Revealing the Superpowers of PrimPol: rescuing replicating microsatellites

Abstract

R‐loops are potentially mutagenic three‐stranded structures where RNA has hybridized to one strand of DNA and displaced the other, exposing ssDNA. Long repeated R‐loop‐forming sequences are known to cause genomic instability and are associated with disease. Šviković et al (2019) show that even short tandem (microsatellite) repeats, abundant in the vertebrate genome, do form R‐loops and present a barrier to replication. However, the replication fork can move past these short R‐loop‐forming repeats through the re‐priming action of primase–polymerase (PrimPol), thus avoiding the loss of epigenetic information or DNA damage.

Error‐free replication of repetitive stretches of DNA is crucial for human health, as more than 30 hereditary developmental and neurological diseases are linked to changes in length of microsatellites, where the repeating unit is < 9 nucleotides. Friedreich's ataxia (FRDA) is one such repeat expansion disease, exhibiting a large number of GAA repeats in the first intron of the FXN gene. Expansion of (GAA)_n results in reduced levels of the FXN protein, which leads to degeneration of the dorsal columns of the spinal cord and hypertrophic cardiomyopathy (Mirkin, 2007). Normal FXN alleles have < 12 repeats, while disease‐associated alleles often have 600–900 repeats. Therefore, it appears there is a limit to the number of (GAA)_n repeats that can be present before this becomes damaging.

GAA repeats and many other disease‐causing tandem repeat sequences can form DNA secondary structures, such as hairpins, triplexes (H‐DNA) or G‐quadruplexes (G4). The DNA strand not participating in the formation of these structures is exposed as single‐stranded DNA (ssDNA). After these sequences are transcribed in the correct direction, this exposed ssDNA is prone to hybridize with the complementary RNA transcript, leading to the formation of an R‐loop. R‐loops can in turn strongly stabilize secondary structures of such repeats, resulting in a synergistic stabilization of R‐loops and DNA structures. Recent work by Neil et al (2018) examined expansion of 100 GAA repeats in yeast and found that deletion of the R‐loop‐processing RNase H enzymes strongly promoted repeat expansion upon transcription of the sequence. The quantity of R‐loops correlates with the number of (GAA)_n repeats (Groh et al, 2014; Šviković et al, 2019), leading to the suggestion that shorter repeats might not cause DNA damage because they do not form stable R‐loops.

Šviković et al (2019) set out to study the potential effects of short GAA repeats that are frequently found throughout the genome, but are generally not considered as impediments to replication. They used a sensitive reporter set‐up in the BU‐1 locus in the chicken B‐cell line DT40 cells, where replication delay can be detected by monitoring BU‐1 expression levels in single cells (Schiavone et al, 2016). In this reporter system, the leading strand of the replication fork enters the locus from the 3′ end and then encounters the structure‐forming DNA being tested, which is 3.5 kb downstream of the transcription start site. Importantly, this set‐up represents a head‐on orientation of the replication and transcription machinery. If processive replication is interrupted while crossing the structure‐forming DNA in the reporter system, DNA unwinding and synthesis become uncoupled, recycling of the histones is disturbed, and this leads to inaccurate propagation of epigenetic information at the promoter region (Schiavone et al, 2014). At the BU‐1 locus, this replication‐dependent instability converts the normal “high” expression state into a lower expression state (Bu‐1a loss variants) that can be monitored by flow cytometry, with more Bu‐1a loss variants detected indicating greater replication impediment. Using this reporter system, the authors previously demonstrated the importance of the then newly discovered enzyme primase–polymerase (PrimPol) in the replication of G‐quadruplexes (Schiavone et al, 2016).

In their current paper, Šviković et al inserted various lengths of (GAA)_n repeats into the BU‐1 locus and monitored the frequency of Bu‐1a loss variant appearance, which signals the interruption of the processive replication process at the inserted repeats. In wild‐type cells, ten copies of a tandem repeat [(GAA)₁₀] did not impede replication which is unsurprising given that short tandem repeats are common in the vertebrate genome. Since their previous work had revealed the importance of PrimPol in the replication of sequences forming G4 structures, they decided to investigate the effect of PrimPol on (GAA)_n repeats, also known to form secondary DNA structures. Surprisingly, when PrimPol was deleted, (GAA)₁₀ showed a significant increase in the rate of Bu‐1a loss variants, indicating that even short tandem repeats can lead to replication impediments in the absence of PrimPol. This effect was orientation‐specific and occurred only when the (GAA)_n was transcribed. PrimPol has several proposed functions, but the authors used complementation experiments with separation‐of‐function mutants to establish PrimPol's re‐priming functionality as the necessary activity for restoring processive replication in the presence of the short GAA repeats. Taken together, these results show that a short purine‐rich repeat, (GAA)₁₀, which normally does not significantly impact on transcription or replication, is actually an impediment for replication in vivo, which is however overcome by the re‐priming activity of PrimPol.

