ISSN 0006-2979, Biochemistry (Moscow), 2024, Vol. 89, No. 4, pp. 601-625 © Pleiades Publishing, Ltd., 2024.
601
REVIEW
Cohesin-Dependent Loop Extrusion:
Molecular Mechanics and Role in Cell Physiology
Arkadiy K. Golov
1,2,a
* and Alexey A. Gavrilov
1,b
*
1
Institute of Gene Biology, Russian Academy of Sciences, 119334 Moscow, Russia
2
Technion – Israel Institute of Technology, 3525433 Haifa, Israel
a
e-mail: golovstein@gmail.com 
b
e-mail: aleksey.a.gavrilov@gmail.com
Received October 21, 2023
Revised December 29, 2023
Accepted February 15, 2024
AbstractThe most prominent representatives of multisubunit SMC complexes, cohesin and condensin, are best
known as structural components of mitotic chromosomes. It turned out that these complexes, as well as their bac-
terial homologues, are molecular motors, the ATP-dependent movement of these complexes along DNA threads
leads to the formation of DNA loops. In recent years, we have witnessed an avalanche-like accumulation of data on
the process of SMC dependent DNA looping, also known as loop extrusion. This review briefly summarizes the cur-
rent understanding of the place and role of cohesin-dependent extrusion in cell physiology and presents a number
of models describing the potential molecular mechanism of extrusion in a most compelling way. We conclude the
review with a discussion of how the capacity of cohesin to extrude DNA loops may be mechanistically linked to its
involvement in sister chromatid cohesion.
DOI: 10.1134/S0006297924040023
Keywords: cohesin, SMC complexes, loop extrusion, cohesion, DNA gripping state
Abbreviations: CAR, cohesin associated region; E-P, en-
hancer-promoter (interactions); FRET, Förster resonance en-
ergy transfer; HAWK,HEAT protein associated with Kleisin;
SMC,structural maintenance of chromosomes.
* To whom correspondence should be addressed.
INTRODUCTION
Cohesin is a protein complex, which is absolute-
ly essential for reproduction of eukaryotic cells [1, 2].
First cohesin subunits were discovered more than
25years ago as factors participating in pairing of sis-
ter chromatids in mitosis [3-5]. It has been found out
later that this phenomenon termed ‘cohesion’ is based
on the fact that the pairs of sister chromosomes af-
ter replication end up being threaded through multi-
ple cohesion complexes (each of which having ring-
shaped structure with a relatively large intersubunit
pore), similar to two threads passing through a se-
ries of beads (Fig. 1a) [6-8]. Maintenance of cohesion
during the G2-phase of cell cycle and its controlled re-
lease in anaphase occurring due to proteolysis of the
RAD21 subunit of cohesin ensures correct attachment
of spindle microtubules to kinetochores and subse-
quent equal distribution of genetic material between
the two daughter cells [9].
No less important activity of cohesin besides co-
hesion is its ability to form DNA loops via the mecha-
nism called extrusion [2, 10, 11]. Extrusion begins with
cohesin binding to small DNA fragment followed by
the ATP-dependent movement of the complex along
the DNA resulting in processive pulling of the flank-
ing DNA inside the growing loop, the length of which
could eventually reach hundreds of thousands of base
pairs (kbp). Theoretically flanking DNA can be contin-
uously pulled inside from one side of the loop held by
the complex (in the case of unidirectional extrusion) or
from both sides (in the case of bidirectional extrusion).
Extrusion is typical not only for cohesin, but for an en-
tire group of protein complexes, known as SMC com-
plexes (structural maintenance of chromosomes pro-
teins), with cohesin being one of the representatives
of this group [12-14]. Synergistic activity of the typeII
DNA topoisomerases and SMC-dependent extrusion is
required for post-replicative individualization of sis-
ter genomes in all cells (prokaryotic and eukaryotic).
GOLOV, GAVRILOV602
BIOCHEMISTRY (Moscow) Vol. 89 No. 4 2024
Fig. 1. Cohesin structure and its participation in sister chromatid cohesion. a)Topological loading of cohesin rings onto sis-
ter chromatids during cohesion. b)Schematic representation of core trimer forming the cohesin ring; recruitment of auxiliary
HAWK subunits to the core trimer. c)Dimerization of head domains of SMC subunits required for ATP hydrolysis.
