Science in Society Archive

New Genetics, Evolution & Hazards of GMOs

RNA Inheritance of Acquired Characters

RNA not only registers epigenetic change as the organism responds to the environment, it also transmits acquired genetic information to subsequent generations independently of DNA, highlighting the potential perils of using RNA interference in GMOs Dr. Mae-Wan Ho

‘Epigenetic’ inseparable from ‘genetic’

The Oxford Dictionary currently defines ‘epigenetic’ as “relating to or arising from non-genetic influences on gene expression”, reflecting a lingering attachment to the outmoded idea that ‘genetic’ means genomic DNA. Genomic DNA of the germ line is still supposed to remain insulated from environmental influences, while everything else, including change in RNA or chemical markings of DNA and histone proteins (that do not change DNA base sequence), can then be assigned to ‘epigenetic’, and by implication, not strictly heritable, because most if not all epigenetic marks are removed at reproduction. This assumption is false, as chemical markings can indeed be stably inherited through many generations, apparently without the help of additional factors. For example, a significant proportion of 962 differentially methylated regions (DMRs) in 71 near-isogenic maize lines were found to be stably inherited (and without the action of small RNAs for maintenance) [1]. Second, the RNA species responsible for change in gene expression can itself be perpetrated over many generations, without being reverse transcribed and incorporated into the genome (which can also happen).

Acquired immunity to viral infection in nematode

An interesting example is the Induced expression of the Flock House virus in the nematode Caenorhabditis  elegans, which results in the production of virus-derived, small-interfering RNA that silence the viral genome. This viRNA-mediated viral silencing is transmitted in a non-Mendelian manner to many subsequent generations. It works in trans (via some mobile mediator) to silence viral genomes present in animals that are deficient in producing their own viRNA [2].

Lab strains of C. elegans such as N2 cannot be infected efficiently with a nodavirus that infects a wild strain, which is unable to mount an effective antiviral RNAi (RNA interference) response. Lab strains can fight viruses by processing the viral double-stranded (ds)RNA into virus-derived small interfering RNAs (viRNAs) that give specific immunity to the virus via Argonaute protein-dependent RNAi. The ability to respond to specific viruses by the production of targeted viRNAs is an acquired trait. This acquired trait is transmitted across generations in the absence of the genetic template, and even in the absence of functional small RNA-generating machinery. It relies on a RNA-dependent RNA polymerase that replicates the viRNA, which is encoded by all known nodaviruses as well as by the C. elegans genome.

Extragenomic RNA inheritance

The best-known epigenetic mechanisms include chemical modification of DNA and histone proteins to make the respective regions more and less accessible to the transcription machinery. However, extragenomic factors such as RNA are involved in many cases of non-Mendelian epigenetic inheritance (reviewed in [3]). The classical example is the white-spotted Kit mouse ‘paramutant’ phenotype. Heterozygous Kit mice mated with wild type mice results in Kit offspring with characteristic white tails and feet. When these mice are again mated with wild-type mice, a fraction of the offspring retains the white spotted phenotype even with a wild-type genotype. The phenotype could also be induced by microinjection of RNA into fertilized oocytes, suggesting that RNA plays an important role in inheritance.

Comparable epigenetic variations were generated at other loci by microinjection of microRNAs and transcript fragments. Heart hypertrophy could be induced by injecting miR, which causes an increased expression of the key effector Cdk9 in heart muscle-cell precursors. Similarly, when miR-124 or fragments of its Sox9 target transcript were injected, Sox9 became over-expressed during early embryonic development with increased proliferation of embryonic stem cells, resulting in increased body sizes during postnatal development and twin pregnancies. In those cases, the mutant phenotype was associated with increased rate of transcription of the target locus (rather than the usual inhibition in RNA interference). They are all distinct from Mendelian inheritance in being induced at far greater frequencies, and although eventually reversed, the changes are transmitted both paternally and maternally for three or more generations in crosses with wild-type partners and with close to 100 % efficiency, and paternal inheritance is related to the presence of sperm transcripts of the target gene and/or the cognate miRNAs.

