RESEARCH ARTICLE
Molecular basis of product recognition during
PIP5K-mediated production of PI(4,5)P2 with positive
feedback
Received for publication, February 20, 2024, and in revised form, July 20, 2024 Published, Papers in Press, August 3, 2024,
https://doi.org/10.1016/j.jbc.2024.107631

Benjamin R. Duewell1,2, Katherine A. Faris1,2, and Scott D. Hansen1,2,*
From the 1Department of Chemistry and Biochemistry, and 2Institute of Molecular Biology, University of Oregon, Eugene, Oregon,
USA

Reviewed by members of the JBC Editorial Board. Edited by Phillip A. Cole
The ability for cells to localize and activate peripheral
membrane-binding proteins is critical for signal transduction.
Ubiquitously important in these signaling processes are phos-
phatidylinositol phosphate (PIP) lipids, which are dynamically
phosphorylated by PIP lipid kinases on intracellular mem-
branes. Functioning primarily at the plasma membrane, phos-
phatidylinositol-4-phosphate 5-kinases (PIP5K) catalyzes the
phosphorylation of PI(4)P to generate most of the PI(4,5)P2

lipids found in eukaryotic plasma membranes. Recently, we
determined that PIP5K displays a positive feedback loop based
on membrane-mediated dimerization and cooperative binding
to its product, PI(4,5)P2. Here, we examine how two motifs
contribute to PI(4,5)P2 recognition to control membrane as-
sociation and catalysis of PIP5K. Using a combination of single
molecule TIRF microscopy and kinetic analysis of PI(4)P lipid
phosphorylation, we map the sequence of steps that allow
PIP5K to cooperatively engage PI(4,5)P2. We find that the
specificity loop regulates the rate of PIP5K membrane associ-
ation and helps orient the kinase to more effectively bind
PI(4,5)P2 lipids. After correctly orienting on the membrane,
PIP5K transitions to binding PI(4,5)P2 lipids near the active
site through a motif previously referred to as the substrate or
PIP-binding motif (PIPBM). The PIPBM has broad specificity
for anionic lipids and serves a role in regulating membrane
association in vitro and in vivo. Overall, our data supports a
two-step membrane-binding model where the specificity loop
and PIPBM act in concert to help PIP5K orient and produc-
tively engage anionic lipids to drive the positive feedback
during PI(4,5)P2 production.

Phosphatidylinositol phosphate (PIP) lipids are important
second messengers that function as molecular scaffolds for
numerous membrane-binding proteins. PIP lipids are
dynamically interconverted between seven different species by
specialized kinases and phosphatases (1). Research has
demonstrated that each PIP lipid species exists in spatially
distinct subcellular membranes and function to recruit specific
enzymes to their necessary location for proper signaling (2, 3).
* For correspondence: Scott D. Hansen, shansen5@uoregon.edu.

© 2024 THE AUTHORS. Published by Elsevier Inc on behalf of American Society for
BY license (http://creativecommons.org/licenses/by/4.0/).
The most abundant PIP lipid that mediates signaling at the
plasma membrane is phosphatidylinositol-4,5-bisphosphate
(PI(4,5)P2), comprising 2 to 5% of total plasma membrane
lipids (4–6). Unlike most PIP lipids, PI(4,5)P2 is omnipresent at
the plasma membrane and is critical for several constitutive
cellular functions, including cytoskeletal assembly (7, 8), ion
channel gating (9, 10), and endocytosis (11). Without a con-
stant and stable pool of PI(4,5)P2, these cellular signaling
events are hindered and cells adopt disease-like states (12).

Phosphatidylinositol-4-phosphate 5-kinase (PIP5K) family
of enzymes are responsible for creating the bulk of PI(4,5)P2
lipids in eukaryotic cells (13). PIP5K enzymes function by
binding their substrate, phosphatidylinositol-4-phosphate
(PI(4)P), then catalyzing the phosphorylation of the 5-
hydroxyl group on the inositol lipid head group in the pres-
ence of Mg(+2)-ATP (14, 15). This activity has been described
as specific to the PI(4)P substrate, although PIP5K can also
phosphorylate the 5-OH of other PIP lipid species with lower
efficiency (16–20). In mammals, PIP5K exists as three paralogs
(i.e. a, b, and g) that display tissue-specific expression patterns
(21–23). Localization studies in macrophages have shown that
PIP5K paralogs predominately localize to the plasma mem-
brane (24). PIP5K has been implicated in many cellular path-
ways, with specific regulatory importance in the Wnt pathway
(25), phagocytosis (26, 27), and focal adhesions (28, 29). More
recently, medical studies have begun to target PIP5K with
potential as a cancer therapeutic (30). Studies that have
knocked out PIP5K from both prostate and breast cancer
models showed diminished tumorigenesis and metastatic
properties (31). The drugs designed to inhibit PIP5KA activity
show promise reducing tumor growth and proliferation;
however, these drugs have low specificity for PIP5K (32). As
our understanding of PIP5K’s role in tumor proliferation
evolves, it is important that we continue to decipher the mo-
lecular mechanisms that regulate PIP5K membrane localiza-
tion and activity in cells.

Recently, our lab established a supported lipid bilayer (SLB)
and total internal reflection fluorescence microscopy (TIRF-
M) assay to directly visualize the membrane-binding dynamics
of purified PIP5K with single molecule resolution in vitro. We
previously reported that human PIP5KB binds cooperatively to
J. Biol. Chem. (2024) 300(9) 107631 1
Biochemistry and Molecular Biology. This is an open access article under the CC

https://doi.org/10.1016/j.jbc.2024.107631
Delta:1_given name
Delta:1_surname
Delta:1_given name
https://orcid.org/0000-0001-7005-6200
mailto:shansen5@uoregon.edu
http://crossmark.crossref.org/dialog/?doi=10.1016/j.jbc.2024.107631&domain=pdf
http://creativecommons.org/licenses/by/4.0/


Molecular basis of PIP5K product recognition
its product, PI(4,5)P2, which enhances PIP5K membrane
localization through a PI(4,5)P2-mediated positive feedback
loop (33). Positive feedback is based on an increased affinity of
PIP5K for PI(4,5)P2 compared to PI(4)P lipids (Fig. S1), leading
to a greater density of PIP5K on the membrane as PI(4,5)P2 is
generated. Further, we have shown that when PIP5K crosses a
threshold membrane surface density of � 100 molecules/mm2,
it can dimerize with another membrane bound kinase,
potentiating lipid kinase activity (34). Membrane-bound
PIP5K has an increased likelihood to encounter PI(4)P lipids
in a conformational state that is ideal for productive binding
and catalysis, leading to an increased apparent activity of
PIP5K enzymes. This suggests that the site of PI(4,5)P2 binding
is an important step for plasma membrane localization in cells,
as PI(4,5)P2 lipids exist almost exclusively at the plasma
membrane. Currently, the residues or motif(s) that regulates
high affinity PI(4,5)P2 binding remains unknown.

To date, much work has been done to define PIP5Ks
enzymatic activity, including identifying motifs that regulate
substrate specificity, ATP binding, and hydrolysis (17, 18, 35).
Structural biochemistry has helped elucidate the mechanism
controlling PIP5K substrate recognition (18, 35). X-ray crys-
tallography of zebrafish PIP5K (zPIP5K) determined that
PIP5K binds to its substrate, PI(4)P, through the DLKGSxxxR
motif, which is also referred to as the PI(4)P or substrate-
binding motif (18, 36). Mutations in the PI(4)P-binding
motif (PIPBM) of PIP5Ks abolishes lipid kinase activity (18).
PIP5K substrate specificity has been found to be mediated by a
structural motif referred to in the literature as either the
activation loop or specificity loop. Structural analysis of
zPIP5KA using NMR suggests that the specificity loop ex-
changes between a disordered and an alpha helical confor-
mation (35). Membrane–lipid interactions are hypothesized to
shift the equilibrium to favor the alpha helical conformation,
allowing hydrophobic residues in the specificity loop to insert
into a lipid bilayer (35). Mutating lysine residues in the spec-
ificity loop have been shown to disrupt plasma membrane
localization (16). Mutations that break the alpha helical char-
acter, but preserve the overall basic charge, have also been
shown to disrupt lipid kinase activity (35). In both cases, loss of
PIP5K function has been attributed to perturbing the PI(4)P-
binding interaction. While these studies have revealed which
motifs are critical for PIP5K lipid kinase activity, we do not
currently understand how these motifs contribute to the
cooperative PI(4,5)P2 binding, membrane-mediated dimeriza-
tion, and positive feedback. Deciphering these mechanisms
could help explain why PIP5K is predominantly localized and
activated at the plasma membrane.

Here we present a mutational analysis and single molecule
TIRF-M study of PIP5K, which reveals the sequence of mo-
lecular interactions that control PIP5K membrane docking and
PI(4,5)P2 engagement. The initial membrane interaction is
regulated by the specificity loop, which plays an important first
step for membrane binding but is not essential for productive
docking on PI(4,5)P2-containing membranes. Using a set of
chimeric proteins that contain different specificity loops, we
show that the initial binding step of PIP5K can be mediated by
2 J. Biol. Chem. (2024) 300(9) 107631
interactions with either PI(4)P or PI(4,5)P2 lipids. Our findings
are consistent with the previously reported substrate specificity
for PIP5K. Interactions with the lipid product, PI(4,5)P2, were
also paradoxically mediated through a region of the kinase
commonly referred to as the PIPBM or substrate-binding
motif. We discovered that the PIPBM is a broad anionic
lipid sensor that can associate with lipids including phospha-
tidylserine (PS) and all types of PIP lipids. Overall, the strength
of lipid interactions with the PIPBM is strongly correlated with
the valency of phosphorylation on the PIP lipid head group,
with triply phosphorylated PIP lipids exhibiting the strongest
interaction. Supporting that the PIPBM controls membrane
localization in addition to catalysis, we find that PIPBM mu-
tants can still phosphorylate PI(4)P when localized to bilayers
artificially. Finally, we used new insight about PIP5Ks’
membrane-binding mechanism to validate separation of
function mutants that are catalytically dead but retain their
ability to localize to PI(4,5)P2-containing membranes. Overall,
this work improves our understanding of the molecular
mechanisms underlying PI(4,5)P2-mediated recruitment of
PIP5K to membranes.
Results

The specificity loop regulates PIP5K localization to PI(4,5)P2-
containing membranes

Previous studies indicate that PIP5K substrate recognition is
regulated by the specificity loop (16–18, 35). Positioned within
�20 Å of the active site, the specificity loop is thought to help
organize the active site to orient the substrate, PI(4)P, for
catalysis (18). NMR experiments also suggest that membrane
docking of PIP5K causes the unstructured specificity loop to
adopt an alpha helical conformation, which reportedly inserts
into lipid membranes (35). Membrane insertion of an
amphipathic helix is predicted to significantly enhance the
membrane dwell time of PIP5K, which could contribute to the
positive feedback mechanism involving cooperative membrane
binding of PIP5K to the lipid product, PI(4,5)P2 (33) (Fig. S1).
To determine how the specificity loop contributes to PIP5K
localization and positive feedback, we disrupted alpha helix
formation in the specificity loop by introducing a W365P
mutation in PIP5KB (Fig. 1A). A homologous mutation (i.e.
W393P) was previously shown to reduce the catalytic activity
of zPIP5KA 25-fold relative to the WT kinase (35). Consistent
with previous observations, PIP5KB (W365P) displayed a 50-
fold reduction in lipid kinase activity compared to WT
PIP5KB when measured on SLBs using TIRF microscopy
(Fig. 1B). By contrast, the PIP5KB (W365A) mutant displayed a
less dramatic decrease in catalytic activity (Fig. S2A), sug-
gesting that preservation of the specificity loop alpha helical
character and/or hydrophobicity is important for PIP5K
catalysis. This is consistent with the zPIP5KA (W393G)
mutant reportedly having similar kinase activity to WT
zPIP5KA (35).

To determine the extent to which the reduced activity of
PIP5KB (W365P) is due to defects in membrane localization,
we used sortase-mediated peptide ligation to fluorescently



A

W365P

*

SOLUTION

MEMBRANE

PIP5KB

*

B

0 200 400 600 800 1000
0

10000

20000

30000

40000

50000

60000

0.5µM W365P

0.2µM W365P1µM W365P
1nM 5KB

P
I(

4
,5

)P
2
 l

ip
id

s
/

m

C

D

2 µm

E F G

H I J

AF647-5KB
(W365P)

AF647-5KB

AF647-5KB(W365P)

AF647-5KB

PIP5KB

PIP5KB (D51R)

PIP5KB (W365P)

0 5 10
0.001

0.01

0.1

1

0 1 2 3 4 5

0

0.05

0.10

0.15

0.20

0.25

Dwell time (seconds)

F
re

q
u

e
n

cy

0 5 10

0 1 2 3 4 5

0

0.02

0.04

0.06

0.08

Dwell time (seconds)

F
re

q
u

e
n

cy

Dwell time (sec)

lo
g

1
0
(1

-C
D

F
)

0.001

0.01

0.1

1

Dwell time (sec)

log
10(

1-C
DF)

0 25 50 75 100

0

0.002

0.004

0.006

0.008

0.010

[PIP5KB], pM

AF647-5KB

AF647-5KB (W365P)

0 1 2 3 4

0

0.2

0.4

0.6

0.8

1.0

% PI(4,5)P
2

N
o

rm
a

li
z

e
d

 P
IP

5
K

m
e

m
b

ra
n

e
 b

in
d

in
g

 (
A

.U
.)

