Article

An alternative cell cycle coordinates

multiciliated cell differentiation

https://doi.org/10.1038/s41586-024-07476-z Semil P. Choksi1 ✉, Lauren E. Byrnes1, Mia J. Konjikusic1, Benedict W. H. Tsai1, Rachel Deleon1,
Quanlong Lu2, Christopher J. Westlake2 & Jeremy F. Reiter1,3 ✉
Received: 19 April 2023

Accepted: 26 April 2024

The canonical mitotic cell cycle coordinates DNA replication, centriole duplication
Published online: 29 May 2024
and cytokinesis to generate two cells from one1. Some cells, such as mammalian
Check for updates
trophoblast giant cells, use cell cycle variants like the endocycle to bypass mitosis2.
Differentiating multiciliated cells, found in the mammalian airway, brain ventricles
and reproductive tract, are post-mitotic but generate hundreds of centrioles, each
of which matures into a basal body and nucleates a motile cilium3,4. Several cell cycle
regulators have previously been implicated in specific steps of multiciliated cell
differentiation5,6. Here we show that differentiating multiciliated cells integrate cell
cycle regulators into a new alternative cell cycle, which we refer to as the multiciliation
cycle. The multiciliation cycle redeploys many canonical cell cycle regulators,
including cyclin-dependent kinases (CDKs) and their cognate cyclins. For example,
cyclin D1, CDK4 and CDK6, which are regulators of mitotic G1-to-S progression, are
required to initiate multiciliated cell differentiation. The multiciliation cycle amplifies
some aspects of the canonical cell cycle, such as centriole synthesis, and blocks
others, such as DNA replication. E2F7, a transcriptional regulator of canonical
S-to-G2 progression, is expressed at high levels during the multiciliation cycle. In the
multiciliation cycle, E2F7 directly dampens the expression of genes encoding DNA
replication machinery and terminates the S phase-like gene expression program.
Loss of E2F7 causes aberrant acquisition of DNA synthesis in multiciliated cells
and dysregulation of multiciliation cycle progression, which disrupts centriole
maturation and ciliogenesis. We conclude that multiciliated cells use an alternative
cell cycle that orchestrates differentiation instead of controlling proliferation.

Throughout the body, multiciliated cells propel overlying fluids4,7. previously linked to multiciliated cell differentiation5,6,20. The mitotic
These cells are distinguished by their ability to generate as many as 500 oscillator, comprising CDK1 and anaphase-promoting complex/cyclo-
centrioles, each of which serves as a basal body to template a motile some (APC/C), directs centriole growth and disengagement5. CDK2
cilium3,4,8. activity regulates both early stages of multiciliated cell differentiation
Multiciliated cells differentiate from precursors that progress and ciliogenesis6. In addition, multiciliated cell differentiation depends
through stages I–IV9–12 (Extended Data Fig. 1). Multiciliated cell pre- on cell-cycle-associated transcriptional regulators, such as E2F4, E2F5
cursors induce transcription factors such as MYB9,11,13,14. During stage I, and MYB9,13,21,22. These parallels between the cell cycle and multiciliated
multiciliated cells express proteins required for centriole biogenesis15. cell differentiation led us to propose that an alternative cell-cycle-like
During stage II, specialized structures called deuterosomes generate program may participate in multiciliated cell development.
centrioles16,17. During stage III, these centrioles disengage, migrate,
mature into basal bodies and dock to the apical membrane. During
stage IV, these basal bodies produce motile cilia (Extended Data Fig. 1). Multiciliated cells express cell cycle genes
We investigated how cells orchestrate these complex steps to generate To begin to test this hypothesis and to map transcriptional changes
a functional multiciliated cell. during multiciliated cell differentiation, we performed single-cell RNA
Most cells have two centrioles in the G1 phase of the mitotic cycle, sequencing (scRNA-seq) of differentiating mouse tracheal epithelial
duplicate their centrioles once and only once during S phase and cells (mTECs). We combined cells differentiated by air–liquid interface
distribute the four resulting centrioles equally to their daughters in culture for various lengths of time (Methods) to capture a range of dif-
mitosis18,19. Multiciliated cells bypass the limits of regular centriole ferentiation stages. Marker gene and pseudotime analyses identified a
synthesis. Several regulators of the canonical cell cycle have been differentiation trajectory initiating within basal stem cells, branching

Department of Biochemistry and Biophysics, Cardiovascular Research Institute, University of California, San Francisco, San Francisco, CA, USA. 2Laboratory of Cell and Developmental Signaling,
1

Center for Cancer Research, National Cancer Institute, National Institutes of Health, Frederick, MD, USA. 3Chan Zuckerberg Biohub, San Francisco, CA, USA. ✉e-mail: [email protected];
[email protected]

214 | Nature | Vol 630 | 6 June 2024

 a b c Proliferating stem cells Multiciliated cells
Proliferating stem cells Post-mitotic differentiating 5 0.8
multiciliated cells 3

Ccnd1 expression

Ccnb1 expression
Ccne1 expression
2 0 0
2 1 1

0 0 0
Pseudotime

G1/G0 S G2/M G1/G0 G1/G0 S G2/M G1/G0 G1/G0 S G2/M G1/G0

d Stages Stages Stages

Precursor I II III IV Precursor I II III IV Precursor I II III IV

S
G2/M

CEP43

CEP43

CEP43
G1/
Cell cycle phase

G0

MYB

MYB

MYB
Cyclin D1

Cyclin B1
CDK1
e Centrioles (CEP43) Cilia (ααTUBAc)
((αTUB mCherry–GMNN(1–110) CDT(1–17)–mVenus mCherry–GMNN(1–110) CDT(1–17)–mVenus f Disengagement,
Normalized fluorescence (a.u.) Protein Centriole mmigration and
Stage I Stage II Stage III Stage IV build-up synthesis docking Ciliogenesis

0 Precursor Stage I Stage II Stage III Stage IV

I

II

III

IV
G1/G0 S G2/M G1/G0
e
ag

e
ag

ag
St

ag
St

St

St
Fig. 1 | Differentiating post-mitotic multiciliated cells transit through a Ccnd1, Ccne1 and Ccnb1 across cycle phases for basal stem cells and multiciliated
cell-cycle-like program. a, Cell cycle phase of basal stem cells. Insets show cells. d, Representative immunofluorescence images of differentiating mTECs
uniform manifold approximation and projection (UMAP) plots of the combined stained for centrioles (CEP43), an early multiciliated cell transcription factor
scRNA-seq data of differentiating mTECs, whereby orange indicates Krt5- (MYB) and cell cycle regulators cyclin D1 (left), CDK1 (centre) and cyclin B1
expressing and Trp63-expressing basal stem cells selected for pseudotime (right). n = 3 biological replicates. e, Left, immunofluorescence images of
analysis. Top, subclustered basal stem cells coloured by cluster identity. PIP–FUCCI mTECs stained for centrioles (CEP43) and cilia (acetylated tubulin
Bottom, cell cycle phase predicted by tricycle. Pseudotime infers a trajectory (αTUBAc)). Bottom panels depict mCherry–GMNN(1–110) (magenta) and
(arrow) corresponding to cell cycle phases. b, Identification of cycle phase of CDT(1–17)–mVenus (green) native fluorescence. Stage based on centriole
multiciliated cells. Blue in inset depicts Gmnc-expressing and Foxj1-expressing and cilia status. Right, quantification of nuclear mCherry–GMNN(1–110)
multiciliated cells selected for pseudotime analysis. Top, subclustered fluorescence and nuclear CDT(1–17)–mVenus fluorescence, minimum and
multiciliated cells coloured by cluster identity. Bottom, tricycle analysis reveals maximum normalized and plotted against the stage of multiciliated cell
differentiating multiciliated cell expression reflective of G1/G0 (blue), S (pink) differentiation. Points are means of three biological replicates, with bars
and G2/M phases (yellow-green). Pseudotime infers a trajectory (arrowhead) indicating ± s.e.m. f, Multiciliated cell differentiation, divided into stages and
corresponding to progression from a G1/G0-like phase, through an S-like phase corresponding phases of the multiciliation cycle. Scale bars, 5 μm (d,e).
and a G2/M-like phase to a second G1/G0-like phase. c, Expression profiles of

in intermediate cells and ending in differentiated multiciliated cells Extended Data Fig. 2e,f). Mature multiciliated cells displayed gene
(Extended Data Fig. 2a–c and Supplementary Table 1), similar to those expression patterns reflective of a second G1/G0-like phase (Fig. 1b
previously described for the upper airway11,20,23,24. Intermediate cells, and Extended Data Fig. 3). For example, cyclins associated with G1,
so called because of their intermediate position on the differentiation S and G2/M phases were expressed successively across these phases
trajectory, contained the precursors that give rise to mature multi- of multiciliated cell differentiation (Fig. 1c and Extended Data Fig. 3).
ciliated cells11. The multiciliated cell pseudotime trajectory reflected Thus, differentiating multiciliated cells exhibit a cell-cycle-like tran-
the stages of multiciliated cell differentiation 11 (Extended Data scriptional program.
Fig. 2c,e). To establish whether the cell-cycle-like transcriptional program is
Different phases of the canonical cell cycle exhibit distinct gene reflected at the protein level, we examined differentiating mTECs for
expression signatures25. Analyses of basal stem cells (defined by Krt5 cell cycle regulator proteins at different stages of multiciliated cell
expression) for cell-cycle-phase-dependent gene expression revealed differentiation, defined by MYB expression and centriole morphology.
that subsets of stem cells expressed genes reflective of the G1/G0, S and Cyclin D1 was expressed in multiciliated cell precursors and then down-
G2/M phases, consistent with their engagement in the canonical cell regulated during differentiation (Fig. 1d). CDK1 localized to the nucleus
cycle (Fig. 1a and Extended Data Fig. 2d,f). in stage I multiciliated cells and then became cytoplasmic as multicili-
Although multiciliated cells are post-mitotic, we also analysed dif- ated cells matured (Fig. 1d). Cyclin B1 was expressed in stages II and III
ferentiating multiciliated cells for gene expression patterns indicative of differentiating multiciliated cells (Fig. 1d). The dynamic expression
of cell cycle phases9,13. Notably, cells early in the multiciliated lineage and localization of these cell cycle regulators suggested that they may
expressed genes reflective of a G1/G0 phase (Fig. 1b and Extended Data coordinate the stages of multiciliated cell differentiation.
Fig. 3). As multiciliated cells differentiated, they expressed genes reflec- To begin to test this possibility, we transduced differentiating
tive of S phase and, later, genes reflective of G2/M phase (Fig. 1b and mTECs with a reporter of cell cycle activity: the PIP-degron-containing