Extrusion can also facilitate orderly compaction of DNA;
formation of mitotic chromatids in prophase/metaphase
in the vertebrate cells is the best example of such com-
paction. Chromatids are elongated structures consist-
ing of the densely packed chromatin loops anchored
on the proteinaceous axial core. The main structural
component of this axis is condensin– SMC complex re-
sponsible for mitotic loop extrusion, which eventually
results in formation of chromatids. The cohesin-depen-
dent loop extrusion contributes to individualization of
eukaryotic chromosomes, in many cases is responsible
for chromatin compaction and also has a number of
secondary functions.
Cohesion, unlike extrusion, is typical exclusively
for cohesin, and not for any other representatives of
SMC complexes[1,2]. Cohesive activity likely emerged
in the primordial cohesin during early eukaryogen-
esis [15], this acquisition did not result, however, in
the loss of extrusive activity of the complex [16-18].
This commonly accepted scenario raises a number of
interesting questions on mechanistic, functional, and
evolutionary interrelationships between the cohesion
phenomenon and the process of loop extrusion.
In this review we summarize current under-
standing of cohesin-dependent loop extrusion, its role
in cellular processes, and its molecular mechanisms.
We also briefly discuss disparate data indicating pos-
sible mechanistic relationship between extrusion and
cohesion. Detailed description of the cohesin complex
structure and principles of its interactions with chro-
matin can be found in the first part of the review pub-
lished in the same issue of the journal [19].
STRUCTURE OF COHESIN COMPLEX
Cohesin is a protein complex with a ring-shaped
structure, which is based on the trimer of core pro-
teins: SMC1 (Smc1)
1
, SMC3 (Smc3), and RAD21 (Scc1).
All three proteins have elongated shape and interact
with each other through terminal globular domains
(Fig. 1b). Such organization results in formation of
the extensive intersubunit pore, which is able to let
through globular particles with diameter up to around
10 nm [20, 21]. Presence of the intersubunit pore makes
possible topological entrapment of DNA within the
complex with DNA being threaded through the protein
ring [6-8].
SMC1 and SMC3 subunits are paralogs belong-
ing to the family of ATPases called SMC proteins [22].
SMC1 and SMC3 form a stable V-shaped heterodimer
via homotypic interaction between the hinge domains
of two subunits [20,23]. Head domains, which are lo-
cated at the opposite end of the rod-shaped molecule
from the hinge domains, are responsible for ATPase
activity. In the presence of ATP intermittent engage-
ment of the head domains of two SMC subunits occurs,
such dimerization is required for hydrolysis of bound
1
In the paper names of human proteins are presented in the main text; names of Saccharomyces cerevisiae homologs are
shown in parenthesis (at first mention).
COHESIN-DEPENDENT EXTRUSION 603
BIOCHEMISTRY (Moscow) Vol. 89 No. 4 2024
ATP molecules (Fig. 1c). The hinge and head domains
of each of the SMC subunits are separated by the long
and relatively flexible coiled-coil arm domain. Flexibil-
ity of the arm domain is to a large degree associated
with the presence of evolutionary conserved defect in
the regular coiled-coil structure, the so-called elbow
region (Fig. 1b). Bending of ‘elbows’ has rather large
amplitude; and simultaneous bending of elbow sites
in both SMC subunits could facilitate direct physical
interaction of the hinge and head domains of the com-
plex. The RAD21 protein, also called kleisin subunit,
forms a constant bridge between the head domains of
the two SMC subunits, thus closing the ring structure.
Auxiliary subunits belonging to the family of
HAWK proteins (HEAT protein associated with Klei-
sin) including STAG1/2
2
(Scc3), NIPBL (Scc2), and
PDS5A/B (Pds5) bind to the core trimer [16, 20, 23].
Theprimary site of interaction of the HAWK subunits
with the core trimer is the kleisin subunit, however,
only STAG1/2 forms a stable contact with RAD21 and,
hence, is a constitutive component of the complex.