M
e

m
b

ra
n

e
b

in
d

in
g

 f
re

q
u

e
n

cy

(m
o

le
cu

le
s/

µ
m

2
•s

e
c

) 

slow kinetics
[PIP2]

fast kinetics
[PIP2]

PI(4,5)P
2

PI(4)P
PIP5K

POSITIVE 
FEEDBACK
based on 

product binding

0

0.1

0.2

0.3

0.4

0.5
T

im
e

co
n

st
a

n
t

(
1
, s

e
c)

5KB 5KB
(W365P)

p=0.0002

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

5KB 5KB
(W365P)

p<0.0001

%
 m

E
o

s-
5

K
B

 p
a

r t
ic

le
s

tr
a

ck
e

d
 <

 1
0

4
 m

s

K
mEos-5KB

(pre UV)

EM
 = 516 nm

5 µm

mEos-5KB 
(post UV)

EM
 = 580 nm

5KB 5KB
(W365P)

0

0.05

0.10

0.15

0.20

0.25

0.30

0.35

A
v

e
. d

is
p

la
ce

m
e

n
t 

(µ
m

)

m
E

o
s-

5
K

B
 p

e
r 

f r
a

m
e

p<0.0001

Time (seconds)

Figure 1. The PIP5K specificity loop regulates the membrane association and dissociation rate constants. A, cartoon illustrating how the W365P
mutation disrupts the specificity loop alpha helical conformation in PIP5KB. B, representative lipid kinase activity traces measured in the presence of PIP5KB
and PIP5KB (W365P). Production of PI(4,5)P2 was monitored in the presence of 20 nM Cy3-PLCd and the indicated concentration of PIP5KB. C, PIP5KB (WT
and D51R) but not PIP5KB (W365P) display cooperative PI(4,5)P2-dependent changes in supported membrane localization. Phosphorylation of PI(4)P and
membrane localization of PIP5KB were monitored in the presence of 5 nM AF647-PIP5KB, 5 nM mNG-PIP5KB (D51R), or 5 nM AF647-PIP5KB (W365P). The
production of PI(4,5)P2 was visualized using 20 nM Cy3-PLCd in each kinase reaction. B and C, initial membrane composition: 96% DOPC, 4% PI(4)P. D,
representative TIRF-M images showing membrane localization in the presence of 100 pM AF647-PIP5KB or AF647-PIP5KB (W365P). E, experimental measure
of membrane-binding frequency in the presence of 10 to 50 pM AF647-PIP5KB (kON = 184/mM⋅mm2⋅sec) and 20 to 100 pM AF647-PIP5KB (W365P, kON = 9.68/
mM�mm2�sec). F and G, single molecule dwell time distributions measured in the presence of either AF647-PIP5KB (s1 = 0.815 ± 0.054s, s2 = 3.09 ± 0.35 s,
a = 0.95 ± 0.014, n = 14,950) or AF647-PIP5KB (W365P) (s1 = 0.296 ± 0.11s, s2 = 2.82 ± 0.71s, a = 0.77 ± 0.056, n = 2389). D–G, membrane composition: 98%
DOPC, 2% PI(4,5)P2. Errors equal SD from n ≥ 3 technical replicates. H, representative TIRF-M images of mEos-PIP5KB plasma membrane localization before
and after photoconversion with 405 nm light in HEK293T cells. I, a larger fraction of mEos-PIP5KB W365P (40%) display transient membrane-binding
events < 104 ms, compared to the WT mEos-PIP5KB (26%). Each data point equals the average number of transient molecules detected per cell
(N = 14–23 cells). J, the mean dwell time of mEos-PIP5KB (W365P) (s1 = 0.289 ± 0.05 s, N = 14 cells, n = 5662 molecules) is shorter than mEos-PIP5KB
(s1 = 0.353 ± 0.04 s, N = 23 cells, n = 19,087 molecules). Each data point is derived from a single exponential fit to a dwell time distribution from a sin-
gle cell. I and J, each data point equals the average number of transient molecules or time constant (s1, sec) calculated per cell (N = 14–23 cells). K, mEos-
PIP5KB (W365P) is more diffuse in the cellular plasma membrane than mEos-PIP5KB. Average single molecule displacement during 52 ms time interval is
greater for mEos-PIP5KB W365P (0.223 ± 0.03 mm, N = 14 cells) than mEos-PIP5KB (0.161 ± 0.01 mm, N = 23 cells). I–K, errors equal SD. Each data point is the
average displacement measured for a single cell. p value calculated by unpaired two-tailed Student’s t test.

Molecular basis of PIP5K product recognition
label PIP5KB with Alexa Fluor 647 (AF647) and visualized the
mutant kinase localization in vitro on reconstituted SLBs.
Unlike WT and dimerization-deficient (i.e. D51R) AF647-
PIP5KB, localization of AF647-PIP5KB (W365P) did not
exhibit cooperative membrane recruitment that was coupled
to the production of PI(4,5)P2 lipids (Figs. 1C, S1, B and C, and
S2B). Overall, the reduced PIP5KB (W365P) activity was
strongly correlated with a decrease in membrane localization.
Since membrane binding of AF647-PIP5KB (W365P) was
unresponsive to the generation of PI(4,5)P2 lipids, we hy-
pothesized that the specificity loop directly senses PI(4,5)P2
lipids, in addition to playing a critical role in substrate
recognition. When we compared the single molecule locali-
zation of AF647-PIP5KB and AF647-PIP5KB (W365P), we
observed a striking difference in the total number of
membrane-binding events on PI(4,5)P2-containing bilayers
J. Biol. Chem. (2024) 300(9) 107631 3



Molecular basis of PIP5K product recognition
(Fig. 1D). To quantify this difference, we measured the single
molecule membrane-binding frequency of AF647-PIP5KB and
AF647-PIP5KB (W365P) (Figs. 1E and S2, C and D). Mutating
the specificity loop resulted in a 19-fold decrease in the asso-
ciation rate constant (kON) compared to WT PIP5KB (Fig. 1E).
A similar reduction in the membrane-binding frequency was
observed for mNG-PIP5KB (W365A) (Fig. S2E). In addition to
the observed reduction in kON for AF647-PIP5KB (W365P),
disrupting the specificity loop reduced the single molecule
dwell time in the presence of PI(4,5)P2 (Fig. 1, F and G).
Considering that the specificity loop has been proposed to
penetrate directly into membranes (35), we expected to
observe a more dramatic decrease in the membrane dwell
time. Mutations that preserve the alpha helical character of the
specificity loop (i.e. W365A and L61E/W365R) but reduce the
hydrophobicity displayed similarly modest decreases in the
dwell time on PI(4,5)P2-containing membranes (Fig. S2F).
Together, these findings suggest that while binding frequency
and dwell time were diminished, PIP5K specificity loop mu-
tants can still dock on PI(4,5)P2-containing membranes.

The single molecule supported membrane-binding experi-
ments provide a useful platform to measure individual
protein–lipid interactions; however, they lack some of the
complexity of a cellular plasma membrane. For example, the
supported bilayers described here lack sterol lipids and various
peripheral membrane-binding proteins that could modulate
membrane localization of PIP5K. To determine whether
membrane binding of PIP5KB (W365P) was altered in living
cells, we established conditions to visualize the plasma mem-
brane localization of PIP5KB in human embryonic kidney
(HEK)293T cells with single molecule resolution using TIRF-
M. For these experiments, genes encoding either mEos-
tagged PIP5KB or PIP5KB (W365P) were transiently
expressed in HEK293T cells under a heat shock promoter,
which yielded a near single molecule membrane density of
fluorescently tagged kinase localized to the plasma membrane
(Fig. 1H). Following a short pulse of 405 nm light, a fraction of
the mEos-tagged PIP5K were photoconverted from the green
to red fluorescent state, which provided the spatial resolution
needed to track individual PIP5K molecules (Fig. 1H). Based
on measurements across multiple cells, we classified signifi-
cantly more transient plasma membrane localization events
(i.e. < 104 ms) in the presence of mEos-PIP5KB (W365P)
(Fig. 1I). Excluding these brief localization events from our
data set, we still consistently measured average dwell times for
mEos-PIP5KB (W365P) that were shorter than mEos-PIP5KB
(Fig. 1J). Consistent with the W365P mutant having fewer
membrane lipid interactions, the average single molecule
displacement of mEos-PIP5KB (W365P) was significantly
larger than mEos-PIP5KB (Fig. 1K). The observed membrane-
binding frequency of mEos-PIP5KB (W365P) was also
reduced. However, calculating the kON for mEos-PIP5KB (WT
and W365P) in cells was impractical due to cell-to-cell varia-
tion in protein expression and differences in the fraction of
photoconverted mEos, which make it challenging to estimate
the total cellular concentration of mEos-PIP5K. Overall, our
cellular measurements agree with our in vitro observations
4 J. Biol. Chem. (2024) 300(9) 107631
that the specificity loop facilitates PI(4,5)P2 binding but is not
solely responsible for membrane localization.
The PIP5K specificity loop enhances PI(4)P and PI(4,5)P2
specificity of chimeric proteins

Previous research indicates that the specificity loop controls
membrane docking of PIP5K through substrate specificity (16,
18, 35). This is supported by molecular dynamics simulations
that indicate membrane docking relies on the initial engage-
ment of PI(4)P by the specificity loop, which then transitions
to the active site for catalysis (37). Based on our single mole-
cule dwell time analysis of AF647-PIP5KB on SLBs, the
specificity loop regulates the rate of membrane association
(kON) on PI(4,5)P2-containing membranes (Fig. 1E). Given
these findings, we hypothesized that the specificity loop can
directly interact with PI(4,5)P2 lipids, a function previously
uncharacterized. One question that arises with this hypothesis
is whether the specificity loop functions as a broad anionic
lipid-binding domain or if it has specificity for certain PIP
lipids. Given that the specificity loop was previously shown to
favor PI(4)P lipids as a substrate for catalysis, we hypothesized
that the specificity loop does not act as a broad anionic lipid-
binding domain. Testing these hypotheses in the context of the
full-length PIP5K protein, however, would be difficult to
interpret due to the highly electro-positivity surface proposed
to interact with anionic lipids. To isolate the role the specificity
loop alone serves as a PI(4)P or PI(4,5)P2 lipid sensor, we
created a chimeric mNeonGreen (mNG) fusion protein con-
taining the PIP5K specificity loop fused to the pleckstrin ho-
mology (PH) domain derived from phospholipase C-d1 (i.e.
mNG-5K(SL)-PLCd) (Fig. 2A). Characterization of the mNG-
5K(SL)-PLCd membrane-binding properties in the context of
PLCd provided an approach for measuring the potential gain
in PIP lipid specificity and enhanced membrane-binding af-
finity due to the addition of the PIP5KB specificity loop. Using
single molecule TIRF-M, we examined whether mNG-5K(SL)-
PLCd acquired membrane-binding specificity for either PI(4)P
or PI(5)P lipids (Fig. 2B). Consistent with PH-PLCd lacking
specificity for either PI(4)P or PI(5)P, mNG-PLCd did not
strongly interact with either lipid on supported membranes
(Fig. 2B). By contrast, we observed an increase in the single
molecule dwell time for mNG-5K(SL)-PLCd in the presence of
PI(4)P, but not PI(5)P (Fig. 2B). This result suggests that the
PIP5K specificity loop can discriminate between PI(4)P and
PI(5)P, even when fused to a different peripheral membrane-
binding protein. Note that we also attempted to visualize
membrane interactions of mNG fused to the PIP5K specificity
loop (i.e. mNG-5K-SL) alone; however, this chimeric protein
displayed very transient membrane interactions that were not
quantifiable. We assume that the additional weak electrostatic
interactions provided by PH-PLCd facilitate membrane asso-
ciation of 5K(SL). As a control, we performed similar single
molecule tracking experiments on PI(4)P- and PI(5)P-
containing membranes in the presence of both full length
AF647-PIP5KB and the AF647-PIP5KB (W365P) specificity
loop mutant (Fig. S3, A–C). Similar to the trends in Figure 2B,



mNG

PIP4K PIP4K(5K-SL)

5K-SL

PI(4)P PI(5)P
PI(4,5)P

2

A B C

hPIP5KB -- QSYRLMKKLEHSWKALVYDGDTV
hPIP5KC -- QSYRFIKKLEHTWKALVHDGDTV

hPIP5KA -- QSYRFVKKLEHSWKALVHDGDTV

hPIP4KB -- TPYDTKKKAAHAAKTVKHGAGAEI
hPIP4KC -- TQYDAKKKAAHAAKTVKHGAGAEI

hPIP4KA -- THYDAKKKAAHAAKTVKHGAGAEI

D E F

G

loop swap

PLC

PI(4,5)P
2

PIP lipid binding?
0 1 2 3 4 5 6

0.001

0.01

0.1

1

Dwell time (seconds)

lo
g

1
0
(1

-C
D

F
)

0 0.4 0.8 1.2 1.6

0.001

0.01

0.1

1 PI(4)P, mNG-PLC

PI(4)P, mNG-(SL)-PLC

PI(5)P, mNG-PLC

PI(5)P, mNG-(SL)-PLC

Dwell time (seconds)
lo

g
1

0
(1

-C
D

F
)

mNG-PLC

mNG-(SL)-PLC

0 1 2 3 4

0

50

100

150

200

250

molar % PI(4,5)P
2

F
o

ld
 i

n
cr

e
a

se
 C

y
5

-4
K

B
m

e
m

b
r a

n
e

 r
e

cr
u

it
m

e
n

t Cy5-PIP4K(5K-SL)

Cy5-PIP4K

enhanced 
PI(4,5)P

2

0 5 10 15 20

0.001

0.01

0.1

1

Dwell time (seconds)

lo
g

1
0
(1

-C
D

F
)

