Reply to “An alternative miRISC targets a cancer‐associated coding sequence mutation in FOXL2”

Abstract

To the Editor,

In “An alternative miRISC targets a cancer‐associated coding sequence mutation in FOXL2”, published by Shin et al (2020), the authors propose that the somatic missense variant (c.402C>G) affecting the gene encoding the transcription factor FOXL2 creates a target site for miR‐1236, inducing its somatic haploinsufficiency.

The authors claim that the levels of transcripts carrying c.402C>G mutated (MUT) alleles were lower than for wild type (WT) in adult‐type granulosa cell tumors (AGCTs) and an AGCT‐derived cell line (KGN (Nishi et al, 2001)). Shin et al used three methods for quantifying the allelic ratio including allele‐specific qPCR, pyrosequencing, and deep sequencing. Allele‐specific PCR, prone to biases due to different efficiency of primer annealing and/or extension (Howard et al, 1999), yielded the strongest imbalance while pyrosequencing, which produces less biased results, showed a lower imbalance and finally deep sequencing yielded 61.9% of WT and 38.1% for the MUT allele. Although the authors state that this difference is statistically significant, its biological relevance is, in our view, unconvincing. Tumor cell purity was not reported by Shin et al; the presence of non‐tumor cells can create a bias toward the WT allele.

Two recent studies suggest that the C134W variant leads to a gain of function through altering FOXL2 DNA‐binding specificity (Carles et al, 2020), probably by fostering a strong interaction with SMAD4 (Weis‐Banke et al, 2020). It is still possible that FOXL2 is a tumor suppressor where somatic haploinsufficiency occurs (Benayoun et al, 2010). However, previous RNAseq data on four tumors (Shah et al, 2009) show no evidence of allelic imbalance (figure 2B of Shah et al, 2009). We have sequenced the cDNA and matching gDNA of KGN cells, which displays no allelic imbalance (Fig 1). Finally, with the proviso that immunohistochemistry is not quantitative, previous results on heterozygous and homo/hemizygous AGCTs and normal tissue show roughly similar expression densities (figure 4C, D, E; Shah et al, 2009).

Figure 1. Sanger electropherograms of the gDNA and cDNA of FOXL2 in KGN cells.

No allelic imbalance is apparent (C in blue and G in black are basically superposed, the arrowhead points to the mutated position). Source data are available online for this figure.

Shin et al (2020) propose that miR‐1236 targets the c.402C>G FOXL2 mRNA, enhancing its degradation. This is supported by a series of in vitro assays. Overexpressed miR‐1236 can indeed target the mutated FOXL2 allele; however, whether this has an in vivo relevance given the endogenous levels of miR‐1236 is, in our view, unclear. The negative correlation between the levels of variant mRNA (pyrosequencing) and of miR‐1236 is consistent with some degree of inhibition. It would have been interesting to test whether this correlation stands when using variant allele frequency data from deep sequencing. However, there is no clear correspondence in the paper between the numbering of the patients analyzed by pyrosequencing and by deep sequencing. Regarding pyrosequencing, the estimation of the proportions is often hard to reconcile with the pyrograms (e.g., Patients 6 and 10 in Shin’s fig S1B).

In sum, the authors aimed at showing that the “mutation (c.402C>G; p.C134W) in FOXL2,..., introduces a target site for miR‐1236, which causes haploinsufficiency of the tumor‐suppressor FOXL2”. We, however, conclude that this claim is unsupported by their and our findings, and that their study suggests at most, that an miRNA can modulate mutated FOXL2 expression.

Materials and Methods

The RNA from KGN cells was treated with DNAse before cDNA synthesis (performed using SuperScript II [Invitrogen] and random hexamers). FOXL2 was amplified from cDNA or gDNA with primers F1 5′ atcagaacagagcgaggctcc 3′ and primer R 5′ cacgagttgttgaggaagcca 3′ using HiFi Pfx from Invitrogen (35 cycles). PCR fragments were purified (NucleoSpin Gel and PCR Clean‑up, Macherey‐Nagel) and sequenced with primer 5′ cgtacgtggcgctcatcg 3′.

Supporting information

Source Data for Figure 1

The EMBO Journal (2021) 40: e107517.