Connections between sister and non-sister telomeres of segregating chromatids maintain euploidy

Summary

The complete separation of sister chromatids during anaphase is a fundamental requirement for successful mitosis. Therefore, divisions with either persistent DNA-based connections or lagging chromosome fragments threaten aneuploidy if unresolved. Here, we demonstrate the existence of an anaphase mechanism present in normally dividing cells in which pervasive connections between telomeres of segregating chromosomes aid in rescuing lagging chromosome fragments. We observe that in a large proportion of Drosophila melanogaster neuronal stem cell divisions, early anaphase sister and non-sister chromatids remain connected by thin telomeric DNA threads. Normally, these threads are resolved in mid-to-late anaphase via a spatial mechanism. However, we find that the presence of a nearby unrepaired DNA break recruits histones, BubR1 kinase, Polo Kinase, Aurora B kinase, and BAF to the telomeric thread of the broken chromosome, stabilizing it. Stabilized connections then aid lagging chromosome rescue. These results suggest a model in which pervasive anaphase telomere-telomere connections that are normally resolved quickly can instead be stabilized to retain wayward chromosome fragments. Thus, the liability of persistent anaphase inter-chromosomal connections in normal divisions may be offset by their ability to maintain euploidy in the face of chromosome damage and genome loss.

Graphical Abstract

eTOC Blurb

Warecki et al. show that prevalent DNA threads connecting telomeres of anaphase chromosomes are stabilized upon chromosome breakage and recruit proteins that retain the resulting fragment. These findings suggest that otherwise dangerous connections between anaphase chromosomes can be beneficial by preventing potential loss of genetic material.

Introduction

Successful mitosis creates two genetically identical daughter cells. This is primarily achieved at the metaphase-to-anaphase transition, when sister chromatids separate and segregate to opposing poles. Incomplete separation of chromatids can result in daughter cells with either missing or extra genetic material. For example, during the breakage-fusion-bridge cycle—when two distinct chromosomes fuse, form a bridge in anaphase, and break—daughter cells have non-identical genomes if the break occurs asymmetrically ¹. Multiple rounds of these cycles result in extensive aneuploidy characteristic of cancer ^2–6. Similarly, catastrophic aneuploidy results from cells that either mis-segregate chromosomes or fail to segregate chromosome fragments ^7–9.

To minimize such deleterious outcomes, evolved mechanisms promote entry into anaphase once separation of the duplicated genome is likely to be complete. For instance, the spindle assembly and DNA damage checkpoints prevent entry into anaphase when chromosomes are improperly attached to the mitotic spindle, incompletely replicated, or damaged ^10–12. Additionally, the ends of chromosomes are capped by telomeres, which prevent fusion between separate chromosomes and bridging during anaphase ^13,14. These mechanisms are highly conserved, underscoring their critical role in maintaining genomic integrity.

However, eukaryotic cells can still enter anaphase with lingering attachments between segregating chromatids or with unrepaired DNA breaks. Anaphase inter-chromosomal connections have been observed in yeast ^15,16, mammalian ^17–21, and insect ^22–24 divisions and may be the norm in most dividing animal cells ²⁵. For example, anaphase chromosomes can remain attached via ultrafine bridges. Ultrafine bridges are DNA strands that link separating sister chromatids and are likely present in most divisions^18,26. While centromere-based ultrafine bridges are most common, ultrafine bridges can originate from multiple locations along a chromosome, including at telomeres ²⁶. Although ultrafine bridges may contribute to sister chromatid tension and telomere maintenance ^27,28, their functional significance, if any, remains unclear. Mutants in proteins that localize to and resolve ultrafine bridges--such as BLM, PICH, Rif1, or TopBP1--cause chromosome instability, aneuploidy, cell death, and potentially cancer ^{15,17,18,29,30}. Thus, ultrafine bridges and other anaphase inter-chromosomal connections threaten the genomic integrity of dividing cells, and their resolution is required for euploidy ^{15,26,31–33}.

Additionally, entering anaphase with chromosome fragments, as observed in some cancers and cells lacking DNA damage checkpoints ^34,35, can be particularly dangerous for dividing cells. If a lagging fragment is excluded from daughter nuclei, it forms a micronucleus ⁸. Micronuclei undergo extensive DNA damage altering gene copy number and driving aneuploidy ^9,36. To maintain euploidy in response to division with broken chromosomes, some cells activate mechanisms that aid integration of chromosome fragments into daughter nuclei ^37–40. For example, in Drosophila neuroblasts, protein-coated DNA “tethers” form to specifically connect the broken ends of chromosome fragments to segregated nuclei, resulting in efficient rescue of the fragments ^23,41–43. The proteins localized to this tether—histones, BubR1, Bub3, Fizzy, Polo, Aurora B, INCENP, and BAF ^23,43,44—are recruited to the tether by recognition of the double-stranded DNA break ⁴⁵. Similar mechanisms retain broken chromosomes in Drosophila papillar cells ^35,46 and cultured human cells ^20,47. Although the proteins involved in retention differ, each mechanism depends upon recognizing DNA damage. Failure to enact these mechanisms results in micronuclei formation, aneuploidy, and cell death ^{23,35,41,43,46}.

Understanding of the mechanisms that act in anaphase to promote euploidy in the face of such division abnormalities remains limited. Here, we find evidence for a mechanism acting in normally dividing Drosophila neuroblast cells in which the first form of division errors (persistent inter-chromosomal connections) rescues the second (chromosome fragments). We observe: 1) early anaphase chromosomes separating to opposing poles remain connected by thin DNA threads predominantly originating from telomeric regions; 2) in the absence of a double-stranded DNA break, these connections are resolved in mid-to-late anaphase as the ends of chromosomes move apart; 3) the presence of a nearby double-stranded DNA break can stabilize these telomere connections; and 4) stabilized connections may aid broken chromosome segregation and incorporation into daughter nuclei. Collectively, these results suggest a novel mechanism in which telomeric connections between segregating chromosomes during early anaphase are tolerated and may allow for efficient retention of lagging chromosomes.

Results

DNA threads connect chromosomes segregating to opposing poles in early anaphase

Given the prevalence of anaphase inter-chromosomal connections in other species ^25,26, we wished to determine if similar connections linked the segregating chromosomes of Drosophila neuroblasts. Therefore, we fixed neuroblasts derived from control y1v1 3^rd instar larvae and stained DNA with the dye 4’6-diamidino-2-phenylindole (DAPI). We performed Airyscan super-resolution imaging on neuroblasts fixed in early-to-late anaphase and focused on detecting the presence of DNA threads connecting segregating sister chromatids.

Anaphase chromosome segregation appeared normal in all neuroblasts imaged, with no chromosome bridges present in any division (N=53; Figure 1, left panels). However, post-acquisition adjustment of brightness and contrast revealed the existence of DAPI-stained threads connecting chromosomes segregating to opposing poles in a substantial proportion of anaphases (Figure 1, middle and right panels, arrows). Overall, we observed inter-chromosomal threads in 38% of anaphases imaged (Figure 1A-B). Notably, the ends of chromosome arms connected by threads were directly facing one another (Figure 1A-B, arrowheads), suggesting the ends of these chromosomes may be physically connected under tension.

Figure 1. Anaphase chromosomes segregating to opposing poles remain connected by telomere-telomere DNA threads in normal divisions.

For each image panel, left=unadjusted Airyscan max projection; middle=brightness/contrast enhanced; right=zoom of yellow boxed region. (A-D) Anaphase DAPI-stained y1v1 neuroblasts showing representative examples of each thread class observed. Arrowheads indicate the connected ends. Arrows indicate threads. (A) Complete, uninterrupted (++) thread=DAPI signal with no gaps connecting chromosome ends. (B) Complete, interrupted (+) thread=DAPI signal with gaps connecting chromosome ends. (C) Stretched/erstwhile (−) thread=DAPI signal from only one chromosome end. (D) No thread (−−)=no DAPI signal. (E) Representative images of early, mid, and late anaphase neuroblasts (early=chromosome ends separated by < half the length of a chromosome arm, mid=chromosome ends separated by between half to twice the length of a chromosome arm, or late=chromosome ends separated by greater than twice the full length of a chromosome arm or chromosome arms are decondensed). (E’) Comparison of the observed frequencies of thread classes for early, mid, and late anaphase neuroblasts. (F) PI-stained anaphase y1v1 neuroblast with a complete, interrupted (+) thread (arrow) connecting chromosome ends (arrowheads). (G) DAPI-stained y1v1 anaphase neuroblast with a complete, uninterrupted (++) thread between one chromosome’s end (arrow) and near the other chromosome’s end (arrowhead). Quantification for the frequencies of thread origins at each location is also displayed. (H) Neuroblasts expressing the telomere capping protein HOAP-GFP fixed in anaphase. The origin points (arrowheads) of DAPI-stained threads (arrows) were generally from the location where HOAP was detected, although threads sometimes emerged from regions where we did not detect HOAP. (H’) Quantification of the distance of thread origins from the nearest detected HOAP signal. Each dot represents one thread origin. Scale bars are 1 μm and 0.5 μm for unzoomed and zoomed images respectively. See also Figure S1.

