Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Tissue curvature and apicobasal mechanical tension imbalance instruct cancer morphogenesis


Tubular epithelia are a basic building block of organs and a common site of cancer occurrence1,2,3,4. During tumorigenesis, transformed cells overproliferate and epithelial architecture is disrupted. However, the biophysical parameters that underlie the adoption of abnormal tumour tissue shapes are unknown. Here we show in the pancreas of mice that the morphology of epithelial tumours is determined by the interplay of cytoskeletal changes in transformed cells and the existing tubular geometry. To analyse the morphological changes in tissue architecture during the initiation of cancer, we developed a three-dimensional whole-organ imaging technique that enables tissue analysis at single-cell resolution. Oncogenic transformation of pancreatic ducts led to two types of neoplastic growth: exophytic lesions that expanded outwards from the duct and endophytic lesions that grew inwards to the ductal lumen. Myosin activity was higher apically than basally in wild-type cells, but upon transformation this gradient was lost in both lesion types. Three-dimensional vertex model simulations and a continuum theory of epithelial mechanics, which incorporate the cytoskeletal changes observed in transformed cells, indicated that the diameter of the source epithelium instructs the morphology of growing tumours. Three-dimensional imaging revealed that—consistent with theory predictions—small pancreatic ducts produced exophytic growth, whereas large ducts deformed endophytically. Similar patterns of lesion growth were observed in tubular epithelia of the liver and lung; this finding identifies tension imbalance and tissue curvature as fundamental determinants of epithelial tumorigenesis.


Mechanical alterations in cancer cells have an important role in tumorigenesis5. Here we address how changes in cellular forces contribute to tumour formation in the context of an intact epithelium, using the pancreatic duct as a model. To preserve the geometric complexity of the adult pancreas, we developed a technique for rapid, whole-organ 3D immunostaining and imaging, which we term ‘fast light-microscopic analysis of antibody-stained whole organs’ (FLASH; see Methods); this technique enables robust quantitative assessment of organ architecture at the single-cell and tissue level. FLASH maintained pancreatic compartmentalization and tissue integrity (Extended Data Fig. 1a, b). We visualized the adult pancreatic ductal system by inducing expression of tdTomato in all duct cells (of Rosa26CAG-tdTomato;Hnf1bcreERT2 mice). FLASH imaging of whole pancreata revealed an intricate hierarchy of tubules that span the exocrine lobules (Fig. 1a, b, Extended Data Fig. 1c, d and Supplementary Video 1). Ductal segments varied in diameter (Fig. 1c, d), with smaller ducts composed of elongated cells and large ducts of cuboidal cells (Fig. 1e, f). Confetti labelling showed clonal expansion predominantly along the cell elongation axis (Extended Data Fig. 1e, f). These findings illuminate the complexity and heterogeneity of the ductal system of the pancreas (Fig. 1g).

Fig. 1: FLASH imaging of the intact pancreas and visualization of the ductal tree.

ac, Three-dimensional views of an intact pancreas from tamoxifen-treated Rosa26CAG-tdTomato;Hnf1bCreERT2 mouse. n = 3 mice. a, Whole pancreas attached to duodenum and spleen. b, Enlargement of area outlined by the dashed rectangle in a. c, Ductal segments corresponding to numbered rectangles in b, showing main duct (1), interlobular duct (2), intralobular duct (3) and intercalated ducts (4). d, Segmental heterogeneity of duct diameter. Mean ± s.d.; n = 11 main ducts, n = 67 interlobular ducts, n = 34 intralobular ducts and n = 72 intercalated ducts from 3 mice. e, Pancreas tissue section stained for Cdh1 and DNA to highlight shapes of duct cells. n = 4 mice. f, Segmental heterogeneity of duct cell width (black), height (red) and length (blue); averages of 5 cells per duct, n = 115 ducts from 6 mice. Fitted lines were obtained using nonlinear regression. g, Segmental heterogeneity of the ductal tree. L, cell length; W, cell width; H, cell height. Scale bars: 5 mm (a), 500 μm (b), 100 μm (c), 10 μm (e).

