Saccharomyces cerevisiae strains are MATa/MATα derivatives of wild-type SK1 as follows: scp1::tetO array::LEU2/SCP1, leu2::URA3p-TetR-mEGFP::LEU2/leu2::hisG (TNY570); scp1::tetO array::LEU2/SCP1, leu2::URA3p-TetR-mEGFP::LEU2/leu2::hisG, ndj1Δ/ndj1Δ (TNY669), ABP140-4xGFP::KanMX/ABP140-4xGFP::KanMX (YKK389), SPC42-YFP::URA/SPC42 (TNY866). TetO array system is based on Kim et al. (2010).

All operations were performed at 30°C. Strains maintained in glycerol stock at −80°C are patched onto YEPG plates (3% w/v glycerol, 2% w/v bactopeptone, 1% w/v yeast extract, 2% w/v bactoagar) overnight. Cells were struck out to single colonies on YEPD plates (2% w/v bactopeptone, 1% w/v yeast extract, 2% w/v glucose, 2% w/v bactoagar) and grown for 2 days. A single colony is transferred to 4 ml YEPD liquid medium (2% w/v bactopeptone, 1% w/v yeast extract, 2% w/v glucose) and grown overnight. A 1/100 dilution of the culture was made with YEPA medium (1% w/v potassium acetate, 2% w/v bactopeptone, 1% w/v yeast extract, 2 drops per liter antifoam) and grown for 13.5 h. Meiosis was initiated by transfer of cells to 1% sporulation medium (SPM) (1% w/v potassium acetate, 0.02% w/v raffinose, 2 drops per liter antifoam). We note that this series of preparation conditions is specifically defined in such a way that, at the time of transfer to the SPM, cells have completed ongoing mitotic cell cycles and are becoming larger without initiating a new mitotic cycle and thus are present primarily as large unbudded cells. Upon transfer to SPM, this population initiates meiosis synchronously and efficiently (e.g., Kim et al., 2010).

Latrunculin B (LatB) was from Santa Cruz Biotechnology Inc. and was dissolved in dimethyl-sulfoxide (DMSO). LatB was added to a final concentration of 30 μM at 2 h after initiation of meiosis by transfer to SPM and imaged after an additional 2 h (t = 4 h of meiosis).

Cell samples from an experimental culture were vortexed at full speed and 1 μl quickly spread onto a glass base dish (MatTek) coated with Concanavalin A (ConA). A premade agarose pad (1%) was placed on top of the cell drop and excess media was absorbed by a piece of Kimwipe. Cells were observed at 30°C using a Ti microscope (Nikon) equipped with GFP filters (Semrock), a sCMOS camera (Hamamatsu Photonics), and a piezo device (Physik Instrumente) for acquiring Z stacks. Cells were exposed to the LED light (Lumencor) through an objective lens (60× PlanApo, NA 1.40; Nikon). The microscopy system was controlled, and images were acquired, through μ-Manager software and MATLAB.

For short time scaled imaging, movies of 15 z-stack steps with 389-nm step size and 720 sequential frames were acquired using μ-Manager software with 25 ms exposure time and 8.3 ms camera acquisition time, totaling 500 ms for one set of z-stack images. For long timescale imaging, movies of 13 z-stack steps with 389-nm step size were acquired using μ-Manager software with 10 ms exposure time and 16.6 ms camera acquisition time. Z-stacks were acquired at 1 min interval for 10 h, giving a total of 601 z-stack images.

Images were denoised quantitatively by a home-made program that accurately defines the presence and positions of fluorescent spots at very low signal-to-noise ratios (Chang, 2018). This algorithm dramatically extends the ability to capture many images over long periods of time, as illustrated in this work. Spots were detected and tracked by ImageJ Fiji plug-in TrackMate (Schindelin et al., 2012; Tinevez et al., 2017). The distribution of spot displacements per time interval and the mean square displacement (MSD) of the fluorescent signals were calculated in 3D based on this TrackMate data. The MSD was calculated by the following formula. MSD = <(X_t − X_{t + Δt})²>. X_t was the three-dimensional position of a fluorescent spot at time t (Seeber et al., 2013). The multi-component Gaussian mixture model is fitted to the step size data of the spot motion using mixtools in R packages (Benaglia et al., 2009). To eliminate the influence of large step size (Group 1, see below) to the 5 s interval analysis, the ± 10 frames (5 s) from the frame that shows the large step in 500 ms were eliminated. We note that, due to specific features of our imaging system and spot detection algorithm, the spatial precision of spot detection in our system is high and that there is no detectable photobleaching or phototoxicity, as discussed in the text and also in Supplementary Notes.

For fixed cell analysis, cells were transferred to 4% paraformaldehyde at 4 h after initiation of meiosis as described above. Cells were then spun down and the cell pellet was resuspended in 0.1 M potassium phosphate for imaging. Other procedures are the same as for live-cell imaging.

