Processive translocation of cohesive and non-cohesive cohesin in vivo

Cohesin is a central architectural element of chromosomes that regulates numerous DNA-based events. The complex holds sister chromatids together until anaphase onset and organizes individual chromosomal DNAs into loops. Purified cohesin translocates along DNA in a diffusive fashion that can be propelled by transcribing RNA polymerase. The complex also extrudes DNA loops in a process that consumes ATP. In this study we examine processive genomic translocation of cohesin in vivo. To this end, obstacles of increasing size were tethered to DNA to act as roadblocks to complexes mobilized by transcription in yeast. The obstacles were built from a GFP-lacI core fused to one or more mCherries. A chimera with four mCherries blocked cohesin passage in late G1. During M phase, the threshold barrier depended on the state of cohesion: non-cohesive complexes were also blocked by four mCherries whereas cohesive complexes were blocked by as few as three mCherries. Furthermore, cohesive complexes that were stalled at obstacles, in turn, blocked the passage of non-cohesive complexes. That synthetic barriers alter cohesin redistribution demonstrates that the complex translocates processively on chromatin in vivo. The approach provides a relative measure of the maximum size of the DNA binding chamber(s) of cohesin. Together, this study reveals unexplored limitations to cohesin movement on chromosomes. Significance Statement Cohesin is an architectural protein that brings distant chromosomal DNA sites together. The complex links sister chromatids cohesion but it also binds to single pieces of DNA in ways that do not generate cohesion. One class of non-cohesive complexes organizes chromosomal DNA into loops. All cohesin complexes move on DNA but the constraints on such movement are not fully explored. Here, we use size-calibrated obstacles in yeast to interrogate cohesin and the properties of its movement on DNA. We show that both cohesive and non-cohesive complexes translocate processively on chromosomes. In addition, we show that cohesive and non-cohesive complexes are blocked by obstacles of different size. Lastly, we show that stalled cohesive complexes block passage of non-cohesive complexes.


Introduction
block escape of cohesin ( Figure 1A). We hypothesized that the bacterial protein (lacI fused to 124 GFP) was smaller than the chamber of the complex through which DNA passes. If topologically-125 bound cohesin truly redistributes by translocation in vivo, then DNA-bound proteins of sufficient 126 size should block passage of the complex. To begin to explore this concept, we used a GFP-lacI 127 chimera linked to FFAT, a tetrapeptide motif that associates with the inner nuclear membrane 128 (36). When simultaneously tethered to DNA and the nuclear membrane, GFP-FFAT-lacI 129 represents an immobile obstacle of essentially infinite size ( Figure 1A). When GFP-lacI was expressed, cohesin at 5'lacO was low both before and after URA3 induction 143 ( Figure 1C). The result is consistent with the inability of GFP-lacI to block translocating cohesin 144 complexes. Expression of GFP-FFAT-lacI, on the other hand, increased the level of cohesin at 145 5'lacO following transcription. This finding serves as a proof-of-principle that large immobile 146 objects block translocation of cohesin complexes in yeast. In B. subtilis, tethering a DNA-bound 147 repressor to the plasma membrane similarly blocked translocation of bacterial SMC proteins 148 To determine if large proteins could similarly block cohesin passage, a series of size-calibrated 152 obstacles was generated by fusing increasing numbers of mCherry to GFP-lacI (Figure 2A-C). 153 This red fluorescent protein is ideally suited as a size standard because it has a compact 154 monomeric structure unlike its tetrameric progenitor, . Like GFP-lacI, the 155 mCherry chimeras bind lacO arrays, albeit with reduced efficiency, as demonstrated by weaker 156 GFP foci and higher background nuclear fluorescence ( Figure S1). 157 158 Accumulation of cohesin at 5'lacO was measured in M phase-arrested cells following three 159 different URA3 induction protocols ( Figure 2D). In the first, URA3 was never induced. 160 Accordingly, cohesin levels at 5'lacO remained low and comparable to the cohesin-free control 161 site, which is represented by a dashed line in the figures. In the second protocol, URA3 was 162 induced following imposition of an M phase arrest. In this case, accumulation of cohesin at 163 5'lacO depended on the identity of the GFP-lacI chimera expressed ( Figure 2C). When GFP-lacI 164 was fused to one or two mCherries or none at all, cohesin levels did not rise above the baseline. 