In order to determine whether R‐loops play a role in the replication impediment caused by (GAA)₁₀ repeats, the authors monitored R‐loop formation at the BU‐1 locus using DNA:RNA immunoprecipitation (DRIP) experiments. While wild‐type cells showed low levels of R‐loops around (GAA)₁₀ repeats, PrimPol‐deleted cells exhibited a significantly increased R‐loop signal around this region; while without GAA repeats in BU‐1, PrimPol deletion did not increase R‐loop signals at this locus. In addition, overexpression of R‐loop‐degrading RNase H1 completely bypassed the requirement for PrimPol to restore processive replication of BU‐1 locus, indicating that R‐loop formation is necessary for the replication‐impeding effect of (GAA)₁₀ repeats. Conversely, stabilizing DNA:RNA hybrids by overexpressing the hybrid binding domain of RNase H1 that can bind but not process hybrids increased the Bu‐1a loss variants caused by (GAA)₁₀.

To extend their findings beyond the BU‐1 locus, the authors mapped DNA:RNA hybrids genome‐wide in wild‐type and PrimPol‐deleted DT40 cells and human pluripotent stem cells. This showed that loss of PrimPol leads to general R‐loop signal increases around genomic loci with secondary structure‐forming sequences, such as H‐DNA‐forming sequences and G4 motifs. These results suggest that while R‐loops are present at these loci in wild‐type cells, they are more persistent in the absence of PrimPol.

The authors propose a detailed molecular model that provides a potential explanation of the observed results: when short (GAA)_n repeats on the leading strand (upstream of the replication origin) are transcribed, the purine‐rich RNA transcript can form a stable R‐loop with the template DNA strand, which represents the lagging strand during replication. These R‐loops are probably necessary to stabilize the secondary structure of the GAA repeats on the leading strand. Once the replication fork reaches these short repeats, the replicative helicase moving along the leading strand can pass the obstacle formed by the GAA repeats, but synthesis of the leading strand is arrested. These events lead to the uncoupling of helicase activity from DNA synthesis and expose ssDNA segments ahead of the stalled replicative polymerase. In wild‐type cells, the arrested leading strand synthesis is quickly re‐primed upstream of the GAA repeats by PrimPol, restoring the processive replication complex. The exposed ssDNA is swiftly replicated and histone recycling remains undisturbed, preserving the epigenetic information at the promoter region of the BU‐1 locus (see Fig 1A).

Figure 1. Replication of R‐loop‐prone structure‐forming DNA repeats.

(A) Replication of short microsatellite repeats such as (GAA)₁₀ or a G4 repeat. In wild‐type cells, re‐priming by PrimPol maintains coupling of the replicative helicase and polymerase and thus prevents the loss of epigenetic information. In the absence of PrimPol, the replicative polymerase is impeded by the structural element while the helicase continues to progress, resulting in an extended stretch of single‐stranded DNA and defects in histone recycling. (B) Replication of extended repeats, such as (GAA)₁₀₀. A highly stable R‐loop may stall RNAPII resulting in a collision of the replication and transcription machinery which leads to replication fork collapse and DNA damage.

In PrimPol‐deleted cells, leading strand DNA synthesis is persistently blocked, but the replicative helicase continues to unwind the DNA duplex, resulting in chromatin disassembly and ssDNA exposure far upstream of the stalled polymerase. If this process reaches the promoter region of the BU‐1 locus, loss of epigenetic information can flip the expression level of BU‐1 from high to low. In addition, the extended ssDNA segments trigger additional R‐loop formation in the BU‐1 locus, potentially leading to replication fork collapse and the generation of DNA damage.

Interestingly, this model suggests that the primary effect of the short microsatellite repeats is the blockage of the leading strand synthesis, however, the causative R‐loops form on the lagging strand. R‐loops might “only” be needed to trigger and stabilize the H‐DNA formation of short GAA repeats; alternatively, they might have a more direct role in the replication impediment effect of these microsatellite repeats, which however remains to be fully understood.

Many nucleotide expansion diseases have a “safe” number of repeats, up to which the allele is phenotypically inert, and above which localized genetic instability occurs. Therefore, there is clearly a limit to the extent of structured DNA that PrimPol can rescue. This limit may depend on the stability of both the R‐loop (Landgraf et al, 1995) and the secondary DNA structure, as well as on the fate of the RNA polymerase that transcribes these repeats. When the nascent RNA transcript forms an R‐loop behind the transcribing polymerase, it can slow down or completely stall the transcription complex (Belotserkovskii et al, 2013). If the replication fork collides with such stalled RNA polymerase in a head‐on direction, the replication fork will collapse, causing DNA damage and potentially further amplification of the repeat region (Hamperl et al, 2017). It is unlikely that PrimPol could restore processive replication in the case of a full‐blown collision between the replication and transcription machinery. However, the short R‐loops at microsatellite repeats with only a few repeats may only slow or temporarily arrest the RNA polymerase, thus avoiding a full head‐on collision with the replication machinery (Fig 1B).

Overall, the study of Šviković et al (2019) strongly suggests that transcription of short (GAA)_n repeats, and more generally, microsatellite repeats with secondary structure‐forming sequences, can lead to R‐loop formation and interruption of the progressive replication process. However, such interruptions are masked by the re‐priming superpower of the PrimPol enzyme that can restart the replication process and prevent genomic instability around microsatellite repeats. These new findings might have important implications in the understanding of the expansions of short microsatellite repeats to disease‐associated alleles with a high number of repeats.

The EMBO Journal (2019) 38: e101298