We next classified the observed threads into two categories based on the strength of their DAPI signals. Complete, uninterrupted (++) threads had continuous DAPI staining stretching from the end of one chromosome to the end of an opposing chromosome and were observed in 14% of anaphases (Figure 1A). Complete, interrupted (+) threads had intermittent DAPI staining stretching from the end of one chromosome to the end of an opposing chromosome and were observed in 24% of anaphases (Figure 1B).

The 62% of divisions scored as not containing threads fell into two categories. Stretched/erstwhile threads (−) showed limited evidence of possible threads, with intermittent DAPI staining emerging from the end of one chromosome towards the end of an opposing chromosome (Figure 1C). This classification could include 1) threads that had become highly stretched to a point beyond our detection, and/or 2) threads that had recently been resolved. As we cannot distinguish between these possibilities, we classified these divisions as not containing threads. However, this classification may contain threads we cannot visualize. We observed this classification in 17% of anaphases. Additionaly, we detected no evidence (−−) of inter-chromosomal threads in 45% of anaphases imaged (Figure 1D).

To determine if inter-chromosomal threads were more likely to be present early in anaphase, we binned each division based on its stage (see STAR Methods; Figure 1E) and compared the strengths of threads observed at each stage (Figure 1E’). This revealed inter-chromosomal threads were present in 78% of early anaphases (N=9), 39% of mid anaphases (N=28), and 12% of late anaphases (N=16). These differences were statistically significant (two-sided Fisher’s exact test p=0.005). Furthermore, while we observed our strongest classification of thread (++) in 67% of early anaphases, we detected ++ threads in 4% and 0% of mid and late anaphases respectively (two-sided Fisher’s exact test p<0.001). Additionally, we observed a considerable percentage (36%) of the stretched/erstwhile (−) classification of threads in mid anaphase, suggesting threads may frequently become highly stretched and/or resolved during this time. Thus, inter-chromosomal threads are common in early anaphase but are mostly resolved by late anaphase.

Given threads were visible soon after chromosome ends separated in early anaphase, we wondered if threads may exist before anaphase. We analyzed neuroblasts fixed in prometaphase and metaphase. While limited spatial separation of chromosomes during these stages prevented accurate quantification of the number of threads, we observed several pre-anaphase cells (6/14) with apparent DAPI-stained connections between separate chromosomes (Figure S1A, arrow). We observed DAPI-stained threads between chromosome arms as well as chromosome ends, including apparent connections between the ends of non-homologous chromosomes. The presence of DNA threads in early mitotic cells suggests some threads may form before anaphase.

To demonstrate the observed inter-chromosomal threads are not an artifact of DAPI staining, we fixed and stained neuroblasts with the nucleic acid dye propidium iodide (PI) (Figure 1F). Whereas DAPI binds the minor groove of DNA, PI intercalates between the bases. As PI detects both DNA and RNA, we performed an RNA digestion step before PI staining. As with DAPI-stained neuroblasts, we observed inter-chromosomal threads (arrow) connecting the ends of chromosomes segregating to opposing poles (arrowheads) in 30% (N=33) of PI-stained anaphase neuroblasts (Figure 1F). Thus, inter-chromosomal threads are DNA-based and detectable by dyes that bind DNA through different mechanisms.

Next, to examine how inter-chromosomal connections affect chromosome dynamics, we live imaged dividing y1v1 neuroblasts expressing histone H2Av-RFP, enabling chromosome visualization. In 9/30 divisions, a segregating chromosome arm seemingly moved backward towards the opposite nucleus after having already begun poleward motion (Figure S1B, arrows). The affected chromosome arm then rejoined the other chromosomes segregating poleward. These movements are consistent with a persistent physical connection between the ends of chromosomes segregating to opposing poles, possibly due to unresolved inter-chromosomal threads. Alternatively, these movements could be caused by merotelic attachments pulling the affected chromatids to both poles. We did not detect any histone signal between chromosomes segregating to opposing nuclei (0/30 divisions), suggesting anaphase inter-chromosomal threads consist of DNA but not histones.

DNA threads originate from or near capped telomeres

To determine where on chromosomes inter-chromosomal threads form, we analyzed the locations from which inter-chromosomal DNA threads originated. We mainly observed inter-chromosomal DNA threads originating from chromosome ends (65%, N=66) (Figure 1G, arrow) or near chromosome ends (within 5% of the length of the chromosome arm) (27%, N=66) (Figure 1G, arrowhead). We did not detect any threads obviously emanating from the arms of chromosomes, although 5/66 (8%) of thread ends had indeterminant origins hidden in the mass of chromatin.

To test if threads were originating from capped telomeres, we fixed neuroblasts expressing HOAP-GFP, a telomere capping protein ⁴⁸, and stained with DAPI and antibodies against GFP (Figure 1H-H’). We observed numerous threads (arrows) emerging directly from or adjacent to HOAP-capped telomeres (arrowheads indicate thread origins) (Figure 1H). Rarely, we observed threads that seemingly emerged from areas with no HOAP (Figure 1H). The average distance between the thread origin and the HOAP-capped telomere was 0.25 μm, with multiple origins closer than the z-pixel width used in our imaging (Figure 1H’). Collectively, these data suggest inter-chromosomal DNA threads are not specific to a particular locus. However, their existence during anaphase seems mainly limited to areas at or near capped telomeres.

DNA threads may depend upon the activity of the single-stranded DNA binding protein RpA-70

We reasoned DNA threads between separating sister chromatids may involve single-stranded DNA during their formation, be it DNA replication, homologous recombination, and/or DNA repair. Therefore, we analyzed anaphases in which the single-stranded DNA binding protein RpA-70 ⁴⁹ (RPA1 in humans) was reduced by expressing RNA interference (RNAi) under control of transgenic Upstream Activating Sequences (UAS).

Driving UAS-RpA-70-RNAi in neuroblasts by expressing elav-Gal4 resulted in small larval brains with few mitotic cells. The mitotic cells we could observe often underwent aberrant divisions (Figure S1C), likely due to RpA-70’s role in multiple essential processes ⁵⁰. In the rare relatively normal mitoses, cells were fixed in late anaphase with no detectable threads (Figure S1C). The lack of mitotic cells upon strong RpA-70 reduction precluded in depth analysis of RpA-70 involvement in thread formation. However, larvae lacking Gal4 still exhibited small brains at low frequency, suggesting “leaky” RpA-70-RNAi expression in the absence of Gal4. These brains contained more mitotic cells. Therefore, we analyzed mitotic neuroblasts expressing low levels of UAS-RpA-70-RNAi in the absence of Gal4.

In total, we observed threads in 49% (N=75) of control and 37% (N=73) of UAS-RpA-70-RNAi neuroblasts fixed concurrently (Figure S1D). As threads are largely undetectable by late anaphase (Figure 1E’), we focused on neuroblasts fixed in early and mid anaphase (Figure S1E). We observed a moderate, significant reduction in the percentage of threads in early and mid anaphase UAS-RpA-70-RNAi neuroblasts (y1v1=69%, N=49; UAS-RpA-70-RNAi=48%, N=52; two-sided Fisher’s exact test p=0.04). Additionally, the number of threads per division was slightly but significantly reduced in UAS-RpA-70-RNAi neuroblasts (y1v1=1.2 threads/division; UAS-RpA-70-RNAi=0.79 threads/division; Mann-Whitney test p=0.038) (Figure S1F), suggesting threads may depend upon RpA-70 activity. However, further research is required to test if RpA-70 indeed contributes to anaphase DNA thread formation and/or stability.

Inter-chromosomal DNA threads are mainly resolved by a spatial mechanism

Despite the presence of inter-chromosomal DNA threads in a large proportion of early divisions, y1v1 flies are healthy, suggesting the observed DNA threads are efficiently resolved during mitosis. The dramatic reduction in detectable DNA threads in late anaphase (Figure 1E) is consistent with resolution of anaphase DNA threads before mitotic exit.

Inter-chromosomal DNA threads could be resolved by a spatial or temporal mechanism. To distinguish between these possibilities, we fixed and DAPI-stained neuroblasts containing extra-long chromosome arms with ends that would be near to one another even during mid-to-late anaphase. We imaged neuroblasts containing either the In(3LR) chromosome, in which an inversion of the 3^rd chromosome results in an elongated chromosome arm ⁵¹, or the C(2)EN chromosome, in which both homologs of the 2^nd chromosome are on a single centromere ⁵² (Figure 2A). As with normal-length chromosomes from previously imaged y1v1 neuroblasts, we observed threads connecting the telomeres of extra-long sister chromatids segregating to opposing poles in both In(3LR) and C(2)EN-bearing neuroblasts (Figure 2B; y1v1=38%, N=53; In(3LR)=55%, N=49; C(2)EN=34%, N=47).

Figure 2. Inter-chromosomal DNA threads are primarily resolved by a spatial mechanism.

