Opposing Wnt signals regulate cervical squamocolumnar homeostasis and emergence of metaplasia

Chumduri, Cindrilla; Gurumurthy, Rajendra Kumar; Berger, Hilmar; Dietrich, Oliver; Kumar, Naveen; Koster, Stefanie; Brinkmann, Volker; Hoffmann, Kirstin; Drabkina, Marina; Arampatzi, Panagiota; Son, Dajung; Klemm, Uwe; Mollenkopf, Hans-Joachim; Herbst, Hermann; Mangler, Mandy; Vogel, Jörg; Saliba, Antoine-Emmanuel; Meyer, Thomas F.

doi:10.1038/s41556-020-00619-0

Download PDF

Article
Open access
Published: 18 January 2021

Opposing Wnt signals regulate cervical squamocolumnar homeostasis and emergence of metaplasia

Nature Cell Biology volume 23, pages 184–197 (2021)Cite this article

14k Accesses
53 Citations
104 Altmetric
Metrics details

Subjects

Abstract

The transition zones of the squamous and columnar epithelia constitute hotspots for the emergence of cancer, often preceded by metaplasia, in which one epithelial type is replaced by another. It remains unclear how the epithelial spatial organization is maintained and how the transition zone niche is remodelled during metaplasia. Here we used single-cell RNA sequencing to characterize epithelial subpopulations and the underlying stromal compartment of endo- and ectocervix, encompassing the transition zone. Mouse lineage tracing, organoid culture and single-molecule RNA in situ hybridizations revealed that the two epithelia derive from separate cervix-resident lineage-specific stem cell populations regulated by opposing Wnt signals from the stroma. Using a mouse model of cervical metaplasia, we further show that the endocervical stroma undergoes remodelling and increases expression of the Wnt inhibitor Dickkopf-2 (DKK2), promoting the outgrowth of ectocervical stem cells. Our data indicate that homeostasis at the transition zone results from divergent stromal signals, driving the differential proliferation of resident epithelial lineages.

Gene trajectory inference for single-cell data by optimal transport metrics

Article 05 April 2024

Rihao Qu, Xiuyuan Cheng, … Yuval Kluger

A single-cell atlas enables mapping of homeostatic cellular shifts in the adult human breast

Article Open access 28 March 2024

Austin D. Reed, Sara Pensa, … Walid T. Khaled

Simultaneous single-cell three-dimensional genome and gene expression profiling uncovers dynamic enhancer connectivity underlying olfactory receptor choice

Article Open access 15 April 2024

Honggui Wu, Jiankun Zhang, … X. Sunney Xie

Main

Despite extensive self-organization abilities, the mucosal epithelial homeostasis is maintained by the local microenvironment, defined by complex interactions between epithelium and stroma. However, the microenvironment, when altered by various extrinsic and intrinsic factors, can result in disease. The boundaries between two different epithelial types (for example, between squamous stratified and columnar epithelia), termed transition zones, are found, for example, at endo–ectocervical and gastro–oesophageal junctions^1,2. Transition zones are particularly susceptible to infections and development of neoplasia^1,2. Cancers of transition zones occur as two major histological types: adenocarcinomas (ADC) and squamous cell carcinomas (SCC). Carcinogenesis at transition zones is often preceded by metaplasia—the replacement of one epithelium type by another³.

Cervical cancer is the fourth most common cancer in women⁴, more than 90% of which originate at the transition zone^5,6. The cervix connects the uterine cavity with the vagina and consists of the ectocervix and the endocervix. The columnar epithelium lining the endocervix is contiguous with the uterine cavity and separated by the internal os, whereas the squamous epithelium of the ectocervix projects into the vagina. The external os marks the transition from the ectocervix to the endocervix, where the columnar epithelium meets the squamous epithelium, forming the squamocolumnar junction (SCJ). Under certain physiological or pathological conditions, the columnar epithelium at the SCJ is replaced by squamous metaplasia, creating a transformation zone^7,8,9, but the underlying mechanisms remain unclear. Active metaplasia of the cervical transition zone is associated with risk of human papillomavirus (HPV) infection, an aetiological agent of cervical cancers^10,11,12.

A diet deficient in vitamin A induces squamous metaplasia in mice¹³ and is associated with squamous metaplasia¹⁴ and a higher incidence of cervical cancer¹⁵ in humans. The earliest recognized sign of squamous metaplasia is the appearance of cuboidal cells beneath the columnar endocervical epithelium next to the transition zone. The origin of these cells has been variously assigned to undifferentiated subcolumnar reserve cells^16,17,18, residual KRT7⁺ embryonic cells¹⁹, a basal cell ingrowth from the adjacent stratified epithelium, or even transdifferentiation of endocervical columnar or stromal cells¹⁸.

To identify the cellular subsets in the cervix and understand their interplay, we used single-cell RNA sequencing (scRNA-seq), to decipher regulatory relationships between individual cells in their niche context. Further, we used in vivo lineage tracing, tissue-mimetic epithelial 3D organoid models, single-molecule RNA in situ hybridization (smRNA-ISH) and a vitamin A-deficient mouse model of squamous metaplasia. We provide mechanistic insights into how cervical transition zone homeostasis is maintained, the cellular and molecular alterations that drive development of cervical squamous metaplasia and identify the cells that give rise to cervical ADC and SCC. Our study unravels the cervical cell subsets and reveals two committed adult epithelial stem cell types, which give rise to either squamous or columnar epithelial lineages. Homeostasis of these different epithelia at the SCJ is regulated by opposing Wnt gradients, and a shift towards a Wnt-repressive microenvironment drives squamous metaplasia.

Results

Distinct cellular origins of squamous and columnar epithelium

To determine the basic cellular features of the mouse endocervix, the ectocervix and the transition zone, we performed scRNA-seq of isolated cells from each tissue region (Fig. 1a, left). The generated data were combined to perform unsupervised clustering, and cluster identity was assigned on the basis of cell-type-specific marker expression (Fig. 1b,c, Extended Data Fig. 1a,b and Supplementary Table 1). Whereas the pattern of cell types in the endocervix was similar to that of the transition zone, subpopulations of epithelial, stromal, immune and smooth muscle cells in the ectocervix showed distinct transcriptional profiles (Fig. 1b and Extended Data Fig. 1a). An independent cluster analysis on the epithelial cell population to characterize their heterogeneity revealed two major epithelial cell types: Krt5^hi/Krt14^hi squamous (Sq) and Krt8^hi/Krt19^hi columnar (Co) cells. These were further divided into six transcriptionally distinct subclusters (Sq1, Sq2A, Sq2B, Sq3, Co1 and Co2) (Fig. 1d,e, Extended Data Fig. 1c and Supplementary Table 2). We designated myoepithelial cells (Me) as a Krt5^hi/Krt14^hi subcluster that also expressed fibroblast markers, including Col6a2 (Fig. 1d,e). The Sq1 subcluster was enriched for mitotic genes and expressed Trp63, Birc5, Mki67, Cks2 and Hmgb2, indicating basal stem cells of the stratified ectocervix (Fig. 1f and Extended Data Fig. 1d–f). The Sq2A and Sq2B subclusters were enriched for skin development and stress or wounding response genes and expressed Trp63, Krt15, Dkk3 and Notch1. The Sq3 subcluster was enriched for keratinocyte differentiation and skin development genes and expressed Fam25c, Gm94, Krt6a and Krt10, indicating differentiated squamous cells. The Co1 subcluster was enriched for genes for morphogenesis of branching epithelium and lipid catabolic processes, and expressed Anpep, Cxcl17, Krt8 and Krt18. The Co2 subcluster was enriched for genes for stress response, apoptosis and response to unfolded proteins, and expressed Krt19, Ltf, Muc1 and Psca (Fig. 1f, Extended Data Fig. 1g,h and Supplementary Table 3).

**Fig. 1: scRNA-seq of the cervix reveals distinct KRT5⁺ stratified and KRT8⁺ columnar epithelial lineages.**

Fluorescence immunohistochemistry and smRNA-ISH confirmed that KRT5 and KRT8 mark the entire squamous stratified and columnar epithelia, respectively (Fig. 1g,h and Extended Data Fig. 2a–d). To test whether these two epithelia originate from distinct cell types, we used Krt5-Cre^ERT2;Rosa26-tdTomato and Krt8-Cre^ERT2;Rosa26-tdTomato mice to perform lineage tracing. Twelve weeks after induction, KRT5⁺ cells exclusively labelled the stratified epithelium, whereas KRT8⁺ cells exclusively labelled the endocervical epithelium (Fig. 1i,j). Both epithelia merged at the transition zone, with KRT5⁺ cells appearing to displace overlying KRT8⁺ columnar cells (Fig. 1k and Extended Data Fig. 2a,b). Thus, these two major epithelial cell types of the postnatal cervix originate from two distinct lineages.

Opposing stromal Wnt signals define the epithelial borders at the transition zone

To identify which niche-derived signalling maintains these two lineages, we established stem cell-derived organoid models under defined conditions that facilitate long-term propagation. We tested various factors known to play a role in the maintenance of diverse adult stem cells, including the canonical Wnt agonists WNT3A and R-spondin-1 (RSPO1), FGF10, EGF, hydrocortisone, the cAMP pathway agonist forskolin (FSK), the BMP signalling inhibitor noggin, nicotinamide and the TGF-β pathway inhibitor A83-01^{20,21,22,23,24,25}. EGF, FGF10, A83-01 and active BMP signalling were essential for the long-term maintenance of squamous stratified organoids derived from human and mouse ectocervix. By contrast, the presence of WNT3A and RSPO1 was detrimental for both the formation and long-term expansion of ectocervical organoids (Figs. 2a,b and Extended Data Fig. 3a,b). Growth was further increased in the presence of FSK (Fig. 2c). Because cAMP signalling is essential for EGF-mediated neuronal stem cell proliferation²⁶, we speculate that FSK also synergizes EGF signalling in ectocervical stem cells. Ectocervical organoids from both humans and mice could be maintained for more than six months (Extended Data Figs. 3c and 6a). They fully recapitulated the in vivo tissue architecture with stratified layers decorated with E-cadherin (CDH1) (Fig. 2d). The outer layer consisted of p63⁺ basal cells that expressed the proliferation marker Ki67; differentiation into parabasal cells was consistent with p63 labelling decreasing towards the lumen (Fig. 2d). Cells derived from human endocervix gave rise to hollow organoids of a simple columnar epithelial layer when cultured in the presence of Wnt-proficient medium containing WNT3A and RSPO1 (Fig. 2e,f). These organoids faithfully resembled the in vivo tissue architecture with sporadic Ki67 staining (Extended Data Fig. 4a). Endocervical organoids could be maintained for more than seven months (Extended Data Fig. 4b).

**Fig. 2: Wnt-signalling pathway agonists and antagonists have a crucial role in ecto- and endocervical development.**

Of note, endocervix-derived cells cultured in a Wnt-deficient medium (that is, without WNT3A and RSPO1) gave rise to p63⁺, KRT5⁺ stratified organoids, resembling ectocervix organoids (Fig. 2e–g and Extended Data Fig. 4c,e–g). Conversely, ectocervix-derived cells gave rise to only squamous organoids independently of Wnt (Fig. 2a). Since formation of columnar organoids was Wnt-dependent, unlike that of squamous organoids, we investigated the source of Wnt signals in the cervix. Microarray analysis of organoids shows that the transcriptional regulation of Wnt in endocervical cells diverged from that in the ectocervix; Wnt agonists were upregulated in the columnar epithelium, whereas the Wnt antagonists DKK1 and KREMEN1 were upregulated in the squamous epithelium (Fig. 2h and Supplementary Table 5). Further, transcriptional profiling and confocal microscopy of ecto- and endocervix-derived human organoids revealed distinct patterns of keratin expression in congruence with data from the respective tissues (Fig. 1e and Extended Data Figs. 1e–h and 4d). Ectocervical organoids expressed KRT5, p63, KRT17 and the luminal cell marker loricrin, whereas endocervical columnar organoids expressed KRT8 and KRT19 (Fig. 2i–k and Extended Data Fig. 4f).