Several studies showed that a complex and diverse set of RNAs is present in germ cells of both sexes as well as in early embryos. Sperm cells are capable of delivering RNAs into the oocyte as part of fertilization (see [4] Sperm Mediated Inheritance of Acquired Character, SiS 63). A fertilized egg is thus equipped with a diverse and complex complement of RNA inherited from both male and female germ lines. Because gametes are transcriptionally inactive, the RNA complement must be sufficiently stable to last for the life time of a gamete, through RNA binding proteins or antisense RNAs and post transcriptional modifications such as cytosine-5-methylation, which have been shown to be present in sperm RNAs (see [3]). Stabilization of RNAs may also be required during the maternal-to-zygote transition (MZT), which involves extensive changes in RNA profiles from the maternal transcripts to transcripts of zygotic genes. Specific RNAi effectors are capable of both destabilizing and stabilizing certain transcripts during the MZT, and are essential for the MZT. Because different classes of small RNAs play key roles in the MZT, inherited small non coding RNAs could be involved in remodelling RNA profiles during early embryogenesis.

Several studies indicate that miRNAs may be inherited thereby mediating inheritance of acquired phenotypes. Another possible mechanism could be the newly discovered circular (circ)RNAs [5], stable RNA molecules that can have marked effects on the transcriptome profile through binding miRNAs and their effector proteins, acting as miRNA sponges.

Another class of small regulatory RNAs with a well-established heritability is small interference (si)RNAs. Endogenous siRNAs are present in mouse germ cells of both sexes. It has been reported that more than half of the small RNAs in mature mouse sperm are tRNA fragments. These fragments can alter gene expression by functioning as siRNA mimics or inhibitors of translation initiation.

RNA methylation and epigenetic inheritance

RNA methylation may play a role in epigenetic inheritance by stabilizing regulatory RNAs thereby influencing MZT during embryogenesis. For example, adenine-6 methylation is the most prominent modification of mammalian mRNAs [5]. Adenine methylation marks have been identified in several thousand mRNAs with a distinct enrichment near the stop codons and in internal exons, and further analyses indicated an association with RNA splicing and/or miRNA binding. Cytosine-5 methylation is another prominent modification of RNA, which can be detected at single-base resolution by bisulphite sequencing. Early transcriptome-wide mapping suggested that m5C is prevalent in tRNAs but can also be found in other RNA species.

The methyltransferases, NSUN2 and DNMT2, were found primarily associated with tRNA. NSUN2 was also found to methylate mRNA, rRNAand several lncRNAs. Vault ncRNAs, which can be processed into small RNAs that regulate gene expression, are methylated by NSUN2. Loss of vault RNA methylation in Nsun2-deficient mice caused aberrant vault processing into Argonaute-associated small RNAs and aberrant expression of several mRNAs thought to be targets of small RNAs derived from vault.

DNMT2 RNA methyltransferase is a close relative of the well-known DNA methyltransferases DNMT1 and DNMT3, but it does not methylate DNA. On the contrary, it has strict specificity towards a certain set of tRNAs. DNMT2 methylation appears to protect the tRNAs from cleavage by endonuclease. Mice that lack both Dnmt2 and NSun2 have a significant reduction in protein translation associated with hypomethylation of tRNA. DNMT2 plays an important role in the generation of tRNA fragments that are known to affect the efficiency of small RNA silencing. The paramutant mouse phenotypes (see earlier) require Dnmt2 gene, suggesting DNMT2 methylation is required in RNA-dependent inheritance. Compared to other tissues, DNMT2 is highly expressed in mouse and human testes and ovaries.

Recent findings suggest that RNA-mediated inheritance is involved in human diseases. Variants in RNA methylation as well as demethylating factors have been associated with pathological phenotypes. Variants of the gene coding for FTO, a nonheme FeII/a-ketoglutarate-dependent dioxygenase that catalyzes the demethylation of m6A in RNA have been associated with high body mass index, risk of obesity and type 2 diabetes. Mutations in NSUN2 methyltransferase has been shown to cause autosomal recessive intellectual disability, and a diagnostic tRNA appeared clearly hypomthylated in fibroblasts from the mutation carriers.

