Unbiased transcription factor CRISPR screen identifies ZNF800 as master repressor of enteroendocrine differentiation

DIFFERENTIATION

Unbiased transcription factor CRISPR screen

identifies ZNF800 as master repressor of

enteroendocrine differentiation

Lin Lin

1,2,3

*, Jeff DeMartino

2,3

, Daisong Wang

1,2

, Gijs J. F. van Son

2,3

, Reinier van der Linden

1,2

Harry Begthel

1,2

, Jeroen Korving

1,2

, Amanda Andersson-Rolf

1,2

, Stieneke van den Brink

1,2

Carmen Lopez-Iglesias

, Willine J. van de Wetering

, Aleksandra Balwierz

, Thanasis Margaritis

Marc van de Wetering

2,3

, Peter J. Peters

, Jarno Drost

2,3

, Johan H. van Es

1,2

, Hans Clevers

1,2,3

*†

Enteroendocrine cells (EECs) are hormone-producing cells residing in the epithelium of stomach,

small intestine (SI), and colon. EECs regulate aspects of metabolic activity, including insulin levels,

satiety, gastrointestinal secretion, and motility. The generation of different EEC lineages is not

completely understood. In this work, we report a CRISPR knockout screen of the entire repertoire

of transcription factors (TFs) in adult human SI organoids to identify dominant TFs controlling

EEC differentiation. We discovered ZNF800 as a master repressor for endocrine lineage commitment,

which particularly restricts enterochromaffin cell differentiation by directly controlling an endocrine

TF network centered on PAX4. Thus, organoid models allow unbiased functional CRISPR screens

for genes that program cell fate.

nteroendocrine cells (EECs) are special-

ized epithelial cells of the intestinal tract

that, like all other epithelial cell lineages,

derive from regionally specified Lgr5

in-

testinal stem cells (ISCs) (1–5). Balanced

differentiation of EEC lineages from ISCs is

governed by a network of transcription fac-

tors (TFs) (6–12). RNA sequencing and organ-

oidtechnologyhaveilluminatedthetemporal

hierarchy of gene expression profiles during

EEC development (13, 14). By using a time-

resolved reporter allele of Neurog3 (master

regulator of endocrine development) (15, 16),

a real-time and lineage-specific map of mouse

EEC differentiation in vivo was constructed at

single cell–level resolution (12). Overexpres-

sion of NEUROG3 can generate endocrine

cells for functional characterization in human

pancreatic duct cells (17). Leveraging human

gut organoids, this strategy enabled profiling

of EEC subtypes along the proximal-distal

gastrointestinal axis (18). TFs known to specify

EEC subtypes in mice were further examined in

human organoids, revealing a predominantly

conserved regulatory mechanism downstream

of NEUROG3 activatio n. However , discrepanc ies

between mouse and human data were also ob-

served. The regulatory mechanisms upstream

of NEUROG3 and the endogenous repressive

factors that control NEU R O G 3 expression at

the adult stage remain largely unknown. In this

work, we established an organoid-based plat-

form in combination with high-throughput ge-

netic screening for the unbiased discovery of TFs

that govern human EEC lineage commitment.

Results

TFome-wide CRISPR knockout screen in

human SI organoids

Optimized human SI organoids exhibit the

spontaneous gener ation of all major cell line-

ages, including EEC subtypes, without TF over-

expression (19). For the current study, we

aimed to establish a CRISPR screening platform

spanningtheentirerepertoireoftranscription

factors (TFome) in human intestinal organoids

to allow for unbiased and systematic discov-

ery of TFs that govern cell lineage commitment

from adult LGR5

intestinal stem cells. We

used a CRISPR knockout library comprising

7210 single-guide RNAs (sgRNAs) targeting

1800 human TFs, alongside 100 negative con-

trols (20–22). In the human ileum organoid

lineused,EECsweremarkedbyCHGA-IRES-

iRFP670, goblet cells by MUC2-mNeonGreen,

and Paneth cells by DEFA5-IRES-DsRed (fig.

S1A). We performed lentiviral-mediated library

integration in expanding SI organoids (Fig. 1A).

Library coverage was assessed by sequencing

the integrated sgRNAs, comparing the organ-

oids after transduction with the plasmid pool.

Fate separation of CHGA

EECs and MUC2

cells occurs early in secretory progenitors,

whereas DEFA5

Paneth cells emerge later from

MUC2

progenitors (19), illustrating sequential

fate decisions of ISCs (23, 24). We aimed to

track the early CHGA

/MUC2

binary fate

decision. After two-step organoid differentia-

tion (19), we analyzed bulk populations as well

as sorted EECs (CHGA

), goblet cells (MUC2

)

and the CHGA

–

/MUC2

–

/DEFA5

–

population

(fig. S1B). 0 to ~0.7% sgRNAs yielded zero

reads in bulk populations, indicating robust

coverage (fig. S2A). Normalized read-count

distribution and Pearson correlations demon-

strated high concordance between biological

replicates (fig. S2, B and C). sgRNAs that target

essential ISC genes [KLF5 (25) and TCF7L2/

TCF4 (26–29)] were depleted, along with

“generic” cell fi tness g enes (POLR2L, MYC,

and RAD51) at both expansion and differen-

tiation culturing stages (b score < –1; FDR <

0.05 versus plasmid library) (fig. S2D and table

S1). Furthermore, sgRNAs targeting IRF2,an

interferon pathway TF, were highly enriched

upon differentiation, consistent with existing

literature (30, 31).

sgRNAs were then assessed in CHGA

EECs

and MUC2

goblet cells versus the triple

reporter –negative cell population (Fig. 1B

and table S2). As expected, NEUROG3, SOX4,

and INSM1 sgRNAs were depleted in EECs

(8, 15, 16, 32). Additionally, NFIC, TEF, and

ZHX2 appeared essential for EEC commitment.

For each of these, clonal knockout organoids

revealed a mild yet significant reduction of

CHGA

EECs (Fig. 1, B and C, and fig. S3, A to

C). Conversely, GFI1 sgRNAs were enriched in

EECs (Fig. 1, B and C). Mouse Gfi1 is crucial

for goblet cell differentiation by suppression

of Neurog3-driven EEC cell fate (23, 33).

