Tristetraprolin inhibits macrophage IL-27-induced activation of antitumour cytotoxic T cell responses

IFN-γ-producing cytotoxic T lymphocytes are essential for host defense against viral infection and cancer. Here we show that the RNA-binding tristetraprolin, encoded by Zfp36, is needed for CD8+ T-cell production of IFN-γ in vivo. When activated in vitro, however, IFN-γ production by naive wild type and tristetraprolin-deficient CD8+ T-cells is comparable. IL-27 is overproduced by tristetraprolin-deficient macrophages and increased systemically in tristetraprolin-deficient mice. Tristetraprolin suppresses IL-27 production by promoting p28 mRNA degradation. Importantly, deletion of IL-27 receptor WSX-1 in tristetraprolin-deficient mice (WSX-1/tristetraprolin double knockout) leads to a reduction in cytotoxic T lymphocyte numbers. Moreover, tumor growth is accelerated, not only in tristetraprolin-deficient mice after cytotoxic T lymphocyte depletion, but also in WSX-1/tristetraprolin double knockout mice, with substantial reduction in the number of tumor cytotoxic T lymphocytes. This study describes a regulatory pathway for IL-27 expression and cytotoxic T lymphocyte function mediated by tristetraprolin, contributing to regulation of antitumour immunity.

Major Points 1. The authors refer to "naïve" T cells throughout the manuscript but they show in Fig 1 that they have quite a large population of CD69+ CD8 T cells that are already present in the KO. Since it is not clear from the description of the experiments if they truly obtained naïve (not antigen experienced cells) or used bulk CD8 populations they need to either redo many of the experiments using truly naïve cells OR they need to modify how the describe an interpret the data presented. For example, is the increased CFSE dilution they see in Fig 1 just because they start with a larger population of antigen experienced cells? This makes it difficult to interpret the data as presented.
2. At many points the authors show representative flow plots -but this is not accompanied by any further analysis that shows statistics or collated data sets. This needs to be addressed in Figs 1, 2 and 5.
3. Fig 2c/d the IFN-g production seems very, very low for a CD8 T cell population. Is this still true if they use more polarizing conditions that might include IL-12? 4. Figure 5A needs the WT control supes combined with the WSX-1KO CD8. Fug 5B seems to have a major error in labeling. Its not clear what they are gating on. The authors do note that these supes still enhance IFN-g production even in the absence of WSX-1. Since they do have an idea of what other pro-inflammatory cytokines are over produced in the cultures -can they show the impact of neutralization of TNF or IL-23 for example that have also been shown to promote IFN-g production? In other words IL-27 alone is not sufficient for this phenotype but works with other cytokines.
5. Fig 6. Is the increased resistance to the tumor cell challenge of the Zfp36 KO dependent on CD8 and/or IFN-g? This is a key point if the authors want to close the loop showing that the 27dependent effect is due to CD8. This can be tested using depleting Abs.
Introduction -the authors might consider modifying the sentence that notes that the mechanisms that regulate the development and function of CTLS remain elusive. I think the literature is quite rich in examples of mechanisms that regulate the development and function of CTLS Reviewer #3 (Remarks to the Author) Very little is known about the role of zfp36 in lymphocytes. Unfortunately the work presented here does little to enlighten us. The links to IL27 are novel but unsurprising. In general, the presentation of the data is extremely poor with little or no consideration given to biological variation or proper quantitation. Some statistical tests are used but it is not clear which test is being applied to what data.
The data in Suppl Fig 1a showing CD4 and CD8 staining of the thymus is of very poor quality. Better quality data must be shown. The age and sex of the mice is an important consideration when interpreting this data and this information should be clearly provided in the figure legend. In general showing percentages is unhelpful unless we are also give the total numbers of cells as well.
Figure 1 A would be better with dot plots. To conclude anything about surface antigen expression we need to see more than the Flow cytometry. Overlay histograms can help visualize the point better. The median fluorescence intensity must be quantitated from a number of different mice and then compared. Here information on the numbers of cells in each population is essential to properly describe the phenotype.
The data leading to claims about differences in the production of cytokines etc need to be similarly analysed. Moreover if the ratio of naïve to effector cells is change then the results are due to a population effect and say nothing interesting about the role of zfp36 in regulating T cell cytokines.
The isolation and activation of T cells for three days could show and effect on proliferation. To be sure of this we would need to know that the initial starting populations put into culture were equivalent. As the authors show that the population of effector cells is different in proportion this seems to be the likely explanation. No information is given on biological variation within experimental groups in this assay.
The same problems of initial different population effect and lack of biological variability exists with Figure 1D.
Figure 2 reinforces some of these criticisms with the authors own data. The lack of variability persists. Moreover the conclusion that there is no effect on cytokine mRNA or protein is premature as it is based on only a single stimulation and timepoint. If the data shown is from a representative experiment what is n ? For the statistics why not pool the data from the three experiments with similar results? The results in Figure 4 are convincing but they are unsurprising. Any ARE containing transcript will give similar results in such as system. Figure 5 also needs to show more than representative flow cytometry. The effects claimed need to be properly quantitated. Roles for IL27 in T cells reported by others should be discussed in this context (https://www.ncbi.nlm.nih.gov/pubmed/25267651). Data in figure 6 i?s difficult to interpret. For example what is being shown in 6C is far from clearwe need to be clear on how many mice how many tumors are. 6D need quantitation of the effect. 6 E-F It would be better to the makeup of cellular infiltrate in absuolute numbers. What are the statistics here? 5a). Since there are still enhanced IFN-γ production by TTP -/supernatants even in the absence of WSX-1, as suggested by the reviewer, other pro-inflammatory cytokines, such as TNF-α and IL-23, may promote IFN-γ production in addition of IL-27. To answer this question, we performed new experiments by treating the TTP -/supernatants with TNF-α and IL-23 neutralizing antibody prior to culturing it with WT naïve CD8 T cells. Neutralizing TNF-α and IL-23 reduced the IFN-γ production in WT CD8 T cells induced by the TTP -/supernatants, with a little stronger effects of TNF-α blockade than IL-23 blockade (new Fig. 5b). Though neutralizing TNF-α and IL-23 reduced IFN-γ production, there still remained a large amount of IFN-γ production compared with it in WSX-1 -/-CD8 T cells culturing with the TTP -/supernatants, suggesting that IL-27 plays a major role, while TNF-a and IL-23 play a minor role, in inducing IFN-γ production in CD8 T cells mediated by TTP -/supernatants. It also suggests that IL-27 can induce IFN-γ production in CD8 T cells; other cytokines (such as TNF-α and IL-23) may work with IL-27 to further enhance IFN-γ production.
The data shown in the previous Fig. 5B was gated on CD4 negative T cells. In this revision, we replaced it with new data by gating directly on CD8 + T cells (now shown as Fig. 5c).

