Response to Heard et al

Reply to: KN Heard et al (in this issue)

Heard et al (2015) generated cIap1^−/− Xiap^−/− mice and were surprised to find them to be viable and fertile, because we had reported (Moulin et al, 2012) that cIap1^−/−Xiap^−/− mice died by day E12.5 of embryogenesis (Moulin et al, 2012 Figs 1B and 2B and Supplementary Fig S1A). We are working with Heard et al (2015) in an attempt to determine why.

It is, however, clear that failure of our cIap2^FRT/FRTcIap1^−/−Xiap^−/− mice to survive past E12.5 is not due to non-functional cIap2^−/− genes. Three types of cross indicate that the cIap2^FRT/FRT locus, which is carried by our cIap1^−/− mice, does produce functional cIAP2. Firstly, comparison of the phenotypes of the cIap2^FRT/FRTcIap1^−/− mice, which are viable and fertile, with the cIap2^−/−cIap1^−/− mice, which die at E12.5, indicates that the cIap2^FRT/FRT locus can function, at least to the extent needed to allow normal development when Xiap is present (Moulin et al, 2012). Secondly, when specific deletion of cIap1 in B cells was combined with whole body cIap2 deletion, it led to more profound B-cell expansion than deletion of either IAP alone (Gardam et al, 2011). Thirdly, deletion of cIap1 in myeloid cells on either a cIap2^−/− or cIap2^−/− Xiap^−/− background triggered splenomegaly, increased neutrophils and monocytes, inflammatory cytokine production and spontaneous inflammatory arthritis, whereas deletion of Xiap, cIap1 or cIap2 alone did not (Wong et al, 2014; Lawlor et al, 2015).

In addition to the differences in viability of the cIap1^−/−Xiap^−/− mice, Heard et al (2015) found much higher levels of cIAP2 protein in their cIap1^−/− mouse embryonic fibroblasts (MEFs) than we reported in our cIap1^−/− MEFs. Furthermore, Heard et al (2015) confirmed this difference: when they directly compared our cIap1^−/− MEFs with their cIap1^−/− MEFs, they saw that ours had very low to undetectable levels of cIAP2 protein (like wild-type MEFs), whereas theirs had much higher levels of cIAP2 (Fig 1F, compare lanes 1, 2 and 4).

Consistent with their finding that levels of cIAP2 rise in the absence of cIAP1 in MEFs, they also found elevated levels of cIAP2 protein in several tissues of cIap1^−/−Xiap^−/− mice.

Although we did not observe elevated cIAP2 in our cIap2^FRT/FRTcIap1^−/− MEFs, their finding of increased cIAP2 in their cIap1^−/− MEFs is consistent with data from several laboratories (including our own) showing that absence or depletion of cIAP1 leads to activation of non-canonical NF-κB and cIAP2 up-regulation (Varfolomeev et al, 2007; Vince et al, 2007; Darding et al, 2011). Indeed, as Heard et al (2015) show, our cIap2^FRT/FRTcIap1^−/− MEFs have elevated cIap2 mRNA expression when compared with their cIap1^loxP/loxP cIap2^FRT/FRT, cIAP1-proficient counterparts. This indicates a potential defect in translation or stability of the cIAP2 protein in our MEFs.

Note, however, that in our hands, immortalised MEFs are highly genetically variable, with a tendency to lose the expression of proteins, often seemingly at random (Cook et al, 2014). Thus, it remains possible that the particular line of immortalised MEFs that we shared with Heard et al (2015) are not truly representative of the situation elsewhere in the mice.

