science.org SCIENCEILLUSTRATION: CHRIS GARCIA AND ERIC SMITHthey do not lead to regression, suggesting a
need for new strategies to inhibit JAKs. The
major limitation in informing new therapeu-
tic approaches has been sufficient insight
into JAK regulatory mechanisms, particu-
larly how the V617F mutation causes ligand-
independent JAK2 signaling.
JAK proteins, which bind to the cytoplas-
mic region of cytokine receptors, comprise
four domains: a FERM (band 4.1, ezrin, ra-
dixin, moesin) domain, a Src homology-2–
like (SH2L) domain, a pseudokinase domain
(PKD), and a carboxyl-terminal tyrosine ki-
nase domain (TKD). The first step in JAK
signaling is activation of the TKD, which
occurs through reciprocal transphosphory-
lation of two tyrosines in the TKD activa-
tion loop, mediated by cytokine-dependent
receptor dimerization. Once activated, JAKs
phosphorylate the cytokine receptor itself
and subsequently the signal transducer and
activator of transcription (STAT) proteins
that are recruited to the tyrosine-phosphor-
ylated receptors.
There have been considerable efforts to
determine the three-dimensional structure
of a full-length JAK. As often is the case, this
work initially went piece by piece, starting
with the crystal structure of the JAK3 TKD
in 2005 ( 5 ), followed by the JAK2 PKD in
2012 ( 6 ), and then the crystal structure of the
integrated FERM-SH2L domains in 2014 ( 7 ).
However, in the absence of a full-length JAK
structure, it was unclear how the various do-
mains cooperated to regulate JAK activity.
A crystal structure and molecular dynam-
ics–derived model of the PKD-TKD of TYK2
( 8 ) and JAK2 ( 9 ), respectively, revealed an
autoinhibitory interaction that rational-
ized activating mutations found in human
cancers, such as Arg^683 Gly (R683G) in the
PKD and Asp^873 Asn (D873N) in the TKD.
However, V617F (in the PKD) was conspicu-
ously absent from the PKD-TKD interface.
Several mutagenesis studies established
that the hyperactivity of V617F and other
mutants was quashed by additional mu-
tations in the PKD, such as Phe^595 Ala
(F595A) ( 10 ) or those that destabilized ad-
enosine triphosphate (ATP) binding to the
PKD ( 11 ). These studies suggested that the
PKD is directly involved in JAK2 dimeriza-
tion and that V617F enhances dimerization,
a hypothesis confirmed with single-mol-
ecule fluorescence studies in cells, which
demonstrated that expression of the JAK2-
V617F mutant caused a substantial increase
in the basal (no cytokine) level of dimerized
JAK2 on cytokine receptors ( 12 ).
What does the cryo–electron microscopy
(cryo-EM) structure of full-length JAK1
from Glassman et al. tell us? The structure
provides the mechanism by which V658F in
human JAK1 (equivalent to V617F in JAK2)
leads to cytokine-independent signaling. The
structure shows a parallel JAK1 homodimer
that is mediated solely by a PKD-PKD inter-
action, with F658 in the heart of the dimer
interface. The authors show that for wild-
type JAK1 (V658), the PKD-PKD interaction
would not be as snug as with F658, which
explains why normal signaling by wild-type
JAKs requires the assistance of cytokine-me-
diated receptor dimerization for activation.
The structure also shows how the PKD
interacts with the FERM-SH2L domains—
experimental information that was com-
pletely lacking. Notably, AlphaFold [arti-
ficial intelligence-based protein structure
prediction ( 13 )] predicted the correct
FERM-SH2L-PKD interaction. Now that
this prediction has been verified by the
JAK1 cryo-EM structure, there is a reliable
structural model for the autoinhibited form
of JAKs, based on the AlphaFold structure
prediction for TYK2.The JAK1 cryo-EM structure also depicts
a previously unknown interaction between
the TKD and the PKD with a limited inter-
face. It is conceivable that the addition of
a nanobody to promote receptor dimeriza-
tion could have influenced the positions of
the TKDs in the structure. The AlphaFold
predictions for full-length JAK1, JAK2, and
JAK3 feature an elongated structure, simi-
lar to the JAK1 cryo-EM structure, but with
a different PKD-TKD interaction. Further
structural and mutagenesis analyses will
be required to determine whether and how
the TKD interacts with the PKD to facilitate
various phosphorylation events.