Source Data

To trigger epithelial deformations, we induced conditional mosaic activation of the KrasG12D oncogene with concomitant deletion of either p53 or Fbw7 tumour suppressors1,2,6. FLASH analysis of KrasG12D;Fbw7flox/flox;Ck19creERT (hereafter, KFCk19; Fbw7 is also known as Fbxw7, and Ck19 is also known as Krt19) and KrasG12D;Fbw7flox/flox;Hnf1bcreERT (hereafter, KFH) mice revealed two morphologically distinct lesion types, which co-occurred in all pancreata we analysed. Transformed ducts either evaginated basally away from the duct lumen (designated ‘exophytic’), or invaginated apically towards the duct lumen (designated ‘endophytic’) (Fig. 2a and Extended Data Fig. 2a–d). Exophytic lesions extended the duct lumen and formed globular structures (Fig. 2b, c, Extended Data Fig. 2e–g and Supplementary Video 2), which progressed to back-to-back gland-like ductal neoplasia (Fig. 2d). By contrast, endophytic lesions grew into the duct lumen in a papillary manner (Fig. 2e, f, Extended Data Fig. 2e and Supplementary Video 3) and progressed to intratubular neoplasia with local obstruction of the duct lumen (Fig. 2g). Activation of KrasG12D and deletion of Fbw7 or p53 in acinar cells (located at the tips of small-calibre ducts) induced acinar-to-ductal metaplasia, which led to Krt19-positive globular lesions that were continuous with the ductal tree (Extended Data Fig. 3a–h and Supplementary Videos 4, 5). In mice with duct-specific (referred to as KPCk19 mice) or pancreas-wide (referred to as KPC mice) KrasG12D activation and p53 deletion1,2,7, we also identified exophytic and endophytic lesions, which indicates that these observations are not dependent on a particular oncogene combination (Extended Data Fig. 4a–e). Lesions with endophytic and exophytic morphology were also observed in the pancreatic tissue of patients with pancreatic ductal adenocarcinoma (Extended Data Fig. 4f).

Fig. 2: Heterogeneity of neoplasia induction in the pancreatic ducts.

a, Exophytic and endophytic ductal deformations. bg, FLASH comparison with 2D histology. b, e, Three-dimensional view (left) and optical section (right) of KFCk19 exophytic (b) and endophytic (e) deformations ten days after recombination; n = 3 mice. c, d, f, g, Haematoxylin and eosin (H & E) staining of exophytic (c, d) and endophytic (f, g) lesions 10 days (c, f) and 21 days (d, g) after recombination; 10 days, n = 4 mice; 21 days, n = 5 mice. h, pMLC2, Cdh1 and EYFP staining of a transformed, but un-deformed, KFCk19 clone and position of pixel intensity measurement (dotted lines). Cdh1 demarcates lateral cell surfaces. n = 7 mice. i, Basal–apical pMLC2 pixel intensity profiles measured as in h. Mean ± s.e.m. (exophytic, n = 45 transformed cells and n = 35 wild-type cells; endophytic, n = 79 transformed cells and n = 71 wild-type cells, from 7 mice). j, Basal and apical pMLC2 intensities (maxima of cell profiles from endophytic lesions). n = 71 wild-type cells and n = 79 transformed cells, from 7 mice. Mean ± s.d.; P < 0.0001 (two-sided Mann–Whitney test). k, Basal–apical F-actin intensity profiles of cells from KPC mice or wild-type littermates, measured as in h. Mean ± s.e.m. (wild type, n = 120 cells from 4 mice; transformed, n = 212 cells from 3 mice). l, pMLC2 and Cdh1 staining of pancreatic wild-type and transformed (KPC) organoids. Apical cell surfaces face inward. n = 3 wild-type and 3 KPC organoid lines. m, Apical–basal pMLC2 intensities of single cells from KPC organoids (T1, T2 and T3) or wild-type organoids (W1, W2 and W3). Mean ± s.d. (W1, n = 95 cells; W2, n = 166 cells; W3, n = 131 cells; T1, n = 88 cells; T2, n = 80 cells; T3, n = 118 cells). n, Summary of mechanical changes in transformed cells. Scale bars: 50 μm (b, e), 100 μm (c, d, f, g), 10 μm (h), 20 μm. (l).