Why might MEFs derived from our cIap2^FRT/FRTcIap1^−/− and cIap2^FRT/FRTcIap1^−/− Xiap^−/− mice have much lower levels of cIAP2 than the MEFs from their cIap1^−/− and cIap1^−/−Xiap^−/− mice? If the differences in cIAP2 levels in the MEFs are reflected in vivo, one reason their cIap1^−/−Xiap^−/− mice are viable, whereas our cIap1^−/−Xiap^−/− mice die in mid-embryogenesis, might be differing levels of cIAP2 present during embryogenesis. In a number of molecular pathways minimum threshold levels of protein are required for normal development. As we have only observed one morphological anomaly, namely defects in the integrity of the atrial walls of the heart (Moulin et al, 2012), it is possible that in one experimental system there is enough IAP2 protein to avoid this lethal defect, whereas in another there is not. Furthermore, if this is the case, is the amount of cIAP2 aberrantly low in our mice, or is it aberrantly high in theirs, or both?

If there are differences in the production of cIAP2 protein, it might be due to the way the closely linked cIap1 locus was deleted in each of the strains. Heard et al (2015) used cIap1^−/− mice as described in Conze et al (2005). These were generated from 129/Sv E14 embryonic stem (ES) cells by homologous recombination of a neomycin (Neo) resistance gene in reverse orientation in place of the transcription initiation start codon and the first BIR domain of cIap1 (see Fig 1A of Conze et al, 2005). These mice were backcrossed to C57BL/6 mice for multiple generations. We generated cIap2^FRT/FRTcIap1^loxP/loxP mice by sequentially targeting the same chromosome in BRUCE embryonic stem cells, which were derived from C57BL/6 mice (Koentgen et al, 1993). In these mice, an FRT site is inserted 5′ of the ATG of cIap2, and an FRT-flanked Neo gene is inserted into the intron between exons 3 and 4 (see Fig 1A of Moulin et al, 2012).

Because the cIap2 gene is so close to the cIap1 gene, in some circumstances, the Neo gene or the promoter driving it in Heard et al’s cIap1^−/− mice might enhance the expression of the linked cIap2 gene, or it is possible that cIap2 regulatory sequences were inadvertently altered during homologous recombination. On the other hand, in our cIap2^FRT/FRTcIap1^−/− mice, it is possible that the intronic Neo gene and Pgk promoter sometimes decrease the expression of cIap2 or the efficiency with which its mRNA is spliced.

Another possible explanation for the differences between the two sets of cIap1^−/−Xiap^−/− mice is the presence or absence of 129/Sv versus C57BL/6 polymorphic genes, especially those physically linked to the cIap2-cIap1 locus. In our mice, the genes are of C57BL/6 origin, as the mice were generated from C57BL/6 BRUCE ES cells, whereas even with extensive backcrossing, the genes linked to the cIap2-cIap1 locus in Heard et al’s mice will be of 129/Sv origin. In addition to the mutation in caspase-11 already described (Kenneth et al, 2012), there is a very high probability that the Mmp1a gene is also mutated in these strains of mice (Vanden Berghe et al, 2015). We know that even minor differences in the expression of other genes can have a major effect on the survival of cIap1^−/−Xiap^−/− embryos. For example, when we crossed our cIap1^−/−Xiap^−/− mice onto a heterozygous Ripk1^+/− background, rather than dying at E12.5, some survived until weaning (Moulin et al, 2012, Fig 6C and Supplementary Fig S3).

There are several lines of experimentation that might reveal why Heard et al’s cIap1^−/−Xiap^−/− mice are viable, whereas our cIap1^−/−Xiap^−/− mice die by day E12.5. Sequencing the cIap1-cIap2 locus in the two cIap1^−/− lines might reveal unexpected changes in the parental cIap2^FRT/FRTcIap1^−/− or cIap1^−/− mice. Using CRISPR/Cas9 technology to mutate the cIap1^−/− gene in cell lines or C57BL/6 zygotes might show if our cIap2^FRT/FRTcIap1^−/− mice have aberrantly low levels of cIAP2, or their cIap1^−/− mice have aberrantly high levels of cIAP2. These mice could be crossed with Xiap^−/− knockouts to test their viability. We welcome any other suggestions for experiments and are happy to provide mice or cell lines to other investigators.