Super-enhancer-driven ZFP36L1 promotes PD-L1 expression in infiltrative gastric cancer

Reviewer #1 (Public Review):

In this paper, the authors provide data to support the existence of a regulatory pathway starting with SPI1-driven ZFP36L1 expression, that goes on to downregulate HDAC3 expression at the transcript level, leading to PD-L1 upregulation due to implied enhanced acetylation of its promoter region. This is therefore an interesting pathway that adds to our understanding of how PD-L1 expression is controlled in gastric cancer. However, this is likely one of many possible pathways that impact PD-L1 expression, which is likely equally important. Thus, while potentially interesting, this is more additive information to the literature rather than a fundamentally new concept/finding.

Overall, there are many experiments presented, which appear to be of good quality, however, there are a number of issues with this that need attention. Moreover, the text is often difficult to follow, partly due to the standard of English, but mainly due to the sparsity of detail in the results section and figure legends. Thus providing an overall assessment of data conclusiveness is not possible at this time. This is exacerbated by frequently extrapolating conclusions beyond what is actually shown in an individual experiment.

Major issues:

(1) All the figure legends need to expand significantly, so it is clear what is being presented. All experiments showing data quantification need the numbers of independent biological replicates to be added, plus an indication of what the P-values are associated with the asterisks (and the tests used).

(2) Related to point 1, the description of the data in the text needs to expand significantly, so the figure panels are interpretable. Examples are given below but this is not an exhaustive list.

(3) The addition of "super-enhancer-driven" to the title is a distraction. This is the starting point but the finding is portrayed by the last part of the title. Moreover, it is not clear why this is a super enhancer rather than just a typical enhancer as only one seems to be relevant and functional. I suggest avoiding this term after initial characterisations.

(4) The descriptions of Figures 1B, C, and D are very poor. How for example do you go from nearly 2000 SE peaks to a couple of hundred target genes? What are the other 90% doing? What is the definition of a target gene? This whole start section needs a complete overhaul to make it understandable and this is important as is what leads us to ZFP36L1 in the first place.

(5) It is impossible to work out what Figures 1F, H, and I are from the accompanying text. The same applies to supplementary Figure S1D. Figure 1G is not described in the results.

(6) What is Figure 2A? There is no axis label or description.

(7) Why is CD274 discussed in the text from Figure 2E but none of the other genes? The rationale needs expanding.

(8) Figure 2G needs zooming in more over the putative SE region and the two enhancers labelling. This looks very strange at the moment and does not show typical peak shapes for histone acetylation at enhancers.

(9) The use of JQ1 does not prove something is a super enhancer, just that it is BRD4 regulated and might be a typical enhancer.

(10) An explanation of how the motifs were identified in E1 is needed. Enrichment over what? Were they purposefully looking for multiple motifs per enhancer? Otherwise what it all comes down to later in the figure is a single motif, and how can that be "enriched"?

(11) A major missing experiment is to deplete rather than over-express SPI1 for the various assays in Figure 4.

(12) The authors start jumping around cell lines, sometimes with little justification. Why is MGC803 used in Figure 4I rather than MKN45? This might be due to more endogenous SPI1. However, this does not make sense in Figure 5M, where ZFP36L is overexpressed in this line rather than MKN45. If SPI1 is already high in MGC803, then the prediction is that ZFP36L1 should already be high. Is this the case?

(13) In Figure 5, HDAC3 should also be depleted to show opposite effects to over-expression (as the latter could be artefactual). Also, direct involvement should be proven by ChIP.

(14) Figure 5G and H are not discussed in the text.

(15) Figure 6C needs explaining. Why are three patients selected here? Are these supposed to be illustrative of the whole cohort? What sub-type of GC are these?

(16) In Figure 6E onwards, they switch to MFC cell line. They provide a rationale but the key regulatory axis should be sown to also be operational in these cells to use this as a model system.

Reviewer #2 (Public Review):

Summary:

This manuscript by Wei et al studies the role of ZFP36L1, an RNA-binding protein, in promoting PD-L1 expression in gastric cancer (GC). They used human gastric cancer tissues from six patients and performed H3K27ac CUT&Tag to unbiasedly identify SE specific for the infiltrative type. They identified an SE driving the expression of ZFP36L1 and immune evasion through upregulation of PD-L1. Mechanistically, they show that SPI1 binds to ZFP36L1-SE and ZFP36L1 in turn regulates PD-L1 expression through modulation of the 3'UTR of HDAC3. This mechanism of PD-L1 regulation in gastric cancer is novel, and ZFP36L1 has not been previously implicated in GC progression. However, the data presented are largely correlations and no direct proof is presented that the identified SE regulates ZFP36L1 expression. Furthermore, the effect of ZFP36L1 manipulation elicited a modest effect on PDL1 expression. In fact, several cell lines (XGC1, MNK45) express abundant ZFP36L1 but no PD-L1, suggesting the ZFP36L1 per se is not a key stimulant of PD-L1 expression as IFNg is. Thus, the central conclusions are not supported by the data.

Strengths:

Use of human GC specimens to identify SE regulating PD-L1 expression and immune evasion.

Weaknesses:

Major comments:

(1) The difference in H3K27ac over the ZFP36L1 locus and SE between the expanding and infiltrative GC is marginal (Figure 2G). Although the authors establish that ZFP36L1 is upregulated in GC, particularly in the infiltrative subtype, no direct proof is provided that the identified SE is the source of this observation. CRISPR-Cas9 should be employed to delete the identified SE to prove that it is causatively linked to the expression of ZFP36L1.

(2) In Figure 3C the impact of shZFP36L1 on PD-L1 expression is marginal and it is observed in the context of IFNg stimulation. Moreover, in XGC-1 cell line the shZFP36L1 failed to knock down protein expression thus the small decrease in PD-L1 level is likely independent of ZFP36L1. The same is the case in Figure 3D where forced expression of ZFP36L1 does not upregulate the expression of PDL1 and even in the context of IFNg stimulation the effect is marginal.

(3) In Figure 4, it is unclear why ELF1 and E2F1 that bind ZFP36L1-SE do not upregulate its expression and only SPI1 does. In Figure 4D the impact of SPI overexpression on ZFP36L1 in MKN45 cells is marginal. Likewise, the forced expression of SPI did not upregulate PD-L1 which contradicts the model. Only in the context of IFNg PD-L1 is expressed suggesting that whatever role, if any, ZFP36L1-SPI1 axis plays is secondary.

(4) The data presented in Figure 6 are not convincing. First, there is no difference in the tumor growth (Figure 6E). IHC in Figure 6I for CD8a is misleading. Can the authors provide insets to point CD8a cells? This figure also needs quantification and review from a pathologist.