For each image panel, left=unadjusted Airyscan max projection; middle=brightness/contrast enhanced; right=zoom of yellow boxed region. (A) The extra-long C(2)EN chromosome is made by combining both homologs of the 2^nd chromosome onto one centromere. The extra-long In(3LR) chromosome is made by an inversion that places the centromere closer to one telomere, resulting in an elongated chromosome arm. (B) Anaphase DAPI-stained y1v1 (top), In(3LR)-bearing (middle), and C(2)EN-bearing (bottom) neuroblasts showing representative examples of complete, uninterrupted (++) threads (arrows). Scale bars are 1 μm and 0.5 μm for unzoomed and zoomed images respectively. (C) In mid-to-late anaphase, a temporal resolution model predicts no threads would be observed in divisions with either normal-length or extra-long chromosomes. In contrast, a spatial resolution model predicts threads would be observed in mid-to-late anaphase divisions with extra-long chromosomes, because the distance between the extra-long chromatid ends (X) is comparable to the distance between the ends of normal-length chromatids earlier in anaphase. (D) Comparison of the observed frequencies of thread classes for y1v1, In(3LR)-bearing, and C(2)EN-bearing neuroblasts fixed in mid and late anaphase. (E) Comparison of the measured lengths of the observed threads for y1v1, In(3LR)-bearing, and C(2)EN-bearing neuroblasts. Each dot (magenta=++; pink=+) represents one thread color-coded for thread class. There is no significant difference in the measured lengths of threads between the control and extra-long chromosomes. (E’) Density distributions of the combined measured thread lengths from (E) for ++ (top) and + (bottom) threads, suggesting threads become stretched as chromosome ends move apart. See also Figure S2.

With a temporal mechanism of thread resolution, DNA threads should be present only in early anaphase, regardless of the cell containing extra-long chromosomes (Figure 2C). In contrast, a spatial mechanism of resolution predicts DNA threads should also be present in mid-to-late anaphase for cells bearing extra-long chromosomes. This is because the distance between the ends of extra-long chromatids later in anaphase is comparable to the distance between the ends of normal-length chromatids earlier in anaphase (Figure 2C).

Therefore, we compared the percentages of threads in mid and late anaphase between neuroblasts with extra-long or normal-length chromosomes. Consistent with a spatial mechanism of resolution, we detected inter-chromosomal threads in mid anaphase neuroblasts containing either extra-long In(3LR) or C(2)EN chromosomes at significantly higher rates than in mid anaphase y1v1 neuroblasts with normal-length chromosomes (y1v1=39%, N=28; In(3LR)=72%, N=32; C(2)EN=50%, N=22) (Figure 2D; two-sided Fisher’s exact test p=0.04). Similarly, the frequencies of the strongest classification of threads (++) were significantly increased when mid anaphase neuroblasts contained extra-long chromosomes (y1v1=4%, N=28; In(3LR)=41%, N=32; C(2)EN=32%, N=22; two-sided Fisher’s exact test p=0.002). By late anaphase, the difference in frequencies of threads was no longer statistically significant (two-sided Fisher’s exact test p=0.77), although the frequencies of threads in In(3LR)- and (C(2)EN-bearing neuroblasts were still increased compared to y1v1 controls (y1v1=12%, N=16; In(3LR)=24%, N=17; C(2)EN=20%, N=25). The persistence of threads in neuroblasts containing extra-long chromosomes indicates inter-chromosomal threads are mainly resolved by a spatial, not temporal, mechanism.

Consistent with spatial resolution, we observed similar distances between the ends of connected chromosomes in y1v1, In(3LR)-, and C(2)EN-bearing neuroblasts (Figure 2E). This suggests the main determinant for the presence of a thread is the physical distance between chromosome ends. However, while these distributions were not statistically significant (Kruskal-Wallis test p=0.23), we were unable to detect >2 μm threads in C(2)EN-bearing neuroblasts despite detecting them in y1v1 control and In(3LR)-bearing neuroblasts (Figure 2E). Anaphase C(2)EN chromosome arms are slightly more elongated than In(3LR) chromosome arms (Figure S2A-B, ⁵¹). Therefore, there may be an upper temporal limit on thread persistence, perhaps due to cytokinesis.

Nevertheless, for all genotypes combined, we found the stronger classification of complete, uninterrupted (++) threads were generally observed when the distances between connected chromosome ends were <1 μm (Figure 2E’). The weaker classification (+) was mainly observed when the distances between connected chromosome ends were >1 μm (Figure 2E’). These results suggest inter-chromosomal DNA threads stretch and then resolve as chromosome ends move apart.

DNA threads between the ends of extra-long chromosomes may recruit histones

Our inability to detect histones between the ends of segregating normal-length chromosomes (Figure S1B) may be because DNA threads are highly stretched. As the ends of extra-long chromosomes remain near one another during early and mid anaphase, threads connecting extra-long chromosomes may be under less tension and more suitable for histone recruitment. Therefore, we live imaged neuroblasts bearing the extra-long In(3LR) chromosome and expressing H2Av-RFP.

While we were unable to detect histones linking the ends of separating extra-long chromosomes in most divisions (Figure S2C-C’), in rare cases, we observed faint H2Av-RFP signal connecting the ends of extra-long chromosomes that had just separated (Figure S2D-E’, arrows). These results suggest DNA threads under less tension can recruit histones, albeit infrequently. We observed histone-coated threads in 3/28 divisions imaged (Figure S2F).

Separated extra-long sister chromatids exhibit coordinated movements

Consistent with a physical connection linking sister chromatids, the ends of extra-long chromosomes appeared to exhibit coordinated movements as if one end physically pulled on the other (Figure S2C timepoints 40–300; S2D timepoints 320–480; S2E timepoints 20–350). We observed coordinated movements in 10/28 divisions (Figure S2F), including in all divisions in which we could detect histone-coated threads. Chromosomes exhibiting coordinated movement should have similar angles between the chromosome ends and the division axis (Figure S2G). We measured the angles between the long chromosome ends and the division axes for 200 seconds following chromosome end separation (Figure S2G’). For a division exhibiting uncoordinated movement (Figure S2G’, dashed lines), the measured angles behaved independently. However, for a division exhibiting coordinated movement between chromosome ends (Figure S2G’, solid lines), the measured angles tracked one another. These coordinated movements suggest DNA threads may transmit force between chromosomes separating to opposing poles.

DNA threads connect the telomeres of segregated chromosomes to sister and non-sister telomeres of lagging chromosomes and chromosome fragments

Given the spatial resolution of threads, we wondered if lagging chromosomes and chromosome fragments could remain connected to nearby segregated chromosomes. Similar to DNA-protein tethers that link the broken ends of chromosome fragments ^23,47, persistent telomere-telomere connections may secure lagging chromosomes to segregated chromosomes. Therefore, we generated lagging chromosomes in neuroblasts by creating acentric chromosomes with intact telomeres and no centromeres and asked if we could detect threads connecting these lagging acentrics to segregated centric chromosomes.

To generate acentric lagging whole chromosomes, we expressed the recombinase FLP in neuroblasts containing transgenic chromosomes with inverted FRTs ¹⁴. In this system, FLP-mediated recombination of the inverted FRTs results in a dicentric chromosome and an unbroken acentric, analogous to a whole chromosome that failed to attach to the spindle (Figure 3A). Threads connected the unbroken acentrics to the centric chromosomes in 34% of divisions imaged (N=44), 18% in total from the strongest classification of thread (++; complete, uninterrupted) (Figure 3B, arrows; Figure 3F). Thus, persistent DNA threads link the telomeres of lagging chromosomes to the telomeres of segregated chromosomes. Additionally, because generation of the unbroken acentric results in sister telomeres being rearranged onto the same DNA molecule (Figure 3A), the observed threads connecting unbroken acentrics to segregated chromosomes demonstrate that DNA threads connect non-sister as well as sister telomeres.

Figure 3. Lagging broken and unbroken acentrics are connected to segregating chromosomes by DNA threads.

For each image panel, left=unadjusted Airyscan max projection; middle=brightness/contrast enhanced; right=zoom of yellow boxed region. (A) FLP-mediated recombination between sister chromatids bearing inverted FRTs (arrowheads) results in a dicentric chromosome and an unbroken acentric with intact telomeres. a and a’ denote sister telomeres of the short arm. b and b’ denote sister telomeres of the long arms. Sister telomeres are recombined onto the same DNA molecule. During anaphase, the acentric chromosome lags on the metaphase plate. (B) Two DAPI-stained FLP-expressing neuroblasts showing DNA threads (arrows) connecting unbroken acentrics (asterisks) to segregated chromosomes. For clarity, a single z-slice is shown for the neuroblast on the left. (C) I-CreI endonuclease recognizes its cut site (square) at the base of the X and Y chromosomes, resulting in an acentric fragment with a broken end and an intact telomere. During anaphase, the acentric chromosome fragment lags on the metaphase plate. (D-E) Anaphase DAPI-stained I-CreI-expressing neuroblasts showing DNA threads connected to broken acentrics (labeled 1–4). Threads connect broken acentrics to segregating chromosomes (arrows) and to other acentrics (arrowhead). Some acentrics have threads emanating from both their ends (E). Scale bars are 1 μm and 0.5 μm for unzoomed and zoomed images respectively. (F) Comparison of the observed frequencies of thread classes connecting unbroken and broken acentrics. (G) Comparison of the normalized thread pixel intensity for threads connecting unbroken and broken acentrics. Each dot represents one thread. See also Figure S3.

