On 2014 Oct 27, Maria Flavia Di Renzo commented:

As mentioned in the paper and shown in Supplementary Figure 2, we also used different single shRNAs. Actually, we also used five shRNAs specific for different sequences of CDT2 and different from the sequences of the four siRNAs of the pool. We obtained regulation of CDT2 targets and commitment towards cell death in cancer cells but not in normal cells. However, the in-depth analysis of molecular targets and cell cycle was not feasible in cells stably committed to death. Thus we carried out experiments with siRNAs that resulted in similar, but transitory, biochemical and functional effects. In conclusion, we are confident that the observed phenotypes might be attributed to CDT2 suppression.

On 2014 Oct 17, Eugen Buehler commented:

The authors cite our paper (Buehler E, 2012) as evidence that “…pools of siRNA targeting different mRNAs, such as those used in libraries, results in increased off-target effects”. Unfortunately, that sentence is not supported by our paper. The method demonstrated in our paper was applied to an arrayed screen of individual siRNAs in which there were no pools, and no conclusions about pooling of siRNAs can be reached from that manuscript. Furthermore, the authors state that “The use of this siRNA pool allows avoiding too high concentration of each single siRNA and thus prevents off-target effects” and cite Jackson AL, 2010 to support this. Again, this assertion is not supported by the cited reference. Jackson AL, 2010 states that “Pooling of multiple siRNAs to the same target may help to reduce off-target silencing, due to competition among the siRNAs in the pool” (emphasis added). No evidence exists to support the claim that pools 3 or 4 siRNAs will prevent off-target effects. To the contrary, we have demonstrated previously that pooling siRNAs does not generate more reproducible phenotypes than single siRNAs (see Marine S, 2012, Figure 1, B and C) and that pools of siRNAs can and do generate many off-target phenotypes, which can then be confirmed by deconvolution of those pools (see Marine S, 2012, Figure 2). Unfortunately, the authors base most of their conclusions on experiments using a single reagent (a pool of four different siRNAs targeting CDT2) and a single non-silencing control. This means that many or all of the phenotypes they observed experimentally may be due to siRNA off-target effects (see Chung HY, 2014 for a case study in how off-target effects can result in false positives). This is not to say that their conclusions are incorrect, only that I believe that RNAi experiments require multiple independently tested reagents and/or matched controls to guard against the false positives due to off-target effects that are so frequent in these experiments.

On 2014 Oct 17, Eugen Buehler commented:

The authors cite our paper (Buehler E, 2012) as evidence that “…pools of siRNA targeting different mRNAs, such as those used in libraries, results in increased off-target effects”. Unfortunately, that sentence is not supported by our paper. The method demonstrated in our paper was applied to an arrayed screen of individual siRNAs in which there were no pools, and no conclusions about pooling of siRNAs can be reached from that manuscript. Furthermore, the authors state that “The use of this siRNA pool allows avoiding too high concentration of each single siRNA and thus prevents off-target effects” and cite Jackson AL, 2010 to support this. Again, this assertion is not supported by the cited reference. Jackson AL, 2010 states that “Pooling of multiple siRNAs to the same target may help to reduce off-target silencing, due to competition among the siRNAs in the pool” (emphasis added). No evidence exists to support the claim that pools 3 or 4 siRNAs will prevent off-target effects. To the contrary, we have demonstrated previously that pooling siRNAs does not generate more reproducible phenotypes than single siRNAs (see Marine S, 2012, Figure 1, B and C) and that pools of siRNAs can and do generate many off-target phenotypes, which can then be confirmed by deconvolution of those pools (see Marine S, 2012, Figure 2). Unfortunately, the authors base most of their conclusions on experiments using a single reagent (a pool of four different siRNAs targeting CDT2) and a single non-silencing control. This means that many or all of the phenotypes they observed experimentally may be due to siRNA off-target effects (see Chung HY, 2014 for a case study in how off-target effects can result in false positives). This is not to say that their conclusions are incorrect, only that I believe that RNAi experiments require multiple independently tested reagents and/or matched controls to guard against the false positives due to off-target effects that are so frequent in these experiments.