Source Data

Pancreatic cancer cells co-opt pancreatic stellate cells as cancer-associated fibroblasts, which support tumour invasion and metastasis8,9. Both mouse and human exophytic lesions—which invade the surrounding parenchyma—recruited cancer-associated fibroblasts more efficiently than the endophytic lesions that grow into the duct lumen, although both lesion types showed rapid proliferation (Extended Data Fig. 5a–c). Advanced exophytic lesions displayed more prominent epithelial-to-mesenchymal transition, as judged by the number of tumour-traced mesenchymal cells (Extended Data Fig. 5d, e). Thus, exophytic growth fosters more aggressive lesion development.

Because tissue morphology depends on cellular tension and contractility, which is driven by the actomyosin cortex10, we quantified the distribution of the key cortex components F-actin and pMLC2 (phosphorylated myosin light chain 2)11,12,13 in single cells in KFCk19 and KPC mice, shortly after oncogene activation and before the occurrence of neoplastic lesions. The intensity of F-actin and pMLC2 staining was higher apically than basally in wild-type cells. Upon transformation, this gradient was strongly reduced (Fig. 2h–k and Extended Data Fig. 6a–h). Acinar-derived lesions, which show redistributed F-actin14,15, displayed a similar apical–basal shift of pMLC2 (Extended Data Fig. 6i, j); this supports the notion that the effect of transformation is independent of cell type. Phosphorylated focal adhesion kinase and vinculin were redistributed in the same manner as F-actin in these lesions, and the integrins Itga2, Itga6 and Itgb116 were overexpressed basally (Extended Data Fig. 6k, l). When cultured in vitro as organoids, pancreatic cells from KPC mice displayed disruptions of the pMLC2 gradient that were similar to those observed in vivo (Fig. 2l, m).

We then investigated possible molecular mechanisms that may underpin pMLC2 perturbation in transformed cells. Phosphatase inhibition in wild-type organoids abolished the apical–basal pMLC2 gradient, which suggests that phosphatase activity contributes to pMLC2 polarization (Extended Data Fig. 7a–d). Disruption of apical–basal polarization has previously been reported in Kras transformed cells, and oncogenic Kras signalling stimulates pMLC217,18. ERK, a target of MEK, regulates actin polymerization19 and activates myosin light-chain kinase, which phosphorylates MLC220,21. We found that KrasG12D;Pdx1cre (hereafter, KC) mice contained ductal structures that had an equal apical–basal pMLC2 distribution, which is consistent with the observed distribution of F-actin in KC mice14 (Extended Data Fig. 7e, f). Furthermore, apical pMLC2 accumulation could be partially restored by inhibition of MEK in KPC organoids. Overall, these results identify oncogenic hyperactivation of Kras–MEK–ERK as a driver of shifts in pMLC2, possibly by overriding phosphatase activity (Fig. 2n and Extended Data Fig. 7g–j). By contrast, ROCK inhibition led to an overall reduction of pMLC2, without affecting its relative distribution (Extended Data Fig. 7k–n).

