High density of unrepaired genomic ribonucleotides leads to Topoisomerase 1-mediated severe growth defects in absence of ribonucleotide reductase

Abstract Cellular levels of ribonucleoside triphosphates (rNTPs) are much higher than those of deoxyribonucleoside triphosphates (dNTPs), thereby influencing the frequency of incorporation of ribonucleoside monophosphates (rNMPs) by DNA polymerases (Pol) into DNA. RNase H2-initiated ribonucleotide excision repair (RER) efficiently removes single rNMPs in genomic DNA. However, processing of rNMPs by Topoisomerase 1 (Top1) in absence of RER induces mutations and genome instability. Here, we greatly increased the abundance of genomic rNMPs in Saccharomyces cerevisiae by depleting Rnr1, the major subunit of ribonucleotide reductase, which converts ribonucleotides to deoxyribonucleotides. We found that in strains that are depleted of Rnr1, RER-deficient, and harbor an rNTP-permissive replicative Pol mutant, excessive accumulation of single genomic rNMPs severely compromised growth, but this was reversed in absence of Top1. Thus, under Rnr1 depletion, limited dNTP pools slow DNA synthesis by replicative Pols and provoke the incorporation of high levels of rNMPs in genomic DNA. If a threshold of single genomic rNMPs is exceeded in absence of RER and presence of limited dNTP pools, Top1-mediated genome instability leads to severe growth defects. Finally, we provide evidence showing that accumulation of RNA/DNA hybrids in absence of RNase H1 and RNase H2 leads to cell lethality under Rnr1 depletion.


INTRODUCTION
In eukaryotes, undamaged nuclear DNA is replicated by three members of the B family of DNA polymerases (Pols), Pol ␣, Pol ε and Pol ␦, whose catalytic subunits are Pol1, Pol2 and Pol3, respectively (for a review, see e.g. (1)). Pol ␣-RNA primase complex initiates synthesis of both leading and lagging strands. On the leading strand, Pol ␣ is then replaced by Pol ε, which synthesizes long stretches of DNA in a processive manner. On the lagging strand, Pol ␦ takes over from Pol ␣ and synthesizes Okazaki fragments (henceforth referred to as 'OF'), which are short segments of about 200 nt that are processed and ligated after polymerization (for a review, see e.g. (2)). Recent in vivo analyses in Saccharomyces (S.) cerevisiae and Schizosaccharomyces pombe, and in vitro reports, indicate that Pol ␦ contributes to leading strand synthesis (3)(4)(5)(6)(7)(8)(9).
Pols ε and ␦ are extremely accurate in copying the genome and have high substrate selectivity for the base and sugar components of deoxyribonucleoside tri-phosphates (dNTPs). However, the stringency of selection against the incorporation of ribonucleoside monophosphates (rNMPs) varies among replicative Pols ␣, ε and ␦ ((10); for reviews, see e.g. (11,12)). As cellular rNTP concentrations in eukaryotes are generally one to two orders of magnitude higher than those of the corresponding dNTPs, this potentially affects the frequencies of rNMP incorporation by the replicative Pols (10,13).
In S. cerevisiae, rNTP abundances are relatively constant throughout the cell cycle (14). In contrast, the levels of dNTPs increase 3-6-fold during DNA replication in S-phase in normal/unperturbed growth conditions and a further 3-5-fold upon DNA damage (14,15). The ribonucleotide reductase (henceforth referred to as 'RNR') complex catalyzes the rate-limiting step in de novo dNTP synthesis by reducing ribonucleotides into deoxyribonucleotides and balancing the concentrations of all four dNTPs. In all eukaryotes, the RNR complex is formed by a large subunit R1 that harbors both the catalytic and allosteric sites, and a small subunit R2 that houses the diferric-tyrosyl radical cofactor that is essential for the initiation of nucleotide reduction. In S. cerevisiae, R1 is a homodimer formed of two copies of the major catalyticsubunit Rnr1, and R2 is a heterodimer formed of Rnr2 and Rnr4 (for a review of yeast RNR complex, see e.g. (16)). The expression, activity and localization of the yeast RNR complex are exquisitely regulated during the cell cycle in unperturbed cells, and also in conditions of DNA damage and replicative stress (henceforth, these two conditions are collectively referred to as 'genotoxic stress') (see Supplementary Figure S1). Notably, Rnr3, the minor catalytic-subunit of the yeast RNR complex, is expressed at low levels during the cell cycle in unperturbed cells, but is highly upregulated under genotoxic stress (17).
Single ribonucleotides incorporated in nuclear DNA by Pols can be removed by the error-free Ribonucleotide Excision Repair (RER) pathway. This is initiated by RNase H2, which incises at the scissile phosphate upstream of the rNMP, thereby creating a nick whose ends have a 3 OH and a 5 RNA-DNA junction. The 3 OH end is subsequently extended by the OF maturation machinery ( (18); for a review see e.g. (12)). Genomic ribonucleotides that accumulate in absence of RNase H2 are henceforth referred to as 'unrepaired rNMPs'. Loss of RNase H2-dependent-RER causes DNA damage that leads to embryonic lethality in mice (19)(20)(21), but is tolerated in S. cerevisiae (see e.g. (22,23)). RNase H1, the other major RNase H in eukaryotes, does not play a role in RER (18), as it needs at least four contiguous rN-MPs in DNA for cleavage (for reviews, see e.g. (24,25)). However, both RNase H1 and RNase H2 (henceforth both enzymes are referred to as 'RNases H1 and H2') can process the RNA moiety of RNA/DNA hybrids (henceforth referred to as 'hybrid-removal activity'; for reviews, see e.g. (24,25)), which can be found as part of R-loops in the nuclear and mitochondrial genomes (for R-loops in S. cerevisiae, see e.g. (26)(27)(28)). Notably, transcription-associated-R-loops can block replication fork (henceforth referred to as 'RF') progression, thereby threatening the stability of the genome (for reviews, see e.g. (29,30)).
Single genomic rNMPs can also be cleaved by DNA topoisomerase 1 (Top1), particularly unrepaired rNMPs. Top1 incises the scissile phosphate downstream of a single rNMP in duplex DNA, which could lead to an unligatable ribonucleoside-2 , 3 cyclic phosphate-terminated end (henceforth referred to as 'un-ligatable nick'; for reviews, see e.g. (12,31)). The nick could be reversed by Top1 (32,33), or be repaired by Apn2 and Srs2-Exo1 pathways (34,35). Alternatively, Top1 could make a second incision in the same strand, upstream of the un-ligatable nick, thereby resulting in a short gap, which can either be filled by error-free repair pathways (33), or lead to a deletion of 2-5 bp if the incision occurs within a short tandemly repeated sequence (henceforth the 2-5 bp deletion is referred to as ' 2-5 bp'; see e.g. (32,33,36)). Another possibility is that Top1 could make a second in-cision in the complementary strand, opposite to the unligatable nick, thereby creating a DNA double strand break (DSB) that can either be repaired by the cellular Rad51/52dependent homologous recombination (HR) machinery, or lead to Top1-mediated illegitimate recombinations (37). Top1-mediated DNA damage at sites of single genomic rN-MPs is henceforth referred to as 'Top1-mediated RNA-DNA damage'.
In this study, we sought to deplete Rnr1 in S. cerevisiae in order to analyze the consequences of reduced dNTP pools on genome integrity and cell viability of mutants lacking RNase H1, RNase H2, or both enzymes. We found that the removal of RNA/DNA hybrids by RNases H1 and H2 is essential for the growth of Rnr1-depleted cells. Importantly, we found that single genomic rNMPs are highly enriched in double mutants lacking both Rnr1 and RNase H2. This was further exacerbated in triple mutants that are depleted of Rnr1, lack RNase H2, and also harbor a steric gate replicative Pol variant with reduced discrimination against utilization of rNTPs as compared to its WT parent enzyme (henceforth referred to as 'rNTP-permissive Pol'). Furthermore, our Southern blotting data led us to infer that, in cells depleted of Rnr1 and lacking RNase H2, Top1-mediated cleavages occur in both the leading and lagging strands when rNMPs are excessively incorporated by an rNTP-permissive form of Pol ␦ or ␣; but only in the leading strand by an rNTP-permissive form of Pol ε. Accordingly, triple mutants that are depleted of Rnr1, lack RNase H2, and harbor an rNTP-permissive form of Pol ε or ␦ showed severe growth defects that are likely to be caused by deleterious Top1-mediated RNA-DNA damage. Finally, we provide evidence to support the proposed role of Pol ␦ in leading strand synthesis (here particularly observed under replicative stress induced by Rnr1 depletion), in addition to its major role in lagging strand synthesis.

Drop test growth assays
Strains harboring the RNR1 gene under the control of its native promoter were pre-grown in liquid medium containing YEPD (medium contains 1% yeast extract [Formedium YEA02], 2% bacto-peptone [Formedium PEP02], and 2% dextrose [Sigma-Aldrich D9434]). Cells were spotted on solid medium containing either YEPD with 2% agar (Formedium AGA02), or YPGS with 2% agar (composition of YPGS is as for YEP but supplemented with 2% galactose [Sigma-Aldrich G0750] and 1% sucrose [Sigma-Aldrich 84097]). Strains carrying P GAL :3HA-RNR1 were pre-grown in liquid medium containing YPGS, which is permissive for Rnr1 expression. Cells were spotted on solid medium containing either YEPD, which is non-permissive for Rnr1 expression, or YPGS. Strains carrying P GAL :3HA-RNR1 together with a plasmid were pre-grown in liquid minimal medium lacking leucine with 2% galactose and 1% sucrose, which is permissive for Rnr1 expression (medium contains SD-Leu-Glucose [Sunrise Science Products 1799; 'SD-Leu-Glucose' stands for 'synthetic defined minus leucine minus glucose'], galactose and sucrose). Cells were spotted on solid minimal medium lacking leucine, with either 2% glucose, which is none-permissive for Rnr1 expression (medium contains SD-Leu [Sunrise Science Products 1707; glucose included]), or 2% galactose and 1% sucrose. 10-fold dilutions of overnight, saturated liquid cultures were spotted on the plates, starting from 0.4 OD 600 of yeast cells.

