Downstream transcription promotes human recurrent CNV associated AT-rich sequence mediated genome rearrangements in yeast

Summary

AT-rich sequence can cause structure variants such as translocations and its instability can be accelerated by replication stresses. When human 16p11.2 or 22q11.2 recurrent copy number variant (reCNV) associated AT-rich sequence was inserted upstream GAL1 promoter in yeast genome, we found that downstream transcription could promote AT-rich forming cruciform structure and mediate gross genome rearrangements. When genes were flanked with direct repeats containing AT-rich sequence, copy number loss of these genes would be stimulated. Transcription-mediated AT-rich instability can be alleviated by disrupting MUS81 or YEN1 and exacerbated by disrupting RAD1/10. Deletion of homologous recombination-associated genes can not only increase AT-rich fragility but also alter the breakpoint positions. AT-rich stability was also affected by DNA topoisomerase poisons. Our results reveal that transcription can promote AT-rich-mediated de novo genome rearrangement, which might be helpful for understanding the mechanism of reCNV formation in humans.

Graphical abstract

Highlights

• •Transcription exacerbates the fragility of upstream AT-rich sequence
• •AT-rich sequences mediate copy number loss in yeast
• •MUS81, YEN1, RAD1/10, RAD51, and Rnase H regulate the fragility of AT-rich
• •DNA topoisomerase poisons increase AT-rich instability

Properties of biomolecules; Techniques in genetics; Molecular genetics; Model organism

Introduction

In both prokaryotes and eukaryotes, palindromic DNA sequences with inverted repeats can form specific cruciform-like structures. The cruciform extrusion of palindromic DNA is promoted by either DNA replication or transcription. At the early S phase, DNA replication can form approximately 3 × 10⁵ cruciforms in a single mammalian cell.¹ During transcription, promoter activation induces negative supercoiling of DNA, which can promote the transcription-mediated cruciform formation.^2,3,4 In mouse G2 phase-like oocytes, cruciform DNAs can be observed in growing oocytes which are transcriptionally active, but are collapsed in fully grown oocytes which are transcriptionally silent.⁵ Evidence shows that palindromic DNA composed of AT base pairs is most vulnerable to cruciform formation, whereas palindromes with centric GC pairs or imperfect palindromes require more driving forces for cruciform extrusion.⁴ Newly formed cruciform structures, in turn, may affect both DNA replication and transcription. Purified cruciform DNA is highly enriched in DNA fragments containing replication origins.⁶ Depletion of the cruciform binding domain of 14-3-3 proteins would retard replication initiation in yeast.⁷ Promoter upstream cruciforms can enhance promoter activity in prokaryotic systems,⁸ suggesting that cruciforms may play a role in transcriptional regulation. Although cruciform DNAs have been discovered for decades and are widespread in somatic and germline cells, their biological functions are still not well understood.

In addition to their beneficial effects, cruciform structures can also have some adverse effects on genome stability. Like other non-B DNA structures, such as G-quadruplexes and Z-DNA,⁹ unsolved cruciforms during DNA transactions may induce DNA double-strand breaks (DSBs). Long palindromic AT-rich repeats (PATRR) have been found located at the breakpoint regions of recurrent chromosomal translocations in humans, such as t(8; 22)(q24; q11),¹⁰ t(17; 22)(q11; q11),¹¹ and t(11; 22)(q23; q11).¹² These PATRR regions are enriched with DSB repair-associated protein markers such as RAD51 and γH2A.X,¹³ indicating that AT-rich indeed induces DSBs and may be repaired by RAD51-mediated homologous recombination.

Palindromic AT-rich sequences have been reported to be cleaved by structure-specific DNA endonucleases, resulting in the generation of DSBs. Study in a plasmid-based system suggested that GEN1 is responsible for the PATRR cruciform cleavage.¹⁴ In microsatellite instability-associated cancer cells, the four-way junction-specific endonuclease MUS81 can cleave long AT-rich sequences when the RecQ DNA helicase gene WRN is depleted.¹⁵ As the formation of cruciforms is promoted by transcription-induced negative supercoiling,¹⁶ cell specific gene expression might induce different cruciform profiles in different cells. AT-rich induced recurrent t(11; 22)(q23; q11) translocations are highly frequently formed in human sperm but couldn’t be detected in lymphoblasts or fibroblasts,¹⁷ which might be caused by testicular specific transcription of genes upstream or downstream of the PATRR sequences.⁵ However, whether the fragilities of AT-rich sequences are aggravated by transcription hasn’t been analyzed.

In addition to translocations, the integrity of the human genome is also disrupted by copy number variations (CNVs). CNVs can form sporadically or recurrently in the genome and are mainly associated with common complex human diseases such as neuropsychiatric disorders.^18,19 The formation of recurrent CNVs (reCNVs) is mediated by non-allelic homologous recombination (NAHR) repairing of DSBs at genomic segmental duplication regions.²⁰ In these reCNVs, 16p11.2 and 22q11.2 reCNVs are associated with diseases such as autism, epilepsy, intellectual disability, schizophrenia, and other brain-related disorders.²¹ According to the UK Biobank,²² these two reCNVs affect approximately 0.056% and 0.067% of people, respectively. Although the pathogenesis of 16p11.2 and 22q11.2 reCNVs has been comprehensively studied, mechanisms regarding the de novo generation of these reCNVs still lack investigation. It has been estimated that 90–95% of 22q11.2 deletions are de novo generated during reproduction and affect about 1 in 1000 fetuses.²³ For 16p11.2 reCNVs, de novo cases are found in 67.8% of deletion carriers and 25.0% of duplication carriers, respectively.²⁴ Deep investigation of the formation mechanisms of these reCNVs is critical for preventing birth defect associated disorders.

In this study, we found that AT-rich sequences were associated with CNV formation and located at the breakpoint regions of 16p11.2 and 22q11.2 reCNVs. Using the yeast system, we analyzed the fragility of the 16p11.2/22q11.2 reCNV associated AT-rich sequences and found that they could form cruciform structures, which made them unstable in yeast. The genome fragility of AT-rich sequence can be induced either by transcription-dependent or transcription-independent mechanisms, and is influenced by various structure-specific endonucleases and DSB repair proteins. These results are helpful for understanding the formation mechanism of 16p11.2 and 22q11.2 reCNVs in human germline cells.

Results

Promoter upstream AT-rich sequences induce gross genome rearrangements

To analyze the genome stability of AT-rich sequences in the human genome, we extracted AT-rich sequences from the human genome (hg19) that meet three criteria: they contain at least one (AT)3 sequence, the content of A and T bases is over 90% in a 100 bp DNA region, and they contain more than 25 AT units in the AT-rich sequence. As a result, 7340 AT-rich sequences (Table S3) were identified in human genome (2.37 AT-rich seq per 1 Mbp of DNA). To analyze whether the genome stability of AT-rich regions decreases with an increase in AT units, these AT-rich sequences were divided into 10 groups based on the number of AT units (26–50, 51–75, 76–100, 101–125, 126–150, 151–175, 176–200, 201–225, 226–250, and >250). Then the number of AT-rich sequences that overlapped with DSB peaks which were detected in the normal MCF7 cell line²⁵ was counted. These DSB peaks which located within the region of 500 bp before and after AT-rich sequence were marked as AT-rich associated DSB peaks. As a result, the proportion of AT-rich sequences overlapped with DSB peaks increases with their AT unit numbers (Figure 1A), indicating that an increase in AT numbers in AT-rich sequences is detrimental to their stability.

Figure 1.

Transcription promotes upstream AT-rich instability in yeast

(A) In the human MCF7 cell line, the stability of AT-rich decreases as the AT units increases. As the AT unit number increases in AT-rich sequences, the proportion of AT-rich sequences overlapping with DSB peaks also increased.

(B) Within the 2500 bp flanking regions, AT-rich associated with more ClinVar CNV breakpoints (ClinVar BP) than the randomly selected sequences. ∗∗, p < 0.01, Wilcoxon test.

(C) An AT-rich sequence with 101–125 AT units (within the red dashed box) is located at the breakpoint 5 (BP5) region of the 16p11.2 recurrent CNV.

(D) The AT-rich sequence and a control sequence are inserted into the upstream region of the GAL1 promoter in BY4742 yeast.

(E) Method for evaluating the gross genome rearrangement rate in yeasts. Yeasts that lose their chrV ends (containing URA3 and CAN1 genes) can exhibit resistance to 5-FOA and Can, forming double-resistant (RR) clones.

(F) The CRE gene expression is upregulated by the YPGal-induced GAL1 promoter activation, and the CRE gene expression levels are comparable in AT-rich and control yeasts (upper). YPGal culturation increases the RR clone number in AT-rich yeast but not in Control yeast. RR clones can also be generated in YPGlu AT-rich yeasts, but their numbers are not significantly different from those in YPGlu Control yeasts. ∗∗, p < 0.01; ns, not significant; Student’s t test.