To gain insights into how stromal factors contribute to these two distinct cervical epithelial homeostases, we determined the heterogeneity of stromal populations in the endocervix, ectocervix and transition zone. Unsupervised clustering analysis of combined scRNA-seq data revealed five transcriptionally distinct subclusters of stromal populations (Fig. 3a and Supplementary Table 4). The stromal 1 (S1), S2, and S3 subclusters represented cells from endocervix and transition zone, commonly expressing Hes1, Nfkbia and Egr1. The S2 subcluster highly expressed Wnt16 and Axin2, and the S3 subcluster highly expressed Mmp3, Adh1, Ces1d, Cxcl16, Rspo1 and Rspo3. Subclusters S4A and S4B represented ectocervical stromal cells; subcluster S4A showed preferential expression of Sfrp1 and Bmp4, and subcluster S4B expressed Dkk2 and Crabp1 (Fig. 3b,c).

**Fig. 3: Stromal compartments of ecto- and endocervix show distinct patterns of expression.**

smRNA-ISH analysis showed that the spatial distribution of Wnt agonists and antagonists in the underlying stroma defined the boundary between the two epithelia, thus confirming the scRNA-seq data (Fig. 3d). Rspo1 and its downstream target Axin2 were highly expressed in the stroma beneath the columnar epithelium, and Rspo3 was expressed in the endocervical muscularis (Fig. 3d and Extended Data Fig. 5a,b,d). Notably, the expression of the gene for the Wnt antagonist Dkk2 was restricted to the stroma proximal to the ectocervical basal cells, while the squamous epithelium highly expressed Dkk3 (Fig. 3d and Extended Data Fig. 5g,h). By contrast, in the endocervix, Dkk3 expression was high throughout the stroma (Fig. 3d and Extended Data Fig. 5h). Expression levels of Rspo2, Rspo4, Dkk1 and Dkk4 did not show notable regional variation (Extended Data Fig. 5c,e,f,i). To further investigate the effects of DKK2 and DKK3, endocervical organoids were reseeded in medium containing WNT3A and RSPO1 plus DKK2 or DKK3, or both DKK2 and DKK3. The Wnt-deficient medium served as a control. The presence of DKK2 resulted in a substantial decrease in columnar organoid size, similar to that found in Wnt-deficient medium, confirming that DKK2 exerts an inhibitory effect. This was prevented when the medium additionally contained DKK3 (Fig. 3e). These findings corroborate earlier studies that, in contrast to other DKK members, DKK3 either has no effect on Wnt signalling or functions as a Wnt agonist^27,28,29,30. Blocking Wnt signalling using inhibitor of Wnt production 2 (IWP2)^31,32, suppressed endocervical organoid growth but did not affect ectocervical organoids (Fig. 3f,g). This indicates that squamous epithelium does not depend on cell-autonomous Wnt signalling. Moreover, there was no difference in the expression of the lineage-specific markers KRT5, KRT8 and p63 in columnar and squamous organoids upon treatment with DKK2 or IWP2 (Fig. 3h). Together, these data show that the epithelium of the cervix is maintained by two distinct stem cell populations whose fate is determined by opposing Wnt-signalling microenvironments, with a defined switch at the transition zone (Fig. 3i).

Wnt antagonists, Notch and EGFR signalling maintain ectocervical stemness and differentiation

Next, we sought to identify the mechanisms that control self-renewal and differentiation in human ectocervix. Microarray analysis showed a higher expression of Notch-related genes in ectocervical squamous organoids than in endocervical columnar organoids (Fig. 4a). To find out how differentiation is controlled, we performed a comparative analysis of 2D-grown ectocervical cells (2D-ecto), three-day-old early organoids (EO-ecto), and two-week-old mature organoids (DO-ecto). Cultures of 2D-ecto were positive for CDH1 and p63, and exhibited organoid-forming potential (Fig. 4b and Extended Data Fig. 6a). EO-ecto consisted of 8–16 undifferentiated cells positive for Ki67 and p63 (Fig. 4c and Extended Data Fig. 6b). DO-ecto consisted of several stratified layers, more than two-thirds of which were differentiated cells, and the rest were proliferating cells (Fig. 4d and Extended Data Fig. 6b). Gene expression patterns of 2D-ecto and EO-ecto showed high similarity and displayed a set of differentially expressed genes compared with DO-ecto (Fig. 4e and Supplementary Table 5–8). We performed a comparative analysis with a stem cell signature (that is, frequently upregulated genes in stem cells from diverse tissues³³). The results confirmed that the expression profiles of early cells (2D-ecto and EO-ecto) showed a high similarity to this stem cell signature, but mature DO-ecto did not (Fig. 4f and Supplementary Table 6–9). This was further supported by comparisons with expression profiles of ground-state stem cells versus their respective differentiated cells (Extended Data Fig. 6c and Supplementary Tables 6–8 and 10). Thus, we conclude that cells from 2D-ecto and EO-ecto correspond to ectocervical stem cells.

**Fig. 4: Stemness and differentiation of ectocervix depend on Wnt antagonist, Notch and EGFR signalling.**

A survey of genes upregulated in ectocervical stem cells versus differentiated cells revealed high expression of the Notch ligands delta-like ligand 3 (DLL3) and manic fringe (MFNG) (Fig. 4g). By contrast, differentiated cells expressed higher levels of NOTCH2 and NOTCH3 receptors and their targets, including the transcription factor HES1 and presenilin 1 (PSEN1), a core component of γ-secretase (Fig. 4g). Ectocervical stem cells also expressed DKK1 and its receptor KREMEN2 (Fig. 4g). Accordingly, blocking Notch activation using the γ-secretase inhibitor DBZ reduced growth of organoids (Fig. 4h and Extended Data Fig. 6d), which failed to differentiate and stratify (Fig. 4i). Thus, it appears that the ectocervical stem cells act as a source of Notch signals, whereas the differentiated cells act as Notch signal-receiving cells, a trans-activating interaction that facilitates differentiation and epithelial stratification.

Further, gene set enrichment analysis (GSEA) revealed that genes regulated by transcription factors downstream of Notch and EGFR–RAS–MAPK signalling, were highly enriched in ectocervical stem cells, including targets of the transcription factors AP1, CREB, ETS, NERF, ELK1, E2F, SRF, MYC and YY1^34,35,36,37 (Fig. 4j). These two pathways function together to regulate proliferation and differentiation^38,39,40, in congruence with the essential role of EGF in the formation of stratified organoids (Fig. 2b and Extended Data Fig. 3a). Conversely, genes belonging to the RAS-antagonistic NF1 pathway⁴¹ were enriched in differentiated cells. Together, these observations indicate that Wnt antagonists, along with EGFR and Notch-inducing pathways, regulate ectocervical stemness and differentiation.

Remodelling of the stromal compartment drives the emergence of KRT5⁺ squamous metaplasia

Having illuminated the signalling pathways and cellular components involved in stemness and differentiation of the ectocervical squamous epithelium, we next examined their role in the emergence of squamous metaplasia. When primary human endocervix-derived cells were grown in 2D in a Wnt-proficient medium, such cultures contained only a few KRT5⁺ or p63⁺ cells; after a transfer to organoid conditions with the same medium, they gave rise to columnar organoids (Extended Data Fig. 7a,b). However, these 2D cells, when cultured in a Wnt-deficient medium, showed a clear enrichment for KRT5⁺ and p63⁺ cells. After a transfer to organoid conditions, these cells produced only squamous organoids, including the characteristic basal and parabasal p63⁺ cells, even in the presence of WNT3A and/ RSPO1 (Extended Data Fig. 7a,b). By contrast, human ectocervical cells that were first grown as 2D in either Wnt-proficient or -deficient medium gave rise only to squamous organoids after a transfer to a Wnt-deficient medium (Extended Data Fig. 7c). Together, this indicates the presence of squamous stem cells in the endocervix that can give rise to KRT5⁺p63⁺ cells given a suitable microenvironment.

We aimed to test whether the squamous and columnar organoids originate from distinct lineage-specific stem cells, or rather a transdifferentiation of columnar to squamous epithelial cells occurs in the absence of Wnt factors. We induced lineage tracing in Krt5-Cre^ERT2;Rosa26-tdTomato and Krt8-Cre^ERT2;Rosa26-tdTomato mice. Five weeks later, epithelial cells were isolated from the ecto- and endocervix of both genotypes and grown as organoids in Wnt-deficient or -proficient media. Endocervical organoids derived from Krt8-Cre^ERT2;Rosa26-tdTomato mice were found to be labelled, whereas matched ectocervical squamous organoids were not. However, when grown in Wnt-deficient medium, endocervical lineage-labelled cells did not give rise to any labelled KRT5⁺p63⁺ squamous organoids, excluding transdifferentiation ability of these cells (Fig. 5a,b).

**Fig. 5: Two distinct stem cells from the endocervix give rise to columnar or squamous stratified lineages depending on the microenvironment.**

Conversely, ectocervical squamous organoids derived from Krt5-Cre^ERT2;Rosa26-tdTomato mice were labelled, whereas matched columnar endocervical organoids were not (Fig. 5a). Our results show that the postnatal columnar and stratified cervical epithelia are derived from two different lineages, each of which respond to a particular microenvironment. Changing the Wnt signals does not induce the transdifferentiation of columnar cells to the stratified cells or vice versa.

To analyse metaplasia development in vivo, we used a model of mice fed with a vitamin A-deficient diet, in which abnormal foci of squamous metaplasia develop in the endocervix¹³ (Fig. 5c). To identify changes in the epithelial and stromal compartments, we performed scRNA-seq and clustering analysis of cells isolated from the endocervix, ectocervix, transition zone and endocervix metaplasia (that is, metaplastic foci) (Fig. 5c–e and Supplementary Table 11). The epithelial populations were assigned to the Sq, Co and Me subclusters (Fig. 5g and Supplementary Table 12), as done for the healthy tissue (Fig. 1c). A phylogenetic tree, generated on the basis of the expression profiles of each epithelial subcluster, revealed that the squamous subpopulations (Sq1, Sq2A, Sq2B and Sq3) and Me cells (Me) were transcriptionally similar but distinct from the columnar subpopulations (Co1 and Co2), as shown by the distance between nodes (Fig. 5h). We next generated a Sankey diagram to analyse the transcriptional contribution of epithelial cells from each tissue region to different subclusters. The endocervix metaplasia demonstrated amplification of expression profiles similar to the squamous and myoepithelial types compared with a healthy endocervix (Fig. 5i). To characterize the alterations in the stromal population, we performed unsupervised clustering of stromal cells from the endocervix, ectocervix, transition zone and endocervix metaplasia (Fig. 6a), and defined stromal subclusters according to gene expression (Fig. 6b and Supplementary Table 13). The stromal compartment of endocervix metaplasia had a distinct clustering profile from those of the healthy endocervix, ectocervix and transition zone (Figs. 3a and 6b). In agreement, a Sankey diagram revealed the emergence of distinct stromal cell subclusters in the endocervix metaplasia (S2B, S2C and S2D) compared with healthy endocervix and ectocervix (Fig. 6c). We also observed an increased expression of Dkk2 in a subcluster of the endocervix metaplasia (Fig. 6a,b,d,e), which primarily corresponds to the ectocervical S4B in healthy mice. These data indicate an extensive remodelling of the stroma during metaplastic development.