CAN1 forward mutation assay
CAN1 forward mutation assay was performed according to (40), with modifications. Briefly, strains were streaked out on solid medium containing either YEPD (for strains WT and rnh201Δ), or YPGS (for strains P GAL :3HA-RNR1 and P GAL :3HA-RNR1 rnh201Δ). For composition of YEPD and YPGS media see section 'Drop test growth assays'. Plates were incubated for 2 days at 30 • C until single colonies appeared. Then, 12-20 patches were made on solid medium containing YEPD (1 single colony per patch). After incubation at 30 • C for 24 h (for strains WT and rnh201Δ), or 48 h (for strains P GAL :3HA-RNR1 and P GAL :3HA-RNR1 rnh201Δ), each patch was re-suspended in 500 l sterile water. To detect Can resistant (Can R ) colonies, an aliquot from the cell suspension was plated on solid media supplemented with 60 mg l −1 L-canavanine (Sigma-Aldrich C9758), as follows: (i) For strains WT and rnh201Δ, 200 l cells were plated on minimal medium lacking arginine with 2% glucose (DOBA [Sunrise Science Products 1651; this medium contains glucose] and CSM-Arg [Sunrise Science Products 1031]). (ii) For strains P GAL :3HA-RNR1 and P GAL :3HA-RNR1 rnh201Δ, 200 and 100 l cells, respectively, were plated on minimal medium lacking arginine with 2% galactose and 1% sucrose (DOBA-glucose w/2% galactose [Sunrise Science Products 1653; this medium does not contain glucose], CSM-Arg and sucrose). For each strain, four independent experiments were performed, each including 12-20 patches. Total mutation rates and 95% confidence intervals were calculated for each independent experiment by the Lea and Coulson method of median (41,42), using a template kindly provided by Nayun Kim (University of Texas Health Science Center at Houston) (see (42)). To determine CAN1 specific mutation rates, one Can R colony was randomly picked from each plate and re-suspended in 50 l water. A 20 l aliquot of cell suspension was used for PCR amplification of CAN1 using Herculase II Fusion DNA polymerase (Agilent Technologies 600679). For primers used for PCR and sequencing of CAN1 see Supplementary Table S2. Specific mutation rates were calculated according to (43).

Growth conditions for Rnr1 depletion in liquid media
Day1: Strains carrying P GAL :3HA-RNR1 were pre-grown overnight at 30 • C in liquid minimal medium lacking histidine with 2% galactose and 2% sucrose, which is permissive for Rnr1 expression (medium contains yeast nitrogen base without amino acids and with ammonium sulphate [Formedium CYN0410], synthetic complete mixture Kaiser drop-out -His [Formedium DSCK1003], galactose [Acros , and sucrose [Fisher 57-50-1]). Day 2: In the morning, saturated pre-cultures were diluted with the same medium to OD 600 ∼0.05. Growth was maintained in exponential phase by dilution with the same pre-warmed medium for 24 h. Day 3: In the morning, cells were harvested at OD 600 ∼0.2, then washed with pre-warmed liquid minimal medium lacking histidine with 2% glucose, which is non-permissive for Rnr1 expression (same medium composition as above but supplemented with glucose [Fisher 50-99-7] as the sole carbon source). Cells were subsequently re-suspended in the same pre-warmed medium containing glucose to OD 600 ∼0.1-0.2. Growth was maintained in exponential phase by dilution with the same pre-warmed medium containing glucose. Cells were collected at the indicated time points. Note that for Figure 1B, D and E, and Supplementary Figures S2, S3B and S8-S10, an aliquot of exponentially growing cells was also collected from medium containing galactose plus sucrose, in which Rnr1 is moderately over-expressed.

Growth conditions in absence or presence of drugs in liquid media
WT cells (BY4741) were pre-grown overnight in YEPD medium at 30 • C (for composition of YEPD medium see section 'Drop test growth assays'). The next morning, saturated pre-cultures were diluted to OD 600 ∼0.05 in the same medium. When OD 600 reached ∼0.3, cells were split in three portions: one for control in absence of drugs, one for treatment with 200 mM hydroxyurea (HU; Acros Organics 127-07-1), and one for treatment with 0.03% methyl methane sulfonate (MMS; Sigma 129925). Control cells were harvested at OD 600 ∼0.5-0.6. Cells treated with HU or MMS were kept in exponential phase by dilution in the same medium and finally harvested at OD 600 ∼0.5-0.6 after 3 h in presence of the drug.

Western-blotting
Total protein extraction from ∼5 OD 600 of yeast cells was performed by NaOH lysis and trichloroacetic acid (TCA) precipitation according to (44), with minor modifications. Dissolved cell pellets (50 l) were heated at 95 • C for 10 min, and then spun for 10 min at 10 000 g at room temperature. Protein extracts (10% volume of the supernatant) were then resolved, together with a protein ladder (SeeBlue Plus2 Pre-stained Standard, ThermoFisher Scientific LC5925), by SDS-PAGE (4-20% Mini-Protean TGX Precast gel [Bio-Rad 456-1096] in Figure 1B, and standard 6.5% SDS-polyacrylamide gel in Figure 1D and Supplementary Figure S3B). Proteins were electro-transferred from the gel onto a nitrocellulose membrane (Thermo Fisher Scientific 88018). The membrane was sequentially treated as follows: Step 1: Stained briefly with Ponceau Red and then washed with distilled water. Step 2: Blocked for 30 min at room temperature in 1× PBS plus 0.1% tween (referred to as 'PBST'; 1× PBS contains 137 mM NaCl, 2.7 mM KCl, 8 mM Na 2 HPO 4 and 2 mM KH 2 PO 4 ) with 5% (w/v) milk (Skim milk powder, OXOID LP0031).
Step 3: Incubated with primary antibody in PBST with 5% milk for 1 h at room temperature followed by overnight incubation at 4 • C.
Step 4: Washed with PBST for 3 × 10 min at room temperature.
Step 5: Incubated with secondary antibody in PBST with 5% milk for 1 h at room temperature.
Step 6: Washed with PBST for 3 × 10 min at room temperature.

Fluorescence-activated cell sorting (FACS) analysis
FACS was performed essentially as described in (45), with minor modifications. Propidium-iodide-stained cells were sonicated for 2 × 10 s at 4 • C (Sonicator Bioruptor PICO, Diagenode) and subsequently analyzed using a flow cytometer. DNA profiles were generated using the FLOWJO software.
Step 2: 700 l GTC-phenol were added and the whole mixture was vortexed briefly, then incubated at 65 • C for 5 min, and subsequently cooled down on ice.
Step 6: 450 l (from the upper phase) were mixed with 1 l glycoblue co-precipitant (Thermofisher Scientific AM9515) and 1 ml 100% ethanol, then incubated for 1 h at −80 • C, and subsequently spun as described above.
Step 7: The RNA pellet was washed once with 1 ml 70% ethanol, then 'air-dried', and finally resuspended in distilled water.
RT-qPCR reactions were performed as follows: Step 1: To digest genomic DNA, an aliquot of total RNA (∼30 g) was incubated with 7 u RQ1 DNase (Promega M6101) and 40 u ribonuclease inhibitor RNasin (Promega N251 A), in a total volume of 60 l, at 37 • C for 30 min. DNA-free RNA was extracted with phenol:chloroform:isoamyl alcohol as described above in steps 4-7.
Step 2: An aliquot of DNAfree total RNA (∼1 g) was incubated with 1 l of 0.2 g l −1 random hexamers (Thermofisher Scientific SO142) and 1 l of 10 mM dNTPs (equimolar mixture of dATP, dCTP, dGTP and dTTP), in a total volume of 14.25 l, at 65 • C for 5 min, and subsequently cooled down on ice.
Step 3: For RT reaction, the mixture from the previous step was incubated with 100 u Superscript III Reverse Transcriptase (Thermo Fisher Scientific 18080093), 1× first-strand buffer, 5 mM DTT and 10 u RNasin, in a total volume of 20 l, at 25 • C for 15 min, and then incubated at 50 • C for 1 h. RT reaction was stopped by heating at 70 • C for 15 min.
Step 4: qPCR reactions were performed in triplicate, using 4 l of 10-fold dilution of complementary DNAs (from previous step) with 1× SYBR premix (TB Green® Premix Ex Taq™ II [Tli RNase H Plus], Takara Bio Europe RR820W) and 0.4 M primers, in a total volume of 10 l, as described previously (46). For primer sequences see Supplementary  Table S2. To generate RT-qPCR data, the average of triplicates of Ct values was used in the formula Ct = 2(Ct 'target mRNA' -Ct 'ACT1 mRNA control').

Detection of genomic ribonucleotides by alkaline-gel electrophoresis combined with Southern hybridization
Detection of ribonucleotides in genomic DNA by alkalinegel electrophoresis was performed according to (48), with some modifications. Briefly, total DNA was extracted from ∼50-100 OD 600 of yeast cells with MasterPure™ Yeast DNA Purification Kit (Epicentre-Lucigen/Cambio.co.uk MPY80200), by omitting RNase A (included in the kit) from cell lysis step. Total DNA was treated with 0.14 g l −1 RNase A in 1× TE at 37 • C for 30 min (RNase A with no/low salt degrades single-stranded RNA, doublestranded RNA, the RNA moiety of RNA/DNA hybrids and genome-embedded single ribonucleotides; e.g. see (46,(49)(50)(51)). An aliquot of total DNA (∼2 g) was heated in presence of alkali (0.3 M KOH) at 55 • C for 2 h. Alkali-denatured-total DNA samples were separated, together with the DNA ladder (1 kb plus DNA ladder, Invitrogen 10787018), on an alkaline (50 mM NaOH and 1 mM EDTA, pH 8.0) 1% agarose gel (length 15.5 cm), in alkaline electrophoresis buffer (50 mM NaOH and 1 mM EDTA, pH 8.0), at 1 V cm −1 , for ∼18-22 h, at room temperature, using Owl separation system model A2 (Thermo Fisher scientific). Note that the buffer was allowed to recirculate using a pump at low flow rate setting (KNF Lab Liquiport 100) to avoid heating during alkaline-gel electrophoresis. The gel was washed in neutralization buffer I (1 M Tris-HCl and 1.5 M NaCl) and then washed briefly in deionized water. The gel was stained for 1 h with 1× SYBR gold (Thermo Fisher scientific S11494) in 0.5× TE, and then washed for 2 × 30 min in 0.5× TE, in a light-protected container, with gentle shaking. SYBR-stained, alkali-fragments from total DNA (henceforth referred to as 'Afts') were visualized using Fuji FLA-5100 PhosphorImager. Raw densitometry of SYBR-staining signal was obtained by using AIDA Image Analyzer v.4.15 densitometry software. For the determination of the numbers of total genomic rNMPs see the section 'Quantitation of genomic ribonucleotides'. Southern-blotting was performed according to (48), with some modifications. Briefly, the gel from the previous step (with ∼5 g total DNA in each lane) was washed in alkaline transfer buffer (0.4 N NaOH and 1 M NaCl) for 20 min, and then capillary-transferred onto a nylon Hybond-N+ membrane (GE Healthcare RPN203B), in alkaline transfer buffer, at room temperature, overnight. The membrane was washed in neutralization buffer II (0.5 M Tris-HCl, pH 7.2 and 1 M NaCl), and then DNA was immobilized to the membrane by UV-crosslinking (120 mJ/cm 2 ; UV Stratalinker 1800, Stratagene). Strand-specific singlestranded probes were synthesized by PCR using 1 single primer with the AGP1 double-stranded PCR amplicon as template, in the presence of ␣-32P-dCTP (for the sequences of primers, see Supplementary Table S2). After 16-24 h hybridization at 65 • C, the membrane was washed and exposed to a phosphor imaging screen. Raw densitometry of radioactivity signal was obtained with AIDA Image Analyzer v.4.15 densitometry software.