(G) The breakpoints in both YPGal and YPGlu RR AT-rich clones are located at the AT-rich regions.

(H) YPGal culture promotes the cruciform structure formation of AT-rich sequence. ∗∗, p < 0.01; ns, not significant; Student’s t test. DNA marker sizes, see STAR Methods.

Then, we analyzed whether AT-rich sequences are associated with the formation of CNVs in human genome. CNV breakpoints information was archived by ClinVar.²⁶ The number of ClinVar CNV breakpoints located within 2.5 Kbp regions flanking the AT-rich or randomly selected control sequences was counted. As a result, we found AT-rich sequences associated with higher number of ClinVar CNV breakpoints than that of their corresponding random sequence groups (mean value 0.31 for AT-rich group vs. 0.27 for random sequence group, p < 0.01, Wilcoxon test, Figure 1B), indicating that AT-rich sequences are more likely to cause structure variants like CNVs than random control sequences. Then the genes located within the 5 Kbp flanking DNA regions of AT-rich sequences were analyzes. We found that these AT-rich associated genes are enriched in the nervous system development pathway, regardless of whether these AT-rich sequences overlapped with ClinVar CNV breakpoints (Table S4). Then, we analyzed the AT-rich sequences that overlapped with DSB peaks in MCF7 cells and were located at the breakpoints of reCNVs. As a result, 29 AT-rich sequences were found within 8 breakpoint regions associated with reCNVs. These reCNVs included 10q22.3q23.2 (1 AT-rich region), 16p13.11 (3 AT-rich regions), 16p11.2 (1 AT-rich region), 22q11.21 (11 AT-rich regions), and 22q11.2 (24 AT-rich regions) reCNVs (Table S4).

In the breakpoint region of 16p11.2 reCNV, the AT-rich sequence (chr16:30201399-30201815) locates downstream of CORA1A and BOLA2B genes and upstream of SLX1A and SLX1A-SULT1A3 genes (Figure 1C). To analyze whether the 16p11.2 reCNV associated AT-rich sequence could be a source of structural variant formation, we investigated its genome stability in a budding yeast system. In the URA3 inactive BY4742 yeast, the HXT13 gene was replaced with an active URA3 gene. AT-rich and a control sequence (chr16:30200828-30201319) (Data S1) were then inserted into the CIN8 gene region, respectively. Both AT-rich and control sequences were positioned upstream of a GAL1 promoter, which controls the expression of an exogenous gene CRE (Figure 1D). Then, the AT-rich and Control yeasts that lost their chrV ends containing the CAN1 gene, which makes yeast sensitive to Canavanine (Can), and the URA3 gene, which makes yeast sensitive to 5-fluoroorotic acid (5-FOA), could be selected using the 5-FOA + Can plates. To analyze the effects of transcription on AT-rich sequence stability, both AT-rich and Control yeasts were cultured in synthetic defined medium (SD) + 2% Glycerol + G418 for more than 3 h and 10⁶ yeasts were cultured in 10 mL YPGal or YPGlu media for 24 h. Subsequently, water drops containing 10⁷ yeasts were spotted on the 5-FOA + Can selection plates. After 8 days of culturing, the RR yeast clone numbers were counted to evaluate the de novo gross genome rearrangement induced by the AT-rich or control sequence (Figure 1E). As a result, the GAL1 promoter was activated in both the AT-rich and Control yeasts by YPGal (Figure 1F), but no RR clone was found in the Control yeasts. There were only 3 clones found in 18 drops of YPGlu AT-rich yeasts, but an average of 1.39 clones per 10⁷ yeasts (C/10⁷) was found in YPGal AT-rich yeasts, which was significantly higher than that in the other groups (p < 0.01, Figure 1F).

To analyze the breakpoint position where gross genome rearrangement was generated in AT-rich yeasts, we amplified the DNA fragment covering the AT-rich sequence (X2 in Figure 1D) and an adjacent upstream fragment of AT-rich sequence (X1 in Figure 1D). As a result, we identified the breakpoints of RR AT-rich yeast clones induced by YPGal and YPGlu, which were situated in the AT-rich regions (Figure 1G). Then, we analyzed whether the GAL1 promoter mediated transcription promotes cruciform DNA formation at AT-rich regions. Using the cruciform DNA antibody-based immunoprecipitation-PCR method, we found that cruciform structures were highly enriched in the AT-rich regions in YPGal-cultured AT-rich yeasts (Figure 1H). This indicates that transcription promotes the extrusion of cruciform structures from upstream AT-rich regions.

With similar methods, we analyzed the genome stability of another human AT-rich sequence which locates at the breakpoint D (BP-D) region of the 22q11.2 reCNV. For this AT-rich sequence (chr22:21681177-21681911, 22q11.2 AT-rich), there is a noncoding gene FAM230H (NR_136559.2) located at its 251 bp downstream region (Figure 2A; Table S4). This 22q11.2 AT-rich and its associated control sequence (chr22:21675551-21677460, 22q11.2 Control) were combined with the GAL1 promoter and inserted into the yeast genome (Figure 2B; Data S2). When the 22q11.2 Control yeasts were cultured in YPGal or YPGlu for 24 h, we found that both YPGal and YPGlu could induce RR yeast clone formation, but there was no significant difference in RR clone number between these two groups (2.61 C/10⁷ when cultured by YPGal and 2.00 C/10⁷ when cultured by YPGlu, Figure 2C). However, for the 22q11.2 AT-rich yeast, the YPGal induced RR clone numbers were significantly larger than that induced by YPGlu (6.11 C/10⁷ for YPGal cultured yeasts and 0.50 C/10⁷ for YPGlu cultured yeasts, p < 0.01, Figure 2C), indicating YPGal induced GAL1 promoter activation increased the fragility of 22q11.2 AT-rich sequence. Then the breakpoints of the 22q11.2 AT-rich and Control sequences induced RR yeasts were analyzed. For the 22q11.2 AT-rich yeasts, we found all the breakpoints in YPGlu and YPGal induced RR yeasts located at the AT-rich regions (23/23 RR clones in YPGlu group and 23/23 RR clones in YPGal group). For the 22q11.2 Control yeasts, we found all the breakpoints in YPGlu and YPGal induced RR yeasts located in the downstream regions of the 22q11.2 Control sequence (23/23 RR clones in YPGlu group and 23/23 RR clones in YPGal group, Figure 2D). Similar with 16p11.2 reCNV associated AT-rich sequence, we found GAL1 promoter activation could also promote the cruciform structure formation in the 22q11.2 AT-rich region (Figure 2E).

Figure 2.

Transcription promotes upstream 22q11.2 reCNV associated AT-rich instability

(A) Location of 22q11.2 reCNV associated AT-rich sequence (within the red dashed box) in human genome.

(B) The 22q11.2 reCNV associated AT-rich sequence and control sequence are inserted into the upstream region of the GAL1 promoter in BY4742 yeast.

(C) For the 22q11.2 AT-rich yeast, YPGal induced GAL1 promoter activation increased the 22q11.2 AT-rich fragility. For the 22q11.2 control yeast, no significant difference of the 22q11.2 Control sequence fragility has been found between the YPGlu and YPGal group.

(D) For the mutated 22q11.2 AT-rich yeasts induced by either YPGlu or YPGal, the breakpoints locate in the AT-rich regions. For the mutated 22q11.2 Control yeasts, the breakpoints locate in the downstream region of the Control sequence.

(E) YPGal culturation promotes the cruciform structure formation of 22q11.2 associated AT-rich sequence. ∗, p < 0.05; ∗∗, p < 0.01; ns, not significant; Student’s t test. DNA marker sizes, see STAR Methods.

AT-rich sequences mediate the copy number loss in yeast

In human genome, reCNV formation are mostly mediated by the NAHR.²⁷ To analyze whether 16p11.2 reCNV associated AT-rich sequence (hereafter the AT-rich and Control sequences refer in particular to the 16p11.2 reCNV associated AT-rich and Control sequences respectively) could mediated CNV in yeast, AT-rich-GAL1p-CRE-LEU2 and Control-GAL1p-CRE-LEU2 cassettes were inserted into the AT-rich yeasts by replacing the AVT2 gene on chrV (Figure 3A). The new yeast strains were named as ATrich-ATrich yeast and ATrich-Control yeasts respectively. In the ATrich-ATrich yeasts, CAN1 gene were surrounded by KanMX-ATrich-GAL1p-CRE and ATrich-GAL1p-CRE-LEU2 cassettes. In the ATrich-Control yeasts, CAN1 gene were surrounded by KanMX-ATrich-GAL1p-CRE and Control-GAL1p-CRE-LEU2. If AT-rich induced DSBs could be repaired by NAHR and caused the loss of CAN1 gene, then the mutated yeasts could be selected by the Can plates. After cultured in SD ARG-/URA-/LEU-/G418 media overnight, 10⁶ ATrich-ATrich and ATrich-Control yeasts were cultured in 10 mL YPGal or YPGlu media for 24h respectively, and then the Can resistant yeasts were selected by two kinds of plates: ARG- Can plates, which select the yeast whose CAN1 is lost or mutated, regardless whether LEU2 and URA3 are lost; and the ARG-/URA-/LEU-/Can plates, which select the yeast whose CAN1 is lost or mutated but LEU2 and URA3 are still exist (Figure 3B).