NIPBL and PDS5A/B compete for the shared binding
site at the kleisin subunit, and both these proteins
interact transiently with the stable cohesin tetramer
(SMC3–SMC1–RAD21–STAG1/2) [24-26]. Therefore, at
each particular moment the complex could contain
one (STAG1/2) or two (STAG1/2 + NIPBL or STAG1/2 +
+ PDS5A/B) HAWK subunits(Fig.1b). Binding of NIPBL
to the core cohesin complex (and presence of DNA)
isrequired for effective hydrolysis of ATP, replacement
of NIPBL with PDS5A/B dramatically changes activity
of the complex [26].
CELLULAR CONTEXT
OF COHESIN-DEPENDENT LOOP EXTRUSION
Cohesin-dependent loop extrusion throughout
the cell cycle. Cohesin-dependent extrusion is realized
in eukaryotic cells throughout interphase and mitosis
(Fig.2a). Various estimates indicate that in the G1-phase
vertebrate cells there are around 100,000 cohesin rings,
and this number doubles in the G2-phase [27, 28].
Nucleoplasm of the cells in G1-phase contains two co-
existing subpopulations of cohesin complexes of ap-
proximately the same size, which are in dynamic equi-
librium: (1) freely diffusing cohesin and (2) cohesin
involved in extrusion (with chromatin residence time of
about 10-30min in vertebrate cells and around 1min
in yeast cells) [28-31]. When cohesion is established
in the S-phase a third subpopulation emerges: stably
bound cohesive rings excluded from extrusion initia-
tion/termination cycle [28, 29, 32]. Complete arrest of
cohesin-dependent extrusion occurs during mitosis. In
vertebrates phosphorylation of the HAWK subunits in
prophase induces termination of cohesion in chromo-
some arms; most likely this phosphorylation also leads
to dissociation of extruding cohesin from DNA [29,33].
Resumption of cohesin-dependent extrusion occurs at
the end of telophase [34,35]. In the budding yeasts the
cohesin-dependent extrusion is realized in mitotic cells
up to the start of anaphase [36], when the separase-de-
pendent proteolysis of the kleisin subunits results in
degradation of the whole cellular pool of cohesin com-
plexes and is resumed only at the end of G1-phase with
restoration of the population of intact cohesin rings
[4,36].
Genomic distribution of initiation sites. Initi-
ation of the cohesin-dependent loop extrusion is not
strictly localized to particular genomic sites, howev-
er, there are some preferences in extrusion complex
binding to DNA. Centromeres and regions adjacent to
them are examples of such sites [8, 26]. Initiation of ex-
trusion, however, also occurs constantly outside of the
centromere regions. It has been assumed up until now
that initiation of the cohesin-dependent extrusion in
the chromosome arms is mainly associated with open
chromatin [37-39]. However, this model is questioned
now. The recently reported experimental data on the
genomic distribution of cohesin binding, as well as the
results of computer modeling indicate that probability
of extrusion initiation is distributed evenly along the
genome outside of the centromeres [40].
WAPL and PDS5A/B are negative regulators of
processivity, NIPBL – positive. The final size of the
loops formed in the process of extrusion is determined
by processivity of the cohesin complexes and genom-
ic distribution of the extrusion pause sites [36, 41-45],
itvaries from several dozens of kb in yeasts [44, 46] to
hundreds of kb in vertebrates[41, 42].
Processivity of the cohesin-dependent extrusion is
suppressed by the HAWK subunit PDS5A/B, as well as
by the WAPL protein recruited by this subunit. Deple-
tion of PDS5A/B and WAPL (together or individually)
significantly increases chromatin residence time of
the extruding cohesin, as well as length of the form-
ing DNA loops [36, 41, 42, 44,45]. Interestingly enough,
both these proteins also participate in non-proteolyt-
ic termination of cohesion as a part of the so-called
prophase cascade: recruitment of the WAPL protein
to cohesin complexes containing PDS5A/B subunit re-
sults in non-proteolytic opening of the protein rings
in the prophase cells of vertebrates and their removal
from the chromosomal arms [47,48]. This dual activity
2
Vertebrate genomes generally encode a pair of somatically expressed paralogs for both Scc3 and Pds5 HAWK proteins:
STAG1/STAG2 and PDS5A/PDS5B. These paralogs in the majority of cases are structurally and functionally equivalent, hence,
here and further in the text the designations STAG1/2 and PDS5A/B are used.