**Fig. 6: Endocervical stroma undergoes extensive remodelling during metaplasia.**

To further consolidate the lineage properties of stratified and columnar epithelia and spatial changes of the cervical microenvironment during metaplasia, we performed lineage tracing, smRNA-ISH and immunohistochemistry in our metaplasia mouse model. The mice exhibited upregulation of Dkk2 in the stroma of the endocervix and uterine horns (Fig. 6e,o compared with Fig. 3d and Extended Data Fig. 5g). We observed an emergence of subcolumnar KRT5⁺, KRT17⁺ and p63⁺ cells, which were negative for KRT8 (Fig. 6f–n and Extended Data Fig. 7d). These cells were similar to healthy ectocervical squamous and endocervical subcolumnar reserve cells, which appeared to develop into a metaplastic stratified epithelium. Axin2, which is typically expressed in the endocervix, remained unaltered and expression of KRT8 and KRT7 was restricted to the columnar epithelium (Fig. 6f,k–m and Extended Data Fig. 7e,f). By inducing lineage tracing in Krt5-Cre^ERT2;Rosa26-tdTomato and Krt8-Cre^ERT2;Rosa26-tdTomato mice fed with a vitamin A-deficient diet, we supported our notion that KRT5⁺ cells gave rise to the endocervix-localized squamous metaplasia, whereas KRT8⁺ cells gave rise to the columnar epithelium as in healthy mice (Fig. 6p,q). We conclude that while Wnt agonists support formation of columnar epithelium, the local upregulation of the Wnt antagonist DKK2 in the stroma drives the proliferation of KRT5⁺ reserve cells, resulting in squamous metaplasia.

Gene expression patterns of cervical SCC and ADC correlate with squamous and columnar lineage organoid signatures

Adult stem cells are susceptible to transformation and often constitute the cells of origin for various cancers⁴². To date, it is unclear from which cell types ADC and SCC originate. Thus far, their possible origin has been ascribed to phenotypic markers that characterize physiological cell types^{19,43,44,45,46}. We used the gene expression signatures of squamous and columnar cervical organoids to determine the possible cells of origin of cervical cancers. We retrieved publicly available gene expression data—from The Cancer Genome Atlas (TCGA) study⁴⁷—of 178 cervical cancers, including tumour classes based on an integrated clustering of mRNA, miRNA, genomic copy number and methylation data with iCluster, which divides cervical cancers into ADC and keratin-high and keratin-low SCC groups, in which the keratin-low SCC group shows a lower expression of squamous-lineage-specific markers⁴⁷. We selected genes that were differentially expressed between ectocervical squamous and endocervical columnar organoids to classify the cancer samples into squamous-like and columnar-like. We found an agreement between cancers classified as columnar-like or squamous-like and their histological classification as ADC or SCC, respectively; this was also evident from comparison with the TCGA clusters of ADC or keratin-high and keratin-low groups (Fig. 7a and Extended Data Fig. 8d,e). A group of cases with a low tumour content, which showed no apparent similarity with either ectocervical or endocervical signatures, were classified as undetermined (Fig. 7a, Extended Data Fig. 8c and Supplementary Table 14).

**Fig. 7: Transcription profiles of cervical SCCs and ADCs correlate with squamous and columnar epithelial lineages.**

Nevertheless, 59 out of 62 of the undetermined cases fell in the TCGA-defined keratin-low or keratin-high SCC clusters. Notably, cancer samples classified as columnar-like were mainly KRT5^low, KRT7^high and TP63^low, whereas samples classified as squamous-like and undetermined were mainly KRT5^high and TP63^high with a mixed KRT7 status (Fig. 7a and Extended Data Fig. 8d,e). This suggests that the undetermined group may consist of SCCs within or outgrown into the endocervix, leading to the presence of contaminating KRT7⁺ endocervical columnar cells in the samples.

It was suggested that a small cell population located in the transition zone (called SCJ cells) expresses KRT7, CD63, GDA, AGR2 and MMP7 and that these cells are the precursors of both SCC and ADC^19,48. We thus investigated the gene expression levels in organoids and TCGA cervical cancer samples, measuring the proposed SCJ markers¹⁹ as well as KRT5. The expression of these markers was significantly higher in endocervical organoids and ADC than in ectocervical organoids and SCC; the opposite trend was shown for KRT5 (Fig. 7c,d). Immunohistochemistry and smRNA-ISH of human and mouse cervix tissue sections confirmed that KRT5 was expressed throughout the squamous epithelium. KRT7 was expressed throughout the columnar epithelium and, to a lesser extent, in the squamous epithelium, rather than being restricted to the transition zone alone (Fig. 8a,b and Extended Data Fig. 9a). Ectocervical organoids were positive for the squamous epithelial markers KRT5 and CTSA, whereas endocervical organoids were positive for the columnar epithelial markers KRT7, AGR2, GDA and KRT18 (Fig. 8c and Extended Data Fig. 9b–d). This implies that the reported SCJ cells are not distinct from the endocervical columnar lineage and are not the cells of origin for SCC.

**Fig. 8: Molecular expression patterns of lineage markers in healthy tissue, organoids and cervical cancers.**

In agreement with the notion that both keratin-high and keratin-low SCC cases derive from the squamous lineage, our analysis also revealed a set of differentially expressed genes between keratin-high and keratin-low, and ADC groups. We observed bimodal gene expression in columnar-like and squamous-like cancers, including MUC5B, KRT5 and CSTA, that differentiates ADC from SCC (Fig. 7b). By contrast, the proposed SCJ markers KRT7, AGR2 and GDA exclusively labelled ADC but not SCC sections (Fig. 8d). In summary, the majority of cervical cancers can be divided into two groups on the basis of molecular signatures: SCCs most probably originate from the KRT5⁺ squamous lineage of the ectocervix, whereas ADCs most probably originate from the KRT7⁺KRT8⁺ columnar lineage of the endocervix.

Discussion

Adult tissue homeostasis is maintained by niches established by complex interactions between stem cells and their surrounding microenvironment⁴⁹. When tissue integrity is disturbed due to infections or other assaults, the disturbance is usually followed by substantial reorganizations of the niche, facilitating the restoration of tissue homeostasis⁵⁰. The transition zones of the mucosal epithelium constitute critical zones of enhanced susceptibility to infections and carcinogenesis^51,52,53,54. The cervical transition zone appears to be particularly vulnerable to HPV infections, the prime aetiological agent of cervical cancer. HPV is thought to infect the reserve stem cells at the transition zone, and the resulting dysregulation may lead to neoplasia⁵⁵. Pre-existing metaplasia at the cervical transition zone may develop into neoplasia under the influence of HPV infection^10,11. Thus, revealing the principles of niche homeostasis and its cellular composition is crucial to understanding the effects of intrinsic and extrinsic disturbances, including viral and bacterial infections.

In this study, we show that the cervical squamous and columnar epithelia arise from two distinct lineage-specific stem cell types and define the subpopulations of these lineages. Further, the endo- and ectocervix harbour distinct stromal subpopulations that constitute a microenvironmental gradient maintaining the transition zone. The regulation of stem cell maintenance, differentiation and tissue patterning by signalling gradients is a universal principle in development⁵⁶, in which Wnt signalling is indispensable^20,57. We demonstrate the presence of a Wnt gradient in the underlying stroma that selectively drives the columnar lineage while imposing quiescence on squamous-lineage-specific stem cells present in the same tissue. In squamous metaplasia, the endocervical stroma undergoes extensive remodelling. Here, we observed an upregulation of the Wnt inhibitor DKK2 in a stromal subpopulation. This transition to a Wnt-repressive microenvironment may activate quiescent squamous-lineage stem cells that invade the columnar epithelia at the transition zone or as metaplastic foci within the endocervix (Fig. 8e). Since oestrogen acts as a cofactor during HPV-driven cervical carcinogenesis⁵⁸, an interplay between the oestrous cycle and Wnt signalling at the transition zone might occur and influence metaplasia and cancer development.

In concordance with previous studies^16,17,18, we also found that subcolumnar KRT5⁺ cells are variably present at the transition zone in humans and mice. Furthermore, our data suggest these precursors of squamous metaplasia are present throughout the cervix and become activated in response to a Wnt-repressive environment. We also detected a divergent epithelial subpopulation in the cervix, classified as myoepithelium, which shares gene expression profiles with KRT5⁺ cells and myofibroblasts. These cells probably represent candidate precursors of the squamous lineage.

This study reveals a delineation of cervical ADC and SCC, implying that these cancers also originate from our two identified distinct epithelial lineages. Prophylactic ablation of the SCJ, where HPV infections and neoplasia are often observed, has been proposed to prevent neoplasia development⁵⁹, but such efforts have failed to do so^59,60,61. Our results suggest that this preventive ablation alone may not eliminate potential cervical cancer precursors, as new SCJs can develop upon activation of quiescent KRT5⁺ stem cells, which could present target sites for HPV infection and carcinogenesis. Furthermore, our organoid models reveal differences in the regulation of those two epithelial lineages similar to their respective cancers, thus providing an opportunity to study carcinogenesis and to identify specific therapeutic targets for each lineage. Nevertheless, organoids may not fully recapitulate the multifaceted interactions between various cell types and may thus represent an approximation of greater complexity in vivo.

From the molecular and cellular points of view, the present elucidation of the mechanisms that maintain cervical epithelial junctions provides an important conceptual advance. It suggests that homeostasis at these sites is not maintained by the transdifferentiation of one epithelial cell type to another. Instead, the adult tissue resembles a mosaic of different lineage-specific stem cell populations activated by the microenvironment in response to extrinsic or intrinsic signals. This concept of mucosal transition zone homeostasis fits well with other recent observations on mucosal stem cell identity⁶² and provides a basis for future investigations into transition zone homeostasis in other tissues with high clinical relevance.

Methods

Antibodies and chemicals

The following antibodies and chemicals were used: mouse anti-p63 (Abcam, ab375), rabbit anti-p63 (Abcam, ab53039), mouse anti-E-cadherin (BD Biosciences, 610181), rabbit anti-Ki67 (Abcam, ab16667), rat anti-Ki67–FITC (eBioscience, 11-5698), rabbit anti-KRT5 (Abcam, ab52635), rabbit anti-KRT5–Alexa488 (Abcam, ab193894), mouse anti-KRT7 (Santa Cruz, sc-23876), rabbit anti-KRT7 (Abcam, ab181598), rabbit anti-KRT7–Alexa555 (Abcam, ab209601), rabbit anti-CSTA (Sigma, HPA001031), rabbit anti-AGR2 (Proteintech, 12275-1-AP), mouse anti-MUC5B (Abcam, ab77995), rabbit anti-GDA (Sigma, HPA019352), Hoechst (Sigma, B2261), rabbit anti-KRT17 (Abcam, ab109725), mouse anti-KRT19 (Abcam, ab7754), mouse anti-KRT18 (Abcam, ab668), donkey anti-rabbit–Alexa Fluor 488 (Jackson Immuno Research, 711-546-152), donkey anti-rabbit–Cy3 (Jackson Immuno Research, 711-166-152), donkey anti-rabbit–Alexa Fluor 647 (Jackson Immuno Research, 647 711-605-152), donkey anti-mouse–Cy5 AffiniPure (Jackson Immuno Research, 715-175-151), Draq5 (Cell Signaling, 4085), γ-secretase inhibitor XX (DBZ) (Calbiochem 565789) and p38 inhibitor SB202190 (Sigma, S7067). Antibody concentrations and the link to validation by the provider are shown in Supplementary Table 15.