165 By contrast, when GFP-lacI was fused to three or four mCherries cohesin accumulated at 5'lacO. 166 A more thorough analysis of the cohesin distribution around the engineered locus is shown in 167 figure S2. In the three mCherry strain, the most dramatic changes in cohesin binding occurred 168 only between URA3 and the lacO array, as expected for complexes travelling from the induced 169 gene into a downstream roadblock. No dramatic changes were found with GFP-lacI alone, 170 including at sites downstream of the array. This suggests that cohesin distributes throughout the 171 region in the absence of a single strong barrier. To be certain that an aberrant property of 172 mCherry was not responsible for 5'lacO accumulation, the experiment was repeated with a 173 chimera bearing three copies of monomeric yeGFP. Similar results were obtained ( Figure S3). 174 These results indicate that DNA bound proteins of sufficient size block passage of translocating 175 cohesin complexes. 176

177
In the final experimental protocol, URA3 was induced at the outset of asynchronous growth. 178 Previously we found that a downstream convergently-oriented gene could not block translocation 179 of cohesive complexes under these conditions. The basis for this phenomenon was not 180 understood (35). When URA3 was induced from the outset here, most of the barrier chimeras 181 similarly failed to accumulate cohesin at 5'lacO. Some binding, however, was detected when 182 GFP-lacI was fused to four mCherries ( Figure 2E). The data suggests that different sized barriers 183 act as obstacles under different conditions. Additional experiments elaborate on this concept 184 below. 185 186 Trapped cohesin complexes remain cohesive 187 To determine whether the complexes at 5'lacO hold sister chromatids together, we evaluated 188 cohesion of the chromosomal domain. In our strains, target sites for the R site-specific 189 recombinase flank the entire region ( Figure 3A). Induction of the recombinase during an M 190 phase arrest generates a pair of DNA circles, one from each sister chromatid. Owing to the bound 191 GFP-lacI chimeras, a single dot of fluorescence appears if the DNA circles are cohesed whereas 192 a pair of dots appears if they are not ( Figure S1). URA3 transcription was induced according to 193 the three protocols described above and cohesion was evaluated at the end of the growth protocol 194 ( Figure 3B). In the absence of induction, cohesion of DNA circles was observed in all strains due 195 to the presence of cohesive cohesin on URA3 ( Figure 3C; (35)). When transcription was induced 196 following M phase arrest, cohesion was greatly reduced in strains with chimeras bearing one or 197 two mCherries or none at all. This result indicates that these chimeras are too small to block 198 translocating cohesin. By contrast, cohesion persisted when transcription was induced in strains 199 expressing chimeras with three and four mCherries. Thus, consistent with the results in figure 2, 200 DNA-bound chimeras equal to or larger than three mCherries block passage of translocating 201 cohesive cohesin. We conclude that at least some of the trapped cohesin complexes detected by 202 ChIP-qPCR are engaged in sister chromatid cohesion. 203

204
When transcription was induced from the outset of cell growth, cohesion of the DNA circles was 205 abolished in all strains, irrespective of the chimera expressed. This result recapitulates the earlier 206 observation that convergent genes downstream of URA3 also do not trap cohesive complexes 207 under identical induction conditions (35). These data suggest that the ability of an obstacle to 208 block passage of cohesin depends on the interplay between transcription and cell cycle stage. 209 Data supporting this view is presented below. 210 211 A larger barrier is required to block cohesin translocation in late G1 phase 212 The studies above examined translocation of cohesin on newly replicated chromatids in M 213 phase-arrested cells. To study translocation on unreplicated chromatin, cells were arrested in late 214 G1 by expressing a non-degradable allele of Sic1, a Cdk1 inhibitor that prevents S phase entry 215 (41). By definition, the chromatin-bound complexes at this stage of the cell cycle are not 216 cohesive. The steps of the growth protocol are depicted in figure 4A and validation of the growth 217 arrest is shown in figure S4. In this scenario, transcriptional induction caused accumulation of 218 cohesin at 5'lacO only in a strain expressing the four mCherry barrier. In strains expressing three 219 mCherries or GFP-lacI alone, cohesin accumulation matched the background levels of the 220 uninduced controls ( Figure 4B). This data suggest that different size barriers are required to 221 block cohesin at different stages of the cell cycle. Apparently, something happens during S phase 222 that affects the how cohesin translocates past obstacles on DNA. 223