Indeed, GFI1 knockout in human organoids

abrogated goblet cell formation while increas-

ingEECnumbers(fig.S3,AandC,andfig.S14,

CandD).

ZNF800 represses EEC differentiation

The strongest repressor of EEC differentiation

in the screen, ZNF800,isaC

zinc-finger TF

of unknown function (Fig. 1, B and C) (34).

Human ZNF800 is broadly expressed, includ-

ing in the SI and colon epithelium (fig. S4, A

and B). ZNF800

−/−

organoids contained in-

creased EECs and strongly reduced goblet and

Paneth cell numbers (Fig. 1, D and E, and fig.

S4, C and G). Transmission electron micros-

copy confirmed the absence of goblet and

Paneth cells with the characteristic apical

secretory vesicles and the increase of EECs

with the basolateral secretory granules (Fig. 1F).

Similarly, ZNF800 knockout in colon organoid

lines from two different donors also resulted

in a significant increase in EECs (fig. S4, D to G).

We next performed single-cell RNA sequenc-

ing (scRNA-seq) to study the phenotype of

ZNF800

−/−

organoids. For equal representa-

tion, CHGA

cells were sorted from wild type

(WT) and ZNF800

−/−

organoids and pooled

with CHGA

–

cells from the same lines in a 1:4

ratio (Fig. 2A and fig. S5A). Major intestinal

celltypeswereidentifiedbygraph-basedclus-

tering analysis (Fig. 2B and fig. S5B). We

observed the expected reduction of goblet cells

RESEARCH

Hubrecht Institute, Royal Netherlands Academy of Arts and

Sciences (KNAW) and UMC Utrecht, Utrecht, Netherlands.

Oncode Institute, Utrecht, Netherlands.

Princess Maxima

Center for Pediatric Oncology, Utrecht, Netherlands.

The

Maastricht Multimodal Molecular Imaging Institute,

Maastricht University, Maastricht, Netherlands.

*Corresponding author. Email: [email protected] (L.L.);

[email protected] (H.C.)

†Current address: Pharma Research and Early Development of

F. Hoffmann–La Roche Ltd., Basel, Switzerland.

Lin et al., Science 382, 451–458 (2023) 27 October 2023 1of8

Downloaded from https://www.science.org on December 15, 2023

EEC

(CHGA

)

Goblet Cell

(MUC2

)

Paneth Cell

(DEFA5

)

Wild type

ZNF800

-/-

MUC2

DEFA5

CHGA

Human ileum organoid

Cell Percentage (%)

ZNF800

-/-

***

****

Library integration

Expansion

Patterning

Maturation

Differentiated stage

Reporter

negative

Sequencing

gRNA

Enrichment

Log

FoldChange(CHGA

or CHGA

−

)

Density

GFI1

INSM1

NEUROG3

NFIC

SOX4

TEF

ZHX2

ZNF800

−1

−1 0 1

MUC2

Goblet cells (β score)

CHGA

Enteroendocrine cells (β score)

Gene knockout enriched in EECs

Gene knockout depleted in EECs

0.0

0.2

0.4

0.6

TEF

ZHX2

NFIC

INSM1

NEUROG3

SOX4

GFI1

ZNF800

NO-TARGET

−4 −2 0 2

sgRNA enrichment

Sorting

Wild type ZNF800

-/-

Basolateral

Apical

Basolateral

Apical

Fig. 1. TFome-wide CRISPR screen for endocrine differentiation in human

SI organoids. (A) Schematic of TFome CRISPR screen. (B) Scatter plot of

enrichment b score of each TF gene in CHGA

EECs and MUC2

goblet cells

versus the triple reporter–negative cell population. (C) Individual sgRNA

enrichment for genes of interest presented by log

-fold changes in CHGA

EECs

versus triple reporter–negative cell population. Density plot (top) represents the

distribution of nontargeting sgRNAs. (D) Representative confocal images of

WT and ZNF800

−/−

human SI organoids. Representative marker genes for EECs

(CHGA, magenta), goblet cells (MUC2, green), and Paneth cells (DEFA5, red) are

highlighted by fluorescent reporters. Scale bars, 100 mm. (E) Proportion of

different differentiated cell types as determined by FACS analysis of the

respective reporters. Data are shown as mean ± SEM. **P < 0.01; ***P < 0.001;

****P < 0.0001 by multiple t tests with two-stage linear step-up procedure of

Benjamini, Krieger, and Yekutieli with Q = 5% and n =3.(F) Transmission

electron microscopy images of WT and ZNF800

−/−

human SI organoids. Goblet

and Paneth cells are indicated with asterisks in WT organoids, and EECs are

indicated with asterisks in ZNF800

−/−

organoids. Scale bars, 20 mm and

10 mm (zoom in).

RESEARCH | RESEARCH ARTICLE

Lin et al., Science 382, 451–458 (2023) 27 October 2023 2of8

Downloaded from https://www.science.org on December 15, 2023

80% 20%

ZNF800

-/-

DOX

Serotonin secretion (ng/mL

)

****

***

80% 20%

CHGA CHGA

ISCs

TA1

TA2

Early Enterocytes

Enterocytes

Secretory progenitors

Goblet cells

Early EECs

Enterochromaffin cells

other EECs

WT ZNF800

Serotonin ELISA

100

(n = 328)

ZNF800

-/-

(n = 490)

Percentage of cell clusters in

CHGA

cell population (%)

Early EECs

Enterochromaffin cells

Other EECs

12.5%

47.6%

39.9%

33.9%

62.9%

3.3%

tight TRE

3XFLAG

ZNF800 ORF

P2A T2AhPGK

BFP

PuroR rtTA

Doxycycline

MUC2 TagBFP CHGA DEFA5

DOX

ZNF800

DOX

EC cells - CHGA

EC cells - TPH1

L cells - GCG

L cells - PYY

N cells- NTS

D cells - SST

I cells - CCK

G cells - GAST

M/X cells - MLN

M/X cells - GHRL

0.00001

0.0001

0.001

0.01

0.1

ZNF800

-/-

sorted CHGA

cells

****

***

Relative Gene Expression

(vs GAPDH

)

n.d. n.d. n.d.