H I

0 20 40 60 80

0

0.02

0.04

0.06

[PIP4K], pM

Cy5-PIP4K(5K-SL)
Cy5-PIP4K

M
e

m
b

ra
n

e
 b

in
d

in
g

 f
re

q
u

e
n

cy
 (

m
o

le
cu

le
s/

µ
m

2
•s

e
c

) 

0 10 20 30 40 50 60

0

0.5

1.0

1.5

2.0

20pM

40pM

80pM

C
u

m
u

la
ti

v
e

 M
e

m
b

ra
n

e
 B

in
d

in
g

(m
o

le
c

u
le

s/
µ

m
2
) 

Time (seconds)

Cy5-PIP4K(5K-SL)
Cy5-PIP4K

Figure 2. The PIP5K specificity loop enhances PI(4,5)P2 lipid binding. A, cartoon showing the structure of the mNG-PIP5K specificity loop-PLCd chimeric
protein. The expected PIP lipid-binding interactions are indicated. B, single molecule dwell time distributions of mNG-PLCd measured on membranes
containing either PI(4)P (s1 = 20 ± 1 ms, n = 6456) or PI(5)P (s1 = 26 ± 10 ms, n = 2872). Single molecule dwell time analysis of mNG-(5K-SL)-PLCd yield the
following time constants: PI(4)P (s1 = 42 ± 34 ms, s2 = 314 ± 271 ms, a = 0.81 ± 0.13, n = 4274) or PI(5)P (s1 = 82 ± 13 ms, n = 855). In both cases, dwell times
were measured on membrane containing 4% PI(4)P or 4% PI(5)P. C, single molecule dwell time distributions of mNG-PLCd (s1 = 0.346 ± 0.041 s, n = 5414)
and mNG-(5K-SL)-PLCd (s1 = 0.127 ± 0.059 s, s2 = 0.797 ± 0.066 s, a = 0.47 ± 0.06, n = 3082) measured on membranes containing 4% PI(4,5)P2. D, sequence
alignment of PIP5K and PIP4K specificity loops. E, cartoon showing the PIP4K to PIP5K specificity loop swap. F, quantification of the fold increase in the
localization of Cy5-PIP4K and Cy5-PIP4K(5K-SL) on membranes containing different molar concentrations of PI(4,5)P2. The fold change was calculated
relative to the measured membrane density in the absence of PI(4,5)P2. G, single molecule dwell time distributions of Cy5-PIP4K (s1 = 0.248 ± 0.021 s,
s2 = 1.82 ± 0.16 s, a = 0.81 ± 0.04, n = 13,117 molecules) and Cy5-PIP4K(5K-SL) (s1 = 0.431 ± 0.067 s, s2 = 8.77 ± 4.9 s, a = 0.84 ± 0.05, n = 12,720 molecules)
measured on membranes containing 4% PI(4,5)P2. H, plots showing the cumulative membrane-binding events measured in the presence of 20 to 80 pM
Cy5-PIP4K. The slope of each curve was calculated by linear regression yielding a binding frequency for each concentration of Cy5-PIP4K. I, experimental
measure of membrane-binding frequency in the presence of 20 to 80 pM Cy5-PIP4K (kON = 0.56/nM⋅mm2⋅sec) or 5 to 10 pM Cy5-PIP4K(5K-SL) (kON = 2.93/
nM⋅mm2⋅sec). B, C, and G, errors equal SD from n ≥ 3 technical replicates.

Molecular basis of PIP5K product recognition
mutating the specificity loop had a more dramatic effect on the
dwell times measured in the presence of PI(4)P compared to
PI(5)P (Fig. S3, B and C).

In addition to regulating substrate recognition, our results
indicate that the specificity loop controls PI(4,5)P2 lipid
binding and contributes to the positive feedback loop observed
during PIP5K-mediated production of PI(4,5)P2. To determine
if the PIP5K specificity loop can generally enhance the affinity
of other membrane-binding proteins, we compared the single
molecule dwell times of mNG-PLCd and mNG-5K(SL)-PLCd
on PI(4,5)P2-containing membranes (Fig. 2C). Both mNG-
PLCd and mNG-5K(SL)-PLCd displayed transient dwell times
that were nearly identical, due to both proteins sharing the
common PH-PLCd domain (Fig. 2C). Looking at the dwell
time distribution of mNG-5K(SL)-PLCd, we observed a second
population of longer dwelling molecules, consistent with the
specificity loop enhancing PI(4,5)P2 association (Fig. 2C).

Our recent membrane-binding study of the type II phos-
phatidylinositol-5-phosphate 4-kinase (PIP4K) revealed that a
threshold membrane density of �3% PI(4,5)P2 lipids was
J. Biol. Chem. (2024) 300(9) 107631 5



Molecular basis of PIP5K product recognition
required to observe robust PIP4K localization on SLBs (38). By
contrast, PIP5K strongly localizes to membranes containing
only 1% PI(4,5)P2 (33, 38). To determine whether the differ-
ence in PI(4,5)P2-dependent membrane association is related
to the specificity loop sequence variation between PIP4K and
PIP5K (Fig. 2D), we generated a PIP4KB to PIP5KB specificity
loop swap referred to as PIP4K(5K-SL) (Fig. 2E). We quantified
the difference in Cy5-PIP4K and Cy5-PIP4K(5K-SL) bulk
membrane recruitment on supported membranes containing
varying concentrations of PI(4,5)P2 (Figs. 2, F and G and S4).
In the presence of 1% PI(4,5)P2, Cy5-PIP4K(5K-SL) reached a
final equilibrium membrane surface density that was 160-fold
greater than Cy5-PIP4K (Fig. 2F). Visualization of the single
molecule membrane-binding properties of Cy5-PIP4K and
Cy5-PIP4K(5K-SL) revealed an increase in the mean dwell
time (0.56 and 0.66 s, respectively) due to the specificity loop
(Fig. 2G). Given the extremely long dwell times displayed by
Cy5-PIP4K and Cy5-PIP4K(5K-SL), we are potentially under-
estimating the total specificity loop–dependent increase in
membrane dwell time due to Cy5 photobleaching. Consistent
with the specificity loop enhancing PIP5K membrane locali-
zation to PI(4,5)P2-containing membranes, swapping the
specificity loop also increased the association rate constant
(kON) of Cy5-PIP4K(5K-SL) by 5-fold (Fig. 2, H and I).
Considering that the PIP4KB to PIP5KB specificity loop swap
reduces the isoelectric from 9.7 to 8.4, the observed
enhancement in membrane localization for Cy5-PIP4K(5K-SL)
cannot be attributed solely to a change in net positive charge.
Overall, these data confirm that the PIP5K specificity loop can
sense both PI(4)P and PI(4,5)P2 lipids. This data also provides
one explanation for why PIP4K has a reduced capacity to
localize to PI(4,5)P2 lipids compared to PIP5K both in vivo and
in vitro (38).
Molecular basis of the PIP5K high affinity PI(4,5)P2 lipid
interaction

Having established that the specificity loop serves an
important role in regulating the initial membrane docking step
of PIP5K, we sought to identify residues responsible for the
high affinity PI(4,5)P2 lipid-binding interaction. We hypothe-
sized the existence of this interaction site based on the ability
of AF647-PIP5K (W365P) to stably associate with PI(4,5)P2-
containing membranes (Fig. 1). PIP5K contains numerous
basic amino acids on the membrane-binding interface that
could promote interactions with anionic lipids (24, 39).
Structural biochemistry studies of zPIP5KA protein crystals
soaked with selenate resolved a SeO4

2- molecule bound to
basic residues (i.e. K238 and R244) near the ATP-binding
pocket (18). These residues are part of a conserved
sequence, DLKGSxxxR, which is referred to as the PIPBM or
substrate-binding motif (18) (Fig. 3A). The PIPBM is consid-
ered part of the kinase domain active site and plays an
important role in positioning PI(4)P near the ATP g-phos-
phate for efficient lipid phosphorylation (18). Consistent with
this model, molecular dynamic simulations have shown that
zPIP5KA interacts with PI(4)P lipids through the PIPBM (37).
6 J. Biol. Chem. (2024) 300(9) 107631
Mutation in R244 has been shown to reduce zPIP5KA lipid
kinase activity but paradoxically does not affect the intrinsic
ATP hydrolysis rate of zPIP5KA in solution (18).

Does the PIPBM regulate PI(4,5)P2 binding? Structural
modeling of PI(4)P bound to the PIPBM previously suggested
that the inositol 5-phosphate of PI(4,5)P2 would sterically clash
with the g-phosphate of ATP (18). However, when we
measured the single molecule dwell times AF647-PIP5KB in
TIRF-M imaging buffer containing either ADP, ATP, or
nonhydrolyzable ATPgS, we found that the kinase could
associate with PI(4,5)P2 membranes independent of the
nucleotide chemistry (Fig. 3B). In fact, the presence of ATP
caused AF647-PIP5KB to dwell slightly longer
(s1 = 0.463 ± 0.16 s) on membranes thano kinases visualized in
the presence of ADP (s1 = 0.365 ± 0.11 s) or ATPgS
(s1 = 0.329 ± 0.004 s). This suggests that nucleotide binding
does not sterically occlude the interaction between PIP5K and
PI(4,5)P2, which led us to hypothesize that the conserved
PIPBM could potentially regulate PI(4,5)P2 lipid binding. To
test whether the PIPBM is required for the PI(4,5)P2 lipid in-
teractions, we introduced mutations that were homologous to
zPIP5KA (i.e. K238 and R244) into human PIP5KB (i.e. K205A
and R211A). We found that disrupting the PIPBM strongly
diminished bulk membrane binding of PIP5K on PI(4,5)P2-
containing membranes (Fig. 3C). Consistent with our bulk
membrane localization experiment, single molecule charac-
terization of AF647-PIP5KB (K205A/R211A) revealed a 2.4-
fold decrease in the average dwell time compared to AF647-
PIP5KB (Fig. 3D). While we did see greatly diminished dwell
time, we did not observe a complete loss of PI(4,5)P2 binding.
This shorter dwell time could be explained by the remaining
specificity loop-binding interaction. To determine if our
in vitro membrane-binding results are consistent with in vivo
membrane-binding dynamics, we expressed mEos-PIP5KB
(K205A, R211A) in HEK293T cells and quantified the single
molecule plasma membrane dwell time (Fig. 3E). Consistent
with the mutant having a reduced frictional drag coefficient in
the plasma membrane due to fewer lipid interactions, the
average single molecule displacement of mEos-PIP5KB
(K205A/R211A) was significantly larger than mEos-PIP5KB
(Fig. 3F). The average single molecule dwell time for mEos-
PIP5KB (K205A/R211A) was also shorter than mEos-
PIP5KB, reflecting a loss of molecular interactions (Fig. 3G).
Fitting all the observed single molecule-binding events
revealed that a larger fraction of mEos-PIP5KB (K205A/
R211A, 34%) molecules displayed a transient membrane-
binding behavior than mEos-PIP5KB (22%) (Fig. 3H). This
data suggests that the PIPBM serves a role in regulating PIP5K
plasma membrane localization in cells. However, other
membrane localization mechanisms such as PIP5K dimeriza-
tion (34, 36) and recruitment by PIP4K (38, 40) could mask the
severity of the K205A/R211A localization phenotype observed
in vivo.

Previous biochemical characterization of the zPIP5KA
(R244A) mutant revealed that the PIPBM is critical for phos-
phorylating PI(4)P to generate PI(4,5)P2 (18). Our data also
suggests that mutations in the PIPBM reduce PIP5K



F

SpyCatcher

SpyTag

PIP5KB
(K205A, R211A)

rescue catalysis?

PI(4,5)P
2

PI(4)P

MCC-PE PI(4)P PI(4,5)P
2

A B C

D

SIDE BOTTOM

loop

PIPBM
K205

R211

0 20 40 60 80 100

0

2500

5000

7500

10000

12500

15000

Time (seconds)

M
e

m
b

ra
n

e
in

t e
n

si
ty

m
N

G
-P

IP
5

K
B

(A
.U

.)

mNG-5KB

mNG-5KB (K205A, R211A)

0 100 200 300 400 500

0

2000

4000

6000

8000

10000

12000

Time (seconds)

P
I(

4
,5

)P
2
 l

ip
id

s/
µ

m
2

5KB(KR->AA)

5KB(KR->AA)-SpyTag

AF647-5KB
AF647-5KB (K205A, R211A)

H

DLKGSxxxR

I

0 1 2 3 4 5 6

0.01

0.1

1

Time (seconds)

0 1 2 3 4

0.001

0.01

0.1

1

ATP

ATP S

ADP

Dwell time (seconds)

lo
g

1
0
(1

-C
D

F
)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

M
e

a
n

d
w

e
ll

ti
m

e
 (

se
c)

5KB 5KB
(KR->AA)

E

lo
g

1
0
(1

-C
D

F
)

G p<0.0001

0

0.05

0.10

0.15

0.20

0.25

5KB 5KB
(KR->AA)

p<0.0001

A
v

e
. d

is
p

la
ce

m
e

n
t 

(µ
m

)

m
E

o
s-

5
K

B
 p

e
r 

fr
a

m
e

0 1 2 3 4 5 6

0.001

0.01

0.1

1

lo
g

1
0
(1

-C
D

F
)

mEos-5KB
mEos-5KB (K205A, R211A)

J

5 µm

mEos-5KB mEos-5KB (K205A, R211A)

post 405 nm pulse, 
EM

 = 580 nm 

Dwell time (seconds)

