increase in CD4+ and CD8+ T cells in SOME tumors treated with anti-GARP:Tgfb1 ( Figure 4a). However, this effect is lost, statistically, by mixing data from responding and non-responding tumors in a population wherein only 40% of tumors respond.
How do the authors explain the increase in Tregs, particularly GARP+ Tregs, after treatment with anti-GARP:TGFB1-FcD? Could this be due to increased proliferation of immature Tregs? Ki67 staining might address this. Can the authors isolate these Tregs and show that they are functionally less immunosuppressive, compared to those isolated from control-treated tumors? or at least stain use cellular markers indicative of Treg differentiation?
In two places, the authors refer to "Foxp3+ non-Treg T cells" but provide no citation to this cell type. Can the authors provide the reference or delete the statement.
Reviewer #2 (Remarks to the Author): The aim of this study was to analyze the effect of a selective blockade of TGF-β1 production by activated Tregs with antibodies against GARP:TGF-β1 in tumor models. Overall the authors found that this complexes is sufficient to induce regressions of mouse tumors otherwise resistant to anti-PD1 immunotherapy. Mechanistically they linked this with increased effector functions of anti-tumor CD8+ T cells without increasing immune cell infiltration or depleting Tregs within tumors. Finally they found GARP-expressing Tregs in about one third of human cutaneous melanoma metastases. Thus, this study builds the basis for a novel therapy using anti-GARP:TGF-β1 mAbs to overcome resistance to PD-1/PD-L1 blockade in patients with cancer.
Overall the authors have responded well to the reviewer comments. Specifically they performed additional experiments to show that their data are robust, generalizable, and statistically significant in multiple mouse tumor models. Thus, in my opinion this study does provide intriguing preclinical rationale for targeting this pathway for the immunotherapy of human cancers.
I agree with Reviewer 1 that this study lacks detailed mechanistic insights in vivo. But, I feel that the required experiments would go beyond the scope of this manuscript. However, these limitations should be discussed in this mansucript. This is especially the case for point 4 raised by reviewer 1.

Point by point reply to reviewer comments.
We would like to thank the reviewers for their positive reception of our work and their constructive input. In the following pages, each reviewer remark appears verbatim in black font, with our response immediately below in blue italic font.
Reviewer #1 (Remarks to the Author): This is a resubmission of an article investigating the potential of effector-negative anti-GARP-Tgfb1 IgG2a antibodies in murine preclinical immunotherapeutic oncology models. The manuscript shows that anti-GARP:TGFB1 has a similar efficacy to that of anti-pan-TGFβ ligand antibodies in potentiating anti-PD-1 activity, albeit that the increase in efficacy is from only 2/10 complete responders with anti-PD-1 monotherapy versus 4/10 for anti-PD-1 combined with the anti-GARP:TGFB1 drug. This is a small differential that is unlikely to be significantly different.
The reviewer cites partial results from Fig. 2 Fig. 3); ii) use a log2 scale to report tumor size, as this better illustrates not only CR but also partial responders (PR), which are now clearly enumerated in response to another comment of the reviewer below. Finally, to address another comment of the reviewer below, we have used an additional statistical approach for the meta-analysis of data shown in revised Fig. 3. P values for differences between anti-PD-1 monotherapy versus anti-PD-1 combined with anti-GARP:TGF-β1 in these meta-analyses are = 0.005. Our conclusion, unlike that of the reviewer, is that it is highly unlikely (<5% probability) that there is no difference between the treatments.
The authors have responded to both reviewers' requests on the statistical analysis of growth curves by stating that they used an analysis that takes into account longitudinal growth data. However, they still miss the point made by the reviewers, that there is missing data that artificially flattens the average growth curves, particularly of those lines wherein larger tumors are removed from study.
There is no missing data in the figures, as larger tumors are not removed from the studies to calculate average tumor sizes. The average growth curves are therefore not artificially flattened by missing data, in Fig. 2  For example, in Figure 2, the "average" growth curve data goes out to 45 days, yet the mice with the largest tumors are removed from the study at 15 days, thus the "average" tumor size after 15 days will be distorted downwards, leading to an artificial flattening of the growth rate in those arms from which tumors are removed.
Again, no mice are removed from the study to calculate average tumor size at any point in Fig. 2 or anywhere. When average tumor growth curves are "distorted downwards", it is because tumors start to regress and will be completed rejected in some mice, as can be seen on graphs of individual tumor growth shown on all revised figures.
Therefore, average growth rates may only be compared up to the time when the first mouse reaches its endpoint (of excessive tumor size).