RNA memory of gene expression maintains integrity of genome

Oded Rachevi at Tel Aviv University in Israel proposes that the integrity of the germline is maintained by transmitted RNA memories that record ancestral gene expression patterns and distinguish ‘self’ from ‘foreign’ sequences. This is achieved by ‘black listing’ invading nucleic acids and ‘guest listing’ endogenous genes. The memory is used by the next generation of small RNAs to act as ‘inherited vaccines’ that permit transcription of the approved ‘self’ sequences [6].

In the example of the inheritance of acquired immunity via RNA interference in C. elegans mentioned earlier [2], mutants defective in the RNA dependent RNA polymerase (RdRP) RRF-1, the inherited RNAi diminishes after two generations. RRF-1 RDRP is required for effective transgenerational targeting of chromatin remodelling, which may persist for more than 80 generations. C. elegans is considered for many years to be resistant to all known viruses, and RNAi is essential for resistance against virus.

In many organisms including bacteria, plants and animals, inherited small RNAs establish the foundation for trans-generational genome immunity. Apart from silencing viruses, heritable endogenous small interfering (endo-si)RNAs and piwi interfering (pi)RNAs also protect the genome from transposons. Although piRNAs act primarily in germ cells, in Drosophila, somatic follicle cells in the ovary produce antisense piRNAs from a Flamenco locus to repress gypsy elements (long terminal repeat retroviruses), which can move from somatic cells to germ cells in viral particles. piRNAs are maternally inherited in Drosophila to protect F1 progeny from transposons and hybrid dysgenesis (caused by the activation of mobile genetic elements after interspecific hybridization). piRNAs can be amplified, and unlike maternal proteins which dilute out, they persist into adulthood, and subsequent generations beyond, by RNA-dependent RNA polymerase or some other mechanisms.

Viruses encode proteins called VSRs (viral suppressors of RNA silencing) to block the host’s RNAi defence. For example, a class 1 RNase III VSR from sweet potato chlorotic stunt virus inhibits host antiviral RNAi by disabling siRNA amplification. Also, certain epigenetic effects favour invading viruses and establish an inherited susceptibility towards them. Thus, seeds from injured plants were found to be more susceptible to infections by a plant virus. Some of the signalling cascades that viruses hijack are used by plants to regulate endogenous genes across generations. In mammals, certain viruses hijack the host’s RNAi system directly to silence endogenous immunity genes and succeed in invading the host.

Virus-host interactions are complicated in persistent infections. The host may benefit by using the cell’s own reverse-transcription to integrate RNA viruses into the genome. Indeed, the human genome contains bits and pieces of non-retrovirus RNA viruses. In Drosophila, integration achieved through the reverse transcriptase activity of transposons results in establishing an antiviral RNAi response.

In addition to inherited viRNAs and piRNAs, other small RNA species confer resistance against animal predators. For example, small RNAs were demonstrated to confer resistance against lepidopteran predators for at least two generations in wild radish.

In general, prior infections are used in two ways to protect progeny. According to Rachevi, the organism can assign the foreign genes to a ‘black list’ consisting of all foreign genes encountered by its ancestor, and which must be suppressed; or it can add to a ‘guest list’ consisting of all those integrated into the genome and converted to self-genes that need to be expressed, and which can also repel further infections by the same agents.

The black list consists of effectively inherited vaccines. In Drosophila germ cells, specific genomic loci act as traps for transposons that land in them. Once trapped, the transposon is immobilized, because the DNA is wrapped in heterochromatin, and piRNAs synthesized from the transposon’s landing site continuously recruit chromatin-remodelling genes to maintain the heterochromatin state. And once piRNAs have been synthesized, this memory is inherited across generations. In addition, the dsRNA that transposons and RNA viruses produce serve as a substrate for the RNAi machinery, leading to the production of heritable inhibitory viRNA and piRNA memories. In bacteria, invading DNA phages can be cut up and packed into specific genome locations - clustered regulatory interspaced short palindromic repeats (CRISPRs) - where they can be silenced. In subsequent bacterial generations, CRISPR RNAs transcribed from these loci will prime the bacteria’s defence against newly infecting phages, thus eliminating them.