Figure 3. PIP5K can engage PI(4,5)P2 lipids through the substrate-binding motif. A, AlphaFold structure of monomeric human PIP5KB validated against
zPIP5Ka (5E3U.pdb). Structure highlights the specificity loop (red) and the PIP lipid (or substrate) binding motif (PIPBM, blue). B, single molecule dwell time
distributions of AF647-PIP5KB measured in the buffer containing 1 mM ATP (s1 = 0.463 ± 0.16 s, n = 6377), 1 mM ADP (s1 = 0.365 ± 0.11 s, n = 10,543), or 1
mM ATPgS (s1 = 0.329 ± 0.004 s, n = 6231). Errors equal SD from n = 3 technical replicates. C, bulk membrane recruitment of either 25 nM mNG-PIP5KB or
mNG-PIP5KB (K205A/R211A). B and C, membrane composition: 98% DOPC, 2% PI(4,5)P2. D, single molecule dwell time distribution comparing AF647-PIP5KB
(s1 = 0.815 ± 0.054, s2 = 3.09 ± 0.35, a = 0.95 ± 0.01, n = 14950) and AF647-PIP5KB (K205A/R211A, s1 = 0.306 ± 0.19 s, n = 684). Membrane composition:
98% DOPC, 2% PI(4,5)P2. E, representative TIRF-M images showing the localization of the mEos-PIP5KB (WT and K205A/R211A) expressed in HEK293T cells. F,
mEos-PIP5KB (K205A, R211A) is more diffuse in the cellular plasma membrane than mEos-PIP5KB. Average single molecule displacement during 52 ms time
interval is greater for mEos-PIP5KB K205A,R211A (0.203 ± 0.02 mm, N = 18 cells) than mEos-PIP5KB (0.161 ± 0.01 mm, N = 23 cells). Each data point is the
average displacement measured for a single cell. G, the mean dwell time of mEos-PIP5KB K205A/R211A (0.516 ± 0.06 s, N = 24 cells) is shorter than mEos-
PIP5KB (0.632 ± 0.06s, N = 27 cells). Each data point is derived from a single cell. F and G, p value calculated by unpaired Student’s t test. H, significantly more
transient interactions observed for mEos-PIP5KB K205A/R211A (s1 = 0.142 s, s2 = 0.645 s, a = 0.34, n = 15,162 molecules, N = 24 cells) than mEos-PIP5KB
(s1 = 0.159 s, s2 = 0.719 s, a = 0.22, n = 19450 molecules, N = 27 cells [a represents the fraction of short dwelling molecules (s1)]). Dwell time distribution was
fit using all single molecule-binding events measured across all cells. I, schematic showing SpyTag-PIP5KB (K205A/R211A) and SpyCatcher kinase activity
assay. J, kinetics of lipid phosphorylation measured for membrane tethered and nontethered SpyTag-PIP5KB (K205A/R211A). Production of PI(4,5)P2 was
monitored in the presence of 20 nM Cy3-PLCd. Initial membrane composition: 98% DOPC, 2% PI(4)P. Errors equal SD.

Molecular basis of PIP5K product recognition
localization to membranes containing PI(4,5)P2. This led us to
hypothesize that the previously observed loss in zPIP5KA ki-
nase activity could be partially due to a disruption in mem-
brane binding. This would be consistent with zPIP5KA
(R244A) displaying no change in the ATP hydrolysis rate
measured in solution (18). To test this hypothesis, we utilized
the SpyTag-SpyCatcher interaction (41, 42) to rescue PIP5KB
(K205A/R211A) membrane localization and measure lipid
kinase activity (Fig. 3I). For these experiments, SpyCatcher was
conjugated to a bilayer via a maleimide lipid. SpyTag-mNG-
PIP5KB (K205A/R211A) was then flowed over the supported
membrane in buffer lacking ATP and allowed to form an
isopeptide bond with membrane-conjugated SpyCatcher. Un-
bound kinase in solution was washed out of the sample
chamber before initiating the kinase reaction by adding ATP-
containing buffer and the PI(4,5)P2 biosensor, Cy3-PLCd.
Consistent with the previous characterization of zPIP5KA
(R244A), we found that nonmembrane-tethered SpyTag-
PIP5KB (K205A/R211A) mutant was unable to catalyze PI(4,5)
P2 production in solution (Fig. 3J). However, upon tethering to
SpyTag-PIP5KB (K205A/R211A) to SLBs via SpyCatcher, we
were able to localize and rescue PI(4,5)P2 production. Taken
J. Biol. Chem. (2024) 300(9) 107631 7



Molecular basis of PIP5K product recognition
together, these data confirm that the site of high affinity
binding and positive feedback come from binding PI(4,5)P2
lipids in the PIPBM. This high affinity binding interactions
controls membrane localization of PIP5K both in vitro and
in vivo.
The PIPBM displays broad specificity for anionic lipids

While the canonical activity of PIP5K is the phosphorylation
of PI(4)P to produce PI(4,5)P2, PIP5K can also phosphorylate
PI(3)P and PI(5)P, albeit less efficiently compared to PI(4)P
(17, 18). In addition, PIP5K has also been shown to catalyze the
synthesis of the important signaling lipid PI(3,4,5)P3 using
PI(3,4)P2 as a substrate (18). Having established that the
PIPBM regulates membrane interactions with PI(4,5)P2, we
aimed to determine whether the PIPBM controls broad spec-
ificity for anionic lipids. For this, we created SLBs and per-
formed single particle tracking of mNG-PIP5K in the presence
of various types of PIP lipids (Fig. 4A and Table 1). We found
that mNG-PIP5K bound singly phosphorylated PI(4)P and
PI(5)P lipids with a similarly low affinity, while interactions
with either doubly or triply phosphorylated PIP lipids (PI(3,4)
P2, PI(4,5)P2, and PI(3,4,5)P3) displayed longer dwell times
A B

D E

0 2 4 6 8 10 12 14

0.001

0.01

0.1

1
PI(5)P

PI(4)P

PI(3,4)P2

PI(3,4,5)P3

PI(4,5)P2

Dwell time (seconds)

lo
g

1
0
(1

-C
D

F
)

20% DOPS
4% PI(4,5)P

2

Dwell time (seconds)

lo
g

1
0
(1

-C
D

F
)

R
e

a
ct

io
n

h
a

lf
-t

im
e

(s
e

c)

0

200

400

600

800

1000

1200

mNG-5KB

0 2 4 6 8 10 12
0.001

0.01

0.1

1

Figure 4. The PIP5K PIPBM displays broad specificity for anionic lipids. A,
PIP5KB on SLBs containing 4% of the indicated PIP lipid. B, kinetics of PIP5KB-de
All membranes contain 2% PI(4)P and reactions were monitored using 20 nM
completion. C, quantification of the fold increase of membrane recruitment me
DOPS. The fold increase in membrane binding we calculated relative to the me
molecule dwell time analysis of mNG-PIP5KB in the presence of 4% PI(4,5)P
localization of 15 nM mNG-PIP5K or mNG-PIP5K (K205A/R211A). Membrane co
of images shown in (E). See Table 1 for single molecule dwell times (s1 and s

8 J. Biol. Chem. (2024) 300(9) 107631
(Fig. 4A and Table 1). To test whether PIP5KB prefers to bind
doubly and triply phosphorylated PIP lipids due to the struc-
ture of the inositol head group versus the total negative charge,
we examined the membrane-binding properties of mNG-
PIP5KB in the presence of PS lipids. We chose PS because
its headgroup has a distinct chemical structure compared to
PIP lipids but retains a negative charge. PS lipids have also
previously been shown to activate PIP5K, though the molec-
ular basis of this activation was not understood (43). In
agreement with published data, we found that increasing the
molar percentage of PS enhanced PIP5K activity, with a
physiologically relevant concentrations of 20% PS lipids
increasing the kinase activity 42.5-fold compared to mem-
branes that contained 2% PI(4)P and no PS lipids (Fig. 4B).
Correlated with the observed increase in kinase activity, mNG-
PIP5KB localization displayed a nonlinear response to PS
membrane density. At the physiologically relevant concentra-
tion of 20% PS, mNG-PIP5KB membrane localization
increased 60-fold compared to membranes lacking PS lipids
(Fig. 4C). Next, we compared the single molecule dwell times
of mNG-PIP5KB associated with supported membranes con-
taining either 4% PI(4,5)P2 or 20% PS lipids. In the presence of
either of these lipid species, mNG-PIP5KB displayed similarly
C

5 µm

mNG-5KB

mNG-5KB (K205A, R211A)

M
e

m
b

ra
n

e
 i n

te
n

si
ty

m
N

G
-P

IP
5

K
B

(A
.U

.x
1

0
3
) 

F
o

ld
 i

n
c r

e
a

se
m

N
G

-5
K

B
m

e
m

b
ra

n
e

re
cr

u
it

m
e

n
t

molar % DOPS

0 5 10 20

molar % DOPS

F

0 5 10 20

0

10

20

30

40

50

60

70

mNG-5KB 
(K205A, R211A)

0 25 50 75 100 125 150

0

5

10

15

20

25

30

35

40

Time (seconds)

single molecule dwell time distributions measured in the presence of mNG-
pendent lipid phosphorylation measured in the presence of 0 to 20% DOPS.
Cy3-PLCd. Kinetics were quantified by comparing the half-time for reaction
asured in the presence of 15 nM mNG-PIP5KB on SLBs containing 0 to 20%
mbrane intensity of mNG-PIP5K measured in the absence of DOPS. D, single
2 or 20% DOPS. E, representative TIRF-M images showing the membrane
mposition: 20% DOPS, 80% DOPC. F, kinetics of bulk membrane recruitment
2) and statistics.



Table 1

Protein visualized Membrane composition s1 ± SD (sec) s 2 ± SD (sec) a ± SD N n

mNG-PIP5KB 4% PI(4)P 0.469 ± 0.22 — — 3 6251
mNG-PIP5KB 4% PI(5)P 0.557 ± 0.14 — — 3 21,425
mNG-PIP5KB 4% PI(3,4)P2 0.854 ± 0.32 — — 3 5461
mNG-PIP5KB 4% PI(4,5)P2 1.05 ± 0.13 — — 3 3993
mNG-PIP5KB 4% PI(3,4,5)P3 0.503 ± 0.07 1.419 ± 0.028 0.42 ± 0.085 3 16,930
mNG-PIP5KB 20% PS 0.322 ± 0.19 1.414 ± 0.49 0.74 ± 0.23 3 11,351
mNG-PIP5KB 2% PI(4,5)P2 0.468 ± 0.019 — — 3 7075
mNG-PIP5KB (K138A) 2% PI(4,5)P2 0.285 ± 0.004 — — 3 7423
mNG-PIP5KB (D266K) 2% PI(4,5)P2 0.619 ± 0.071 — — 3 4667
mNG-PIP5KB (D350A) 2% PI(4,5)P2 0.228 ± 0.004 — — 3 3546

SD = standard deviation from the indicated number of technical replicates.
N = # of SLBs or cells used for calculating the mean dwell times (i.e. technical replicates).
n = total number of molecules tracked across all of the indicated number of technical replicates (N).
alpha (a) = fraction of molecules with characteristic dwell time (s1).
Membrane composition equals DOPC plus the indicated molar percentage of PIP or PS lipids.

Molecular basis of PIP5K product recognition
long membrane dwell times (Fig. 4D and Table 1). To deter-
mine if the interaction with PS lipids was mediated by the
PIPBM, we measured the bulk membrane localization of
mNG-PIP5K (K205A, R211A). Like our characterization of
mNG-PIP5K (K205A, R211A) membrane binding in the
presence of PI(4,5)P2 (Fig. 3C), the PIPBM mutant was unable
to localize to membranes containing 20% PS lipids (Fig. 4, E
and F). Taken together, these data demonstrate that PIP5K is
localized to PS through association of the lipid with the
PIPBM, supporting a model in which the PIPBM functions as a
broad specificity anionic lipid sensor.
Deciphering the relationship between PIP5K membrane
recruitment and catalysis

Our membrane-binding studies of the PIPBM mutant
revealed that residues that have previously been shown to be
critical for lipid kinase activity also regulate PIP5K membrane
localization. This raised questions about whether other
commonly studied “kinase dead” mutants are defective in
membrane binding. For this reason, we aimed to identify a
PIP5K “kinase dead” mutant that lacks lipid kinase activity
but retains the membrane-binding dynamics characteristic of
the WT enzyme. To validate a separation of function mutant,
we rationalized that PIP5K must have a functional specificity
loop and PIPBM. Looking at existing structural biochemistry
data of zPIP5KA (18), we found that residue D350 helps co-
ordinate a Mg2+/Mn2+ ions in the active site in PIP5KB
(Fig. 5A). Mutations in this residue are predicted to destabi-
lize the transition state following nucleophilic attack of the
ATP gamma phosphate. In addition, residues K138 and D266
of PIP5KB serve important roles in stabilizing ATP in the
active site. Using the mNG-PIP5K single molecule TIRF
assay, we quantified the membrane-binding properties of
fluorescently labeled PIP5KB (K138A, D266A, and D350A)
on supported membrane-containing 2% PI(4,5)P2 (Fig. 5B).
We found that only the D266K mutant exhibits membrane-
binding properties similar to WT PIP5KB (Fig. 5C). This
was observed when we measured the single molecule dwell
times (Fig. 5C and Table 1) and bulk membrane recruitment
(Fig. 5D). Next, we measured the catalytic efficiency of each
mutant on supported membranes containing an initial con-
centration of 4% PI(4)P and no additional anionic lipids.
Under these conditions, none of the mutants displayed sig-
nificant levels of activity compared to WT PIP5KB (Fig. 5E).
When we repeated these experiments on supported mem-
branes containing 20% PS lipids, we were able to stimulate the
activity of PIP5KB (D350A) through enhanced membrane
localization (Fig. 5F). However, the K138A and D266K mu-
tants still displayed weak activity even though their mem-
brane localization dynamics were similar to WT PIP5KB.
Together, these results indicate that some PIP5K “kinase
dead” mutants display reduced catalytic efficiency due to the
inability to bind to anionic lipids, which controls membrane
localization.
Discussion

Previous structural biochemistry studies have implicated
both the specificity loop and PIPBM (also known as the
substrate-binding motif) of PIP5K in regulating substrate
specificity and kinase activity (16–18, 35). These molecular
details contribute to two potential models that describe PIP5K
membrane docking and catalysis. In a single binding pocket
model, the specificity loop adopts a helical conformation
positioned near the PIPBM. These two motifs function in
concert to form a single unified PIP lipid-binding site
responsible for substrate specificity. This model is supported
by mutational analysis and chemical crosslinking studies that
show the specificity loop and PIPBM move into close prox-
imity of each other following substrate binding (18). Alterna-
tively, in a two-step binding model, the specificity loop
regulates the initial membrane sensing and docking step, while
the substrate binding and PI(4)P phosphorylation is mediated
by the PIPBM. In this model, the specificity loop holds the
PIP5K enzyme at the membrane until the PI(4)P lipid inter-
action is repositioned to interact with the PIPBM. This model
is supported by molecular dynamics simulations showing that
PI(4)P binds to the specificity loop prior to engaging the
PIPBM (37). Through the direct visualization of fluorescently
labeled PIP5K on supported lipid bilayers using TIRF micro-
scopy, we found that mutating either the specificity loop or
J. Biol. Chem. (2024) 300(9) 107631 9



B C

5 µm

mNG-5KB mNG-5KB (D266K) mNG-5KB (D350A)mNG-5KB (K138A)

D266

ATP

Mn2+

D350

ATP

Mn2+

K138

ATP

Mn2+

A

Dwell time (seconds)

lo
g

1
0
(1

-C
D

F
)

0 1 2 3 4

0.001

0.01

0.1

1
WT

K138A

D266K

D350A

0 100 200 300 400 500 600

0

5000

10000

15000

20000

25000

Time (seconds)

0 50 100 150 200 250

0

2000

4000

6000

8000

10000

12000

14000

Time (seconds)

m
N

G
-5

K
B

 m
e

m
b

ra
n

e
in

t e
n

si
t y

(A
.U

.)

