Abstract
I
Perhaps the authors refer to the lack of conservation between the most commonly disrupted region of the Rian locus and the orthologous region of MEG8. Of 37 HCC-associated rAAV integration sites in Rian from three independent mouse studies, 3 –5 21 (57%) occurred within Mir341, an intronic microRNA not present in the human genome. It is important to note, however, that nearly half of the HCC-associated integrations occurred elsewhere in the gene. Additionally, validation experiments established that HCC could be induced in mice by targeted insertion of an rAAV genome via in vivo homologous recombination at a site downstream of Mir341, 6 demonstrating that the precise integration site within the Rian locus is unlikely to be critical for transformation. Consistent with this hypothesis, unbiased insertional mutagenesis screens conducted by myself and others have detected recurrent HCC-associated mutations distributed throughout the Rian locus. 7,8
HCC-associated integrations within Rian are consistently associated with altered expression of adjacent genes, regardless of whether they occur within Mir341 or elsewhere. 3,4,6 –8 The precise mechanism behind these local effects on transcription has not been determined, but may involve disruption of normal imprinting within the Dlk1-Dio3 domain, which includes Rian/MEG8 and several other lncRNAs, mRNAs, snoRNAs, and microRNAs. Interestingly, the association between rAAV integration within Rian and HCC depends on the promoter and enhancer elements included within the vector, 3 suggesting specific interactions between exogenous and endogenous regulatory elements. Many Dlk1-Dio3 domain members have been linked to HCC in humans and animal models, 9 –11 with activation of Rtl1 in particular validated as a driver of hepatocarcinogenesis in mice. 8 Thus, neoplastic transformation initiated by integration within the Rian locus is likely to be functionally driven, at least in part, by other elements within the domain, nearly all of which have human orthologs.
Given that rAAV vectors integrate with a largely random pattern throughout the genome, aside from an apparent preference for actively transcribed genes, it is reasonable to assume that integrations within the MEG8 locus will occur at some rate in treated patients. If such integrations functionally recapitulate the transcriptional disruption of Dlk1-Dio3 domain elements observed in mice, they could similarly promote hepatocarcinogenesis. While the potential benefits of treating genetic liver diseases with rAAV-based gene therapy are great and this risk is probably low, as supported by a lack of evidence for association between rAAV and cancer in human clinical trials conducted to date, it should not be dismissed out of hand under the inaccurate assumption that “the human genome lacks the Rian locus.” Similarly, presenting this example as an illustration of how “rodent models are poor surrogates for humans, as well as poor predictors for human diseases” is unjustified, particularly given that molecular profiling of mouse HCCs driven by targeted rAAV integration at the Rian locus identified strong resemblance to a specific subclass of human HCC associated with poor prognosis. 6 In fact, valuable insights may be gained from mouse models of HCC caused by Rian integrations, such as guidance on which regulatory elements to include in therapeutic vectors and which Dlk1-Dio3 domain members have oncogenic or tumor suppressive potential.