Fig 6. Is the increased resistance to the tumor cell challenge of the Zfp36 KO dependent on CD8 and/or IFN-g? This is a key point if the authors want to close the loop showing that the 27-dependent effect is due to
CD8. This can be tested using depleting Abs. Response: We thank the reviewer for the suggestion on using depletion Ab. We performed new animal experiments using the antibody for CD8 T cell depletion (InVivoMAb anti-mouse CD8α Clone: 53-6.72 purchased from Bio X cell). The CD8 T cell depleting antibody was administrated five times in total (on day 0, 4, 8, 12 and 15) after tumor cell inoculation. In consistent with the results in Fig. 6a, Zfp36 -/mice showed retarded tumor growth compared with WT control mice (new Fig. 6j,k). Importantly, depletion of CD8 T cells in Zfp36 -/mice resulted in acceleration of tumor growth, even a little faster than it in WT mice with CD8 depletion (new Fig. 6j,k). The efficiency of CD8 T cell depletion was checked on day 18 after tumor cell inoculation. The percentages of CD8 T cells were significantly reduced by the CD8 depleting antibody (new Supplementary Fig. 4b). These data close the loop showing that IL-27-dependent antitumor effect is mediated by CD8 T cells in Zfp36 -/mice.

The authors are correct that little is known about what limits IL-27 production. This itself raises the question of what regulates TTP and one obvious candidate might be IL-10. So, does IL-10 limit IL-27 or upregulate TTP? At this point, if IL-10 does not impact this pathway that is fine -but if it does, it really closes the loop on this study.
Response: We performed new experiments by stimulating WT macrophages with LPS, IL-10 or both, then measured p28 and EBI3 mRNA expression. As reported previously, IL-10 inhibited the p28 (new Supplementary Fig. 5a) but not EBI3 (new Supplementary Fig. 5b) mRNA expression induced by LPS. In the same setting, LPS strongly induced TTP expression which is in line with our previous report (new Supplementary Fig. 5c). While IL-10 induced minimal level of TTP expression by itself, it inhibited LPSinduced TTP expression (new Supplementary Fig. 5c). IL-10 does not upregulate TTP in activated macrophages, suggests that IL-10-mediated suppression of IL-27 is independent of TTP. Moreover, we treated WT and Zfp36 -/-BMDMs with LPS in the presence or absence of IL-10 and then measured p28 mRNA expression. As shown in Supplementary Fig. 5d, IL-10 inhibited p28 expression in both BMDMs, indicating that IL-10 does not impact this pathway.