We next asked if telomere-telomere threads persist on lagging chromosome fragments resulting from chromosome breakage. We generated broken acentric fragments by expressing the endonuclease I-CreI (Figure 3C), which cuts at the base of the sex chromosomes ⁵³ to create 4 distinct fragments (labeled 1–4 in Figure 3D-E). We observed DNA threads connecting acentric fragments in 87% (N=54) of analyzed neuroblasts, 70% from the strongest classification of thread (++, complete, uninterrupted) (Figure 3D, arrow; Figure 3F). Some observed threads correspond to the DNA component of protein-coated “tethers” that link the broken end of the acentric to the broken end of the centric fragment ²³ (from here on, tethers=DNA threads + “tether” proteins). However, we also observed divisions in which DNA threads emanated from both ends of broken acentrics simultaneously (Figure 3E). In these cases, one thread connected the acentric to the nearby segregated chromosomes (Figure 3E, arrow) while the second thread connected the other end of the acentric to either another acentric (Figure 3E, arrowhead) or chromosomes segregated to the opposite pole. Because one end of the broken acentric is the telomere end, DNA threads emanating from both ends of acentrics demonstrate telomere-telomere connections persist after a chromosome is broken. Strong DNA threads between acentrics segregating to opposing nuclei in Drosophila neuroblasts were reminiscent of inter-chromosomal DNA connections we could detect between lagging chromosomes in human cancer cells (Figure S3A-B).

These results indicate lagging chromosomes remain connected to segregated chromosomes via telomere-telomere DNA connections. Additionally, the significantly increased frequency (unbroken acentrics=34%; broken acentrics=87%; two-sided Fisher’s exact test p<0.001) and strength (unbroken acentrics=18% ++ threads; broken acentrics=70% ++ threads; two-sided Fisher’s exact test p<0.001) of threads connected to broken acentrics compared to those connected to unbroken acentrics (Figure 3F) suggests chromosome breakage is correlated with more robust threads. Consistent with this view, threads connecting to broken acentrics were quantifiably brighter than threads connecting unbroken acentrics (Figure 3G; Mann-Whtiney test p<0.001). Increased thread frequency and strength are likely due to the formation of protein-coated tethers upon recognition of the broken end ^23,45.

Histones are recruited to DNA threads emanating from sister and non-sister telomeres of broken acentrics

Given our observation of frequent DNA threads emanating from both ends of broken acentrics, we wondered if telomere-telomere DNA threads present in normal divisions could be stabilized into protein-coated “tethers” when a chromosome is broken. Support for this idea comes from 1) the high frequency of protein-coated tether formation between acentric fragments and segregated chromosomes ²³, and 2) the surprisingly high frequency (~40%) that the telomere end of the broken acentric is positioned toward the segregated chromosomes ⁵⁴. These studies, and our observation of telomere threads, suggest protein-coated tethers may link to the telomere end of broken acentrics if that end is closest to the segregated chromosomes.

To test if telomere-telomere DNA threads are stabilized into protein-coated tethers upon chromosome breakage, we live imaged dividing neuroblasts co-expressing I-CreI, H2Av-RFP, and the telomere protein HOAP-GFP. As histones are one of the proteins recruited to tethers ²³, expression of H2Av-RFP allowed visualization of both chromosomes and tethers. We found histones were recruited to tethers extending from the broken ends of acentrics (Figure 4A, magenta arrow; Video S1) and also from the HOAP-GFP-marked telomere ends of acentrics (Figure 4B, magenta arrowhead; Video S2). We observed histone-coated tethers between the telomeres of acentric fragments and daughter nuclei that connected non-sister telomeres, as the sister acentric fragments were either segregating to the opposing nucleus or co-segregating with the tethered acentric and thus not in the daughter nucleus (Figure 4B). Acentric fragments connected by their telomeres to segregated chromosomes could still move poleward. Additionally, we sometimes observed histone-coated tethers extending off both broken and telomere ends of acentrics simultaneously (Figure 4C), sometimes connecting sister acentrics. We observed H2Av-RFP tethers from the broken ends of 51/90 acentrics (57%), from the telomere ends of 28/90 acentrics (31%), and from both ends of 15/90 acentrics (17%) (Figure 4D). 36/90 (40%) and 13/90 (14%) of acentrics had detectable H2Av-RFP tethers off only their broken or telomere end, respectively (Figure 4D). These results indicate sister and non-sister telomere DNA threads can recruit histones following chromosome breakage.

Figure 4. Histones are recruited to DNA threads emanating from the telomeres of broken acentrics.

(A-C) Max projection stills from time lapse movies of neuroblasts expressing I-CreI, the chromosomal marker H2Av-RFP (magenta), and the telomere marker HOAP-GFP (green). (A) Histones are recruited to tethers (magenta arrows) that connect the broken ends (white arrows) of acentrics to segregating chromosomes ²³. See also Video S1. (B) Upon chromosome breakage, histones can also be recruited to the DNA threads (magenta arrowheads) emanating from the telomere ends (asterisks) of the acentrics. The histone-coated DNA thread in this cell connects non-sister telomeres. See also Video S2. (C) Some acentrics had histone-coated tethers originating from both their ends simultaneously. (D) Quantification of the frequencies of H2Av-RFP tethers emanating from broken, telomere, or both ends of individual acentrics. For each image panel, zoomed images of the merged and H2Av-RFP channels are shown for the yellow boxed region. Time is seconds before/after acentric segregation. Scale bars are 2 μm and 1 μm for unzoomed and zoomed images respectively.

Telomere DNA threads recruit tether proteins upon chromosome breakage

Histone-coated telomere threads connecting broken acentrics may serve as a platform upon which “tether” proteins localize. Tethers connecting the broken ends of acentric and centric fragments recruit a set of proteins: BubR1, Bub3, Fizzy, Polo, Aurora B, INCENP, and BAF ^23,43,44. These proteins aid in the segregation and incorporation of acentric fragments into daughter nuclei ^23,41–44. As acentric fragments segregated to the nuclei to which they were connected by their telomeres (Figure 4B), we wondered if the tether proteins--specifically BubR1, Polo, Aurora B, and BAF--localized to the telomere threads of broken acentrics. We used single acentrics forming tethers from both their broken and telomere ends (Figure 4C) as the readout for tether protein recruitment to telomere threads. As I-CreI expression generates 4 distinct acentric fragments, we first identified the 4 acentrics, then tracked individual acentrics to assay tether formation from both ends (Figure 5A, labeled 1–4 Figure 5B-E).

Figure 5. Telomere DNA threads recruit the tether proteins BubR1, Polo, Aurora B, and BAF upon chromosome breakage.

(A) Expression of I-CreI in female larvae results in four broken X chromosome fragments. Single acentrics can be individually tracked through anaphase and assayed for the formation of protein-coated tethers simultaneously from both of their ends (broken and telomeric). (B-E) Max projection stills from time lapse movies of neuroblasts expressing I-CreI, H2Av-RFP (magenta), and either BubR1-EGFP (B, green; See also Video S3), GFP-Polo (C, green; See also Video S4), GFP-Aurora B (D, green; See also Video S5), or GFP-BAF (E, green; See also Video S6). BubR1, Polo, Aurora B, and BAF tethers can be observed connecting lagging acentrics to both segregating chromosomes (arrows) and to other acentrics (arrowheads). Each of the 4 individual acentrics in a division is marked by a number. Recruitment of these proteins to tethers off both ends of a single acentric simultaneously indicates their recruitment to the DNA thread emanating from the telomere end of the acentric. Time is seconds after acentric segregation. (F-F’) Max projection stills from time-lapse movies of neuroblasts expressing H2Av-RFP (magenta) and BubR1-EGFP (green). Neither BubR1 nor histones are detected coating telomere-telomere DNA threads in divisions without chromosome breaks, even in divisions when a chromosome is slightly lagging and its tip is directly opposed to its sister’s tip (F’, arrows). See also Video S7. Time is seconds before/after acentric segregation. Scale bars are 2 μm and 1 μm for unzoomed and zoomed images respectively. (G) Comparison of the frequency of observed BubR1-coated tethers between divisions in which I-CreI-induced broken chromosomes are present and normal divisions with unbroken chromosomes. For divisions with broken chromosomes, the overall frequency of BubR1-coated tethers and the frequency of acentrics with BubR1-coated tethers emanating from both ends is also compared. See also Figures S3 and S4.

To determine if tether proteins are recruited to threads extending from the telomeres of broken acentrics, we live imaged dividing neuroblasts co-expressing I-CreI, H2Av-RFP, and BubR1-EGFP. Consistent with previous reports ²³, BubR1-EGFP formed tethers connecting broken acentrics to daughter nuclei (Figure 5B, arrows; Video S3). However, we also observed BubR1-EGFP form tethers between acentrics segregating to opposing poles (Figure 5B, arrowhead). Importantly, we observed instances of single acentrics with BubR1-EGFP tethers extending from both their ends (Figure 5B, zoom). We observed similar results with GFP-Polo (Figure 5C; Video S4), GFP-Aurora B (Figure 5D; Video S5), and GFP-BAF (Figure 5E; Video S6). Therefore, telomere DNA threads recruit BubR1, Polo, Aurora B, and BAF upon chromosome breakage.