, a first mouse reaches the endpoint on day 14. It belongs to the group receiving the isotype control antibody. Comparisons need to be continued after day 14 because i) treatments are still ongoing, and ii) complete responses (tumor rejections) in mice treated with antibody combinations occur in a vast majority of the cases after day 20. Complete responses, even if delayed by comparison to first death, are the most interesting and significant events observed in response to immunotherapies.
A standard and alternative way of dealing with this type of data is to plot Kaplan Meyer survival graphs and undertake statistical analysis using a Mann-Whitney U test or Wilcoxon rank-sum test.
We agree that Kaplan Meyer (KM) survival graphs and log rank tests are an interesting alternative, particularly well-suited for analyses of survival data derived from large numbers of patients or mice. In Fig. 2, the relatively low number of mice per group (n=9-10) precludes use of this approach for this given single experiment. This experiment is nevertheless crucial to our manuscript because it contains several important control groups that are not present in the other confirmatory experiments shown later in the manuscript. To address this comment and another one below, we propose to use KM representation for metaanalyses of data pooled from 7 experiments comparing anti-PD-1 alone to anti-PD-1 combined with anti-GARP:TGF-β1, and to include these graphs in a new main figure (revised Fig. 3) which would replace former Supplementary Fig. S2. Groups in the meta-analyses comprise 29-39 mice, which allows for KM plots and statistical analysis using a Wilcoxon test. P values for differences in survival between anti-PD-1 combined with anti-GARP:TGF-β1 and anti-PD-1 alone are 0.005, whether the antibodies are used as WT antibodies or as Fc-dead variants. We thank reviewer #1 for this excellent suggestion, as this allows us to demonstrate our point more clearly on lines 130-131 of the revised manuscript.
From data presented in Figure S2, it is impossible to tell how many tumors, from each arm of each of the five experiments displayed, shows complete tumor regression.
We agree that this was impossible to tell in the original Supplementary

tumor regressions or responses (CR), but also partial responses (PR, see other comment below). Due to the large amount of data and to follow another suggestion of the reviewer, revised Fig. 3 is presented as 2 meta-analyses of a total of 7 experiments (4 pooled experiments with WT antibodies, 3 pooled experiments with Fc dead antibodies).
The reader is therefore left with the impression that, whereas anti-PD-1 FcS monotherapy results in 2 to 3 complete tumor regressions per 10 mice, combination with anti-GARP:Tgfb1 FcS only raises this efficacy to 4 per ten mice (Main Fig 2).
This comment relates to partial results from one experiment shown in Fig. 2, but fails to acknowledge results from 10 other independent experiments. It has been mostly addressed above, except for the reference to the "third" tumor regression, which we address below.
[Note that in main Fig. 2 there is a third tumor in the anti-PD-1 FcS monotherapy arm that is showing a downward trajectory in growth at termination, which on a Kaplan Meyer graph would be considered a complete response or censuring. . Spider plots on mouse data yield graphs that are very difficult to read and interpret. We provide an example below, in which a spider plot is used to illustrate partial data shown in revised main Fig. 2.
To the best of our knowledge, spider plots are very well suited to plot evolution of tumor volumes in patients, but unfortunately not in mouse tumor models.
The authors make the claim that anti-GARP:TGFB1 acts via effects in blunting the immunosuppressive activity of Tregs, however they do not provide any mechanistic data proving this to be the case. IN fact GARP is also clearly expressed at high levels in endothelial cells. However, the correct experiment to undertake to address this point would be to treat tumor-bearing Foxp3-Garp-KO mice with anti-PD-1 in order to determine whether loss of Treg-associated GARP can synergize with anti-PD-1 in promoting tumor regression. I agree that it would take too long now to undertake the genetic KOs experiments for the purpose of this manuscript.   Fig. 7)

described on lines 273-294 in the revised manuscript.
We can now firmly conclude that targeting GARP:TGF-β1 on Tregs, but not on platelets, is necessary to increase the anti-tumor activity of anti-PD-1.
Our results also imply that targeting GARP:TGF-β1 on endothelial cells is not sufficient to observe anti-tumor activity.

We believe that these new data are the most important part of our reply to reviewers. They firmly establish the mechanism and specificity of anti-tumor activity of anti-GARP:TGF-β1 mAbs, which are now clearly demonstrated to act by targeting Tregs.
An alternative strategy to address whether GARP expression on other cell types, such as endothelial cells, contribute to tumor regression, would be to immunologically deplete Tregs by, for example, using anti-CD25 depleting antibodies, and then determine whether anti-GARP-TGFB1 has any additional effect on tumor growth (see ).