GOLOV, GAVRILOV604
BIOCHEMISTRY (Moscow) Vol. 89 No. 4 2024
Fig. 2. Cohesin activity throughout the cell cycle and chromatin compaction due to SMC-dependent loop extrusion. a)Intact
cohesin complex quantity and their activity throughout mitotic cycle in vertebrate(1) and S. cerevisiae(2) cells. b)Metaphase
chromosomes of vertebrates are formed due to extrusion activity of condensin complexes eventually accumulated in axial struc-
tures. Typical X-shaped structure is maintained due to residual cohesion in centromeres of two metacentric sister chromosomes.
c)Compact chromatid-like “vermicelli” structures formed due to cohesin-dependent loop extrusion in interphase vertebrate
cells with suppressed WAPL activity. Cohesin is the primary structural component of the axial structures in such chromosomes.
ofPDS5A/B and WAPL implies existence of mechanis-
tic relatedness between the process of loop extrusion
and the phenomenon of cohesion.
Another HAWK subunit, NIPBL, on the contrary,
is a positive regulator of the extrusion processivity.
NIPBL is commonly considered as a cohesin loader;
however, this point of view probably requires recon-
sideration as the new data have been reported demon-
strating that NIPBL could be dispensable for the pri-
mary loading of cohesin onto chromatin [49]. It has
been firmly established that recruitment of NIPBL to
the complex is necessary for active extrusion: func-
tional depletion of NIPBL results in suppression of co-
hesin-dependent loop interactions in vertebrates and
inhibition of the cohesin translocation from the pri-
mary loading sites in yeasts [26, 42, 50]. During each
round of extrusion initiation/termination cycle NIPBL
is repeatedly recruited to and released from the com-
plex: chromatin residence time of NIPBL subunit in
the vertebrate G1-cells is around 1 min, which is an
order of magnitude less than average duration of each
round of extrusion [25, 31]. During NIPBL absence its
binding site can be occupied by the PDS5A/B subunit
recruiting WAPL, which, likely, leads to extrusion ter-
mination by not yet elucidated mechanism [2].
Cohesin-dependent interphase extrusion, unlike
mitotic condensin-dependent extrusion, usually does
not result in formation of condensed chromatid-like
structures with proteinaceous axial cores to which
DNA loops are anchored. This can be a consequence
of low processivity of the cohesin-dependent extru-
sion. Suppression of activity of PDS5A/B and WAPL in
the vertebrate cells results in interphase condensation
of chromatin accompanied by the formation of micro-
scopically visible elongated structures with axial cores
containing cohesin [41, 42, 51]. Such compact structures
with characteristic shape resembling metaphase chro-
matids (Fig.2b) have been called ‘vermicelli’(Fig.2c).
Interphase ‘vermicelli’ in a structural sense is simi-
lar to chromatids formed in the meiosisI prophase[52].
Compaction of meiotic chromosomes is achieved by the
extruding activity of cohesin that eventually accumu-
COHESIN-DEPENDENT EXTRUSION 605
BIOCHEMISTRY (Moscow) Vol. 89 No. 4 2024
lates in the axial structures, which later become an im-
portant component of synaptonemal complex. Discov-
ery of cohesin-mediated formation of the condensed
structures similar to metaphase chromatids in the
meiosis I prophase became one of the early indica-
tions that extrusion of DNA loops might be a universal
activity of all SMC complexes [51,53].
Site-specific arrest of loop extrusion. Cohesin-
dependent extrusion differs from extrusion mediated
by other SMC complexes by the ability for a regulat-
ed arrest at the specific genome loci. It is not known
exactly whether such arrest is a temporary pause or
final termination of extrusion. In any case, stability of
the formed DNA loop is maintained by the extrusion
complex for a certain time after arrest until dissoci-
ation of the complex from DNA [31, 54]. At least two
mechanisms of site-specific arrest of extrusion cohes-
in complexes have been described: CAR-dependent
(Cohesin Associated Region) and CTCF-dependent.
Thefirst one, which is evolutionary more ancient and
typical for the cells of lower eukaryotes, is associated
with extended genome regions of cohesin accumula-
tion, CAR-regions [44, 46, 55]. CAR-regions are prefer-
ably immunoprecipitated by the antibodies against
cohesin subunits and, as a rule, are located at the
3′-ends of the convergently transcribed genes [56, 57].