Quantitation of genomic ribonucleotides
Ribonucleotide incorporation abundances were estimated using a slightly modified version of the method described previously in (20). Raw densitometric histograms of SYBRstained Afts were obtained as described in section 'Detection of genomic ribonucleotides by alkaline-gel electrophoresis combined with Southern hybridization'. After subtracting the background intensity, a smoothing spline with 15 optimally placed internal knots was applied to each lane of the gel by running the SLM tool (D'Errico, 2017*) in Matlab version 9.2. The smoothened intensity curves were resampled at intervals of width d = 1mm. For each interval, the characteristic fragment size (sz) was calculated following the equation sz = exp((d − a)/b), where d represents the electrophoretic distance in the middle point of the interval. Parameters a and b were inferred by fitting the linear model lm(d ∼ log(sz)) to the peaks of the size reference lane (i.e. 1 Kb plus DNA ladder, Invitrogen 10787018). The fragment count associated with each interval was estimated as n sz = I sz /sz, where I sz is the smoothened densitometric intensity in that interval. To make the results independent of the total amount of DNA loaded in the lane, the fragment count per size interval per 1Gb of total genomic DNA was obtained as n sz ( per 1Gb) = n sz × 10 9 / (sz n sz ), where the sum extends over all size intervals ( Figure 5B). Note that the choice of 1Gb as the unit of measurement is arbitrary and does not change the results by any means. Because the conversion from densitometry intensity to fragment count is highly sensitive to small, noisy fluctuations in the far bottom end of the electrophoretic gel, a cutoff at an electrophoretic distance d max was introduced. The value of d max was determined under supervision, as the point where fluctuations in the original (non-smoothened) intensities of the loaded lanes became similar in magnitude to those observed in empty lanes, indicating a poor signal-to-noise ratio. Setting a distance cutoff indirectly defined a minimum detectable fragment size equal to s z mi n = exp((d max − a)/b).
Binned distribution of fragment sizes. Binned distributions of fragment sizes were obtained by adding the fragment counts per size interval per 1Gb in bins covering 50 nucleotides (nt) each. The distributions were normalized by dividing the value in each bin by the sum of values in all bins. To account for the fact that sz mi n cutoffs differ across gels, all comparisons between gels were restricted to bins spanning fragment sizes above sz mi n . This sub-section is part of section 'Quantitation of genomic ribonucleotides' and is related to Supplementary Figure S4.
Estimate of numbers of total genomic rNMPs. A preliminary estimate of the number of ribonucleotides per genome, N, was obtained by adding, for all intervals, the fragment counts per 1Gb, dividing by 10 9 and multiplying by the size of the yeast haploid genome (∼24 Mb).
Nucleic Acids Research, 2020, Vol. 48, No. 8 4279 To account for small fragments that had migrated beyond d max (i.e. with sizes below sz mi n ), the total fragment count was corrected under the assumption that break points are randomly distributed with uniform probability along the genome. Thus, the corrected estimate of the total number of ribonucleotides per genome became N corr = N × (2 − exp{−sz mi n × n sz ( per 1Gb)/10 9 }), where the sum extends over all intervals. Note that if the distribution of break points is not uniform along the genome (52), this formula provides a conservative estimate for (i.e. it does not overestimate) the total number of genomic rNMPs. This sub-section is part of section 'Quantitation of genomic ribonucleotides' and is related to Figure 5C and Supplementary Table S5.
Calculation of the contributions of replicative Pols α, δ and ε to synthesis of S. cerevisiae nuclear genome. The percentage of contribution of each replicative Pol was calculated by applying the mathematical formula that we adapted from Reijns et al. (53) (see also Supplementary Table S6 (Reprinted by permission from Copyright Clearance Centre: Springer Nature; Nature; Lagging-strand replication shapes the mutational landscape of the genome; Martin A.M. . 'N Polx ' represents the subtraction of the number of total genomic rNMPs incorporated in vivo by an rNTP-permissive Polx ('x' indicates ␣-L868M, ␦-L612M or ε-M644G) in a given strain lacking RNase H2, from the number of total genomic rNMPs of the corresponding strain bearing the three WT replicative Pols (␣, ␦ and ε) and lacking RNase H2, within the same gel. 'F Polx ' represents the in vitro frequency of rNMP incorporation by Polx, i.e. 1 rNMP per 40, 100 and 300 dNMPs for Pols ␣-L868M, ε-M644G, and ␦-L612M, respectively (frequencies from (23,54)). This sub-section is part of section 'Quantitation of genomic ribonucleotides' and is related to Supplementary Figure S6 and Supplementary Table S6.

Resolution of formamide-denatured genomic DNA on neutral gel
The protocol was adapted from (20,49), with some modifications. Total DNA was extracted from ∼50 OD 600 of yeast cells with MasterPure™ Yeast DNA Purification Kit, as described in section 'Detection of genomic ribonucleotides by alkaline-gel electrophoresis combined with Southern hybridization'. An aliquot of total DNA (∼30 g) was treated with 0.02 g l −1 RNase A in 1× TE with high salt (0.5 M NaCl), in a total volume of 175 l, at 25 • C for 1 h (RNase A with high salt degrades selectively single-stranded RNA, while avoiding degradation of double-stranded RNA, the RNA moiety of RNA/DNA hybrids and genome-embedded single ribonucleotides; see e.g. (46,(49)(50)(51)). DNA was purified with an equal volume of AMPure XP beads (Beckman Coulter A63880). DNA aliquots (∼0.5 g) were incubated in 1× ThermoPol buffer (New England Biolabs B9004S), either in absence or presence of 25 u of recombinant E. coli RNase HII (New England Biolabs M0288S), or in presence of both 25 u RNase HII and 0.1 g l −1 RNase A, in a total volume of 50 l, at 37 • C for 2 h. As a control for DNA fragmentation, DNA aliquots (∼0.5 g) were incubated in 1× CutSmart buffer in presence of 2.5 u of Nb.BtsI (New England Biolabs R0707S), either in the absence or presence of 0.1 g l −1 RNase A, in a total volume of 50 l, at 37 • C for 1 h. DNA was extracted with phenol:chloroform:isoamyl alcohol as described in steps 4-7 in section 'Reverse transcription of total RNA combined with quantitative PCR (RT-qPCR)'. DNA was re-suspended in 2 l water and then incubated in presence of 90% formamide and 20 mM EDTA, pH 8, in a total volume of 25 l, at 37 • C for 1 h. Formamide-denatured DNA samples, together with the DNA ladder, were separated by neutral gel-electrophoresis at ∼5.7 V cm −1 , for 4.5 h, at room temperature, with recirculation of buffer (1% agarose gel in 1× TBE; length of gel 15.5 cm). The gel was subsequently stained with SYBR gold. For other details, see section 'Detection of genomic ribonucleotides by alkaline-gel electrophoresis combined with Southern hybridization'.