010 010

−4

UMAP_1

UMAP_2

Paneth cells

ISC

ly Enterocyte

Enterocyte

Secreto

progenitor

Gob

let cell

Early EEC

Enterochromaffin

other EECs

Percent expressed

100

Relative expression

−1

LGR5

NFIA

OLFM4

CCDC85B

MCM6

MKI67

ABP1

RT20

IL32

GDF15

HES1

RNF186

MUC2

REG4

PLA2G2A

FA6

NEUROG3

AX4

CHGA

TPH1

ARX

GCG

Paneth cell

EEC

(CHGA

)

Goblet Cell

(MUC2

)

Paneth Cell

(DEFA5

)

0.0

0.2

0.4

0.6

0.8

1.0

Cell Percentage (%)

***

CHGA CHGA

-/-

RESEARCH | RESEARCH ARTICLE

Lin et al., Science 382, 451–458 (2023) 27 October 2023 3of8

Downloaded from https://www.science.org on December 15, 2023

in ZNF800

−/−

organoids ( 2.3%) compared

with WT organoids (10.8%) (fig. S5C). Clonal

formation efficie ncy of WT and ZNF800

−/−

and colon orga noids revealed no significant

differences (fig. S6, A and B). Cell-cycle pro-

gression in WT versus ZNF800

−/−

organoids

using a Fucci cell-cycle reporter (35) showed

no signif icant differenc es (fig. S6C). The com-

bined results indicate that ZNF800 does not

affect ISC homeostasis in organoids.

EECs were further subclustered based on

markers of early EECs (NEUROG3 and PAX4),

enterochromaffin cells (ECs, CHGA and TPH1)

and the other EEC subtypes (D cells, SST;Lcells,

GCG and PYY; M/X cells, MLN and GHRL;

I cells, CCK; G cells, GAST; N cells, NTS) (Fig.

2B and fig. S7A). Notably, early EECs and ECs

were increased in ZNF800

−/−

organoids, where-

as other EEC subtypes were depleted (Fig. 2,

C and D, and fig. S7A). The same phenotype

was also observed in human colon organoids

(fig. S8, A and B). Thus, ZNF800 loss induces

robust EEC differentiation at the expense of

goblet cells while also driving an EC-biased

trajectory.

We reexpressed WT ZNF800 protein in

ZNF800

−/−

organoids by doxycycline (dox)–

inducible overexpression (Fig. 2E and fig. S8C),

which effectively rescued goblet and Paneth cell

differentiation while repressing EEC lineage

commitment (Fig. 2F). Notably, the EC-biased

differentiation pattern was also reversed upon

ZNF800 rescue, which resulted in decreased

serotonin secretion, a hallmark hormone of

ECs (Fig. 2G and fig. S8D).

ZNF800 represses the endocrine T F

regulatory network

Differential gene expression (DGE) analysis in

CHGA

cell populations of WT and ZNF800

−/−

organoids highlighted distinct expression pat-

terns of EEC-specific TFs (fig. S9, A and B). To

unravel these gene regulatory mechanisms,

we performed single-cell regulatory network

inference and clustering (SCENIC) analysis

on our scRNA-seq dataset (36), which iden-

tified 249 regulons activated across different

cell clusters (table S4). Regulons controlling

EEC commitment displayed higher activity

in ZNF800

−/−

organoids, including SOX4,

NEUROD2,andPAX4 (Fig. 3A). Notably, TEF

and NFIC, discovered in the TFome CRISPR

screen as EEC regulators, also showed regulon

activity in EEC lineages (fig. S9C), supporting

the functional relevance of our hits.

Chromatin immunoprecipitation sequencing

(ChIP-seq) with anti-ZNF800 and anti-FLAG

antibodies revealed 11,565 consensus peaks in

WT organoids and 7085 in rescued ZNF800

−/−

organoids (q value < 0.01) (fig. S10A). Most

ZNF800-binding sites localized within ±5 kb of

transcription start sites (TSS) (fig. S10B). Com-

parison between ZNF800 peak-proximal genes

in WT organoids and in rescued ZNF800

−/−

organoids showed a clear overlap (Fig. 3B).

Of the consensus 3625 ZNF800-binding genes,

we captured 3085 (85%) in our scRNA-seq

dataset (fig . S11A). DGE analysis (adjusted

P value < 0.05) revealed that 461 (15%) of the

ZNF800-bound genes were significantly up-

regulated upon ZNF800 knockout, whereas 59

(2%) were significantly down-regulated (fig.

S11B and table S5), indicating that ZNF800

functions as a transcriptional repres sor.

Gene ontology enrichment analysis of 870

genes (corresponding to the top 1000 anti-

ZNF800 ChIP-seq peak loci) revealed enrichment

of neural and endocrine-gland development

pathways (fig. S11C and table S6). We then

constructed an enrichm ent map for gene sets

involved in gland development, endocrine sys-

tem development, and pancreas develop ment

(Fig. 3C). In particular, INSM1, NEUROG3,and

PAX4 were found to be central TFs in each node

with the highest ZNF800 binding activity. As

predicted from SCENIC (Fig . 3A), ZNF800

bound SOX4 and NEUROD2 loci (Fig. 3D and

fig. S12A). EGR2 and DLL3 were also priori-

tized as top hits by ZNF800 peak abundance;

both are implicated in neurogenesis (37–41)

and specifically expressed in endocrine cell

lineages (fig. S12B). These findings were highly

correlated with their respective gene expres-

sion profiles (fig. S10C and S12B). Notably, we

also identified mild enrichment of ZNF800

binding activity on the NEUROD1 locus (fig.