To determine if the recruitment of tether proteins to telomere threads only occurs following chromosome breakage, we live imaged neuroblasts with unbroken chromosomes co-expressing H2Av-RFP and BubR1-EGFP. We were unable to detect either BubR1-EGFP or H2Av-RFP localization between segregating chromosomes (Figure 5F), consistent with previous reports ²³. Even in instances when chromosomes were slightly lagging with their ends directly opposed to one another, conditions in which we readily observed DNA threads by DAPI staining, we were unable to detect BubR1-EGFP tethers or BubR1-EGFP on telomeres (Figure 5F’; Video S7). Collectively, we detected BubR1-EGFP tethers in 16/16 cells dividing with broken acentrics, and 42/58 (72%) of acentrics in those cells contained at least one BubR1-EGFP tether (Figure 5G). Of those 42 acentrics with at least one BubR1-EGFP tether, we detected BubR1-EGFP emanating from the other end for 9 acentrics (21%). 0/24 divisions with unbroken chromosomes contained BubR1-EGFP tethers (Figure 5G). These results suggest tether proteins are recruited to telomere threads upon chromosome breakage.

Tether proteins are recruited before anaphase

We next analyzed the timing with which tether proteins localized to the telomere threads of broken acentrics. As previously reported ^23,45, BubR1 (Figure S3C, green arrowheads), Polo, and Aurora B all associated with acentrics (Figure S3C, magenta arrows) before anaphase (Figure S3D, grey dots, mean times=144 (BubR1), 116 (Polo), and 186 (Aurora B) before anaphase onset). These pools became tethers (Figure S3C, green arrows) soon after anaphase entry (Figure S3D, green dots, mean=101 (BubR1), 82 (Polo), and 54 (Aurora B) sec after anaphase onset). In contrast, GFP-BAF localized to acentrics (Figure 3E, asterisks, mean=244 sec after anaphase onset) then tethers (Figure S3E, green arrows, mean=276 sec after anaphase onset) well after anaphase onset (Figure S3D-E). This delay in BAF localization is because VRK1 phosphorylation disrupts the interaction between BAF and DNA in early mitosis ^55,56. While we could not discern if the observed BubR1, Polo, and Aurora B tethers originated from the broken or telomere ends of acentrics, we regularly observed BAF simultaneously localize to tethers off both ends of single acentrics (Figure S3E, green arrows and arrowheads). Additionally, we did not observe multiple patterns in the timing with which BubR1, Aurora B, and Polo formed visible tethers (Figure S3D) despite acentrics segregating with their telomere end first ~40% of the time ⁵⁴. Instead, tethers consistently became visible in the first timepoint when acentrics were spatially distinct from segregating chromosome arms. These results suggest tether proteins localize to the broken and telomere ends of acentrics with similar dynamics. Therefore, the recruitment of tether proteins to telomere threads likely occurs before anaphase.

Broken chromosomes result in persistence and stabilization of telomere DNA threads

To quantify how chromosome breakage affects resolution of telomere threads, we compared the timing of thread resolution for 700 seconds after anaphase onset between divisions with and without acentrics. Because BAF, which is enriched on acentrics and tethers ⁴³, also localizes generally to DNA in late anaphase ⁵⁷, we analyzed GFP-BAF localization to telomere DNA threads in divisions with unbroken chromosomes. No detection of BAF on threads in normal divisions would suggest threads had been resolved before BAF localizes to DNA at the end of mitosis. Consistent with thread resolution occurring in mid-to-late anaphase, we never observed GFP-BAF threads connecting BAF-coated daughter nuclei in divisions with unbroken chromosomes (Figure S3F-G). In contrast, in divisions with broken acentrics, we observed GFP-BAF tethers emanating from both ends of single acentrics that persisted for up to 700 sec after anaphase onset (Figure S3G). Thus, chromosome breakage dramatically increases the persistence of telomere threads.

Broken end tether integrity requires BubR1 activity ^23,44. To determine if global BubR1 activity affects the telomere threads of broken acentrics, we monitored threads in neuroblasts co-expressing I-CreI and a truncated, dominant negative version of BubR1 (UAS-BubR1-DN). In UAS-BubR1-DN neuroblasts, threads connected acentrics in 16/19 of divisions imaged (Figure S3H). 5/19 divisions had threads emanating from both ends of individual acentrics (Figure S3H, arrow and arrowhead), suggesting telomere threads may not require global BubR1 activity to form. However, BubR1 interacts with Bub3 to localize to the broken end tether ⁴⁴. The truncated version of BubR1 lacks its full Bub3 binding domain and so may not robustly affect tether integrity. Therefore, future research should determine how BubR1 and other broken end tether proteins contribute to the persistence of telomere threads after chromosome breakage. Nevertheless, these results, and our observations of threads originating from the telomeres of undamaged chromosomes lacking detectable BubR1 (Figures 1, 5), are consistent with a model in which telomere threads on acentrics precede protein-coated tethers and do not rely on the activity of tether proteins for their initial formation. Collectively, these results suggest normally resolved telomere threads may be stabilized into functional tethers upon chromosome breakage.

Telomere DNA threads from nearby chromosomes may also be stabilized upon chromosome breakage

Recruitment of tether proteins to telomere DNA threads upon chromosome breakage indicates tether formation is not limited to broken ends. However, in divisions with broken chromosomes, we did not observe any tether proteins localize to the DNA threads emanating from the telomeres of segregating unbroken chromosomes (Figure 5B-E). This suggests thread-to-tether stabilization does not occur globally following chromosome breakage but instead is limited to nearby chromosome ends. Recognition of a double-stranded DNA break promotes broken end tether formation ⁴⁵. Therefore, recruitment of tether proteins to the DNA threads on the telomeres of broken acentric fragments represents cis-stabilization.

To evaluate trans-stabilization of threads from the telomeres of nearby unbroken chromosomes, we expressed I-CreI in cells bearing the extra-long In(3LR) chromosome. Unlike the ends of normal-length chromosomes, the ends of extra-long In(3LR) chromatids are close to the broken ends of acentric fragments (Figure S4A). Should trans-stabilization occur, normally-resolved threads between the telomeres of the extra-long chromatids may persist, potentially causing bridging, aneuploidy, cell cycle arrest, or death (Figure S4A). We found larvae expressing I-CreI or bearing the In(3LR) extra-long chromosome alone both developed into flies with normal tissues. In contrast, In(3LR) larvae expressing I-CreI developed into flies with nicked wings (Figure S4B, arrows) at a small, significantly increased frequency (I-CreI=3%, N=136; In(3LR)=0%, N=46; I-CreI+In(3LR)=11%, N=176; two-sided Fisher’s exact test p=0.003). Wing nicks result from defective proliferation during development ⁵⁸. Therefore, I-CreI+In(3LR) cells may arrest or die, potentially from aneuploidy after stabilization of telomere-telomere threads between the extra-long chromatids (Figure S4C). These results suggest trans-stabilization of telomere threads may occur rarely. Additionally, the low rate of wing defects in adults developed from larvae with either extra-long chromosomes (threads resolved) or broken acentrics (threads stabilized) illustrates how selective stabilization of telomere threads efficiently maintains genomic integrity.

Discussion

Successful mitosis requires accurate segregation of sister chromatids. Improper segregation leads to aneuploidy ³². Traditionally, the metaphase-to-anaphase transition is considered the point at which duplicated sister chromatids separate. However, our observation that separating chromosomes remain connected by telomere-telomere DNA threads in most early anaphase Drosophila neuroblasts (Figure 1) supports a model that genome separation is largely incomplete until late anaphase ^25,26. Anaphase inter-chromosomal connections often originate at centromeres and telomeres and have been observed in diverse species ^{17,18,21,22,59}. If unresolved, these connections threaten the genomic integrity of dividing cells ^{15,26,31–33}. The functional significance of anaphase inter-chromosomal connections is poorly understood.

Here, we propose a model in which pervasive DNA threads between sister and non-sister telomeres can be co-opted to retain lagging chromosomes that would otherwise result in aneuploidy (Figure 6A). Our results indicate inter-chromosomal DNA threads are resolved in mid-to-late anaphase of normal divisions (Figure 6B) but are stabilized into protein-coated tethers if they are near a double-stranded DNA break--such as on a lagging broken fragment. Stabilized tethers may then mediate efficient rescue of the lagging chromosome (Figure 6C). The retention of a potentially lost chromosome may partially balance the threat of aneuploidy these anaphase inter-chromosomal connections pose.

Figure 6. General telomere inter-chromosomal connections capture and retain lagging chromosomes to promote euploidy.

(A) Summary of when threads (DNA + histones) and tethers (threads + proteins) connect anaphase chromosomes. Tethers/threads we directly observed are boxed. Question marks indicate histones were rarely detected. (B) In normal divisions, DNA threads connect the telomeres of most segregating sister chromatids. As the cell progresses through anaphase, threads stretch and are resolved, leading to complete genome separation and euploidy. (C) In divisions with broken chromosomes, broken ends (yellow-blue stars) produce a signal so that nearby telomere threads are coated with proteins (histones, BubR1, Polo, Aurora B, BAF) and stabilized to form tethers. This results in tethers from both the broken and telomere ends of acentrics (potentially connecting both sister and non-sister telomeres). Tethers persist in anaphase when threads are resolved and mediate segregation to the daughter nucleus. Thus, euploidy is efficiently maintained.

