mechanisms in the context of human tumors
is unclear and requires further study.
Mechanism of action of CDK4/6 inhibitors
Three small-molecule CDK4/6 inhibitors have
been extensively characterized in preclinical
studies: palbociclib and ribociclib, which are
highly specific CDK4/6 inhibitors, and abe-
maciclib, which inhibits CDK4/6 and other
kinases (Table 1). It has been assumed that
these compounds act in vivo by directly in-
hibiting cyclin D–CDK4/6 ( 9 ). This simple
model has been recently questioned by ob-
servations that palbociclib inhibits only cyclin
D–CDK4/6 dimers, but not trimeric cyclin D–
CDK4/6-p27KIP1( 44 ). However, it is unlikely
that substantial amounts of cyclin D–CDK4
dimers ever exist in cells, because nearly all
cyclin D–CDK4 in vivo is thought to be com-
plexed with KIP/CIP proteins ( 11 , 14 , 44 ).
Palbociclib also binds monomeric CDK4 ( 44 ).
Surprisingly, treatment of cancer cells with
palbociclib for 48 hours failed to inhibit CDK4
kinase, despite cell cycle arrest, but it inhibited
CDK2 ( 44 ). Hence, palbociclib might prevent
the formation of active CDK4-containing com-
plexes (through binding to CDK4) and in-
directly inhibit CDK2 by liberating KIP/CIP
inhibitors. This model needs to be reconciled
with several observations. First, treatment
of cells with CDK4/6 inhibitors results in
a rapid decrease of RB1 phosphorylation on
cyclin D–CDK4/6-dependent sites, indicating
an acute inhibition of CDK4/6 ( 45 – 47 ). More-
over, CDK4/6 immunoprecipitated from cells
can be inhibited by palbociclib ( 48 )andp21CIP-
associated cyclin CDK4/6 kinase is also in-
hibited by treatment of cells with palbociclib
( 49 ). Lastly, CDK2 is dispensable for prolifer-
ation of several cancer cell lines ( 50 , 51 ), hence
the indirect inhibition of CDK2 alone is un-
likely to be responsible for cell cycle arrest.
Palbociclib, ribociclib, and abemaciclib were
shown to block binding of CDK4 and CDK6 to
CDC37, the kinase-targeting subunit of HSP90,
thereby preventing access of CDK4/6 to the
HSP90-chaperone system ( 52 ). Because the
HSP90-CDC37 complex stabilizes several ki-
nases ( 53 ), these observations suggest that
CDK4/6 inhibitors, by disrupting the inter-
action between CDC37 and CDK4 or CDK6,
might promote degradation of CDK4 and
CDK6. However, depletion of CDK4/6 is typi-
cally not observed upon treatment with CDK4/6
inhibitors ( 54 ). More studies are needed to
resolve these conflicting reports and to es-
tablish how CDK4/6 inhibitors affect the cell
cycle machinery in cancer cells.
Validation of CDK4/6 inhibitors as
anticancer agents
Consistent with the notion that RB1 represents
the major rate-limiting substrate of cyclin D–
CDK4/6 in cell cycle progression ( 55 – 57 ),
palbociclib, ribociclib, and abemaciclib were
shown to block proliferation of several RB1-
positive cancer cell lines, but not cell lines that
have lost RB1 expression ( 46 , 58 , 59 ). Breast
cancer cell lines representing the luminal,
estrogen receptor–positive (ER+) subtype were
shown to be most susceptible to cell prolifer-
ation arrest upon palbociclib treatment ( 45 ).
Palbociclib, ribociclib, abemaciclib, and another
CDK4/6 inhibitor, lerociclib, were demon-
strated to display potent antitumor activity in
xenografts of several tumor types, including
breast cancers ( 46 , 60 – 62 ). Palbociclib and
abemaciclib cross the blood-brain barrier and
inhibit growth of intracranial glioblastoma
(GBM) xenografts, with abemaciclib being
more efficient in reaching the brain ( 63 , 64 ).
Recently, additional CDK4/6 inhibitors were
shown to exert therapeutic effects in mouse
xenograft models of various cancer types,
including SHR6390 ( 65 ), FCN-437 ( 66 ), and
compound 11 ( 67 ); the latter two were re-
ported to cross the blood-brain barrier. In
most in vivo studies, the therapeutic effect
was dependent on expression of intact RB1
protein in tumor cells ( 46 , 63 ). However, anti-
tumor effects of palbociclib were also reported
in bladder cancer xenografts independently of
RB1 status; this was attributed to decreased
phosphorylation of FOXM1 ( 68 ).Tumor cell senescence upon
CDK4/6 inhibition
In addition to blocking cell proliferation, in-
hibition of CDK4/6 can also trigger tumor cell
senescence ( 63 ), which depends on RB1 and
FOXM1 ( 35 , 54 ). The role of RB1 in enforcing
cellular senescence is well established ( 69 ). In
addition, cyclin D–CDK4/6 phosphorylates and
activates FOXM1, which has anti-senescence
activity ( 35 , 70 ). Senescence represents a pre-
ferred therapeutic outcome to cell cycle arrest,
as it may lead to a durable inhibition of tumor
growth.