Nature | Vol 630 | 6 June 2024 | 215

Article
fluorescent ubiquitination-based cell cycle indicator PIP–FUCCI26. However, inhibiting CDK4/6 activity altered gene expression in
PIP–FUCCI uses two fluorescent reporters, CDT1(1–17)–mVenus and intermediate cells. That is, CDK4/6 inhibition reduced the expres-
mCherry–GMNN(1–110), which are degraded at different phases of the sion of several important transcriptional effectors of early multicili-
cell cycle. In proliferating cells, mCherry fluorescence without mVenus ated cell differentiation (for example, Gmnc, Mycl, Myb and Foxj1;
indicates S phase, whereas high levels of both mCherry and mVenus Fig. 2h,i and Supplementary Table 4). These results further sug-
are associated with G2/M, and mVenus with no mCherry expression gest that CDK4/6 activity is required for initiating multiciliated cell
indicates G1/G0 phase26. In multiciliated cells, PIP–FUCCI fluorescence differentiation.
levels were dynamic across the stages of differentiation (Fig. 1e). Stage I CDK4/6 activity also promoted the multiciliated cell expression of
multiciliated cells exhibited some mCherry expression with low lev- Ccne1 (Fig. 2i). Therefore, we proposed that CDK4/6 drives progression
els of mVenus. Stages II and III multiciliated cells showed increased from the G1/G0-like to the S-like phase of the multiciliation cycle. To test
mVenus and mCherry levels. Stage IV multiciliated cells maintained this hypothesis, we compared the cell cycle gene expression profiles of
high mVenus levels but no mCherry (Fig. 1e). multiciliated cells treated with ribociclib or DMSO. CDK4/6 inhibition
Together, these data suggest that multiciliated cells display hallmarks reduced the proportion of cells in the S-like phase of the multiciliation
of cell cycle progression during post-mitotic differentiation (Fig. 1f and cycle (Fig. 2j,k). Thus, CDK4/6 activity is important for progression
Supplementary Table 2). The multiciliated cell precursors exhibited from the G1/G0-like phase to the S-like phase of the multiciliation cycle.
characteristics of G1/G0, including expression of Ccnd1 (which encodes To test whether CDK4/6 activity is sufficient to drive multiciliated
cyclin D1) and of Cdk4 and Cdk6 (Cdk4/6) (Extended Data Fig. 3). The cell differentiation, we overexpressed a GFP-tagged cognate cyclin
stage I and early-stage II multiciliated cells displayed hallmarks of for CDK4/6, cyclin D1, in mTECs. Cyclin D1–GFP increased the propor-
S phase, including Cdk1 and Ccne1 (which encodes the S phase regu- tion of multiciliated cells (Fig. 2l,m). These additional multiciliated
lator cyclin E1) expression. Late-stage II and stage III multiciliated cells cells differentiated with kinetics comparable to control multiciliated
were similar to cells in G2/M, expressing cyclin B1 and both PIP–FUCCI cells (Extended Data Fig. 5i,j). Therefore, cyclin D1–CDK4/6 drives
reporter proteins. Stage IV multiciliated cells returned to a second G1/ the progression of cells from the G1/G0-like to the S-like phase of the
G0-like phase, exemplified by Cdk4/6 expression and high PIP–FUCCI multiciliation cycle to initiate multiciliated cell-specific gene expres-
mVenus-to-mCherry ratios. We call this cell-cycle-like progression sion and multiciliated cell differentiation (Fig. 2n).
during multiciliated cell differentiation the multiciliation cycle.
To evaluate proliferation in differentiating multiciliated cells, we
assessed mTECs for ethynyl-2′-deoxyuridine (EdU) incorporation. The multiciliation cycle is a variant cell cycle
Few differentiating multiciliated cells (marked by FOXJ1 expression27) To begin to discern what distinguishes the multiciliation cycle, we
exhibited signs of DNA synthesis (Extended Data Fig. 4a,b). Assess- compared expression profiles between multiciliated cells and pro-
ment of mTECs for a marker of mitosis, phosphorylated histone H3 liferating stem cells. The expression of many cell cycle genes was
(H3S10P), revealed a sharp decline in the number of mitotic cells at dampened during the multiciliation cycle relative to the canonical
the time of differentiation initiation (Extended Data Fig. 4c–e). These cell cycle (Fig. 3a, Extended Data Fig. 6a,b and Supplementary Table 2).
results confirm previous studies showing that multiciliated cells do not For example, S phase genes encoding DNA replication regulators, such
proliferate during differentiation9,13. Therefore, we hypothesized that as Gmnn, Mcm5 and Gins2, were expressed at lower levels in the mul-
multiciliated cells use a modified cell cycle not for generating daughter ticiliation cycle than in the cell cycle (Fig. 3a). Similarly, expression of
cells but to regulate differentiation. G2/M phase-associated genes encoding the centralspindlin complex,
important for chromosome segregation and cytokinesis (for example,
Kif23, Racgap1 and Ect2)32, were expressed at lower levels during the
CDK4/6 initiate the multiciliation cycle multiciliation cycle than the cell cycle (Fig. 3a).
Ccnd1 and Cdk4/6 were expressed in multiciliated cell precursors To identify candidate multiciliation cycle regulators, we identified
(Figs. 1c,d and 2a). During the canonical cell cycle, D-type cyclins genes expressed at higher levels during the multiciliation cycle than
activate CDK4/6 to direct the progression from early G1/G0 into S during the canonical cell cycle. As expected, multiciliated cells prefer-
phase28–30. To test whether these cell cycle regulators function in entially expressed genes encoding regulators of centriole synthesis,
multiciliated cell differentiation, we treated mTECs with inhibitors including Plk4, Sass6 and Deup1 (Extended Data Fig. 6c). In addition,
of CDK4/6, palbociclib and ribociclib31. Inhibiting CDK4/6 activity we identified several cell cycle regulators enriched in the multicili-
during mTEC differentiation blocked centriole synthesis and ciliogen- ation cycle, including Ccno (which encodes a noncanonical cyclin
esis (Fig. 2b,d and Extended Data Fig. 5a,b). To determine whether previously implicated in multiciliated cell differentiation33), Cdkn1a
the effect of CDK4/6 inhibition was specific to differentiation, we (which encodes the CDK inhibitor p21CIP1) and E2f7 (Fig. 3a, Extended
assessed cell density and found no change in palbociclib-treated or Data Fig. 6d).
ribociclib-treated mTECs relative to vehicle control across these treat- E2f7 encodes an atypical repressor in the E2F transcription factor
ment regimens (Extended Data Fig. 5c). To assess how CDK4/6 activity family that represses S phase gene expression to promote the transition
promotes multiciliated cell differentiation, we examined the expres- from the S to G2 phase in proliferating cells34,35. We proposed that E2F7
sion of early multiciliated cell transcription factors, MYB and FOXJ1, may also regulate progression through the multiciliation cycle. E2F7
in CDK4/6 inhibited cells. CDK4/6 activity was required for MYB and was upregulated specifically in mouse adult tracheal cells undergoing
FOXJ1 expression, which suggests that CDK4/6 activity may be required centriole synthesis (Fig. 3b). Similarly, E2F7 was expressed transiently
early in multiciliated cell differentiation (Fig. 2c,e). during stages I and II of mTEC multiciliated cell differentiation (Fig. 3c).
To further investigate the time at which CDK4/6 activity is required As stages I and II correspond to S-like and early G2/M-like phases of the
during multiciliated cell differentiation, we compared the scRNA-seq multiciliation cycle and as E2F7 promotes the transition from S to G2
profiles of mTECs treated with ribociclib or vehicle control (DMSO) in proliferating cells34,35, we speculated that E2F7 terminates the S-like
(Extended Data Fig. 5d–h). Marker gene analysis identified the major phase of the multiciliation cycle.
airway cell types and confirmed that CDK4/6 inhibition decreased
the differentiation of multiciliated cells (Fig. 2f,g and Supplementary
Table 3). CDK4/6 inhibition did not affect the proportion of other cell E2F7 advances the multiciliation cycle
types, such as intermediate cells, which contain multiciliated cell pre- To investigate the role of E2F7 in the multiciliation cycle, we generated
cursors11 (Fig. 2f,g and Extended Data Fig. 5d–h). mice with mutated E2f7 (Extended Data Fig. 6e). E2f7 –/– mice were

216 | Nature | Vol 630 | 6 June 2024

 a b DMSO Palbociclib Ribociclib c DMSO Ribociclib d
Ccnd1 Cdk4 Cdk6
DMSO Palbociclib Ribociclib

MYB
αTUBAc)
*

Cilia
**

((αTUB
20

Multiciliated
cell (%)
FOXJ1
Expression

10

Centrioles
(CEP43)
0

Nuclei (Hoechst)
****

expressing cells (%)

(Hoechst)
***

MYB- or FOXJ1-
Nuclei 20
Intermediate Multiciliated
cells cells 10
Pseudotime

f g h i
Intermediate cells DMSO Ribociclib
e
ted
t
m
dia

ry
ste

lia

eto
me

ici Gmnc Mycl Ccne1

DMSO

al

cr
er

ult

0.2
s

Se
Ba

Int

1.2
log2[fold change (ribociclib/DMSO)]

NS NS NS 0.4
* * * 200

Expression
–log10(P value)
Intermediate
cells Multiciliated
0 Mycl
cells
100
Foxj1 Gmnc
0
0
Ribociclib

0
Myb
0 Basal Multiciliated
–3 0 3 stem cells cells
log2(fold change)
–2

j Cycle phase
k l NLS–GFP Cyclin D1–GFP m *
40
(αTUBAc) CEP43
0

/M
/G

Multiciliated
G1

G2

NLS–GFP
S

cell (%)
*** *** 20 Cyclin D1–GFP
DMSO

log2[fold change (ribociclib/DMSO)]

Cilia
GFP (αTUB

0 0
G1/G0
n
S
Cyclin D1
G2/M
CDK4/6
–1
Nuclei (Hoechst)
Ribociclib

–2 Precursor Stage I Stage II St

Stage III Stage IV
G1/G0 S G2/M
G2/ G1/G0

Fig. 2 | Cyclin D1–CDK4/6 initiates multiciliated cell differentiation. quasi-likelihood F-test with Benjamini–Hochberg correction). h, Genes
a, Minimum and maximum normalized scRNA-seq expression across differentially expressed by ribociclib-treated intermediate cells compared
multiciliated cell differentiation pseudotime. Grey bars indicate 95% with DMSO-treated cells. Red, fold-change > 1.5 and P < 0.0005 (two-tailed
confidence intervals. Coloured x-axis indicates cluster identity. b, mTECs Wald test with Benjamini-Hochberg correction). i, scRNA-seq expression across
treated with DMSO, palbociclib or ribociclib and stained for centrioles (CEP43), multiciliated cell differentiation pseudotime (DMSO or ribociclib). Grey bars
cilia (αTUBAc) and nuclei. c, mTECs treated with DMSO or ribociclib and stained indicate 95% confidence intervals. Colours indicate cluster identity. j, scRNA-
for MYB, FOXJ1 and nuclei. d, Percentage of multiciliated cells after DMSO, seq UMAP of tricycle-based cycle phase after DMSO or ribociclib treatment.
palbociclib or ribociclib treatment. Horizontal lines indicate means ± s.e.m. k, Change in cycle phase proportion after ribociclib treatment. Bars indicate
of 3 biological replicates, **P = 0.0118, *P = 0.0183 (one-way analysis of variance means ± s.e.m. of 3 biological replicates, ***log 2(fold-change) < −0.5 and FDR <
(ANOVA) with Dunnet’s correction). e, Percentage of MYB-expressing or FOXJ1- 0.0008 (two-tailed moderated t-test with Benjamini–Hochberg correction).
expressing cells after DMSO, palbociclib or ribociclib treatment. Horizontal l, mTECs expressing NLS–GFP or cyclin D1–GFP stained for centrioles (CEP43)
lines indicate means ± s.e.m. of 3 biological replicates, ***P = 0.000002, and cilia (αTUBAc). Bottom panels show GFP and nuclei staining. m, Proportion
****P = 0.0000007 (one-way ANOVA with Dunnet’s correction). f, scRNA-seq of multiciliated cells after cyclin D1–GFP expression. Horizontal lines indicate
UMAP of mTEC intermediate and differentiating multiciliated cells after DMSO means ± s.e.m. of 3 biological replicates, *P = 0.0102 (unpaired two-tailed t-test).
or ribociclib treatment. Arrow indicates differentiation trajectory. g, Change in n, Cyclin D1–CDK4/6 promotes the transition of precursor cells from the G1/G0-
cell cluster proportion after ribociclib treatment. Bars indicate means ± s.e.m. like phase to S-like phase of the multiciliation cycle. Scale bars, 10 μm (b,c,l).
of 3 biological replicates, *false discovery rate (FDR) < 0.02 (two-tailed Bayes

viable, as previously reported35, and mTECs generated from mutant many genes encoding regulators of DNA replication were upregulated
mice did not produce detectable levels of wild-type E2f7 transcripts in stages II and III E2f7 –/– multiciliated cells (Fig. 3e, Extended Data
(Extended Data Fig. 6f,g). Both E2f7 –/– mTECs and trachea exhibited Figs. 7h,i and 8a and Supplementary Table 7). For example, genes
attenuated E2F7 protein (Extended Data Fig. 6h–j). To identify genes encoding DNA replication machinery, which were expressed at low
regulated by E2F7 in multiciliated cells, we analysed mTECs from E2f7 –/– levels in the multiciliation cycle (for example, Mcm5 and Gins2; Fig. 3a),
and littermate controls using scRNA-seq (Extended Data Fig. 7a–g and were derepressed in multiciliated cells lacking E2F7 (Fig. 3e,f). We
Supplementary Table 5). We observed no alteration in the expression conclude that in the multiciliation cycle, E2F7 restricts the expression
of multiciliated cell transcription factors or genes involved in centriole of genes expressed during the late G1 and S phases of the canonical
synthesis or ciliogenesis (Fig. 3d and Supplementary Table 6). Instead, cell cycle (Extended Data Fig. 8a).

Nature | Vol 630 | 6 June 2024 | 217

Article
a b d E2f7+/+ E2f7–/–
f g
Proliferating stem cells Multiciliated cells
NS NS NS E2f7+/+ E2F7–GFP

CEP43
E2f7–/– 64
Centralspindlin Cell cycle 0.2

Expression score

Mcm5 expression
DNA replication complex regulators 0.4

Gmnn Kif23 Ccno 0

RPGC
Nuclei
2 6 0.2 NLS–GFP
2 64

0 0

E2F7
0 0 0 Multiciliated cell Centriolar Ciliary 0
transcription genes genes Multiciliated cell Mcm5
Gins2 Racgap1 Cdkn1a differentiation
1 factors
Expression

2 20 E2F7–GFP
c Stage I Stage II Stage III Stage IV e 104
100 Hells
5

CEP43
0.6

Gins2 expression
0 0
0
Mcm7 Mcm4

RPGC
–log10(P value)
Mcm5 Ect2 E2f7 Stmn1 NLS–GFP
2 1.5 Pcna 104
1.5 50

Nuclei
Mcm3
Gins2 Cdt1
E2f1 0 0
0 0 0 0 Mcm5 Multiciliated cell
E2F7
Gins2
S

0
M

differentiation
0

–2.5 0 2.5 5.0

G

G
G

G
2/

2/

2/

Gse1
1/

1/

1/
1/

1/

1/
G

G
G

G
G

log2[fold change (relative expression)]

h Nuclei EdU FOXJ1 i j S phase score G2/M phase score k

E2f7+/+
0.08 0.15 S G2/M

Expression score
1

80 **
E2f7+/+

E2f7+/+

0 0
EdU+ multiciliated cells (%)

60
0
Multiciliated cell differentiation
40 Intermediate (pseudotime)
cells Multiciliated
E2f7–/–
20 cells
S
G2/M
0 1