The Hunter JI paper on the IL-27 transgenics shows a massive activation of CD8 T cells while the Kedl PNAS paper is a good non-cancer model that illustrates the impact of endogenous IL-27 on CD8 T cell responses.
Citing these manuscripts would further strengthen/broaden the overall message of the study. Response: We are sorry for missing the two important papers related to the impact of IL-27 on CD8 T cells. The two papers are cited in the revised manuscript.

8.
The authors need to provide additional information on the ELISA kits being used. Do they detect IL-27p28 or the heterodimer. If its only p28 then the graphs need to reflect this. Response: The ELISA kit used in this study is to measure the heterodimer. We added more detailed information about the ELISA kits in this revision (mouse IL-27 ELISA kit. Catalog#: 88-7274, eBioscience).

Introduction -the authors might consider modifying the sentence that notes that the mechanisms that regulate the development and function of CTLS remain elusive. I think the literature is quite rich in examples of mechanisms that regulate the development and function of CTLS.
Response: We sincerely appreciate the advice and have modified the sentence in the text.

Reviewer #3
1. Very little is known about the role of zfp36 in lymphocytes. Unfortunately the work presented here does little to enlighten us. The links to IL27 are novel but unsurprising. In general, the presentation of the data is extremely poor with little or no consideration given to biological variation or proper quantitation. Some statistical tests are used but it is not clear which test is being applied to what data. Response: We are sorry for the confusion and sincerely appreciate the comments on biological variation. We have performed statistical analyses on all data in this revised manuscript. The methods of statistical analysis, variation and sample size are now included in figure legends. Suppl Fig 1a showing CD4 and CD8 staining of the thymus is of very poor quality. Better quality data must be shown. The age and sex of the mice is an important consideration when interpreting this data and this information should be clearly provided in the figure legend. In general showing percentages is unhelpful unless we are also give the total numbers of cells as well. Response: We performed new staining of the CD4 and CD8 T cells in thymus and obtained similar results. Both the percentages and total numbers of CD4 and CD8 T cells in thymus are comparable between WT and Zfp36 -/mice (new Supplementary Fig. 1a,b). The WT and Zfp36 -/mice used in the study are female aged 6-8 weeks old. This information is added in the figure legend. Figure 1 A would be better with dot plots. To conclude anything about surface antigen expression we need to see more than the Flow cytometry. Overlay histograms can help visualize the point better. The median fluorescence intensity must be quantitated from a number of different mice and then compared. Here information on the numbers of cells in each population is essential to properly describe the phenotype. Response: We thank the reviewer for the invaluable advice. In this revision, we provide the flow data with dot plots (new Fig. 1a) and with histogram (new Supplementary Fig. 1d). The median fluorescence intensities from three different mice are quantitated and shown in new Supplementary Fig. 1d. The numbers of CD8 T cells in each population are also quantitated and shown as new Supplementary Fig. 1c. All data support the conclusion that there are more effector CD8 T cells in Zfp36 -/mice compared to WT mice.

Response:
We have re-analyzed the difference in cytokine production and presented the quantitative results in new Fig. 1c. Zfp36 -/-CD8 T cells produced more IFN-γ and TNF-α, but similar levels of IL-2, granzyme B and perforin compared with WT CD8 T cells (Fig. 1b,c). As the reviewer pointed out, the effector CD8 T cells are increased in Zfp36 -/mice (Fig. 1a-c and Supplementary Fig. 1c,d). To exclude the inference of effector cells, we used the purified naïve CD8 T cells in the subsequent experiments, and found that there was no intrinsic defect in producing IFN-γ by Zfp36 -/-CD8 T cells ex vivo (Fig. 2a-d and Supplementary Fig.  2a-d). The increased CTL in Zfp36 -/mice is mediated by IL-27 as confirmed with the double knockout mice (Fig. 5a-e).