S12A). Knockout of ZNF800 resulted in increased

gene expression of NEUROD1 (fig. S11B). How-

ever, NEUROD1 did not appear as a hit in our

CRISPR screen (fig. S12C). Neurod1 acts as a late-

stage EEC TF in mice (12). In human EECs,

NEUROD1 was also found to mark late pro-

genitors and mature EECs (18). We therefore

generated NEUROD1

−/−

organoids, which ex-

hibited no effect on EEC or goblet cell differ-

entiation (fig. S12, D to F), whereas profiling of

EEC subtypes demonstrat e d a mil d decr e as e in

ECs, G cells, and M/X cells and a mild increase

in N cells (fig. S12G). These results confirmed

that NEUROD1 is a late-stage TF regulator,

which does not affect the EEC or goblet binary

cell–fate decision.

Most ZNF800-binding loci presented a bi-

valent (H3K4me3- and H3K27me3-containing)

chromatin signature in the SI and colon (Fig.

3D and fig. S13A). Such marks are associated

with stem-cell different iation, allowing timely

responsiveness to TF regulation (42, 43). The

reversibility of gene activation or suppression

by poised ZNF800-binding loci was tested in

organoids with inducible ZNF800 expression

(fig. S13B). Sequential cycles of dox induction

demonstrated that ZNF800 alone can trigger

a dynamic yet reversible equilibrium between

EECs and goblet cells.

GFI1 and ZNF800 function independently in

repressing EEC differentiation

Given that both ZNF800 and GFI1 were discovered

i n our screen as repressors of EEC differentia-

tion, we sought to understand their regulatory

interactions by generating a double knock-

out of GFI1 and ZNF800 (GFI1

−/−

;ZNF800

−/−

)

inhumanSIorganoids(fig.S14A).Furtherab-

rogation of goblet and Paneth cells was ob-

served in GFI1

−/−

;ZNF800

−/−

organoids (fig. S14,

C and D), mirrored by further induction of

EECs. The GFI1

−/−

single knockout did not lead

to EC bias (fig. S14E). Instead, we observed a

mild decrease of EC cells (TPH1)andasubtle

increase in L cells (PYY). Notably, ZNF800

expression was not affected by GFI1 loss (fig.

S14, B and E). Because GFI1 is expressed in

goblet and Paneth cells (33)andtheknockout

of ZNF800 significantly depleted both cell

types, we assayed GFI1 expression in goblet cells

sorted from WT (12 ± 1.0%) and ZNF800

−/−

(3.4 ± 0.14%) organoids to avoid biased cell

heterogeneity of the organoids, which revealed

no differences in GFI1 expression (fig. S14F).

Overall, our results indicated that ZNF800 and

GFI1 function independently as repressors of

EEC differentiation.

Fig. 2. Phenotypic characterization of ZNF800

−/−

organoids. (A) Uniform

manifold approximation and projection (UMAP) of WT and ZNF800

−/−

organoids

from scRNA-seq. Descriptive cluster labels are shown. TA1, transit-amplifying cell

stage 1; TA2, transit-amplifying cell stage 2. (B) Dot plot showing the relative

expression and the percentage of cells expressing selected markers across

scRNA-seq clusters. Two representative markers for each cluster are plotted.

−/−

organoids. (D) Reverse transcription (RT)–qPCR quantification of various EEC

markers in CHGA

cell populations of WT and ZNF800

−/−

organoids. (E) (Top)

Schematic of dox-inducible overexpression construct of ZNF800 . (Bottom left)

Representative immunohistochemical staining images of ZNF800 in ZNF800

−/−

organoids with or without dox-induced ZNF800 expression. (Bottom right)

Representative confocal images of organoids with fluorescent reporters. Scale

bars, 100 mm. (F) Proportion of EECs and goblet and Paneth cells as determined

by fluorescence-activated cell sorting (FACS) analysis of the respective

reporters in ZNF800

−/−

organoids with or without dox-induced ZNF800

expression. (G) Enzyme-linked immunosorbent assay (ELISA) quantification of

serotonin secretion of organoids in different conditions. Data in this figure are

shown as mean ± SEM. N.d. not detected; ns, not significant. *P < 0.05;

**P < 0.01; ***P < 0.001; ****P < 0.0001 by multiple t tests with two-stage

linear step-up procedure of Benjamini, Krieger, and Yekutieli with Q =5%

and n =3.

RESEARCH | RESEARCH ARTICLE

Lin et al., Science 382, 451–458 (2023) 27 October 2023 4of8

Downloaded from https://www.science.org on December 15, 2023

anti-ZNF800

anti-FLAG

Small Intestine

H3K4me3

Transverse colon

H3K4me3

Small intestine

H3K27me3

Transverse colon

H3K27me3

Gene

6kb

INSM1

[0 - 300]

[0 - 24]

[0 - 34]

[0 - 30]

[0 - 84]

SOX4

[0 - 49]

[0 - 15]

[0 - 270]

[0 - 10]

12kb

PAX4

[0 - 209]

[0 - 10]

[0 - 19]

[0 - 31]

6kb

NEUROG3

[0 - 170]

[0 - 10]

[0 - 30]

[0 - 31]

[0 - 90]

8kb

anti−ZNF800

nti−F

3980

(49.8%)

391

(4.9%)

3625

(45.3%)