Figure 3.

16p11.2 reCNV associated AT-rich sequence induces copy number loss in yeast

(A) The yeast models used for copy number variant formation analysis.

(B) Method for selection of yeast clones which lost the CAN1 gene.

(C) YPGal increases the number of Canavanine (Can) resistant clones in both ATrich-ATrich and ATrich-Control yeasts.

(D) Detection of whether CAN1 gene is deleted in the Can resistant yeast clones by PCR method.

(E) PacBio HiFi sequencing of the mutated regions of Can resistant yeasts. For each yeast group, 8 clones are chosen for HiFi sequencing. For ATrich-ATrich yeasts, both YPGlu and YPGal induced Can resistant clones (8 in 8 clones) lost their genome regions containing AT-rich sequence, CRE and CAN1 genes. For ATrich-Control yeasts, 8 in 8 YPGal induced Can resistant clones and 1 in 8 YPGlu induced Can resistant clone lost their genome regions containing CRE, CAN1 and Control sequence; 7 in 8 YPGlu induced Can resistant clone lost their genome regions containing AT-rich sequence, CRE and CAN1.

(F) The genome elements of ATrich-Control yeast.

(G) The HiFi sequencing results of YPGlu induced Can resistant ATrich-Control yeast clones which lost the AT-rich sequence. ∗∗, p < 0.01; ns, not significant; Student’s t test. DNA marker sizes, see STAR Methods.

As a result, when ATrich-ATrich yeasts were selected by ARG- Can plates or ARG-/URA-/LEU-/Can plates, YPGal media significantly increased the Can resistant yeast clone number comparing with that caused by YPGlu (when yeasts were selected by ARG-/Can plates, 98.22 C/10⁴ for YPGal group vs. 34.83 C/10⁴ for YPGlu group, p < 0.01; when yeasts were selected by ARG-/URA-/LEU-/Can plates, 101.33 C/10⁴ for YPGal group vs. 37.89 C/10⁴ for YPGlu group, p < 0.01; Figure 3C). when ATrich-Control yeasts were selected by Can plates, YPGal media also significantly increased the Can resistant yeast clone number (when yeasts were selected by ARG-/Can plates, 7.61 C/10⁶ for YPGal group vs. 2.11 C/10⁶ for YPGlu group, p < 0.01; when yeasts were selected by ARG-/URA-/LEU-/Can plates, 7.17 C/10⁶ for YPGal group vs. 2.50 C/10⁶ for YPGlu group, p < 0.01; Figure 3C). For both ATrich-ATrich and ATrich-Control yeasts, when yeasts were selected by different Can plates (Figure 3B), the Can resistant clone numbers had no significant difference (Figure 3C).

Then we analyzed whether CAN1 gene were lost and URA3 gene still exit in the Can resistant yeast clones using PCR method. As a result, for all the Can resistant yeasts had been analyzed (23 clones for each group), the CAN1 gene fragment couldn’t be amplified but URA3 gene fragments had been amplified (Figure 3D). To validate whether CAN1 genes were lost in these Can resistant yeast clones, we amplified the genome region from KanMX to the LEU2 (Figure 3E) from the YPGal and YPGlu induced Can resistant ATrich-ATrich and ATrich-Control yeasts (8 clones for each group). After single molecule sequencing, we found both YPGal and YPGlu induced Can resistant ATrich-ATrich yeasts had lost their CAN1 gene by an NAHR manner,²⁷ indicating AT-rich sequence in yeast could induce copy number loss. For the YPGal induced Can resistant ATrich-Control yeasts, we found all 8 clones lost their CAN1 genes and the Control sequences. For the YPGlu induced Can resistant ATrich-Control yeasts, we found one clone lost their CAN1 gene and the Control sequence, but the other 7 clones lost their AT-rich sequences and CAN1 genes (Figure 3E; Data S3). The exact mechanism for how CAN1-Control sequences were deleted in the ATrich-Control yeast is not known. It might be caused by NAHR-repair of DSBs occurring in the GAL1p-CRE region. In the ATrich-Control yeast, there is a short homologous sequence (green in Figure 3F) in the ATrich-GAL1p-CRE and Control-GAL1p-CRE cassettes, which might mediate the NAHR-mediated ATrich-CAN1 loss in the YPGlu induced Can resistant ATrich-Control yeast (Figure 3G).

Deletion of RAD1, RAD2, or RAD10 promotes transcription-mediated AT-rich instability

It has been reported that AT-rich formed DNA structures could be cleaved by MUS81²⁸ and GEN1 (also referred as YEN1 in yeast).¹⁴ MUS81 has been reported to be associated with AT-rich genome instability induced by replication stress,^28,29 but it was unknown whether it causes the transcription-mediated AT-rich instability. To comprehensively analyze which DNA structure-specific endonucleases caused the transcription-mediated AT-rich instability, we deleted structure-specific endonuclease genes MUS81, YEN1, RAD1 (also referred as ERCC1 or XPF), RAD2 (also referred as ERCC5 or XPG), and RAD27 (also referred as FEN1)³⁰ in the AT-rich yeast, and evaluated the stability of AT-rich regions. As a result, compared to normal AT-rich yeast (1.67 and 0.11 RR C/10⁷ when cultured by YPGal and YPGlu respectively), the deletion of RAD1, RAD2, YEN1, or MUS81 had no obvious effects on AT-rich stability when the GAL1 promoter was suppressed by YPGlu. However, the RR clones were decreased in yen1Δ (0.94 C/10⁷, p < 0.05) and mus81Δ (0.50 C/10⁷, p < 0.01) AT-rich yeasts, but were increased in rad1Δ (12.33 C/10⁷, p < 0.01) and rad2Δ (2.67 C/10⁷, p < 0.05) AT-rich yeasts (Figure 4A). Using breakpoint analysis by PCR, we found that 22 out of 23 (22/23) RR rad1Δ clones were generated by DNA cleavage at the AT-rich region, while 1/23 clone was generated by DNA cleavage between the AT-rich region and the CRE gene body (Figures 4B and S1). It has been reported that RAD1 participates in homologous recombination repair of DSB when resected 3′ overhang contains AT-rich induced stem loops and other non-B structures.³¹ The increase in RR clones in rad1Δ AT-rich yeasts might be caused by interference in the repair of DSBs generated at AT-rich regions.

Figure 4.

RAD1/10 complex plays roles in AT-rich stability

(A) Disruption of RAD1 or RAD2 can significantly increase the YPGal-induced RR clone numbers, whereas disruption of MUS81 or YEN1 decreased the RR clone numbers. RAD27 disruption increased RR clone numbers in both YPGal and YPGlu cultured yeasts.

(B) The breakpoints in rad1Δ RR clones are mostly (22 out of 23 clones) located in AT-rich regions, whereas in rad27Δ YPGal and YPGlu RR yeasts, the number of breakpoints located in AT-rich regions is 3 and 0, respectively (Figure S1). First lane: DNA marker; second lane: normal AT-rich yeasts; third and other lanes: RR yeasts.

(C) Disruption of YEN1 or MUS81 in rad1Δ yeast decreases the YPGal-induced RR clone numbers.

(D) Disruption of SLX1 or SLX4 has no obvious effects on the AT-rich stability in YPGal culturation conditions.

(E) YPGal-cultured rad10Δ yeasts can generate more RR clones than normal AT-rich yeasts but fewer than rad1Δ yeasts. The YPGal-induced RR clone numbers are comparable between rad1Δ yeasts and rad1Δ rad10Δ yeasts. ∗, p < 0.05; ∗∗, p < 0.01; ns, not significant; Student’s t test. DNA marker sizes, see STAR Methods.

When RAD27 was deleted, we observed a significant increase in RR clones in both YPGal (115.89 C/10⁷, p < 0.01) and YPGlu (27.72 C/10⁷, p < 0.01) cultured yeasts compared to normal AT-rich yeasts (Figure 4A). However, we found that only 3/23 of RR rad27Δ clones were generated by AT-rich cleavage in the YPGal group, and no clone was generated by AT-rich cleavage in the YPGlu group (Figure 4B). In the YPGal rad27Δ RR clones, 1/23 clone was generated by DNA cleavage between AT-rich sequence and its upstream KanMX gene (Figure S1). In the YPGlu RR rad27Δ clones, we found that the breakpoints in 10/23 clones were located between the AT-rich sequence and KanMX gene (Figure S1). These data indicate that RAD27 deletion had more unfavorable influences on the overall yeast genome stability, making it difficult to determine whether rad27Δ directly affects AT-rich stability.