To test whether the pMLC2 perturbation in transformed cells could explain the observed lesion deformation, we developed a computational model of the pancreatic duct using a 3D vertex model simulation, which integrates duct cell geometry, apical, basal and lateral surface tensions, and cell volume conservation22,23 (Fig. 3a). We assumed that apical and basal surface tensions in wild-type and transformed cells were proportional to the relative fluorescence intensities of pMLC211,12,24 (Fig. 3b and Supplementary Modelling Procedures). To mimic oncogenic transformation, we introduced a single cell with surface tensions that were modified in accordance with changes in experimentally measured average pMLC2 fluorescence intensity (Fig. 2j), and simulated successive rounds of cell division (Fig. 3b) to match cell counts that were experimentally observed at different time points. Simulations were performed for tubes of different diameters by varying the number of circumferential cells (Extended Data Fig. 8a). Notably, without further fitting, we found evaginating and invaginating deformations that closely resembled those observed in the pancreas (Fig. 3c). Furthermore, simulations showed a transition from evaginating deformations in small ducts to invaginating deformations in large ducts at about 17 μm diameter (Fig. 3d and Extended Data Fig. 8b). In simulations in which only cellular proliferation was considered (that is, without changes in the surface tension of the transformed cells), deformations were exophytic and no transition was observed (Extended Data Fig. 8c).

Fig. 3: Duct diameter instruction of lesion morphology.

a, Schematic of the 3D vertex model. Apical and basal vertices represent tissue geometry. Effective energy (W)—which takes into account intracellular hydrostatic pressures (Pα) constraining cell volumes (Vα), and apical, basal and lateral surface tensions (\({T}_{{\rm{a}}}^{{\rm{\alpha }}}\), \({T}_{{\rm{b}}}^{{\rm{\alpha }}}\)and \({T}_{{\rm{l}}}^{{\rm{\alpha }}}\), respectively)—is minimized to obtain mechanically equilibrated shapes. \({A}_{{\rm{a}}}^{{\rm{\alpha }}}\), \({A}_{{\rm{b}}}^{{\rm{\alpha }}}\) and \({A}_{{\rm{l}}}^{{\rm{\alpha }}}\) denote apical, basal and lateral surface areas, respectively. b, For wild-type duct simulations, pMLC2 intensity and cell aspect ratio measurements determine mechanical parameters (see Supplementary Modelling Procedures). A single transformed cell then undergoes cycles of cell division. Cellular interfacial tension changes are introduced in simulated transformed cells, in line with changes in pMLC2 intensity (Fig. 2j). c, Representation of simulated deformed ducts with diameters below 17 μm (exophytic deformation) and above 17 μm (endophytic deformation), and comparison with experimental data from KFCk19 mice. n = 7 mice. Scale bars, 50 μm. d, Quantification of ductal deformation of the pancreas ductal tree as a function of duct diameter. Relative deformation defined by (dtf − dwt)/(dtf + dwt), in which dtf and dwt denote the apical-to-apical ductal diameter within and outside the lesion, respectively. Positive deformation indicates an evaginating (exophytic) lesion, and negative deformation indicates an invaginating (endophytic) lesion. In both simulations and experiments, a transition from exophytic to endophytic lesions occurs at about 17 μm. Left, experimental measurements (smaller dots) and binned mean ± s.e.m. (larger dots) (10 days, n = 112 ducts from 3 mice; 3 weeks, n = 88 ducts from 4 mice). Grey dotted line demarcates transition through 0, as estimated in Extended Data Fig. 8d. Right, predicted ductal deformation as function of duct diameter, from vertex model simulations. Dots are mean ± s.e.m., n = 10 simulations.

Source Data

To assess whether lesion shape correlated with ductal diameter in vivo, we measured the radial deformation of transformed pancreatic ducts of KFCk19 mice throughout the ductal tree. Notably, transformed ducts with a diameter of less than 17 μm deformed exophytically, whereas ducts with a diameter of greater than 17 μm deformed endophytically, consistent with the model predictions (Fig. 3d and Extended Data Fig. 8d). These observations, together with our computational model, strongly suggested that ductal deformations are dependent on duct diameter.