Expression score
E2f7–/–

E2f7–/–
/–
7 +/ ol

f7 –
2f tr
+
)
(E on

E2
C

0
Multiciliated cell differentiation
(pseudotime)

Fig. 3 | E2F7 blocks DNA replication and promotes progression through NLS–GFP CUT&RUN in mTECs, presented as reads per genomic content
the multiciliation cycle. a, Average read counts across cycle phase of stem (RPGC). h, E2f7 +/+ and E2f7 –/– mTECs stained for EdU and FOXJ1. White arrows
(orange) and multiciliated (blue) cells from the wild-type scRNA-seq dataset indicate E2f7 –/– cells that express FOXJ1 and are EdU-positive. i, Percentage of
(Fig. 1). b, Representative image of adult mouse trachea immunostained for E2f7 +/+ and E2f7 –/– mTECs expressing FOXJ1 that are EdU-positive. Horizontal
E2F7, centrioles (CEP43) and nuclei. n = 3 mice. c, mTECs immunostained for lines indicate means ± s.e.m. of 3 biological replicates, **P = 0.001 (unpaired
E2F7, centrioles (CEP43) and nuclei. Cells representative of each stage from two-tailed t-test). j, S phase (left) and G2/M phase (right) gene signature scores
three biological replicates are shown. d, Composite expression of multiciliated derived from normalized sum rank of gene sets projected onto UMAPs of
cell transcription factors, centriolar and ciliary genes in E2f7 +/+ and E2f7 –/– E2f7 +/+ and E2f7 –/– multiciliated cells. Colours indicate expression score of S
multiciliated cells. Scores are the normalized Mann–Whitney U-statistic of or G2/M gene sets. Arrow indicates differentiation trajectory. Red and black
gene set expression. NS, not significant (multiple unpaired two-tailed t-tests arrowheads indicate the end of the S and G2/M phases, respectively, as defined
with Holm–Sidak correction). e, Genes differentially expressed between E2f7 –/– by the half-maximal phase score. k, E2f7 +/+ and E2f7 –/– multiciliated cell
and E2f7 +/+ differentiating multiciliated cells. Blue, fold-change > 1.5 and minimum and maximum normalized S phase and G2/M phase gene signature
P < 0.00001 (two-tailed Wald test with Benjamini–Hochberg correction). scores across pseudotime. Grey bars indicate 95% confidence intervals Scale
f, scRNA-seq expression in E2f7 +/+ and E2f7 –/– multiciliated cells across bars, 10 μm (b,h), 5 μm (c) or 1 kb (g).
pseudotime. Grey bars indicate 95% confidence intervals. g, E2F7–GFP or

To identify which genes E2F7 might directly regulate in the multicilia- few E2f7 +/+ multiciliated cells showed EdU incorporation (Fig. 3h,i). By
tion cycle, we determined where E2F7–GFP binds the genome in mTECs contrast, E2f7 –/– mTECs showed an 8.7-fold increase in the percentage
using cleavage under targets and release using nuclease (CUT&RUN)36. of multiciliated cells incorporating EdU (Fig. 3h,i). Thus, E2F7 restricts
E2F7–GFP bound near genes encoding DNA replication machinery, DNA replication gene expression to repress DNA synthesis during the
including Mcm5 and Gins2 (Fig. 3g, Extended Data Fig. 8a–c and Sup- multiciliation cycle, and loss of E2F7 restores aspects of the canonical
plementary Table 8), which suggested that E2F7 directly represses cell cycle to the multiciliation cycle.
DNA replication-related genes during the multiciliation cycle S-like Proliferating cells display a sharp cessation of S phase gene expres-
phase. Comparing E2F7 target genes in mTECs to previously identified sion at the transition from S phase to G2/M phase37. Similarly, wild-type
E2F7 target genes in proliferating cells revealed partial overlap, which differentiating multiciliated cells displayed a sharp cessation of S-like
indicated that E2F7 regulates a similar set of genes in the multicilia- phase gene expression before the initiation of G2/M-like gene expres-
tion cycle and the canonical cell cycle (Extended Data Fig. 8d,e and sion (Fig. 3j,k). In addition to increased levels of S phase gene expres-
Supplementary Table 9). sion, differentiating E2f7 –/– multiciliated cells displayed delayed
A key difference between the cell cycle S phase and the multiciliation termination of S phase gene expression (Fig. 3j,k). By contrast, the
cycle S-like phase was DNA synthesis (Extended Data Fig. 4a,b). We timing of G2/M phase-related gene expression was unaffected in
proposed that increased expression of E2F7 in multiciliated cells may differentiating E2f7 –/– multiciliated cells (Fig. 3j,k). Thus, without
be responsible for this difference. To test this hypothesis, we performed E2F7, differentiating multiciliated cells aberrantly express S-like and
EdU labelling of differentiating E2f7 –/– and E2f7 +/+ mTECs. As expected, G2/M phase gene programs simultaneously. This overlap of S-like and

218 | Nature | Vol 630 | 6 June 2024

 a b c d *
NS
3

Nuclei (Hoechst)

intensity, a.u.)
Cilia (αTUBAc
(αTUBAc)
Cilia (αTUB
E2f7+/+

(α
0

/+

–
7 +/ ol

/
f7 –
f7 –
2f tr
+
(αTUBAc)

)
( E on

E2
E2
C
Cilia (αTUB
(α
****

Centriolar area (μm2)

150 NS

Centrioles (CEP43)
Centrioles (CEP43)
E2f7–/–

/+

–
7 +/ ol

/
f7 –
f7 –
2f tr
+
)
(E on

E2
E2
C
e Centrioles (CEP43) CCP110 Centrioles (CEP43) DEUP1 Centrioles (CEP43) CEP164
E2f7+/+
E2f7–/–

f g h i j k

Apically docked centrioles (%)

NS *** NS *** **
*** **** ****
E2f7+/+
(CCP110 intensity per cell, a.u.)

(CEP164 intensity per cell, a.u.)

circular morphology (%)

Stage III–IV multiciliated cells

6
Stage II–III multiciliated cells

60 8
Stage I–II multiciliated cells

(% expressing DEUP1)

Centrioles with
100 100

50 50
0 0 0
E2 +
/–

/+

E2 +
/–

/+

/–

E2 +
/–

/+

–
/

0 0
/

/
f7 –

f7 –

f7 –

f7 –

f7 –

f7 –

E2f7–/–
f7 +

f7 +

f7 +

f7 +

f7 +

f7 +
E2

E2

E2
E2

E2

E2

E2

E2

E2

/–
/–
7 +/ ol

7 +/ ol

f7 –
f7 –
2f tr

2f tr
+
+
)

)
(E on

(E on
E2

E2
Day 7 Day 21 Day 7 Day 21 Day 7 Day 21
C

C
Fig. 4 | E2F7 coordinates centriole synthesis during multiciliated cell P = 0.0689 (one-way ANOVA with Sidak’s correction). g, Percentage of
differentiation. a, Images of E2f7 +/+ and E2f7 –/– littermate mice, with coronal multiciliated cells possessing DEUP1-containing deuterosomes after 7 or
brain sections stained with haematoxylin and eosin. Scale bar, 1mm. 21 days. Horizontal lines indicate means ± s.e.m. of 3 biological replicates,
b, Representative images of trachea immunostained for centrioles (CEP43), ***P = 0.0002, NS indicates P = 0.3808 (one-way ANOVA with Sidak’s correction).
cilia (αTUBAc) and nuclei. Insets show magnification of centrioles in boxed cells. h, Intensities of CEP164 immunofluorescence per multiciliated cell after 7 or
n = 3 mice per genotype. Scale bar, 10 μm. c, Left, mTECs immunostained for 21 days. Horizontal lines indicate means ± s.e.m. of 3 biological replicates,
centrioles (CEP43) and cilia (αTUBAc). Scale bar, 10 μm. Right, magnifications ***P = 0.0088, **P = 0.0015 (one-way ANOVA with Sidak’s correction). i, Left,
of boxed cells. Scale bar, 5 μm. d, Quantification of ciliary intensity (αTUBAc) transmission electron micrographs (TEMs) of multiciliated cells. Scale bar,
and centriolar area in mTECs. Horizontal lines indicate means ± s.e.m. of 3 1 μm. Right, undocked centrioles. Scale bar, 100 nm. j, Percentage of apically
biological replicates, *P = 0.0148, ****P = 0.000017 (one-way ANOVA with docked centrioles in TEMs of multiciliated cells. Horizontal lines indicate
Dunnet’s correction). e, Left, mTECs immunostained for centrioles (CEP43) and means ± s.e.m. of 30 E2f7 +/+ cells (234 centrioles) or 30 E2f7 –/– cells (212
CCP110 (left), deuterosomes (DEUP1, centre) and distal appendages (CEP164, centrioles), ****P < 0.00000001 (unpaired two-tailed t-test). k, Percentage of
right) after 7 days of differentiation. Right, magnifications of boxed cells, with undocked centrioles that are circular in TEMs of multiciliated cells. Horizontal
individual channels in small panels. Scale bars, 5 μm. f, Intensities of CCP110 lines indicate means ± s.e.m. of 18 E2f7 +/+ cells (37 undocked centrioles) or 29
immunofluorescence per multiciliated cell after 7 or 21 days. Horizontal lines E2f7 –/– cells (199 undocked centrioles), ****P = 0.00000001 (unpaired two-tailed
indicate means ± s.e.m. of 3 biological replicates, ***P = 0.0004, NS indicates t-test).

G2/M-like phases in E2f7 –/– multiciliated cells indicates that E2F7 helps numbers of cytoplasmic, undocked centrioles (Fig. 4b and Extended
enforce the transition from the S-like phase to the G2/M-like phase of Data Fig. 9c–g). Similarly, E2f7 –/– mTECs showed fewer cilia per cell and
the multiciliation cycle. clustered, undocked centrioles (Fig. 4c,d). These results suggested that
E2F7 is crucial for multiciliated cell maturation. Several genes encod-
ing cytoskeletal regulators were dysregulated in E2f7 –/– multiciliated
Multiciliated cell differentiation requires E2F7 cells (Extended Data Fig. 8a) and therefore may contribute to centriole
E2f7 –/– mice exhibited hydrocephalus (Fig. 4a and Extended Data docking defects in cells lacking E2F7.
Fig. 9a,b), a condition that can be caused by defective multiciliated In the canonical cell cycle, E2F7 and a paralogue, E2F8, have over-
ependymal cells in the brain38. In E2f7 –/– brain ventricles, oviducts lapping functions35. In differentiating multiciliated cells, E2f8 was
and airways, multiciliated cells exhibited fewer cilia and increased expressed and became further upregulated in E2f7 –/– multiciliated

Nature | Vol 630 | 6 June 2024 | 219

Article
cells (Extended Data Fig. 10a–c). To assess whether E2f8 regulates Stage III:
APC/C Disengagement,
multiciliated cell differentiation, we used CRISPR–Cas9 to generate Cyclin B1 migration and
mTECs lacking E2f8 (Extended Data Fig. 10d,e). E2f8 sgRNA multicili- Stage II: CDK1
docking

ated cells did not exhibit defects in differentiation or multiciliation Centriole

synthesis
(Extended Data Fig. 10f–i). E2f7 –/– E2f8 sgRNA multiciliated cells exhib-
ited defects in differentiation and multiciliation comparable
with those of E2f7 –/– multiciliated cells (Extended Data Fig. 10f–i). E2F7 G2
These data suggest that E2f8 is dispensable for multiciliated cell /M CDK2
differentiation.
We sought to understand how the discoordination of the multicili-
Stage
age I:

S
ation cycle disrupts multiciliated cell differentiation by examining Protein build-up
mTECs from E2f7 –/– mice and control littermates at 7 days of air–

G1/
liquid interface culture. We examined the expression of proteins asso-

G0
ciated with each stage of multiciliated cell differentiation (Extended

0
CDK2

G
Stage IV:
Data Fig. 1).