The isolation and activation of T cells for three days could show and effect on proliferation. To be sure of this we would need to know that the initial starting populations put into culture were equivalent. As the authors show that the population of effector cells is different in proportion this seems to be the likely explanation. No information is given on biological variation within experimental groups in this assay.
Response: We are sorry for the confusion. Since there are more effector CD8 T cells in Zfp36 -/mice, for fair comparison we isolated naïve CD8 T cells from spleens of WT and Zfp36 -/mice, used same number of naïve CD8 T cells for ex vivo stimulation with plate-coated anti-CD3/CD28 Abs, and then performed CFSE dilution assay and Ki67 detection on day three. The levels of CFSE and Ki67 at the beginning of culture (0h) are now included in new Fig. 1d. The results of statistical analysis are also included in new Fig. 1d as well. Figure 1D. Response: Same number of naïve CD8 T cells isolated from WT and Zfp36 -/mice was used in the previous Figure 1D. The levels of PI + and Annexin-v + CD8 T cells at the beginning of culture (0h) and the results of statistical analysis are shown in new Fig. 1e. The representative dot plots are shown as new Supplementary  Fig. 1e. Figure 2 reinforces some of these criticisms with the authors own data. The lack of variability persists. Moreover the conclusion that there is no effect on cytokine mRNA or protein is premature as it is based on only a single stimulation and time point. Fig. 2a. The results of statistical analyses are shown in new Fig. 2b,d. To further confirm the effects of TTP on cytokine production, we performed new experiments by stimulating the naïve CD8 T cells with PMA/Ionomycin and IL-12, in addition of TCR activation. The production of IFN-γ was similar between WT and Zfp36 -/-CD8 T cells in response to different stimuli at different time points (new Supplementary Fig. 2b,c). These data support the conclusion that CD8 T cells deficient of TTP have no intrinsic defect in IFN-γ production.

If the data shown is from a representative experiment what is n ? For the statistics why not pool the data from the three experiments with similar results?
Response: We pooled the data from three experiments (n=3) and performed statistical analysis. The quantitative results are now included in new Fig. 2b,d, and in Supplementary Fig. 2d.

Fig 3 A-C is uninterpretable as we are not told what N is.
Response: A detailed description including the N and statics are now included in figure legend. Figure 4 are convincing but they are unsurprising. Any ARE containing transcript will give similar results in such as system. Response: Though TTP is known to bind to ARE in the 3'UTR, our data demonstrate that not all AREs are used by TTP to exert its function. TTP promotes p28 mRNA decay through several AREs but not ARE2. This inhibitory effect is mediated by the two zinc fingers in TTP protein. To the best of our knowledge, this is the first work showing how IL-27 is down-regulated by RNA-binding protein TTP. Figure 5 also needs to show more than representative flow cytometry. The effects claimed need to be properly quantitated. Roles for IL27 in T cells reported by others should be discussed in this context (https://www.ncbi.nlm.nih.gov/pubmed/25267651). Response: Quantitative results of the flow cytometry data are now included in Fig. 5. We are sorry for missing the important paper and have cited it in the revised manuscript. figure 6 i?s difficult to interpret. For example what is being shown in 6C is far from clear-we need to be clear on how many mice how many tumors are. 6D need quantitation of the effect. 6 E-F It would be better to the makeup of cellular infiltrate in absuolute numbers. What are the statistics here?

Data in
Response: The data shown in Fig. 6c is the quantitative results of the images shown in Fig. 6b, which includes four mice and four tumors in each group. To better visualize the infiltrates, we stained CD8 T cells on different tumor sections and quantitated the number of tumor-infiltrating CD8 T cells in each group. More infiltrating CD8 T cells were observed per field in tumors of Zfp36 -/mice compared with tumors of other groups; while the increased CD8 T cells were reduced significantly in tumors of the DKO mice (new Fig.  6d,e). Because the tumor mass (sizes) are markedly different among four groups (Fig. 6a-c), for fair comparison we normalized the infiltrates against tumor mass. The percentages and number of CD8 T cells in tumors are quantitated and shown in Fig. 6f-i. One-way ANOVA (and Nonparametric) with Tukey was used to analyze the difference among four groups. ***: p<0.001; **: p<0.01;*: p<0.05 between indicated groups. In addition, we performed new in vivo experiments using CD8 depletion antibody. In consistent with the results shown in Fig. 6a, Zfp36 -/mice showed retarded tumor growth compared with WT control mice (new Fig. 6j,k). Importantly, depletion of CD8 T cells in Zfp36 -/mice resulted in acceleration of tumor growth (new Fig. 6j,k). Taken together, these data confirm that IL-27-dependent antitumor effect is mediated by CD8 T cells in Zfp36 -/mice.
The authors have made several revisions to the manuscript that help close the loop (i.e the anti-CD8 experiment on Fig 6) and have added in many additional experimental details that help clarify how the experiments are performed. These clarifications themselves are useful but have raised a couple of questions that need to be considered -specifically about cell intrinsic vs extrinsic effects ofTTP that just need to be thought through. In addition, there remain several gaps between the last reviews and the revision that also need to be addressed.
1. Figure 1D shows that purified naive CD8 T cells from Zfp36KO mice have an increased ability to proliferate. This implies that there is a T cell intrinsic role for this transcription factor on the regulation of T cell proliferation and survival. This stands in contrast to the author's conclusion that TTP modulates CD8 T cell activation in a T-cell extrinsic fashion. This looks like a major effect and (as noted by reviewer 2) there are concerns about the data previously presented and the authors need to provide data on the numbers of T cells not just percentages in the later figures. This was requested but has not been addressed adequately.
2. Figure  3. Figure 5. The authors show that supernatants from KO macrophages promote T cell production of IFN-g. There was a previous suggestion that they neutralize over cytokines that might affect this process and they have neutralized TNF and IL-23 and see a modest effect. The details of the Abs are not apparent in the materials and methods and it seems like it is critical to know if the authors are blocking IL-23 p19 or the p40 subunit shared with IL-12. IL-12 seems like the most likely candidate for a macrophage derived cytokine that promotes IFN-g and it may or may not act in concerns with IL-27.