0 100 200 300 400 500 600

0

10000

20000

30000

40000

50000

60000

Time (seconds)

mNG-5KB

mNG-5KB (K138A)

mNG-5KB (D266K)

mNG-5KB (D350A)

mNG-5KB

mNG-5KB (D266K)

mNG-5KB (K138A)

mNG-5KB (D350A)

mNG-5KB

mNG-5KB (D266K)

mNG-5KB (K138A)

mNG-5KB (D350A)

PI(4,5)P
2

PI(4)P
+PS

PI(4,5)P
2

PI(4)P
D E F

P
I(

4
,5

)P
2
 l

ip
id

s
/

m

P
I(

4
,5

)P
2
 l

ip
id

s
/

m

Figure 5. Deciphering the relationship between PIP5K catalysis and membrane recruitment. A, structure of zPIP5KA (5E3U.pdb) showing the position
of residues that are important for catalysis. Residue numbers represent human PIP5KB amino acids that are homologous to the zPIP5KA residues shown in
the structure. B, representative TIRF-M images showing the localization of mNG-PIP5KB (WT and mutants). C, single molecule dwell time distributions for
mNG-PIP5KB (WT and mutants) measured. See Table 1 for single molecule dwell times (s1) and statistics. B and C, membrane composition: 98% DOPC, 2%
PI(4,5)P2. D and E, lipid kinase activity measured using TIRF-M in the presence of 10 nM PIP5KB (WT and mutants). Production of PI(4,5)P2 was monitored in
the presence of 20 nM Cy3-PLCd. D, membrane composition: 96% DOPC, 4% PI(4)P. E, membrane composition: 78% DOPC, 2% PI(4)P, 20% DOPS. F, bulk
membrane recruitment measured by TIRF-M in the presence of 20 nM mNG-PIP5KB (WT and mutants). Membrane composition: 98% DOPC, 2% PI(4,5)P2.

Molecular basis of PIP5K product recognition
PIPBM disrupts PIP5K lipid kinase activity, membrane local-
ization, and positive feedback. In support of the two-step
membrane recruitment mechanism, we find that mutating
the specificity loop reduces the membrane-binding frequency
(kON) but only moderately affects the dissociation rate constant
(kOFF) of PIP5K. The ability of the PIPBM to facilitate mem-
brane association in the absence of a functional specificity loop
provides support for a two-step membrane recruitment
mechanism. Although membrane recruitment of PIP5K is
more robust in the presence of both a functional specificity
loop and PIPBM, each motif can individually contribute to
membrane binding.

Previous research has established the specificity loop as the
determinative motif that controls substrate specificity of the
10 J. Biol. Chem. (2024) 300(9) 107631
type I and II PIP kinases, PIP4K and PIP5K (16–18, 35).
Studies have identified specific residues in the PIP4K and
PIP5K specificity loops that can be swapped to switch sub-
strate recognition and modulate catalysis (17, 18). The struc-
tural basis of sequence-dependent changes in lipid specificity
PIP4K and PIP5K, however, remains unclear. Our character-
ization of chimeric fusion proteins revealed that the PIP5K
specificity loop can confer specificity for either PI(4)P and
PI(4,5)P2 when attached to another peripheral membrane-
binding protein. Fusing the PIP5KB specificity loop to the
PH domain of PLCd enhanced the ability of the mNG-(5K-
SL)-PLCd chimeric protein to interact with PI(4)P and PI(4,5)
P2 lipids, but not PI(5)P (Fig. 2). These findings support a
model in which the specificity loop has preferred PIP lipid



Molecular basis of PIP5K product recognition
interactions, rather than functioning as a broad lipid mem-
brane sensor.

The PIPBM has previously been described as the active site
of the type I, II, and III PIP lipid kinases (18). The finding that
the PIPBM also strongly binds PI(4,5)P2 lipids raises questions
concerning how PIP5K can exhibit positive feedback based on
product binding when PI(4,5)P2 appears to function as
competitive inhibitor. If PI(4,5)P2 binds to the “active site,” we
expect to observe attenuated lipid kinase activity following
PI(4)P phosphorylation. One mechanism that could resolve
this paradox is through membrane-mediated dimerization of
PIP5K. We previously reported that PIP5K dimerizes in a
density-dependent manner and that dimerization potentiates
lipid kinase activity through a mechanism consistent with
allosteric regulation (34). Molecular dynamics simulations
have shown that only a single kinase domain of a PIP5K dimer
is able to associate with membrane lipids (37). Quantitative
analysis of monomeric and dimeric mNG-PIP5K membrane
surface densities also support a model in which only a single
kinase domain can engage PI(4,5)P2 on a membrane (34). This
may allow dimeric PIP5K to bypass PI(4,5)P2-dependent in-
hibition by toggling membrane engagement of kinase domains
between cycles of catalysis. Additional biochemical and mo-
lecular dynamics simulations are required to understand the
mechanism for bypassing competitive inhibition due to PI(4,5)
P2 binding.

Catalytically dead PIP5K mutants are often used in cellular
PIP5K studies (38, 44–46). Many studies, however, used mu-
tations to the PIPBM to achieve their “catalytically dead”
PIP5K mutants. This is unideal, as we show in this study that
mutating the PIPBM leads perturbs membrane localization.
We sought to identify candidate catalytically dead PIP5K
mutants for the field to use in future in vivo studies. We used
single molecule TIRF microscopy to characterize several
catalytically dead PIP5K mutants (i.e. K138A, D266K, and
D350A) with mutations outside of the PIPBM that have pre-
viously been used in a variety of cell biological studies (38,
44–54). We found that PIP5KB (D266K) was the ideal sepa-
ration of function mutant because it displayed single molecule
membrane-binding properties that were nearly identical to the
WT kinase while showing no discernable activity on PI(4)P
membranes and greatly reduced activity in the presence of a
physiologically relevant concentrations of PS lipids. It is
important to note that in the presence of 20% phosphati-
dylserine, none of the “catalytically dead” PIP5K mutants were
fully inactive. Considering that the cellular environment con-
tains a variety of antagonists, including PIP lipid phosphatases
and PIP4K, the basal level of activity displayed by the PIP5KB
(D266K) mutant will likely be suppressed in vivo.

Previous studies have shown that PIP5K enzymes target the
inner leaflet of the cellular plasma membrane (24, 38). The
prevailing hypothesis in the field is that PIP5K targets the
membrane through direct association with the lipid substrate,
PI(4)P, and interactions with peripheral membrane-binding
proteins (14, 25, 36, 55, 56). However, if PIP5K preferentially
binds to its substrate lipid, PI(4)P, why does it target the
inner leaflet of the plasma membrane and not other PI(4)
P-containing intracellular membranes? Previous characteriza-
tion by immunofluorescence has shown that a large abundance
of cellular PI(4)P exists at the Golgi cisternae (57–59). The
overall lack of strong PIP5K localization to the Golgi suggests
that its substrate, PI(4)P, is not the major lipid interaction that
promotes plasma membrane localization of PIP5K. Our data
provides a different explanation for preferential inner leaflet
targeting. First, we show that PIP5K binds PI(4,5)P2 lipids with
higher affinity than PI(4)P lipids (Fig. 4A), consistent with
previous findings (33, 34). Previous work, however, demon-
strated that disrupting PI(4,5)P2 lipid concentrations di-
minishes but does not fully abolish plasma membrane
localization of PIP5K (24). Our study explains this result by
establishing that PIP5K recruits to membranes containing high
molar concentrations of anionic lipids like PS. However, unlike
PI(4,5)P2 binding, which occurs even at low molar percentages,
PS must be present at high molar concentrations >15% before
we detect robust membrane recruitment of PIP5K (Fig. 4C).
This switch-like recruitment to PS provides a possible expla-
nation for PIP5K inner leaflet targeting. In mammalian cell
membranes, despite PS being present at all intracellular
membranes, it is only highly abundant (>15%) in the inner
leaflet of the plasma membrane (60, 61).

Researchers recently discovered that PIP4K negatively reg-
ulates PIP5K activity, which establishes a homeostatic mech-
anism for maintaining stable PI(4,5)P2 levels at the plasma
membrane (38, 40). Plasma membrane recruitment of PIP4K
requires elevated levels of PI(4,5)P2, where it then attenuates
PIP5K catalytic activity through an unknown mechanism (38).
In vitro experiments confirmed that PIP4K requires a higher
molar concentration of PI(4,5)P2 lipids for membrane locali-
zation compared to PIP5K (38). Why PIP4K had a reduced
sensitivity to PI(4,5)P2 lipids compared to PIP5K was unclear.
Here, we report that the difference in PI(4,5)P2 lipid sensitivity
can be attributed to the divergence in the PIP4K and PIP5K
specificity loop sequences. Swapping specificity loops
enhanced the membrane-binding frequency of PIP4K, even on
SLBs containing 1% PI(4,5)P2. Overall, our results show that
the PIP4K and PIP5K specificity loop is a major factor that
controls sensitivity to PI(4,5)P2 lipids. These findings reveal
intricate mechanisms underlying membrane targeting and
lipid kinase activity within the phosphoinositide signaling
pathway, offering insights into potential avenues for modu-
lating cellular PI(4,5)P2 levels by PIP4K and PIP5K during cell
signaling.
Experimental procedures

Molecular biology

The gene coding the PH domain derived from human
phospholipase C-d1 (PLCd Accession #P51178.2), human
phosphatidylinositol 4-phosphate 5-kinase type-1 beta
(hPIP5KB; Uniprot #O14986), and human phosphatidylinosi-
tol 5-phosphate 4-kinase type-2 beta (hPIP4KB; Uniprot
#P78356) were derived from codon-optimized genes synthe-
sized by GeneArt (Invitrogen). The gene encoding mEos3.2
(62) was PCR-amplified and cloned in-frame with PIP5KB to
J. Biol. Chem. (2024) 300(9) 107631 11



Molecular basis of PIP5K product recognition
create a N-terminal fusion for mammalian cell expression.
Gene sequences were subcloned into either bacterial, insect
cell, or mammalian expression vectors using Gibson assembly
(63). Plasmids containing PIP5KB mutations (i.e. W365A,
W365P, D51R, K205A, R211A, K138A, D350A, D266K, etc.)
were generated through site-directed mutagenesis using the
PfuUltra High-Fidelity DNA polymerase (Agilent, cat#
600380). Chimeric mNG-PIP5K, mNG-PLCd, and mNG-
Specificity Loop-PLCd were created using PCR amplification
of genes via AccuPrime Pfx Master Mix (Thermo Fisher Sci-
entific, Cat#12344040) then combined with a digested plasmid
using Gibson Assembly. The complete ORF of all vectors used
in this study were sequenced by Azenta (formerly Genewiz)
and Plasmidsaurus (University of Oregon) to ensure constructs
lacked deleterious mutations. Each protein expression
construct was screened for optimal yield and solubility in
either bacteria (BL21 DE3 Star, Rosetta, etc.) or Spodoptera
frugiperda (Sf9) insect cells. See Supplementary file for the
amino acid sequences used during recombinant protein
expression or transient transfection in HEK293T cells. Note
that there have previously been inconsistencies in nomencla-
ture between human and mouse PIP5K paralogs. In this
manuscript, PIP5KB refers to the human PIP5KB paralog. All
plasmids created and used within this manuscript function
under this human nomenclature.
Protein purification