1/
G
Ciliogenesis
CCP110 caps newly synthesized centrioles and is removed before Cyclin D1
centriole maturation and ciliogenesis39. During multiciliated cell dif- CDK4/6
ferentiation, CCP110 is removed from centrioles by stage III (Extended
Precursor
Data Fig. 1a). Assessing E2f7 –/– and control mTECs revealed that CCP110
persisted on the centrioles of differentiating E2f7 –/– multiciliated cells Fig. 5 | The multiciliation cycle is a cell cycle variant that coordinates
(Fig. 4e,f). differentiation. A model of how the multiciliation cycle coordinates
Deuterosomes are present specifically during stages II and III of multiciliated cell differentiation. Multiciliated cell precursors initiate
multiciliated cell differentiation (Extended Data Fig. 1b). Examining differentiation in a G1/G0-like phase. Precursors progress into an S-like phase
E2f7 –/– and control mTECs for DEUP1, a deuterosome component16, encompassing stage I and early stage II. Cyclin D1–CDK4/6 and CDK2 regulate
revealed that differentiating E2f7 –/– multiciliated cells exhibited per- entry into the S-like phase (this work and ref. 6). E2F7 suppresses DNA synthesis
sistent deuterosomes (Fig. 4e,g). during the S-like phase and promotes the S-like to G2/M-like transition. During
the G2/M-like phase, cyclin B1–CDK1 promotes the growth of newly forming
By stage IV, multiciliated cell centrioles have acquired distal append-
centrioles and APC/C controls centriole number and progression to stage III
ages through which they dock to the apical membrane40 (Extended
of multiciliated cell differentiation, when centrioles dock to the membrane 5.
Data Fig. 1c). Examining E2f7 –/– and control mTECs for CEP164, a distal
From the G2/M-like phase, differentiating multiciliated cells transition into the
appendage component, revealed that differentiating E2f7 –/– multicili- G1/G0-like phase corresponding to stage IV, ciliogenesis. CDK2 promotes this
ated cells were defective in recruiting CEP164 (Fig. 4e,h). final stage of multiciliated cell differentiation6.
Thus, after 7 days of air–liquid interface culture (a time point at which
most E2f7 +/+ multiciliated cells have progressed to stage IV), E2f7 –/–
multiciliated cells were in stages II and III. To assess whether E2f7 –/– multiciliated cell differentiation to generate several hundred centrioles
multiciliated cells were delayed in their differentiation, we repeated that mature into basal bodies and produce motile cilia (Fig. 5).
the analysis after air–liquid interface culture for 21 days. At this late Like the canonical cell cycle, the multiciliation cycle involves the
time point, E2f7 –/– multiciliated cells had removed CCP110 and deuter- sequential expression of a network of cycle regulators. Many of these
osomes, but still had not recruited CEP164 (Fig. 4f–h), which indicated regulators, identified in this and in previous work5,6, are involved in both
that E2F7 promotes advancement from the multiciliated S-like phase to the cell cycle and the multiciliation cycle, but with different effects. For
the G2/M-like phase. Moreover, in the absence of E2F7, dysregulation example, cyclin D1-CDK4/6 is expressed in multiciliated cell precursors
of this advancement disrupts the centriolar maturation essential for and promotes progression of these precursors from a G1/G0-like to an
generating motile cilia. S-like phase to initiate the earliest steps of multiciliated cell differen-
To better understand how E2F7 is required for the differentiation tiation. CDK2 then initiates centriole synthesis during the G1-like to
of mTECs, we used transmission electron microscopy to examine S-like phase transition6. From the end of the S-like phase through the
E2f7 –/– and E2f7 +/+ mTECs cultured by the air–liquid method for 7 days. G2/M-like phase, the mitotic oscillator CDK1–APC/C regulates centri-
E2f7 +/+ multiciliated cells displayed abundant docked centrioles with ole growth and disengagement5. CDK2 then initiates ciliogenesis as
associated cilia (Fig. 4i,j). By contrast, multiciliated cells lacking E2F7 multiciliated cells move into a final G1/G0-like phase, characterized
displayed undocked centrioles and decreased ciliogenesis (Fig. 4i,j), by re-expression of Cdk4 and Cdk6.
consistent with defects in centriole maturation. Furthermore, many This shared use of components raises the question of how the multi-
centrioles in multiciliated cells lacking E2F7 were short or incomplete ciliation cycle redeploys canonical cell cycle regulators while bypassing
(Fig. 4i,k), which suggested that centriole synthesis is also disrupted key steps of the cell cycle such as DNA replication and cell division. The
in cells lacking E2F7. levels of several cell cycle regulators, including E2F7, differ between dif-
Together, these data indicate that E2F7 is required for the coordi- ferentiating multiciliated cells and proliferating cells. Investigating the
nated progression from the multiciliation cycle S-like phase into the role of E2F7 in the multiciliation cycle revealed two important functions.
G2/M-like phase. Termination of the S-like phase program is necessary First, E2F7 prevents expression of DNA replication genes in the S-like
for the timely execution of specific steps of multiciliated cell differen- phase and blocks aberrant DNA synthesis in differentiating multicili-
tiation (for example, CCP110 removal and deuterosome disassembly), ated cells. As E2F7 is elevated in the S-like phase of the multiciliation
without which later steps (for example, centriole maturation and cili- cycle to prevent DNA synthesis, CDK1 activity is dampened during the
ogenesis) are compromised. G2/M-like phase of the multiciliation cycle to prevent cytokinesis5. The
multiciliation cycle uses non-canonical cyclins, such as cyclin A1 and
cyclin O, likely to modulate CDK activity during the differentiation6,33.
Discussion Still other cell cycle regulators and related proteins may participate
Several cell cycle regulators have previously been implicated in in the multiciliation cycle. For example, Cdkn1a, which encodes the CDK
post-mitotic differentiating multiciliated cells5,6. We report an alterna- inhibitor p21CIP1, is upregulated during the S-like and G2/M-like phases
tive version of the cell cycle, the multiciliation cycle, which coordinates of the multiciliation cycle (Fig. 3a), which raises the possibility that it

220 | Nature | Vol 630 | 6 June 2024

may restrict CDK1 activity to prevent cell division. Cdc20b and Fbxo43 16. Zhao, H. et al. The Cep63 paralogue Deup1 enables massive de novo centriole biogenesis
for vertebrate multiciliogenesis. Nat. Cell Biol. 15, 1434–1444 (2013).
(also known as Emi2) encode cell cycle-related proteins required for 17. Klos Dehring, D. A. et al. Deuterosome-mediated centriole biogenesis. Dev. Cell 27, 103–112
centriole disengagement during multiciliated cell differentiation41,42. (2013).
Thus, numerous cell cycle regulators may be repurposed by the mul- 18. Breslow, D. K. & Holland, A. J. Mechanism and regulation of centriole and cilium
biogenesis. Annu. Rev. Biochem. 88, 691–724 (2019).
ticiliation cycle to coordinate the construction of a multiciliated cell. 19. Gönczy, P. Centrosomes and cancer: revisiting a long-standing relationship. Nat. Rev.
The second function of E2F7 in the multiciliation cycle is to terminate Cancer 15, 639–652 (2015).
the S-like gene expression program. Without E2F7, S-like phase expres- 20. García, S. R. et al. Novel dynamics of human mucociliary differentiation revealed by
single-cell RNA sequencing of nasal epithelial cultures. Development 146, dev177428
sion extends into the G2/M-like phase, hallmarks of the G2/M-like phase (2019).
(such as CCP110 removal and deuterosome disassembly) are delayed, 21. Ma, L., Quigley, I., Omran, H. & Kintner, C. Multicilin drives centriole biogenesis via E2f
and later events (such as distal appendage formation and ciliogenesis) proteins. Genes Dev. 28, 1461–1471 (2014).
22. Kim, S., Ma, L., Shokhirev, M. N., Quigley, I. & Kintner, C. Multicilin and activated E2f4
are disrupted. Consequently, E2F7 was required for multiciliation in induce multiciliated cell differentiation in primary fibroblasts. Sci. Rep. 8, 12369 (2018).
the lung, oviduct and brain. As these three tissues are endodermal, 23. Plasschaert, L. W. et al. A single-cell atlas of the airway epithelium reveals the CFTR-rich
mesodermal and ectodermal derivatives, respectively, the multicili- pulmonary ionocyte. Nature 560, 377–381 (2018).
24. Montoro, D. T. et al. A revised airway epithelial hierarchy includes CFTR-expressing
ation cycle is likely to be a general regulator of multiciliated cell dif- ionocytes. Nature 560, 319–324 (2018).
ferentiation. 25. Zheng, S. C. et al. Universal prediction of cell-cycle position using transfer learning.
E2F family members are implicated in two other cell cycle variants, Genome Biol. 23, 41 (2022).
26. Grant, G. D., Kedziora, K. M., Limas, J. C., Cook, J. G. & Purvis, J. E. Accurate delineation of
the endocycle and endomitosis2,43–46. Therefore, differential regulation cell cycle phase transitions in living cells with PIP-FUCCI. Cell Cycle 17, 2496–2516 (2018).
of E2F transcription factors may be a common mechanism by which the 27. Choksi, S. P., Babu, D., Lau, D., Yu, X. & Roy, S. Systematic discovery of novel ciliary genes
canonical cell cycle is repurposed in non-mitotic cells. The endocycle through functional genomics in the zebrafish. Development 141, 3410–3419 (2014).
28. Rubin, S. M., Sage, J. & Skotheim, J. M. Integrating old and new paradigms of G1/S
blocks cell division and centriole duplication, whereas the multicilia- control. Mol. Cell 80, 183–192 (2020).
tion cycle omits cell division and chromosome duplication. It will be 29. Matsushime, H., Roussel, M. F., Ashmun, R. A. & Sherr, C. J. Colony-stimulating factor 1
of interest to assess whether alternative cell cycles coordinate the dif- regulates novel cyclins during the G1 phase of the cell cycle. Cell 65, 701–713 (1991).
30. Xiong, Y., Connolly, T., Futcher, B. & Beach, D. Human D-type cyclin. Cell 65, 691–699
ferentiation of other post-mitotic cell types. (1991).
31. Fassl, A., Geng, Y. & Sicinski, P. CDK4 and CDK6 kinases: from basic science to cancer
therapy. Science 375, eabc1495 (2022).
32. Basant, A. & Glotzer, M. Spatiotemporal regulation of RhoA during cytokinesis. Curr. Biol.
Online content 28, R570–R580 (2018).
Any methods, additional references, Nature Portfolio reporting summa- 33. Funk, M. C. et al. Cyclin O (Ccno) functions during deuterosome-mediated centriole
amplification of multiciliated cells. EMBO J. 34, 1078–1089 (2015).
ries, source data, extended data, supplementary information, acknowl-
34. Westendorp, B. et al. E2F7 represses a network of oscillating cell cycle genes to control
edgements, peer review information; details of author contributions S-phase progression. Nucleic Acids Res. 40, 3511–3523 (2012).
and competing interests; and statements of data and code availability 35. Li, J. et al. Synergistic function of E2F7 and E2F8 is essential for cell survival and embryonic
development. Dev. Cell 14, 62–75 (2008).
are available at https://doi.org/10.1038/s41586-024-07476-z.
36. Skene, P. J. & Henikoff, S. An efficient targeted nuclease strategy for high-resolution
mapping of DNA binding sites. eLife 6, e21856 (2017).
37. Fischer, M., Grossmann, P., Padi, M. & DeCaprio, J. A. Integration of TP53, DREAM,
1. Morgan, D. O. The Cell Cycle: Principles of Control (New Science Press, 2007).
MMB-FOXM1 and RB-E2F target gene analyses identifies cell cycle gene regulatory
2. Orr-Weaver, T. L. When bigger is better: the role of polyploidy in organogenesis. Trends
networks. Nucleic Acids Res. 44, 6070–6086 (2016).
Genet. 31, 307–315 (2015).
38. Kumar, V. et al. The regulatory roles of motile cilia in CSF circulation and hydrocephalus.
3. Choksi, S. P., Lauter, G., Swoboda, P. & Roy, S. Switching on cilia: transcriptional networks
Fluids Barriers CNS 18, 31 (2021).
regulating ciliogenesis. Development 141, 1427–1441 (2014).
39. Spektor, A., Tsang, W. Y., Khoo, D. & Dynlacht, B. D. Cep97 and CP110 suppress a cilia
4. Spassky, N. & Meunier, A. The development and functions of multiciliated epithelia. Nat.
assembly program. Cell 130, 678–690 (2007).
Rev. Mol. Cell Biol. 18, 423–436 (2017).
40. Graser, S. et al. Cep164, a novel centriole appendage protein required for primary cilium
5. Al Jord, A. et al. Calibrated mitotic oscillator drives motile ciliogenesis. Science 358,
formation. J. Cell Biol. 179, 321–330 (2007).
803–806 (2017).
41. Kim, S., Chien, Y.-H., Ryan, A. & Kintner, C. Emi2 enables centriole amplification during
6. Vladar, E. K. et al. Cyclin-dependent kinase control of motile ciliogenesis. eLife 7, e36375
multiciliated cell differentiation. Sci. Adv. 8, eabm7538 (2022).
(2018).
42. Revinski, D. R. et al. CDC20B is required for deuterosome-mediated centriole production
7. Walentek, P. Signaling control of mucociliary epithelia: stem cells, cell fates, and the
in multiciliated cells. Nat. Commun. 9, 4668 (2018).
plasticity of cell identity in development and disease. Cells Tissues Organs 211, 736–753
43. Chen, H.-Z. et al. Canonical and atypical E2Fs regulate the mammalian endocycle. Nat.
(2022).
Cell Biol. 14, 1192–1202 (2012).
8. Mahjoub, M. R., Nanjundappa, R. & Harvey, M. N. Development of a multiciliated cell.
44. Lammens, T., Li, J., Leone, G. & De Veylder, L. Atypical E2Fs: new players in the E2F
Curr. Opin. Cell Biol. 77, 102105 (2022).
transcription factor family. Trends Cell Biol. 19, 111–118 (2009).
9. Tan, F. E. et al. Myb promotes centriole amplification and later steps of the multiciliogenesis
45. Peterson, N. G. & Fox, D. T. Communal living: the role of polyploidy and syncytia in tissue
program. Development 140, 4277–4286 (2013).
biology. Chromosome Res. 29, 245–260 (2021).
10. Vladar, E. K. & Stearns, T. Molecular characterization of centriole assembly in ciliated
46. Edgar, B. A., Zielke, N. & Gutierrez, C. Endocycles: a recurrent evolutionary innovation for
epithelial cells. J. Cell Biol. 178, 31–42 (2007).
post-mitotic cell growth. Nat. Rev. Mol. Cell Biol. 15, 197–210 (2014).
11. Byrnes, L. E., Deleon, R., Reiter, J. F. & Choksi, S. P. Opposing transcription factors MYCL
and HEY1 mediate the Notch-dependent airway stem cell fate decision. Preprint at bioRxiv
https://doi.org/10.1101/2022.10.05.511009 (2022). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
12. Vladar, E. K. & Brody, S. L. in Methods in Enzymology Vol. 525 (ed. Marshall, W. F.) 285–309 published maps and institutional affiliations.
(Academic Press, 2013).
13. Pan, J.-H. et al. Myb permits multilineage airway epithelial cell differentiation. Stem Cells
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this
32, 3245–3256 (2014).
article under a publishing agreement with the author(s) or other rightsholder(s); author
14. Lewis, M. & Stracker, T. H. Transcriptional regulation of multiciliated cell differentiation.
self-archiving of the accepted manuscript version of this article is solely governed by the
Semin. Cell Dev. Biol. 110, 51–60 (2021).
terms of such publishing agreement and applicable law.
15. Gomes Pereira, S., Dias Louro, M. A. & Bettencourt-Dias, M. Biophysical and quantitative
principles of centrosome biogenesis and structure. Annu. Rev. Cell Dev. Biol. 37, 43–63
(2021). © The Author(s), under exclusive licence to Springer Nature Limited 2024