Depletion of Rnr1 mildly induces the S-phase checkpoint, greatly reduces dNTP levels and significantly slows cell growth in S-phase
Deletion of the RNR1 gene is not lethal in the S. cerevisiae BY4741/SC288 background. However, rnr1Δ mutants are slow growing, relative to the otherwise isogenic wild-type (WT) and suffer from both limited and imbalanced dNTP pools (55,56). Spontaneous suppressor mutations arise in the gene CRT1, whose product represses the transcription of the genes RNR2-4 and HUG1 during the cell cycle in unperturbed cells ((57,58); see also Supplementary Figure S1). These can reverse the growth defects in rnr1Δ strains, likely due to the expansion of dNTP pools (56). To avoid selection for crt1 suppressors, we constructed the strain P GAL :3HA-RNR1, in which the RNR1 gene is under the control of the P GAL1/10 promoter (59,60). We also constructed the strain P GAL :3HA-RNR1 crt1Δ lacking Crt1. 3HA-Rnr1 expression can be either induced under permissive conditions in galactose-containing medium (plus sucrose to limit Rnr1 over-expression and facilitate yeast growth) or repressed under non-permissive conditions in glucose-containing medium ( Figure 1A). We next determined the effects of Rnr1 depletion on the DNA damage and replication checkpoint (henceforth referred to as 'Sphase checkpoint'; see Supplementary Figure S1), dNTP levels, cell cycle progression and cell growth.
RT-qPCR analyses showed elevated levels of RNR1 mRNA under permissive conditions (0 h time-point) in both the single mutant P GAL :3HA-RNR1 and double mutant P GAL :3HA-RNR1 crt1Δ, relative to the WT strain and single mutant crt1Δ, which were both cultured in rich YEPD medium (Supplementary Figure S2A, compare lanes e and i with a and d). Following transfer of P GAL :3HA-RNR1 or P GAL :3HA-RNR1 crt1Δ to glucose medium, RNR1 transcripts were greatly decreased by 2 h (Supplementary Figure S2A, lanes e-h and i-l). Consistent with this, Western blotting ( Figure 1B) showed robust depletion of 3HA-Rnr1 protein following transfer of P GAL :3HA-RNR1 or P GAL :3HA-RNR1 crt1Δ strains to The promoter is induced in galactose plus sucrose (gal + suc) media and inhibited in glucose (glu) media. Rnr1 is epitope-tagged with 3x hemagglutinin (3HA) at its N-terminus. (B) Rnr3 protein is mildly expressed in single mutant P GAL :3HA-RNR1 depleted of Rnr1. Strains P GAL :3HA-RNR1 and P GAL :3HA-RNR1 crt1Δ were grown at 30 • C in liquid minimal medium lacking histidine with 2% galactose and 2% sucrose. To trigger Rnr1 depletion, cells at OD 600 ∼0.2 were transferred to liquid minimal medium lacking histidine with 2% glucose. Cells were harvested before transfer (0 h) and 2, 4 and 6 h after transfer to glucose-containing medium (see also Materials and Methods). Strain WT was grown in rich YEPD (2% glucose) medium at 30 • C, in absence of drugs (i.e. unperturbed conditions), or for 3 h in presence of either 200 mM HU (labeled +HU) or 0.03% MMS (labeled +MMS) (see also Materials and Methods). Strain crt1Δ was grown in rich YEPD (2% glucose) medium at 30 • C. Note that all cell samples were harvested in exponential phase. Total proteins were separated on a 4-20% SDS-polyacrylamide gel and then electro-blotted. The filter was stained with Ponceau Red (bottom sub-panel). The same filter was probed separately with antibodies against HA tag (3HA-Rnr1), Rnr3, PGK1 and Sml1. Relevant protein molecular weights (KDa) are indicated at the left-hand. For the ease of comparison, each well is allocated a unique Latin letter, which is repeated in each sub-panel. The length of Rnr1 depletion in hours (hr) is indicated above the wells a-d and e-h. One representative experiment is shown of at least three independent ones. (C) dNTP levels, particularly dGTPs, are greatly decreased in single mutant P GAL :3HA-RNR1 depleted of Rnr1. The WT strain was cultured in rich YEPD medium (2% glucose) at 30 • C and harvested at OD 600 ∼0.4. The single mutant P GAL :3HA-RNR1 was cultured as explained in (B) and harvested 6 h after transfer to glucose-containing media at OD 600 ∼0.4. dNTP levels were normalized to rNTP levels and values were adjusted to the total number of cells used for the preparation. Error bars reflect S.E.M. of 2 independent repeats. Symbols on the organigram: + and -indicate that Rnr1 is present or absent, respectively. For the ease of comparison, WT strain and single mutant glucose medium ( Figure 1B, 3HA-Rnr1, lanes a-d and eh).
To induce genotoxic stress, we treated WT cells with hydroxyurea (HU) or methyl-methane sulfonate (MMS). HU inhibits RNR activity by scavenging the tyrosyl free radical in Rnr2, thereby slowing DNA synthesis (see e.g. (60)), while MMS is a DNA alkylating agent that leads to RF arrests (see e.g. (61)). These genotoxic agents led to increased RNR1 mRNA levels (Supplementary Figure S2A, compare lanes b and c with a), as previously reported for Rnr1 protein (62).
Depletion of Rnr1 in strain P GAL :3HA-RNR1 after 6 h transfer to glucose-containing medium reduced the dNTP pools >3-fold, particularly the levels of dGTP, as compared to the WT strain ( Figure 1C, Supplementary Figure S3A and Supplementary Table S4). These results are consistent with previously published data for mutant rnr1Δ (56). Limited dNTP pools in the P GAL :3HA-RNR1 strain depleted of Rnr1 are predicted to slow the progression of RFs. RF stalls would in turn trigger the activation of the S-phase checkpoint kinase cascade Mec1-Rad53-Dun1. Activation of this checkpoint can be monitored by western-blotting analysis of Rad53 phosphorylation (phospho-Rad53), visible via reduced electrophoretic mobility (see e.g. (63)). Following depletion of Rnr1 for 6 h, total protein extracts from P GAL :3HA-RNR1 strain showed a noticeable phospho-Rad53 mobility upshift, but this was less marked in the double mutant P GAL :3HA-RNR1 crt1Δ ( Figure 1D, compare lanes a-d with e-h). Phospho-Rad53 was virtually absent in the WT and crt1Δ strains but was strongly induced by treatment of the WT strain with HU or MMS ( Figure  1D, compare lanes i and l with j and k).
Activation of the S-phase checkpoint kinase cascade Mec1-Rad53-Dun1 under genotoxic stress leads to Dun1mediated-inhibition of Crt1, thereby leading to the upregulation of the expression of RNR2-4 and HUG1 genes ( (57,58); see also Supplementary Figure S1). Depletion of Rnr1 in P GAL :3HA-RNR1 strain increased the levels of these mRNAs but, except for RNR2, mRNA levels were lower than in crt1Δ or P GAL :3HA-RNR1 crt1Δ at all timepoints (Supplementary Figure S2A and B, compare lanes e-h with i-l and d). In WT cells, RNR3 and HUG1 mR-NAs were virtually absent, and RNR2 and RNR4 mRNAs were expressed at low levels; however, RNR2-4 and HUG1 mRNAs were all strongly induced by genotoxic stress following treatment with HU or MMS (Supplementary Figure  S2A and B, compare lanes b and c with a). There was good concordance between the abundance of RNR3 mRNA and Rnr3 protein, which was mildly elevated following Rnr1 depletion in P GAL :3HA-RNR1 strain, and strongly elevated in strains lacking Crt1 or the WT strain treated with HU or MMS ( Figure 1B, Rnr3, lanes d, e-h, l and j-k; compare with Supplementary Figure S2A, lanes h, i-l, d and b-c, respectively). The phosphorylation of Rad53 and the induced expression of the RNR3 and HUG1 genes both indicate that the S-phase checkpoint is modestly activated in P GAL :3HA-RNR1 strain depleted of Rnr1.
Activation of the S-phase checkpoint kinase cascade Mec1-Rad53-Dun1 in unperturbed cells that are replicating their DNA, or in cells that are under genotoxic stress, leads to Dun1-mediated-degradation of Sml1, which is the protein repressor of Rnr1 (see Supplementary Figure S1). Conversely, increased Rnr1-Sml1 association due to overexpression of Rnr1 stabilizes Sml1 (64,65). Western blotting showed that, as expected, Sml1 totally disappeared from WT cells following treatment with HU or MMS ( Figure  1B, Sml1, compare lanes j and k with i). In addition, Sml1 was lost upon depletion of Rnr1 in P GAL :3HA-RNR1 strain ( Figure 1B, Sml1, lanes a-d), as previously reported for rnr1Δ strain (55). Moreover, Sml1 was degraded upon depletion of Rnr1 in P GAL :3HA-RNR1 crt1Δ strain ( Figure  1B, Sml1, lanes e-h). Together, these results suggest that depletion of Rnr1 led to the disappearance of Sml1 in both strains P GAL :3HA-RNR1 and P GAL :3HA-RNR1 crt1Δ.
Fluorescence-activated cell sorting (FACS) analysis showed that cells from single mutant P GAL :3HA-RNR1 significantly accumulated in S-phase at 6 h depletion of Rnr1, as revealed by the peak between the 1C and 2C positions (lower-part of Figure 1E), which is consistent with a previous report that analyzed rnr1Δ cells (55). Loss of Crt1 in the P GAL :3HA-RNR1 crt1Δ double mutant strain, however, substantially reduced cell accumulation in S-phase after depletion of Rnr1 for 6 h (upper-part of Figure 1E).
Finally, in drop test growth assays, P GAL :3HA-RNR1 strain showed WT growth in galactose plus sucrose medium, but grew slower than the WT strain in glucose medium ( Figure 1F, compare rows e with a). In contrast, growth of P GAL :3HA-RNR1 crt1Δ strain was similar to the WT strain and the single mutant crt1Δ in glucose medium ( Figure 1F, compare rows f with a and b). Loss of Sml1 did not improve the growth of the double mutant P GAL :3HA-RNR1 sml1Δ relative to the single mutant P GAL :3HA-RNR1 in glucose medium, which is in accordance with our western blotting results showing that Sml1 protein is de-  Table S4). (D) Rad53 is mildly phosphorylated in single mutant P GAL :3HA-RNR1 depleted of Rnr1. Strains and growth conditions are as in (B). Total proteins were separated on a 6.5% SDS-polyacrylamide gel. The Filter was probed with antibody against Rad53 (P-Rad53 represents the phosphorylated form of Rad53). For other details, see (B). (E) Cells depleted of Rnr1 accumulate significantly in S-phase in the presence of Crt1. Flow cytometry (FACS) histograms of cells from strains P GAL :3HA-RNR1 and P GAL :3HA-RNR1 crt1Δ, before (0 h), and 2, 4 and 6 h after transfer to glucose medium. One representative experiment is shown of at least three independent ones. Note that cell samples from the same cultures were used for FACS, western-blotting in (B) and (D), and RT-qPCR in Supplementary Figure S2. (F) Cells depleted of Rnr1 grow much slower than the WT strain, but their growth is fully restored in the absence of Crt1. Drop test growth assays of strain WT, strains deleted of the gene CRT1, SML1, or DUN1, and strains carrying P GAL :3HA-RNR1 without gene deletion (labeled "none"), or with deletion of the gene CRT1, SML1 or DUN1. Cells were pre-grown in YPGS (2% galactose and 1% sucrose) liquid medium overnight. Serial dilutions were plated on YEPD (2% glucose) and YPGS solid media. Plates were incubated at 25 • C. Photographs were taken at the indicated number of days (d).
For the ease of comparison, a unique Latin alphabet letter is allocated for each row. 'Glu' stands for glucose and 'Gal + Suc' stands for galactose plus sucrose. The horizontal line across the images is included for clarity. See Supplementary Table S1 for the list of strains. One representative experiment is shown of at least three independent ones. in mutants lacking RNase H1, RNase H2 or both enzymes has different effects on their growth. Drop test growth assays of strain WT, and strains carrying P GAL :3HA-RNR1 without gene deletion (labeled "none"), or with deletion of the gene RNH1, RNH201, RNH202, or RNH203, or both genes RNH1 and RNH201, or RNH1 and RNH202, or RNH1 and RNH203. Cells were grown in YPGS (2% galactose and 1% sucrose) liquid medium overnight at 30 • C. Serial dilutions were plated on YEPD (2% glucose) and YPGS solid media. Plates were incubated at 30 • C. Photographs were taken at the indicated number of days (d). 'Glu' stands for glucose. 'Gal + Suc' stands for galactose plus sucrose. The horizontal line across the images is included for clarity. See Supplementary Table S1 for the list of strains. For the ease of comparison, a unique Latin alphabet letter is allocated for each row. One representative experiment is shown of at least three independent graded in the single mutant P GAL :3HA-RNR1 upon depletion of Rnr1 (compare Figure 1F, rows e and g with Figure  1B, Sml1, lanes a-d). The double mutant P GAL :3HA-RNR1 dun1Δ was non-viable on glucose medium ( Figure 1F, row h), suggesting that induced expression of Rnr3 via the activated S-phase checkpoint kinase cascade Mec1-Rad53-Dun1 (see Supplementary Figure S1) is essential for the survival of single mutant P GAL :3HA-RNR1 depleted of Rnr1. This result is consistent with a previous report showing that Dun1 is essential for the viability of rnr1 hypomorphic mutants with limited dNTP pools (66).
Collectively, these results indicate that depletion of Rnr1 mildly activates the S-phase checkpoint, greatly reduces and imbalances dNTP levels, and significantly slows S-phase progression and cell growth. Constitutive replicative stress in Rnr1-depleted strains is likely to reflect a combination of limited and imbalanced dNTP pools, as previously reported for rnr1 hypomorphic mutants (66,67). The additional loss of Crt1 in Rnr1-depleted cells should expand and balance dNTP pools, as previously reported for the double mutant rnr1Δ crt1Δ (56), which would mitigate replicative stress and restore cell growth.

