Referee 1's review (Christine Perret):

In this manuscript Quelard et al, demonstrated that a cryptic frizzled module is present in collagen 18, named FZC18, that is able to bind Wnt 3a, to inhibit the Wnt/beta-catenin signaling and decrease the tumor growth of colorectal and hepatoma cell lines. They also studied the potential physiological role of the FZC18 module in the context of hepatocellular carcinoma and provide evidence that a processed form of collagen 18 containing the FZC18 module may down-regulate the Wnt/beta-catenin signaling that is aberrantly activated in the tumor.

Altogether the results are original and significant and provide an original contribution since they identified a new matricryptin that derives from collagen 18, FZC18. This matricryptin has SFRP properties, it is able to inhibit the Wnt signaling, has tumor suppressor activity and may reduce cancer growth during liver carcinogenesis.

The experiments are in general well performed and the conclusions are justified.

The major concern with the work as it stands lies in the structure of the manuscript that is hard to understand for a non-specialist.

I would like to propose an organization od the manuscript that could improve it:

1. To begin, a schematic structure of the collagen 18 comprising the variant chains with the different non collagenous and collagenous domains together with their expected MW would help greatly the reading of the text. It is surprising to have the description of the V2/V3 forms at Figure 3 while the expression studies of these different variant are presented in Figure 1. It is also confusing to read either V3 or FZC18 mRNA throughout the text.
2. Then in vitro data showing that V3 need to be processed to have a sFRP-like activity (Figure 4 and Figure S2) and that the small FZC18 has the highest SFRP-like activity (Fig 7AC) could be presented.
3. Then, functional data showing the tumor suppressor role of FRC18 (Figure 5 and Figure 7D) in comparison to the V2 variant.
4. In vitro data showing the binding of FZC18 to Wnt ligand (Figure S4 and Figure 8)
5. Finally, the presentation of the data supporting the view that the matricryptin could have a physiological role in liver tumorigenesis (figure 2 , Figure S1, Figure 3)

Other comments
1. p7. The conclusion that FZC18 inhibits the Wnt/beta-catenin activity in the tumor microenvironment of the liver is not always correct. Indeed, in normal liver the authors show that only the V3 full-length form is present (see Figure 3). Thus, the results presented in Fig2AB show that the FZC18 full-length variant (that have in contrast to the processed forms, a positive action on the Wnt signaling, see Figure S2) is expressed in the periportal domain of the liver known to be a region where the Wnt signaling is repressed. This data could suggest that the full-length FZC18 is a negative target of the Wnt signaling and also explain the staining observed in the tumoral section present in Figure 2D-E. Results presented in Figure A-E support rather the conclusion that FZC18 is a negative target of the Wnt signaling.
However, the staining of the tumoral case presented in Figure G-H supports the view that there are in this tumor sample, processed forms of the FZC18 that inhibit the Wnt signaling in the tumor microenvironnement.
Case TL325 have probably an activating mutation of beta-catenin with a strong expression of GS, while the case TL04 reveals a lower activation of the Wnt signaling with a weaker GS staining, that could be not linked to b-catenin mutation, but perhaps to FZ overexpression and thus could be inhibited by a SFRP-like activity. Thus caution should be taken for the interpretation of the correlation data between FZC18 (full-leght and processed forms) expression data and activation of the beta-catenin signaling shown by GS staining.

2. p9. Figure S1E. It is not clear why antibody against DUF-99 will detect only V2 variant since from figure 2, it appears that the most frequent processed FZC18 variant is the 50KDa form that do contains the DUF domain. With this notion in mind it is hard to understand the staining of the tumoral section present in Figure S1D-E.

3. Why the V3FL is not secreted when overexpressed in cell line while the small VNter is secreted?

4. Figure 7. The data presented in Fig7E are surprising, inhibition by FZC18 of the TOP activity induced by an oncogenic mutant of beta-catenin. Since they are not crucial, they should be deleted.

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication the manuscript has been revised in light of these comments and to address other editorial requirements.