Telomere motion was examined analogously in premeiotic G1/G0 cells (defined as large unbudded cells observed immediately upon transfer to sporulation medium; “Materials and Methods,” i.e., t = 0 h, Figures 1C,E,G) and in meiotic prophase cells observed at t = 4 h after initiation of meiosis, the time at which dramatic chromosome movements are known to occur (Koszul et al., 2008; see Supplementary Figure 4 and Figures 1D,F,H). Parallel control experiments confirm that cells are at late zygotene or early pachytene at this time, as judged by Zip1-GFP patterns (T.N. unpublished).

In accord with previous observations, telomere motion is limited at pre-meiotic G1/G0 (t = 0 h) and dramatic at meiotic mid-prophase (t = 4 h). This difference is qualitatively apparent in representative kymographs which show the position of the spot as projected onto the X-, Y-, or Z-plane over a 3 min time period (Figure 1C vs. Figure 1D) and in corresponding 3D trajectories (Figure 1E vs. Figure 1F). At G1/G0, the telomere locus never moves far from its starting point in 3 min (Figure 1E) while at mid-prophase, it traverses distances comparable to the diameter of the nucleus (∼2 μm) (Figure 1F).

The difference between the two situations can be described quantitatively. Plots of mean square displacement (MSD) over time are widely used with single-particle tracking to quantify chromatin and chromosome dynamics (Heun et al., 2001; Sage et al., 2005; Miné-Hattab and Rothstein, 2012; Hajjoul et al., 2013; Nozaki et al., 2017). For normal diffusion, the MSD of a tracked “spot” varies linearly with time. In contrast, at both G1/G0 and mid-prophase, the marked telomere exhibits a rising curve which reaches a plateau. This behavior, referred to as “anomalous diffusion,” implies that the movement of the tracked particle is somehow constrained, with the distance reached at the plateau defined as the “radius of constraint” (Rc) (Neumann et al., 2012). Confinement is severe at G1/G0 (Rc = 0.64 μm; Figure 1G, n = 10 cells). Interestingly, this pattern is very similar to that observed by analogous MSD analysis in yeast mitotic interphase cells (Sage et al., 2005). Confinement is much less severe at mid-prophase, with Rc = 1.46 μm, roughly comparable to the diameter of the nucleus (Figure 1H, n = 17 cells), in accord with the fact that telomeres are actively moved around the nuclear periphery during this period (Koszul et al., 2008). The mid-prophase MSD curve also rises more sharply than the G1/G0 MSD curve. This difference implies that the telomere locus tends to move farther in a given period of time in mid-prophase vs. G1/G0, in accord with the presence and absence of active actin/Myo2-mediated movements in the two situations, respectively.

Mid-prophase telomere motions are substantially reduced by deletion of meiotic telomere protein Ndj1 (which connects telomeres to the LINC complex; Background) (ndj1Δ; Figures 1I,K,M). These movements are also dramatically reduced by treatment with LatB, which blocks actin polymerization (Figures 1J,L,N). Thus, as expected, the dynamic prophase motions of telomere-located scp1::tetO-TetR observed at mid-prophase are derived from the force generated by the linkage of the telomere to cytoskeletal actin fiber(s) via LINC complexes.

Interestingly, the overall behavior of telomeres in ndj1Δ is similar to that observed at G1/G0 (compare Figures 1I,K with Figures 1C,E; further discussion below). Also, the effect of LatB is more severe than that of ndj1Δ, perhaps because telomeres are now locked onto immobile LINC complexes rather than being free of such complexes.

It is well known that cytoskeletal forces can also provoke whole nucleus movements during meiotic prophase (e.g., in budding yeast, Conrad et al., 2008). To evaluate the potential contribution of such movements to telomere locus dynamics, we tracked the SPB component SPC42, tagged with YFP (Koszul et al., 2008). Since the SPB is embedded in the nuclear envelope throughout prophase (Koszul et al., 2008), its motions are often used to define whole nucleus movements within the cell in meiosis (Conrad et al., 2008) and in mitotic cells (Heun et al., 2001). This analysis suggests that whole-nucleus movements do not contribute significantly to either “directed” or “straight” motions; however, they likely account for some of the motions seen as “pauses.” This conclusion is validated by three findings:

Finally, we note that additional global whole nucleus movements may tend to occur over longer time scales than the 6 min duration of imaging used in the present analysis.