To validate the effects of MUS81 and YEN1 on AT-rich stability, these two genes were also deleted in the rad1Δ AT-rich yeasts. After culturing in YPGal, we found that the mean RR clone numbers in rad1Δ yen1Δ yeasts (3.11 C/10⁷) and rad1Δ mus81Δ yeasts (1.17 C/10⁷) were significantly less than that in rad1Δ yeasts (10.72 C/10⁷, p < 0.01, Figure 4C). The number of clones in rad1Δ mus81Δ yeasts was significantly lower than that in rad1Δ yen1Δ yeasts (p < 0.01, Figure 4C). These data confirm that both YEN1 and MUS81 participate in the cleavage of AT-rich formed DNA structures, and indicate that MUS81 plays a more critical role in cleaving AT-rich structures than YEN1 in yeasts.

It has been reported that SLX1/4 and RAD1/10 complexes participate in the MUS81 cleavage of replication stress-induced AT-rich instability.²⁸ SLX4 complexed with RAD1/10 participates in the removal of 3′ overhangs during single-strand annealing repair of DSB. It can also form a complex with SLX1 as a co-activator to enhance the 5′-flap endonuclease activity of SLX1.³² To analyze whether SLX1 and SLX4 are associated with transcription-mediated AT-rich instability, the SLX1 and SLX4 genes were deleted in normal AT-rich yeast and rad1Δ AT-rich yeast. After being cultured in YPGal, the RR clone numbers generated in slx1Δ (2.05 C/10⁷) and slx4Δ (1.94 C/10⁷) yeasts showed no significant difference compared to normal AT-rich yeasts (2.00 C/10⁷). Similarly, the RR clone numbers in rad1Δ slx1Δ (11.56 C/10⁷) and rad1Δ slx4Δ (10.56 C/10⁷) yeasts also did not significantly differ from rad1Δ AT-rich yeasts (12.28 C/10⁷, Figure 4D), indicating that SLX1 and SLX4 may not be involved in the transcription-mediated AT-rich instability.

After deleting of RAD10 in normal AT-rich yeast and rad1Δ AT-rich yeast, we assessed the stability of AT-rich. As a result, deletion of RAD10 also increases the mean RR clones in the YPGal group (6.78 C/10⁷ vs. 1.06 C/10⁷ in normal AT-rich yeasts, p < 0.01). However, the mean clone numbers in rad10Δ yeasts were lower than those in rad1Δ yeasts (12.44 C/10⁷). There was no obvious effect on RR clone numbers in YPGlu-cultured yeasts by RAD10 deletion (0.28 C/10⁷ in rad10Δ AT-rich yeasts vs. 0.11 C/10⁷ and 0.28 C/10⁷ in normal and rad1Δ AT-rich yeasts, Figure 4E). When cultured in YPGal, RR clones generated in the rad1Δ rad10Δ AT-rich yeasts (12.61 C/10⁷) showed no significant difference compared to those in rad1Δ AT-rich yeasts but were significantly higher than in rad10Δ AT-rich yeasts (p < 0.01). When cultured in YPGlu, RR clones generated in the rad1Δ rad10Δ AT-rich yeast showed a slight increase (1.17 C/10⁷) compared to that in rad1Δ, rad10Δ, and normal AT-rich yeasts (p < 0.01, Figure 4E).

Homologous recombination participates in transcription-mediated AT-rich DSB repair

As the RAD1/10 complex might attenuate AT-rich sequence instability by promoting homologous recombination repair of the AT-rich-generated DSB,³¹ we analyzed whether the homologous recombination associated factor RNase H could increase AT-rich instability. When RNH1 and RNH201 were double deleted in normal AT-rich yeast and rad1Δ AT-rich yeast, the methyl methanesulfonate (MMS) resistance capability, which mostly represents the efficiency of homologous recombination repair,^33,34 was decreased in these yeasts (Figure 5A). In addition, RNH1 and RNH201 double deletion also slightly increased the sensitivity of AT-rich yeasts to DSB inducer Zeiocin, but it hadn’t obviously increase yeast sensitivity to H₂O₂. No obvious difference of yeast sensitivity to H₂O₂ and Zeiocin had been found between RAD1 deleted and normal AT-rich yeast. But deletion of RAD1 in rnh1Δ rnh201Δ AT-rich yeasts slightly increased the yeast sensitivity to H₂O₂ (Figure S2). When cultured in YPGal, rnh1Δ rnh201Δ AT-rich yeasts generated more RR clones (4.94 C/10⁷) than rnh1Δ (1.61 C/10⁷), rnh201Δ (1.33 C/10⁷), and normal (1.06 C/10⁷) AT-rich yeasts (p < 0.01, Figure 5B). When cultured in YPGlu, rnh1Δ rnh201Δ AT-rich yeasts also generated slightly more RR clones (0.61 C/10⁷, p < 0.05) than rnh201Δ (0.22 C/10⁷) and normal (0.22 C/10⁷) AT-rich yeasts, but not more than that in rnh1Δ (0.33 C/10⁷) AT-rich yeasts (Figure 5B). Similarly, rad1Δ rnh1Δ rnh201Δ triple disrupted AT-rich yeasts generated significantly more RR clones (34.06 C/10⁷) than rad1Δ rnh1Δ (13.44 C/10⁷), rad1Δ rnh201Δ (11.56 C/10⁷), and rad1Δ (11.61 C/10⁷) AT-rich yeasts (p < 0.01, Figure 5B) when cultured in YPGal. When cultured in YPGlu, rad1Δ rnh1Δ rnh201Δ AT-rich yeasts produced more RR clones (1.00 C/10⁷) than rad1Δ rnh1Δ (0.39 C/10⁷, p < 0.05), rad1Δ rnh201Δ (0.22 C/10⁷, p < 0.01), and rad1Δ (0.11 C/10⁷, p < 0.01) AT-rich yeasts (Figure 5C).

Figure 5.

Homologous recombination proteins are involved in the repair of transcription-mediated DSB in AT-rich

(A) The sensitivity of AT-rich yeasts to methyl methanesulfonate (MMS) is increased by double deletion of RNH1/201 or single deletion of RAD1. Triple deletion of RAD1 and RNH1/201 further increases the MMS sensitivity of AT-rich yeasts.

(B and C) RNH1/201 double deletion significantly increases the number of RR clones in normal AT-rich yeasts and rad1Δ AT-rich yeasts cultured in YPGal.

(D) Breakpoint identification of YPGal-induced rnh1Δ rnh201Δ RR yeasts and rnh1Δ rnh201Δ rad1Δ RR yeasts.

(E) Deletion of RAD51, but not YKU70, increases the sensitivity of normal AT-rich yeasts and RNH1/201 double-deleted AT-rich yeasts to MMS.

(F) The effects of RAD51 or YKU70 deletion on the gross genome rearrangement rates of normal and rnh1Δ rnh201Δ AT-rich yeasts.

(G) Compared to RAD51 single deletion or RNH1/201 double deletion, RAD51 and RNH1/201 triple deletion significantly increases the transcription induced RR clone numbers in AT-rich yeasts.

(H) Breakpoint identification of YPGal- or YPGlu-induced RR clones generated from rad51Δ, rad51Δ rnh1Δ rnh201Δ, and yku70Δ rnh1Δ rnh201Δ AT-rich yeasts. First lane: DNA marker; second lane: normal AT-rich yeasts; third and other lanes: RR yeasts. ∗, p < 0.05; ∗∗, p < 0.01; ns, not significant; Student’s t test. DNA marker sizes, see STAR Methods.

The breakpoints of RR AT-rich yeasts resulting from the double deletion of RNH1/201 and the triple deletion of RNH1/RNH201/RAD1 were analyzed using PCR. As a result, we found breakpoints in 9/23 rnh1Δ rnh201Δ RR clones located at AT-rich sequences, whereas in 14/23 clones the breakpoints were located between AT-rich sequence and the KanMX gene (Figure 5D). However, all breakpoints in YPGal-induced rad1Δ rnh1Δ rnh201Δ RR clones were located at the AT-rich region (Figure 5D). Similar to the breakpoint positions in YPGal-induced RR clones, in the YPGlu-induced rnh1Δ rnh201Δ RR clones, 11/23 breakpoints were located in AT-rich regions, and 12/23 breakpoints were situated between the AT-rich sequence and the KanMX gene. For YPGlu-induced rad1Δ rnh1Δ rnh201Δ RR clones, 20/23 breakpoints were located in AT-rich regions, 1/23 located between AT-rich sequence and CRE gene, and 2/23 located downstream of the CRE gene (Figure S1).