To confirm our computational prediction that hyperproliferation alone (that is, without mechanical changes) is insufficient to induce the formation of endophytic lesions (Extended Data Fig. 8c), we used Fbw7flox/flox;Krt19creERT (FCk19) mice, in which Fbw7 inactivation induces hyperproliferation of duct cells without leading to oncogenic transformation2,25. Fbw7flox/flox duct cells had no alterations in pMLC2 intensity or integrin levels, as compared with wild-type cells (Fig. 4a and Extended Data Fig. 8e–h). FLASH imaging identified local duct-cell crowding and a mild expansion of the duct lumen without neoplastic deformations (Fig. 4b, c). These findings support our model predictions that hyperproliferation alone cannot account for the diameter-dependent formation of exophytic versus endophytic lesions, and that only endophytic lesion growth requires altered apical–basal tension.

Fig. 4: Lesion morphology dependence on tension imbalance and tissue curvature.

a, Quantification of pMLC2 distribution in recombined (hyperproliferative) cells and wild-type neighbours from FCk19 mice; n = 93 wild-type cells, n = 64 hyperproliferative cells from 6 mice. Mean ± s.d. NS, not significant, P = 0.1712 (two-sided Mann–Whitney test). b, Measured (small blue dots), binned experimental (large blue dots, mean ± s.e.m.) and simulated (large red dots, mean ± s.e.m.) ductal deformations of hyperproliferating ducts with mechanical parameters as in wild-type cells of Fig. 3d. Measurements from n = 216 FCk19 clones from 4 mice. Hyperproliferation simulated by selecting one duct cell and performing cell divisions until a clone of 21 cells is formed; n = 10 simulations. c, Three-dimensional view of small and large ducts with local hyperproliferative cell patches (red). n = 4 mice. d, Distribution of endophytic and exophytic neoplasia in KFCk19 pancreatic ducts (n = 62 lesions from 6 mice), intrahepatic biliary tree (n = 24 lesions from 3 mice) and bronchiolar system (n = 29 lesions from 3 mice), as a function of tube diameter outside the lesion. e, FLASH stainings for Krt19 and tdTomato on liver (n = 3) and lung (n = 3). f, H & E staining of tissue sections through endophytic and exophytic lesions from liver (n = 3) and lung (n = 3). g, Model of ductal transformation into exophytic and endophytic neoplastic lesions. Scale bars: 50 μm (c), 100 μm (e, f).

Source Data

To determine whether cytoskeletal alterations could be used to distinguish between hyperproliferative benign and malignant ductal reactions, we compared cellular pMLC2 localization and Itga2 expression in caerulein-induced pancreatitis26, genetic hyperproliferation of FCk19 mice and early neoplastic lesions of KPC mice. Although pMLC2 was apically depleted in transformed duct cells, it was maintained in reactive and hyperproliferative ducts (Extended Data Fig. 9a–c). Itga2 was strongly expressed in transformed lesions, but not in reactive or hyperproliferative ducts (Extended Data Fig. 9d). Thus, cytoskeletal redistribution distinguishes oncogene-induced ductal transformation from benign hyperproliferation.

To test our hypothesis that ductal deformation is dependent on duct diameter, we developed a continuum theory of a circular epithelium subjected to asymmetric apical–basal surface tension27,28,29 (Extended Data Fig. 10a and Supplementary Modelling Procedures). We considered the deformation induced by a localized patch of transformed epithelium. Reflecting the cytoskeletal changes in transformed cells, we assumed that the patch had increased basal and decreased apical tension, which leads to a preferred inward curvature of the transformed tissue (Extended Data Fig. 10b). In small tubes, this tension imbalance is not large enough to overcome the resistance of the surrounding epithelium and the lesion grows outwards in an exophytic manner (Extended Data Fig. 10c, left). In large tubes, early tumours are more free to adopt their preferred invaginating shape, and the lesion grows inward (Extended Data Fig. 10c, right). Our generic continuum model therefore predicts a transition from exophytic to endophytic lesion at a threshold tube radius that depends on the bending modulus of the tissue and the magnitude of the changes in apical–basal tension in the transformed region (Extended Data Fig. 10d).