Triple mutants depleted of Rnr1 and lacking RNases H1 and H2 are non-viable, but cell growth is restored by the presence of Rnh201-RED
We hypothesized that reduced dNTP pools in cells depleted of Rnr1 would increase the load of genome-embedded single rNMPs in mutants lacking the RNase H2-dependent-RER pathway, thereby compromising genome stability and cell growth. We further hypothesized that accumulation of persistent RNA/DNA hybrids (e.g. R-loops) in absence of RNases H1 and H2 in cells depleted of Rnr1 would aggravate replicative stress and compromise genomic integrity and cell viability. In principal, RER activity, hybrid-removal activity, or both RNase H activities might be important for growth of cells depleted of Rnr1. To assess this, we deleted one of the genes encoding for the heterotrimeric enzymatic complex RNase H2 (yeast RNase H2 is formed of the catalytic subunit Rnh201 and the accessory subunits Rnh202 and Rnh203 (68)), and/or the gene encoding for the monomeric enzyme RNase H1 in strains carrying P GAL :3HA-RNR1. We then performed drop test growth assays to determine viability.
The variant Rnh201-G42S suppresses less well the growth defects of Rnr1-depleted triple mutants lacking RNases H1 and H2 than the variant Rnh201-RED. Drop test growth assays of strains P GAL :3HA-RNR1 rnh201Δ and P GAL :3HA-RNR1 rnh201Δ rnh1Δ that have an empty vector, or a plasmid expressing WT Rnh201, variant Rnh201-G42S, or variant Rnh201-RED. Cells were grown overnight in liquid minimal medium lacking leucine with 2% galactose and 1% sucrose at 30 • C. Serial dilutions were plated on solid minimal medium lacking leucine with either 2% glucose, or 2% galactose and 1% sucrose. Photographs were taken after 7 days of incubation at 30 • C. For other details, see (A). (C) Cells depleted of Rnr1 and lacking RNases H1 and H2 grow like the WT strain in absence of Crt1. Drop test growth assays of strain WT, and strains carrying P GAL :3HA-RNR1 without gene deletion (labeled "none"), or with deletion of the gene RNH201 or CRT1, or both genes RNH201 and CRT1, or RNH1 and RNH201, or the three genes RNH1, RNH201 and CRT1. For other details, see (A).
Nucleic Acids Research, 2020, Vol. 48, No. 8 4283 We found that on glucose medium (Figure 2A): (i) The three double mutants carrying P GAL :3HA-RNR1 together with rnh201Δ, rnh202Δ or rnh203Δ grew slower than the single mutant P GAL :3HA-RNR1 (compare rows d-f with b). (ii) The double mutant P GAL :3HA-RNR1 rnh1 grew similarly to the single mutant P GAL :3HA-RNR1 (compare rows c with b). (iii) The three triple mutants carrying P GAL :3HA-RNR1 rnh1Δ together with rnh201Δ, rnh202 or rnh203 did not grow at all (rows g-i). This led us to infer that loss of RNases H1 and H2 in cells depleted of Rnr1 induced synthetic lethality. Finally, strains carrying P GAL :3HA-RNR1 without RNase H1, or RNase H2, or both enzymes grew similarly to the WT and single mutant P GAL :3HA-RNR1 in galactose plus sucrose medium, showing that the lack of one or both of these enzymes does not clearly affect cell growth in the presence of Rnr1, which is consistent with previous reports (see e.g. (27,69)) ( Figure 2A, compare lanes c-i with  a and b).
To determine which of the two RNase H2 activities is important for preventing the growth defects observed in strains depleted of Rnr1 and lacking RNase H2 in presence/absence of RNase H1, we made use of two mutant variants of RNase H2: (i) Rnh201-P45D-Y219A which has no RER activity, but retains ∼50% of its hybrid-removal activity on long RNA/DNA hybrids (70) (henceforth designated as 'Rnh201-RED'; RED stands for Ribonucleotide Excision Defective). (ii) Rnh201-G42S, which has ∼2% and <10%, RER and hybrid-removal activities, respectively (70). Note that S. cerevisiae Rnh201-G42S is homologous to the human mutant RNase H2 G37S , which is associated with Aicardi-Goutières Syndrome (AGS)--a rare neuroinflammatory autoimmune disorder in humans (71).
We transformed the double mutant P GAL :3HA-RNR1 rnh201Δ and triple mutant P GAL :3HA-RNR1 rnh201Δ rnh1Δ with empty vector, p-RNH201, p-rnh201-G42S or p-rnh201-RED. We found that on glucose medium, the variant Rnh201-RED suppressed the growth defects in both of these strains slightly less well than the WT Rnh201 (Figure 2B, compare rows a with b and d, and e with f and h). Suppression of the growth defects in glucose medium by the variant Rnh201-G42S was similar to Rnh201-RED in the P GAL :3HA-RNR1 rnh201Δ strain, but was much weaker in the P GAL :3HA-RNR1 rnh1Δ rnh201Δ strain ( Figure 2B, compare rows c and d with g and h). These results indicate that both single genomic rNMPs and RNA/DNA hybrids are detrimental for the growth of strains depleted of Rnr1 and lacking RNase H2 or RNases H1 and H2. Because the variant Rnh201-RED, which has much higher hybridremoval activity than the variant Rnh201-G42S, better alleviated the growth defects of the triple mutant depleted of Rnr1 and lacking RNases H1 and H2, we concluded that removal of RNA/DNA hybrids is the critical factor for survival of this triple mutant. Note that it is possible that, because RER is absent (i.e. in presence of plasmid vector) or defective (i.e. in presence of plasmid expressing Rnh201-G42S), persistent RNA/DNA hybrids, e.g. R-loops, could increase and/or become highly toxic in cells depleted for Rnr1 and lacking RNases H1 and H2.
Finally, we tested the effects of the loss of Crt1, which leads to the expansion of dNTP pools (56), on the growth of strains P GAL :3HA-RNR1 rnh201Δ and P GAL :3HA-RNR1 rnh1Δ rnh201Δ in glucose medium. Drop test growth assays showed that the absence of Crt1 fully suppressed the growth defects in both strains in glucose medium, relative to the WT strain ( Figure 2C, compare rows c with e, and f with g, and e and g with a). Increased dNTP synthesis in absence of Crt1 should improve DNA synthesis (both replication and repair) (14,60,72) and reduce utilization of rNTPs by WT replicative Pols (10,13), thereby mitigating genomic instability defects that can be induced by unrepaired single genomic rNMPs and persistent RNA/DNA hybrids.
We analyzed total mutation rates and specific mutation rates (i.e. transitions, transversions, 1 bp indel and Δ2-5 bp) for the CAN1 gene in the WT, rnh201Δ, P GAL :3HA-RNR1 and P GAL :3HA-RNR1 rnh201Δ strains, grown in glucose medium (Figure 3; see also Supplementary Table S3). We found that total mutation rates and Δ2-5 bp rates in single mutant rnh201Δ were 2.4-fold and 19.3-fold higher than in the WT strain, respectively ( Figure 3A and B; compare samples 2 with 1). These data are in agreement with earlier reports using other yeast backgrounds (see e.g. (23,34,73,76)). Additionally, the single mutant P GAL :3HA-RNR1 depleted of Rnr1 showed a modest 2-fold increase in total mutation rates, and slightly higher specific mutation rates, relative to the WT strain ( Figure 3A and B, compare samples 3 with 1). Strikingly, the double mutant P GAL :3HA-RNR1 rnh201Δ depleted of Rnr1 showed 23-fold and 1039.3-fold increase in total mutation rates and Δ2-5 bp rates, respectively, relative to the WT strain ( Figure 3A and B, compare samples 4 with 1). These results are consistent with published mutation rates (using CAN1 and other reporters) for the double mutant pol2-M644G rnh201Δ, which accumulates high loads of single rNMPs in the leading strand (see e.g. (23,54,77)). This leads us to infer that Top1-mediated RNA-DNA damage is greatly increased in the double mutant P GAL :3HA-RNR1 rnh201Δ depleted of Rnr1.

