The Retraction of a Plasmid-Phage Chimera Genome Assembly Study Leaves Serious Issues Unaddressed—Essentially, Nolo Contendere

LETTER

This retraction (1) reads as though the authors themselves identified two unintentional sequencing errors, which they also report correcting. These two errors did not affect the basic tenet of the study, the reported discovery of a plasmid-phage chimera in one replicon (pKP12226, GenBank entry KP453775), leaving readers to question why this warrants retraction. In fact, I first raised these errors and multiple other anomalies in their assembly with the journal, who shared my concerns with the authors. These anomalies raise serious doubts about the accuracy and completeness of even their newly corrected, but not currently public, sequence assembly. To my knowledge, the authors have not addressed all of these other issues, which I summarize here to enable a more complete interpretation of their assembly synthesis by readers (for full details, see the supplemental material).

1. The fusion of the plasmid and phage replicons is questionable. The authors postulated that IS2 insertion sequences (IS) may have played “an important role in recombination between the plasmid and the phage,” but while sequence-divergent and inverted IS2 elements are present in both the phage and plasmid sequences, they are both 10 kb removed from the replicon fusion sites. While their plasmid backbone sequence has different, and partial or highly incomplete, IS sequences at both ends, these provide no support for IS-mediated recombination. Instead, the plasmid genome is directly inserted into the phage genome, internal to, and therefore inactivating, a phage gene encoding the gp7 structural protein. This gene is, however, unannotated in their assembly at 94,096←94,315′ and ′1←509.

2. The inversion of an approximately 56-kb region of the plasmid backbone—also directly joined to the phage genome—is similarly in doubt. It is flanked by direct repeats of full 768-bp IS1 elements in their comparator plasmid pIP1206, but this region in pKP12226 has IS1 sequences trimmed to only 8 bp at the 5′ end and the 3′ homology ends 2 bp before the pIP1206 3′ IS1. Both genome fusion and backbone segment inversion junctions are reported as being confirmable by PCR, which the authors restate in the retraction notice, although specific primer locations or sequences were not given. Nevertheless, detailed analysis of their sequence raises doubts about both of these events and the overall assembly.

3. They report 15× coverage of the 267,645-bp pKP12226 genome from 51,656 Roche 454 reads, which, if correct, produces a very short 78-bp average read length (coverage × size/total reads). The assembly has more than 30 identical repeat sequences larger than 100 bp but only 5 larger than 400 bp. The authors report generating five contigs, joined by PCR and gap-filling (presumably by Sanger sequencing). However, there are at least nine regions (including all three plasmid-phage and backbone inversion junctions) where non-contiguous sequence matches to other GenBank entries are joined. Each of these is unique, with no other GenBank entries having these junctions (detailed in the supplemental material). Additionally, severely truncated ends of IS1 or IS26 elements are found at six of these junctions (only 4, 5, or 8 bp of one end of the IS element’s terminal inverted repeats remain in their assembly, when complete IS1 or 820 bp IS26 elements are present at these exact insertion points in other plasmid genomes), and other unique junctions end precisely with coding sequences. This is consistent with nine assembled contigs, excessively end-trimmed of repetitive or noncoding sequences, being somewhat randomly and directly joined, with no linking sequences. I am seriously concerned with the authors’ stated method of connecting (five) contigs by PCR and gap-filling.

4. Their plasmid assembly ends with a 100% identical match to the final 2,100 bp of an otherwise unique 2.9-kb 2010 GenBank entry (GQ274928) (2) for a plasmid-borne bla_CTX-M-15 locus and its flanking sequences (ISEcp1 and a partial IS26 locus). This region of the GenBank entry has three unique identifiers: (i) an erroneously inverted bla_CTX-M-15 gene (compared to its published plasmid map [2] and all other bla_CTX-M-15 loci), and (ii) deletions of two single-G bases (present in all other IS26 sequences) near the end of (iii) a partial 780-bp IS26′ which ends both the GQ274928 and pKP12226 sequences. That these three unique signatures are also present and identical in their assembly is concerning, as is the direct insertion of this sequence into a structural gene of the phage that resulted in the fusion of the plasmid and phage replicons. The start of this 2,100-bp sequence (the bla_CTX-M-15 gene) appears to have been spliced with a typical bla_CTX-M-15 locus (i.e., preceded by an IS26 insertion in an ′ISEcp1 element), resulting in a bla_CTX-M-15 gene directly flanked by 477-bp inverted repeats of ′ISEcp1 and bla_CTX-M-15 proximal sequences (normally found only 5′ of the bla_CTX-M-15 gene); as a result, the bla_CTX-M-15 locus is also flanked by inverted IS26 elements in a totally unnatural arrangement. Presumably, this is the gene inversion error they report correcting. But merely (re)inverting the bla_CTX-M-15 coding sequence in their assembly does not correct this error or explain the GQ274928 identity. It is also unclear to which sequences (or other plasmids) they refer in their retraction regarding the 17 to 18 kb erroneously incorporated (or how they were initially inserted, or subsequently corrected). I have noted similar IS-related anomalies, other assembly issues, and the same problems with their reported assembly methodologies, in two other plasmid genome assembly studies by this group (3–5); assemblies in a third study (6) are also under investigation, and two more are noted in Expressions of Concern (7, 8) published by Antimicrob Agents Chemother concurrently with this retraction.

The authors’ retracting of their study while claiming their tenet to still be valid (by restating PCR- and sequence read-based claims for a plasmid-phage replicon fusion) is disconcerting. PCR is not infallible and should be confirmed by Sanger sequencing. Stating that they “reexamined the primary sequence data” implies that no corrections other than the two noted points were necessary, but they do not acknowledge any of my other noted anomalies. Any revision of the sequence that does not address all issues remains incorrect. Resolution of these issues would best be accomplished by an independent analysis of the primary sequence data. Requiring submission of this primary data to the SRA at NCBI along with the assembly sequence would help to resolve these issues. The science community as a whole would benefit from an open and candid analysis of the data in question, along with a full disclosure of all of the issues relating to a retraction.

Ed. Note: The author of the published article did not respond.