Reviewer #3 (Remarks to the Author)
The revised manuscript attempts to clarify what role TTP has in the cellular immune response to tumors and concludes and major role for the IL27 axis but no intrinsic role in CD8T cell activation. The enhanced anti-tumor immunity of TTP KO mice is not evident when IL27 signaling is abolished in WSX-1 mutant mice, or upon depletion of CTL with anti-CD8 antibodies. I think the conclusion that IL27 overproduction by TTP KO macrophages affects CD8 T cells is likely to be correct and that loss of IL-27 signaling or depletion of CD8 T cells will diminish anti-tumor immunity. I would have bee surprised if either of these manipulations did not diminish anti-tumor immunity. I think the conclusion that TTP has no role in CD8 T cells is likely to be incorrect and I ask the authors to carefully reconsider this conclusion. The conclusion stated on line 132 that "TTP modulates CD8 T cell activation in a T cell extrinsic fashion." is not supported by the authors data. The authors own results (data in Fig1 d and e) contradict this conclusion. It would appear the authros base their conclusion on the observation that IFNg is not different at the time tested, but to extrapolate from this specific instance to such a strong general conclusion is unwarranted. From what is shown in Figure 1a it is uninformative to state that CD8 cells express higher levels of activation markers because the populations being compared are not the same. Separation of naive and memory subsets is required then the median fluorescence intensity should be compared. I do not think that this is what has been done in supplementary Fig1d. Showing cell % is uninformative and misleading and could well be removed. Showing activation marker expression on naïve and memory cells could be helpful if there is a difference and the biological significance of the difference were stated. In Fig1 b the differential cytokine production form a mixed population is not very informative reflecting again the comparison being between mixtures of cells that differ in the proportions of naïve and activated memory cells. The proliferation data is of interest but the analysis is rather superficial. It is possible to determine cell number at each division and the display of % CFSE low tells us nothing quantitatively about the number of divisions that appears from the figure to be different. This is potentially interesting but is not developed well enough for publication. It isn't clear whether this has any in vivo relevance to the tumor immunity results. What is the justification to use quadrants to gate in figure 2a and how are positive and negative thresholds for setting gates in the flow cytometry analysis determined? In 2b some attempt at quantitation is made but it looks like there is no effect. The text accompanying this figure suggests there may be a decrease. This could be more precisely described.
The new data on IFNg mRNA are interesting but this study is still incomplete. Why not measure other cytokines such as TNF? Also the accumulation of cytokines by ELISA over a timecourse? In supplemental figure 2d the nature of the cell mixture is unclear and the means of stimulation so strong, and limited to a single timepoint, that it is hard to be convinced that IFNg production by CD8+ T cells under physiological conditions of T cell stimulation is TTP independent.
The addition of supernatants from activated macrophages to the CD8T cells and the use of receptors deficient cells does indicate IL27 plays a role but the point is somewhat labored and other factors are also active in those sups.
From 6a I am not convinced that it is correct to say tumors grow more quickly in WSX or DKO mice. For the first 20 days no clear difference is evident. Are there other data that back this conclusion up?
Is the % of DKO CD8 in figure 6h less than in the zfp36 single KO?
The identification of IL-27 as a TTP target is not novel and has been found in the major analysis of TTP targets in macrophages from Kovarik's group. This can be found on line in their web tool that navigates through the data. This needs to be acknowledged.
http://ttp-atlas.univie.ac.at/ There are many spelling and grammatical errors throughout the manuscript that will need to be corrected.