On 2017 Jul 07, Wei Wang commented:

Several other problems have been reported: https://pubpeer.com/publications/DAA39CA8966C9B4DAB38EDAA343B4C

On 2014 Dec 15, Ivan Shatsky commented:

General comment: The phenomenon studied in this paper is very exciting. I read the article with a great interest. Unfortunately, I regret to say that the underlying mechanism remained uncovered. Although I agree that the authors identified some curious structures within the 5’UTRs of HOXA mRNAs which might be implicated in the regulation of these mRNAs by RPL38, I did not find sufficient evidence for existence of IRES-elements in these mRNAs. As in numerous other similar investigations, to identify IRES-elements the authors employed the method of DNA bicistronic constructs, the approach that had been repeatedly shown to be associated with almost unavoidable artifacts (see Jackson Cold Spring Harb Perspect Biol. 2013 Feb 1;5(2); Lemp et al. Nucleic Acids Res. 2012 Aug;40(15):7280-90.). And I suspect this paper is not free of those artifacts either (see below). Several crucial control experiments necessary to support or to exclude the IRES-mediated mechanism have been recently described (for references see Shatsky et al. Mol Cells. 2010 Oct; 30(4):285-93). One of them, for instance, is the ratio of translational activities for m7G capped versus uncapped (A-capped) monocistronic constructs. This value estimates contribution of the cap to the translational potential of a 5’UTR under selected conditions. If this contribution is very high (as is the case of cap-dependent mRNAs) one may exclude the presence of a true IRES. I think that this and other obligatory controls are feasible to perform with cells C3H10T1/2 used in this paper but they were not done.

Some specific points:

1. The authors regard the cap-independent and IRES-dependent modes of translation initiation as synonymous mechanisms and support this notion with reference 21. I should stress that not all specialists in eukaryotic translation would share this opinion since nobody has ever shown that a 5’end dependent translation initiation cannot be regulated by specific structures within the respective 5' UTR. The opposite has recently been demonstrated (Terenin et al. Nucleic Acids Res. 2013 Feb 1;41(3):1807-16.)
2. When listing examples of cellular IRESs identified to date (second paragraph), the authors mention c-myc, Apaf-1, XIAP. To the best of my knowledge, these cellular IRESs have been disproved (Andreev et al. Nucleic Acids Res. 2009 Oct;37(18):6135-47; Bert et al. RNA. 2006 Jun;12(6):1074-83; Baranick et al. Proc Natl Acad Sci U S A. 2008 Mar 25;105(12):4733-8; van Eden et al. RNA. 2004 Apr;10(4):720-30); Lemp et al. Nucleic Acids Res. 2012 Aug;40(15):7280-90. )
3. “Extended data Figure 1a” raises a great concern: the HCV IRES should not be used as a normalizing construct for testing bicistronic DNAs since the HCV IRES has been reported to harbor a cryptic promoter (Dumas, E. et al. 2003 Nucleic Acids Res. 31 (4): 1275-1281) and hence may produce capped monocistronic mRNAs . By the way, among viral IRESs characterized to date the HCV IRES is regarded as one of the weakest.
4. The control test with Rluc shRNA (Extended data Figure 1 b,c) strongly suggests that some significant amount of monocistronic (and therefore capped) Fluc mRNAs is present in transfected cells since the residual Fluc activity after RNA interference is too high. This may also be the case for the control bicistronic mRNA containing the HCV IRES (see point 3). Otherwise, the pictures 1b and 1c must be similar. At least, in the analogous test performed in our lab, the Rluc and Fluc activities fall down to the similar background levels (Fig. 2D in Dmitriev et al. Mol Cell Biol. 2007 Jul;27(13):4685-97).
5. The authors suggest that the IRES elements are mostly confined within ~300 nts proximal to the start codon. As a support to this conclusion, they note that some of HOXA mRNAs possess a 5’ UTR of that size. If so, why the activity of HOXA9 construct 944-1266 is much lower than that for the full length 5’UTR (Fig. 1d)?<br>
6. “Extended Figure 1d”. This control is useless. It only shows that the bicistronic mRNA of the expected size is present in transfected cells but unable to show the presence of monocistronic mRNAs starting within the intercistronic region. The corresponding bands won’t be seen.
7. On the base of pull-down experiments the authors claim that 80S ribosomes are specifically formed on their IRES-elements. The problem is that they use 10mM of magnesium in these expts, i.e. the concentration at which the assembly of translation initiation complexes in mammalian systems should not occur.
8. The mode of action of TIE element looks absolutely puzzling. It is even difficult to imagine any mechanism for its operation. My question: is it specific to these particular cells?