Our new data in Treg-specific Garp KO mice clearly establish that it is necessary to target GARP on Tregs but not sufficient to target GARP on endothelial cells to observe anti-tumor activity with anti-GARP:TGF-β1 mAbs. Notwithstanding this, we considered also the alternative strategy proposed by reviewer #1, but unfortunately had to conclude that it is not technically feasible with the currently available anti-CD25 antibodies, including with variants of the most widely used anti-CD25 clone PC-61. This clone has been used under its original format (rat IgG1, commercially available), or under a format with reduced binding to inhibitory FcgRIIb (mouse IgG2a variant constructed and described by Arce-Vargas et al, Immunity 2017; variant not commercially available). Technical limitations are the following: -When used as a rat IgG1, anti-CD25 PC-61 partially depletes Tregs in the blood and peripheral lymphoid organs, but not within tumors because of high intra-tumoral expression of inhibitory FcgRIIb (Arce-Vargas et al, 2017, Immunity). This explains why anti-CD25 PC-61 rIgG1 exerts anti-tumor activity when administered in a "
prophylactic" setting, i.e. before or soon after tumor challenge (Golgher et al., 2002;Jones et al., 2002;Onizuka et al., 1999;Quezada et al., 2008;Shimizu et al., 1999), but not when administered in a "therapeutic" setting, i.e. when mice carry established tumors (Golgher et al., 2002;Jones et al., 2002;Onizuka et al., 1999;Shimizu et al., 1999).

Yet another strategy to deplete Tregs consists in administering diphteria toxin (DT) to C57BL/6 mice carrying a human Diphteria Toxin Receptor (DTR) transgene in the Foxp3 locus (Foxp3 DTR mice). Administration of DT to MC38 tumor-bearing mice as late as 7 days after tumor cell transplantation depletes >95% of the Tregs. But again, Treg depletion in this therapeutic setting exerts very potent anti-tumor activity (100% survival), leaving no room to examine additional effects of anti-GARP:TGF-β1 on MC38 tumor growth (see our data in the figure for reviewers below).
Figure 2: what is the X axis of the middle column of graphs shown on the far left of the page? This is unnecessary, and makes it appear as though there is missing data.

As shown at the bottom left of panel b, all X axes indicate "Days" after tumor cell transplantation and all Y axes indicate "Tumor volume". X and Y axes have identical scales and tick marks in all graphs, with numbers indicated for X axes below the most bottom graph in each column, and for Y axes next to the far left graphs in each row. We do not understand what the reviewer means by "middle column of graphs on the far left of the page"… We do not understand either what exactly is "unnecessary".
Finally, there is of course no missing data, as already explained above. Could we call upon the editor for help in interpreting the reviewer comment in this particular case? We are more than willing to make our figure as clear as possible.
Another unusual way that the authors have presented data is seen Figure 4. Here they present the number of cells per mm3, whereas it is more usual to present flow cytometry data as a percentage of live cells, or as a percentage of another cellular population. How do they calculate this parameter?

DT counted after tumor collection and mechanical dissociation using an automated cell counter that discriminates live and dead cells with a dead cell marker. A sentence has been added to Material and Methods to clarify this point (lines 481-482). -total number of TILs / mm3 of tumor = total number of TILs / tumor volume. The latter is calculated from large and small tumor diameters measured with a caliper, as indicated in the Material and Methods.
We could of course also represent numbers of TILs/g. We have also calculated this parameter, and the conclusions are the same, as expected.
As TIL numbers / mm3 of tumor are calculated from flow cytometry data, it is of course also possible to represent percentages of TILs within live cells or another cell population (such as CD45 + cells). We were already citing some of these percentages in the text of the submitted manuscript.
To address the referee comment, we have revised our manuscript as follows: -We have added a new panel to revised Fig. 5 (i.e. original Fig. 4)  Here again, we observed no significant difference between the various treatment groups (Supplementary Fig. 7a-c). A statistically non-significant trend towards increased densities of T cells, apparent in both the periphery and center of tumors, was observed in mice treated with anti-PD-1, whether it was combined or not with anti-GARP:TGF-β1 ( Supplementary Fig. 7c)." On page 10, the authors highlight the difference between their Treg-mediated mechanism of blocking GARP:TGFB1 with that of an anti-TGFβ study in a mammary tumor model that exhibits spatial T cell exclusion from the tumor parenchyma, due to physical blockade by TGFB-responsive CAFs (Mariathasan et al. 2018). But their proposed mechanism of action of anti-GARP may be more similar to that described by Dodagatta-Marri et al 2019 for anti-TGFβ1-3 antibodies, i.e. blocking active TGFβ on Tregs. (Fig. 1); ii) it exerts antitumor activity when used as an Fc-dead variant, i.e. a variant able to block active TGF-β1 production but not to kill GARP-expressing cells (Fig. 2, 3, 6, 7, S2, S3); iii) it exerts antitumor activity without depleting Tregs or reducing Treg densities within tumors (Fig. 5); and finally iv) it does not exert anti-tumor activity in mice carrying a Treg-specific deletion of Garp (Fig. 7). Altogether, this shows that anti-GARP:TGF-β1 works by blocking active TGF-β1 production by Tregs. This mechanism is different from the mechanism of action proposed by Mariathasan et al, who suggest that anti-TGF-β1,2,3 blocks TGFβ signaling in stromal cells and facilitates T-cell penetration into tumors. We do not observe increased penetration of T cells within tumors in mice treated with anti-GARP:TGF-β1 (Fig. 5, S5-7). It is also different from the mechanism of action of anti-TGF-β1,2,3 proposed by Dodagatta-