Mice

All procedures involving animals were approved by the national legal, institutional and local authorities at Max Planck Institute for Infection Biology. This study is compliant with all relevant ethical regulations regarding animal research. All animals were maintained in autoclaved micro isolator cages and provided with sterile drinking water and chow ad libitum. Four- to twenty-week-old female mice were used for this study. Wild-type C57BL/6, Krt5-Cre^ERT2 (ref. ⁶³) and Krt8-Cre^ERT2 (ref. ⁶⁴) mice (Jackson Laboratory) were bred to Rosa-tdTomato⁶⁵ mice to generate mice labelled in Cre-expressing cells. For KRT5⁺ or KRT8⁺ lineage analysis, Cre recombinase was induced in female mice by administering tamoxifen (Sigma) intraperitoneally (0.25 mg per g body weight in 50 μl corn oil) at 4 weeks of age for 3 consecutive days. Mice were euthanized at 14–20 weeks of age, and the genital tracts were removed for further analysis. Experiments were performed in at least three biological replicates per condition. Mice were randomly allocated to experimental groups in all experiments.

Single-cell isolation and sequencing

Following FGT extraction, the endocervix, transition zone and ectocervix from control mice and endocervix from mice with metaplasia were cut out. Tissue samples were washed in sterile PBS (Gibco, 14190-094) and minced with scissors. Minced tissue was processed separately by incubating in 0.5 mg ml⁻¹ collagenase II (Calbiochem, 234155) in a shaker (45 min, 37 °C). Tissue and dissociated cells were pelleted (7 min, 1,000g, 4 °C), the supernatant was discarded, cells were resuspended in warm TrypLE Express (Gibco, 12604021) and incubated in a shaker (15 min, 37 °C). The pellet was resuspended in Advanced DMEM/F-12 (ADF) medium (Invitrogen) and passed through a 40 µM cell strainer (BD Falcon, 352340). For single-cell sequencing, we combined cells isolated from three mice and processed each tissue separately. Cells were pelleted (7 min, 1,000g, 4 °C) and resuspended in PBS containing 0.04% w/v BSA (400 μg ml⁻¹) at a concentration of 1,000 cells per μl. A 10x Chromium Controller was used to partition single cells into nanolitre-scale Gel-Bead-In-EMulsions (GEMs) and Single-Cell 3′ Reagent Kit v2 for reverse transcription, cDNA amplification and library construction according to the manufacturer’s protocol. Approximately 13,200 cells per sample were loaded onto the controller. A SimpliAmp Thermal Cycler (Applied Biosystems) was used for amplification and incubations. Libraries were quantified by Qubit^TM 3.0 Fluorometer (ThermoFisher), and quality was checked using a 2100 Bioanalyzer with a High Sensitivity DNA-kit (Agilent). Sequencing was performed in paired-end mode with S1 100-cycles kit using Novaseq 6000 sequencer (Illumina). Approximately 400 million reads were allocated per sample with at least 70,000 reads per cell.

Computational analysis of scRNA-seq data

Cellranger pipeline

The Cell Ranger v3.0.1 software suite was obtained from 10x Genomics (https://support.10xgenomics.com/single-cell-gene-expression/software/downloads/latest). Raw sequencing data were first de-multiplexed and quality checked using the ‘mkfastq’ script. For all libraries, alignment and transcript quantification were performed with the standard ‘count’ script against the mm10 mouse genome assembly. All samples were aggregated using the ‘aggr’ script with the default normalization step (by downsampling) enabled. Only the healthy cervix samples were aggregated to form a control dataset under the same settings. The pre-filtered count matrices of the combined datasets were imported into R for further processing. A total of 25,932 cells were detected by Cell Ranger for all samples.

Quality control

The expression matrices were filtered, and potential doublets were removed by excluding barcodes with less than 250 genes, more than 4,000 genes, and more than 15,000 unique molecular identifiers. Barcodes with more than 10% of mitochondrial genes were excluded. Genes that were not detected for any barcode were removed. Potential doublets were scrutinized using Scrublet from the scran package (1.14.6)⁶⁶, and marker gene expression was used to assess whether a hybrid transcriptome could mimic the cell type in focus; these analyses did not lead to further removal of cell-associated barcodes.

Clustering

Initial clustering was performed using the R package Seurat v2.3.4. Count data was log-normalized before identification of highly variable genes based on the following criteria: 0.0125 < mean of non-zero values < 4 and s.d. > 0.5. Unwanted variation due to library size and proportion of mitochondrial gene content was regressed out during data scaling. Features were selected using PCA: the first 21 principal components were selected for dimensional reduction and clustering of the whole datasets, 8 components were selected for the stromal subsets, and 6 components were selected for the epithelial subsets. A 2D representation was computed using the UMAP algorithm⁶⁷. For clustering, a shared nearest neighbours graph was constructed using the Seurat functions BuildSNN without pruning and FindClusters using the SLM algorithm⁶⁸. UMAPs (Fig. 1b and Extended Data Fig. 1a,b) were derived from analysing a total of 13,773 cells from ectocervix, endocervix and transition zone. From these, 1,626 epithelial cells (Fig. 1d and Extended Data Fig. 1c–h) and 6,380 stromal cells (Fig. 3a,c) were analysed. UMAPs (Fig. 5d,e) were derived from analysing a total of 22,590 cells from ectocervix, endocervix, transition zone and metaplasia samples. From these, 1,833 epithelial cells (Fig. 5f,g and Extended Data Fig. 1c–h) and 11,634 stromal cells (Fig. 6a–e) were analysed.

Cell cycle annotation

Cell cycle stage annotation was based on a list of cell cycle markers⁶⁹. The Seurat function CellCycleScoring was used to compute quantitative scores for G2M- and S-phase and assign qualitative labels to each barcode.

Differential gene expression analysis

Differential expression analysis was performed between cell types/clusters using the FindAllMarkers function from the R package Seurat using default settings.

Batch-effect assessment

Technical variability was assessed by comparing the cell distribution across each technical batch within each transcriptional profile of assigned cell types using UMAPs. Cells derived from the endocervix of metaplasia mice were sequenced as two technical replicates, equally distributed across the identified clusters (Fig. 5d, right).

Further analysis

GO enrichment of cluster markers and differentially expressed genes was performed using the R package clusterProfiler. Dendrograms displaying the relationship of clusters were computed with the Seurat function BuildClusterTree.

Vitamin A-deficient diet

From birth onwards, experimental mice and their mothers were fed with a vitamin A-deficient diet (SAFE, U8978P-0074) or a control diet with added Vitamin A at physiological levels (6 IU g⁻¹, SAFE, U8978P-0075), following the described protocol¹³. Three-week-old littermates were weaned and maintained on the vitamin A-deficient or control diets for 14–20 weeks before being killed for further analysis.

Mouse ecto- and endocervical medium

Endocervical cells were cultured in ADF medium (Invitrogen, 12634) supplemented with 12 mM HEPES, 1% GlutaMax, 1% B27, 1% N2, 50 ng ml⁻¹ mouse EGF (Invitrogen, 15630-056, 35050-038, 17504-044, 17502048, PMG8043), 100 ng ml⁻¹ mouse noggin, 100 ng ml⁻¹ human FGF10 (Peprotech, 250-38-100, 100-26-25), 1.25 mM N-acetyl-l-cysteine, 10 mM nicotinamide, 10 µM ROCK inhibitor (Y-27632) (Sigma, A9165-5G, N0636, Y0503), 2 µM TGF-β receptor kinase inhibitor IV (Calbiochem, 616454), 1% penicillin–streptomycin (Gibco, 15140-122) with 25% WNT3A- and 25% R-spondin-1-conditioned medium. Ectocervical cells were cultured in endocervical medium but without 25% WNT3A- and 25% R-spondin 1-conditioned medium.

Wnt-deficient human ectocervical medium

Ectocervical cells were cultured in ADF medium supplemented with 12 mM HEPES, 1% GlutaMax, 1% B27, 1% N2, 0.5 µg ml⁻¹ hydrocortisone (Sigma, H0888-1G), 10 ng ml⁻¹ human EGF (Invitrogen, PHG0311), 100 ng ml⁻¹ human noggin, 100 ng ml⁻¹ human FGF10 (Peprotech, 120-10 C, 100-26-25), 1.25 mM N-acetyl-l-cysteine, 10 mM nicotinamide, 2 µM TGF-β receptor kinase Inhibitor IV, 10 µM Y-27632, 10 µM forskolin (Sigma, F6886) and 1% penicillin–streptomycin.

Wnt-proficient human endocervical medium

Endocervical cells were cultured in ADF medium supplemented with 12 mM HEPES, 1% GlutaMax, 1% B27, 1% N2, 10 ng ml⁻¹ human EGF, 100 ng ml⁻¹ human noggin, 100 ng ml⁻¹ human FGF10, 1.25 mM N-acetyl-l-cystein, 10 mM nicotinamide, 2 mM TGF-β receptor kinase inhibitor IV and 10 µM Y-27632 with 25% WNT3A- and 25% R-spondin-1-conditioned medium.

Epithelial stem cell isolation from human and mouse cervix

Human ecto- and endocervix samples were provided by the Department of Gynecology, Charité University Hospital, Berlin, Germany. Usage for scientific research was approved by their ethics committee (EA1/059/15); informed consent was obtained from all subjects. The study is compliant with all relevant ethical regulations regarding research involving human participants. Tissue biopsies from anonymous donors (Supplementary Table 16) were processed within 2–3 h after removal. Biopsies were sourced from standard surgical procedures. Mouse cervix was removed from euthanized 10- to 20-week-old female wild-type BALB/c mice (Charles River). Tissue samples were washed in sterile PBS (Gibco, 14190-094) and minced with surgical scissors. The minced human cervix tissue was incubated in 0.5 mg ml⁻¹ collagenase II (Calbiochem, 234155) (2.5 h, 37 °C). Tissue and dissociated cells were pelleted (5 min, 1,000g, 4 °C), cells were resuspended in TrypLE Express (Gibco, 12604021) and incubated in a shaker (15 mins, 37°). Minced mouse cervix tissue was incubated in TrypLE express in a shaker (2.5 h, 37 °C). The pellet was resuspended in ADF medium and passed through a 40 µM cell strainer (BD Falcon, 352340). Cells were pelleted (5 min, 1,000g, 40 °C), resuspended in either human ecto- or endocervical medium or mouse cervical medium, and cultured either directly as organoids or as 2D cultures.

Human epithelial 2D cell culture

Isolated human epithelial stem cells were resuspended in either ecto- or endocervical medium, plated and incubated in collagen-coated flasks. Upon 70–80% confluence, cells were detached using TrypLE Express and centrifuged (5 min, 1,000g, 40 °C). Then, cells were cultured as organoids or as stem cell-enriched 2D cultures. The 2D cultures were maintained by seeding 2D cells from P1 into flasks containing lethally irradiated J2-3T3 fibroblast feeder cells in ecto- or endocervical media. The medium was replaced, and irradiated fibroblasts added every 4 d until 60–70% confluence, at this stage, cells were reseeded onto freshly irradiated feeders or cryopreserved.

Organoid culture and maintenance

Tissue-isolated or 2D culture cells were mixed with 50 µl ice-cold Matrigel (BD, 356231); Matrigel droplets were placed in prewarmed 24-well plates to allow polymerization (10 min, 37 °C). Freshly isolated endocervical cells were seeded at a higher density, as they had approximately 1% organoid-forming efficiency compared with 10% observed with ectocervical epithelial cells. The Matrigel droplet was then overlaid with 500 µl of prewarmed human or mouse ecto- or endocervical medium. Cultures were incubated for 2–3 weeks (37 °C, 5% CO₂), and the medium was replaced every 4 d. For organoid passaging, organoids in Matrigel drops were first dissolved in 1 ml ice-cold ADF, thoroughly pipetted, transferred to 15 ml tubes, to which an additional 4 ml ice-cold ADF medium was added until the Matrigel was dissolved thoroughly, followed by centrifugation (5 min, 300g, 4 °C). Pelleted organoids were incubated in 1 ml TrypLE Express (30 min, 37 °C), followed by mechanical fragmentation with vigorous pipetting using fire-polished glass Pasteur pipettes to generate single cells. For organoid expansion, cells were reseeded into a fresh Matrigel (ectocervical organoids at 1:10 ratio and endocervical organoids at 1:5 ratio). Matrigel was allowed to polymerize as described above.