Loss of Top1 reverses the severe growth defects of Rnr1depleted RER-deficient Pol -M644G or ␦-L612M triple mutants
The data in Figures 2 and 3 suggested that the accumulation of single genomic rNMPs and the associated Top1mediated DNA damage are greatly increased in the doublemutant P GAL :3HA-RNR1 rnh201Δ depleted of Rnr1. Accordingly, drop test growth assays showed that the triple  Table S3. Symbols on the organigram below the plot: + and -indicate that the protein is present or absent, respectively. (B) 2-5 bp specific mutation rate is highly increased in cells lacking RNase H2 and depleted of Rnr1. Specific mutation rates of CAN1 (mutation-spectra) for the same strains and growth conditions as in (A). Specific mutation rates are plotted on the Yaxis. The different types of mutations are color-coded. '1 bp Indel' stands for 1 base pair insertion/deletion. ' 2-5 bp' stands for 2-5 base pairs deletion. See also Supplementary Table S3. For other details see (A). mutant P GAL :3HA-RNR1 rnh201Δ top1Δ grew better than the double mutant P GAL :3HA-RNR1 rnh201Δ in glucose medium, whereas both mutants grew similarly to each other in galactose plus sucrose medium ( Figure 4A, compare rows a-d).
Previously published data indicate that survival of yeast mutants harboring Pol ε-M644G requires the expansion of dNTP pools, by constitutive activation of the S-phase checkpoint (23,84). Deletion of RNH201 in these mutants further increases the dNTP levels, which is indicative of further exacerbation of replicative stress (23). However, Williams et al. (54) found that the presence of Pol ␣-L868M or ␦-L612M in yeast cells does not lead to increased dNTP abundance, either in the presence or absence of RNase H2. Here, we found that, as described previously (23,54), the pol2-M644G rnh201Δ double mutant has ∼4-fold higher dNTP pool levels than the single mutant rnh201Δ and the two double mutants pol1-L868M rnh201Δ and pol3-L612M rnh201Δ. The latter three strains showed only slightly increased dNTP concentrations as compared to the WT strain (Supplementary Figure S3A and Supplementary Table S4, compare samples 1-5). Interestingly, depletion of Rnr1 for 6 h decreased dNTP pools > 3-fold in strains carrying P GAL -3HA:RNR1, regardless of the status of RER, as compared to the WT strain (Supplementary Figure S3A and Supplementary Table S4, compare samples 6 and 7 with 1). Moreover, the >3-fold decrease in dNTP levels in Rnr1-depleted RER-deficient strains remained even in combination with Pol ε-M644G, ␦-L612M or ␣-L868M (Supplementary Figure S3A and Supplementary Table S4, compare samples 7-10). These observations are in accordance with western blotting data for the activation of the S-phase checkpoint in Supplementary Figure S3B. Induction of Rnr3 expression and phosphorylation of Rad53 were modest in strains carrying P GAL -3HA:RNR1 and depleted of Rnr1 for 6 h, with or without active RER, or both lacking RER and harboring an rNTP-permissive Pol (lanes b, f, h, j and l, Rnr3 and P-Rad53). We conclude that the presence of an rNTPpermissive Pol (ε-M644G, ␦-L612M or ␣-L868M) does not affect dNTP concentrations in cells lacking RNase H2 and depleted of Rnr1.
Drop test growth assays with strains harboring Pol ε-M644G or ε-M644L showed that, in glucose medium: (i) The two double mutants P GAL :3HA-RNR1 pol2-M644G and P GAL :3HA-RNR1 pol2-M644L grew slower than the   (23) indicate that purified Pols ε-M644G and ε-M644L have lower catalytic efficiency than their WT parent enzyme. We therefore hypothesize that these Pol variants are less active following Rnr1 depletion, due to limited and/or unbalanced dNTP concentrations, than with WT levels of Rnr1. (ii) The growth of the triple mutant P GAL :3HA-RNR1 pol2-M644G rnh201Δ was severely affected with respect to P GAL :3HA-RNR1 pol2-M644G ( Drop test growth assays with mutants harboring Pol ␦-L612M or ␣-L868M ( Figure 4A) showed that, in glucose medium: (i) The two double mutants P GAL :3HA-RNR1 pol3-L612M and P GAL :3HA-RNR1 pol1-L868M grew similarly to the single mutant P GAL :3HA-RNR1 (compare rows k and o with a). (ii) The triple mutant P GAL :3HA-RNR1 pol1-L868M rnh201Δ grew only slightly slower than P GAL :3HA-RNR1 pol1-L868M (compare rows q with o). (iii) The triple mutant P GAL :3HA-RNR1 pol3-L612M rnh201Δ showed severe, synergistic growth impairment, as compared to P GAL :3HA-RNR1 pol3-L612M (compare rows m with k). Note that growth impairment in P GAL :3HA-RNR1 pol3-L612M rnh201Δ was slightly less marked than in P GAL :3HA-RNR1 pol2-M644G rnh201Δ (compare rows m with g). (iv) Loss of Top1 in the triplemutant P GAL :3HA-RNR1 pol3-L612M rnh201Δ rescued the severe co-synthetic growth defect (compare rows n with m). (v) The quadruple mutant P GAL :3HA-RNR1 pol1-L868M rnh201Δ top1Δ grew better than P GAL :3HA-RNR1 pol1-L868M rnh201Δ TOP1 + (compare rows r with q). Together, these results suggest that following depletion of Rnr1 the excessive incorporation of rNMPs by Pol ␦-L612M in lagging strand leads to detrimental Top1mediated RNA-DNA damage in RER-deficient cells. However, rNMP incorporation by Pol ␣-L868M in lagging strand is much less damaging.  Figure 2B. Cells were grown overnight in liquid minimal medium lacking leucine with 2% galactose and 1% sucrose at 30 • C. Serial-dilutions were plated on solid minimal medium lacking leucine with 2% glucose, or with 2% galactose and 1% sucrose. Photographs were taken after incubation for 7 days at 30 • C. The images for each growth condition are from the same plate. The horizontal line across the images is included for clarity. For other details, see (A). Top1-mediated cleavage occurs at RNA-DNA junctions of unrepaired embedded single or stretches of rN-MPs. Both can be incorporated by rNTP-permissive Pols in triple mutants P GAL :3HA-RNR1 rnh201Δ pol2-M644G and P GAL :3HA-RNR1 rnh201Δ pol3-L612M in glucose medium, thereby compromising genome stability and cell viability. To determine whether single rNMPs alone or a combination of both single and stretches of rNMPs contributed to Top1-mediated RNA-DNA damage, we transformed these triple mutants with empty vector, p-RNH201, p-rnh201-G42S or p-rnh201-RED. Drop test growth assays showed that, in glucose medium, in sharp contrast to WT Rnh201, neither the variant Rnh201-RED nor the variant Rnh201-G42S suppressed the severe growth defects of the triple mutants ( Figure 4B, compare rows a-d and e-h). Because Rnh201-RED processes RNA/DNA hybrids but cannot process single rNMPs (70), we concluded that unrepaired single genomic rNMPs are likely to constitute the major source of Top1-induced DNA damage and growth defects in the triple-mutants in glucose medium. Supporting this conclusion, endogenous RNase H1 failed to support the growth of these triple mutants in glucose medium, despite processing RNA/DNA hybrids ( Figure 4A rows g and m and 4B rows a, c-d, e and g-h). Additionally, we found that, in sharp contrast to P GAL :3HA-RNR1 pol2-M644G rnh201Δ, the triple mutant P GAL :3HA-RNR1 pol2-M644G rnh1Δ grew similarly to double mutant P GAL :3HA-RNR1 pol2-M644G in glucose medium (Supplementary Figure  S3C, compare rows d-f). Finally, one cannot exclude the possibility that, in addition to single rNMPs, very short stretches of rNMPs that are not substrates for RNase H1 and are not cleaved efficiently by Rnh201-RED (e.g. 2-3 consecutive rNMPs; see (70)), might also contribute to Top1-mediated RNA-DNA damage in the triple mutants P GAL :3HA-RNR1 rnh201Δ pol2-M644G and P GAL :3HA-RNR1 rnh201Δ pol3-L612M in glucose medium.

Measurements of the numbers of genomic rNMPs by alkalifragmentation of total DNA
To gain more insight into the role of Rnr1 depletion in increasing the incorporation of rNMPs in genomic DNA, we sought to determine the numbers of genomic rNMPs in mutants lacking both Rnr1 and RER, or that are also combined with an rNTP-permissive Pol in presence/absence of Top1. We made use of a panel of pairs of strains (see organigram in Figure 5A), of which each pair comprises one strain harboring the RNR1 gene under the control of its native promoter and expressing Rnr1 at WT levels (henceforth referred to as 'Rnr1 [+]'; odd numbers on the organigram), and an equivalent strain carrying P GAL :3HA-RNR1 and depleted of Rnr1 for 6 h in glucose-containing medium (henceforth referred to as 'Rnr1 [−]'; even numbers on the organigram). Total cellular DNA was heated in the presence of alkali, which should both denature the DNA duplex and hydrolyze the backbone downstream of embedded ribonucleotides (see e.g. (20,23)). Alkali-fragments from total DNA (Afts) were resolved by alkaline-gel electrophoresis ( Figure 5A). Based on densitometry of SYBR-staining, we determined for each strain the distribution of Afts sizes (i.e. densitometry normalized to the length of the fragment; Figure 5B and Supplementary Figure S4), and the numbers of rNMPs present in the 24 Mb haploid yeast genome (i.e. 'total genomic rNMPs'; Figure 5C and Supplementary Table S5). For the calculation of the numbers of total genomic rNMPs, in order to account for DNA breaks that may occur independently of incorporated ribonucleotides during alkaline heat-treatment, the number of DNA breaks in the WT strain was subtracted from every other Rnr1 [+] strain. Replicating DNA may be more susceptible to breakage during alkaline heat-treatment than the rest of the genome, so the number of DNA breaks in the single mutant P GAL :3HA-RNR1 depleted of Rnr1, which shows severe blockage of cells in S-phase ( Figure 1E Table S5). Together, these results are consistent with earlier studies reporting that Top1-mediated processing of unrepaired single genomic rNMPs is mainly associated with ribonucleotides incorporated by Pol ε in leading strand (37,54,73,74,85).
The increased alkali-fragmentation observed in genomic DNA from P GAL :3HA-RNR1 rnh201Δ (Rnr1 [−]) mutants without or with an rNTP-permissive Pol (ε-M644G, ␦-L612M or ␣-L868M) could potentially reflect incomplete DNA replication/repair, independently of incorporated ribonucleotides ( Figure 5A, lower-part of Figure 5B, Supplementary Figure S4B, Figure 5C and Supplementary Table S5, samples 4, 8, 10, 14, 16, 20 and 22). To assess this, formamide-denatured genomic DNA was resolved on non-denaturing, neutral 1% agarose gels (Supplementary Figure S5). This revealed that P GAL :3HA-RNR1 rnh201Δ  Figure S5, compare samples 1 with 4, and 9 with 12). Together, these results indicate that most Afts detected in P GAL :3HA-RNR1 rnh201Δ (Rnr1 [−]) mutants without or with an rNTP-permissive Pol are due to cleavage at sites of genome-embedded ribonucleotides ( Figure 5A, lower-part of Figure 5B, Supplementary Figure S4B, Figure 5C and Supplementary Table S5,  Collectively, the results presented above led us to conclude that genomic rNMPs are greatly increased in the presence of limited dNTP concentrations in RER-deficient strains depleted of Rnr1, and this is further exacerbated in cells that also harbor an rNTP-permissive Pol. Even though Afts of shorter sizes accumulated in the triple mutant P GAL :3HA-RNR1 pol2-M644G rnh201Δ (Rnr1 [−]) as compared to the triple mutant P GAL :3HA-RNR1 pol3-L612M rnh201Δ (Rnr1 [−]), the amounts of total genomic rNMPs, were, unexpectedly, similar for both mutants (compare samples 20 with 14 in Figure 5A, lowerpart of Figure 5B, and Supplementary Figure S4B, with same samples in Figure 5C [for P-values, see Supplementary  Table S5]). To gain more insight into this question, we calculated the contributions of replicative Pols ␣, ␦, and ε to the synthesis of genomic DNA, by applying the mathematical formula from Reijns et al. (53), for four different conditions (Supplementary Figure S6 and Table S6 The slightly higher contribution of Pol ␦ relative to Pol ε to synthesis of genomic DNA in cells expressing Rnr1 at WT levels (conditions i and iii) is consistent with the notion that Pol ␦, besides synthesizing the bulk of lagging strand, plays a role in leading strand synthesis (see e.g. (5)). The minor contribution of Pol ␣, regardless of the state of Rnr1 (conditions i-iv), is in agreement with the role of Pol ␣ in synthesizing only a small fraction of the yeast nuclear genome (see e.g. (53)). Notably, the contribution of Pol ␦ was ∼3fold higher than Pol ε in cells depleted of Rnr1 (conditions ii and iv). This suggests that Pol ε might synthesize much less leading strand DNA under limited dNTP availability, with Pol ␦ taking a larger role. Both triple mutants P GAL :3HA-   Figure 5A, lower-part of Figure 5B, Supplementary Figure S4B, and Figure 5C [for P-values, see Supplementary  Table S5]). This was unexpected, since the loss of Top1 reversed the severe growth defects of RER-deficient mutants that are depleted of Rnr1 and contain Pol ␦-L612M or ε-M644G (see Figure 4A) and will be addressed in the following section.