Genes annotated from

ChIP-seq peaks

endocrine system development

endocrine system developmentendocrine system development

gland development

gland developmentgland development

pancreas development

pancreas developmentpancreas development

INSM1

INSM1INSM1

WNT11

WNT11WNT11

NEUROG3

NEUROG3NEUROG3

TBX1

TBX1TBX1

NOG

NOGNOG

PAX6

PAX6PAX6

ONECUT1

ONECUT1ONECUT1

BMP2

BMP2BMP2

GLI2

INHBB

INHBBINHBB

PDGFRA

PDGFRAPDGFRA

WNT5A

WNT5AWNT5A

BMP6

BMP6BMP6

CDH2

CDH2CDH2

ASCL1

ASCL1ASCL1

DLL1

DLL1DLL1

SHH

SHHSHH

ONECUT2

ONECUT2ONECUT2

NKX6−2

NKX6−2NKX6−2

ALDH1A3

ALDH1A3ALDH1A3

NTN1

NTN1NTN1

WNT7B

WNT7BWNT7B

LEF1

LEF1LEF1

SIX4

SIX4SIX4

NKX2−3

NKX2−3NKX2−3

MAFB

MAFBMAFB

OXTR

OXTROXTR

NKX2−8

NKX2−8NKX2−8

NRG1

NRG1NRG1

BMP7

BMP7BMP7

PKD2

PKD2PKD2

PLXND1

PLXND1PLXND1

SERPINE2

SERPINE2SERPINE2

FOXF1

FOXF1FOXF1

BCL2

BCL2BCL2

WNT3A

WNT3AWNT3A

ID4

ID4ID4

CSF1

CSF1CSF1

GLI3

GLI3GLI3

MSX2

MSX2MSX2

LRP5L

LRP5LLRP5L

NEURL1

NEURL1NEURL1

WNT1

WNT1WNT1

PA M

PA MPA M

NRP1

NRP1NRP1

IGF2

IGF2IGF2

FOXB1

FOXB1FOXB1

TFCP2L1

TFCP2L1TFCP2L1

PTF1A

PTF1APTF1A

PA X 4

Gene set size

100

150

Peak enrichment

Condition

ZNF800

-/-

TF activity

(z-score of AUC)

WTZNF800

-/-

-4

-2

240

ISC

TA1

TA2

Early Enterocyte

Enterocyte

Secretory progenitor

Goblet cell

Paneth cell

Early EEC

Other EECs

Enterochromaffin

SAP30

HMGB3

TCF3

HDAC2

NFYB

MYCN

HES6

KLF13

USF1

ETS1

PAX4

DMRTA2

HOXA9

NEUROD2

TEAD2

SOX4

FOXL2

SNAI1

FOXA2

UNCX

FOXA3

Early EEC

Fig. 3. ZNF800 binds to chromatin regions of TFs involved in endocrine

differentiation. (A) Heatmap with unsupervised clustering of gene regulatory

network activity in different cell clusters within WT and ZNF800

−/−

organoids

and visualized as row z-scores of mean area-under-the-curve (AUC) values.

Zoom-in plot highlighting the top TF regulons specifically activated in early EECs.

Regulons with significant differences between WT and ZNF800

−/−

are indicated

with asterisks, assessed by Wilcoxon rank sum test. (B) Venn diagram of

overlapping genes annotated from ChIP-seq datasets generated by anti-ZNF800

antibody in WT organoids and anti-FLAG antibody in ZNF800

−/−

organoids with

dox-induced ZNF800 expression. (C) An enrichment-network plot depicting

the linkages of gene sets and three selected pathways by gene ontology analysis.

(D) ChIP-seq tracks at the INSM1, PAX4, NEUROG3, and SOX4 loci.

RESEARCH | RESEARCH ARTICLE

Lin et al., Science 382, 451–458 (2023) 27 October 2023 5of8

Downloaded from https://www.science.org on December 15, 2023

PAX4 drives EC differentiation directly

downstream of ZNF800

A panel of genes was found to be up-regulated

across multiple cell types upon ZNF800 knockout

( table S7), including COL4A2, TMEM178B,and

BTB D 1 1 (fig. S15A). ZNF800 binds to these genes,

but not to the down-regulated genes such as

SULT1C2, HMGCS1,andMS4A8 (fig.S15B).De-

spite their modest up-regulation, the expression

pattern and the annotated function of these

genes did not directly relate to the EEC phe-

notype. Therefore, we focused on EEC TFs

with stronger functional implications (fig. S9,

AandB).

We focused on three direct ZNF800 target

genes, INSM1, SOX4, and PAX4 (Fig. 3D), by

performing double knockouts (fig. S16A). EEC

induction was effectively suppressed, but to

different extents (Fig. 4A). Similar to the single

depletion of INSM1 (fig. S3C), INSM1 knockout

in ZNF800

−/−

organoids caused an essentially

complete reversal of EEC induction, confirming

its pivotal role in the TF cascade (8, 44). Addi-

tional knockout of SOX4 or PAX4 in ZNF800

−/−

organoids both reduced EEC numbers.

Next, we analyzed EEC subtypes in the CHGA

cell population from different mutant back-

grounds (Fig. 4B and fig. S17, A and B). Ad-

ditional knockout of INSM1 or SOX4 caused

a mild reduction of EC cells (TPH1)inCHGA

cells while further suppressin g L cells (GCG and

PYY) and D cells (SST). Loss of PAX4 directed

a robust conversion of EC-biased EECs into all

other EEC subtypes. Basal secretion of subtype

hormones (L cell, GLP-1; M/X cell, Ghrelin),

confirmed their functionality (Fig. 4C). We also

performed a single knockout of SOX4 and

PAX4 (fig. S3A, S16B and S18A). The SOX4

−/−

single knockout significantly repressed L cells

(GC G), N cells (NTS), D cells (SST), and G

cells (GAST), whereas MLN-expressing M/X cells

werenotaffected,agreeingwithfindingsin

Sox4

−/−

mice (12)(fig.S18,BtoD).ThePAX4

−/−

single knockout repressed ECs while mildly up-

regulating other EEC subtypes. Overall, PAX4,

as a direct target of ZNF800, appeared respon-

sible for the EC-biased EEC differentiation

trajectory seen upon loss of ZNF800.

Mouse ECs are exclusively Pax4-dependent

(45 ), whereas Arx is essential for L, G, I, and N

cells (10, 45). We therefore probed a possible

interaction between PAX4 and ARX in our

organoids. ARX expression was down-regulated

in ZNF800

−/−

organoids and up-regulated in

EC cells - CHGA

EC cells - TPH1

L cells - GCG

L cells - PYY

N cells- NTS

D cells - SST

I cells - CCK

G cells - GAST

M/X cells - MLN

M/X cells - GHRL

ZNF800

-/-

ZNF800

-/-

; SOX4

-/-

ZNF800

-/-

; INSM1

-/-

ZNF800

-/-

; PAX4

-/-

ZNF800

-/-

ZNF800

-/-

;INSM1

-/-

ZNF800

-/-

;SOX4

-/-

ZNF800

-/-

;PAX4

-/-

CHGA

Cell Percentage (% )

****

0.0

0.5

1.0

1.5

L cell: GLP-1

Secretion (ng/mL)

n.d.