Formation and resolution of inter-chromosomal DNA threads

We observed DNA threads emanating from or near the capped telomeres of anaphase chromosomes (Figure 1). While this suggests telomeres may be conducive to threads, further research is required to determine if threads are composed of telomere, telomere associated, and/or other DNA sequences. Threads may be due to the unique nature of Drosophila telomeres. Drosophila lack telomerase and maintain their telomeres through retrotransposon activity and homologous recombination ⁶⁰. Drosophila telomeres can associate with one another, sometimes forming “ectopic fibers” in polytene chromosome preparations ^61–63. Additionally, strong associations between chromosome ends in Drosophila strains with long telomeres persist into anaphase before resolving ⁵⁹. However, while long telomere associations form pronounced chromosome bridges in anaphase ⁵⁹, we do not observe any chromosome bridging in our fixed divisions. Instead, linked telomeres are separated by large distances (Figure 2). Furthermore, the DNA telomere-telomere threads we observe lack detectable histones (Figures S1, 5), suggesting threads may be highly stretched. Consistent with this view, threads between extra-long chromosomes, which are presumably under less tension, can occasionally recruit histones (Figure S1). Thus, telomeric DNA threads may be more similar to ultrafine bridges than to traditional chromosome bridges.

Ultrafine bridges commonly originate from centromeres but can also form at telomeres ^26,64,65. Telomere-based ultrafine bridges may contribute to telomere maintenance through “telomere sharing” between telomere-free and telomere-containing chromosome ends ²⁷, which might be more common in telomerase-lacking Drosophila. One cause of ultrafine bridges is replication-induced catenation between sister chromatids ^17,18,66. Our observation of DNA threads between the telomeres of extra-long sister chromatids (Figure 2) is consistent with threads resulting from replication-induced catenation. However, creation of unbroken acentrics by FLP-mediated recombination places both sister telomeres on the same unbroken acentric (Figure 3). Therefore, threads connecting the intact, non-sister telomeres of unbroken acentrics and the telomeres of segregated chromosomes (Figure 3) suggests other mechanisms besides catenation contribute to the formation of these threads. Other potential mechanisms include homologous recombination or late replication intermediates ^29,67.

Regardless of how threads form, normally dividing cells must resolve the threads to preserve euploidy. We believe that before resolution, threads become so stretched they become undetectable. It is likely highly stretched, but invisible, threads are included in our (−) classification. We expect stretched threads would be susceptible to nucleases, such as Ankle1 ⁶⁸, or factors that resolve bridges, such as BLM helicase ¹⁸. Thus, recently resolved threads likely contribute to the (−) classification as well. Since we cannot differentiate between highly stretched, invisible threads and recently resolved threads, we consider the (−) classification to span the period of resolution. Thread resolution occurs in mid-to-late anaphase (Figure 1) primarily through a spatial mechanism (Figure 2). The spatial resolution mechanism likely accounts for the persistence of DNA threads connecting segregated chromosomes to closely associated lagging chromosomes (Figure 3).

Stabilization of telomere DNA threads into protein-coated tethers

Upon chromosome breakage, sister and non-sister telomere DNA threads recruit a set of tether proteins: histones, BubR1, Polo, Aurora B, and BAF (Figures 4–5). Previous research has demonstrated these proteins play important roles in the segregation and incorporation of broken chromosomes into daughter nuclei ^23,41,42,44. Considering 1) telomere threads can ostensibly transmit force between unbroken chromosome ends (Figure S2) and 2) threads on broken acentrics persist through late anaphase well after they would have normally been resolved (Figures 3 and S3), chromosome breakage may preserve threads and provide a platform for the recruitment of tether proteins that subsequently retain the broken chromosome fragment. Indeed, acentrics often segregate toward and reintegrate into daughter nuclei with their telomere ends leading ⁵⁴. It seems plausible the integration of these telomere-leading acentrics could be mediated by telomere tethers. Thus, telomere DNA threads that have been stabilized into tethers could aid in persevering genome integrity.

We do not know how tether proteins are recruited to telomere DNA threads upon chromosome breakage. One possibility is the telomere end is recognized as a DNA break. Recognizing a double-stranded DNA break is key for the recruitment of proteins to the broken end of chromosome fragments ⁴⁵. However, HOAP-GFP localization on the natural ends of broken acentrics suggests the acentric telomeres are intact and capped (Figure 4) and should not be recognized as broken ends ⁶⁹. An alternate possibility is the initial recruitment of proteins to the broken ends results in “leaky” recruitment of proteins to the telomere DNA threads, either by 1) direct recruitment to any chromosome thread near a recognized broken end or 2) recruitment first to the broken end then relocation to the telomere thread.

Like their recruitment to the broken end ^23,45, tether protein recruitment to telomere end threads probably initiates before anaphase (Figure S3). Additionally, tether proteins localize to damage induced specifically during anaphase ⁴⁵, indicating recruitment continues throughout anaphase. Accordingly, BAF localizes to both the broken end and telomere tethers simultaneously during late anaphase (Figure S3). Sustained recruitment of tether proteins to the telomere end before and during anaphase could contribute to the increased persistence of telomere threads following chromosome breakage.

General telomere inter-chromosomal DNA threads may lead to efficient recovery of broken chromosome fragments

Retention of genetic material that would otherwise be lost in division is not a unique feature of Drosophila (for review, see ³⁷). Retention can occur through multiple mechanisms. For example, proteinaceous filaments retain chromosome fragments in human cells ⁴⁷. Additionally, double minute chromosomes are transported by attaching to the arms of segregating centric chromosomes ^70,71. Furthermore, lateral microtubule connections also mediate segregation of acentric chromosomes in insects and trypanosomes ^54,72,73. Efficiency of these mechanisms vary. Overall, many chromosome fragments are not retained and instead form damage-prone micronuclei ⁸. However, chromosome fragment retention in Drosophila occurs with remarkable fidelity ^23,35.

Selective stabilization of prevalent telomere DNA threads into protein-coated tethers offers a potential explanation for the high efficacy of broken chromosome retention in Drosophila. In addition to lateral connections with microtubules ⁵⁴, tether proteins are important for both the timely poleward segregation of acentrics ²³ and their integration into daughter nuclei ^41–43. Thus, the ability to form tethers off either the broken or telomere end of an acentric may enhance the ability to retain that fragment (Figure 6C). Additionally, a pre-formed network of telomeric DNA threads may allow for a more rapid response to retaining broken chromosomes. However, until the function of telomere threads can be experimentally tested in both normal divisions and in divisions with broken chromosomes, it is possible telomere threads are only involved passively in chromosome segregation. Future research is therefore required to test these ideas. Collectively, our results suggest a novel mechanism in which pervasive anaphase inter-chromosomal connections could be co-opted to retain wayward chromosomes that would otherwise threaten aneuploidy.

STAR Methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Brandt Warecki (bwarecki@ucsc.edu).

Materials availability

This study did not generate new unique reagents

Data and code availability

• All data reported in this paper will be shared by the lead contact upon request.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Drosophila stocks

All Drosophila stocks were raised on standard brown food ⁷⁴ at 25°C with a 12h light/dark cycle. The following stocks from the Bloomington Drosophila Stock Center (BDSC) were used in this study: y1v1 (RRID:BDSC_1509), In(3LR) (RRID:BDSC_1222), C(2)EN (RRID:BDSC_1020), hs-I-CreI (RRID:BDSC_6936), UAS-BubR1-EGFP (RRID:BDSC_91697), UAS-RpA-70-RNAi (RRID:BDSC_35426), UAS-BubR1-DN (RRID:BDSC_8382; generated by Exelixis, Inc.). elav-Gal4 ⁷⁵ was used to drive expression of UAS-BubR1-EGFP, UAS-GFP-Aurora B, UAS-RpA-70-RNAi, and UAS-BubR1-DN. GFP-Polo and GFP-Aurora B stocks were provided by Dr. Joseph Lipsick (Stanford University, Stanford, CA) ⁷⁶. The GFP-BAF stock was provided by Dr. Pamela Geyer (University of Iowa, Iowa City, IA) ⁷⁷. hs-I-CreI (III), hs-FLP, and inverted FRT stocks were provided by Dr. Kent Golic (University of Utah, Salt Lake City, UT) ^53,78. HOAP-GFP stocks were provided by Dr. John Tamkun (University of California, Santa Cruz, Santa Cruz, CA). Timing data for acentric localization and tether formation (Figure S3) was supplemented with additional experiments using GFP-BubR1 ⁷⁹ and GFP-Polo (RRID:BDSC_84275) stocks.

The genotypes of larvae used for fixing and staining experiments were: y1v1=male and female y1,v1/y1,v1 or y1,v1/Y (Figures 1, 2, S1); HOAP-GFP=male and female w*/w*;;HOAP-GFP/HOAP-GFP or w*/Y;;HOAP-GFP/HOAP-GFP (Figure 1); elav-Gal4 + UAS-RpA-70-RNAi=female elav-Gal4/y1,sc*,v1,sev21;UAS-RpA-70-RNAi/H2Av-RFP (Figure S1); female y1v1 =+/y1,v1 (Figure S1); female UAS-RpA-70-RNAi=+/y1,sc*,v1,sev21;UAS-RpA-70-RNAi/+ (Figure S1); In(3LR)=male and female In(3LR)264,mv1/Tm6B,Tb (Figures 2 and S3); C(2)EN=male and female C(2)EN,bw1,sp1 (Figures 2 and S3); hs-FLP + inverted FRTs=male and female 105(inverted FRTs)/w1118;hs-FLP/+ or 105/Y;hs-FLP/+ (Figure 3); I-CreI=male and female hs-I-CreI.2A,v1/hs-I-CreI.2A,v1;;ry506/ry506 or hs-I-CreI.2A,v1/Y;;ry506/ry506 (Figures 3 and S3); I-CreI + elav-Gal4 + UAS-BubR1-DN=female elav-Gal4/y1,w1118;H2Av-RFP/+;UAS-BubR1-DN/I-CreI,Sb (Figure S3); and I-CreI + In(3LR)=female hs-I-CreI.2A,v1/+;;In(3LR)264,mv1/ry506 (Figure S4).