To further study the repair mechanism of AT-rich-induced DSBs, we deleted the critical homologous recombination repair factor gene RAD51 and the non-homologous end-joining factor gene YKU70 in normal and rnh1Δ rnh201Δ AT-rich yeast. As a result, deletion of RAD51 but not YKU70, increased the sensitivity of normal and rnh1Δ rnh201Δ AT-rich yeasts to MMS (Figure 5E). In addition, deletion of RAD51 obviously increased the sensitivity of normal AT-rich and rnh1Δ rnh201Δ AT-rich yeasts to Zeocin, whereas deletion of YKU70 had no obvious effect of yeast sensitivity to either H₂O₂ or Zeocin (Figure S2).

When cultured in YPGal, the RR clone numbers significantly increased progressively in rnh1Δ rnh201Δ (4.83 C/10⁷), rad51Δ (10.39 C/10⁷), and rad51Δ rnh1Δ rnh201Δ (53.39 C/10⁷) AT-rich yeasts, compared to normal AT-rich yeast (1.39 C/10⁷, p < 0.01 for every comparison, Figure 5F). When cultured in YPGlu, rad51Δ rnh1Δ rnh201Δ AT-rich yeasts generated significantly more RR clones (10.17 C/10⁷) than normal AT-rich yeasts (0.22 C/10⁷). However, rad51Δ (0.83 C/10⁷) and rnh1Δ rnh201Δ (0.89 C/10⁷) AT-rich yeasts only produced slightly more RR clones than normal AT-rich yeasts (Figure 5F). Then we compared the numbers of RR clones generated by transcription alone between rad51Δ single disrupted, rnh1Δ rnh201Δ double disrupted, and rad51Δ rnh1Δ rnh201Δ triple disrupted AT-rich yeasts by subtracting the mean value of RR clone numbers caused by YPGlu from the RR clone numbers caused by YPGal. As a result, transcription alone induced RR clones were significantly more abundant in rad51Δ rnh1Δ rnh201Δ AT-rich yeasts than in rad51Δ or rnh1Δ rnh201Δ AT-rich yeasts (Figure 5G).

Unlike RAD51 deletion, we found that YKU70 single deletion did not affect the AT-rich stability in both YPGlu (0.06 C/10⁷) and YPGal (1.44 C/10⁷) cultured yeasts compared to normal AT-rich yeasts. However, when YKU70 was deleted in rnh1Δ rnh201Δ AT-rich yeasts, the RR clone numbers significantly increased compared to yku70Δ AT-rich yeasts in both YPGlu (35.39 C/10⁷, p < 0.01) and YPGal (32.50 C/10⁷, p < 0.01) cultured yeasts (Figure 5F). In addition, no significant difference in yku70Δ rnh1Δ rnh201Δ RR clone numbers was found between the YPGlu and YPGal groups (Figure 5F).

In YPGal-generated rad51Δ RR clones, breakpoints of 21/23 clones were located between AT-rich sequence and KanMX gene, while 2/23 were located at the AT-rich region. In YPGlu-generated rad51Δ RR clones, 18/23 were located between AT-rich sequence and KanMX gene, while 5/23 were located in the AT-rich region, with no significant difference compared to the YPGal clones (p > 0.05, Fisher’s exact test). However, for the YPGal-generated rad51Δ rnh1Δ rnh201Δ RR yeasts, the number of clones with breakpoints located between AT-rich sequence and KanMX or located in the AT-rich region was 22/23 and 1/23, respectively. In contrast, the number of clones for YPGlu-generated rad51Δ rnh1Δ rnh201Δ RR yeasts changed significantly to 9/22 and 14/22, respectively (p < 0.01, Fisher’s exact test, Figure 5H). For the YPGal and YPGlu generated yku70Δ rnh1Δ rnh201Δ RR yeasts, the breakpoints were mostly located between AT-rich sequence and KanMX gene (23/23 for the YPGal group and 22/23 for the YPGlu group, p > 0.05, Fisher’s exact test, Figure 5H).

The effects of topoisomerase poisons on the stability of AT-rich sequences

It has been reported that TOP2 plays critical roles in the formation of negative supercoiling at gene boundaries¹⁶ and contributes to the specific DNA structure-mediated DSB formation.³⁵ Unlike TOP2, although TOP1 also regulates DNA transcription, it mainly function in removing positive supercoils generated ahead of RNA polymerase II.³⁶ To analyze whether RR clone numbers generated in AT-rich yeasts were associated with DNA topoisomerases, we treated the rad1Δ and rad51Δ AT-rich yeasts with a low dose of the TOP2 inhibitor Etoposide (Etop, 1 μM or 2 μM) or the TOP1 inhibitor Camptothecin (CPT, 0.5 μM).

Comparing with control rad1Δ AT-rich yeasts, when treated with a low dose of Etop or CPT for 24 h in YPGal, there was no significant change in the RR clone number in each group. When threated with Etop or CPT in YPGlu, only the RR clone number in 2 μM Etop treatment group was slightly higher (0.72 C/10⁷) than that in the control group (0.11 C/10⁷, p < 0.05, Figure 6A). All the breakpoints (23/23) of rad1Δ RR clones generated in the control group, 2 μM Etop treatment, and CPT treatment groups in YPGal were found to be located at the AT-rich regions (Figure 6B), indicating that Etop and CPT hadn’t affected the mutation features in rad1Δ RR yeasts.

Figure 6.

The effects of topoisomerase inhibitors on the AT-rich stability

(A) A low dose of the topoisomerase II inhibitor Etoposide (Etop) or the topoisomerase I inhibitor camptothecin (CPT) has no significant effect on the YPGal-induced RR clone numbers in rad1Δ AT-rich yeasts.

(B) Breakpoint identification of rad1Δ RR clones induced by YPGal and treated with inhibitors.

(C) Both 2 μM of Etop and 0.5 μM of CPT can increase the YPGal-induced RR clone numbers in rad51Δ AT-rich yeasts.

(D) But only CPT treatment increased the transcription-mediated RR clone numbers in rad51Δ AT-rich yeasts.

(E) Breakpoint identification of rad51Δ RR clones induced by YPGal or YPGlu and treated with Etop.

(F) Breakpoint identification of rad51Δ RR clones induced by YPGal or YPGlu and treated with CPT. For all DNA gel electrophoresis results, first lane: DNA marker; second lane: normal AT-rich yeasts; third and other lanes: RR yeasts. ∗, p < 0.05; ∗∗, p < 0.01; ns, not significant; Student’s t test. DNA marker sizes, see STAR Methods.

Unlike the results in rad1Δ AT-rich yeasts, for the rad51Δ AT-rich yeasts, we found that 2 μM Etop treatment could significantly increase the RR clone numbers in both YPGal (14.00 C/10⁷) and YPGlu (3.56 C/10⁷) groups when compared to that in the control yeasts (10.94 C/10⁷ in YPGal, p < 0.05; 1.39 C/10⁷ in YPGlu, p < 0.01) or 1 μM Etop-treated yeasts (9.89 C/10⁷ in YPGal, p < 0.01; 1.61 C/10⁷ in YPGlu, p < 0.01, Figure 6C). However, after deducing the mean value of RR clone numbers induced by YPGlu, we found no significant difference in the transcription-mediated RR clone numbers between 2 μM Etop-treated and control rad51Δ AT-rich yeasts (Figure 6D). When compared to the control rad51Δ AT-rich yeasts, we observed an increase in RR clone numbers when treated with CPT in both YPGlu (6.94 C/10⁷, p < 0.01) and YPGal (21.83 C/10⁷, p < 0.01) groups (Figure 6C). After deducing the mean value of RR clone number caused by YPGlu, the RR clone number in CPT-treated group was still larger than that in the control group (Figure 6D), indicating that CPT might have increased the transcription-mediated AT-rich fragility.

We then detected the breakpoints of rad51Δ RR clones in the yeast treated with 2 μM Etop. In the YPGal group, we found the breakpoints in 16/23 RR clones situated between AT-rich sequence and KanMX and 7/23 located in AT-rich regions, with no significant difference when compared to normal rad51Δ RR clones (p > 0.05, Fisher’s exact test, Figures 5H and 6E). In the YPGlu group, 5/23 breakpoints were found located between AT-rich and KanMX, and 18/23 located at AT-rich regions. This distribution was significantly different from that in normal rad51Δ RR clones (p < 0.01, Fisher’s exact test, Figures 5H and 6E).

For the breakpoints of rad51Δ RR clones in the CPT-treated YPGal-cultured yeasts, 5/23 located between AT-rich and KanMX gene, while 18/23 were distributed even farther, between the KanMX gene and PRB1 gene (Figure 1D). This distribution was significantly different from that in normal YPGal-cultured rad51Δ RR clones (p < 0.01, Chi-squared Test, Figures 5H and 6F). In the YPGlu group, 18/23 breakpoints were located between AT-rich and KanMX gene and 5/23 located at AT-rich regions, which was consistent with the findings in normal YPGlu-cultured rad51Δ RR clones (Figures 5H and 6F).