The simplicity of the continuum model suggested that the dependency of lesion shape on ductal diameter was a universal principle. Keratin 19 is a general marker of ductal epithelia30, and in KFCk19 mice—in addition to pancreatic cancer—we observed neoplastic lesions that resembled squamous cell carcinoma precursors in the airways of the lung, and biliary neoplasia in the intrahepatic biliary tree. In both organs, transformed cells showed apical–basal redistribution of pMLC2, which indicates that oncogenic Kras-driven cytoskeletal disruption is tissue-independent (Extended Data Fig. 10e, f). The hepatic ductal system comprised both small and large ducts, which gave rise to exophytic and endophytic lesions, respectively. However, the bronchiolar tree of the lung is built only of airways with diameters above 30 μm, which gave rise to endophytic neoplasia only (Fig. 4d–f). Thus, a tube diameter of about 20 μm appears to be the threshold for the transition from exophytic to endophytic tumours during lesion initiation in different organs and cancers (Fig. 4g). Together, our findings indicate that the local curvature of the source epithelium—combined with mechanical changes within the transformed cells—instructs the direction of epithelial deformation and ultimately directs tumour morphology.


No statistical methods were used to predetermine sample size. Unless indicated otherwise, the experiments were not randomized and investigators were not blinded to allocation during experiments and outcome assessment.


All experiments were approved by the London Research Institute Animal Ethics Committee or the Animal Welfare and Ethical Review Body of the Francis Crick Institute, and conformed to UK Home Office regulations under the Animals (Scientific Procedures) Act 1986 including Amendment Regulations 2012. Tumour experiments are permitted until mice experience excessive weight loss or display overt signs of illness. In no experiments were permitted endpoints exceeded; all mice were humanely killed and analysed at the onset of transformation before macroscopic tumour development. No longitudinal tumour measurements were performed as this was technically not feasible for microscopic lesions.

Ela1creERT2 (ref. 31), Krt19creERT (ref. 30), Hnf1bcreERT2 (ref. 32), Pdx1cre (ref. 33), Ptf1acreERT (ref. 34), LSLKrasG12D/wt (ref. 35), Fbw7flox/flox (ref. 36), MIPGFP (ref. 37), p53flox/flox (ref. 38; p53 is also known as Trp53), Rosa26Confetti (ref. 39), Rosa26CAG-tdTomato (ref. 40) and Rosa26LSL-YFP (ref. 41) mouse lines have previously been described. Mice were genotyped using Transnetyx and intercrossed to yield the desired genotypes. For Cre recombinase activation, adult mice received daily intraperitoneal injections of tamoxifen (Sigma-Aldrich) in peanut oil (Sigma-Aldrich) at a dose of 100 mg/kg bodyweight as follows: Rosa26CAG-tdTomato;Hnf1bcreERT2 mice received 5 injections, Rosa26Confetti;Hnf1bcreERT2 mice received 2 injections and Krt19creERT mice received 2 injections. Mice were analysed 7 days, 10 days or 3 weeks after the last tamoxifen injection before macroscopic tumours were apparent. Ela1creERT2 and Ptf1acreERT mice received 5 tamoxifen injections. To induce acute pancreatitis, mice were subjected to a short caerulein treatment. Mice received intraperitoneal injections of caerulein (Sigma-Aldrich) in PBS at a dose of 50 μg/kg bodyweight (7 times a day, at hourly intervals, for 2 days). Ela1creERT2 and Ptf1acreERT mice were analysed after 2 to 5 months to study transformation. Wild-type mice were analysed 4 days after caerulein treatment to assess acute pancreatitis.