ZNF800

-/-

0.001

0.01

0.1

M/X cell: Ghrelin

n.d.

ZNF800

-/-

;PAX4

-/-

ZNF800

-/-

;PAX4

-/-

ZNF800

-/-

Stem cell

ZNF800

Early EEC

Goblet cell

Enterochromaffin

INSM1

SOX4

PAX4

NEUROG3

PAX4

LN D

M/X

other EEC subtypes

sorted CHGA

cells

0.001

0.01

0.1

100

Relative fold change vs WT

(normalized by GAPDH)

n.d.

****

** ***

***

****

***

****

***

n.d.

ARX gene expression

in CHGA

sorted cells

Relative fold change vs WT

(normalized by GAPDH)

***

ZNF800

-/-

;PAX4

-/-

ZNF800

-/-

PAX4

-/-

ARX

Fig. 4. PAX4 is responsible for EC-biased differentiation trajectory as a

direct downstream target of ZNF800. (A) Proportions of EECs as determined

by FACS analysis of organoids from different genotypes by CHGA reporter.

(B) RT-qPCR quantification of various EEC markers in CHGA

cell populations of

organoids from different genotypes. (C) ELISA quantification of GLP-1 and Ghrelin

secretion of organoids from different genotypes. (D) RT-qPCR quantification of

ARX expression in CHGA

cell populations of organoids from different conditions.

(E) Schematic of ZNF800-driven repression model during endocrine cell

differentiation. Data in this figure are shown as mean ± SEM. N.d. not detected;

ns, not significant. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 by

multiple t tests with two-stage linear step-up procedure of Benjamini, Krieger,

and Yekutieli with Q = 5%, n =3.

RESEARCH | RESEARCH ARTICLE

Lin et al., Science 382, 451–458 (2023) 27 October 2023 6of8

Downloaded from https://www.science.org on December 15, 2023

PAX4

−/−

single-knockout and ZNF800

−/−

;PAX4

−/−

double-knockout organoids (Fig . 4D and fig.

S19, A and B). Reintroduction of PAX4 in

ZNF800

−/−

;PAX4

−/−

organoids significantly

repressed ARX expression (fig. S20, A and B).

Furthermore, as expected, the EEC differentia-

tion tr a je c tor y i n Z N F800

−/−

;PAX4

−/−

organoids

was redirected to an EC-biased phenotype upon

PA X4 reexpression.

Reciprocal transcriptional repression exists

between Pax4 and Arx during mouse pancrea-

tic development (46, 47). We assessed whether

ARX also inhibits intestinal PAX4 expression

by overexpressing ARX in ZNF800

−/−

organoids

(fig. S21A). ARX overexpression was insuffi-

cient to reverse the EC-biased EEC cell type

specification induced by the ZNF800-PAX4

axis (fig. S21, B and C). Furthermore, PAX4

expression was not affected upon ARX over-

expression in the ZNF800

−/−

condition (fig. S21D).

W e next performed ChIP-qPCR of ARX on the

PAX4 locus using FLAG-tagged ARX in human

SI organoids (fig. S21E). Unlike ZNF800, ARX

did not bind to the PAX4 upstream enhancer

region(fig.S21F).WealsoperformedChIP-qPCR

of PAX4 on the ARX locus using FLAG-tagged

PAX4 (fig. S22A), focusing on six ultraconserved

enhancers (48–50) (fig . S22B). PAX4 did not

bind across these enhancers, including hs121,

which exh ibits PAX4 binding activity in the

mouse pancreas (fig. S22C) (49). Overall, our

results strongly supported that PAX4 unilater-

ally inhibits ARX, whereas the inhibitory effect

does not involve direct chromatin interaction

through the previously described regulatory

elements.

Discussion

The human SI epithelium consists of at least

14 main cell types, including six EEC lineages

(51, 52). Leveraging the near-physio logical cell

heterogeneity of human SI organoids, we per-

formed a CRISPR screen for positive and neg-

ative TF regulators of EEC lineage commitment.

Our findings defin e a ZNF800 -repressive

mechanism upstream of the classic endocrine

gene regulatory network (Fig. 4E). Among its

direct downstream targets, INSM1, SOX4,and

NEUROG3 are well described central players

that drive early EEC commitment. We also

found that PAX4, in the absence of ZNF800,

drives an EC-specific cell fate by suppressing

differentiation of all other EEC subtypes.

The PAX4 knockout rescued somatostatin-

producing D cells in ZNF800

−/−

organoids

(Fig. 2D and fig. S7A). Although Pax4 con-

trols both b- and somatostatin-producing d-cell

specification in the mouse pancreas (46), our

single PAX4 knockout had no effect on SST-

expressing D cells (fig. S18C). This prompts

questions into the generalizability of endo-

crine fate regulation between different diges-

tive organs. Double depletion of Pax4 and Arx

promoted d-cell specification in mice (49),

suggesting that a third TF is involved in d-cell

cell differentiation, and that Pax4 is likely to in-

duce a b-cell fate at the expense of d cells. There-

fore, the up-regulation of PAX4 by ZNF800

−/−

in human SI organoids might drive a further

binary cell–fate decision between ECs and

D cells.

ZNF800 is ubiquitously expressed along the

crypt-villus axis in the adult human gut epi-

thelium (fig. S4A), marked by H3K4me3 en-

richment and low H3K27me3 levels (fig. S4B),

indicating an open chromatin state and active

expression. The scRNA-seq dataset of the devel-

oping human gut revealed that ZNF800 is

already expressed at the earliest assayed time

point of 6.1 postconception weeks (53). Given

the downstream TF network described in our

study, it appears likely that ZNF800 also plays

an important function in embryonic develop-

ment. A recent coexpression network study in

mouse pancreatic development (54) revealed

that expres sion of the mouse ortholog zfp800

correlates with endocrine specification from

embryonic day 8 (E8) until E15.5. Global knock-

out of zfp800 caused postnatal lethality and

disrupted early pancreatic development (in-

cluding both endocrine and acinar cells at

E18.5). Thus, the constitutive null phenotype

hindered a mechanistic study of mouse zfp800

function in the endocrine lineages. It is con-

ceivable that ZNF800 might regulate b-cell

differentiation in the human pancreas through

the TF network described in this study.