PIP5KB and mNG-PIP5K

WT and mutant PIP5K proteins were recombinantly
expressed and purified as previously described (33, 34). Gene
sequences encoding PIP5KB and mutants (i.e. W365P, D51R,
and K205A/R211A) were cloned into FastBac1 vectors in
frame with an N-terminal his6-MBP-TEV-GGGGG or his6-
TEV-mNG-GGGGG and expressed under the polyhedrin (pH)
promoter. BACMIDS and baculoviruses were generated as
previously described (33). For large protein expression, high
five cells were infected with baculovirus using an optimized
multiplicity of infection, typically 2% vol/vol. Infected insect
cells were grown for 48 h at 27 �C in ESF 921 serum-free insect
cell culture medium (Expression Systems, Cat# 96-001-01).
Cells were then harvested by centrifugation, washed with 1×
PBS [pH 7.2], resuspended in cell storage buffer (1× PBS [pH
7.2], 10% glycerol, 2× Sigma protease inhibitor table), and then
stored in the −80 �C freezer. For purification, frozen insect cell
pellets harvested from 2 to 4 L of liquid cell culture were
thawed at room temperature in a water bath and lysed into
buffer containing 50 mM Na2HPO4 [pH 8.0], 10 mM imid-
azole, 400 mM NaCl, 1 mM PMSF (added twice, once before
homogenization and once after), 5 mM beta-mercaptoethanol
(BME), 100 mg/ml DNase, SIGMAFAST protease inhibitor
cocktail tablets, EDTA-free (Sigma, Cat# S8830-20TAB) per
100 ml lysis buffer. Cells in this buffer were lysed using a
dounce homogenizer. Lysate was clarified by centrifugation at
36,000 rpm (140,000g) for 60 min under vacuum using a
Beckman Ti-45 rotor at 4 �C. Lysate was then batch bound to
5 ml of Ni-NTA Agarose (Qiagen, Cat# 30230) resin at 4 �C
12 J. Biol. Chem. (2024) 300(9) 107631
for 2 h in a beaker set on a stir plate. Resin was then collected
in 50 ml tubes, centrifuged, and washed with buffer containing
50 mM Na2HPO4 [pH 8.0], 10 mM imidazole, 400 mM NaCl,
and 5 mM BME and centrifuged again before being transferred
to gravity flow column in more wash buffer. Ni-NTA resin
with his6-MBP-TEV-GGGGG-PIP5K bound was then eluted
into buffer containing 500 mM imidazole. Peak fractions were
pooled, combined with 200 mg/ml his6-TEV(S291V) protease,
and dialyzed against 4 L of buffer containing 20 mM Tris [pH
8.0], 200 mM NaCl, 2.5 mM BME for 16 to 18 h at 4 �C.
Dialysate was then combined 1:1 with 20 mM Tris [pH 8.0],
1 mM TCEP (�100 mM NaCl final). Precipitation was
removed by centrifugation and 0.22 mm syringe filtration.
Clarified dialysate was then bound to a MonoS cation ex-
change column (GE Healthcare, Cat# 17-5168-01) equilibrated
in 20 mM Tris [pH 8.0], 100 mM NaCl, 1 mM TCEP buffer.
Proteins were resolved over a 10 to 100% linear gradient
(0.1–1 M NaCl, 45 CV, 45 ml total, 1 ml/min flow rate). PIP5K
homologs and paralogs typically eluted from the MonoS col-
umn in the presence of 370 to 450 mM NaCl. Peak fractions
containing PIP5K were pooled, concentrated in a 30 kDa
MWCO Vivaspin 6 centrifuge tube (GE Healthcare, Cat# 28-
9323-17), and loaded onto a 24 ml Superdex 200 10/
300 Gl (GE Healthcare, Cat# 17-5174-01) size-exclusion col-
umn equilibrated in 20 mM Tris [pH 8.0], 200 mM NaCl, 10%
glycerol, 1 mM TCEP. Peak fractions were concentrated in a
30 kDa MWCO Vivaspin 6 centrifuge tube and snap frozen at
a final concentration of 10 to 40 mM using liquid nitrogen.
GGGGG-PIP5KB (WT and mutants) were labeled with
Alexa647-LPETGG using sortase-mediated peptide ligation as
previously described (33).
PIP4K2B and PIP4K2B (5K-SL)

Codon-optimized gene sequence encoding human PIP4K2B
isoform 2 (Uniprot #P78356) was cloned into a pETM-derived
bacterial expression vector to create the following fusion
protein: his6-SUMO3-GGGGG-PIP4K2B (1–416aa). Recom-
binant PIP4K2B was expressed in BL21(DE3) Star Escherichia
coli as previously described (38). Using 2 to 4 L of terrific
broth, bacterial cultures were grown at 37 �C until A600 = 0.6.
Cultures were then shifted to 18 �C for 1 h to cool down.
Protein expression was induced with 50 mM IPTG and
bacteria-expressed protein for 20 h at 18 �C before being
harvested by centrifugation. For purification, cells were lysed
into buffer containing 50 mM Na2HPO4 [pH 8.0], 400 mM
NaCl, 0.4 mM BME, 1 mM PMSF (add twice, 15 min in-
tervals), DNase, 1 mg/ml lysozyme using a microtip sonicator.
Lysate was centrifuged at 16,000 rpm (35,172g) for 60 min in a
Beckman JA-17 rotor chilled to 4 �C. Lysate was circulated
over 5 ml HiTrap Chelating column (GE Healthcare, Cat# 17-
0409-01) that had been equilibrated with 100 mM CoCl2 for
1 h, washed with MilliQ water, and followed by buffer con-
taining 50 mM Na2HPO4 [pH 8.0], 400 mM NaCl, 0.4 mM
BME. Recombinant PIP4K2B was eluted with a linear gradient
of imidazole (0–500 mM, 8 CV, 40 ml total, 2 ml/min flow
rate). Peak fractions were pooled, combined with 50 mg/ml of



Molecular basis of PIP5K product recognition
his6-SenP2 (SUMO protease), and dialyzed against 4 L of
buffer containing 25 mM Na2HPO4 [pH 8.0], 400 mM NaCl,
and 0.4 mM BME for 16 to 18 h at 4 �C. Following overnight
cleavage of the SUMO3 tag, dialysate containing his6-SUMO3,
his6-SenP2, and GGGGG-PIP4K2B was recirculated for at
least 1 h over a 5 ml HiTrap(Co+2) chelating column. Flow-
through containing GGGGG-PIP4K2B was then concen-
trated in a 30 kDa MWCO Vivaspin 6 before loading onto a
Superdex 200 size-exclusion column equilibrated in 20 mM
Hepes [pH 7], 200 mM NaCl, 10% glycerol, 1 mM TCEP. In
some cases, cation exchange chromatography was used to in-
crease the purity of GGGGG-PIP4K2B before resolving on the
Superdex 200 column. In those cases, we equilibrated a MonoS
column 20 mM Hepes [pH 7], 100 mM NaCl, 1 mM TCEP
buffer. PIP4K2B (pI = 6.9) bound to the MonoS was resolved
over a 10 to 100% linear gradient (0.1–1 M NaCl, 30 CV, 30 ml
total, 1.5 ml/min flow rate). Peak fractions collected from the
Superdex 200 were concentrated in a 30 kDa MWCO Vivaspin
6 centrifuge tube and snap frozen at a final concentration of 20
to 80 mM using liquid nitrogen.

Codon-optimized gene sequence encoding human PIP4K2B
isoform 2 (Uniprot #P78356) was modified by PCR-based
insertion to swap the PIP4K2B specificity loop (372–384aa;
DTKKKAAHAAKTVKHGAGAEI) for the PIP5KB specificity
loop (353–373aa; RLMKKLEHSWKALVYDGDTV). This
produced a chimeric PIP4K2B enzyme with a PIP5K specificity
loop, which is referred to as PIP4K2B (5K-SL). The his6-
SUMO3-GGGGG-PIP4K2B (5K-SL) protein was purified as
described above for WT PIP4K2B. Both PIP4K2B and
PIP4K2B (5K-SL) were labeled on an N-terminal GGGGG
motif with Cy5-LPETGG using sortase-mediated peptide
ligation (33, 34, 64). Following labeling, Cy5-PIP4KB and Cy5-
PIP4KB (5K-SL) were resolved on a Superdex 200 size-
exclusion column equilibrated in 20 mM Hepes [pH 7],
200 mM NaCl, 10% glycerol, 1 mM TCEP. This removed free
Cy5-LPETGG, Sortase (16 kDa), and a small amount of
aggregated Cy5-PIP4KB. Peak fractions were pooled from the
Superdex 200 column and concentrated using a 30 kDa
MWCO Vivaspin 6 centrifuge tube before being snap frozen in
liquid nitrogen at a final concentration of 10 to 20 mM.
PLCd-PH domain

This protein was expressed and purified as previously
described (33). Briefly, human PLCd-PH domain (11–140aa)
was expressed in BL21 (DE3) Star E. coli as a his6-SUMO3-
(Gly)5-PLCd (11–140aa) fusion protein. Following growth at
37 �C in terrific broth to an A600 of 0.8, cultures were shifted to
18 �C for 1 h, induced with 0.1 mM IPTG, and allowed to
express protein for 20 h at 18 �C before being harvested. Cells
were lysed into 50 mM Na2HPO4 [pH 8.0], 300 mM NaCl,
0.4 mM BME, 1 mM PMSF, 100 mg/ml DNase using a
microfluidizer. Lysate was then centrifuged at 16,000 rpm
(35,172g) for 60 min in a Beckman JA-17 rotor chilled to 4 �C.
Lysate was circulated over 5 ml HiTrap chelating column (GE
Healthcare, Cat# 17-0409-01) charged with 100 mM CoCl2 for
1 h. Bound protein was then eluted with a linear gradient of
imidazole (0–500 mM, 8 CV, 40 ml total, 2 ml/min flow rate).
Peak fractions were pooled, combined with SUMO protease
(50 mg/ml final concentration), and dialyzed against 4 L of
buffer containing 50 mM Na2HPO4 [pH 8.0], 300 mM NaCl,
and 0.4 mM BME for 16 to 18 h at 4 �C. Dialysate containing
SUMO-cleaved protein was recirculated for 1 h over a 5 ml
HiTrap chelating column. Flow-through containing (Gly)5-
PLCd (11–140aa) was then concentrated in a 5 kDa MWCO
Vivaspin 20 before being loaded on a Superdex 75 size-
exclusion column equilibrated in 20 mM Tris [pH 8.0],
200 mM NaCl, 10% glycerol, 1 mM TCEP. Peak fractions
containing (Gly)5-PLCd (11–140aa) were pooled and concen-
trated to a maximum concentration of 75 mM (1.2 mg/ml)
before snap freezing with liquid nitrogen and storage at −80
�C. As previously described, GGGGG-PLCd (11–140aa) was
labeled with either AF488 or Cy3 using sortase-mediated
peptide ligation (33, 34).

mNG-PLCd-PH and mNG-(5K-SL)-PLCd-PH

Genes encoding chimeric mNeonGreen-hPLCd-PH domain
(11–140aa) and mNeonGreen-hPIP5K Specificity Loop
(353–373aa)-hPLCd-PH domain (11–140aa) were cloned into
bacterial expression vectors. Each construct was expressed in
BL21 (DE3) Star E. coli as his10-TEV fusion proteins. Bacteria
were grown at 37 �C in terrific broth to an A600 of 0.8. Cultures
were induced with 0.1 mM IPTG and allowed to express
protein for 20 h at 18 �C before being harvested. Cells were
lysed into 50 mM Na2HPO4 [pH 8.0], 400 mM NaCl, 0.5 mM
BME, 1 mM PMSF, 100 mg/ml DNase using tip sonication
(45% amplitude, 5 s on, 10 s off). Lysate was then centrifuged
at 16,000 rpm (35,172g) for 60 min in a Beckman JA-17 rotor
chilled to 4 �C. Lysate was circulated over 5 ml HiTrap
chelating column (GE Healthcare, Cat# 17-0409-01) charged
with 100 mM CoCl2 for 1 h. Bound protein was then eluted
with a linear gradient of imidazole (0–500 mM, 8 CV, 40 ml
total, 2 ml/min flow rate). Peak fractions were pooled and
combined with 200 mg/ml his6-TEV(S291V) protease and
dialyzed against 4 L of buffer containing 50 mM Na2HPO4 [pH
8.0], 400 mM NaCl, and 0.4 mM BME for 16 to 18 h at 4 �C.
Dialysate containing cleaved His10-TEV protein was recircu-
lated for 1 h over a 5 ml HiTrap chelating column. Flow-
through containing mNG-PLCd or mNG-(5K-SL)-PLCd were
then concentrated in a 50 kDa MWCO Vivaspin 20 before
being loaded on a Superdex 75 size-exclusion column equili-
brated in 25 mM Tris [pH 8.0], 150 mM NaCl, 10% glycerol,
1 mM TCEP. Peak fractions containing either mNG-PLCd or
mNG-(5K-SL)-PLCd were pooled and concentrated to 4 to
20 mM before snap freezing with liquid nitrogen and storage
at −80 �C.

