Reply to Liu et al.: ALK5-mediated tumor suppressor signaling through SMAD2 and SMAD3 in the uterus

Our recent publications in PNAS (1, 2) describe the phenotypic effects of loss of activin-like kinase 5; transforming growth factor β receptor 1 (ALK5;TGFBR1) or Smad2 and Smad3 in the mouse uterus and are accompanied by a commentary by Li (3). Our conclusions are that this pathway acts as a tumor suppressor pathway, findings that are also supported by a recent paper (4). We appreciate the letter to the editor by Liu et al. (5); however, their analysis is performed utilizing data from uterine cancer samples from The Cancer Genome Atlas (TCGA) and should be cautiously interpreted for the following reasons:

• i)Our analysis of TCGA expression data using interquartile range (Fig. 1) yields different results than those presented by Liu et al. (5). In our analysis, statistical significance between human endometrial cancer and gene expression was obtained only for SMAD3 and TGFBR1, but not for SMAD2. The method of analysis that we present in Fig. 1 has been published by Chandrashekar et al. (6) and is preferable because it applies the same cutoff (top 25% versus the rest) to all genes. This analysis removes some of the ambiguity that could arise in analyses with borderline statistical significance.
• ii)Liu et al. (5) use survival trends from TCGA samples as evidence of the TGF-β signaling pathway promoting tumor formation. While these correlations can be useful in hypothesis development, they require careful testing in appropriate experimental models that are representative of the specific tumor type. In fact, the TCGA Uterine Corpus Endometrial Carcinoma (UCEC) dataset is highly heterogeneous and includes various histological endometrial cancer subtypes (e.g., endometrioid and serous), menopause status, stage, and patient age (6), which do not apply to our mouse models.

Fig. 1.

TCGA gene expression analysis of ALK5, SMAD2, and SMAD3 in normal and uterine cancer tissues. (A–C) Box and whisker plots for TGFBR1/ALK5 (A), SMAD3 (B), and SMAD2 (C) showing gene expression levels in TCGA samples of UCEC patients. The samples represent RNA sequencing data corresponding to the primary tumor and normal samples. The plots indicate the minimum, 25th percentile, median, 75th percentile, and maximum values. Transcripts per million, which were used to measure expression, were also used to estimate significant difference in gene expression levels.

Regarding the theory of haploinsufficiency, we refer the authors to the text and figure S1 A and B in SI Appendix of Kriseman et al. (1), where we present details about the mouse lines used in our study. Our breeding strategy first requires us to produce double heterozygous floxed male mice carrying the progesterone receptor-cre (Pgr-cre) allele, which are crossed with double heterozygous floxed female mice. In the next round of breeding, homozygous Smad2^flox/flox;Smad3^flox/flox females are intercrossed with Smad2^flox/flox;Smad3^flox/flox male mice containing one copy of Pgr-cre (i.e., Pgr^cre/+) to produce the experimental mice. All experimental female mice were verified to contain a genotype with double floxed Smad2 and Smad3 alleles with one copy of Pgr-cre (Smad2^flox/flox;Smad3^flox/flox-Pgr^cre/+) and to display significantly decreased Smad2 and Smad3 expression in the uterus [figure S1C in SI Appendix of Kriseman et al. (1)]. Therefore, haploinsufficiency is ruled out because of the setup of our genetic crosses and is supported by several published mouse models using the Pgr-cre deleter strain to conditionally delete genes that are also on chromosome 9 (7–9).