REFERENCES AND NOTES

1. N. Barker et al., Nature 449, 1003–1007 (2007).

2. Y. Sei, X. Lu, A. Liou, X. Zhao, S. A. Wank, Am. J. Physiol.

Gastrointest. Liver Physiol. 300, G345–G356 (2011).

3. S. J. A. Buczacki et al., Nature 495,65–69 (2013).

4. O. Basak et al., EMBO J. 33, 2057–2068 (2014).

5. O. Basak et al., Cell Stem Cell 20, 177–190.e4 (2017).

6. H. Mutoh et al., Proc. Natl. Acad. Sci. U.S.A. 94, 3560–3564

(1997).

7. F. J. Naya et al., Genes Dev. 11, 2323–2334 (1997).

8. M. S. Gierl, N. Karoulias, H. Wende, M. Strehle, C. Birchmeier,

Genes Dev. 20, 2465–2478 (2006).

9. S. Desai et al., Dev. Biol. 313,58–66 (2008).

10. A. Du et al., Dev. Biol. 365, 175–188 (2012).

11. S. Gross et al., Development dev.130682 (2016).

12. H. Gehart et al., Cell 176, 1158–1173.e16 (2019).

13. T. Sato et al., Nature 459, 262–265 (2009).

14. T. Sato et al

., Gastroenterology 141, 1762–1772 (2011).

15. G. Gradwohl, A. Dierich, M. LeMeur, F. Guillemot, Proc. Natl.

Acad. Sci. U.S.A. 97, 1607–1611 (2000).

16. M. Jenny et al., EMBO J. 21, 6338–6347 (2002).

17. J. Lee et al., eLife 2, e00940 (2013).

18. J. Beumer et al., Cell 181, 1291–1306.e19 (2020).

19. G.-W. He et al., Cell Stem Cell 29, 1333–1345.e6

(2022).

20. J. G. Doench et al., Nat. Biotechnol. 34, 184–191 (2016).

21. S. A. Lambert et al., Cell 172 , 650–665 (2018).

22. L. Lin et al., Cell Rep. 33, 108426 (2020).

23. N. F. Shroyer, D. Wallis, K. J. T. Venken, H. J. Bellen,

H. Y. Zoghbi, Genes Dev. 19, 2412–2417 (2005).

24. T. K. Noah, B. Donahue, N. F. Shroyer, Exp. Cell Res. 317,

2702–2710 (2011).

25. T. Nakaya et al., Cancer Res. 74, 2882–2891 (2014).

26. V. Korinek et al., Nat. Genet. 19, 379–383 (1998).

27. E. Batlle et al., Cell 111

, 251–263 (2002).

28. M. van de Wetering et al., Cell 111, 241–250 (2002).

29. J. H. van Es et al., Mol. Cell. Biol. 32, 1918–1927

(2012).

30. K. Minamide et al., Sci. Rep. 10, 14639 (2020).

31. T. Sato et al., Nat. Cell Biol. 22, 919–926 (2020).

32. A. D. Gracz et al., Gastroenterology 155, 1508–1523.e10

(2018).

33. M. Bjerknes, H. Cheng, Dev. Biol. 345,49–63 (2010).

34. H. S. Najafabadi et al., Nat. Biotechnol. 33, 555–562

(2015).

35. A. Sakaue-Sawano et al., Cell 132, 487–498 (2008).

36. S. Aibar et al., Nat. Methods 14, 1083–1086 (2017).

37. M. H. Sham et al., Cell 72, 183–196 (1993).

38. J. Ghislain, C. Desmarquet-Trin-Dinh, P. Gilardi-Hebenstreit,

P. Charnay, M. Frain, Development 130, 941–953

(2003).

39. A. Desmazières, P. Charnay, P. Gilardi-Hebenstreit, J. Biol. Chem.

284,10831–10840 (2009).

40. S. L. Dunwoodie, D. Henrique, S. M. Harrison,

R. S. P. Beddington, Development 124, 3065

–3076

(1997).

41. R. M. Henke, D. M. Meredith, M. D. Borromeo, T. K. Savage,

J. E. Johnson, Dev. Biol. 328, 529–540 (2009).

42. B. E. Bernstein et al., Cell 125, 315–326 (2006).

43. P. Voigt, W.-W. Tee, D. Reinberg, Genes Dev. 27, 1318–1338

(2013).

44. A. B. Osipovich et al., Development 141, 2939–2949

(2014).

45. A. Beucher et al., PLOS ONE 7, e36449 (2012).

46. P. Collombat et al., Genes Dev. 17, 2591–2603 (2003).

47. P. Collombat et al., Cell 138, 449–462 (2009).

48. G. Bejerano et al., Science 304, 1321–1325 (2004).

49. P. Collombat et al., Development 132, 2969–2980

(2005).

50. D. E. Dickel et al., Cell 172, 491–499.e15 (2018).

51. R. Elmentaite et al., Nature 597, 250–255 (2021).

52. J. Burclaff et al., Cell. Mol. Gastroenterol. Hepatol. 13,

1554–1589 (2022).

53. R. Elmentaite et al.,

Dev. Cell 55, 771–783.e5

(2020).

54. A. B. Osipovich et al., Development 148, dev196964

(2021).

55. ENCODE Project Consortium, Nature 489,57–74

(2012).

56. Y. Luo et al., Nucleic Acids Res. 48 (D1), D882–D889

(2020).