Nature | Vol 630 | 6 June 2024 | 221

Article
Methods (10x Genomics, v.3, PN-1000075). For the analysis of mTECs treated
with DMSO or ribociclib, 16,500 cells from each of 3 independent rep-
Mouse husbandry licate mTECs were loaded onto 6 separate wells of the 10x Chromium
Mice were housed under standard pathogen-free conditions at the Controller. For the analysis of E2F7 function, 16,000 cells from each
University of California, San Francisco (UCSF) animal care facility in of 2 independent replicates of E2f7 +/+ and E2f7 –/– mTECs were loaded
the Cardiovascular Research Institute under a 12 h light–12 h dark cycle, onto 4 separate wells of the 10x Chromium Controller. Manufacturer’s
30–70% humidity and a temperature of 20–26 °C. All animal protocols instructions were followed for gel beads-in-emulsion generation, cDNA
and procedures were approved by the Institutional Animal Care and production and library construction. Libraries were sequenced with an
Use Committee of UCSF. Wild-type mice used in these studies were Illumina NovaSeq and NextSeq. CellRanger 3.0 (time course dataset),
of the strain C57BL/6J, unless otherwise noted, and obtained from 7.1 (DMSO and ribociclib dataset) or 7.0 (E2f7 dataset) was used with
Jackson Laboratory ( JAX stock 000664). All mice were 6–14 weeks default settings to demultiplex, align reads to the mouse genome (modi-
of age at the time of experimentation. All experimental groups were fied mm10 genome 3.0.0, https://www.10xgenomics.com/support/
randomized and consisted of mice of both sexes, with an equal number software/cell-ranger/downloads/cr-ref-build-steps#mouse-ref-3-0-
of male and female mice per experiment when possible. 0-mm10-vdj) and to count unique molecular identifiers. Cellranger
aggr was used to aggregate technical replicates of mTECs on day 1 of
Culturing of mTECs air–liquid interface culture. Doublets were identified and removed
mTECs were purified and cultured as previously described47. Adult from individual datasets with Scrublet (v.0.2.1)48 for the time course
mice were anaesthetized with isoflurane. Trachea were removed and and E2f7 datasets or DoubletFinder (v.2.0)49 for the DMSO and ribociclib
placed into 3 mg ml–1 Pronase (Sigma-Aldrich, 10165921001) in Ham’s dataset with an assumed doublet rate based on loading concentrations.
F-12 medium with 100 U ml–1 penicillin and 100 mg ml–1 streptomycin SoupX (v.1.5.0)50 was used to remove ambient background RNA from
overnight at 4 °C. Trachea were then agitated and cells were plated on droplets of wild-type time course and DMSO and ribociclib datasets.
Primaria plates (Corning, 087724A) in mTEC basic medium (DMEM/F-12 Seurat (v.4)51 was used for downstream analysis of individual data-
with 15 mM HEPES, 4 mM l-glutamine, 3.6 mM NaHCO3, 100 U ml–1 sets, including filtering, dimension reduction, clustering, UMAP and
penicillin and 100 mg ml–1 streptomycin) with 10% FBS for 4 h at 37 °C. differential gene expression analysis.
Airway epithelial cells were collected in mTEC Plus medium (mTEC Individual scRNA-seq datasets were merged using Seurat integration
basic with 10 μg ml–1 insulin (Sigma-Aldrich, I1882), 5 μg ml–1 transfer- to correct for batch effects. One replicate from each of days 1, 3, 9 and
rin (Sigma-Aldrich, T1147), 0.1 μg ml–1 cholera toxin (Sigma-Aldrich, 36 of air–liquid interface culture was integrated to produce the time
C8052), 25 ng ml–1 epidermal growth factor (Corning, CB40001), course dataset. Three replicates of DMSO-treated and ribociclib-treated
30 μg ml–1 bovine pituitary extract (Sigma-Aldrich, P1476), 5% FBS and mTECs were integrated to produce the DMSO and ribociclib dataset.
50 nM retinoic acid (Sigma-Aldrich, R2625) and 10 μM ROCK inhibitor Two replicates of E2f7 +/+ and E2f7 –/– mTECs were integrated to pro-
(Y-27632, Sigma-Aldrich, Y0503)). duce the E2f7 dataset. Dimension reduction and clustering using the
Cells were seeded at a density of 33,000 cells per well on 6.5 mm Tran- Louvain algorithm were performed in Seurat. Basal-stem-cell-focused
swells (Corning, 3470) coated with 50 μg ml–1 rat tail collagen (type I, or multiciliated-cell-focused datasets were subclustered and reana-
Corning, CB354249). Cells were allowed to proliferate for 5 days before lysed as described above. In both the time course and E2f7 datasets, a
initiating differentiation by starting an air–liquid interface culture. subcluster of mature multiciliated cells that shared gene expression
The basal chamber contained mTEC basic medium with 2% Nu-Serum with other cell types was removed from further analysis. Trajectory
(Corning, 355100) and 50 nM retinoic acid, whereas the apical chamber inference was performed with Monocle3 (v.1.2)52, using as.Celldataset
was left empty. Every 2 days, the apical surface was rinsed twice with to import the dataset from Seurat. Marker genes were identified using
PBS and the basal medium was changed to fresh mTEC basic medium FindAllMarkers. Potential differences in cell type proportions were
with 2% Nu-Serum and 50 nM retinoic acid. tested with Speckle (v.0.99.7)53 for the E2f7 dataset and cell cycle phase
proportions in the DMSO and ribociclib dataset and edgeR (v.3.40.2)54
scRNA-seq of mTECs glmQLFit for cell type proportions in the DMSO and ribociclib dataset.
To generate single-cell suspensions of mTECs for scRNA-seq, mTECs Genes annotated with the ‘cell cycle’ GO term (GO:0007049) and
were cultured and differentiated as described above. For the differen- genes from previously generated S and G2M phase gene lists55 were
tiation time course, mTECs generated from C57BL/6J adult mice were tested for enrichment in basal stem cells and multiciliated cells with
collected after culture for 1, 3, 9 and 36 days by the air–liquid inter- FindAllMarkers. Cell cycle phase scores were predicted with Tricy-
face method. For scRNA-seq of mTECs treated with DMSO or ribociclib, cle (v.1.6.0)25, using the default reference dataset. S and G2/M scores
ribociclib was added at a final concentration of 10 μM in mTEC basic were calculated with UCell (v.2.2.0)56, using previously generated S
medium with 2% Nu-Serum and 50 nM retinoic acid at 1 day after air– and G2/M phase gene lists55. Analysis of differential gene expression
liquid interface culture and dissociated on day 3 of air–liquid interface between E2f7 +/+ and E2f7 –/– clusters and DMSO and ribociclib clusters
culture. For scRNA-seq of mTECs lacking E2F7, mTECs were derived was performed using DESeq2 (v.1.38.3)57. Smooth lines depicting gene
from E2f7 –/– adult mice or E2f7 +/+ control littermates and dissociated expression across pseudotime were generated using ggplot2 (v.3.4.2)58
on day 7 of air–liquid interface culture. geom_smooth.
Collection consisted of washing mTECs with PBS three times apically To investigate cytoskeletal factors that may contribute to pheno-
and basally, incubating in apical and basal 0.05% trypsin–EDTA in cell types observed in E2f7 –/– multiciliated cells, 521 genes encoding anno-
dissociation buffer (Gibco, 13151014) for 15 min at 37 °C. Cells were gen- tated cytoskeletal regulators (GO biological process terms containing
tly triturated and trypsin was inactivated by adding mTEC basic medium ‘cytoskeleton’) were intersected with differentially expressed genes
with 2% Nu-Serum. An additional collection of cells was performed between E2f7 –/– and E2f7 +/+ multiciliated cells (Extended Data Fig. 8a
by adding 0.05% trypsin–EDTA in cell dissociation buffer apically for and Supplementary Table 7).
5 min at 37 °C. Both collections were combined, washed, centrifuged
and resuspended in 0.04% BSA in PBS at a density of 1,000 cells per μl. Lentiviral production and transduction of mTECs
For the transcriptional time course, 10,000 mTECs at day 1 of air– The pLenti-PGK-Neo-PIP-FUCCI 26 plasmid was obtained from
liquid interface culture were loaded onto each of 4 separate wells. On Addgene (118616). For expression of NLS–GFP, cyclin D1–GFP or
day 3, 9 or 36, 25,000 mTECs were loaded onto separate wells of a 10x E2F7–GFP, the coding sequence of NLS was synthesized and mouse
Chromium Controller using a Chromium Single Cell 3′ Reagent kit Ccnd1 (ENSMUST00000093962.5) or E2f7 were PCR-amplified
(ENSMUST00000073781.12) and cloned with a sequence encoding a normal donkey serum. The remaining immunofluorescence detection
carboxy-terminal eGFP into pLKO.3G (Addgene, 14748) with the sgRNA protocol proceeded as described above.
expression cassette removed. Lentivirus was generated by transfecting
7.5 μg of each plasmid with 1.5 μg of pCMV-VSV-G (Addgene, 8454) and PIP–FUCCI cell cycle analysis
6 μg of psPAX2 (Addgene, 12260) into a 10 cm plate of Lenti-X 293T cells For mTECs transduced with PIP–FUCCI lentivirus, cells were fixed
(HEK 293T, Takara) at 70–80% confluence using Fugene 6 transfec- with PFA after 3 days of air–liquid interface culture and were stained
tion reagent (Promega, E2691). Medium containing lentiviral particles with antibodies to detect centrioles (CEP43) or cilia (βIV-tubulin) as
were collected at 1, 2 and 3 days after transfection and concentrated described above. For nuclear PIP–FUCCI imaging, images were col-
by centrifugation at 25,000 r.p.m. for 1.5 h at 4 °C, and the pellet was lected at a zoom of ×1 to prevent photobleaching. To stage multiciliated
resuspended in 300 μl PBS. cell differentiation, the same cells were then imaged at ×3 zoom and with
For lentiviral transduction of mTECs, cells were isolated from mouse AIRYscan superresolution imaging. The mean nuclear fluorescence of
trachea as described above. Cells were plated in 6.5 mm Transwells the PIP–FUCCI reporters was assessed in ImageJ (v.1.53c) by manually
and transduced with 1–20 μl of lentivirus in mTEC Plus medium sup- drawing a region of interest around the nucleus of multiciliated cells
plemented with 50 nM retinoic acid and 10 μM ROCK inhibitor as previ- and measuring mean fluorescence of both mVenus and mCherry. Multi-
ously described59. Basal medium was changed 1 day after transduction ciliated cells were independently staged based on the superresolution
and apical medium was changed 2 days after transduction. Medium in imaging of CEP43 and αTUBAc, as described (Extended Data Fig. 1).
both chambers was changed at 3 days after transduction, and mTECs Multiciliated cell staging was then matched with the corresponding
were differentiated by air–liquid interface culture, as described nuclear PIP–FUCCI fluorescence levels. Three independently derived
above. mTEC populations were transduced separately for the 3 biological
replicates of PIP–FUCCI scoring, and at least 44 multiciliated cells were
Immunofluorescence of mTECs quantified per biological replicate.
For paraformaldehyde (PFA) fixation, cells on Transwells were washed
3 times in PBS and fixed in 4% PFA in PBS for 10 min at room tempera- CDK4/6 modulation in differentiating mTECs
ture. Cells were rinsed three times with PBS and Transwell membranes To evaluate the effects of CDK4/6 activity inhibition in mTECs, DMSO
were removed using a scalpel and then cut into smaller pieces. An indi- (vehicle control), palbociclib or ribociclib was added to a final con-
vidual piece of membrane was placed on a slide inside a hydrophobic centration of 10 μM in mTEC basic medium with 2% Nu-Serum and
boundary for staining. Cells were blocked for 1 h at room temperature 50 nM retinoic acid and added to the basal chamber of mTECs during
in PBT (1% BSA, 0.5% Triton X-100 and 0.02% sodium azide in PBS) sup- differentiation. Medium with fresh DMSO, palbociclib or ribociclib
plemented with 10% normal donkey serum. was changed every other day until the day of collection. DMSO, pal-
For the Triton–PFA fixation, cells on Transwells were washed as bociclib or ribociclib treatment was started at different times during
above, then 0.5% Triton X-100 in PBS was added to apical and basal differentiation in an experiment-specific manner. For immunofluo-
chambers. After a 3 min incubation, Triton X-100 was removed and cells rescence experiments, drugs were delivered from days 1–5 of differ-
were fixed in 4% PFA in PBS for 10 min at room temperature. Cells were entiation by air–liquid interface culture (Fig. 2b–e and Extended Data
rinsed and cut as above. Once on the slide, cells were incubated in 0.5% Fig. 5c). For generating the DMSO and ribociclib scRNA-seq dataset,
Triton-X-100 in PBS for 15 min at room temperature. After 3 washes in drugs were delivered from days 1–3 of differentiation by air–liquid
PBS for 5 min each, cells were blocked 1 h at room temperature in PBT interface culture (Fig. 2f–k, and Extended Data Fig. 5d–h). For evalu-
as above. ation of delivery timing, drugs were added at days 0–5 or 2–5 of dif-
For both PFA-fixed and Triton–PFA-fixed cells, cells were then incu- ferentiation by air–liquid interface culture, as indicated (Extended
bated in primary antibody diluted in PBT for 1 h at room temperature. Data Fig. 5a–c).
After 5 PBS washes of 5 min each, secondary antibodies diluted in To evaluate the effects of CDK4/6 activity overactivation in mTECs,
PBT were added for 1 h at room temperature. Secondary antibodies lentivirus encoding NLS–GFP (control) or cyclin D1–GFP was delivered
were washed off as for the primary antibody. Hoechst, phalloidin or as described above. Cells were fixed for immunofluorescence at 5 days
βIV-tubulin were added to stain nuclei, actin or cilia, respectively, either of differentiation by air–liquid interface culture.
during the secondary incubation or as a tertiary stain. Tertiary stains
were removed with 5 PBS washes of 5 min each. CRISPR–Cas9 mutant mouse generation
Cells were then mounted in Prolong Diamond Antifade (Molecular To generate E2f7 –/– mice, 180 pmol of sgRNA targeting the exon encod-
Probes, P36970) medium. Primary and secondary antibodies, tertiary ing the DNA-binding domain of E2F7 was mixed with 2.5 μg of TrueCut
stains, dilutions and fixation conditions are provided in Supplementary Cas9 v2 (Thermo Fisher, A36499) and incubated at room temperature
Table 10. For all mTEC immunofluorescence experiments, biological for 15 min to allow for the formation of ribonucleoprotein (RNP) com-
replicates refer to independently derived mTEC populations. plexes. RNP complexes were brought up to 100 μl with Tris–EDTA and
filtered twice through a 0.1 mm PVDF filter (Millipore, UFC30VV25).
EdU labelling and detection RNPs were microinjected into zygotes from primed and mated CD-1
For EdU labelling, 10 μM EdU or DMSO was added to basal medium of female mice and transferred into recipient females. Pups were screened
mTECs at 1 day after initiation of air–liquid interface culture. Medium by PCR and sequencing, and founder lines were crossed to C57BL/6J
with fresh EdU or DMSO was changed every other day until 5 days of mice to generate heterozygous mutant lines.
air–liquid interface culture, at which point cells on Transwells were
washed 3 times in PBS and fixed in 4% PFA in PBS for 15 min at room Whole-mount brain immunofluorescence
temperature. PFA was washed off cells with three rinses with PBS and Brains were dissected and lateral walls exposed, washed once with PBS,
Transwell membranes were removed, cut and placed on a slide inside then fixed for 10 min in ice cold methanol. Tissue was washed once
a hydrophobic boundary for staining. Cells were permeabilized with with PBS for 5 min and then fixed overnight in 4% PFA at 4 °C overnight.
a 20-min incubation in 0.5% Triton X-100 in PBS at room temperature, Tissue was washed 3 times for 10 min in PBS, then blocked for 1 h in
followed by a 30-min incubation in Click-iT EdU reaction buffer using 10% normal donkey serum with 0.03% Triton X-100 in PBS at room
Azide-AlexaFluor-488 (Thermo Fisher, C10337). Cells were rinsed once temperature. Primary and secondary antibodies were diluted in 10%
with PBS, then blocked for 1 h at room temperature in PBT (1% BSA, 0.5% normal donkey serum with 0.03% Triton X-100 in PBS and incubated
Triton X-100 and 0.02% sodium azide in PBS) supplemented with 10% at 4 °C overnight, with 3 times 10 min washes in PBS containing 0.03%
Article
Triton X-100 after each incubation. Tissue was then mounted and sensitivity D5000 ScreenTape reagents (Agilent, 5067-5593) and the
imaged in Prolong Diamond Antifade medium. Tapestation 4200 (Agilent) were used to assess library quality and pool
libraries for sequencing. Sequencing was done using a NextSeq 500
Tissue cryosectioning and immunofluorescence (Illumina).
Tissues were dissected from mice and fixed overnight in 4% PFA, CUT&RUN analysis was performed using previously described
then were washed 3 times in PBS and placed into 30% sucrose until scripts63,64. In brief, FASTQs were trimmed using trimmomatic (v.0.39)65
tissues descended. Tissues were embedded in OCT and sections with a second trimming step to remove remaining read-through adap-
taken on a Leica CM1900 cryostat at 10 μm thickness. Sections were tors. Reads were aligned with bowtie2 (v.2.4.2)66 using the parameters
collected on glass slides and allowed to dry for at least 1 h before --local --very-sensitive-local --no-unal --no-mixed --no-discordant
staining. For staining, sections were washed 3 times in PBST (0.1% --phred33 -I 10 -X 700. Samples were filtered for reads under 120 bp
Tween-20 diluted in PBS). Next, citrate antigen retrieval was per- and duplicates were marked, but not removed, with samtools (v.1.10)
formed 3 times with boiling 0.1 M citrate buffer with 0.1% Tween-20 and picard (v.2.24.0) MarkDuplicates. MACS2 (v.2.1.2)67 was used to
at pH 6.0. Tissues were then washed once with PBST and blocked call peaks using parameters --keep-dup all -q 0.05 and -c with the cor-
for 1 h in 10% normal donkey serum in PBST at room temperature. responding control being the NLS–GFP sample. Final peak list was gen-