Discussion

In this study, when rad51Δ yeasts were treated with Etop or CPT, no breakpoints outside of the genome region from PRB1 to AT-rich were identified in the RR clones. Since the PRB1 gene locates adjacent to the KanMX-ATrich-GAL1p cassette, and no additional RR clones were induced by Etop or CPT in rad1Δ yeasts, the increase in RR clone numbers in Etop or CPT treated rad51Δ yeasts may be attributed to the combined effects of AT-rich and topoisomerase poisons, rather than the individual effects of topoisomerase poison or AT-rich. In Arabidopsis thaliana, TOP2 plays a role in DSB homologous recombination repair.³⁷ It has been proposed that topoisomerases might also participate in relaxing the DNA donor state to promote DNA synthesis in the D-loop.³⁸ However, the impact of topoisomerases on the AT-rich associated DSB repair still needs to be explored. As CPT can move the breakpoint of rad51Δ RR yeasts forward, we propose that TOP1 might participate in the AT-rich DSB repair in rad51Δ yeasts. TOP1 mainly functions on DNA torsion removal and regulating DNA replication and transcription by nicking and resealing the DNA.^39,40 Deletion of TOP1 would limit the transcription of long genes⁴¹ and decrease the DNA replication speed on long chromosomes.⁴² Unlike TOP1 deletion, TOP1 inhibitor CPT suppresses DNA resealing but not nicking activity, and generates covalent TOP1-DNA crosslink.⁴³ In this study, the effects of CPT on breakpoint positions might be caused by TOP1 poisons but not the accumulation of DNA torsion.

It has been reported that replication stress is a critical driver of AT-rich-mediated genome instability.⁴⁴ In this study, we found transcription-mediated AT-rich instability in yeast and investigated how transcription-mediated AT-rich DSBs were repaired (Figure 7). Based on our results, we hypothesized that transcription promotes the accumulation of negative supercoiling at the gene boundaries.¹⁶ Negative supercoiling then stimulate the transformation of AT-rich regions into cruciform structures (Figure 7A).^2,4 Subsequently, these cruciform DNA structures are cleaved by MUS81 or YEN1, generating DSBs with stem structures at their ends (Figure 7B). Afterward, the DSB end is resected, and the 3′ overhangs are protected by DNA:RNA hybrid formation. If RNH1 and RNH201 are disrupted, the homologous recombination repair will be suppressed, leading to the formation of gross genome rearrangements due to various mechanisms (Figure 7C). Else, the RNA in DNA:RNA hybrids is degraded by RNase H, and the single strand 3′ overhangs invade the sister chromatid with the assistance of RAD51. If RAD51 is disrupted, gross genome rearrangements will form and the breakpoints will be located at the 3′ overhangs (Figure 7D). As there is no RAD51-mediated 3′ overhang invading, prolonged exposure may lead to the coiling of the 3′ overhang. If TOP1 is poisoned, the tangled single-strand DNA might cause DSB end resection to be even longer, leaving a longer 3′ overhang and shifting the breakpoint further forward from the AT-rich region (Figure 7E). If the 3′ overhang has been exchanged with the template DNA strand by RAD51, in order to synthesize new DNA, the stem structure at its end should be removed by the RAD1/10 complex. If RAD1 is disrupted, homologous recombination will still be suppressed, leading to gross genome rearrangement with breakpoints in the AT-rich region (Figure 7F).

Figure 7.

Predicted model of transcription-mediated DSB formation at AT-rich and the possible DSB repair mechanism

(A) Downstream transcription may promote the accumulation of negative supercoiling at AT-rich region, which then induces the cruciform formation.

(B) Cruciform structure is cleaved by the YEN1 or MUS81 endonuclease.

(C) After DSB ends are resected, 3′ overhangs form DNA:RNA hybrids with newly synthesized RNA. If the RNA in DNA:RNA hybrids cannot be degraded by RNase H, then the DSB will lead to gross genome rearrangements with breakpoints located in AT-rich regions or upstream regions of AT-rich.

(D) If RNA is degraded from DNA:RNA hybrids, but there is no RAD51 to mediate 3′ overhang strand exchange, then DSBs will generate gross genome rearrangements with breakpoints located in upstream region of AT-rich.

(E) If TOP1 is poisoned by CPT, the exposed 3′ overhangs of AT-rich DSBs in rad51Δ yeasts might become entangled, potentially impeding homologous recombination. Then the DSB ends might be resected longer and induce gross genome rearrangements with breakpoints located in upstream regions far away from AT-rich.

(F) If the 3′ overhang can form a D loop with template DNA but the stem residua cannot be removed by RAD1/10, normal homologous recombination will be interrupted, leading to gross genome rearrangements with breakpoints located at AT-rich regions. Unverified speculations are marked by question marks.

In addition to transcription-mediated AT-rich instability, we also found that the triple deletion of RNH1, RNH201, and YKU70 could significantly exacerbate the instability of AT-rich in a non-transcription-mediated manner. However, no obvious effect of YKU70 single disruption has been found on the AT-rich stability in either transcription or non-transcription conditions. Further investigations are needed to determine if there are complementary functions of RNase H and non-homologous end-joining in maintaining the stability of AT-rich regions. In our study, YPGlu-cultured normal 16p11.2 AT-rich yeasts and 22q11.2 AT-rich yeasts could also generate gross genome rearrangements with breakpoints located at AT-rich regions. Double deletion of RAD1/10, RNH1/RNH201, or single deletion of RAD51 could promote YPGlu-mediated RR clone formation. These gross genome rearrangements might also be caused by non-transcription-mediated AT-rich instability.

As the BY4742 yeast is haploid and has a different genome background compared to the human genome, the direct analysis of whether AT-rich mediates CNV formation in diploid cells were not conducted in this study. However, as 16p11.2 associated AT-rich region can induce copy number loss in ATrich-ATrich and ATrich-Control yeasts, we hypothesized that this AT-rich region can also generate DSBs and mediate the formation of reCNVs through NAHR in human cells. In this study, the gross genome rearrangements caused by AT-rich in yeasts occur at a rate of approximately 0–100/10⁷/24 h, which is much lower than the rates of reCNV formation (about 1/10000 to 1/1000 per generation) observed at 16p11.2 or 22q11.2 CNVs in humans.⁴⁵ However, for the ATrich-ATrich yeasts, the copy number loss rate dramatically increased to ∼100/10⁴/24 h, which is much faster than the rate of reCNV formation in humans. Why the copy number loss rate is so high in ATrich-ATrich yeasts is not well known in this study. This high formation rate of reCNVs in human might be caused by the persistent DNA replication in male germ stem cells or DNA transcription in female oocytes for ten years to several decades, which increases the risk of transcription-mediated or non-dependent ATrich-caused genome mutations. All these results revealed a new potential source for the formation of 16p11.2/22q11.2 reCNVs, as well as other AT-rich associated reCNVs, structure variants, and DSBs. This finding might be helpful for preventing AT-rich associated birth defect in humans in the future.

In summary, in this study we found the 16p11.2 and 22q11.2 reCNVs associated AT-rich sequences are unstable in yeast genome. The AT-rich sequence can induce gross genome rearrangement, including copy number loss in yeast genome. We found the stability of AT-rich in yeast is accelerated by downstream transcription. Downstream transcription promotes the formation of cruciform structure of AT-rich sequence. Transcription-mediated genome instability of AT-rich can be attenuated by MUS81 or YEN1 disruption and be further accelerated by disruption of homologous recombination associated genes such as RNase H, RAD1, and RAD51. Triple deletion of RNH1/201 and YKU70 can make AT-rich unstable in a transcription-independent manner. In the human genome, 16p11.2 CNV associated AT-rich locates at the downstream region of BOLA2B gene, and its homolog located at the intron region of BOLA2-SMG1P6 fusion gene (chr16:29462040-29462615). The 22q11.2 CNV associated AT-rich locates at the upstream region of FAM230H gene, and it has homologs located at the upstream of FAM230 family genes such as FAM230E, FAM230B, FAM230A, and FAM230D. The NAHR repair of DSB at the 16p11.2/22q11.2 reCNV associated AT-rich sequences might be one of the causes of 16p11.2/22q11.2 reCNV formation. However, whether the 16p11.2/22q11.2 reCNV associated AT-rich sequences could form cruciform structures and be affected by related gene transcription in human cells still requires further research. Nonetheless, the ATrich-mediated copy number change might be a possible mechanism of 16p11.2/22q11.2 reCNV formation and suppression of DSB formation in AT-rich regions would be a possible path to prevent reCNV de novo formation in humans.