Organoid-forming efficiency

Stem cells were counted, and a defined number was resuspended in 50 µl of Matrigel to generate organoids as described above. Two to three weeks later, whole-well images were taken and the number and area of formed organoids were determined using ImageJ to calculate the organoid-forming efficiency.

Immunofluorescent histochemistry

Organoids were washed with cold PBS 5 times to remove Matrigel, fixed with 4% paraformaldehyde (1 h, room temperature) and washed twice in PBS. Organoids were then subjected to dehydration in an ascending ethanol series and placed in isopropanol and acetone (20 min each). Mouse and human tissues were washed in PBS and fixed with 4% paraformaldehyde (overnight, room temperature). Samples were dehydrated in an ascending ethanol series and placed in isopropanol and xylene (60 min each). Following paraffinization using a Leica TP1020 tissue processor, 5 µm sections were cut on a Microm HM 315 microtome. For immunostaining, paraffin sections were deparaffinized and rehydrated, treated with antigen-retrieval solution (Dako, S1699), and blocked using blocking buffer (1% BSA, 2% FCS in PBS) (1 h, room temperature). Sections were incubated with primary antibodies diluted in blocking buffer (90 min, room temperature), PBS-washed 5 times, incubated with secondary antibodies diluted in blocking buffer along with Hoechst or Draq5 (1 h, room temperature), washed with PBS 5 times, and finally mounted using Mowiol.

Fresh epithelial isolates were cultured in 2D on collagen-coated coverslips and fixed with 4% PFA (30 min, room temperature). Cells were permeabilized and blocked with 0.5% Triton X-100 and 1% BSA in PBS, incubated with primary antibodies diluted in 1% BSA in PBS (1 h, room temperature), washed with PSBT (PBS, 0.1% Tween-20) 3 times, incubated with secondary antibodies diluted in 1% BSA in PBS along with Hoechst or Draq5 (1 h, room temperature), washed with PBST three times, washed with PBS once, and finally mounted using Mowiol.

Images were acquired on a Leica TCS SP8 confocal microscope and processed with Adobe Photoshop; 3D reconstruction was done with Volocity 6.3 software (Perkin Elmer).

Whole-mount staining

Matrigel was removed by extensive washing with ice-cold PBS (4 × 45 min). Organoids were allowed to settle by gravity to maintain their 3D structure, fixed using pre-warmed 3.7% PFA in PBS (1 h, room temperature), and washed with PBST 3 times. Permeabilization and blocking was performed using 5% donkey serum, 1% FCS, 0.05% Tween-20, 2% Triton X-100, 0.02% sodium azide in PBS (overnight, 4 °C). Organoids were incubated with primary antibodies diluted in blocking buffer (5% donkey serum, 1% FCS, 0.25% Triton X-100, 0.02% sodium azide in PBS) (3–5 d, 4 °C), washed with PBST (3 × 45 min, room temperature), incubated with secondary antibodies diluted in blocking buffer (2 d, 4 °C), washed with PBST (45 min) and washed with PBS containing 5% glycerol (3 × 45 min, room temperature). Organoids were then carefully transferred to ibidi µ-slides (81822) together with some PBS/glycerol solution. Z-stack images were acquired and processed as described.

smRNA-ISH

Hybridizations of paraffin-embedded 10 µm tissue sections with RNAscope 2.5HD Red Reagent kit (Advanced Cell Diagnostics) were performed according to the manufacturer’s protocol, along with positive (PPIB) and negative (DapB) control probes. Tiled bright-field images were obtained with an AxioScan.Z1 tissue imager (Zeiss), processed with Zen 2.3 (Blue edition) and compiled with Adobe illustrator.

RNA isolation and quality control for microarray analysis

Microarrays were hybridized for cultured cells and organoids. From Wnt-deficient medium: 2D human ectocervical cells in (three biological replicates from two donors and one technical replicate from one of the donors), human EO-ecto (three donors), human DO-ecto (four donors). From Wnt-proficient medium: 2D human endocervical cells (three donors) and human DO-ecto (three donors). Mouse EO-ecto and DO-ecto regardless of medium (two mice per condition). Effect sizes could not be pre-determined; therefore, sample sizes were selected on the basis of availability. Cells and organoids were pelleted and resuspended in 1 ml Trizol (Life Technologies), and RNA was isolated according to the manufacturer’s protocol. RNA was quantified with a NanoDrop 1000 UV-Vis spectrophotometer (Kisker), and quality was assessed by 2100 Bioanalyzer with RNA Nano 6000 microfluidics kit (Agilent Technologies).

Microarray expression profiling

Single-colour hybridizations were performed on custom whole-genome human 8 × 60k arrays (Design ID048908) and feature extraction was applied to obtain probe intensities (Agilent). The extracted raw data were background-corrected, quantile-normalized and analysed for differential gene expression using R and the BioConductor package LIMMA (Supplementary Tables 6–8). R was used for all statistical analyses unless stated otherwise: unpaired tests were used for microarray gene expression comparisons and Mann–Whitney U test was used for comparisons with SCJ marker genes, P < 0.05. Microarray data have been deposited at the Gene Expression Omnibus (GEO) under accession number GSE87076. The signature of differentially expressed genes in cultured cells and organoids was based on all genes with false discovery rate <0.05 and log₂(fold change) <−1.5 or >1.5 for each comparison (2D versus DO-ecto or EO-ecto versus DO-ecto). For each gene, the largest absolute fold change from both comparisons and possible replicate probes was taken.

Analysis of stem cell-related genes

Raw microarray datasets of adult stem cells cultured on feeder cells and corresponding differentiated cells from an air–liquid interface, Matrigel or self-assembly sphere were downloaded from the GEO (GSE57584, GSE66115, GSE69453, GSE65013, GSE32606, GSE69429, GSE49292) and normalized using the RMA-sketch method (Affymetrix Power Tools). We assessed genes that were differentially expressed between stem cells and corresponding differentiated cell cultures: normal oesophagus, Barrett’s oesophagus, gastric cardia, duodenum, jejunum, ileum, colon ascendens, colon transversum, colon descendens, KRT5⁺ and KRT7⁺ fetal oesophagal cells, fallopian tube, nasal turbinated epithelium, tracheobronchial epithelium and distal airway epithelium. We selected significantly up- or downregulated stem cell-related genes (abs(log fold change) >1, adjusted P <0.05) in at least 5 out of 18 comparisons (Supplementary Tables 5, 9 and 10).

Gene set enrichment analysis

We performed a pre-ranked analysis using GSEA software v2.1.0 and applied t-statistics on comparisons of ectocervical organoids (2D versus DO-ecto or EO-ecto versus DO-ecto) to rank probes and enrichment of MSigDB Motif gene sets (c3.all.v5.1.symbols.gmt) under standard settings; the Max_probe method (1,000 permutations) was applied for collapsing probe sets within genes. Motif gene sets (significant in at least 1 gene, FDR <5%) were further analysed. For the heat map visualization, we selected motif gene sets referring to the same transcription factors with the most significant negative log₁₀(P-value).

Cervical cancer data

Gene expression data, including public clinical and molecular annotations, were obtained for 178 unique samples published by The TCGA consortium. Gene expression levels were extracted from ‘*.rsem.genes.normalized_results’ files using custom scripts and normalized by applying DESeq2-generated size factors. To classify samples into squamous-like and columnar-like classes, gene expression levels were log₂-transformed, and Z-scores were applied for comparisons. A squamous versus columnar organoid signature was defined on the basis of fold changes between ectocervical squamous and endocervical columnar organoids (2,834 genes, FDR < 0.05, absolute log₂(fold change) > 1), selecting the probe with the lowest P-value for each gene. Spearman correlation coefficients (referred to as Co–Sq scores) were computed between Z-scored gene expression values from each cancer sample and the corresponding fold change for the same gene from the squamous versus columnar organoid signature. We defined samples with Co–Sq score >0.2 as squamous-like, those with Co–Sq score <−0.2 as columnar-like, and all others as ‘undetermined’ (Extended Data Fig. 8b). Applying the same procedure to 1,000 random gene sets of the same size and fold changes produced correlation coefficients generally lower than |0.06|. Thresholds for sample classification into KRT5^high versus KRT5^low, KRT7^high versus KRT7^low and TP63^high versus TP63^low classes were selected manually to separate the highest cluster from all other samples (Extended Data Fig. 8a). For simplicity, we combined all diagnoses with an adenoma component (ADC and adenosquamous carcinoma) into cervical ADC (Supplementary Table 14).

Statistics and reproducibility

GraphPad Prism (v.8) was used for statistical calculations and generation of plots. The data are displayed as mean ± s.e.m. P < 0.05 was considered to be statistically significant. Each experiment was repeated independently with similar results.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

Microarray and scRNA-seq data that support the findings of this study have been deposited in the GEO under accession codes GSE87076 and GSE128987, respectively. Previously published microarray data that were reanalysed here are available under accession codes GSE57584, GSE66115, GSE69453, GSE65013, GSE32606, GSE69429 and GSE49292. The human cervical cancer data were derived from the TCGA research network (http://cancergenome.nih.gov/). The dataset derived from this resource that supports the findings of this study is available at https://gdc.cancer.gov/about-data/publications/cesc_2017. The quantitative data of this study are available within the paper and its supplementary information files. All other data supporting the findings of this study are available from the corresponding author on reasonable request. Source data are provided with this paper.

Code Availability

All computational codes used for analyses used can be accessed at https://github.com/MPIIB-Department-TFMeyer/Chumduri_Gurumurthy_et_al._Cervical_squamocolumnar_homeostasis.