The second strategy for inherited immunity is to make a small RNA-based guest list of endogenous genes to distinguish them from the black list of invading elements. For example, the unicellular ciliate Oxytrichia trifallax has two genomes packed inside two separate nuclei: somatic genome that encodes all vegetative growth function and a germline genome responsible for sexual reproduction. During development of the somatic nucleus, the ciliate rids itself of all parasitic DNA (> 95 %). RNA guides DNA rearrangements in Oxytrichia, and it was recently shown that piRNAs are central to this process; piRNAs that match genes present in the somatic genome are inherited from the maternal nucleus. Because these piRNAs correspond to genes that were previously present in the soma, they indicate genes that are safe to express and are ‘self’, <5 %. Thus, synthetic piRNAs matching regions otherwise deleted, when injected into Osytrichia led to retention of these DNA regions in later generations. This RNA cache idea was later dubbed ‘RNA licensing when it was demonstrated in C. elegans that an RNA transcript of the gene fem-1, a sex-determining factor, needs to be transcribed in the mother to license its expression in the zygote germline. Introducing a fem-1 deletion that eliminated mRNA production in the mother resulted in a feminized germline in the heterozygous progeny, while injection of fem-1 RNA, even one incapable of coding for a protein into the maternal germlne was sufficient to rescue the defect in the offspring. This suggests fem-1 RNA prevents epigenetic silencing in the next generation and may be a way to protect the identity and integrity of the germline. Several groups have proposed that heritable piRNAs bound to PRG-1 protein scan and silence foreign elements by inducing chromatin remodelling in the germline.

Inheritance of silenced and actively transcribed genes determined by RNA

All organisms use a variety of mechanisms to silence transgenes artificially introduced by genetic modification, and this has accounted for widely observed transgene instability (see [7]).

Recent findings also suggest that organisms use RNAi-related mechanisms to detect foreign sequences by comparing the foreign sequence to a memory of previous gene expression. In many organisms, transgene silencing has been linked to factors also required in the RNAi pathway: a sequence-specific response triggered by double-stranded (ds)RNA, which is processed by RNAse II-related protein Dicer into ~21 nt short interfering RNAs (siRNAs) loaded onto Argonaute (AGO) proteins to form the key effectors of RNA induced silencing complexes (RISCs) (reviewed in [8]).  AGOs are RNAse H-related proteins that use the base-pairing of small RNA cofactors to guide sequence-specific binding to target sequences to cleave the target directly or indirectly, by recruiting further factors for mRNA destruction, or execute other modes of gene regulation.

Despite overlap between the mechanisms that mediate RNAi and the silencing of transposons and transgenes, there are distinct triggering mechanisms. For example, the AGO protein RDE-1 is essential for the dsRNA response in C. elegans, but not required for transposon or transgene silencing. RDE-1 is thought to recruit a cellular RdRP which uses the target mRNA as a template for producing secondary siRNAs, 22G-RNAs (22 nucleotides starting with G). The 22G-RNAs are loaded onto members of an expanded, partially redundant group of worm-specific AGOs (WAGOs). WAGOs that localize to the cytoplasm mediate mRNA turnover, whereas WAGOs that localize to the nucleus mediate transcriptional silencing.

In the germline, RdRPs not only produce 22G-RNAs that interact with WAGOs, but also those that interact with a distinct AGO, CSR-1, which is required for fertility and chromosome segregation. Some factors including RDE-3 and MUT-7 are only required for WAGO 22G-RNA to accumulate, indicating that the CSR-1 and WAGO22 pathways also involve distinct mechanisms. The WAGO and CSR-1 22G pathways together target virtually all germline-expressed mRNAS, but their targets are largely non-overlapping. Unlike the WAGO pathway, the CSR-1 22G pathway does not appear to silence its targets. Instead, the CSR-1 pathway may help to define and maintain euchromatic regions (with actively transcribed genes) in order to support the proper assembly of kinetochores (protein structures on chromosomes where fibres attach during cell division to pull the replicated chromosomes apart).