Previous time-lapse data of prophase movements indicated that telomeres remain associated with the nuclear envelope during periods of movement and during pauses (Koszul et al., 2008). We now show that this association is maintained throughout the entire progression of complex motions described above. Here, the 3D outline of the nucleus can be defined by the “fuzz” of fluorescence from TetR-mEGFP that is nuclear localized but not specifically bound to the chromosomes that reflects its overall nuclear localization (Figures 3A–C). Three-dimensional reconstructions show that during the mid-prophase dynamic motion period (t = 4 h), the tagged telomere locus is attached to (or closely associated with) the nuclear periphery at all times, irrespective of whether it is exhibiting straight, directed or paused motion (as defined by visual inspection; Figures 3D,H,I).

Interestingly, the telomere is also associated with the nuclear periphery in the ndj1Δ mutant (Figure 3E). The same is true during premeiotic G1/G0 (Figure 3G) and is also known to be the case for telomeres in mitotic cells (Sage et al., 2005). In all three cases, such association is independent of the meiosis-specific LINC complex, which is absent in ndj1Δ (Conrad et al., 2008) and unformed during pre-meiosis and in mitotic cells. This condition of LINC-independent nuclear envelope association can thus be inferred to underlie the “jiggling in place” movements observed in all three situations. By extension, telomeres might also be in this state during paused and indirectly promoted movements in wild type meiosis (Discussion). Telomeres also remain associated with the nuclear surface in LatB treated cells (Figure 3F). Whether telomeres remain associated with LINC complexes in this condition or not remains to be determined. It is possible that absence of an actin filament triggers disassembly of LINC complexes to give ndj1Δ-like associations with the nuclear periphery also in this situation (Discussion).

We sometimes observe cases in which a period of straight telomere movement is immediately followed by a rapid reverse motion (Figures 2O, 3I, large arrow). These “transition” effects are reminiscent of the previously described “recoil” of FROS-tagged loci (Conrad et al., 2008). We further find that a transition involving such reverse motion is accompanied by change in the shape of the nuclear surface (Figures 2M, 3H, large arrow). The nucleus initially elongates concomitantly with, and in the same direction as, an outward-directed straight telomere movement; then, beginning at the time of telomere recoil, the nucleus returns to its normal, roughly spherical, shape. We suggest that a complex between a telomere/LINC/nuclear envelope becomes attached to a nucleus-hugging actin fiber and then follows that fiber away from the nucleus, dragging the associated nuclear envelope region with it. If the telomere complex is then released from the actin filament, the result would be recoil of the telomere and restoration of nuclear shape. A general association of nuclear shape changes at points of telomere attachment (Koszul et al., 2008) and during rapid prophase chromosome movements (Conrad et al., 2008) have been described previously. Among these are cases of long actin-mediated nuclear protrusions with (or without) an associated orphan chromosome (Koszul et al., 2008). In such cases, pulling on the nuclear envelope results in separation of the nuclear membrane from constraining features to give a long “membrane tube.” We suggest that the global telomere-linked nuclear deformations identified in the current analysis could be a major factor in limiting the duration of direct Myo2/actin-promoted telomere motion and an important contributor to indirectly promoted telomere movements (Discussion).

Movement of an actin-linked telomere through space will be the net result of the combined effects of Myo2-mediated tracking along an actin fiber and motion of the actin fiber itself. To further elucidate the contributions of the latter effect, we visualized ABP140-GFP signals in mid-prophase of meiosis.

Visualization of actin fibers in mitotic cells of budding yeast has defined two types of movement: dynamic movements of filaments through 3D space and treadmilling, in which subunits are added to one end and lost from the other (Yang and Pon, 2002). The two types of dynamics can be defined and distinguished by analysis of local strongly staining regions which provide fiduciary marks (Yang and Pon, 2002). Movement of the actin fiber through space (irrespective of treadmilling) is indicated by movement of a fiduciary mark. Treadmilling is implied by changes in the lengths of the fiber segments on either side of the fiduciary mark, with growth at the (+) end and shortening at the (−) end. In the present study, by applying analogous methods, we could see both types of processes occurring during meiotic prophase.

We observed movement of a fiduciary mark, and thus the actin fiber, at ∼0.29 μm per sec and 0.26 μm per sec (Figure 4A and Supplementary Figure 1A). This is close to the 0.52 μm/s movement reported in mitotic cells (Yang and Pon, 2002). Movements in mitotic/meiotic cells include cases in which the actin fibers are associated either along the outer nuclear periphery, presumptively by association with the outer nuclear membrane (Koszul et al., 2008; Supplementary Figure 1C), or along the cell periphery, in accord with the well-known association of actin with the plasma membrane (Moseley and Goode, 2006). These dual localizations are illustrated here for meiotic prophase by an example in which an actin fiber moves from association with the edge of the cell (defined by cellular “fuzz”) to the outer surface of the nucleus (defined by the absence of cellular “fuzz”) (Figure 4C).