References

Ajayi, T. A., Cantrell, S., Spann, A. & Garman, K. S. Barrett’s esophagus and esophageal cancer: links to microbes and the microbiome. PLoS Pathog. 14, e1007384 (2018).
Article PubMed PubMed Central CAS Google Scholar
McNairn, A. J. & Guasch, G. Epithelial transition zones: merging microenvironments, niches, and cellular transformation. Eur. J. Dermatol. 21, 21–28 (2011).
Article CAS PubMed Google Scholar
Giroux, V. & Rustgi, A. K. Metaplasia: tissue injury adaptation and a precursor to the dysplasia-cancer sequence. Nat. Rev. Cancer 17, 594–604 (2017).
Article CAS PubMed PubMed Central Google Scholar
World Cancer Research Fund. Worldwide Cancer data. https://www.wcrf.org/dietandcancer/cancer-trends/worldwide-cancer-data
Burghardt, E. & Ostor, A. G. Site and origin of squamous cervical cancer: a histomorphologic study. Obstet. Gynecol. 62, 117–127 (1983).
CAS PubMed Google Scholar
Deng, H., Hillpot, E., Mondal, S., Khurana, K. K. & Woodworth, C. D. HPV16-immortalized cells from human transformation zone and endocervix are more dysplastic than ectocervical cells in organotypic culture. Sci. Rep. 8, 15402 (2018).
Article PubMed PubMed Central CAS Google Scholar
Mitra, A. et al. The vaginal microbiota, human papillomavirus infection and cervical intraepithelial neoplasia: what do we know and where are we going next? Microbiome 4, 58 (2016).
Article PubMed PubMed Central Google Scholar
Singer, A. The uterine cervix from adolescence to the menopause. Br. J. Obstet. Gynaecol. 82, 81–99 (1975).
Article CAS PubMed Google Scholar
Singer, A. The uterine cervix from adolescence to the menopause. Gynakol. Rundsch. 15, 257–278 (1975).
Article CAS PubMed Google Scholar
Hwang, L. Y. et al. Active squamous metaplasia of the cervical epithelium is associated with subsequent acquisition of human papillomavirus 16 infection among healthy young women. J. Infect. Dis. 206, 504–511 (2012).
Article PubMed PubMed Central Google Scholar
Moscicki, A. B., Burt, V. G., Kanowitz, S., Darragh, T. & Shiboski, S. The significance of squamous metaplasia in the development of low grade squamous intraepithelial lesions in young women. Cancer 85, 1139–1144 (1999).
Article CAS PubMed Google Scholar
zur Hausen, H. Papillomaviruses in the causation of human cancers—a brief historical account. Virology 384, 260–265 (2009).
Article CAS PubMed Google Scholar
Darwiche, N., Celli, G., Sly, L., Lancillotti, F. & De Luca, L. M. Retinoid status controls the appearance of reserve cells and keratin expression in mouse cervical epithelium. Cancer Res. 53, 2287–2299 (1993).
CAS PubMed Google Scholar
Sommer, A. Vitamin A Deficiency and its Consequences. A Field Guide to Detection and Control 3rd edn (World Health Organization, 1995); https://www.who.int/nutrition/publications/vad_consequences.pdf
Zhang, X., Dai, B., Zhang, B. & Wang, Z. Vitamin A and risk of cervical cancer: a meta-analysis. Gynecol. Oncol. 124, 366–373 (2012).
Article CAS PubMed Google Scholar
Reich, O. & Regauer, S. Two major pathways of recurrent high-grade squamous intraepithelial lesions of the cervix. Int. J. Cancer 137, 2520–2521 (2015).
Article CAS PubMed Google Scholar
Smedts, F. et al. Keratin expression in cervical cancer. Am. J. Pathol. 141, 497–511 (1992).
CAS PubMed PubMed Central Google Scholar
Vooijs, G. P. The problem of replacement and differentiation of the intestinal epithelium: its relation to squamous metaplasia in the uterine cervix. Cancer 81, 317–322 (1997).
Article CAS PubMed Google Scholar
Herfs, M. et al. A discrete population of squamocolumnar junction cells implicated in the pathogenesis of cervical cancer. Proc. Natl Acad. Sci. USA 109, 10516–10521 (2012).
Article CAS PubMed PubMed Central Google Scholar
Clevers, H., Loh, K. M. & Nusse, R. Stem cell signaling. An integral program for tissue renewal and regeneration: Wnt signaling and stem cell control. Science 346, 1248012 (2014).
Article PubMed CAS Google Scholar
Hsu, Y. C., Li, L. & Fuchs, E. Transit-amplifying cells orchestrate stem cell activity and tissue regeneration. Cell 157, 935–949 (2014).
Article CAS PubMed PubMed Central Google Scholar
Huch, M. et al. Long-term culture of genome-stable bipotent stem cells from adult human liver. Cell 160, 299–312 (2015).
Article CAS PubMed PubMed Central Google Scholar
Rheinwald, J. G. & Green, H. Epidermal growth factor and the multiplication of cultured human epidermal keratinocytes. Nature 265, 421–424 (1977).
Article CAS PubMed Google Scholar
Sato, T. et al. Long-term expansion of epithelial organoids from human colon, adenoma, adenocarcinoma, and Barrett’s epithelium. Gastroenterology 141, 1762–1772 (2011).
Article CAS PubMed Google Scholar
Watanabe, K. et al. A ROCK inhibitor permits survival of dissociated human embryonic stem cells. Nat. Biotechnol. 25, 681–686 (2007).
Article CAS PubMed Google Scholar
Iguchi, H., Mitsui, T., Ishida, M., Kanba, S. & Arita, J. cAMP response element-binding protein (CREB) is required for epidermal growth factor (EGF)-induced cell proliferation and serum response element activation in neural stem cells isolated from the forebrain subventricular zone of adult mice. Endocr. J. 58, 747–759 (2011).
Article CAS PubMed Google Scholar
Mao, B. & Niehrs, C. Kremen2 modulates Dickkopf2 activity during Wnt/LRP6 signaling. Gene 302, 179–183 (2003).
Article CAS PubMed Google Scholar
Mao, B. et al. LDL-receptor-related protein 6 is a receptor for Dickkopf proteins. Nature 411, 321–325 (2001).
Article CAS PubMed Google Scholar
Nakamura, R. E. & Hackam, A. S. Analysis of Dickkopf3 interactions with Wnt signaling receptors. Growth Factors 28, 232–242 (2010).
Article PubMed PubMed Central CAS Google Scholar
Niehrs, C. Function and biological roles of the Dickkopf family of Wnt modulators. Oncogene 25, 7469–7481 (2006).
Article CAS PubMed Google Scholar
Chen, B. et al. Small molecule-mediated disruption of Wnt-dependent signaling in tissue regeneration and cancer. Nat. Chem. Biol. 5, 100–107 (2009).
Article CAS PubMed PubMed Central Google Scholar
Willert, K. et al. Wnt proteins are lipid-modified and can act as stem cell growth factors. Nature 423, 448–452 (2003).
Article CAS PubMed Google Scholar
Wang, X. et al. Cloning and variation of ground state intestinal stem cells. Nature 522, 173–178 (2015).
Article CAS PubMed PubMed Central Google Scholar
Avraham, R. & Yarden, Y. Feedback regulation of EGFR signalling: decision making by early and delayed loops. Nat. Rev. Mol. Cell Biol. 12, 104–117 (2011).
Article CAS PubMed Google Scholar
Cargnello, M. & Roux, P. P. Activation and function of the MAPKs and their substrates, the MAPK-activated protein kinases. Microbiol. Mol. Biol. Rev. 75, 50–83 (2011).
Article CAS PubMed PubMed Central Google Scholar
Li, M. et al. Guanylate binding protein 1 is a novel effector of EGFR-driven invasion in glioblastoma. J. Exp. Med. 208, 2657–2673 (2011).
Article CAS PubMed PubMed Central Google Scholar
Yang, S. H., Sharrocks, A. D. & Whitmarsh, A. J. MAP kinase signalling cascades and transcriptional regulation. Gene 513, 1–13 (2013).
Article CAS PubMed Google Scholar
LaVoie, M. J. & Selkoe, D. J. The Notch ligands, Jagged and Delta, are sequentially processed by α-secretase and presenilin/γ-secretase and release signaling fragments. J. Biol. Chem. 278, 34427–34437 (2003).
Article CAS PubMed Google Scholar
Ranganathan, P., Weaver, K. L. & Capobianco, A. J. Notch signalling in solid tumours: a little bit of everything but not all the time. Nat. Rev. Cancer 11, 338–351 (2011).
Article CAS PubMed Google Scholar
Tsuda, L., Nagaraj, R., Zipursky, S. L. & Banerjee, U. An EGFR/Ebi/Sno pathway promotes delta expression by inactivating Su(H)/SMRTER repression during inductive Notch signaling. Cell 110, 625–637 (2002).
Article CAS PubMed Google Scholar
Yap, Y. S. et al. The NF1 gene revisited—from bench to bedside. Oncotarget 5, 5873–5892 (2014).
Article PubMed PubMed Central Google Scholar
White, A. C. & Lowry, W. E. Refining the role for adult stem cells as cancer cells of origin. Trends Cell Biol. 25, 11–20 (2015).
Article CAS PubMed Google Scholar
Alizadeh, A. A. et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 403, 503–511 (2000).
Article CAS PubMed Google Scholar
Hoadley, K. A. et al. Cell-of-origin patterns dominate the molecular classification of 10,000 tumors from 33 types of cancer. Cell 173, 291–304 (2018).
Article CAS PubMed PubMed Central Google Scholar
O’Shannessy, D. J. et al. Gene expression analyses support fallopian tube epithelium as the cell of origin of epithelial ovarian cancer. Int. J. Mol. Sci. 14, 13687–13703 (2013).
Article PubMed PubMed Central CAS Google Scholar
van Laar, R. K. et al. Implementation of a novel microarray-based diagnostic test for cancer of unknown primary. Int. J. Cancer 125, 1390–1397 (2009).
Article PubMed CAS Google Scholar
The Cancer Genome Atlas Research Network Integrated genomic and molecular characterization of cervical cancer. Nature 543, 378–384 (2017).
Article PubMed Central CAS Google Scholar
Herfs, M. et al. A novel blueprint for ‘top down’ differentiation defines the cervical squamocolumnar junction during development, reproductive life, and neoplasia. J. Pathol. 229, 460–468 (2013).
Article PubMed Google Scholar
Rojas-Rios, P. & Gonzalez-Reyes, A. Concise review: The plasticity of stem cell niches: a general property behind tissue homeostasis and repair. Stem Cells 32, 852–859 (2014).
Article CAS PubMed Google Scholar
Naik, S., Larsen, S. B., Cowley, C. J. & Fuchs, E. Two to tango: dialog between immunity and stem cells in health and disease. Cell 175, 908–920 (2018).
Article CAS PubMed PubMed Central Google Scholar
Rajendra, S. et al. Transcriptionally active human papillomavirus is strongly associated with Barrett’s dysplasia and esophageal adenocarcinoma. Am. J. Gastroenterol. 108, 1082–1093 (2013).
Article CAS PubMed Google Scholar
Wang, X. et al. Residual embryonic cells as precursors of a Barrett’s-like metaplasia. Cell 145, 1023–1035 (2011).
Article CAS PubMed PubMed Central Google Scholar
Yang, E. J. et al. Microanatomy of the cervical and anorectal squamocolumnar junctions: a proposed model for anatomical differences in HPV-related cancer risk. Mod. Pathol. 28, 994–1000 (2015).
Article CAS PubMed PubMed Central Google Scholar
Syrjänen, S., Rautava, J. & Syrjänen, K. HPV in head and neck cancer—30 years of history. Recent Results Cancer Res. 206, 3–25 (2017).
Article PubMed Google Scholar
Doorbar, J. & Griffin, H. Refining our understanding of cervical neoplasia and its cellular origins. Papillomavirus Res. 7, 176–179 (2019).
Article PubMed PubMed Central Google Scholar
Briscoe, J. & Small, S. Morphogen rules: design principles of gradient-mediated embryo patterning. Development 142, 3996–4009 (2015).
Article CAS PubMed PubMed Central Google Scholar
Yamaguchi, T. P. Heads or tails: Wnts and anterior–posterior patterning. Curr. Biol. 11, R713–R724 (2001).
Article CAS PubMed Google Scholar
Brake, T. & Lambert, P. F. Estrogen contributes to the onset, persistence, and malignant progression of cervical cancer in a human papillomavirus-transgenic mouse model. Proc. Natl Acad. Sci. USA 102, 2490–2495 (2005).
Article CAS PubMed PubMed Central Google Scholar
Franceschi, S. Past and future of prophylactic ablation of the cervical squamocolumnar junction. Ecancermedicalscience 9, 527 (2015).
PubMed PubMed Central Google Scholar
Garsia, S. et al. [Initial results of a 15-year follow-up study of a group of patients with risk of carcinoma of the uterine cervix treated in a prevention program]. Ann. Ostet. Ginecol. Med. Perinat. 103, 372–376 (1982).
CAS PubMed Google Scholar
Remotti, G. et al. Follow-up results of a prevention program for cervical cancer. J. Reprod. Med. 31, 4–10 (1986).
CAS PubMed Google Scholar
Yang, H., Adam, R. C., Ge, Y., Hua, Z. L. & Fuchs, E. Epithelial–mesenchymal micro-niches govern stem cell lineage choices. Cell 169, 483–496 (2017).
Article CAS PubMed PubMed Central Google Scholar
Rock, J. R. et al. Basal cells as stem cells of the mouse trachea and human airway epithelium. Proc. Natl Acad. Sci. USA 106, 12771–12775 (2009).
Article CAS PubMed PubMed Central Google Scholar
Van Keymeulen, A. et al. Distinct stem cells contribute to mammary gland development and maintenance. Nature 479, 189–193 (2011).
Article PubMed CAS Google Scholar
Madisen, L. et al. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nat. Neurosci. 13, 133–140 (2010).
Article CAS PubMed Google Scholar
Wolock, S. L., Lopez, R. & Klein, A. M. Scrublet: computational identification of cell doublets in single-cell transcriptomic data. Cell Syst. 8, 281–291 (2019).
Article CAS PubMed PubMed Central Google Scholar
McInnes, L., Healy, J. & Melville, J. UMAP: uniform manifold approximation and projection for dimension reduction. Preprint at https://arxiv.org/abs/1802.03426 (2018).
Waltman, L. & van Eck, N. J. A smart local moving algorithm for large-scale modularity-based community detection. Preprint at https://arxiv.org/abs/1308.6604v1 (2013).
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, 1860–1872 (2015).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We thank C. Dimmler for technical assistance; I. Wagner for the microarrays; S. Beyer, C. Beckmann and A. Iftekhar for initial technical support on mouse ectocervical organoids; G. Rausch and M. Meyer for assistance with the mouse application; E. Vafadarnejad for advice on scRNA-seq analysis; D. Schad for assistance with preparing graphics; and K. Lapid and R. Zietlow for editing the manuscript. O.D. is supported by the Deutsche Forschungsgemeinschaft Graduiertenkolleg 2157 (DFG-GRK 2157). This work was supported by the Federal Ministry of Education and Research (BMBF) through the Infect-ERA project CINOCA (FK031A409A) to T.F.M. The funders had no influence on the study design or analysis of the data.