Top1 processes unrepaired single genomic ribonucleotides in both leading and lagging strands under conditions of limited dNTP concentrations
We analyzed the role of Top1 in the processing of unrepaired single genomic rNMPs in Rnr1-depleted RERdeficient triple mutants bearing Pol ε-M644G, ␦-L612M or ␣-L868M, by strand-specific Southern blotting of Afts (see e.g. (48,54,77,80,81,87)). DNA was processed as for Figure 5, resolved by alkaline-gel electrophoresis and capillaryblotted. We analyzed the gene AGP1, which is situated nearby the strong early-firing bidirectional origin ARS306 (88) (see scheme in Figure 6A; see also Supplementary Table  S2 for the chromosomal coordinates of the probes). Singlestranded PCR probes were complementary to either the lagging strand (bottom strand) (probe A), or leading strand (top strand) (probe B).
In Figure 6, we compare strains harboring the RNR1 gene under the control of its native promoter and expressing Rnr1 at WT levels (Rnr1 [+]), to strains carrying P GAL :3HA-RNR1 and depleted of Rnr1 for 6 h (Rnr1 [−]) (see organigram in Figure 6B and D). In Supplementary  Figures S8-S10, we used the same strains, but for strains carrying P GAL :3HA-RNR1 we included also growth in permissive conditions, in which Rnr1 should be moderately over-expressed (henceforth referred to as 'Rnr1 [*]') (see organigram in panels B and D of Supplementary Figures S8-S10).
Together, the results presented above for strains bearing the alleles pol1-L868M and pol3-L612M lead us to hypothesize that embedded rNMPs increased on the AGP1lagging strand in the quadruple mutants P GAL :3HA-RNR1 pol1-L868M rnh201Δ top1Δ (Rnr1 [−]) and P GAL :3HA-RNR1 pol3-L612M rnh201Δ top1Δ (Rnr1 [−]), relative to their corresponding triple mutants TOP1 + equivalents. In contrast, there was no increase in embedded rNMPs on the AGP1-lagging strand in the triple mutants pol1-L868M rnh201Δ top1Δ (Rnr1 [+]) and pol3-L612M rnh201Δ top1Δ (Rnr1 [+]), relative to their corresponding double mutants TOP1 + equivalents, which is consistent with a previous report (54) (Figure 6B and upper-part of Figure 6C, compare samples 15 with 13, and 11 with 9; panels B and C in Supplementary Figures S8 and S9, compare samples 13 with 10). We presume this difference reflects a much higher density of genomic ribonucleotides in the mutants depleted of Rnr1, which is above the limit for detection of Top1-mediated incisions by Southern analysis. We, therefore, conclude that Top1 processes unrepaired single genomic rNMPs that are incorporated by Pols ␣ and ␦ in lagging strand at the AGP1 locus in absence of Rnr1.
Alkali-fragmentation of the AGP1-leading strand revealed that high mobility fragments <500 nt accumulated in the strain P GAL :3HA-RNR1 pol2-M644G rnrh201Δ (Rnr1 [−]), compared to the corresponding strain pol2-M644G rnrh201Δ (Rnr1 [+]) ( Figure 6D and E, compare samples 6 with 5; Supplementary Figure S10D and E, compare samples 12 with 10). Additionally, alkali-fragmentation was higher in the quadruple mutant P GAL :3HA-RNR1 pol2-M644G rnrh201Δ top1Δ (Rnr1 [−]) relative to the corresponding triple mutant expressing Top1 ( Figure 6D and lower-part of Figure 6E, compare samples 8 with 6; Supplementary Figure S10D and E, compare samples 15 with 12). These results further confirm the elevated incorporation of rNMPs by Pol ε-M644G following Rnr1 depletion and support the model that Top1 processes unrepaired single rNMPs in leading strand (54).   65 Kb distance from the right side of the bidirectional origin ARS306 on chromosome III (the direction of replication, left to right, is indicated by a black horizontal arrow). For clarity, the RF to the left side of ARS306 is omitted. OFs in the nascent lagging strand, which are synthesized by Pols ␣ and ␦, are depicted by small purple arrows. The nascent leading strand, which is mainly synthesized by Pol ε, is depicted by a long brown arrow. Template lagging and leading strands are depicted by black lines. Single-stranded probes A and B (length ∼660 nt) hybridize to AGP1-lagging strand DNA (bottom strand) and AGP1-leading strand DNA (top strand), respectively, at ∼2 Kb distance from the right side of ARS306 (for primer sequences see Supplementary Table S2). (B-E) The following strains (see Supplementary Table S1 for the list of strains) are represented by symbols on the organigrams in (B) and (D). Rnr1 [+] condition: 1. WT; 3. rnh201 ; 5. pol2-M644G rnh201 ; 7. pol2-M644G rnh201Δ top1 ; 9. pol3-L621M rnh201 ; 11. pol3-L621M rnh201 top1 ; 13. pol1-L868M rnh201 ; lacking Top1. The greatest fragmentation was seen in the quadruple mutant P GAL :3HA-RNR1 pol3-L612M rnh201Δ top1Δ (Rnr1 [−]) ( Figure 6D and lower-part of Figure 6E, compare samples 14 with 4, and 10 with 14, and 16 with 14, and 12 with 10; panels D and E in Supplementary  Figures S8 and S9, compare samples 6, 12 and 15). These results suggest that rNMPs are inserted by Pol ␣-L868M or ␦-L612M in AGP1-leading strand in absence of Rnr1 and are processed by Top1 when RNase H2 is also absent.
Finally, there was a clear decrease in alkali-fragmentation of both AGP1-lagging and AGP1-leading strand DNAs in various combinations of RER-deficient mutants carrying P GAL :3HA-RNR1 that were grown in permissive conditions, in which Rnr1 should be moderately over-expressed (Rnr1 [*]) (Supplementary Figures S8B, S8D, S9B, S9D and S10D, compare samples 11 with 10 and 12, and 14 with 13 and 15). Overexpression of Rnr1 may be associated with higher dNTP concentrations (59,60), supporting the expectation that dNTP levels and rNMP incorporation in DNA are inversely correlated, which is consistent with a previous report (13).

DISCUSSION
Herein, we report the effects of high density of unrepaired ribonucleotides in genomic DNA of S. cerevisiae. We have modulated two factors, which influence rNMP incorporation: (i) The dNTP:rNTP ratios, by depleting Rnr1, which controls both the levels and balance of dNTPs (see Figure  1C, Supplementary Figure S3A and Supplementary Table  S4). (ii) The sugar selectivity of replicative Pols, by employing steric gate Pol variants that are rNTP-permissive (for reviews, see e.g. (11,12)). We also deleted the gene RNH201 or both genes RNH201 and TOP1. Here we found that genomic rNMPs and the associated Top1-mediated 2-5 bp are greatly increased in double mutants lacking Rnr1 and RER. Furthermore, when a threshold of single genomic rNMPs is exceeded in RER-deficient Rnr1-depleted cells that also harbor an rNTP-permissive Pol (ε-M644G or ␦-L612M), Top1-mediated RNA-DNA damage leads to severe growth defects. Moreover, we show that under constitutive replicative stress induced by Rnr1 depletion, loss of RNases H1 and H2 leads to cell lethality, likely due to deleterious accumulation of RNA/DNA hybrids (e.g. R-loops).
Finally, we provide evidence showing that the contribution of Pol ␦ to synthesis of genomic DNA is much greater than that of Pol ε in cells depleted of Rnr1.

Use of variants RNase H2-RED and RNase H2-G42S elucidates the growth defects associated with depletion of Rnr1 in the absence of RNase H activity
Multiple studies have shown that the absence of RNase H2, and to a lesser extent RNase H1, induces genomic instability in S. cerevisiae (see e.g. (22,23,74,76,82,(89)(90)(91)). Both RNase H2 activities, hybrid-removal and RER, play important roles in safeguarding genome integrity, confounding assignment of a phenotype to either function.
In our drop test growth assays, the use of the three plasmids expressing WT Rnh201, variant Rnh201-RED, or variant Rnh201-G42S, led us to infer that the accumulation of RNA/DNA hybrids is the critical factor for cell lethality in strain P GAL :3HA-RNR1 rnh1Δ rnh201Δ depleted of Rnr1 ( Figure 2B).
It is established that the accumulation of R-loops in budding and fission yeasts double mutants rnh1Δ rnh201Δ compromises genome stability (see e.g. (27,(91)(92)(93)(94)(95)(96)(97)(98)). We propose that R-loop-mediated transcriptionreplication conflicts (for reviews, see e.g. (29,30,99)) lead to defects in DNA replication, genome stability and growth in strains depleted of Rnr1 and lacking RNases H1 and H2. Limited dNTP pools, induced by Rnr1 depletion, could compromise the restart of stalled RFs at sites of persistent R-loops (98,100), thereby leading to RF collapse/breakage, and/or block repair of R-loop induced DNA damage, e.g. by the HR-repair machinery that would escort RFs (for a review, see e.g. (101)). Because RNR supplies also dNTPs for DNA repair, we speculate that insufficient dNTP supply may additionally compromise repair of R-loop associated-DNA damage independently of DNA replication in cells depleted of Rnr1 (see e.g. (102)).
Rnh201-RED suppressed the growth defects induced by Rnr1 depletion to similar levels in absence of RNase H2 or RNases H1 and H2 ( Figure 2B). This suggests that either endogenous RNase H1 is dispensable to resolve RNA/DNA hybrids, or that only a subset of hybrids is toxic and these are specifically recognized by Rnh201-RED. However, Rnh201-G42S, which has much lower hybrid- removal activity than Rnh201-RED (70), suppressed less well the growth defects of strain P GAL :3HA-RNR1 rnh1Δ rnh201Δ in glucose medium than did Rnh201-RED (Figure 2B). This suggests a possible overlap between RNases H1 and H2 in hybrid removal in conditions of constitutive replicative stress following Rnr1 depletion. Nonetheless, it is important to emphasize that RNases H1 and 2 have different substrates preferences that do not completely overlap. There are R-loops targeted exclusively by RNase H1 (91), and conversely RNase H2 has R-loop substrates that RNase H1 is not able to cleave even when overexpressed, but that the variant Rnh201-RED can degrade (70).
Because the variant Rnh201-RED was almost as effective as WT Rnh201 in strains P GAL :3HA-RNR1 rnh201Δ and P GAL :3HA-RNR1 rnh1Δ rnh201Δ in glucose-containing medium ( Figure 2B), it may be that increased accumulation of unrepaired single genomic rNMPs in these strains has only a slight negative effect on their growth. An alternative possibility is that the accumulation of unrepaired single genomic rNMPs (and the associated Top1-mediated RNA-DNA damage) in strain P GAL :3HA-RNR1 rnh1Δ rnh201Δ in glucose-containing medium greatly exacerbated formation of RNA/DNA hybrids, e.g. R-loops, and/or their impact on genome stability and cell growth. This is in contrast to the situation that occurs in strains P GAL :3HA-RNR1 pol2-M644G rnh201 and P GAL :3HA-RNR1 pol3-L612M rnh201 where the variant Rnh201-RED failed to support growth in glucose medium ( Figure 4B), suggesting that high density of unrepaired single genomic rNMPs, and not RNA/DNA hybrids or a combination of both substrates, caused severe growth defects in these strains.
Finally, Meroni et al. (103) suggested that the translesion synthesis DNA polymerase incorporates multiple contiguous rNMPs that are toxic for genome stability and viability of cells lacking RNases H1 and H2 under limited dNTP pools triggered by HU. We found that the deletion of the gene RAD30 encoding for yeast Pol in strain P GAL :3HA-RNR1 rnh1Δ rnh201Δ (background BY4741) only modestly alleviated cell lethality in glucose medium (our unpublished observations). We therefore hypothesize that, unlike R-loops, genome-embedded multiple contiguous rNMPs may not accumulate at high levels and/or may not be highly toxic in strains depleted of Rnr1 and lacking RNases H1 and H2.