On 2014 Dec 15, Ivan Shatsky commented:

General comment: The phenomenon studied in this paper is very exciting. I read the article with a great interest. Unfortunately, I regret to say that the underlying mechanism remained uncovered. Although I agree that the authors identified some curious structures within the 5’UTRs of HOXA mRNAs which might be implicated in the regulation of these mRNAs by RPL38, I did not find sufficient evidence for existence of IRES-elements in these mRNAs. As in numerous other similar investigations, to identify IRES-elements the authors employed the method of DNA bicistronic constructs, the approach that had been repeatedly shown to be associated with almost unavoidable artifacts (see Jackson Cold Spring Harb Perspect Biol. 2013 Feb 1;5(2); Lemp et al. Nucleic Acids Res. 2012 Aug;40(15):7280-90.). And I suspect this paper is not free of those artifacts either (see below). Several crucial control experiments necessary to support or to exclude the IRES-mediated mechanism have been recently described (for references see Shatsky et al. Mol Cells. 2010 Oct; 30(4):285-93). One of them, for instance, is the ratio of translational activities for m7G capped versus uncapped (A-capped) monocistronic constructs. This value estimates contribution of the cap to the translational potential of a 5’UTR under selected conditions. If this contribution is very high (as is the case of cap-dependent mRNAs) one may exclude the presence of a true IRES. I think that this and other obligatory controls are feasible to perform with cells C3H10T1/2 used in this paper but they were not done.

Some specific points:

1. The authors regard the cap-independent and IRES-dependent modes of translation initiation as synonymous mechanisms and support this notion with reference 21. I should stress that not all specialists in eukaryotic translation would share this opinion since nobody has ever shown that a 5’end dependent translation initiation cannot be regulated by specific structures within the respective 5' UTR. The opposite has recently been demonstrated (Terenin et al. Nucleic Acids Res. 2013 Feb 1;41(3):1807-16.)
2. When listing examples of cellular IRESs identified to date (second paragraph), the authors mention c-myc, Apaf-1, XIAP. To the best of my knowledge, these cellular IRESs have been disproved (Andreev et al. Nucleic Acids Res. 2009 Oct;37(18):6135-47; Bert et al. RNA. 2006 Jun;12(6):1074-83; Baranick et al. Proc Natl Acad Sci U S A. 2008 Mar 25;105(12):4733-8; van Eden et al. RNA. 2004 Apr;10(4):720-30); Lemp et al. Nucleic Acids Res. 2012 Aug;40(15):7280-90. )
3. “Extended data Figure 1a” raises a great concern: the HCV IRES should not be used as a normalizing construct for testing bicistronic DNAs since the HCV IRES has been reported to harbor a cryptic promoter (Dumas, E. et al. 2003 Nucleic Acids Res. 31 (4): 1275-1281) and hence may produce capped monocistronic mRNAs . By the way, among viral IRESs characterized to date the HCV IRES is regarded as one of the weakest.
4. The control test with Rluc shRNA (Extended data Figure 1 b,c) strongly suggests that some significant amount of monocistronic (and therefore capped) Fluc mRNAs is present in transfected cells since the residual Fluc activity after RNA interference is too high. This may also be the case for the control bicistronic mRNA containing the HCV IRES (see point 3). Otherwise, the pictures 1b and 1c must be similar. At least, in the analogous test performed in our lab, the Rluc and Fluc activities fall down to the similar background levels (Fig. 2D in Dmitriev et al. Mol Cell Biol. 2007 Jul;27(13):4685-97).
5. The authors suggest that the IRES elements are mostly confined within ~300 nts proximal to the start codon. As a support to this conclusion, they note that some of HOXA mRNAs possess a 5’ UTR of that size. If so, why the activity of HOXA9 construct 944-1266 is much lower than that for the full length 5’UTR (Fig. 1d)?<br>
6. “Extended Figure 1d”. This control is useless. It only shows that the bicistronic mRNA of the expected size is present in transfected cells but unable to show the presence of monocistronic mRNAs starting within the intercistronic region. The corresponding bands won’t be seen.
7. On the base of pull-down experiments the authors claim that 80S ribosomes are specifically formed on their IRES-elements. The problem is that they use 10mM of magnesium in these expts, i.e. the concentration at which the assembly of translation initiation complexes in mammalian systems should not occur.
8. The mode of action of TIE element looks absolutely puzzling. It is even difficult to imagine any mechanism for its operation. My question: is it specific to these particular cells?

On 2017 Jul 07, Wei Wang commented:

Several other problems have been reported: https://pubpeer.com/publications/DAA39CA8966C9B4DAB38EDAA343B4C