Marri et al, who suggest that anti-TGF-β combined with anti-PD-1 acts by blocking TGF-β activity in Tregs and tumor cells. This suggestion is based on observations that anti-TGF-β1,2,3 reverts the increases in Treg/ Th ratios and phosphoSMAD3 in tumor cells that are induced by anti-PD-1 alone. We do not observe a reduction in Treg / Th ratio in mice treated
with anti-GARP: .
We have modified the discussion to cite the Dogatta-Marri report, in addition to the Mariathasan et al report (lines 358-366 of the revised manuscript).
Despite high GARP on endothelial cells, the author did not address the effect of anti-GARP:TGFβ1 on the vascular system. Are their morphological or functional changes in the vascular or lymphatic networks in vivo? This could be addressed for example by IHC. They mention unpublished negative data on in vitro analysis of GARP function on primary ECs, but this might not reflect the in vivo situation.
As detailed above, our new data in Treg-and platelet-specific Garp KO mice clearly establish that whereas it is necessary to target GARP on Tregs, it is not necessary to target it on platelets and not sufficient to target it on endothelial cells to observe anti-tumor activity with anti-GARP:TGF-β1 mAbs. We did not observe gross morphological changes in CT26 and MC38 tumors from mice treated with any mAb or mAb combination used in this study. Examining morphological and functional changes in the vascular and lymphatic networks in tumor-bearing mice treated with anti-GARP:TGF-β1 mAbs falls far beyond the scope of our manuscript. But we thank the reviewer for this interesting suggestion.
It is a shame that for immune profiling analysis of tumors, the authors only included 5 tumors per arm, and used the parameter of cells per mm2 (Fig 4). It seems that there could a 5 or even 10 fold increase in CD4+ and CD8+ T cells in SOME tumors treated with anti-GARP:Tgfb1 (Figure 4a). However, this effect is lost, statistically, by mixing data from responding and non-responding tumors in a population wherein only 40% of tumors respond.
We discussed above our choice to represent densities of TILs (cell numbers per mm 3 ) in revised Fig. 5. As requested by the reviewer, we now also illustrate proportions (%) of these cells in Supplementary Fig. S5, but this does not change our interpretation and conclusions. Immune profiling in Fig. 5  We have added a panel to Fig. 5 and S7 to illustrate tumor weights and volumes on day 13. These panels show that there is already a statistically significant trend towards reduced tumor weights in groups treated with anti-GARP:TGF-β1 combined with anti-PD-1 by comparison to controls (revised Fig. 5a and S7). However, no rigorous prediction can be made on the future CR, PR or NR status of the corresponding mice.  Fig. 5g), but not with other parameters. We now added more precisely that tumor weights did not inversely correlate with densities of total leukocytes or any leukocyte subset, and show selected examples of correlation analyses in Supplementary Fig. S5b. This is described on lines 247-252 of the revised manuscript.

This is why we had rather chosen to show correlations between immune profiling parameters and tumor weights in our submitted manuscript. We had concluded that tumor weights inversely correlated with proportions of anti-AH1 CD8 + T cells with multiple effector functions (revised
How do the authors explain the increase in Tregs, particularly GARP+ Tregs, after treatment with anti-GARP:TGFB1-FcD? Could this be due to increased proliferation of immature Tregs? Ki67 staining might address this. Can the authors isolate these Tregs and show that they are functionally less immunosuppressive, compared to those isolated from control-treated tumors? or at least stain use cellular markers indicative of Treg differentiation? As described on lines 206-208 of the revised manuscript, "numbers of total Tregs and GARP + Tregs per mm 3 of tumor were not decreased in mice that had received an anti-GARP:TGF-β1 mAb, alone or in combination with anti-PD-1 (Fig. 5c and Supplementary Fig. S6a). If anything, an increase in Treg and GARP + Treg numbers was observed in mice treated with the anti-GARP:TGF-β1 FcD + anti-PD-1 combination in one experiment (Fig. 5c), but this was not confirmed in two others ( Supplementary Fig. S6a)." The important point is that Tregs are not depleted.