The genotypes of larvae used for live imaging experiments were: y1v1=male y1,v1/Y;H2Av-RFP/+ (Figure S1); elav-Gal4 + UAS-RpA-70-RNAi=female elav-Gal4/y1,sc*,v1,sev21;UAS-RpA-70-RNAi/H2Av-RFP (Figure S1); H2Av-RFP + In(3LR)=female elav-Gal4/+;H2Av-RFP/+;In(3LR)264,mv1/+ (Figure S2); I-CreI + HOAP-GFP=female elav-Gal4/w*;H2Av-RFP/+;I-CreI,Sb/HOAP-GFP (Figure 4); I-CreI + BubR1-EGFP=female elav-Gal4/w*;H2Av-RFP/+;hs-I-CreI,Sb/UAS-BubR1-EGFP (Figure 5); I-CreI + GFP-Polo=female elav-Gal4/+;H2Av-RFP/+;hs-I-CreI,Sb/Ubi-GFP-Polo (Figure 5); I-CreI + GFP-Aurora B=female elav-Gal4/+;H2Av-RFP/UAS-GFP-Aurora B;hs-I-CreI,Sb/+ (Figure 5); I-CreI + GFP-BAF=female elav-Gal4/y1,w*;H2Av-RFP/GFP-BAF;hs-I-CreI,Sb/+ (Figure 5 and S3); BubR1-EGFP=female elav-Gal4/w*;H2Av-RFP/+;UAS-BubR1-EGFP/+ (Figure 5 and S3); and GFP-BAF=female elav-Gal4/y1,w*;H2Av-RFP/GFP-BAF (Figure S3). Additionally, female elav-Gal4/w1118;GFP-BubR1/H2Av-RFP;hs-I-CreI,Sb/+ and female elav-Gal4/w*;hs-GFP-Polo/H2Av-RFP;hs-I-CreI,Sb/GFP-Polo larvae were used to supplement timing experiments in Figure S3.

The genotypes of larvae/flies used for the wing defect analysis (Figure S4) were: I-CreI=hs-I-CreI.2A,v1/y1,v1;;ry506/+ and hs-I-CreI.2A,v1/Y;;ry506/+; In(3LR)=y1,v1/Y;;In(3LR),mv1/+ and y1v1/+;;In(3LR),mv1/+; I-CreI + In(3LR)=hs-I-CreI.2A,v1/Y;;In(3LR),mv1/ry506 and hs-I-CreI.2A,v1/+;;In(3LR),mv1/ry506. For experiments involving both male and female larvae/adults, analysis of the influence of sex was not performed as sex was not expected to be pertinent for these experiments.

Cancer cell culture

PC-3 prostate adenocarcinoma cells that had been derived from a 62 year-old male (RRID:CVCL_0035) were grown at 37°C in a 95%-air/5%-CO₂ atmosphere. Cells were grown in F-12K medium (Gibco) supplemented with 10% fetal bovine serum (Gibco). Cells were maintained in logarithmic growth phase by splitting with a Trypsin-EDTA mixture. Cell authentication was not performed prior to experiments.

METHOD DETAILS

Generation of broken and unbroken acentrics

Broken acentrics were generated by expressing hs-I-CreI through a 1 h 37°C heatshock of 3^rd instar larvae. Larvae were allowed to recover for 1 h before dissection. Larvae were dissected for a period of 1–4 h after heatshock. The I-CreI endonuclease recognizes its cut site in the base of the Drosophila X and Y chromosomes ⁵³. Therefore, I-CreI expression in female larvae resulted in 4 acentrics. Unbroken acentrics were generated by expressing hs-FLP through 1 h 37°C heatshock of 3^rd instar larvae containing the 105 inverted FRTs transgene ¹⁴. Larvae were allowed to recover for 1 h before dissection. Larvae were dissected for a period of 1–4 h after heatshock.

Fixed neuroblast cytology

Larval brains were dissected in 0.7% NaCl and fixed in 3.7% formaldehyde for 30 min. Dissected brains were washed in 45% acetic acid in PBS, squashed between siliconized coverslips and glass slides in 60% acetic acid in PBS, then frozen in liquid nitrogen for at least 10 min. Slides were washed in 20% ethanol (10 min at −20°C), PBS + Triton X-100 (PBST; 10 min) and PBS (2 × 5 min). Slides were blocked for 1h in 5% dried milk in PBST. Slides were washed in PBST (3 × 5 min) then stained with DAPI in Vectashield (Vector Laboratories Cat# H-1200, RRID:AB_2336790). For neuroblasts stained with PI, samples were prepared as above with an overnight RNA digestion step (RNAse A 2.5 mg/mL in PBST) prior to blocking. After the final 3 PBST washes, samples were stained with PI (20 ug/mL in 70% glycerol/30% PBS).

Immunofluorescence

Larval brains were dissected in 0.7% NaCl and fixed in 3.2% paraformaldehyde. Brains were immediately squashed between siliconized coverslips and glass slides in 3.2% paraformaldehyde then frozen in liquid nitrogen for at least 10 min. Slides were incubated with chicken anti-GFP antibody (Aves Labs Cat# GFP-1010, RRID:AB_2307313) at 1:500 in 1X PBS 1% BSA + 0.05% Triton X-100 (PBT) for 1h at room temperature. Slides were rinsed in PBT then stained with anti-chicken-Alexa488 (Thermo Fisher Scientific Cat# A-11039, RRID:AB_2534096) at 1:500 in PBT for 1h at room temperature. Slides were rinsed in PBT, then counterstained with DAPI in Vectashield.

Live neuroblast cytology

Larval brains were dissected in PBS and placed in 17 μL of PBS between a slide and a coverslip. PBS was wicked away to induce gentle squashing ⁷⁹. Coverslips were outlined with halocarbon oil, and slides were imaged immediately. We filmed neuroblasts along the periphery of the brain. Slides were imaged for up to 1 h maximum before a new dissection ⁸⁰. Imaging was performed using a spinning disk confocal (Figure 4–5, S1, S2, S3) or a wide-field (Figure S1) microscope.

Cancer cell cytology

Prior to fixing, PC-3 cells were first split into 2-well chamber slides (ThermoFisher Scientific) and allowed to adhere for 2 days (media changed after first night). For fixing, slides were placed on ice for 30 min, after which cells were fixed in 4% paraformaldehyde in PHEM buffer for 20 min. Slides were rinsed with PBST for 2 min and blocked with 5% BSA in PHEM for 1 h. Cells were stained with rabbit-anti-LaminB1 (1:1000 in PBS; Abcam Cat# ab16048, RRID:AB_443298) overnight at 4°C. The following day, slides were washed twice with PBS (5 min each), stained with anti-rabbit-Alexa488 (1:1000; Thermo Fisher Scientific Cat# A-11008, RRID:AB_143165) and washed twice more with PBS (5 min each). Slides were counterstained with DAPI in Vectashield.

Microscopy and image acquisition

Airyscan microscopy

Super-resolution images were acquired with an AxioObserver.Z1 Zeiss 880 confocal microscope equipped with an Airyscan detector and 63× 1.4 NA Plan-Apochromat objective. Imaging was performed at room temperature. For DAPI, samples were illuminated with a 405 nm laser. For PI, samples were illuminated with a 561 nm laser. For Alexa488, samples were illuminated with a 488 nm laser. Data were collected with 420–480 and 495–550 band pass and 605 long pass filters. Images were collected and processed using the Airyscan mode with Zen Black software (Zeiss) and deconvolved using Zen Blue (Zeiss).

Spinning-disk confocal microscopy

Time-lapse imaging was performed on an inverted Eclipse TE2000-E spinning disk (CSL1-X1; Nikon) confocal microscope equipped with a Hamamatsu electron-multiplying charge couple device camera (ImageEM X2) and a 100× 1.4 NA oil immersion objective. 488 and 561 nm lasers were used to excited GFP and RFP respectively. Imaging was performed at room temperature. Successive timepoints were filmed at 10–18 s. Images were acquired with MicroManager 1.4 software.

Wide-field microscopy

Time-lapse imaging was performed with a Leica DM16000B wide-field inverted microscope with a Hamamatsu electron-multiplying charge coupled device camera (ORCA 9100–02) with a binning of 1 and a 100× 1.4 NA Plan-Apochromat objective. Successive timepoints were filmed at 20 s. Imaging was performed at room temperature. Images were acquired with Leica Application Suite Advanced Fluorescence software and 3D deconvolved using AutoQuant X2.2.0 software.

Scanning confocal microscopy

Fixed and stained PC-3 cells were imaged on an inverted Leica DMI6000 SP5 scanning confocal microscope. DAPI was excited with a 405 nm laser, and emission was collected from 410–470 nm. Alexa488 was excited with a 488 nm laser, and emission was collected from 503–538 nm. Imaging was performed at room temperature. Images were acquired with Leica Application Suite Advanced Fluorescence (Leica) software.