In-situ retrograde pancreas ductal perfusion

Mice were euthanized by cervical dislocation, and under a stereo microscope the abdominal wall and peritoneum were opened and the pancreatic head located. A suture was placed around the first bifurcation of the extrahepatic biliary tree. With an insulin gauge, 50 μl of 12 mg/ml FITC-conjugated dextran (Sigma-Aldrich; average molecular mass 2,000,000) was injected into the hepato-pancreatic duct by cannulation of the duct through the duodenal papilla.


FLASH was developed for the rapid detection of a multitude of antigens in intact adult organs by mild non-destructive epitope recovery. Mice were euthanized by cervical dislocation. Cardiac perfusion was carried out with 20 ml PBS. Pancreata were removed attached to spleen and duodenum without perturbing the gland. Samples were fixed in 4% PFA overnight at 4 °C. Specimens were washed in PBT (0.4% Triton X-100 (Sigma-Aldrich) in PBS) twice for one hour and incubated in 200 mM boric acid (Sigma-Aldrich) with 4% SDS (Sigma-Aldrich) pH 7.0 overnight at 54 °C. Samples were washed in PBT for three hours with three volume exchanges. For immunolabelling, samples were incubated in FLASH blocking buffer (1% bovine serum albumin (Sigma-Aldrich), 5% DMSO (Sigma-Aldrich), 10% fetal calf serum (Gibco), 0.02% sodium azide (Sigma-Aldrich), 0.2% Triton X-100 in PBS) for one hour and incubated with antisera (all 1:100) for at least 16 h at room temperature on a nutator. Samples were washed by three volume exchanges PBS and incubated with secondary antibodies (all 1:100) for at least two days at room temperature. Samples were washed by three volume exchanges with PBS and gradually dehydrated in 30%, 50%, 75%, 2× 100% MetOH (Sigma-Aldrich), 1 hour each and—in a glass dish—immersed in methyl salicylate diluted in MetOH: 25%, 50%, 75%, 2× 100% methyl salicylate (Sigma-Aldrich), 30 min each, protected from light.

Fluorescent proteins were stained by immunofluorescence. The following antibodies were used: amylase (goat, SCBT), GFP (goat, Abcam), GFP (mouse, Roche), Krt19 TROMA III (rat, DSHB), Tomato (rabbit, Rockland). All secondary antibodies were Alexa-dye conjugates (ThermoFisher). Nuclei were stained with DRAQ5 (Biostatus).

Three-dimensional imaging

Organs were imaged on a Zeiss LSM 780 confocal microscope equipped with a 405-nm laser, an argon laser, a DPSS 561-nm laser, a HeNe 594-nm laser and a HeNe 633-nm laser using the following objective lenses: 10×/0.45 Ph2 Plan-Apochromat and 25×/0.8 LD LCI Plan-Apochromat. Because the limited working distance of the objectives did not permit imaging throughout the total thickness of the liver, liver lobes were bisected before imaging. Three-dimensional image analysis and quantifications were performed using Imaris software (Bitplane). For display purposes, 3D images were gamma-corrected and clipping planes were used to crop datasets to the relevant areas. Scale bars refer to the centre of the 3D view.

Three-dimensional quantifications of ductal and cellular geometry

Cell height, length and duct diameter were measured using Imaris software. Staining for Krt19 was used to demark the duct cell shape, and length and height were measured for individual cells using the measurement tool in optical sections. Five cells were measured per duct. The cell width was calculated using the quantifications of duct diameter and number of circumscribing cells. The duct diameter was measured in optical sections as the distance of opposing duct walls in one optical section (xy plane) and perpendicular to it (z), and for each duct the average of both values was calculated. The number of duct-circumscribing cells was counted manually using 3D representation and optical sections.