Primary antibodies were diluted in PBST + 10% normal donkey serum erated by the intersection of replicate peak lists. ChIPseeker (v.1.34.1)68
and placed on slides for 1.5 h at room temperature or overnight at was used to annotate peaks to the nearest gene.
4 °C. Slides were washed 3 times in PBST and secondary antibody was
added for 1 h at room temperature. Slides were washed 5 times for Quantitative PCR with reverse transcription
5–10 min each in PBST and mounted with Prolong Diamond Antifade Total RNA was extracted from mTECs at 3 days of differentiation by
medium. air–liquid interface culture using a RNeasy Plus Mini kit (Qiagen, 74134).
Reverse transcription of 300–1,000 ng of RNA was carried out for 1 h
Histological staining and quantitation of brain phenotypes at 42 °C using a iSCRIPT cDNA synthesis kit (Bio-Rad, 1708841BUN).
Slides were washed 3 times in PBS for 5 min each, then washed in ddH2O Quantitative PCR was performed using a QuantStudio 5 real-time PCR
twice for 3 min each. Slides were placed in haematoxylin for 3 min then machine running QuantStudio Design & Analysis software (v.1.5.1)
washed for 1 min in running tap water. Slides were then placed in clari- (Applied Biosystems) and using PowerUp SYBR green master mix
fier for 1 min, washed for 30 s in ddH2O and placed into Bluing Reagent (Applied Biosystems, A25742). A table of primer sequences is pro-
for 1 min. Slides were then washed for 30 s in ddH2O. Slides were then vided in Supplementary Table 10. Relative expression was calculated
placed in 80% ethanol for 1 min and stained in eosin Y for 3 min. The using the ΔΔCT method69. An average of three control gene (Hprt, Actb
slides were progressively dehydrated in ethanol from 95 to 100% for 2 and Rplp0) expression levels was used for internal sample normali-
washes of 2 min each, then dehydrated in xylene twice for 3 min each zation. All quantitative PCR with reverse transcription experiments
and mounted using Cytoseal (Thermo Fisher, 23-244257). Slides were were carried out on at least three independently derived and treated
imaged on a Lecia Widefield DMi8 microscope equipped with a Leica mTEC cultures.
DFC9000 GTC camera.
To quantitate ventricle size, a ratio of the ventricular area to the whole Image acquisition and analysis
brain area was calculated using the line segment tool in ImageJ (v.1.53c). Immunofluorescence images were acquired on a Zeiss 800 laser scan-
Three coronal sections from each animal were averaged to calculate ning confocal microscope with Airyscan superresolution capability
ventricular area of three animals per genotype. and equipped with a ×63 oil objective. ZEN microscopy software
(v.3.7; Zeiss) was used for image acquisition. All images were pro-
CUT&RUN cessed using ImageJ (v.1.53c) and figures were assembled in Adobe
CUT&RUN was performed as previously described36,60, with pro- Illustrator.
tocol and buffers detailed in protocol 2 (ref. 61). In brief, E2F7– At least three randomly selected regions of stained membrane were
GFP-transduced or NLS–GFP-transduced mTECs were grown on imaged for each biological replicate per each condition, with a mini-
24 mm Transwells (Corning, 3450) for 3 days in air–liquid interface mum of three biological replicates per experiment. At each position, a
culture and were dissociated in 0.05% trypsin–EDTA in cell disso- stack of 10–20 images was collected, spaced 0.5 μm apart in the z axis
ciation buffer for 15 min at 37 °C. Cells were collected in mTEC basic with no additional zoom. For superresolution imaging, stacks were
medium with 2% Nu-Serum, washed twice in 5 ml of CUT&RUN wash spaced 0.14 μm apart and collected at ×3 zoom.
buffer and resuspended in CUT&RUN wash buffer. One million cells Gain and laser power were held constant for each channel for each
of each sample were brought up to a volume of 1 ml with CUT&RUN experiment. Images were processed identically using ImageJ (v.1.53c)
wash buffer. Cells were incubated with 10 μl activated wheat germ to generate maximum projections for figure assembly. Sum projections
agglutinin beads (Bangs Laboratories, BP530) at room temperature were created for fluorescence intensity quantification. Significance
for 10 min with gentle rocking. Cells were incubated overnight with testing was performed in Prism (v.9.5.1).
rabbit anti-GFP antibody (Abcam, ab290) diluted 1:100 in 0.05% digi-
tonin in CUT&RUN wash buffer. Following 2 washes in CUT&RUN wash Generating CRISPR knockout mTECs
buffer, samples were incubated for 1 h at room temperature in guinea mTECs were isolated as described above and 175,000 cells were seeded
pig anti-Rabbit IgG (Rockland, 611-201-122) diluted 1:100 in 0.05% digi- on 24 mm collagen-coated Transwells (Corning, 3450) in mTEC Plus
tonin in CUT&RUN wash buffer. Samples were incubated with Protein Expansion medium (mTEC Plus supplemented with 50 nM retinoic
A/G–MNase fusion protein (Addgene, 123461) at 700 ng ml–1 for 1 h at acid, 10 μM ROCK inhibitor, 1 μM A83-01 (Tocris, 2939), 200 nM DMH-1
4 °C. Following 2 washes in 0.05% digitonin in CUT&RUN wash buffer (Tocris, 4126) and 500 nM CHIR99021 (Tocris, 4423)) based on a previ-
and one wash in CUT&RUN low-salt rinse buffer, samples were digested ously described recipe70. Expansion medium was changed daily for
for 10 min at 0 °C in cold CUT&RUN incubation buffer. DNA fragments 4 days. Cells were then dissociated using 0.05% trypsin diluted in cell
were released by incubation at 37 °C for 30 min and then collected in the dissociation buffer (Gibco, 13151014) and collected, counted, washed in
aqueous phase of a phenol–chloroform extraction. Library preparation PBS and resuspended in P3 electroporation buffer (Lonza, V4SP-3096)
was performed using a Next Ultra II DNA Library Prep kit for Illumina at a concentration of 20,000 cells per 1 μl.
(NEB, E7645S) using 25 μl of CUT&RUN DNA as input, as previously Guide RNAs targeting E2f8 or two control nontargeting guides were
detailed62. Adaptors were diluted 1:15 in adaptor dilution buffer. High designed and synthesized (Synthego; sequences in Supplementary
Table 10). RNPs were assembled by incubating 90 pmol TrueCut Cas9
v2 protein (Thermo Fisher, A36499) with 180 pmol of sgRNA at room Code availability
temperature for 15 min. A total of 400,000 cells in 20 μl of P3 buffer Scripts are available at GitHub (https://github.com/lb15/multiciliation_
were added to each RNP and electroporated using program 96-EA- cycle). Scripts and R objects used for analysis are available at Zenodo
104 on an Amaxa 4D-Nucleofector equipped with a 96-well shuttle (https://doi.org/10.5281/zenodo.10896100)71, including CUT&RUN
(Lonza, AAF-1003B/S). After electroporation, cells were collected in analysis scripts (https://doi.org/10.5281/zenodo.10896066)72 and
mTEC Plus Expansion medium and seeded equally onto five 6.5 mm Seurat analysis scripts (https://doi.org/10.5281/zenodo.10896071)73.
collagen-coated Transwells. Expansion medium was changed daily
until cells reached confluence (3–4 days after electroporation) at which
point cells were differentiated similar to standard mTECs as described
47. You, Y. & Brody, S. L. Culture and differentiation of mouse tracheal epithelial cells.
above. Methods Mol. Biol. 945, 123–143 (2013).
To evaluate knockout efficiency, gDNA was extracted from a single 48. Wolock, S. L., Lopez, R. & Klein, A. M. Scrublet: computational identification of cell
doublets in single-cell transcriptomic data. Cell Syst. 8, 281–291.e9 (2019).
6.5 mm Transwell using an AllPrep DNA/RNA Mini kit (Qiagen, 0204).
49. McGinnis, C. S., Murrow, L. M. & Gartner, Z. J. DoubletFinder: doublet detection in
gDNA was amplified using oligonucleotides surrounding the sgRNA single-cell RNA sequencing data using artificial nearest neighbors. Cell Syst. 8, 329–337.
binding site. PCR amplicons were purified and Sanger sequenced using e4 (2019).
50. Young, M. D. & Behjati, S. SoupX removes ambient RNA contamination from
the corresponding forward primer. Sequence data were deconvolved
droplet-based single-cell RNA sequencing data. Gigascience 9, giaa151 (2020).
and analysed using ICE to determine knockout efficiency (https://ice. 51. Stuart, T. et al. Comprehensive integration of single-cell data. Cell 177, 1888–1902.e21
synthego.com/). A minimum of three independently derived and (2019).
52. Cao, J. et al. The single-cell transcriptional landscape of mammalian organogenesis.
electroporated mTEC populations per guide RNA were generated for Nature 566, 496–502 (2019).
analysis of knockout efficiency and phenotypes. 53. Phipson, B. et al. propeller: testing for differences in cell type proportions in single cell
data. Bioinformatics 38, 4720–4726 (2022).
54. Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor package for
Transmission electron microscopy of mTECs
differential expression analysis of digital gene expression data. Bioinformatics 26,
mTECs were differentiated for 7 days by air–liquid interface culture and 139–140 (2010).
fixed in 4% PFA and 2% glutaraldehyde in 0.1 M sodium cacodylate 55. Kowalczyk, M. S. et al. Single-cell RNA-seq reveals changes in cell cycle and
differentiation programs upon aging of hematopoietic stem cells. Genome Res. 25,
buffer. Subsequently, cells were washed 3 times with 0.1 M sodium 1860–1872 (2015).
cacodylate buffer, followed by post-fixation with 1% osmium tetroxide. 56. Andreatta, M. & Carmona, S. J. UCell: robust and scalable single-cell gene signature
This was followed by 3 washes with ddH2O and pre-staining with 0.5% scoring. Comput. Struct. Biotechnol. J. 19, 3796–3798 (2021).
57. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion
uranyl acetate for 1 h, followed by 3 additional washes with ddH2O. Cells for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
were dehydrated with graded ethanol and inserts were removed from 58. Wickham, H., Navarro, D. & Pedersen, T. L. ggplot2: Elegant Graphics for Data Analysis 3rd
the Transwell and placed on a piece of ACLAR film. Poly/Bed 812 resin edn https://ggplot2-book.org/ (Springer, 2024).
59. Horani, A., Nath, A., Wasserman, M. G., Huang, T. & Brody, S. L. Rho-associated protein
(Polysciences, 08792-1) was added directly onto the inserts. A thin resin kinase inhibition enhances airway epithelial basal-cell proliferation and lentivirus
slide, approximately 2 mm thick, was sectioned from the polymerized transduction. Am. J. Respir. Cell Mol. Biol. 49, 341–347 (2013).
resin block near the bottom of the insert using a jeweller’s saw blade. 60. Meers, M. P., Bryson, T. D., Henikoff, J. G. & Henikoff, S. Improved CUT&RUN chromatin
profiling tools. eLife 8, e46314 (2019).
The resin slide was sliced into 1 mm blocks using a razor blade and 61. Janssens, D. & Henikoff, S. CUT&RUN: targeted in situ genome-wide profiling with high
glued onto a supporting resin block, with the surface where the cells efficiency for low cell numbers. protocols.io https://doi.org/10.17504/protocols.io.
had grown positioned sideways. The other sides of the block without zcpf2vn (2019).
62. Liu, N. Library prep for CUT&RUN with NEBNext® UltraTM II DNA Library Prep Kit for
cells were further trimmed, and 70 nm sections were sliced using a Illumina® (E7645). protocols.io https://doi.org/10.17504/protocols.io.bagaibse (2021).
Leica EM UC7 microtome and diamond knife. TEMs were obtained 63. Zhu, Q., Liu, N., Orkin, S. H. & Yuan, G.-C. CUT&RUNTools: a flexible pipeline for CUT&RUN
using a transmission electron microscope (FEI T12) equipped with a processing and footprint analysis. Genome Biol. 20, 192 (2019).
64. Zheng, Y., Ahmad, K. & Henikoff, S. CUT&Tag data processing and analysis tutorial.
CCD camera. protocols.io https://doi.org/10.17504/protocols.io.bjk2kkye (2020).
65. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina
Obtaining biological materials sequence data. Bioinformatics 30, 2114–2120 (2014).
66. Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. Methods
All biological materials generated in this study are available upon 9, 357–359 (2012).
request. 67. Zhang, Y. et al. Model-based analysis of ChIP–seq (MACS). Genome Biol. 9, R137 (2008).
68. Yu, G., Wang, L.-G. & He, Q.-Y. ChIPseeker: an R/Bioconductor package for ChIP peak
annotation, comparison and visualization. Bioinformatics 31, 2382–2383 (2015).
Statistical testing 69. Livak, K. J. & Schmittgen, T. D. Analysis of relative gene expression data using real-time
All statistical testing was performed using Prism (v.9.5.1). Statistical quantitative PCR and the 2( −ΔΔT) method. Methods 25, 402–408 (2001).
70. Mou, H. et al. Dual SMAD signaling inhibition enables long-term expansion of diverse
tests used for each experiment are listed in the accompanying figure
epithelial basal cells. Cell Stem Cell 19, 217–231 (2016).
legend. 71. lb15. multiciliation_cycle: v1.1.3. Zenodo https://doi.org/10.5281/zenodo.10896100
(2024).
Reporting summary 72. lb15. autoCutandRun: v1.1.1. Zenodo https://doi.org/10.5281/zenodo.10896066 (2024).
73. lb15. autoSeurat: v1.1.2. Zenodo https://doi.org/10.5281/zenodo.10896071 (2024).
Further information on research design is available in the Nature Port-
folio Reporting Summary linked to this article.
Acknowledgements We thank D. Erle, K. D. Koh and L. Bonser for advice on developing the
CRISPR knockout protocol; E. Yu for mouse husbandry; D. O. Morgan, P. H. O’Farrell, M. Aydogan,
Data availability P. K. Choksi and members of the Reiter Laboratory for critical discussion; N. Neff, R. Sit and
scRNA-seq and CUT&RUN raw FASTQ files and processed data have M. Tan from the CZ Biohub Genomics platform for sequencing; and staff at the UCSF
Laboratory for Cell Analysis and the UCSF Center for Advanced Cell Technology for use of
been deposited into the Gene Expression Omnibus database under equipment and the Wynton high performance computing cluster for analysis. L.E.B was
accession GSE228110. scRNA-seq datasets have been deposited in supported by Ruth L. Kirschstein National Research Service Awards (5T32HL007731-27 and
the CellxGene single cell browser (https://cellxgene.cziscience.com/ 1F32HL154611-01). This work was supported by a CIRM Discovery grant (DISC1-10475) to S.P.C.
and grants from the NIH (R01AR054396 and R01HD089918) to J.F.R. This research was also
collections/c26ca66a-63ea-4059-a24e-0e0be0a2a173). The mouse supported by the National Cancer Institute, National Institutes of Health, Intramural Research
mm10 reference genome was downloaded for analysis and is avail- Program, and was funded in whole or in part with federal funds from the National Cancer
Institute, National Institutes of Health under contract HHSN26120080001E. The content of this
able from 10x Genomics (https://www.10xgenomics.com/support/
publication does not necessarily reflect the views or policies of the Department of Health and
software/cell-ranger/downloads/cr-ref-build-steps#mouse-ref-3-0- Human Services, nor does mention of trade names, commercial products, or organizations
0-mm10-vdj). Source data are provided with this paper. imply endorsement by the US Government.
Article
Author contributions S.P.C. conceptualized the study. S.P.C., L.E.B. and J.F.R. designed Additional information
experiments. S.P.C. and L.E.B. performed experiments and analysed data. L.E.B. performed Supplementary information The online version contains supplementary material available at
bioinformatics analyses. M.J.K. performed phenotypic analyses of mouse mutants. B.W.H.T. https://doi.org/10.1038/s41586-024-07476-z.
and R.D. assisted with experimentation. Q.L. and C.J.W. performed transmission electron Correspondence and requests for materials should be addressed to Semil P. Choksi or
microscopy analyses. S.P.C., L.E.B. and J.F.R. wrote the manuscript. S.P.C. and J.F.R. supervised Jeremy F. Reiter.
the research. All authors reviewed and edited the manuscript. Peer review information Nature thanks Piotr Sicinski, Bart Westendorp and the other,
anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer
reports are available.
Competing interests J.F.R. cofounded startup companies funded by BridgeBio and 459AM. Reprints and permissions information is available at http://www.nature.com/reprints.
Extended Data Fig. 1 | Multiciliated cells go through sequential stages of multiciliated cell differentiation. Precursors induce the expression of early
differentiation. a, Representative immunofluorescence images of wild-type multiciliated cell transcription factors, such as MYB. Stage I involves the
mTECs cultured at air-liquid interface for three days and stained for centrioles induction of proteins required for centriole biogenesis, such as CEP43 and
(CEP43, cyan) and CCP110 (yellow). Cells at stages I-IV of multiciliated CCP110. Stage II involves the generation of deuterosomes (marked by DEUP1,
differentiation are depicted. n = 3 biological replicates. Scale bar, 5 μm. depicted here as yellow circles) and the synthesis of centrioles (depicted as
b, Representative immunofluorescence images of wild-type mTECs cultured at blue dots). Stage III involves centriole disengagement, migration, acquisition
airliquid interface for three days and stained for centrioles (CEP43, cyan) and of distal appendages (marked by CEP164) and docking at the apical membrane.
deuterosomes (DEUP1, yellow). n = 3 biological replicates. c, Representative Stage IV involves ciliogenesis. Cilia are in red. e, Proteins that are expressed
immunofluorescence images of wild-type mTECs cultured at airliquid interface during each stage of multiciliated cell differentiation used in this study are listed,
for three days and stained for centrioles (CEP43, cyan) and distal appendages with expression in precursors and during multiciliated cell differentiation
(CEP164, yellow). n = 3 biological replicates. d, A schematic of the stages of schematized by color.
Article