AC KN OW LE D GM E NT S

We thank R. Sherwood (Harvard Medical School, Brigham and

Women's Hospital, Division of Genetics) for providing the TFome

sgRNA library in the CRISPR-v2-FE backbone. We thank D. Krueger,

A. de Graaff, and the Hubrecht Imaging Centre for microscopy

assistance and the Hubrecht Flow Cytometry Core Facility for flow

cytometry analysis and cell sorting. We thank the Máxima Single

Cell Genomics Facility and Utrecht Sequencing Facility for

sequencing support. We thank the Microscopy CORE Lab of

the Faculty of Health, Medicine, and Life Sciences of Maastricht

University for its help in transmission electron microscopy

imaging. We thank the ENCODE Consortium and the

ENCODE production laboratories (B. Bernstein, Broad;

J. Stamatoyannopoulos, UW; B. Ren, UCSD) for generating the

related ChIP-seq public datasets. Funding: This work was

supported by the Netherlands Organ-on-Chip Initiative, a Dutch

Research Council (NWO) Gravitation project (024.003.001) funded

by the Ministry of Education, Culture, and Science of the

government of the Netherlands (H.C.); the Oncode Institute (partly

financed by the Dutch Cancer Society) (H.C. and J.H.v.E.); and the

European Research Council under ERC advanced grant no.

101020405 (GutHormones) (H.C.). The project Organoids in time

with project no. 2019.085 of the research program NWO

Investment Large is financed by the NWO (H.C. S.v.d.B.). Author

contributions: Conceptualization: L.L. and H.C.; Methodology: L.L.,

J.De., D.W., G.J.F.v.S., R.v.d.L., H.B., J.K., S.v.d.B., A.A.-R., C.L.-I.,

W.J.v .d.W., A.B., T.M., M.v.d.W., P.J.P., J.Dr ., J.H.v.E., and H.C.; Investigation:

L.L., J.De., G.J.F.v.S., H.B., J.K., A.A.-R., C.L.-I., and W.J.v.d.W.;

Visualization: L.L., J.De., G.J.F.v.S., and C.L.-I.; Funding acquisition:

J.H.v. E. and H.C.; Project administration: J.H.v. E. and H.C.; Supervision:

J.H.v.E. and H.C.; Writing – original draft: L.L. and H.C.; Writing –

review and editing: all authors. Competing interests: H.C. is an

inventor on patents held by the Royal Netherlands Academy of Arts

and Sciences that cover organoid technology. He is currently

head of pharma Research and Early Development (pRED) at

RESEARCH | RESEARCH ARTICLE

Lin et al., Science 382, 451–458 (2023) 27 October 2023 7of8

Downloaded from https://www.science.org on December 15, 2023

Roche, Basel, Switzerland. H.C.’s full disclosure is given at https://

www.uu.nl/staff/JCClevers/. Data and materials availability:

Further information and requests for resources and reagents

should be directed to the corresponding author. Specific and stable

reagents generated in this study are available and can be

requested from the corresponding author; a completed Materials

Transfer Agreement may be required. Sharing organoid lines used

in this study requires approval by our local institutional review

board. CRISPR screen sequencing data, scRNA-seq data, and

ChIP-seq data from this study have been deposited to the Gene

Expression Omnibus (GEO) under accession no. GSE229586.

ChIP-seq data for histone marks H3K4me3 and H3K27me3 from

human small intestine and colon were obtained from the ENCODE

portal. We downloaded the ChIP-seq datasets from the ENCODE

portal (55, 56) (https://www.encodeproject.org/) with the

following identifiers: ENCFF661MCT, ENCFF988AAQ, ENCFF219ZIO,

ENCFF284XHG, ENCFF603QDZ, ENCFF825TAZ, ENCFF745IGA, and

authors, some rights reserved; exclusive licensee American

Association for the Advancement of Science. No claim to original

US government works. https://www.science.org/about/science-

licenses-journal-article-reuse

SUPPLEMENTARY MATERIALS

science.org/doi/10.1126/science.adi2246

Materials and Methods

Figs. S1 to S22

References (57–76)

Tables S1 to S8

MDAR Reproducibility Checklist

Submitted 17 April 2023; resubmitted 9 August 2023

Accepted 8 September 2023

10.1126/science.adi2246

RESEARCH | RESEARCH ARTICLE

Lin et al., Science 382, 451–458 (2023) 27 October 2023 8of8

Downloaded from https://www.science.org on December 15, 2023

Use of this article is subject to the Terms of service

Science (ISSN 1095-9203) is published by the American Association for the Advancement of Science. 1200 New York Avenue NW,

Washington, DC 20005. The title Science is a registered trademark of AAAS.

to original U.S. Government Works

Unbiased transcription factor CRISPR screen identifies ZNF800 as master

repressor of enteroendocrine differentiation

Lin Lin, Jeff DeMartino, Daisong Wang, Gijs J. F. van Son, Reinier van der Linden, Harry Begthel, Jeroen Korving, Amanda

Andersson-Rolf, Stieneke van den Brink, Carmen Lopez-Iglesias, Willine J. van de Wetering, Aleksandra Balwierz,

Thanasis Margaritis, Marc van de Wetering, Peter J. Peters, Jarno Drost, Johan H. van Es, and Hans Clevers

Science 382 (6669), . DOI: 10.1126/science.adi2246

Editor’s summary

Enteroendocrine cells (EECs) reside in the epithelium of the digestive tract and produce various hormones involved

in metabolism. The generation of different EEC lineages is governed by a dedicated network of transcription factors.

However, given the low efficiency of EEC specification from adult stem cells, it has been difficult to elucidate the

components of this regulatory network. Lin et al. used an optimized human small intestinal organoid culture system

to perform an unbiased, systematic screen for transcription factors that regulate EEC differentiation. The screen

implicated ZNF800 as a key factor exerting a dominant repressive role in controlling the endocrine transcription factor

network. This work showcases the use of optimized human organoids for CRISPR-based functional screens, paving

the way for the identification of additional regulators in human gut physiology and pathophysiology. —Stella M. Hurtley

View the article online

https://www.science.org/doi/10.1126/science.adi2246

Permissions

https://www.science.org/help/reprints-and-permissions

Downloaded from https://www.science.org on December 15, 2023