The second mechanism of extrusion arrest associated
with activity of the insulator protein CTCF is realized
in the vertebrate cells [58, 59]. Blocking of extrusion
complexes at the CTCF-sites results in colocalization of
the vast majority of strong cohesin binding sites with
the CTCF binding sites in the vertebrate cells.
CAR-dependent and CTCF-dependent mechanisms
of extrusion arrest have much in common both
with each other and with the processes of stabiliza-
tion of cohesive ring binding to chromatin (Fig. 3a).
In both cases arrest is realized due to the fact that
the blocking sites favor replacement of NIPBL with
PDS5A/B[44, 58]. This replacement is facilitated by the
ESCO1(Eco1)-dependent acetylation of the SMC3 pro-
tein occurring at the arrest sites [43, 45]. Acetylation
of the SMC3 subunit at the conserved lysine residues
K105/K106 (K112 andK113– in yeasts) suppresses ex-
trusion activity of cohesin likely due to decrease in
affinity of the complex to NIPBL[60,61]. Additionally,
acetylation inhibits activity of the WAPL protein [62],
which, as has been mentioned above, removes from
chromatin not only the topologically engaged cohes-
in complexes, but also complexes participating in ex-
trusion [41, 42, 51]. Interestingly enough, stabilization
of the cohesive rings on chromatin in the G2-phase
occurs due to acetylation of SMC3 at the same amino
acid residues (in many species an auxiliary protein so-
rorin also participates in stabilization). Acetylation fa-
cilitating stabilization of cohesion is established in the
S-phase co-replicatively due to activity of the ESCO1
paralog, acetyltransferase ESCO2 (Eco1), and it also
suppresses activity of WAPL towards the acetylated
complexes.
Existence of the genomic elements blocking extru-
sion results in accumulation of cohesin complexes in-
teracting with CAR-elements and CTCF-sites, as well as
in formation of interphase chromatin typical folding
patterns: structural chromatic loops anchored in the
mentioned genomic elements and topologically-associ-
ated domains (TADs) (Fig.3,b andc) [44, 46, 58, 59,63].
In vertebrates the N-terminal fragment of CTCF is re-
sponsible for the arrest of extruding cohesin complex-
es, as well as for inhibition of the activity of WAPL sub-
unit [58, 59,64]. Steric characteristics of CTCF molecule
bound to the binding site allow effective interaction
of its N-terminal fragment with cohesin only if the
extrusion complex approaches the CTCF-motif from
the side of the 3′-end (N-terminal zinc-fingers of CTCF
bind there to DNA); this interaction blocks extrusion
and stabilizes cohesin at the CTCF-bound site (Fig.3a).
At the same time when extrusion complex approach-
es CTCF-motif from the side of the 5′-end it does not
cause prolonged stalling of cohesin. These features of
protein–protein interactions result in the peculiar reg-
ularity in orientation of the CTCF-motifs located at the
bases of structural loops and at the boundaries of to-
pological domains in vertebrates: pairs of CTCF-motifs
located at the bases of structural loops are general-
ly oriented in such a way that their 3′-ends face each
other (convergent orientation); at the same time sev-
eral CTCF-motifs (at least a pair) that are typically
present within each TAD boundary face interior of the
closest topological domain with their 3′-ends (divergent
orientation) (Fig.3,b andc)[63,65-67].
Interplay with transcriptional apparatus of
the cell. Why CAR-sites that stop extrusion in the
yeast cells are located at 3′-ends of the convergently
transcribed genes? Theoretically this could be a con-
sequence of the fact that collision of the transcribing
polymerase with extruding cohesin complexes moving
in the opposite direction results in translocation of
the latter to the end of the transcriptional units. This
model is in agreement with the fact that turning off
transcription leads to the removal of cohesin from the
CAR-sites[56].
The mechanisms underlying transcription-depen-
dent accumulation of cohesin rings at the 3′-end of the
genes are operating not only in yeast cells, they are
universal. For example, despite the fact that the verte-
brate cells do not have classic CAR-sites, simultaneous
depletion of the CTCF and WAPL results in relocation
of cohesin complexes from the CTCF-sites to the 3′-ends
of actively transcribing genes [38, 40]. These new co-
hesin-bound sites differ significantly from the original
genomic peaks colocalized with CTCF; they comprise
extended regions several kbp in length appropriately