Extended Data Fig. 2 | A time course of mTEC differentiation captured (Dnah5) overlaid on the UMAP of the integrated dataset. Color indicates
by scRNA-seq reveals expression of cell cycle regulators. a, Integrated expression level. d, Pseudotime values for proliferating basal stem cells.
scRNA-seq dataset of mTECs on day 1, 3, 9, and 36 at air-liquid interface culture. e, Pseudotime values for differentiating multiciliated cells. f, Heatmap of
Clusters are distinguished by color. b, Individual day 1, 3, 9, and 36 datasets that average expression of select cell cycle-related genes in both basal stem (above)
contribute to the integrated dataset. c, Expression of select marker genes for and multiciliated (below) cells across S and G2/M-binned phases of the cell
basal stem cells (Krt5), secretory cells (Scgb3a1), deuterosome-producing cycle. Color indicates expression (z-score).
differentiating multiciliated cells (Deup1) and mature multiciliated cells
Extended Data Fig. 3 | Differentiating multiciliated cells express cell cycle multiciliation functional categories are listed. Cluster identity of cells is
regulators and ciliogenesis genes. a, Heatmap of cell cycle and ciliogenesis schematized by color below. b, Average minimum and maximum normalized
genes arranged across multiciliated cell differentiation pseudotime. Color of expression across multiciliated cell differentiation pseudotime of select genes
individual boxes in heatmap indicates expression (z-score). S (blue) and G2/M encoding CDK proteins, cyclins, composite scores of S and G2/M phase-related
(green) phase scores represent normalized expression of genes associated genes, multiciliated cell transcription factors, proteins involved in centriole
with each stage of the cell cycle. Select genes associated with cell cycle and synthesis and maturation, or proteins involved in ciliogenesis.
Article