In most animals, the Piwi family AGOs are required for fertility and transposon silencing. In C. elegans, however, the piwi-related gene product PRG1 has only been linked to the silencing of one transposon family, Tc3, by recruiting RdRP and the WAGO22 pathway. Piwi-interacting RNAs are genomically encoded and appear to be expressed as Pol II transcripts whose single-stranded products are processed and loaded onto Piwi proteins. More than 15 000 distinct piRNA species exist in C. elegans and millions are expressed in the testis of mammals. The majority map uniquely to the genome and lack obvious targets, so their function remains entirely unknown.

Researchers at University of Massachusetts Medical School led by Craig Mello used Mos1 inserts (a transposon introduced from Drosophila) at mapped locations in the C. elegans genome to introduce single-copy transgenes into the C. elegans genome. The targeted single copy transgene insertion produced stable transgenic strains, as opposed to multiple copy inserts created by usual methods.  They then carried out a comprehensive series of genetic and molecular analyses to show that transgenes inserted at the same chromosomal site can exhibit opposite and remarkably stable epigenetic fates: either expressed or silenced [8]. Transgenes consisting of an endogenous germline-expressed gene fused to a relatively long foreign sequence (eg gfp) were prone to silencing, but not when fused to a short foreign sequence. Silencing is dependent on nuclear and cytoplasmic WAGOs and correlates with the accumulation of 22G-RNA targeting the foreign portion of the transgene. PRG-1 is required for initiating but not for maintaining silencing, which involves a chromatin component, the methylation of lysine 9 on histone H3 (H3K9me). The researchers proposed that PRG-1 and its 21U-RNA  (21 nt long starting with U) cofactors scan for foreign RNA sequences that initiate WAGO-maintained gene silencing, while endogenous mRNAS are protected from silencing, perhaps by the CSR-1 22G-RNA pathway.

The initiation of silencing in C. elegans involves the comparison of the foreign sequence to an epigenetic memory of previously expressed sequences. A model is proposed in which three AGO pathways function together in a system that maintains an inventory of expressed mRNAs while constantly scanning for foreign sequences. In this system, PRG-1 uses genomically encoded piRNA cofactors to scan, via imperfect base-pairing interactions for foreign RNAs expressed in the germline. Upon targeting, Prg-1 recruits RdRP to produce antisense 22G-RNAs that are loaded onto WAGO Argonautes. In turn WAGOs mediate silencing and establish a memory of nonself RNA. A third as yet unidentified pathway provides a memory of self and is capable of acting as an ante-silencing signal. The CSR-1 22G-RNA pathway provides an attractive candidate. The self-recognition pathway can prevent Prg-1 from recruiting the WAGO pathway providing a function that helps expressed transgenes maintain their expression and endogenous genes to recover from WAGO-mediated silencing induced by RNAi. The initial decision to silence or express represents a stochastic outcome of competition between the establishment of epigenetic self- and nonself memories. This hypothesis has been confirmed.

RNAi maintains memory of actively transcribed genes

A team led by Alla Grishok at Columbia University New York in the United States found that the CSR-1 22G-RNA pathway is responsible for activating genes: it promoted sense-oriented RNA polymerase II (Pol II) transcription [9, 10]. To evaluate genome wide effects of the CSR-1 pathway on Pol II transcription, they used a global run-on sequencing method that maps the position and determines the amount and orientation of transcriptionally engaged Pol II. In csr-1 mutants, and mutants of Dicer-related helicase drh-3 (which are depleted of 22G RNAs), there was a global reduction of transcription at CSR-1 target genes and also a global increase in transcription of non-CSR-1 target genes. This suggests that CSR-1 specifically promotes transcription of its target genes in a 22G-RNA dependent manner. The researchers found that pol II transcription in CSR-1 pathway mutants was reduced along the entire length of genes including the promoter, and CSR-1 interacted with the Pol II complex depending on the RNA. Thus, CSR-1 seems to have a direct effect on the Pol II complex at target genes by associating through 22G-RNAs with nascent transcripts.

Another important finding was a global increase in antisense Pol II transcription in csr-1 and drh-3 mutants. This suggests that by interacting with nascent sense RNAs, CSR-1 associated 22G-RNAs stabilize the sense oriented Pol II machinery, and in turn reduce the probability of pol II initiation in antisense orientation.