Limitations of the study

In this study, there are some limitations to be considered. First, beside GAL1 promoter activation and gene disruption, the different metabolic microenvironments and gene expression profiles between YPGlu and YPGal cultured yeasts might also affect the AT-rich stability or DSB repair, and interfere with the results in this study. Second, in the yeast system, CIN8, HXT13, and/or AVT2 genes were substituted, which might have effects on the stability or topological structure of the yeast genome. Third, double deletion of RNH1/RNH201 will induce the accumulation of R-loops in the yeast genome and make the genome unstable. All these limitations may affect the yeast model used in this study.

Resource availability

Lead contact

Further information will be fulfilled by the lead contact (Jun-Yu Ma, majy@gd2h.org.cn).

Materials availability

The yeast strains are available from the lead contact upon request. Reagents and materials in this study are commercially available.

Data and code availability

• •Data: The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive⁴⁶ in National Genomics Data Center,⁴⁷ China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA: CRA018885) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa. The MCF-7 RNA-seq data was downloaded from SRA (https://www.ncbi.nlm.nih.gov/sra, accession number: SRR19737218) and MCF-7 END_seq data was downloaded from GEO (https://www.ncbi.nlm.nih.gov/geo, accession number: GSE99194). The clinvarCNV data was downloaded from UCSC genome browser website (https://hgdownload.soe.ucsc.edu/goldenPath/hg19/database) and the recurrent CNV data was downloaded from ClinGen website (https://search.clinicalgenome.org/kb/downloads).
• •Code: Codes for AT-rich sequence analysis are listed in Data S4 and any other codes and data underlying this article are available from the lead contact upon request.
• •Other items: All other information required to reanalyze the data reported are available from the lead contact upon request.

Acknowledgments

We thank all the members at Clinical Lab and Fertility Preservation Lab of the Reproductive Medicine Center at Guangdong Second Provincial General Hospital for their supports to this study. This work is supported by the National Key R&D Program of China (2022YFC2703204) and the National Natural Science Foundation of China (82271683).

Author contributions

J.-Y.M., S.Y., L.-N.C., and X.-H.O. designed the research; F.-Y.X., T.-J.X., and J.-Y.M. performed the analysis; F.-Y.X., J.C., X.X., T.-J.X., and S.L. analyzed the results; J.-Y.M. made figures; X.G.Z. and J.Y.M. wrote the article.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
cruciform DNA antibody	GeneTex	Cat#GTX54648
Chemicals, peptides, and recombinant proteins
Mouse IgG (H + L)	Sangon	Cat#D110503
Camptothecin	Acmec	Cat#7689-03-4
Etoposide	Beyotime	Cat#SC0173
Methyl methanesulfonate	Acmec	Cat#66-27-3
Protein A/G beads	MCE	Cat#HY-K0202
G418	Sangon	Cat#A600958
Canavanine sulfate salt	Sangon	Cat#A606173
5-fluoroorotic acid	Macklin	Cat#F832427
Zeocin Selection Antibiotic, Sterile	MCE	Cat#HY-K1053
Hydrogen peroxide solution	Macklin	Cat#H792073
DNA marker III	Tiangen	Cat#MD103
Zirconia beads	Youlisheng	Cat#111178579
Snailase	Solarbio	#S8280
Critical commercial assays
TIANamp Genomic DNA Kit	Tiangen	Cat#DP304
Es Taq Master Mix	CWbio	Cat#CW0690
Universal DNA Purification Kit	Tiangen	Cat#DP214
Hieff Clone Zero TOPO-TA Simple Cloning Kit	Yeasen	Cat#10908ES20
DNA Purification Kit with Magnetic Beads	Beyotime	Cat#D0041M
Rapid Taq Master Mix	Vazyme	Cat#P222-01
EZ-10 DNAaway RNA Mini-Preps Kit	Sangon	Cat#B618133
HiScript III 1st Strand cDNA Synthesis Kit	Vazyme	Cat#R312-01
AceQ qPCR SYBR Green Master Mix	Vazyme	Cat#Q111-02
FastPure Cell/Tissue DNA Isolation Mini Kit	Vazyme	Cat#DC102-01
Deposited data
Pacbio HiFi Seq data deposited on Genome Sequence Archive (https://ngdc.cncb.ac.cn/gsa)	This paper	CRA018885
MCF-7 RNA-seq data	SRA	SRR19737218
MCF-7 END_seq	GEO	GSE99194
clinvarCNV	UCSC	https://hgdownload.soe.ucsc.edu/goldenPath/hg19/database/
Recurrent CNV data	ClinGen	https://search.clinicalgenome.org/kb/downloads
Experimental models: Organisms/strains
Yeast strains	This paper	Table S1
Oligonucleotides
Primers	This paper	Table S2
Recombinant DNA
16p11.2_AT-rich and 16p11.2_Control cassettes in yeast	This paper	Data S1
22q11.2_AT-rich and 22q11.2_Control cassettes in yeast	This paper	Data S2
Software and algorithms
MACS3	Zhang et al.	https://github.com/macs3-project/MACS
Integrative Genomics Viewer	IGV team	https://igv.org/
Jupyter notebook	Jupyter	https://jupyter.org/
Flye-2.9.5	Pavel Pevzner’s lab	https://github.com/mikolmogorov/Flye
AT-rich sequence extration code	This paper	Data S4

Experimental model and study participant details

Amplification of AT-rich sequences

The 16p11.2 reCNV associated AT-rich sequence at chr16:30201379-30201957 (hg19) and the its corresponding control sequence at chr16:30200828-30201319, and the 22q11.2 reCNV associated AT-rich sequence (chr22:21681065-21682882) and its corresponding control sequence (chr22:21675551-21677460) were amplified from HeLa cell DNA extracted using the TIANamp Genomic DNA Kit (Tiangen, DP304). To amplify the AT-rich sequence from HeLa cell DNA, the PCR amplification temperature was set at 60°C. To amplify the control sequence, the amplification temperature was set at 72°C. Es Taq Master Mix (CW0690, CWbio) was used for the PCR amplification, and the PCR products were purified by Universal DNA Purification Kit (DP214, Tiangen). Then AT-rich and control sequences were inserted into the pESI-T vector using the Hieff Clone Zero TOPO-TA Simple Cloning Kit (10908ES20, Yeasen) for Sanger sequencing. After that, we amplified the 16p11.2_AT-rich, 16p11.2_Control, 22q11.2_AT-rich and 22q11.2_Control sequences using the M13 F/R primers on the pESI-T vector for further analysis.

Yeast strains

In this study, the BY4742 strain of yeast, in which the HXT13 gene was replaced by the URA3 gene, was used. Firstly, we amplified the DNA segment containing the GAL1 promoter (GAL1p), CRE coding sequence, and CYC1 terminator (CYC1t) from plasmid pSH63. Then the GAL1p-CRE-CYC1t was ligated with either an AT-rich or its corresponding control sequence at its upstream. The ATrich-GAL1p-CRE-CYC1t and Control-GAL1p-CRE-CYC1t cassettes were further ligated with the KanMX cassette from plasmid pFA6a-kanMX6 at its downstream. Then the KanMX-ATrich-GAL1p-CRE-CYC1t and KanMX-Control-GAL1p-CRE-CYC1t fragments were inserted into the hxt13::URA3 yeast by replacing the CIN8 gene. Finally, we obtained four yeast strains, which were labeled as AT-rich yeast (BY4742 cin8::KanMX-16p11.2_ATrich-GAL1p-CRE-CYC1t hxt13::URA3), Control yeast (BY4742 cin8::KanMX-16p11.2_Control-GAL1p-CRE-CYC1t hxt13::URA3), 22q11.2_AT-rich yeast (BY4742 cin8::KanMX-22q11.2_ATrich-GAL1p-CRE-CYC1t hxt13::URA3) and 22q11.2_Control yeast (BY4742 cin8::KanMX-22q11.2_Control-GAL1p-CRE-CYC1t hxt13::URA3), see Figures 1D and 2B; Datas S1 and S2.

Thereafter, the 3′ ends of 16p11.2_AT-rich and 16p11.2_Control sequences were linked with a GAL1p-CRE-CYC1t-LEU2 cassette, and the 16p11.2_ATrich-GAL1p-CRE-CYC1t-LEU2 and 16p11.2_Control-GAL1p-CRE-CYC1t-LEU2 cassettes were integrated into the AT-rich yeast by substituting the AVT2 gene, which located between the CAN1 and hxt13::URA3 of yeast chrV. These two new yeast strains were termed as ATrich-ATrich yeast (AT-rich yeast, avt2:16p11.2_ATrich-GAL1p-CRE-CYC1t-LEU2) and ATrich-Control yeast (AT-rich yeast, avt2:16p11.2_Control-GAL1p-CRE-CYC1t-LEU2) respectively, see Figure 3A.

The DNA damage associated genes were disrupted in the AT-rich yeasts. PCR-based methods were used for yeast gene disruption, and the LiAc transformation method was employed to transfer PCR products into yeasts.⁴⁸ All yeast strains used in this study are listed in Table S1.