Extended Data Fig. 4 | mTECs cease proliferating during differentiation. (Day -1), the day of transition (Day 0), one day after transition to air-liquid
a, Representative immunofluorescence images of wild-type mTECs cultured interface (Day 1) or two days after (Day 2). H3S10P is a marker of cells in mitosis.
at air-liquid interface for five days and stained for EdU (cyan), FOXJ1 (red) and TP63, also known as p63, is a marker of airway stem cells. Lower panels depict
nuclei (Hoechst, grey). EdU or DMSO was added during days one to five of individual channels. Scale bar, 10 μm. d, Percentage of H3S10P-expressing
culture at air-liquid interface. n = 3 biological replicates. Scale bar, 10 μm. cells in differentiating mTECs at the indicated times. Horizontal lines indicate
b, Percentage of multiciliated cells (expressing FOXJ1) that exhibit EdU means ± s.e.m. of 3 biological replicates, ****P = 0.000002 and NS, not
incorporation. Bar graph quantitates EdU incorporation in 186 multiciliated significant (one-way ANOVA with Sidak’s correction). e, Percentage of TP63-
cells assessed from 3 biological replicates. c, Immunofluorescence images expressing airway stem cells that also express H3S10P in differentiating mTECs
of differentiating wild-type mTECs. mTECs were stained for Histone 3 at the indicated times. Horizontal lines indicate means ± s.e.m. of 3 biological
phosphorylated at serine 10 (H3S10P, cyan), TP63 (yellow) and nuclei (Hoechst, replicates, ***P = 0.0009 and NS, not significant (one-way ANOVA with Sidak’s
grey) two days before transition to air-liquid interface (Day -2), one day before correction).
Extended Data Fig. 5 | Assessment of roles of CDK4/6 and cyclin D1 in mTEC colored by cell cycle phase scores determined by tricycle analysis. g, Integrated
proliferation and differentiation. a, Immunofluorescence images of mTECs scRNA-seq dataset with pseudotime values of cells of the basal stem,
treated with DMSO, palbociclib or ribociclib during days 0-5 or days 2-5 at intermediate and multiciliated cell clusters. Pseudotime infers a differentiation
air-liquid interface and stained for centrioles (CEP43, cyan), cilia (αTubAc, red) trajectory (black arrow). Integrated scRNA-seq dataset with blue indicating
and nuclei (grey). Scale bar, 10 μm. b, Percentage of multiciliated cells in mTECs cells expressing markers of intermediate and multiciliated cells (e.g., Gmnc and
treated with DMSO, palbociclib or ribociclib for given timepoints. Horizontal Foxj1) selected for subclustering and pseudotime analysis. h, Pseudotime
lines indicate means ± s.e.m. of 3 biological replicates, ***P = 0.0004, values of cells of the intermediate and multiciliated cell clusters. Pseudotime
****P = 0.0002, *P = 0.0059, **P = 0.0018 (one-way ANOVA with Sidak’s infers a differentiation trajectory (black arrow). i, Immunofluorescence images
correction). c, Nuclear density of mTECs treated with DMSO, palbociclib or of mTECs transduced with lentivirus expressing NLS-GFP (control) or cyclin
ribociclib for given timepoints. Horizontal lines indicate means ± s.e.m. of D1-GFP and stained for centrioles (CEP43, cyan) and markers of multiciliated
3 biological replicates, NS, not significant (one-way ANOVA with Dunnet’s cell differentiation: CCP110, deuterosomes (DEUP1) and distal appendages
correction). d, UMAP of integrated scRNA-seq data of DMSO- and ribociclib- (CEP164). Individual channels are shown to the right. Scale bars, 10 μm.
treated mTECs. Clusters are distinguished by color. e, Individual scRNA-seq j, Percentage of multiciliated cells expressing CCP110, DEUP1 or CEP164 in
datasets of DMSO- or ribociclib-treated mTECs. Clusters are distinguished by mTECs expressing NLS-GFP or cyclin D1-GFP. Horizontal lines indicate means ±
color. f, Individual scRNA-seq datasets of DMSO- or ribociclib-treated mTECs s.e.m. of 3 biological replicates, NS, not significant (unpaired two-tailed t-test).
Article

Extended Data Fig. 6 | Comparing transcriptional signatures of the E2f7 −/+ mice reveals no significant difference between observed and expected
canonical cell cycle and multiciliation cycle identifies E2f7. a, Proliferating ratios (1 E2f7 +/+:2 E2f7 −/+:1 E2f7 −/−, chi-squared test). n = 126 mice from 16 litters.
stem (orange) and multiciliated (blue) cells in S or G2/M phase scored for the g, Allele-specific (E2f7 + or E2f7 −) quantitative PCR on cDNA from E2f7 +/+ or
expression of S or G2/M phase genes. Scores are normalized Mann-Whitney E2f7 −/− mTECs. Bars indicate means ± s.e.m. of 3 (E2f7 +) or 4 (E2f7 −) biological
U-statistic of gene set expression. ****P = 0.000002, ***P = 0.00001 (unpaired replicates. Points at value 0 indicate undetectable expression. *P = 0.0259
two-tailed t-tests). b, Heatmap of cell cycle-related genes differentially (E2f7 +) or *P = 0.0246 (E2f7 −) (paired two-tailed t-test). h, Immunofluorescence
expressed between proliferating stem and multiciliated cells during each cell images of E2f7 +/+ and E2f7 −/− mTECs stained for E2F7 (yellow), centrioles (CEP43,
cycle phase. Color indicates expression (z-score). c, Average expression of cyan) and nuclei (grey). Right: magnifications of boxed cells. Scale bars, 5 μm.
select genes preferentially expressed during the multiciliation cycle across i, E2F7 nuclear intensity in multiciliated cells in E2f7 +/+ and E2f7 −/− mTECs.
cycle phases of proliferating stem and multiciliated cells. d, Expression of E2f7 Horizontal lines indicate means ± s.e.m. of 3 biological replicates, **P = 0.0013
projected onto the UMAP of the integrated mTEC dataset, proliferating stem (unpaired two-tailed t-test). j, Representative images of adult trachea from
and multiciliated cell subset. Color indicates expression level. e, E2f7 mutation E2f7 +/+ and E2f7 −/− mice immunostained for E2F7 (yellow), MYB (red), centrioles
generation using an sgRNA (red arrow) homologous to mouse exon 4. The (CEP43, cyan) and nuclei (grey). Right: magnifications of boxed cells. n = 3
E2f7 em1Schok allele is predicted to generate a frameshift after codon 165 and biological replicates. Scale bars, 10 μm.
referred to as E2f7 −. Scale bar, 1 kb. f, Genotypes of offspring of intercrossed
Extended Data Fig. 7 | E2F7 regulates S phase-like gene expression during clusters in E2f7 +/+ and E2f7 −/− mTECs. Colors distinguish clusters A-E, NS, not
multiciliated cell differentiation. a, Individual replicate scRNA-seq datasets significant (two-tailed Moderated t-test with Benjamini-Hochberg correction).
of E2f7 +/+ and E2f7 −/− mTECs after culture at air-liquid interface for seven days. h, Box plots depicting S phase gene signature scores of E2f7 +/+ and E2f7 −/− basal
Clusters are distinguished by color. b, Integrated scRNA-seq dataset of E2f7 +/+ stem and multiciliated cell clusters. Bars are colored by cluster. Boxes show
and E2f7 −/− mTECs with cells highlighted from which basal stem- (orange) and interquartile range, horizontal bars indicate medians and whiskers show the
multiciliated- (blue) subclustered scRNA-seq datasets were derived (Fig. 3). minimum and maximum values of 2 biological replicates, with outliers plotted
c, Clusters of E2f7 +/+ and E2f7 −/− mTEC-derived basal stem cells. Colors distinguish individually, *P = 0.0014, **P = 0.0013 (multiple unpaired t-tests with Holm-
clusters. d, Proportion of basal stem cell clusters in E2f7 +/+ and E2f7 −/− mTECs. Sidak correction). i, Box plots depicting G2/M phase gene signature scores of
Colors distinguish clusters, NS, not significant (two-tailed Moderated t-test E2f7 +/+ and E2f7 −/− basal stem and multiciliated cell clusters. Bars are colored by
with Benjamini-Hochberg correction). e, Pseudotime values across multiciliated cluster. Boxes show interquartile range, horizontal bars indicate medians and
cell differentiation. f, Clusters of E2f7 +/+ and E2f7 −/− mTEC-derived multiciliated whiskers show the minimum and maximum values of 2 biological replicates,
cells. Colors distinguish clusters, labeled A-E. Arrows indicate the inferred with outliers plotted individually, NS, not significant (multiple unpaired t-tests
pseudotime differentiation trajectory. g, Proportion of multiciliated cell with Holm-Sidak correction).
Article

Extended Data Fig. 8 | E2F7 directly represses genes encoding DNA replication machinery that are differentially expressed between E2f7 +/+ and
replication proteins. a, Heatmap of DNA replication-, S phase- and E2f7 −/− multiciliated cells are shown. y-axes represent reads per genomic
cytoskeleton-associated genes differentially expressed between E2f7 +/+ and content. Scale bars, 1 kb. c, Distribution of E2F7-GFP peaks relative to gene
E2f7 −/− multiciliated cells derived from mTEC scRNA-seq data. Boxes represent positions. d, Venn diagram of the overlap of the 334 genes differentially
bins of expression arranged across multiciliated cell differentiation pseudotime. expressed between E2f7 +/+ and E2f7 −/− multiciliated cells (blue) and 89 E2F7
Color indicates expression (z-score). S (blue) and G2/M (green) phase scores direct target genes previously identified in proliferating cells by Westendorp
represent normalized expression of genes associated with each cell cycle et al. 34 (grey). e, Venn diagram of the overlap of the 43 genes near E2F7
phase. Genes near E2F7-GFP CUT&RUN peaks are outlined in blue. b, mTECs CUT&RUN peaks in mTECs (blue) and 89 E2F7 direct target genes previously
transduced with E2F7-GFP or a control (NLS-GFP) lentivirus were analyzed identified in proliferating cells by Westendorp et al. 34 (grey).
using CUT&RUN. CUT&RUN peaks for a subset of genes encoding DNA
Extended Data Fig. 9 | E2F7 is required for multiciliated cell differentiation. *P = 0.0381 (unpaired two-tailed t-test). e, Sections of oviducts (left) or brain
a, Hydrocephalus quantification of 34 progeny from 6 litters of E2f7 −/+ or E2f7 −/− ventricles (right) of adult E2f7 +/+ and E2f7 −/− mice immunostained for centrioles
mice crossed with E2f7 −/+ mice. Fractions are the number of mice of each (CEP43, cyan), cilia (αTubAc, red) and nuclei (Hoechst, gray). Lower panels
genotype with hydrocephalus over the total number of mice of that genotype show individual channels. White arrows indicate E2f7 −/− cells with accumulated
assessed. b, Ratio of ventricle area to whole brain area in sections from adult cytoplasmic centrioles. Scale bar, 10 μm. f, Percentages of cells with > 5
E2f7 +/+ and E2f7 −/− mice. Horizontal lines indicate means ± s.e.m. of three centrioles that have multiple cilia in adult mouse tracheas and oviducts.
different mice, *P = 0.0139 (unpaired two-tailed t-test). c, Brain ventricles of Horizontal lines indicate means ± s.e.m. of 3 mice, ***P = 0.0005 and *P = 0.0142
adult E2f7 +/+ and E2f7 −/− mice immunostained for centrioles (CEP43, cyan) and (ordinary one-way ANOVA with Sidak’s correction). g, Percentages of
cilia (αTubAc, red). Right panels show individual channels. Scale bars, 10 μm. multiciliated cells with centrioles undocked to the apical membrane in adult
d, Quantitation of acetylated tubulin (αTubAc) intensity in multiciliated cells mouse tracheas, oviducts and brain ventricles. Horizontal lines indicate
of E2f7 +/+ and E2f7 −/− brain ventricles. Each dot represents the mean αTubAc means ± s.e.m. of 3 mice, ***P = 0.0004 and ****P = 0.00005 (ordinary one-way
intensity of > 100 multiciliated ependymal cells from a mouse brain, with n = 7 ANOVA with Sidak’s correction).
E2f7 +/+ mice and n = 8 E2f7 −/− mice. Horizontal lines indicate means ± s.e.m.,
Article

Extended Data Fig. 10 | E2f8 is dispensable for multiciliation. a, Expression mutant, or E2f7 and E2f8 double mutant mTECs cultured for 7 days at air-liquid
of E2f8 projected onto the UMAPs of basal stem and multiciliated cells interface, stained for centrioles (CEP43, cyan) and cilia (αTubAc, red). Right:
from the mTEC timecourse scRNA-seq dataset. Color indicates expression individual channels. Scale bars, 10 μm. g, αTubAc intensity in control, E2f8
level. b, Average expression of E2f8 across cycle phases of basal stem and mutant, E2f7 mutant, and E2f7 and E2f8 double mutant mTECs cultured for 7 days
multiciliated cells. c, scRNA-seq expression of E2f8 in E2f7 +/+ and E2f7 −/− across at air-liquid interface. Horizontal lines indicate means ± s.e.m. of 3 biological
differentiation pseudotime. Grey bars indicate 95% confidence intervals. replicates, *P = 0.0130 (one-way ANOVA with Sidak’s correction). h, Centriolar
Colors indicate cluster identity. d, Strategy for generating E2f8 knockout area in control, E2f8 mutant, E2f7 mutant, and E2f7 and E2f8 double mutant
mTECs. Red arrows indicate positions of sgRNAs homologous to sequences in mTECs cultured for 7 days at air-liquid interface. Horizontal lines indicate
exon 7 of E2f8. Scale bar, 1 kb. e, Quantitative RT-PCR for wild-type E2f7 (E2f7 +) means ± s.e.m. of 3 biological replicates *P = 0.0121 (one-way ANOVA with
and E2f8 (E2f8 +) transcripts from control (E2f7 +/+ Control sgRNA), E2f8 mutant Sidak’s correction). i, Percentage of control, E2f8 mutant, E2f7 mutant, and
(E2f7 +/+ E2f8 sgRNA), E2f7 −/− (E2f7 −/− Control sgRNA) or E2f7 and E2f8 double mutant E2f7 and E2f8 double mutant multiciliated cells containing deuterosomes
(E2f7 −/− E2f8 sgRNA) mTECs. Bars indicate means ± s.e.m. of 3 biological (DEUP1) in mTECs cultured for 7 days at air-liquid interface. Horizontal lines
replicates. Points at value 0 indicate undetectable expression. *P = 0.0284, indicate means ± s.e.m. of 3 biological replicates *P = 0.0295 (one-way ANOVA
**P = 0.0043, ***P = 0.0003, ****P = 0.000008 (one-way ANOVA with Sidak’s with Sidak’s correction).
correction). f, Immunofluorescence images of control, E2f8 mutant, E2f7