Image processing

The brightness and contrast of images acquired from Airyscan super-resolution experiments were adjusted to score for the presence or absence of inter-chromosomal threads. Often, this meant saturating the signal from the main chromosome masses. Original, unsaturated images of the main chromosome masses are shown with each adjusted image.

Figure preparation

Imaging data was processed in Fiji, and figures were assembled in Adobe Illustrator. Maximum projections are shown for all images unless otherwise indicated. To improve clarity when appropriate, individual images were adjusted for brightness and contrast. Graphs were created using the ggplot2 package ⁸¹ in R.

Experimental design

Specific sample sizes were not estimated prior to experimentation. Instead, each experiment was performed at least 3 independent times. Experiments were performed and analyzed unblinded. For comparisons between different conditions, experiments were performed during the same time period. We did not score 1) anaphase divisions in which the ends of chromosomes were not visually separated or 2) anaphase divisions in which another cell was fixed in plane between separating chromosome masses, as it was impossible to accurately assay for the presence and strength of threads in these situations. All other anaphase divisions were included in analyses.

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantification

Threads were classified and scored as described in Figure 1A-D. We scored divisions as containing threads if we were able to observe ++ (complete, uninterrupted DAPI staining connecting chromosome tips) or + (complete DAPI staining connecting chromosome tips with small gaps in the DAPI staining). We scored divisions as lacking threads if we observed - (DAPI staining from only one chromosome tip) or -- (no apparent DAPI staining) threads. To minimize any confounding background noise, we scored for the presence or absence of threads by analyzing single brightness/contrast enhanced z-slices from an acquired z-stack. Anaphase stage was scored as follows: early=chromosome ends separated by < half the length of a chromosome arm, mid=chromosome ends separated by between half to twice the length of a chromosome arm, or late=chromosome ends separated by greater than twice the full length of a chromosome arm or chromosome arms are decondensed. Representative early, mid, and late anaphases are shown in Figure 1E.

Distances of thread origins to capped telomeres (Figure 1E) were calculated in x-, y-, and z-three-dimensional space in acquired z-stacks by measuring between the points on chromosome ends from where threads emerged to the edge of the nearest HOAP-GFP signal. Thread distances (Figure 2E-E’) were measured by calculating the distance between the ends of connected chromosome tips three-dimensional space in acquired z-stacks. Chromosome lengths (Figure S2A) were measured by calculating the distance in three-dimensional space between the furthest segregated point of a nucleus and the most extended tip of either an extra-long chromosome or a normal-length chromosome.

Histone recruitment to threads (Figure S2C’-E’) was measured using the plot profile feature in Fiji. Angle measurements for coordinated chromosome movement (Figure S2G-G’) were calculated using the angle tool in Fiji. Normalized thread intensity (Figure 3G) was determined in Fiji by measuring the mean grey value of the thread divided by the mean grey value of a region of interest outside the cell. Timing of tether protein recruitment to acentrics (Figure S3D) was determined as the first timepoint at which tether proteins were visible on chromatin clearly identifiable as acentrics. Timing of tether visibility (Figure S3D) was determined as the first timepoint at which tether proteins formed a linear tether between spatially distinct chromosomes. For the timings of BubR1 and Polo recruitment, additional data was gathered with secondary stocks containing GFP-BubR1 and GFP-Polo transgenes.

Analyses of the following experiments contributed to data in multiple figures: fixed and DAPI-stained y1v1 neuroblasts (Figures 1-2, S1), fixed and DAPI-stained In(3LR) neuroblasts (Figures 2 and S2), fixed and DAPI-stained C(2)EN neuroblasts (Figures 2 and S2), and live neuroblasts expressing H2Av-RFP, I-CreI, and BubR1-EGFP (Figures 5 and S3), GFP-Polo (Figures 5 and S3), GFP-Aurora B (Figures 5 and S3), or GFP-BAF (Figures 5 and S3).

Statistical analyses

All statistical analyses were performed in R (R Core Team). Two-sided Fisher’s exact tests (Figure 1E’, Figure 2D, Figure 3F, Figure S1E, Figure S4B) were performed using 2 × 2 and 3 × 2 contingency tables, when appropriate. Mann-Whitney tests (Figure 2E’, Figure 3G, Figure S1F, Figure S2B) and Kruskal-Wallis tests (Figure 2E) were used when comparing two or more continuous datasets respectively. N values refer to the number of neuroblasts analyzed except for: Figure 1G-H’ (N=number of thread origins); Figure 3G (N=number of threads); Figure 4D + 5G (N=number of acentrics); and Figure S4B (N=number of flies). N values are listed in either the figure panels, figure legends, and/or written results sections. To show variation, all data were graphed as individual points in dot plots for non-binary continuous datasets. Statistical significance was defined as p<0.05 (*), p<0.01 (**), p<0.001 (***).

Supplementary Material

Video S1

Video S1. Histones coat tethers connecting the broken ends of acentric fragments to segregated chromosomes. Related to Figure 4.

Video S2

Video S2. Histones are recruited to the DNA threads connecting the telomere ends of acentric fragments to segregated chromosomes. Related to Figure 4.

Video S3

Video S3. BubR1 kinase is recruited to tethers on both broken and unbroken ends of an acentric fragment. Related to Figure 5.

Video S4

Video S4. Polo kinase is recruited to tethers on both broken and unbroken ends of an acentric fragment. Related to Figure 5.

Video S5

Video S5. Aurora B kinase is recruited to tethers on both broken and unbroken ends of an acentric fragment. Related to Figure 5.

Video S6

Video S6. BAF is recruited to tethers on both broken and unbroken ends of an acentric fragment. Related to Figure 5.

Video S7

Video S7. Neither BubR1 kinase nor histones are recruited to telomere DNA threads when chromosomes are unbroken. Related to Figure 5.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
chicken anti-GFP	Aves Lab	Cat# GFP-1010; RRID:AB_2307313
goat anti-chicken-Alexa488	Thermo Fisher Scientific	Cat# A-11039; RRID:AB_2534096
rabbit anti-LaminB1	Abcam	Cat# ab16048; RRID:AB_443298
goat anti-rabbit-Alexa488	Thermo Fisher Scientific	Cat# A-11008; RRID:AB_143165
Experimental Models: Cell Lines
PC-3 cells	Gift from Wang Lab	RRID:CVCL_0035
Experimental Models: Organisms/Strains
D. melanogaster: y1v1: y1,v1	Bloomington Drosophila Stock Center	RRID:BDSC_1509
D. melanogaster: In(3LR): In(3LR)264,mv1/Tm6B,Tb1	Bloomington Drosophila Stock Center	RRID:BDSC_1222
D. melanogaster: C(2)EN: C(2)EN,bw1,speck1	Bloomington Drosophila Stock Center	RRID:BDSC_1020
D. melanogaster: I-CreI: P{v+t1.8=hs-I-CreI.R}2A,v1;ry506	Bloomington Drosophila Stock Center	RRID:BDSC_6936
D. melanogaster: BubR1-EGFP: w*;P{w+mC=UAS-BubR1.EGFP}III.1	Bloomington Drosophila Stock Center	RRID:BDSC_91697
D. melanogaster: UAS-RpA-70 RNAi: y1sc*v1sev21;P{y+t7.7v+t1.8=TRiP.GL00350}attP2	Bloomington Drosophila Stock Center	RRID:BDSC_35426
D. melanogaster: UAS-BubR1-DN: y1w1118;P{w+mC=UAS-BubR1.DN}3	Bloomington Drosophila Stock Center, generated by Exelixis, Inc.	RRID:BDSC_8382
D. melanogaster: elav-Gal4: Elav-Gal4	Lin and Goodman75	N/A
D. melanogaster: GFP-Polo: Ubi-GFP-Polo	Gift from Lipsick Lab76	N/A
D. melanogaster: GFP-Aurora B: UAS-GFP-Aurora B	Gift from Lipsick Lab	N/A
D. melanogaster: I-CreI (III): w1118;P{v+t1.8=hs-I-CreI.R}1A,Sb1/Tm6	Gift from Golic Lab53	RRID:BDSC_6937
D. melanogaster: hs-FLP: w1118;P{ry+t7.2+70FLP}10	Gift from Golic Lab78	RRID:BDSC_6938
D. melanogaster: inverted FRTs: P{FRT(whs)}(8F)105	Gift from Golic Lab78	N/A
D. melanogaster: HOAP-GFP: w*;HOAP-GFP	Gift from Tamkun Lab	N/A
D. melanogaster: GFP-BubR1: w1118;GFP-BubR1	Buffin et al.79	N/A
D. melanogaster: GFP-Polo: w*;P{w+mW.hs=GFP-polo}p2;P{GFP-polo}p31	Bloomington Drosophila Stock Center	RRID:BDSC_84275
D. melanogaster: GFP-BAF: y1w*;GFP-BAF	Gift from Geyer Lab77	N/A
Software and Algorithms
R	R Core Team	Version 4.0.5
ggplot2	Wickham81	Version 3.3.5
Other
DAPI	Vector Laboratories	Cat# H-1200; RRID:AB_2336790

Highlights.

• DNA threads connect sister and non-sister telomeres in most early anaphases

• Telomere threads are normally resolved in mid-to-late anaphase

• Telomere threads recruit a set of proteins upon chromosome breakage

• Persistent protein-coated telomere threads promote retention of broken chromosomes