We have not isolated Tregs to test if they were less suppressive after treatment with anti-GARP:TGF-β1 FcD in vivo because this is technically not feasible (due to low numbers of cells that can be isolated from mouse tumors).
In two places, the authors refer to "Foxp3+ non-Treg T cells" but provide no citation to this cell type.
Can the authors provide the reference or delete the statement.
In contrast to non-Treg mouse T cells, non-Treg human T cells frequently express FOXP3 in response to TCR stimulation. This is a notorious difference between human and mouse T cell biology, repeatedly reported by many groups since the discovery of FOXP3 and its role in Treg biology. We had provided one reference for this observation (reference 18), and we will not delete the statement.
Reviewer #2 (Remarks to the Author): The aim of this study was to analyze the effect of a selective blockade of TGF-β1 production by activated Tregs with antibodies against GARP:TGF-β1 in tumor models. Overall the authors found that this complexes is sufficient to induce regressions of mouse tumors otherwise resistant to anti-PD-1 immunotherapy. Mechanistically they linked this with increased effector functions of anti-tumor CD8+ T cells without increasing immune cell infiltration or depleting Tregs within tumors. Finally they found GARP-expressing Tregs in about one third of human cutaneous melanoma metastases. Thus, this study builds the basis for a novel therapy using anti-GARP:TGF-β1 mAbs to overcome resistance to PD-1/PD-L1 blockade in patients with cancer.
Overall the authors have responded well to the reviewer comments. Specifically they performed additional experiments to show that their data are robust, generalizable, and statistically significant in multiple mouse tumor models. Thus, in my opinion this study does provide intriguing preclinical rationale for targeting this pathway for the immunotherapy of human cancers.
be discussed in this mansucript. This is especially the case for point 4 raised by reviewer 1.
We thank reviewer #2 for his support and comments on the robustness, generalizability and statistical significance of our data. We are also very pleased that we can now provide data in Treg-specific Garp KO mice, as requested in point 4 by reviewer #1 in Nature Medicine. It took us almost a year to obtain robust and reproducible results in tumor experiments with these mice (revised Fig. 7). These data unequivocally support our hypothesis regarding the mechanism of action of blocking anti:GARP:TGF-β1 antibodies.

REVIEWERS' COMMENTS:
Reviewer #1 (Remarks to the Author): The authors have more than adequately addressed the reviewers' comments, and should be congratulated on an excellent study. Their inclusion of tumor studies in cell specific KO of Garp in platelets or Tregs is commendable, but unfortunately no statistics were provided, and it is unlikely that the data in Figure 7b show any statistical significance between groups. This reviewer understands the problems of reaching statistical significance in studies of this type, but to accommodate this issue, the two paragraphs in Results and Discussion relating to Figure 7b should be modified as suggested below, where underlined text are changes to the original:

Line 286 to 290
As shown in Fig. 7b, the anti-tumor efficacy of anti-GARP:TGF-β1 combined with anti-PD-1 trended towards better responses than to anti-PD-1 alone in both platelet-specific Garp KO mice (CR: 42% vs 23%) and their WT littermates (CR: 46% vs 13%), similar to observations in our previous experiments, although statistical significance was not reached. In contrast, Fig. 7c shows that anti-GARP:TGF-β1 combined with anti-PD-1 showed the opposite trajectory to that seen in platelet-specific Garp KO mice or wild type mice. Anti-GARP:TGF-β1 combined with anti-PD-1 in Treg-specific Garp KO mice trended towards inferior responses compared to anti-PD-1 alone (CR: 19% vs 28%), with the opposite trend observed in wild type litter mice (CR: 34% vs 20%).

Lines 352-353
Our experiments in cell-specific Garp KO mice (<establish>) suggest that blocking the activity of TGF-β1 emanating from GARP-expressing Tregs is required for anti-GARP:TGF-β1 to exert antitumor activity.

Line 353-354
Either delete the sentence: "Blocking the activity of TGF-β1 emanating from GARP-expressing platelets is not necessary, and blocking that from other GARP-expressing cells, such as endothelial cells, is not sufficient"; or replace it with "Whether GARP-expressing platelets or GARP-expression from other cell types contributes to tumor rejection requires more extensive studies." Reviewer #2 (Remarks to the Author): The points raised in the previous round of review have been satisfactorily addressed.