224
We hypothesized that the four mCherry threshold seen in figure 4B was related to the same 225 threshold seen during M phase arrest after induction from the outset of growth ( Figure 2E). To 226 understand this line of reasoning, recall that cohesin binding resets after each anaphase passage 227 when separase activity subsides and Mcd1 synthesis resumes (42-44). If URA3 is already 228 transcribing when cohesin binding resumes in late G1 then the ensuing translocation of the 229 complex would occur on an unreplicated chromosome, at least until S phase. Put simply, 230 translocation before DNA replication would bypass the three mCherry limitation normally seen 231 in M phase cells. To test this idea, URA3 was induced from the outset of growth and barrier 232 activity of the obstacles was analyzed by ChIP-qPCR after cells were arrested in late G1. Figure  233 4C shows that cohesin accumulated at 5'lacO in cells that expressed a four mCherry obstacle but 234 not the three mCherry variant or GFP-lacI alone. These data indicate that when URA3 is induced 235 from the outset of growth, transcription triggers translocation of cohesin before DNA replication, 236 leading to mobile complexes that pass obstacles the size of three mCherries but not four 237 mCherries. 238 239 Translocation of complexes that arrive after DNA replication 240 Cohesin loads continuously onto chromosomes from late G1 to anaphase (19,35,45). 241 Complexes that arrive after DNA replication are not cohesive, do not acquire S phase-dependent 242 acetylation and are subject to turnover by Wpl1 (46-50) To gain a more complete picture of 243 cohesin dynamics at URA3, transcription was induced while cells were arrested in late G1 and 244 then the cultures were released for growth into a subsequent M phase arrest. ChIP-qPCR at this 245 stage thus reports on those cohesin complexes that load and translocate to 5'lacO before S phase, 246 as well as those that load and translocate after S phase. Formally, this induction protocol also 247 records translocation of complexes that load during DNA replication of the gene, a time span that 248 is likely too short to affect the results significantly. The novel arrest-and-release procedure was 249 achieved by inducible clearance of AID-tagged, "non-degradable Sic1", using the auxin analog 250 1-naphthaleneacetic ( Figure S4). The ChIP-qPCR data of figure 4D show that when URA3 251 transcription was induced, cohesin only accumulated at 5'lacO in the strain expressing the four 252 mCherry obstacle. That the barrier limit did not revert to three mCherries indicates that late 253 arriving, non-cohesive cohesin was not blocked by the smaller barrier.   Figure 5C shows that the 265 late arriving TAP-tagged cohesin now bypassed the three mCherry obstacle and was instead 266 blocked by four mCherries ( Figure 5C). The accumulation signal was weak but significant. Thus, 267 in the absence of cohesive cohesin, late-arriving, non-cohesive complexes stopped only at the 268 larger barrier to translocation. Importantly, the data show that stalled cohesive complexes block 269 the passage of non-cohesive complexes.  Transcriptional regulation of URA3 was unchanged by the depletions (Figures S5D and S6B). In 277 these analyses, ChIP-qPCR data of transcribed and non-transcribed samples were always 278 compared to avoid impact that the mutants might have on absolute levels of cohesin binding. 279 280 When Eco1 was depleted the smallest obstacle, GFP-lacI alone, blocked cohesin translocation 281 ( Figure 6B). This unexpected result was due specifically to loss of Eco1 because it did not occur 282 when DMSO was added instead of auxin (data not shown) or when Eco1 lacked the AID tag 283 ( Figure 6C). The blockade was also dependent on the synthetic barrier: when Eco1 was depleted 284 from a strain lacking GFP-lacI accumulation of cohesin at the 5'lacO site was abolished (data not 285 shown). It appears that Eco1, presumably through Smc3 acetylation, facilitates translocation past 286 DNA-bound GFP-lacI. 287

288