It is not clear what determines the extent
of senescence upon treatment of cancer cells
with CDK4/6 inhibitors. A recent study showed
that inhibition of CDK4/6 leads to an RB1-
dependent increase in reactive oxygen spe-
cies (ROS) levels, resulting in activation of
autophagy, which mitigates the senescence
of breast cancer cells in vitro and in vivo ( 71 ).
Co-treatment with palbociclib plus autoph-
agy inhibitors strongly augmented the abil-
ity of CDK4/6 inhibitors to induce tumor cell
senescence and led to sustained inhibition of
cancer cell proliferation in vitro and of xeno-
graft growth in vivo ( 71 ). Decreased mTOR
signaling after long-term CDK4/6 inhibition
was shown to be essential for the induction
of senescence in melanoma cells, and activa-
tion of mTORC1 overrode palbociclib-induced
senescence ( 72 ). Others postulated that expres-
sion of the chromatin-remodeling enzymeATRX and degradation of MDM2 determines
the choice between quiescence and senescence
upon CDK4/6 inhibition ( 73 ). Inhibition of
CDK4 causes dissociation of the deubiqui-
tinase HAUSP/USP7 from MDM2, thereby
driving autoubiquitination and proteolytic
degradation of MDM2, which in turn promotes
senescence. This mechanism requires ATRX,
which suggests that expression of ATRX can
be used to predict the senescence response ( 73 ).
Two additional proteins that play a role in
this process are PDLIM7 and type II cadherin
CDH18. Expression of CDH18 correlated with
a sustained response to palbociclib in a phase 2
trial for patients with liposarcoma ( 74 ).Markers predicting response to
CDK4/6 inhibition
Only tumors with intact RB1 respond to CDK4/6
inhibitor treatment by undergoing cell cycle
arrest or senescence ( 9 , 58 ). In addition,“D-
cyclin activating features”(CCND1transloca-
tion,CCND2orCCND3amplification, loss of
theCCND1-3 3 ′-untranslated region, and dele-
tion ofFBXO31encoding an F-box protein im-
plicated in cyclin D1 degradation) were shown
to confer a strong response to abemaciclib in
cancer cell lines ( 58 ). Moreover, co-deletion of
CDKN2AandCDKN2C(encoding p16INK4A/
p19ARFand p18INK4C, respectively) confers
palbociclib sensitivity in glioblastoma ( 75 ).
Thr^172 phosphorylation of CDK4 and Tyr^88
phosphorylation of p27KIP1(both associated
with active cyclin D–CDK4) correlate with
sensitivity of breast cancer cell lines or tumor
explants to palbociclib ( 76 , 77 ). Surprisingly, in
PALOMA-1, PALOMA-2, and PALOMA-3 trials
( 78 – 80 ), and in another independent large-
scale study ( 81 ),CCND1gene amplification or
elevated levels of cyclin D1 mRNA or protein
were not predictive of palbociclib efficacy. Con-
versely, overexpression of CDK4, CDK6, or
cyclin E1 is associated with resistance of tumors
to CDK4/6 inhibitors (see below).Synergy of CDK4/6 inhibitors with
other compounds
Several preclinical studies have documented
the additive or synergistic effects of combining
CDK4/6 inhibitors with inhibitors of the re-
ceptor tyrosine kinases as well as phospho-
inositide 3-kinase (PI3K), RAF, or MEK (Table 2).
This synergism might be because these pathways
impinge on the cell cycle machinery through
cyclin D–CDK4/6 ( 82 – 86 ). In some cases, the
effect was seen in the presence of specific genetic
lesions, such asEGFR,BRAFV600E,KRAS, and
PIK3CAmutations ( 59 , 87 – 89 ) (Table 2). When
comparing different dosing regimens, contin-
uous treatment with a MEK inhibitor with
intermittent palbociclib resulted in more com-
plete tumor responses than other combination
schedules ( 90 ). Treatment with CDK4/6 in-
hibitors sensitized cancer cells to ionizingFasslet al.,Science 375 , eabc1495 (2022) 14 January 2022 4 of 19
RESEARCH | REVIEW