Funding

Open access funding provided by Max Planck Society

Author information

Mandy Mangler
Present address: Klinik für Gynäkologie und Geburtsmedizin, Vivantes Auguste-Viktoria-Klinikum, Berlin, Germany
These authors contributed equally: Cindrilla Chumduri, Rajendra Kumar Gurumurthy.

Authors and Affiliations

Department of Molecular Biology, Max Planck Institute for Infection Biology, Berlin, Germany
Cindrilla Chumduri, Rajendra Kumar Gurumurthy, Hilmar Berger, Naveen Kumar, Stefanie Koster, Volker Brinkmann, Kirstin Hoffmann, Marina Drabkina, Uwe Klemm, Hans-Joachim Mollenkopf & Thomas F. Meyer
Chair of Microbiology, University of Würzburg, Würzburg, Germany
Cindrilla Chumduri, Naveen Kumar & Dajung Son
Institute for RNA-based Infection Research (HIRI), Helmholtz Center for Infection Research (HZI), Würzburg, Germany
Oliver Dietrich, Jörg Vogel & Antoine-Emmanuel Saliba
Core Unit Systems Medicine, University of Würzburg, Würzburg, Germany
Panagiota Arampatzi
Institute of Pathology, Vivantes Klinikum Berlin, Berlin, Germany
Hermann Herbst
Department of Gynecology, Charité University Medicine, Berlin, Germany
Mandy Mangler
Institute of Molecular Infection Biology, University of Würzburg, Würzburg, Germany
Jörg Vogel
Laboratory of Infection Oncology, Institute of Clinical Molecular Biology (IKMB), Christian Albrechts University of Kiel, Kiel, Germany
Thomas F. Meyer

Authors

Cindrilla Chumduri
View author publications
You can also search for this author in PubMed Google Scholar
Rajendra Kumar Gurumurthy
View author publications
You can also search for this author in PubMed Google Scholar
Hilmar Berger
View author publications
You can also search for this author in PubMed Google Scholar
Oliver Dietrich
View author publications
You can also search for this author in PubMed Google Scholar
Naveen Kumar
View author publications
You can also search for this author in PubMed Google Scholar
Stefanie Koster
View author publications
You can also search for this author in PubMed Google Scholar
Volker Brinkmann
View author publications
You can also search for this author in PubMed Google Scholar
Kirstin Hoffmann
View author publications
You can also search for this author in PubMed Google Scholar
Marina Drabkina
View author publications
You can also search for this author in PubMed Google Scholar
Panagiota Arampatzi
View author publications
You can also search for this author in PubMed Google Scholar
Dajung Son
View author publications
You can also search for this author in PubMed Google Scholar
Uwe Klemm
View author publications
You can also search for this author in PubMed Google Scholar
Hans-Joachim Mollenkopf
View author publications
You can also search for this author in PubMed Google Scholar
Hermann Herbst
View author publications
You can also search for this author in PubMed Google Scholar
Mandy Mangler
View author publications
You can also search for this author in PubMed Google Scholar
Jörg Vogel
View author publications
You can also search for this author in PubMed Google Scholar
Antoine-Emmanuel Saliba
View author publications
You can also search for this author in PubMed Google Scholar
Thomas F. Meyer
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

C.C., R.K.G. and T.F.M. conceived the study. C.C. and R.K.G. designed the experiments and analysed the data. C.C., R.K.G., N.K., S.K., M.D., K.H. and D.S. performed experimental work. H.B. performed microarray analysis and in silico comparison with cervical cancer and stem cell datasets with contributions from C.C. and R.K.G. C.C., R.K.G., A.-E.S., J.V., N.K. and P.A. designed and performed scRNA-seq. O.D. performed scRNA-seq bioinformatics analysis with the help of A.-E.S. and C.C. V.B., U.K. and H.-J.M. contributed imaging, mouse breeding and microarray studies, respectively. M.M. and H.H. provided human samples. C.C. and R.K.G. wrote the manuscript with contributions from T.F.M.

Corresponding authors

Correspondence to Cindrilla Chumduri or Thomas F. Meyer.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Peer review information Nature Cell Biology thanks the anonymous reviewers for their contribution to the peer review of this work.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 scRNA-seq profiling of the cervix.

a, UMAP of single cell transcriptomes obtained from healthy mouse endocervix, ectocervix and TZ as in Fig. 1b. Here, cells are labelled by the tissue of origin. b, Normalized expression of selected markers (Sfn, Epcam, Acta2, Col6a2, Cd52, Pcam1, Rgs5 and S100b) colour-coded on the UMAP, representing cellular sub-clusters. c, UMAP of epithelial sub-clusters colour-coded by the tissue of origin. d-h, Normalized expression values of selected markers colour coded on the UMAP representing key epithelial sub-clusters as in Fig. 1d. Color maps of squamous type (Sq_1): Birc5, Mki67, Cks2, Hmgb2; Sq_2A and Sq_2B: Trp63, Krt5, Dkk3, Notch1; Sq_3: Fam25c, Gm94, Krt6, Krt10; Columnar type (Co_1): Anpep, Cxcl17, Krt8, Krt18; Co_2: Krt19, Ltf, Muc1, Psca; n = 3 biologically independent mice experiments.

Extended Data Fig. 2 Cervix consists of KRT5⁺ stratified and KRT8⁺ columnar epithelium.

(a-b) Human (a) and mouse (b) tiled images of cervix tissue sections, including stratified and columnar epithelium, immunolabeled against KRT5 and KRT8; nuclei in blue. c-d, Tiled brightfield images of sections showing the entire mouse female genital system labelled with smRNA-ISH for Krt5 (c) and Krt8 (d); nuclei in blue. Boxed areas are magnified on the right. Images are representative of n = 3 biologically independent mice or human samples.

Extended Data Fig. 3 Culture conditions for human and mouse ectocervical organoids derived from single epithelial stem cells.

a-c, Bright-field images showing (a) two-week-old human ectocervical organoids grown in the absence of indicated growth factors from the complete media; (b) Time course of organoids grown from single ectocervical stem cells; (c) Maintenance of stemness from the passage (P)1 to P17 in mouse ectocervical organoids. Data representative of n = 3 biologically independent mice or human samples.

Extended Data Fig. 4 Wnt microenvironment controls the growth of endocervical organoids.

a, Confocal images show a similar distribution of Ki67 in human endocervical tissue and organoids. b, Brightfield images of human endocervical organoids at P1 and P7. (a-b) Data representative of n = biologically independent experiments from 3 donors. c, Brightfield images of mouse ecto- and endocervical organoids. Cells isolated from ecto- and endocervical tissue were grown in Matrigel with WNT-proficient or -deficient medium. Data representative of n = biologically independent experiments from 3 mice. d, Analysis of differential expression of cytokeratins in human ecto- vs endocervical organoids, revealing a distinct expression profile. e, Confocal images of endocervix tissue-derived organoids immunolabeled for KRT5 and p63; nuclei in blue. Data representative of n = biologically independent experiments from 3 mice. f, Confocal images of organoids-derived from human ectocervix or endocervix that were grown in the presence or absence of WNT3A and RSPO1 and immunolabeled for KRT5 and KRT8; nuclei in blue. Data representative of n = biologically independent experiments from 3 donors. g, Effect of growth factors on endocervical organoid size. Data represented as mean ± s.e.m. n = shown are the number of organoids quantified from a representative of 3 independent biological replicates. Statistical significance was determined using a two-tailed Student’s t-test, and p-values are shown on the graph. Statistical source data are provided in Source Data Extended Data Fig. 4.

Source data

Extended Data Fig. 5 Microenvironmental signaling molecules in the cervix.

Tiled brightfield images of entire mouse genital system sections labeled with smRNA-ISH for (a) Axin2, (b) Rspo1, (c) Rspo2, (d) Rspo3, (e) Rspo4, (f) Dkk1, (g) Dkk2, (h) Dkk3, and (i) Dkk4; nuclei in blue. Boxed areas are magnified on the right. Data representative of n = biologically independent experiments from 3 mice.

Extended Data Fig. 6 Ectocervical stemness and differentiation.

a, Brightfield images of human organoids at P1 or P8 from 2D-ectocervical cells and long-term passage (P5) of organoids in 3D. Percentages indicate organoid-forming efficiency. b, Percentage of Ki67⁺ proliferating cells in EO-ecto and DO-ecto organoid-derived cells. c, Heatmap of genes concordantly up- or downregulated in ectocervical stem cells vs differentiated cells and similar comparisons to the ground state stem cell dataset (see Methods for details). Expression levels in stem cells and their corresponding differentiated cells from 13 different tissue types. DA-Distal Airway; NT-Nasotubular epithelium; TB-Tracheobronchial epithelium; FT-Fallopian tube; CA-Colon ascendens, CD-Colon descendens, CT-Colon transversum, DD-Duodenum, ES-Esophagus, IL-Ileum, JJ-Jejunum, K5-Keratin 5⁺ oesophagal cells, K7-Keratin 7⁺ oesophagal cells. d, Phase-contrast images of human ectocervical organoids in the presence or absence of DBZ. Data representative of n = biologically independent experiments from 3 donors.

Extended Data Fig. 7 Regulation of squamous metaplasia by the tissue microenvironment. Regulation of squamous metaplasia by the tissue microenvironment.

a-b, WNT-deficient medium enriches for p63⁺/KRT5⁺ human endocervical cells, grown in 2D, that can only give rise to stratified organoids. WNT-proficient medium results in predominantly KRT7⁺ and few p63⁺/KRT5⁺ cells, which can give rise to columnar or stratified organoids, respectively, depending on culture conditions. b, Bar graph depicting the number of cells positive for KRT7, p63 or KRT5. Data represented as mean ± s.e.m from biologically independent experiments of 3 donors. c, Confocal images of human ectocervical epithelial cells grown in 2D. p63⁺ cells are present in WNT-deficient medium as well as in WNT-proficient medium, but their growth is reduced. However, organoids are formed only in WNT-deficient medium. d-f, Tiled brightfield images of entire female genital system sections from mice fed vitamin A-deficient diet. smRNA-ISH labelling for Krt5 (d), Krt8 (e) and Axin2 (f); nuclei in blue. Boxed areas are magnified on the right. Data representative of n = biologically independent experiments from 3 mice or human samples. Statistical source data are provided in Source Data Extended Data Fig. 7.