Following Rnr1 depletion Pol ␦ contributes much more to nuclear genome synthesis than Pol
We assessed the contribution of replicative Pols ␣, ␦ and ε to overall DNA synthesis (Supplementary Figure S6 and Supplementary Table S6). We found that Pol ␦ synthesizes only slightly more DNA than Pol ε in the presence of Rnr1, supporting previously reported in vivo data in budding and fission yeasts (see e.g. (4,5)). Interestingly, however, we found that Pol ␦ synthesizes ∼3-fold more DNA than Pol ε in Rnr1-depleted cells. In accordance with this result, Southern analysis of alkali-treated genomic DNA detected increased fragmentation of AGP1leading strand in the triple mutant P GAL :3HA-RNR1 pol3-L612M rnh201Δ (Rnr1 [−]) compared to the double mutant P GAL :3HA-RNR1 rnh201Δ (Rnr1 [−]), with even greater fragmentation in the quadruple mutant P GAL :3HA-RNR1 pol3-L612M rnh201Δ top1Δ (Rnr1 [−]) ( Figure 6 and Supplementary Figure S9).
These results are consistent with recent in vitro experiments (3), reporting that leading strand synthesis at ARS306 is initiated by Pol ␦. This uses lagging strand primers, which are laid down by Pol ␣-primase complex at RFs on opposite sides of the origin. Notably, our probe B (∼660 nt length) hybridizes to AGP1-leading strand DNA ∼2 kb from ARS306 ( Figure 6A and Supplementary Table S2). Moreover, in vitro analyses show that purified Pol ε requires 2-to 4-fold higher dNTP levels for maximum extension of nascent leading strand compared to Pol ␦ (104). We hypothesize that, following Rnr1 depletion, the handoff from Pol ␦ to Pol ε during initiation of leading-strand synthesis at bidirectional origins is delayed by insufficient dNTP supply. It is important to emphasize that probe B detected excessive fragmentation of AGP1-leading strand DNA in strains P GAL :3HA-RNR1 pol2-M644G rnh201Δ (Rnr1 [−]) and P GAL :3HA-RNR1 pol2-M644G rnh201Δ top1Δ (Rnr1 [−]) ( Figure 6 and Supplementary Figure S10), suggesting that the delay in the hand-off from Pol ␦ to Pol ε might occur only in a subset of cells within the same strain.
Finally, Pol ␦ may not only act to initiate leading strand synthesis at origins, but might function in other regions in the genome; e.g. difficult-to-replicate sequences and sites of mismatches, or under conditions of stress (see e.g. (6-9,105,106)). We further hypothesize that, following Rnr1 depletion, Pol ε may frequently uncouple from the nascent, leading strand 3 terminus. Pol ␦ would then take over DNA synthesis until coupled leading strand synthesis by Pol ε is re-established.

Top1-processing of unrepaired single genomic rNMPs is observable by Southern blotting in both leading and lagging strands under high rNMP load
Processing of unrepaired single genomic rNMPs by Top1 was previously reported to predominantly occur on the leading strand, with much less activity on the lagging strand (see e.g. (37,54,73,74,85)). Williams et al. (54) showed by Southern blotting that Top1 activity at sites of unrepaired rNMPs is observable in leading strand in strain pol2-M644G rnh201Δ, but not in lagging strand in strains pol1-L868M rnh201Δ and pol3-L612M rnh201Δ, which all three express Rnr1 at WT levels (results recapitulated herein in Figure 6 and Supplementary Figures S8-S10). Our Southern analyses using mutants depleted of Rnr1 showed, as expected, that processing of unrepaired single genomic rN-MPs by Top1 is observable in leading strand in P GAL :3HA-RNR1 pol2-M644G rnh201Δ (Rnr1 [−]) ( Figure 6 and Supplementary Figure S10). Notably, Top1 activity at sites of rNMPs was observable in both leading and lagging strands in P GAL :3HA-RNR1 pol1-L868M rnh201Δ (Rnr1 [−]), and much more in P GAL :3HA-RNR1 pol3-L612M rnh201Δ (Rnr1 [−]) ( Figure 6 and Supplementary Figures  S8 and S9). This leads us to hypothesize that the combination of Rnr1 depletion, lack of RNase H2 and presence of Pol ␦-L612M or ␣-L868M increased the rNMP load and Top1-dependent rNMP removal in both leading and lagging strands in these triple mutants, sufficiently for Nucleic Acids Research, 2020, Vol. 48, No. 8 4293 detection of the retained rNMPs by Southern analysis in the corresponding top1Δ quadruple mutants counterparts. However, the density of ribonucleotides and Top1-mediated cleavage at unrepaired rNMP sites in leading strand of strains harboring Pol ε-M644G and lacking RNase H2 was sufficiently high for detection of retained rNMPs by Southern analysis, regardless of the presence/absence of Rnr1, as previously suggested by Williams et al. (54) for cells expressing Rnr1 at WT levels. Taken together, we conclude that Top1 incises at unrepaired single genomic rNMPs on leading and lagging strands. However, there is a detection limit for the identification of Top1 cleavages by Southern analyses.

S. cerevisiae has a threshold of tolerance to unrepaired genomic ribonucleotides
The incorporation of ribonucleotides in genomic DNA of mice lacking RNase H2 (RNase H2 null ), or expressing the variant RED (RNase H2 RED ) causes massive genome instability, leading to a p53 DNA damage response and early embryonic death (19)(20)(21). The threshold for embryonic cell death was estimated at ≤240K unrepaired rNMPs per 1 Gb of genomic DNA (21).
Even though the two triple mutants P GAL :3HA-RNR1 pol2-M644G rnh201 (Rnr1 [−]) and P GAL :3HA-RNR1 pol3-L612M rnh201 (Rnr1 [−]) have only ∼20-25% more rNMPs than the triple mutant P GAL :3HA-RNR1 pol1-L868M rnh201 (Rnr1 [−]), the growth of the former two strains is much more compromised than the latter strain (Figures 4A and 5C and Supplementary Table S5). We can connect these findings to indicate that total genomic rNMPs in strain P GAL :3HA-RNR1 pol1-L868M rnh201 (Rnr1 [−]) do not exceed a threshold above which Top1induced RNA-DNA damage leads to severe growth defects. We speculate that a substantial fraction of ribonucleotides incorporated by Pol ␣ and their associated Top1-mediated un-ligatable nicks in cells lacking both RER and Rnr1 are removed by the OF maturation machinery, and/or by the mismatch repair mechanisms, which are very efficient in correcting errors generated by Pol ␣ (53,54,107).
Top1-induced-DSBs at unrepaired single genomic rNMP sites in strain pol2-M644G rnh201Δ are not tolerated in absence of Rad52 or presence of HU (37,77), albeit this strain accumulates ∼590K ± 156K rNMPs per 1 Gb according to (53) (∼400K ±11K rNMPs Gb −1 in our strain background, Figure 5C and Supplementary Table S5), which is below the number of rNMPs we observed in strains P GAL :3HA-RNR1 pol2-M644G rnh201Δ (Rnr1 [−]) and P GAL :3HA-RNR1 pol3-L612M rnh201Δ (Rnr1 [−]). It remains, therefore, to be determined whether the severe growth defects in strains expressing Pol ε-M644G or ␦-L612M in absence of both RNase H2 and Rnr1 arise because the level of Top1mediated-DSBs at single genomic ribonucleotide sites exceeds the capacity of Rad51/Rad52-dependent HR repair. The machinery may be swamped by high numbers of DSBs and/or be compromised by insufficient dNTP supply induced by Rnr1 depletion (see model in Supplementary Figure S11).
We conclude that in budding yeast, unrepaired single genomic rNMPs above a critical threshold lead to deleterious DNA damage in a Top1-dependent manner, particularly in conditions of insufficient dNTP supply (e.g. under Rnr1 depletion or HU). DNA lesions induced by human TOP1 at sites of genomic ribonucleotides in transformed, human RNase H2-deficient cells are efficiently repaired by Poly (ADP-ribose) polymerase (PARP) (108). However, when these cancerous cells were treated with PARPtrapping drugs, unresolved TOP1-mediated DNA lesions led to deleterious defects in DNA replication, genome integrity and cell viability (108). RNase H2-depleted human cells are non-viable in the presence of inhibitors of ATR Mec1 (ATR is the mammalian homolog of yeast Mec1), likely due to deleterious Top1-mediated RNA-DNA damage (109). It is tempting to speculate that inhibition of ATR Mec1 in these RNase H2-depleted cells (109) led to insufficient dNTP supply (for reviews on mammalian RNR regulation, see e.g. (110,111)), thereby both increasing incorporation of rN-MPs in genomic DNA by replicative Pols and compromising the repair of the associated DNA damage. Finally, we hypothesize that the threshold of tolerance to single unrepaired genomic rNMPs in mice embryos lacking RER (19)(20)(21) might be dependent on the deleterious effects of TOP1.

Modulating RNR activity could be beneficial in human diseases associated with RNase H deficiencies
Silencing of RRM2, the gene encoding for mammalian RNR complex subunit R2, hyper-sensitizes cancer cells to camptothecin, an inhibitor of TOP1, by hampering repair of DNA damage caused by collisions of RFs with TOP1 cleavage complexes (112). Indeed, inhibitors of mammalian RNR activity are widely used as chemotherapeutic agents to reduce the dNTP supply that is needed for uncontrolled proliferation in cancer cells (for reviews, see e.g. (110,113)). Our findings in budding yeast leads us to hypothesize that, apart from impeding DNA synthesis (both replication and repair) in cancer cells, RNR inhibitors would have the under-appreciated effect of increasing the incorporation of rNMPs in genomic DNA. Targeted inhibition of RNR activity in cancer cells defective for RNase H2 might hypersensitize them to PARP-trapping drugs, which hamper the repair of TOP1-mediated DNA lesions at sites of single genomic rNMPs (108,114,115). The cytotoxicity of the combination of RNR and PARP inhibitors could be even further enhanced in cancer cells defective for RNase H2 and also bearing an rNTP-permissive Pol mutator (for a review, see e.g. (116)). We also posit that RNR inhibitors would potentiate R-loop-driven DNA damage in cancer cells lacking RNase H2 and/or RNase H1 ( (117,118); for a review, see e.g. (111)). Contrary to cells harboring hypomorphic RNases H, however, cancer cells with WT or hypermorphic RNases H could respond less well to RNR inhibitors. We suggest that cancerous tumors could be tested for defects in RNases H in order to identify those with hypomorphic RNase H mutations as bonafide candidates for treatment with RNR inhibitors (109), and/or to improve anti-RNR strategies and reduce their toxicity.
Increased gene dosage of RRM2 alleviated DNA damage defects in ATR Mec1 -depleted mice (119), and supplementing cells with nucleosides alleviated replication-induced DNA damage in cancer cells (120). We anticipate that increasing dNTP levels exogenously might have a protective effect on proliferating cells in AGS patients with hypomorphic mutations in the RNase H2 complex, by alleviating DNA damage associated with unrepaired single genomic rNMPs and/or R-loops (121,122). RNR might therefore be a promising opportunity to explore in AGS therapies.