To explore the phenomenon further, we examined cohesin translocation in a strain with an smc1-289 D1164E allele. This mutation diminishes ATP-hydrolysis by cohesin, creating chromatin-bound 290 complexes that do not require Eco1 for protection against Wpl1 unloading (16,51,52). 291 Importantly, this mutation reduces Smc3 acetylation by Eco1 (52). In this strain, URA3 292 transcription during an M phase arrest also caused cohesin to accumulate at GFP-lacI alone 293 ( Figure 6D). These results reinforce the notion that acetylation of Smc3 by Eco1 influences how 294 cohesin navigates the GFP-lacI array. 295 296 When Wpl1 was depleted, two notable changes occurred. First, depletion of the protein abolished 297 the three mCherry barrier, yielding a slight increase in accumulation at the four mCherry barrier 298 ( Figure 6E). The small signal amplitude may be related to earlier observations that cohesin levels 299 are lower at enrichment sites in the absence of Wpl1 (9, 17, 18). These data suggest that cohesin 300 translocation is less restrained in the absence of Wpl1. The second notable change in figure 6E is 301 that accumulation of cohesin at the four mCherry chimera occurred even in the absence of URA3 302 induction. This feature is not unique to M phase-arrested cells. During late G1 arrest when 303 cohesin accumulates at the four mCherry chimera, loss of Wpl1 also eliminated the requirement 304 for URA3 induction ( Figure S6A). Misregulation of URA3 in the absence of Wpl1 could explain 305 the results. However, URA3 transcription was found to be normal ( Figure S5D and S6B). 306 Instead, it seems that basal transcription of URA3 is sufficient to cause cohesin accumulation at 307 synthetic obstacles in the absence of Wpl1. If cohesin binds less specifically in Wpl1 mutants 308 (53), and the residence time on DNA is longer (48), basal transcription of URA3 might be 309 sufficient to mobilize mis-targeted complexes toward downstream synthetic barriers. 310 311 When Eco1 and Wpl1 were co-depleted, cohesin bypassed the three mCherry chimera yet the 312 complex was blocked efficiently by the larger four mCherry chimera ( Figure 6F). Cohesin 313 accumulation did not require URA3 induction, as seen before when depleting Wpl1 alone. 314 Interestingly, the paradoxical accumulation of cohesin at GFP-lacI in the absence of Eco1 was 315 abolished when both proteins were co-depleted. Apparently, the impact of Wpl1 depletion is 316 dominant when both proteins are absent. Perhaps with a longer DNA residence time conferred by 317 Wpl1 loss, cohesin eventually bypasses barriers with fewer than four mCherries. In this study we show that synthetic barriers of sufficient size cause ectopic accumulation of the 334 cohesin at synthetic roadblocks downstream of an inducible gene (Figure 2). That a DNA-bound 335 obstacle built entirely of non-yeast proteins blocks the passage of cohesin shows that the 336 complex translocates processively along DNA inside live cells. While our work was underway, a 337 study using nuclease-dead Cas9 as a barrier also showed that cohesin moves processively in vivo 338 (55). Our study goes further in three significant ways. First, we show that both cohesive and non-339 cohesive complexes translocate processively on chromosomes. Second, we show that the two 340 types of complexes are blocked by barriers of different size ( Figure 7A). Third, we show that 341 stalled cohesive complexes block passage of non-cohesive complexes ( Figure 7B). 342

343
Cohesive cohesin has been shown to entrap both sister chromatids topologically within a single 344 complex (3, 4). Presumably, the cohesive cohesin that translocates in our studies is similarly 345 bound to DNA. Crosslinking studies have defined one large DNA chamber formed by the 346 interactions of core subunits (Smc1, Smc3 and Mcd1), as well as two smaller DNA chambers 347 partitioned from the larger chamber by the engaged heads of the Smc proteins ( Figure 7C, (56)). 348 These and additional cryo-EM studies have mapped cohesin within the chambers of cohesin (57, 349 58). Synthetic obstacles of increasing size serve as a molecular ruler of the embrace of cohesin 350 through one or more of these chambers. Our work defines the upper size limit (measured in 351 mCherries) for cohesin's embrace of sister chromatids. 352 353 Non-cohesive cohesin can also bind single chromatids in a topological embrace but whether the 354 non-cohesive complexes in our studies are similarly bound is less clear (3,4,59). Biochemical 355 studies of cohesin translocation on single pieces of DNA have described active and passive 356 movements. In the active case, cohesin moved linearly along DNA and extruded DNA loops in 357 an ATP-dependent manner, the latter of which involved major conformational changes of the 358 complex (27,28,60). This type of movement required the continual action of NIPBL/Scc2, a 359 cofactor that stimulates the intrinsic cohesin ATPase activity (61). In congruence, loop extrusion 360 in vivo, as well as less defined modes of transport, have been linked to NIPBL/Scc2 (33,62,63). 