Source data

Extended Data Fig. 8 Endo- and ectocervical lineage markers as improved cancer classifiers.

a, Distribution of mRNA expression values and selected thresholds used to define high and low KRT5-, KRT7- and TP63-expressing cancer samples. b, Distribution of Squamous-Columnar correlation scores used for the classification of cancer samples as squamous-like (> 0.2), columnar-like (< -0.2) or undetermined (-0.2 to 0.2). Dotted line: distribution of scores for the same samples with 1,000 random gene sets. c, Density plot showing estimated cancer cell purity for samples designated as columnar-like, squamous-like and the remaining undetermined samples. d, Summary table characterizing samples designated as columnar-like, squamous-like and remaining cases. Given are the numbers and percentages for each type. e, Classification of cancer samples according to histopathological diagnosis, TCGA iCluster classes as well as similarity to squamous and columnar lineages as proposed in this work.

Extended Data Fig. 9 Expression of KRT7 and lineage-specific markers.

a, Tiled brightfield images of sections showing the entire mouse genital system labelled with smRNA-ISH for Krt7. Boxes are magnified on the right. Data representative of n = biologically independent experiments from 3 mice. b-d, Confocal images of human ectocervical stratified and endocervical columnar organoids immunolabeled for AGR2, GDA, KRT18, KRT5 and CSTA; nuclei in blue. Images are representative of biologically independent experiments from 3 humans.

Supplementary information

Reporting Summary

Peer review information

Supplementary Table 1

Marker genes for annotated clusters of three healthy murine cervix datasets. Related to Fig. 1b. Marker genes (rows) associated with the population subsets (column 1) are extracted from Seurat (P < 0,01, column 2). Columns 3 to 6 give the default Seurat output (pct1 and pct2 refer to the percentage of cells expressing genes among the studied subset (pct1) and among the whole dataset analysed (pct2)). Ensembl gene number and gene names are given in columns 7 and 8, respectively. Statistical significance was calculated employing the Wilcoxon rank-sum test for genes detected for 10% of barcodes with at least 0.25-fold difference (log scale) across clusters using the Seurat function.

Supplementary Table 2

Marker genes for annotated clusters of the epithelial subsets from three healthy murine cervix datasets. Related to Fig. 1d. Marker genes (rows) associated with the population subsets (column 1) are extracted from Seurat (P < 0,01, column 2). Columns 3 to 6 give the default Seurat output (pct1 and pct2 refer to the percentage of cells expressing genes among the studied subset (pct1) and among the whole dataset analysed (pct2)). Ensembl gene number and gene names are given in columns 7 and 8, respectively. Statistical significance was calculated employing the Wilcoxon rank-sum test for genes detected for 10% of barcodes with at least 0.25-fold difference (log scale) across clusters using the Seurat function.

Supplementary Table 3

Gene Ontology categories for annotated clusters of the epithelial subset from three murine healthy cervix datasets. Related to Fig. 1f. Displayed GO terms for each subpopulation on Fig. 1f are labelled in red. Gene ontology (rows) associated to the epithelial subset from three murine healthy cervix datasets are calculated using clusterProfiler. For each distinct cluster (column 1, as identified in UMAP projection (Fig. 1d), GO terms (columns 2 and 3) are classified by decreasing P-value. The table reproduces the output of clusterProfiler (columns 3 to 7). Genes associated with each GO term are identified by their Entrez number (columm 8). Gene counts are given column 9. Statistical significance was calculated employing the Wilcoxon rank-sum test for genes detected for 10% of barcodes with at least 0.25-fold difference (log scale) across clusters using the Seurat function.

Supplementary Table 4

Marker genes for annotated clusters of the stromal subset from three murine healthy cervix datasets. Related to Fig. 3a. Marker genes (rows) associated with the population subsets (column 1) are extracted from Seurat (P < 0,01, column 2). Columns 3 to 6 give the default Seurat output (pct1 and pct2 refer to the percentage of cells expressing genes among the studied subset (pct1) and among the whole dataset analysed (pct2)). Ensembl gene number and gene names are given in columns 7 and 8, respectively. Statistical significance was calculated employing the Wilcoxon rank-sum test for genes detected for 10% of barcodes with at least 0.25-fold difference (log scale) across clusters using the Seurat function.

Supplementary Table 5

Differentially expressed genes from the heat maps and squamous versus columnar signature. Related to Fig. 4e. Statistical significance was calculated by two-sided moderated t-test adjusted for multiple testing by the method of Benjamini–Hochberg (FDR).

Supplementary Table 6

Genes differentially expressed in 2D versus differentiated organoids from human ectocervix. Related to Fig. 4e. Statistical significance was calculated by two-sided moderated t-test adjusted for multiple testing by the method of Benjamini–Hochberg (FDR).

Supplementary Table 7

Genes differentially expressed in early versus differentiated organoids from human ectocervix. Related to Fig. 4e. Statistical significance was calculated by two-sided moderated t-test adjusted for multiple testing by the method of Benjamini–Hochberg (FDR).

Supplementary Table 8

Genes differentially expressed in differentiated organoids from human ecto- and endocervix. Related to Fig. 4e. Statistical significance was calculated by two-sided moderated t-test adjusted for multiple testing by the method of Benjamini–Hochberg (FDR).

Supplementary Table 9

Differentially expressed genes from the heat maps and squamous versus columnar signature. Related to Fig. 4f. Statistical significance was calculated by two-sided moderated t-test adjusted for multiple testing by the method of Benjamini–Hochberg (FDR).

Supplementary Table 10

Differentially expressed genes from the heat maps and squamous versus columnar signature. Related to Extended Data Fig. 6c. Statistical significance was calculated by two-sided moderated t-test adjusted for multiple testing by the method of Benjamini–Hochberg (FDR).

Supplementary Table 11

Marker genes for annotated clusters of five healthy and metaplastic murine cervix datasets. Related to Fig. 5e. Marker genes (rows) associated with the population subsets (column 1) are extracted from Seurat (P < 0.01, column 2). Columns 3 to 6 give the default Seurat output (pct1 and pct2 refer to the percentage of cells expressing genes among the studied subset (pct1) and among the whole dataset analysed (pct2)). Ensembl gene number and gene names are given in columns 7 and 8, respectively. Statistical significance was calculated employing the Wilcoxon rank-sum test for genes detected for 10% of barcodes with at least 0.25-fold difference (log scale) across clusters using the Seurat function.

Supplementary Table 12

Marker genes for annotated clusters of the epithelial subsets from five healthy and metaplastic murine cervix datasets. Related to Fig. 5g. Marker genes (rows) associated with the population subsets (column 1) are extracted from Seurat (P < 0.01, column 2). Columns 3 to 6 give the default Seurat output (pct1 and pct2 refer to the percentage of cells expressing genes among the studied subset (pct1) and among the whole dataset analysed (pct2)). Ensembl gene number and gene names are given in columns 7 and 8, respectively. Statistical significance was calculated employing the Wilcoxon rank-sum test for genes detected for 10% of barcodes with at least 0.25-fold difference (log scale) across clusters using the Seurat function.

Supplementary Table 13

Marker genes for annotated clusters of the stromal subsets from five healthy and metaplastic murine cervix datasets. Related to Fig. 6b. Marker genes (rows) associated with the population subsets (column 1) are extracted from Seurat (P < 0.01, column 2). Columns 3 to 6 give the default Seurat output (pct1 and pct2 refer to the percentage of cells expressing genes among the studied subset (pct1) and among the whole dataset analysed (pct2)). Ensembl gene number and gene names are given in columns 7 and 8, respectively. Statistical significance was calculated employing the Wilcoxon rank-sum test for genes detected for 10% of barcodes with at least 0.25-fold difference (log scale) across clusters using the Seurat function.

Supplementary Table 14

Classification of cancer samples based on the expression signature of squamous ectocervical and columnar endocervical organoids.

Supplementary Table 15

Table of antibody dilutions and validation links.

Supplementary Table 16

Patient information.

Source data

Source Data Fig. 2

Statistical source data

Source Data Fig. 3

Statistical source data

Source Data Fig. 4

Statistical source data

Source Data Extended Data Fig. 4

Statistical source data

Source Data Extended Data Fig. 7

Statistical source data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Chumduri, C., Gurumurthy, R.K., Berger, H. et al. Opposing Wnt signals regulate cervical squamocolumnar homeostasis and emergence of metaplasia. Nat Cell Biol 23, 184–197 (2021). https://doi.org/10.1038/s41556-020-00619-0

Download citation

Received: 17 December 2019
Accepted: 26 November 2020
Published: 18 January 2021
Issue Date: February 2021
DOI: https://doi.org/10.1038/s41556-020-00619-0

This article is cited by

Decoding spatiotemporal transcriptional dynamics and epithelial fibroblast crosstalk during gastroesophageal junction development through single cell analysis
- Naveen Kumar
- Pon Ganish Prakash
- Cindrilla Chumduri
Nature Communications (2024)
Crosstalk of ferroptosis regulators and tumor immunity in pancreatic adenocarcinoma: novel perspective to mRNA vaccines and personalized immunotherapy
- Yanlong Shi
- Yizhu Wang
- Yewei Zhang
Apoptosis (2023)
Modelling Chlamydia and HPV co-infection in patient-derived ectocervix organoids reveals distinct cellular reprogramming
- Stefanie Koster
- Rajendra Kumar Gurumurthy
- Cindrilla Chumduri
Nature Communications (2022)
Patient-derived and mouse endo-ectocervical organoid generation, genetic manipulation and applications to model infection
- Rajendra Kumar Gurumurthy
- Stefanie Koster
- Cindrilla Chumduri
Nature Protocols (2022)
Integrated analysis of cervical squamous cell carcinoma cohorts from three continents reveals conserved subtypes of prognostic significance
- Ankur Chakravarthy
- Ian Reddin
- Tim R. Fenton
Nature Communications (2022)

Subjects

Abstract

Similar content being viewed by others

Main

Results

Distinct cellular origins of squamous and columnar epithelium

Opposing stromal Wnt signals define the epithelial borders at the transition zone

Wnt antagonists, Notch and EGFR signalling maintain ectocervical stemness and differentiation

Remodelling of the stromal compartment drives the emergence of KRT5+ squamous metaplasia

Gene expression patterns of cervical SCC and ADC correlate with squamous and columnar lineage organoid signatures

Discussion

Methods

Antibodies and chemicals

Mice

Single-cell isolation and sequencing

Computational analysis of scRNA-seq data

Cellranger pipeline

Quality control

Clustering

Cell cycle annotation

Differential gene expression analysis

Batch-effect assessment

Further analysis

Vitamin A-deficient diet

Mouse ecto- and endocervical medium

Wnt-deficient human ectocervical medium

Wnt-proficient human endocervical medium

Epithelial stem cell isolation from human and mouse cervix

Human epithelial 2D cell culture

Organoid culture and maintenance

Organoid-forming efficiency

Immunofluorescent histochemistry

Whole-mount staining

smRNA-ISH

RNA isolation and quality control for microarray analysis

Microarray expression profiling

Analysis of stem cell-related genes

Gene set enrichment analysis

Cervical cancer data

Statistics and reproducibility

Reporting Summary

Data availability

Code Availability

References

Acknowledgements

Funding

Author information

Authors and Affiliations

Contributions

Corresponding authors

Ethics declarations

Competing interests

Additional information

Extended data

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

This article is cited by

Search

Quick links

Remodelling of the stromal compartment drives the emergence of KRT5⁺ squamous metaplasia