Preparation of small unilamellar vesicles

The following lipids were used to generated small unilamellar
vesicles (SUVs): 1,2-dioleoyl-sn-glycero-3-phosphocholine
(18:1 DOPC, Avanti # 850375C), L-a-phosphatidylinositol-4-
phosphate (Brain PI(4)P, Avanti Cat# 840045X), 1,2-dioleoyl-
sn-glycero-3-phospho-(10-myo-inositol-50-phosphate) (PI(5)P,
J. Biol. Chem. (2024) 300(9) 107631 13



Molecular basis of PIP5K product recognition
Avanti Cat# 850152P), L-a-phosphatidylinositol-4,5-
bisphosphate (Brain PI(4,5)P2, Avanti Cat# 840046X), D-myo-
phosphatidylinositol 3,4,5-trisphosphate (PI(3,4,5)P3 diC16,
Echelon Cat# P-3916-100mg), D-myo-phosphatidylinositol 3,4-
bisphosphate (PI(3,4)P2 diC16, Echelon Cat# P-3416-100mg),
1,2-dioleoyl-sn-glycero-3-phospho-L-serine (18:1 DOPS,
Avanti Cat# 840035C), 1,2-dioleoyl-sn-glycero-3-
phosphoethanolamine-N-[4-(p-maleimidomethyl)cyclo-
hexane-carboxamide] (18:1 MCC-PE, Avanti Cat# 780201C).
To make liposomes, 2 mmoles of total lipids are combined in a
35 ml glass round bottom flask with 2 ml of chloroform. Lipids
were dried to a thin film using rotary evaporation with the glass
round-bottom flask submerged in a 42 �C water bath. The lipid
filmwas then resuspended in 2ml of PBS [pH 7.2], getting a final
concentration of 1 mM total lipids. All lipid mixtures expressed
as percentages (e.g. 98% DOPC, 2% PI(4)P) are equivalent to
molar fractions. To generate SUVs, 1 mM total lipid mixture
was extruded through a 0.05 mmpore size 19mmpolycarbonate
membrane (Avanti, Cat# 610002) with filter supports (Avanti,
Cat# 610014) on both sides of the polycarbonate membrane.
Extruding hydrated lipids a total of 11 times achieved the
desired SUV diameter of �50 nm.

Preparation of supported lipid bilayers

SLBs are formed on 25 x 75 mm coverglass (IBIDI, Cat#
10812). Coverglass first is cleaned with 2% Hellmanex III
(Thermo Fisher Scientific, Cat# 14-385-864) heated to 60-70
�C in a glass coplin jar. This was incubated for at least 30 min.
Once incubated, the coverglass was washed thoroughly with
MilliQ water. Once cleaned, the coverglass was then etched in
Piranha solution (1:3, hydrogen peroxide:sulfuric acid) for 10
to 15 min the same day SLBs were formed. Once etched,
coverglass was again thoroughly rinsed with MilliQ water
before being rapidly dried with nitrogen glass. Once dried,
glass was adhered to a 6-well sticky-side chamber (IBIDI, Cat#
80608). SLBs were formed by flowing 30 nm SUVs diluted in
PBS [pH 7.2] to a total lipid concentration of 0.25 mM and
incubated for 30 min. IBIDI chambers were then washed with
5 ml of PBS [pH 7.2] to remove nonabsorbed SUVs. Mem-
brane defects are blocked for 15 min with a 1 mg/ml beta
casein (Thermo Fisher Scientific, Cat# 37528) diluted in 1×
PBS [pH 7.4]. Before use as a blocking protein, frozen 10 mg/
ml beta casein stocks were thawed, centrifuged for 30 min at
21,370g, and 0.22 mm syringe filtered. After blocking SLBs with
beta casein, membranes were washed again with 1 ml of PBS,
followed by 1 ml of kinase buffer before TIRF-M.

Membrane conjugation of SpyCatcher

Following beta-casein blocking of SLBs, membranes that
contained MCC-PE lipids were washed into 1× PBS [pH 7.2]
containing 0.1 mM TCEP. MCC-PE lipids were used to
covalently couple SpyCatcher protein onto SLBs. For these
SLBs, 100 ml of 30 mM SpyCatcher diluted in a 1× PBS [pH 7.2]
and 0.1 mM TCEP buffer was added to the IBIDI chamber and
incubated for 2 h at 23 �C. Once the coupling period passed,
SLBs with MCC-PE lipids were then washed with 2 ml of 1×
14 J. Biol. Chem. (2024) 300(9) 107631
PBS [pH 7.2] containing 5 mM BME and incubated in this
buffer for 15 min to quench the unreacted maleimide head-
groups. SLBs were then washed with 2 ml of 1× PBS to remove
unbound protein. Membranes were stored for up to 2 h in 1×
PBS before being washed into imaging buffer to initiate TIRF-
M measurements.

Assay for measuring the kinetics of PI(4,5)P2 production

The kinetics of PI(4)P phosphorylation were measured on
SLBs formed in IBIDI chambers and visualized using TIRF
microscopy as previously described (33, 34). Imaging or reac-
tion buffer contained 20 mM Hepes [pH 7.0], 150 mM NaCl,
1 mM ATP, 5 mM MgCl2, 0.5 mM EGTA, 20 mM glucose,
200 mg/ml beta casein (Thermo Fisher Scientific, Cat# 37528),
20 mM BME, 320 mg/ml glucose oxidase (Serva, Cat# 22780.01
Aspergillus niger), 50 mg/ml catalase (Sigma, Cat# C40-100 MG
Bovine Liver), and 2 mM Trolox (UV treated as previously
described by Hansen et al. 2019). Perishable reagents (i.e.
glucose oxidase, catalase, and Trolox) were added 5 to 10 min
before image acquisition. For all experiments, we monitored the
change in PI(4)P or PI(4,5)P2 membrane density using a solu-
tion concentration of 20 nM AF488-GGGGG-PLCd or Cy3-
GGGGG-PLCd. Density of PIP lipids (lipids/mm2) was calcu-
lated assuming a footprint of 0.72 nm2 for DOPC lipids (33, 65).

Cell culture and live cell imaging

(HEK) 293T cells (ATCC, #CRL-3216) were cultured in
DMEM + GlutMAX + High Glucose (4.5 g/L) + sodium py-
ruvate (110 mg/L) (Life Technologies, cat #10569010) sup-
plemented with 10% FBS (Sigma, cat# F4135-500 Ml),
penicillin (100 units/ml), and streptomycin (100 mg/ml). This
cell line was acquired from the Barker Hall tissue culture fa-
cility at the University of California at Berkeley. Cells were
grown in 10 cm dishes in humidified incubators at 37 �C in the
presence of 5% CO2 and split at a confluency of 80 to 90%
every 2 to 3 days. HEK293T cells were split using using 1.5 ml
of 0.25% Trypsin. Trypsin was that quenched with 8.5 ml
complete DMEM media containing 10% FBS. Cells were
diluted 1:10 and seeded on a new 10 cm dish containing a total
volume of 10 ml complete DMEM media warmed to 37 �C.

HEK293T cells were prepared for imaging by seeding cells
at a confluency of 30% in an 18-well glass bottom chamber
(IBIDI, Cat# 81817) 48-h before imaging. Cells were trans-
fected 24-h before imaging with 0.1 mg of plasmid DNA using
Lipofectamine 2000 (Thermo Fisher Scientific, Cat#
11668027). Prior to imaging cells, the complete DMEM media
containing phenol red was aspirated from the chamber and
replaced with 1× HBSS (20 mM HEPES [pH 7.2], 150 mM
NaCl, 4 mM KCl, 1 mMMgCl2, 10 mM glucose). The chamber
was mounted on an inverted Nikon Ti2 microscope with an
OKO Touch heated stage. Cells were imaged through a 100×
Nikon objective (1.49 NA) oil immersion TIRF objective with a
heated flexible collar with temperature control. The correction
collar on the 100× objective was adjusted for 37 �C imaging.

Single molecule imaging of PIP5KB in HEK293T utilized an
N-terminal fusion of monomeric Eos3.2 (62), referred to as



Molecular basis of PIP5K product recognition
mEos throughout the manuscript. Photoconversion of mEos-
PIP5KB from the green (516 nm lpeak) to the red fluorescent
(580 nm lpeak) state was accomplished by exposing cells in the
desired field of view to a 50 ms pulse of 0.5 to 1 mW of 405 nm
light. The intensity of 405 nm light was measured through the
objective using a Newport light meter. Following photo-
conversion, single cells were imaged for 30 to 60 s with a time
resolution of 52 to 102 ms to visualize the membrane-binding
dynamics of mEos-PIP5KB. To visualize the localization of
mEos-PIP5KB (W365P), we imaged with 52 ms intervals (19
fps), which allowed us to more easily detect transient mem-
brane interactions and rapid diffusion (see Fig. 1, H–K). To
visualize the localization of mEos-PIP5KB (K205, R211A), we
imaged with 102 ms intervals (10 fps), which allowed us cap-
ture longer time scale differences in membrane dwell times
than WT mEos-PIP5KB (see Fig. 3, E–H).

Microscope hardware and imaging acquisition

Single molecule imaging experiments were performed on an
inverted Nikon Ti2 microscope using a 100× Nikon objective
(1.49 NA) oil immersion TIRF objective. The x-axis and y-axis
positions were manually controlled using a Nikon-motorized
stage and joystick. All images were acquired using an iXion
Life 897 EMCCD camera (Andor Technology Ltd). Fluo-
rescently labeled proteins were excited with either a 405 nm,
488 nm, 561 nm, or 637 nm diode laser (OBIS laser diode,
Coherent Inc) controlled by a Vortran laser drive with
acousto-optic tunable filters control. The power output
measured through the objective for single particle imaging was
1 to 3 mW. For dual color imaging of mNG-PIP5K localization
during Cy3-PLCd–monitored PI(4,5)P2 synthesis, samples
were excited with 1 mW 488 nm and 1 mW 561 nm light, as
measured through the objective. Excitation light was passed
through the following dichroic filter cubes before illuminating
the sample: (1) ZT488/647rpc and (2) ZT561rdc (ET575LP)
(Semrock). Fluorescence emission was detected on an ANDOR
EMCCD camera position after a Sutter emission filter wheel
housing the following emission filters: ET525/50M, ET600/
50M, ET700/75M (Semrock). All in vitro experiments were
performed at room temperature (23 �C). Single molecule im-
aging in living cells was performed at 37 �C using an OKO
Touch heated stage and heated objective collar. Microscope
hardware was controlled using Nikon NIS elements.

Single particle tracking

Fluorescent protein detection and tracking was performed
using the ImageJ/Fiji TrackMate plugin (66). Data in the form
of .nd2 files were loaded in ImageJ. Before being analyzed
using TrackMate, data brightness was adjusted for molecules
to be easily identifiable. TrackMate was then used to identify
and track molecular tracks in these steps: Particles were first
identified using the LoG detector option based on brightness
and signal-to-noise ratio. Once identified, particles were
tracked for their full lifetime using the LAP tracker. This LAP
tracker follows molecular displacement as a function of time.
Particle trajectories were filtered based on Track Start
(removed trajectories that began in first frame), Track End
(removed trajectories present in last frame), Duration
(removed trajectories ≤ 2 frames and singular extra-long
tracks), Track displacement (removed immobilized particles
displacement <0.1), and X - Y location (removed particles near
the edge of the images). The TrackMate output files were
analyzed using PRISM 9 (GraphPad) to calculate characteristic
dwell times and diffusion coefficients.

Step size distribution of single particle trajectories were
plotted in Prism as frequency versus step size (mm). For all
analysis presented in this manuscript, the bin size for the step
size distribution equals 0.01 mm. For curving fitting, the step-
size distributions were plotted as probability density versus
step size (mm). This was achieved by dividing the frequency
distribution (i.e. y-axis values) by the bin size (0.01 mm). The
probability density versus step size plots were fit to the
following one- or two-species distributions.

Single species model

f ðrÞ¼ r
2Ds

e
−

�
r2
4Ds

�

Two species model

f ðrÞ¼a
r

2D1s
e
−

�
r2

4D1s

�
þð1−aÞ r

2D2s
e
−

�
r2

4D2s

�

Variables are defined as the D1 = diffusion coefficient
species 1 (mm2/sec), D2 = diffusion coefficient species 2 (mm2/
sec), alpha (s1 = % of species 1, r = step size (mm), s2 = time
interval between steps (sec). Final step size distribution plots
were generated in PRISM graphing software and using the
following equations: (1 species model): f(x) = x/(2*D1*t)*exp(-
(x^2/(4*D1*t))), (2 species model): f(x) = alpha*(x/(2*D1*t)
*exp(-(x^2/(4*D1*t)))) + (1-alpha)*(x/(2*D2*t)*exp(-(x^2/
(4*D2*t)))).

To calculate the single molecule dwell times for Cy5-
PIP4K2B, Cy5-PIP4K2B(5K-SL), AF647-PIP5K (and mu-
tants), mNG-PIP5K (and mutants), mNG-PLCd, and
mNG-(5K-SL)-PLCd, we generated a cumulative distribution
frequency plot using the frame interval as the bin size (e.g.
50 ms). The log10(1-CDF) was plotted against the dwell time
and fit to either a single or double exponential decay curve.

Single exponential model

f ðtÞ¼ eð−x=sÞ

Two exponential model

f ðtÞ¼a � eð−x=s1Þþð1−aÞ � eð−x=s2Þ

Fitting procedure initiated with a single exponential. In
cases of a low-quality single exponential fit, a maximum of two
J. Biol. Chem. (2024) 300(9) 107631 15



Molecular basis of PIP5K product recognition
species model was used. For double exponential fit, alpha (a)
represents the fraction of fast dissociating population of mol-
ecules characterized by the time constant, s1.

Image analysis, curve fitting, and statistics

Image analysis was performed on ImageJ/Fiji. Prism 9
(GraphPad) was used for generating plots, curve fitting, and
statistics. Single molecule dwell time and step size presented
in this manuscript represent combined data from 2 to 3
movies acquired from multiple fields of view for each tech-
nical replicate. Single molecule dwell time errors equal SD
for ≥ 3 technical replicates. Dwell time distributions and
curve fits were typically generated with n = 500 to 3000
particle trajectories from a single movie. Step size distribu-
tion plots and curve fits represent 10,000 to 30,000 measured
displacements. When fitting single particle dwell time dis-
tributions derived from the visualization of mEos-PIP5KB
(WT and mutants) in HEK293T cells, we generated a sin-
gle dwell time distribution for all the molecules tracked in a
single cell. For single molecule data collected in cells, each
data point represents the average dwell time, time constant,
or displacement for all the molecules detected in a single cell
from a single movie (n = 14–27 cells per condition). p values
(Figs. 1, I–K and 3, F and G) were calculated in Prism 9 using
an unpaired two-tailed Student’s t test. In all cases, the dis-
tributions were Gaussian and had similar SDs for
comparison.