Method details

Human AT-rich sequence stability analysis

Human AT-rich sequences were extracted from the human genome (hg19) using three criteria as described in the results section. The DSB peaks in the MCF7 cell line were analyzed using the bigwig data of the END_seq_MCF7_NT sample from GSE99194,²⁵ and called by the MACS3 program (parameters: --cutoff 0.5 -L 30 -g 100).⁴⁹ The DSB peaks were associated with AT-rich sequences if their positions (from start site to end site) located within the DNA range from AT-rich start position - 500 bp to AT-rich end position +500 bp. The CNV breakpoint information was extracted from the clinvarCNV data on the UCSC website. Position information of reCNVs was downloaded from the clinGen website.⁵⁰ The random sequences corresponding to each AT-rich sequences were randomly extracted from the same chromosome (non-centromere region). The CNV breakpoints located within 2500 bp upstream or downstream of the AT-rich and random sequences were labeled as AT-rich or Random sequence associated breakpoints. Genes located 5 Kbp flanking of AT-rich sequence were extracted as the AT-rich associated genes. Human genome data were visualized using IGV.⁵¹ All scripts used for human genome AT-rich analysis were listed in Data S4.

Yeast genome rearrangement rate analysis

The methods for selecting yeasts with gross genome rearrangements have been described in previous literature⁵² and are summarized in Figure 1E. Specifically, yeasts were firstly maintained on the synthetic defined (SD) URA-plate with 200 μg/mL G418 (A600958, Sangon). For mutation rate analysis, yeasts were cultured in SD URA-media with 2% Glycerol and G418, and shaking at 200 rpm for more than 3 h. Then, 10⁶ yeasts in 100 μL of water were added to 10 mL of YPGlu (YP + 2% glucose) or YPGal (YP + 2% galactose) medium and shaken for 24 h at 30°C. Then the yeasts were counted and resuspended in water to achieve a concentration of 10⁷ per 250 μL of water. Then 250 μL of yeasts (18 repeats) were spotted onto the ARG-selection plates containing 60 mg/L canavanine sulfate salt (Can, A606173, Sangon) and 0.1% 5-fluoroorotic acid (5-FOA, F832427, Macklin). After drying on a sterilized clean bench for 3–5 h at room temperature, the plates were incubated at 30°C for 8 days. Then the selected Can and 5-FOA double-resistant (RR) yeast clones were counted to evaluate the de novo gross genome rearrangement rate in yeast.

To analyze the frequency of copy number loss in ATrich-ATrich and ATrich-Control yeasts, these two strain yeasts were firstly cultured in SD LEU-/URA-/G418 medium with 2% glucose at 30°C overnight. Then 10⁶ yeasts in 100 μL of water were added to 10 mL of YPGlu or YPGal medium and cultured at 30°C for 24 h. Then, for the ATrich-ATrich yeasts, 10⁴ yeasts in 100 μL water were dropped on the SD ARG-plates or SD ARG-/LEU-/URA-plates with 60 mg/L canavanine; for the ATrich-Control yeasts, 10⁶ yeasts in 100 μL water (18 repeats) were dropped on the same selection plates as ATrich-ATrich yeasts. After drying on a sterilized clean bench for 3–5 h at room temperature, the selection plates were incubated at 30°C for 8 days and the yeast clones were counted to evaluate the copy number loss rates in yeast.

Chemical reagent treatment and yeast survival analysis

To test the homologous recombination repair efficiency of yeasts, various quantities of yeasts (approximately 10⁵, 10⁴, 10³, 10², and 10) in 10 μL of water were spotted onto the YPD plate containing 0%, 0.004%, or 0.02% (v/v) methyl methanesulfonate (MMS, 66-27-3, Acmec). To test the yeast sensitivity to the oxidative damage caused by hydrogen peroxide (H₂O₂), the yeasts were firstly treated with 0 mM or 4 mM of H₂O₂ (H792073, Macklin) at 30°C for 30 min, and spotted onto YPD plates. The yeasts treated with 0 mM of H₂O₂ were also spotted on the plates containing 20 or 30 μg/mL or Zeiocin (HY-K1053, MCE) to test the sensitivity of yeast to the Zeiocin induced DSBs. After water drying, the yeasts were cultured at 30°C for 2–3 days for the survival analysis. In this study, Camptothecin (CPT, 7689-03-4, Acmec) and Etoposide (Etop, SC0173, Beyotime) were utilized to inhibit DNA topoisomerase I (TOP1) and topoisomerase II (TOP2), respectively.

Cruciform DNA antibody-based chromatin immunoprecipitation

The chromatin immunoprecipitation (ChIP) method used in this study was based on the literature⁵³ with some modifications. In detail, AT-rich yeasts were first cultured in SD + 2% Glycerol + G418 medium at 30°C overnight. Subsequently, 10⁶ yeasts were transferred into 20 mL YPGlu medium and 20 mL YPGal medium. After 24 h of culturing, yeasts were cross-linked with formaldehyde, and then the cross-linking process was stopped by glycine. Then, yeasts were lysed by beating them three times with zirconia beads (diameter 0.25mm) in lysis buffer at 4°C for 3 min. Then the yeast lysates were sonicated twice using Qsonica (parameters: 30 s on, 30 s off, 80% amplitude, for a total of 4 min). After centrifugation, 10 μL suspensions were kept as Input and the remaining samples were incubated with Protein A/G beads (HY-K0202, MCE) that had bonded with cruciform DNA antibodies (clone ID: 2D3, GTX54648, GeneTex). After overnight incubation at 4°C, the beads were washed and resuspended in 100 μL of TE-SDS as the immunoprecipitation (IP) samples. 90 μL of TE-SDS was added to the 10 μL Input samples. Then both IP and Input samples were treated with thermo mixer (1200 rpm at 65°C, TS100, Hangzhou Ruicheng) for 1 h. Afterward, the samples were treated with proteinase K, and the DNA was purified using the Universal DNA Purification Kit (DP214, Tiangen).

Breakpoint and copy number loss identification

The breakpoints of RR yeast clones were roughly identified using PCR. Clones with a negative PCR bond indicated the loss of the corresponding DNA fragment in yeast. For yeast PCR, the AT-rich, ATrich-ATrich and ATrich-Control yeasts were first treated with 100 μL of 0.1M NaOH at 100°C for 1 h, and yeast DNA was isolated using the DNA Purification Kit with Magnetic Beads (D0041M, Beyotime). For the 22q11.2 AT-rich yeasts, the yeast walls were firstly removed by Snailase (S8280, Solarbio), and the genome DNA was isolated by FastPure Cell/Tissue DNA Isolation Mini Kit (DC102-01, Vazyme). Rapid Taq Master Mix (P222-01, Vazyme) was used for PCR amplification. DNA marker III (MD103, Tiangen) was used for the gel electrophoresis of the PCR products (the marker sizes are 200, 500, 800, 1200, 2000, 3000, and 4500 bp).

Quantitative PCR

To extract RNAs, yeasts were first suspended in an isotonic solution, and the cell wall was removed using Snailase. The RNAs were isolated using the EZ-10 DNAaway RNA Mini-Preps Kit (B618133, Sangon) and then reverse transcribed into cDNAs with the HiScript III 1st Strand cDNA Synthesis Kit (R312-01, Vazyme). The real-time quantitative PCR was used to analyze the relative CRE gene expression (URA3 gene is used as internal reference), and the ChIP-PCR was employed to assess the relative level of ATrich-formed cruciform structures. All quantitative PCR experiments were conducted on the LC480 platform using the AceQ qPCR SYBR Green Master Mix (Q111-02, Vazyme). All the primer sequences used in this study are listed in Table S2.

Pacbio HiFi sequencing

To analyze the copy number loss in ATrich-ATrich and ATrich-Control yeasts, the yeast walls were removed by Snailase and genome DNA were isolated using FastPure Cell/Tissue DNA Isolation Mini Kit. Then the regions occurred copy number loss were amplified by PCR using Rapid Taq Master Mix (primers were listed by Table S2). Then the PCR products were purified by Universal DNA Purification Kit and sequenced by Pacbio HiFi sequencing method (sequel II platform). The raw fastq data were analyzed using Flye-2.9.5⁵⁴ with the parameters: --pacbio-raw -g 20k -m 2000.

Quantification and statistical analysis

The Student’s t test was used as the default hypothesis test method to analyze the significance of the difference in RR yeast clone numbers in this study. The Wilcoxon test was used to analyze the differences in structural variant breakpoint numbers corresponding to different DNA fragments. Fishers’ exact test and Pearson’s Chi-squared test were used to analyze the differences in breakpoint positions among different RR yeast clones. For all hypothesis testing methods, p-values less than 0.01 or 0.05 were considered as significant difference. p < 0.01 was marked by ∗∗; p < 0.05 was marked by ∗; and not significant was marked by ns.

Published: November 30, 2024