Finally, we can also observe treadmilling. Here, the distance between a fiduciary mark and the one end of a filament increased at ∼0.36 μm per sec and ∼0.15 μm per sec (Figure 4B and Supplementary Figure 1A), similarly to the ∼0.3 μm/s rate defined for this process in mitotic cells (Yang and Pon, 2002). We note, however, that (contrary to previous considerations; Koszul et al., 2008; Lee et al., 2020), treadmilling cannot contribute to telomere movement, because addition/subtraction at the ends of a filament will not affect the position of a centrally positioned telomere/LINC/Myo2/actin complex.

A marked telomere sporadically undergoes fast, brief, straight/curvilinear motion. Myosin protein Myo2, which translocates along actin fibers, has been implicated as a major mediator of prophase movements (Lee et al., 2020). The movements defined in the present study as Group 1 by step-size analysis and as “straight” motion in phenotypic analysis, match those attributed to direct, active Myo2/actin-mediated transport in other systems. We observe average speeds of telomere movement of ∼0.5 μm per 500 ms step, with speeds as high as ∼1.3 μm per 500 ms step observed in some cases (Figures 2E,J). The speed of Myo2-mediated motion along a filament varies with its cargo, but has been reported to be as high as ∼1–3 μm/s in mitotic cells (Beach et al., 2000; Schott et al., 2002).

We also demonstrate that actin fiber localization is as dynamic during meiotic prophase as previously reported for mitotic cells. This is important because the speed of a particular Myo2-mediated event will reflect the combined rates of directional Myo2 tracking along an actin fiber, from the + end to the − end of the filament (Förtsch et al., 2011), and of movement of the fiber itself. Thus, actin fiber movement may either add to, or subtract from, the Myo2 tracking rate. This interplay is likely an important contributor to the range of rates of motion observed for very fast (directly promoted) telomere movements.

We also note that myosins are capable of smoothly negotiating actin fiber branches or switching smoothly from one fiber to another, without apparent disruption of movement (Hammer and Wu, 2007). Such effects could explain cases in which two straight trajectories are linked by an abrupt turn (e.g., Figure 2M, single asterisk).

The current analysis provides further evidence that Myo2/actin-mediated motion is, for a given telomere, quite sporadic. Step size analysis suggests that these motions occupy only ∼17% of the telomere’s life (Figure 2H). This finding matches previous evidence from whole chromosome analysis which also suggests that, at any given moment, among the ∼32 telomeres within a given mid-prophase nucleus, only one or two are undergoing actin-mediated motion (Koszul et al., 2008; Supplementary Figure 1B). This feature, which is likely important for the in vivo roles of motion, is not unexpected given that actin fibers only contact the nuclear periphery in a few places and transiently.

These considerations also emphasize the important fact that a given individual telomere spends most of its time in a “resting state” where it may be either motionless (“paused”) or undergoing passive motion promoted indirectly by direct actin-mediated motion of some other (nearby) telomere (“indirect movement”). In the current study, the marked telomere spends ∼85% of its time in one or the other of these states. Moreover, telomeres exhibit the same rates of movement (as defined by 500 ms step sizes) in both conditions. This correspondence suggests, as might be expected a priori, that the only difference between these two states is whether the monitored telomere is, or is not, subjected to indirect effects from active movement of some other telomeres.

Two further, related questions thus arise:

Simultaneous visualization of telomere position and nuclear shape has provided new information regarding the interplay of actin-mediated telomere movement and nuclear shape. We can document cases in which Myo2/actin-mediated movement of the monitored telomere is correlated with, and apparently responsible for, global nuclear envelope deformation. The same relationship is prominent in whole chromosome analysis where telomere-led motion results in indirect movement of multiple nearby chromosomes, thereby changing the entire shape of the chromosome complement and thus, presumptively, the nuclear envelope (Supplementary Figure 1B). The phenomenon of global nuclear envelope deformation raises two important questions.

The methodology used in the current study depends on specialized imaging methods and a novel denoising algorithm in which a two-stage maximum likelihood pipeline is used to accurately detect very faint spots (Kleckner and Chang, 2017; Chang, 2018). These approaches have significant advantages with respect to those described previously for both short time-scale and long time-scale imaging. In the present study we mainly exploited short this method for spot detection for rapid imaging over time scales of minutes (∼500 ms per single 3D stack for ∼6 min). However, it is now technically possible to carry out long time-scale 3D imaging from pre-meiotic G1/G0 to the MI/MII divisions and beyond, over a period of up to 10 h or more, with 3D images taken at 1 min intervals and with low photobleaching and low phototoxicity (Supplementary Figures 4A–C). In addition, two-color 3D imaging is also possible, both for short and long-time scales. Our methods should thus provide a powerful tool to further investigate many processes of interest, including (but not limited to) meiotic pairing, chromosome structure, chromosome movement, and nuclear envelope mechanics and dynamics.