Data availability

All the information needed for interpretation of the data is
presented in the manuscript or the supplemental material.
Plasmids related to this work are available upon request.

Supporting information—This article contains supporting informa-
tion (33, 34).

Author contributions—B. R. D. and S. D. H. conceptualization; B. R.
D., K. A. F., and S. D. H. data curation; B. R. D., K. A. F., and S. D. H.
formal analysis; B. R. D., K. A. F., and S. D. H. investigation; B. R. D.
and S. D. H. methodology; B. R. D. and S. D. H. resources; B. R. D.
and S. D. H. validation; B. R. D., K. A. F., and S. D. H. visualization;
B. R. D., K. A. F., and S. D. H. writing–review and editing; B. R. D.
and S. D. H. writing–original draft; S. D. H. funding acquisition; S.
D. H. project administration; S. D. H. software; S. D. H. supervision.

Funding and additional information—Research was supported by
the University of Oregon Start-up funds (S. D. H.), National Science
Foundation CAREER Award (S. D. H., MCB-2048060), Molecular
Biology and Biophysics Training Program (B. R. D, NIH T32
GM007759), and the John Keana Fellowship from the Department
of Chemistry and Biochemistry at the University of Oregon (B. R.
D.). The content of this article is solely the responsibility of the
authors and does not necessarily represent the official views of the
National Science Foundation.

Conflicts of interest—The authors declare that they have no conflicts
of interest with the contents of this article.
16 J. Biol. Chem. (2024) 300(9) 107631
Abbreviations—The abbreviations used are: HEK, human embry-
onic kidney; PH, pleckstrin homology; PI(4)P, phosphatidylinositol-
4-phosphate; PI(4,5)P2, phosphatidylinositol-4,5-bisphosphate; PIP,
phosphatidylinositol phosphate; PIP4K, phosphatidylinositol-5-
phosphate 4-kinase; PIP5K, phosphatidylinositol-4-phosphate 5-
kinase; PIPBM, PIP-binding motif; PS, phosphatidylserine; SLB,
supported lipid bilayer; SUV, small unilamellar vesicle; TIRF-M,
total internal reflection fluorescence microscopy; zPIP5K, zebrafish
PIP5K.

References

1. Balla, T. (2013) Phosphoinositides: tiny lipids with giant impact on cell
regulation. Physiol. Rev. 93, 1019–1137

2. Di Paolo, G., and De Camilli, P. (2006) Phosphoinositides in cell regula-
tion and membrane dynamics. Nature 443, 651–657

3. Hammond, G. R. V., and Burke, J. E. (2020) Novel roles of phosphoino-
sitides in signaling, lipid transport, and disease. Curr. Opin. Cell Biol. 63,
57–67

4. Mitchell, K. T., Ferrell, J. E., and Huestis, W. H. (1986) Separation of
phosphoinositides and other phospholipids by two-dimensional thin-layer
chromatography. Anal. Biochem. 158, 447–453

5. Nasuhoglu, C., Feng, S., Mao, J., Yamamoto, M., Yin, H. L., Earnest, S.,
et al. (2002) Nonradioactive analysis of phosphatidylinositides and other
anionic phospholipids by anion-exchange high-performance liquid
chromatography with suppressed conductivity detection. Anal. Biochem.
301, 243–254

6. Wenk, M. R., Lucast, L., Di Paolo, G., Romanelli, A. J., Suchy, S. F.,
Nussbaum, R. L., et al. (2003) Phosphoinositide profiling in complex lipid
mixtures using electrospray ionization mass spectrometry. Nat. Bio-
technol. 21, 813–817

7. Janmey, P. A., Bucki, R., and Radhakrishnan, R. (2018) Regulation of
actin assembly by PI(4,5)P2 and other inositol phospholipids: an update
on possible mechanisms. Biochem. Biophys. Res. Commun. 506,
307–314

8. Rozelle, A. L., Machesky, L. M., Yamamoto, M., Driessens, M. H. E.,
Insall, R. H., Roth, M. G., et al. (2000) Phosphatidylinositol 4,5-
bisphosphate induces actin-based movement of raft-enriched vesicles
through WASP-Arp2/3. Curr. Biol. 10, 311–320

9. Hansen, S. B. (2015) Lipid agonism: the PIP2 paradigm of ligand-gated ion
channels. Biochim. Biophys. Acta 1851, 620–628

10. Huang, C. L., Feng, S., and Hilgemann, D. W. (1998) Direct activation of
inward rectifier potassium channels by PIP2 and its stabilization by
Gbetagamma. Nature 391, 803–806

11. Antonescu, C. N., Aguet, F., Danuser, G., and Schmid, S. L. (2011)
Phosphatidylinositol-(4,5)-bisphosphate regulates clathrin-coated pit
initiation, stabilization, and size. Mol. Biol. Cell 22, 2588–2600

12. Mandal, K. (2020) Review of PIP2 in cellular signaling, functions and
diseases. Int. J. Mol. Sci. 21, 8342

13. Wills, R. C., and Hammond, G. R. V. (2022) PI(4,5)P2: signaling the
plasma membrane. Biochem. J. 479, 2311–2325

14. Honda, A., Nogami, M., Yokozeki, T., Yamazaki, M., Nakamura, H.,
Watanabe, H., et al. (1999) Phosphatidylinositol 4-phosphate 5-kinase is a
downstream effector of the small G protein ARF6 in membrane ruffle
formation. Cell 99, 521–532

15. Jenkins, G. H., Fisette, P. L., and Anderson, R. A. (1994) Type I phos-
phatidylinositol 4-phosphate 5-kinase isoforms are specifically stimulated
by phosphatidic acid. J. Biol. Chem. 269, 11547–11554

16. Kunz, J., Wilson, M. P., Kisseleva, M., Hurley, J. H., Majerus, P. W.,
and Anderson, R. A. (2000) The activation loop of phosphatidylino-
sitol phosphate kinases determines signaling specificity. Mol. Cell 5,
1–11

17. Kunz, J., Fuelling, A., Kolbe, L., and Anderson, R. A. (2002) Stereo-specific
substrate recognition by phosphatidylinositol phosphate kinases is
swapped by changing a single amino acid residue. J. Biol. Chem. 277,
5611–5619

http://refhub.elsevier.com/S0021-9258(24)02132-X/sref1
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref1
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref2
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref2
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref3
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref3
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref3
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref4
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref4
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref4
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref5
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref5
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref5
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref5
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref5
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref6
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref6
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref6
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref6
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref7
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref7
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref7
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref7
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref8
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref8
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref8
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref8
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref9
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref9
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref10
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref10
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref10
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref11
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref11
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref11
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref12
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref12
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref13
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref13
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref14
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref14
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref14
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref14
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref15
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref15
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref15
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref16
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref16
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref16
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref16
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref17
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref17
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref17
http://refhub.elsevier.com/S0021-9258(24)02132-X/sref17


Molecular basis of PIP5K product recognition
18. Muftuoglu, Y., Xue, Y., Gao, X., Wu, D., and Ha, Y. (2016) Mechanism of
substrate specificity of phosphatidylinositol phosphate kinases. Proc. Natl.
Acad. Sci. U. S. A. 113, 8711–8716

19. Tolias, K. F., Rameh, L. E., Ishihara, H., Shibasaki, Y., Chen, J., Pre-
stwich, G. D., et al. (1998) Type I phosphatidylinositol-4-phosphate 5-
kinases synthesize the novel lipids phosphatidylinositol 3,5-
bisphosphate and phosphatidylinositol 5-phosphate. J. Biol. Chem.
273, 18040–18046

20. Zhang, X., Loijens, J. C., Boronenkov, I. V., Parker, G. J., Norris, F. A.,
Chen, J., et al. (1997) Phosphatidylinositol-4-phosphate 5-kinase isozymes
catalyze the synthesis of 3-Phosphate-containing phosphatidylinositol
signaling molecules. J. Biol. Chem. 272, 17756–17761

21. Ishihara, H., Shibasaki, Y., Kizuki, N., Katagiri, H., Yazaki, Y., Asano, T.,
et al. (1996) Cloning of cDNAs encoding two isoforms of 68-kDa type I
phosphatidylinositol4-phosphate 5-kinase. J. Biol. Chem. 271,
23611–23614

22. Ishihara, H., Shibasaki, Y., Kizuki, N., Wada, T., Yazaki, Y., Asano, T.,
et al. (1998) Type I phosphatidylinositol-4-phosphate 5-kinases: cloning
of the third isoform and deletion/substitution analysis of members of this
novel lipid kinase family. J. Biol. Chem. 273, 8741–8748

23. Loijens, J. C., and Anderson, R. A. (1996) Type I phosphatidylinositol-4-
phosphate 5-kinases are distinct members of this novel lipid kinase family.
J. Biol. Chem. 271, 32937–32943

24. Fairn, G. D., Ogata, K., Botelho, R. J., Stahl, P. D., Anderson, R. A., De
Camilli, P., et al. (2009) An electrostatic switch displaces phosphatidyli-
nositol phosphate kinases from the membrane during phagocytosis. J. Cell
Biol. 187, 701–714

25. Pan, W., Choi, S.-C., Wang, H., Qin, Y., Volpicelli-Daley, L., Swan, L.,
et al. (2008) Wnt3a-Mediated formation of phosphatidylinositol 4,5-
bisphosphate regulates LRP6 phosphorylation. Science 321, 1350–1353

26. Coppolino, M. G., Dierckman, R., Loijens, J., Collins, R. F., Pouladi, M.,
Jongstra-Bilen, J., et al. (2002) Inhibition of phosphatidylinositol-4-
phosphate 5-kinase Ia impairs localized actin remodeling and sup-
presses phagocytosis. J. Biol. Chem. 277, 43849–43857

27. Mao, Y. S., Yamaga, M., Zhu, X., Wei, Y., Sun, H.-Q., Wang, J., et al.
(2009) Essential and unique roles of PIP5K-g and -a in Fcg receptor-
mediated phagocytosis. J. Cell Biol. 184, 281–296

28. Di Paolo, G., Pellegrini, L., Letinic, K., Cestra, G., Zoncu, R., Voronov, S.,
et al. (2002) Recruitment and regulation of phosphatidylinositol phos-
phate kinase type 1 gamma by the FERM domain of talin. Nature 420,
85–89

29. Ling, K., Doughman, R. L., Firestone, A. J., Bunce, M. W., and Anderson,
R. A. (2002) Type I gamma phosphatidylinositol phosphate kinase targets
and regulates focal adhesions. Nature 420, 89–93

30. Burke, J. E., Triscott, J., Emerling, B. M., and Hammond, G. R. V. (2023)
Beyond PI3Ks: targeting phosphoinositide kinases in disease. Nat. Rev.
Drug Discov. 22, 357–386

31. Semenas, J., Hedblom, A., Miftakhova, R. R., Sarwar, M., Larsson, R.,
Shcherbina, L., et al. (2014) The role of PI3K/AKT-related PIP5K1a and
the discovery of its selective inhibitor for treatment of advanced prostate
cancer. Proc. Natl. Acad. Sci. U. S. A. 111, E3689–E3698

32. Burke, J. E. (2018) Structural basis for regulation of phosphoinositide
kinases and their involvement in human disease. Mol Cell 71, 653–673

33. Hansen, S. D., Huang, W. Y. C., Lee, Y. K., Bieling, P., Christensen, S. M.,
and Groves, J. T. (2019) Stochastic geometry sensing and polarization in a
lipid kinase–phosphatase competitive reaction. Proc. Natl. Acad. Sci. U. S.
A. 116, 15013–15022

34. Hansen, S. D., Lee, A. A., Duewell, B. R., and Groves, J. T. (2022)
Membrane-mediated dimerization potentiates PIP5K lipid kinase activity.
eLife 11, e73747

35. Liu, A., Sui, D., Wu, D., and Hu, J. (2016) The activation loop of PIP5K
functions as a membrane sensor essential for lipid substrate processing.
Sci. Adv. 2, e1600925

36. Hu, J., Yuan, Q., Kang, X., Qin, Y., Li, L., Ha, Y., et al. (2015) Resolution of
structure of PIP5K1A reveals molecular mechanism for its regulation by
dimerization and dishevelled. Nat Commun 6, 8205

37. Amos, S.-B. T. A., Kalli, A. C., Shi, J., and Sansom, M. S. P. (2019)
Membrane recognition and binding by the phosphatidylinositol
phosphate kinase PIP5K1A: a multiscale simulation study. Structure 27,
1336–1346.e2

38. Wills, R. C., Doyle, C. P., Zewe, J. P., Pacheco, J., Hansen, S. D., and
Hammond, G. R. V. (2023) A novel homeostatic mechanism tunes PI(4,5)
P2-dependent signaling at the plasma membrane. J. Cell Sci. 136,
jcs261494

39. Arioka, M., Nakashima, S., Shibasaki, Y., and Kitamoto, K. (2004) Dibasic
amino acid residues at the carboxy-terminal end of kinase homology
domain participate in the plasma membrane localization and function of
phosphatidylinositol 5-kinase g. Biochem. Biophys. Res. Commun. 319,
456–463

40. Wang, D. G., Paddock, M. N., Lundquist, M. R., Sun, J. Y., Masha-
dova, O., Amadiume, S., et al. (2019) PIP4Ks suppress insulin
signaling through a catalytic-independent mechanism. Cell Rep. 27,
1991–2001.e5