The cryo-EM structure of JAK1 provides
the crucial missing piece in the JAK ac-
tivation process: the specific JAK dimer
configuration that triggers TKD transphos-
phorylation and downstream signaling. The
model for JAK activation that has emergedis as follows: JAK molecules bound to mo-
nomeric cytokine receptors undergo con-
formational equilibrium between a closed,
autoinhibited state and an open state in
which the PKD is accessible for homotypic
(JAK2-JAK2) or heterotypic (all JAKs) di-
merization, and the TKD is available to
serve as enzyme or substrate in a transphos-
phorylation event. For wild-type JAKs in the
absence of cytokine, equilibrium favors the
closed state, establishing a low basal kinase
activity. For cytokine-mediated receptor di-
merization, or for activating mutations such
as JAK2 V617F in the absence of cytokine,
the equilibrium is shifted toward the PKD-
mediated dimerized state and concomitant
TKD transphosphorylation.
Initial efforts to inhibit JAKs focused on
the ATP-binding pocket of the TKD, but
attention has recently shifted to the same
pocket in the PKD, and compounds that
bind to the TYK2 PKD are in phase 2 and
phase 3 clinical trials for psoriasis, Crohn’s
disease, and psoriatic arthritis. How these
PKD-targeted compounds actually inhibit
JAKs is not understood, but the cryo-EM
structure of dimeric JAK1 will provide im-
portant clues and enable other therapeutic
strategies for mitigating hyperactive sig-
naling by JAKs. More broadly, the study of
Glassman et al. sets the stage for structural
analyses of additional complexes—includ-
ing cytokine receptors, JAK heterodimers,
and STAT proteins—paving the way for a
comprehensive structure-based approach
to abrogate pathologic JAK-STAT signaling
in a spectrum of human diseases. jREFERENCES AND NOTES- A. Rubbert-Roth et al., N. Engl. J. Med. 383 , 1511 (2020).
- C. R. Glassman et al., Science 376 , 163 (2022).
- C. J a m e s et al., Nature 434 , 1144 (2005).
- S. Verstovsek et al., N. Engl. J. Med. 366 , 799 (2012).
- T. J. Boggon, Y. Li, P. W. Manley, M. J. Eck, Blood 106 , 996
(2005). - R. M. Bandaranayake et al., Nat. Struct. Mol. Biol. 19 , 754
(2012). - H. J. Wallweber, C. Tam, Y. Franke, M. A. Starovasnik, P. J.
Lupardus, Nat. Struct. Mol. Biol. 21 , 443 (2014). - P. J. Lupardus et al., Proc. Natl. Acad. Sci. U.S.A. 111 , 8025
(2014). - Y. Shan et al., Nat. Struct. Mol. Biol. 21 , 579 (2014).
- A. Dusa, C. Mouton, C. Pecquet, M. Herman, S. N.
Constantinescu, PLOS ONE 5 , e11157 (2010). - H. M. Hammarén et al., Proc. Natl. Acad. Sci. U.S.A. 112 ,
4642 (2015). - S. Wilmes et al., Science 367 , 643 (2020).
- J. Jumper et al., Nature 596 , 583 (2021).
ACKNOWLEDGMENTS
R.L.L. is on the supervisory board of Qiagen and is a scientific
advisor to Imago, Mission Bio, Zentalis, Ajax, Auron, Prelude,
C4 Therapeutics, and Isoplexis, for which he has received
equity. R.L.L. receives research support from and has con-
sulted for Celgene and Roche and has consulted for Incyte,
Janssen, Astellas, Morphosys, and Novartis. He has received
honoraria from Roche, Lilly, and Amgen for invited lectures
and from Gilead for grant reviews. S.R.H. is a cofounder and
scientific advisory board member of Ajax Therapeutics, for
which he has received equity.
10.1126/science.abo77 88INSIGHTS | PERSPECTIVES
Model of an activated Janus kinase dimer (blue;
pseudokinase domain highlighted in yellow)
interacting with a dimerized receptor (red) bound
to a ligand (orange).140 8 APRIL 2022 • VOL 376 ISSUE 6589