361 For the following reasons, cohesin transport driven by NIPBL/Scc2 does not appear to involve 362 topological entrapment of DNA. To begin with, loop extrusion in vitro was observed with 363 crosslinked complexes that prevented DNA access to the central chamber (27) The threshold size of an impassible barrier changed depending on when cohesin was loaded onto 379 chromosomes and when it translocated. Complexes loaded and transiting before S phase were 380 blocked by a four mCherry obstacle ( Figure 4). Non-cohesive complexes loaded after S phase 381 were similarly blocked by a four mCherry barrier ( Figure 5). Cohesive complexes, those which 382 were loaded before S phase but translocated after, were blocked instead by a three mCherry 383 barrier (Figures 2 and 3). Collectively, these observations suggest that the cohesive state of a 384 cohesin complex is the determinant of which mCherry barrier will block translocation. In this 385 scenario, the requirement of larger barriers to block non-cohesive complexes may simply reflect 386 the fact that these complexes entrap one less DNA duplex. In other words, non-cohesive 387 complexes translocate past smaller barriers because their DNA chambers are filled with less 388 DNA. 389 390 Both Eco1 and Wpl1 influence the size and distribution of chromosome loops in yeast, as do 391 their homologs in higher eukaryotes (8,9,27,32,33,65,66). Whereas Wpl1 restricts loop size 392 by limiting the residence time and location of cohesin on DNA, Eco1 is thought to restrict loop 393 size through a mechanism(s) independent of cohesion establishment. A simple hypothesis holds 394 that conditions which favor more extensive loop extrusion would also favor translocation past 395 smaller synthetic barriers. Indeed, simultaneous depletion of Eco1 and Wpl1 or just Wpl1 alone 396 increased the threshold size of a barrier in M phase cells from three mCherries to four mCherries 397 ( Figure 6E and 6F). However, we showed that the locus under study was not cohesed in the 398 absence of Eco1 and Wpl1 ( Figure 6G). Thus, a simpler explanation may prevail: the increase in 399 barrier size from three to four mCherries reflects the absence of cohesive complexes in these 400 mutants. This conclusion is entirely congruent with the cell cycle constraints described above.

Yeast strains and plasmids 427
Strains generated for this study were confirmed by PCR and/or DNA sequencing. Strains are 428 listed in Table S2 and primers are listed in Table S1. GFP-lacI with a linked SV40 NLS was 429 expressed from either a single or tandem integration of plasmid pGVH60, as noted in the strain 430 table. These integrants were modified by PCR-mediated gene tagging using multimeric repeats 431 of mCherry from plasmids pMAM12 and pMAM44 or three yeGFP(A206R)s from plasmid 432 pMAM88. Strains with tandem pGVH60 thus yielded simultaneous expression a GFP-lacI 433 chimera and GFP-lacI alone. Strains co-expressing GFP-lacI/GFP-lacI-mCherry chimeras, 434 however, produced identical results to strains expressing the mCherry chimeras alone (compare 435 figures 4C and S7). Integrating plasmid pXDA2-FFAT was assembled by in vivo recombination 436 from pAFS144-FFAT and pXRA2 (69). pXDL2-GAL1p-Sic1(4m)-AID*-9xmyc was similarly 437 assembled from pXRL2 (69) To arrest cells in M phase for ChIP-qPCR, methionine was added (Cf = 2mM) to mid-log 448 cultures grown in SC-met to shut-off MET3p-cdc20. After 2.5 hours (roughly 80% of the cells 449 adopted a dumbbell shape), cells were washed twice with water and resuspended in SC media 450 lacking uracil to induce URA3. Two hours later the cultures were fixed for ChIP, according to 451 (35). To arrest cells in M phase for analysis of cohesion, cultures were grown in SC-met plus 452 dextrose for 6-8 hours and then diluted 200-400 fold into SC-met plus raffinose for overnight 453 growth. When cells reached mid-log the following morning, methionine was added. 150 minutes 454 later, cells were washed twice and resuspended in SC media lacking uracil. Two hours later, 455 galactose was added (Cf = 2%) to induce circularization by the R recombinase. Cells were fixed 456 with paraformaldehyde after two additional hours. To arrest cells for ChIP in late G1, dextrose 457 grown cultures were diluted 200-400 fold into SC-met plus raffinose for overnight growth. When 458 the cultures reached mid-log the following morning, alpha factor was added (Cf = 10 µM), and 459 90 minutes later galactose was added to induce expression of GAL1p-SIC1(4m)-AID*-9xmyc. 460 After one additional hour, cultures were washed twice with water and resuspended in SC-met 461 plus galactose and pronase E (Cf = 0.1 mg/ml) to release from pheromone arrest. One hour later, 462 the cells were washed twice and resuspended in galactose lacking both methionine and uracil. 463 Two hours later, the cultures were fixed for ChIP. For release from late G1 arrest, cells were 464 grown into the arrest as described above but with the following exceptions. After cells were 465 washed to remove alpha factor, they were resuspended in galactose media that lacked uracil but 466 contained methionine and pronase E. Two hours later, dextrose (Cf = 2%) was added to stop