In wild type animals, Pol II was predominantly located at genes that are highly expressed in the germ line, whereas in the mutants, Pol II was depleted at those genes and instead became enriched at genes that should be weakly expressed or silenced, such as domains that normally contain the centromere histone H3 variant CENPA and the H3K27me3 repressive mark.

Together, the results indicate that the CSR-1 RNAi pathway promotes sense-oriented transcription by Pol II and concomitantly reduced antisense transcription. In this way, it contributes to the propagation of silent and active chromatin and to the maintenance of genome organization.

To conclude

Numerous species of RNA encoded in the genome, or independent of the genome, are playing crucial roles in epigenetic inheritance including the transmission of viral resistance, silencing of transposons and other foreign genes as well as in the acquisition and maintenance of actively transcribed genes.  Known mechanisms such as RNA interference, RNA-dependent RNA polymerase, RNA methylation, and RNA memory interact with one another and with proteins and chromatin in circular, mutually reinforcing networks,  greatly expanding the conventional notion of genetic information and further blurring the distinction between epigenetic and genetic inheritance.  This further underscores the perils of all forms of artificial genetic modification, especially those involving RNA interference, complex and flexible in the context of natural genetic modification  (see [11] New GM Nightmares with RNA, SiS 58; [12] RNA Interference "Complex and Flexible", SiS 59), and definitely beyond the control of the artificial process.

Article first published 02/07/14


  1. Li Q, Eichten SR, Hermanson PJ and Springer NM. Inheritance patterns and stability of DNA methylation variation in maize near-isogenic lines. Genetics 2013 Early Online 20 December 2013, doi:10.1534/genetics.113.148980
  2. Rechavi O, Minevich G and Hobert O. Transgenerational inheritance of an acquired small RNA-based antiviral response in C. elegans. Cell 2011, 147, 1248-56.
  3. Liebers R, Rassoulzadegan M and Lyko Frank. Epigenetic regulation by heritable RNA. PLOS Genetics 2014, 10, e10042960.
  4. Ho MW. Sperm mediated inheritance of acquired characters. Science in Society 63 (to appear) 2014.
  5. “Circular RNAs throw genetics for a loop”, Heidi Ledford, Nature News 27 February 2013,
  6. Rechavi O. Guest list or black list: heritable small RNAs as immunogenic memories. Trends in Cell Biology 2014, 24, 212-20.
  7. Ho MW. The new genetics and natural versus artificial genetic modification. Entropy 2013, 15, 4748-81.
  8. Shirayama M, Seth M, Lee H-C, Gu W, Ishidate T, Cont jr D, Mello CC. piRNAs initiate an epigenetic memory of nonself RNA in the C. elegans germline. Cell 2012, 150, 65-77.
  9. Cecere G. Hoersch S, O’Keeffe S, Sachidanandam R and Grishok A. Global effects of the CSR-1 RNA interference pathway on the transcriptional landscape. Nature Structural & Molecular Biology 2014, 21, 358-65.
  10. Bauman K. RNAi as a global transcriptional activator. Nature Reviews Molecular Biology published online 16 April 2014; doi:10.1038/nrm3797
  11. Ho MW. New GM nightmares with RNA. Science in Society 58, 6-7, 2013.
  12. Ho MW. RNA interference “complex and flexible”. Science in Society 59, 1819, 2013.

Got something to say about this page? Comment

Comment on this article

Comments may be published. All comments are moderated. Name and email details are required.

Email address:
Your comments:
Anti spam question:
How many legs on a cat?

There are 1 comments on this article so far. Add your comment above.

dhinds Comment left 3rd July 2014 14:02:24
This a bomb. Somehow I missed the earlier article on this subject: Like the technology itself, the biotech industry is clearly out of control and represents a real and present threat to the future of the biological world we (home sapiens) form part of and yet are destroying, thanks to misguided efforts to harness the foundations of life in ways not compatible with the ongoing evolution process, in order to generate "income" (an articial goal) for corporate entities with little understanding and even less moral fibre that have never-the-less managed to manipulate the governmental agencies charged with overseeing the public good for the benefit of their own self-destructive and life destroying goals. The gravity of the foolish biotech vision can not be overemphasized.