Quantification of the relative division plane

Rosa26Confetti;Hnf1bcreERT2 mice received 2 intraperitoneal tamoxifen injections at a dose of 100 mg/kg bodyweight, which induced recombination and fluorescent protein expression in less than 1% of duct cells as estimated from 3D imaging data. Clones were identified as clusters of directly adjacent cells that had the same fluorescence and subcellular fluorescent protein localization, and were surrounded by non-recombined cells. Using Imaris software, a surface describing the clone was automatically reconstructed in the channel of the fluorescent protein and split manually into surfaces describing each cell. Measurement points were positioned automatically into the centre of each surface (clone cell) and into the centre of the ductal lumen on both sides of the clone. The direction of cell division was determined automatically as the angle between the line connecting the cell surface centres and the line connecting the two measurement points in the ductal centre.

Histopathology, immunohistochemistry and immunofluorescence

Histological and immunohistochemical analyses were performed as previously described2. The following antibodies were used: Cdh1 (rat, Novex), cortactin (rabbit, Abcam), GFP (goat, Abcam), GFP (mouse, Roche), ItgA2 (rabbit, Abcam), ItgA6 (rabbit, Abcam), ItgB1 (rabbit, Abcam), Ki67 (rat, Biolegend), Krt19 TROMA-III (rat, DSHB), MLC2 pSer19 (rabbit, NEB), myosin (rabbit, Abcam), nestin (mouse, BD transduction), Pdgfrβ (rabbit, NEB), PTK2 pTyr397 (rabbit, ThermoFisher), SMA (mouse, Agilent), SMA (mouse, Sigma), Tomato (rabbit, Rockland), Tomato (goat, Biorbyt), vimentin (rabbit, NEB), vinculin (mouse, Sigma-Aldrich). DBA–rhodamine and DBA–FITC were from Vectorlabs. F-actin was stained with phalloidin–TRITC (Sigma-Aldrich) and nuclei with DAPI (Sigma-Aldrich). F-actin staining of LSLKrasG12D;p53flox/flox;Pdx1cre pancreata and pancreata from wild-type littermate controls was performed on cryosections. Samples were embedded fresh in OCT medium and, after sectioning, fixed in 5% NBF for 10 min. Slides were washed in 0.2% Triton X-100 in PBS for 10 min and incubated in FLASH blocking buffer for 30 min. Staining reagent incubations were performed as above. Fluorescent stainings were imaged on a Zeiss LSM 780 confocal microscope. Chromogenic DAB stainings were imaged on a Zeiss Axio Scan Z1 Slide Scanner.

Quantification of staining intensities of actomyosin cortex components

Immunofluorescence was performed as detailed above and imaged by confocal microscopy with a 40×/1.3 Plan-Apochromat oil immersion lens and two-to-fourfold optical zoom, or with a 63×/1.46 Alpha Plan-Apochromat oil immersion objective. Lesions were classified as endophytic or exophytic on the basis of recombination status and morphological presentation. At early stages (10 days after recombination), some recombined non-transformed cells may have been included in the transformed classification, but at later stages transformed and non-transformed cells could be classified unambiguously. Per lesion, one optical section was analysed by intensity quantification of transformed cells and neighbouring wild-type cells using ImageJ software (NIH). Images were converted into composite images in which green and red marked Cdh1 (E-cadherin, to identify lateral membrane) and GFP (to identify genetically recombined cells), respectively, and the blue channel depicted the staining to be quantified. In each image, transformed cells were distinguished from wild type by GFP positivity and all cells with clear lateral Cdh1 staining were included in intensity measurements. Cells with apical or basal Cdh1 positivity were omitted from analysis as such a signal could be an artefact from the lateral surfaces of diagonally cut cells in front or behind the cell of interest. Thus, only cells cut through the middle were analysed. Basal–apical intensity gradients were quantified for each cell. Pixel intensities of the blue channel along a straight line from the basal to the apical edge, equidistant from the two Cdh1-stained lateral cell edges, were recorded. To control for differences in cell height and overall staining intensity, the profiles were aligned for cell height by equally distributing values along the profile of the tallest cell without filling the resulting gaps, and normalized for intensity by subtracting the minimum value from each intensity profile and dividing by its average value. For comparison, normalization to nuclear levels was carried out and the localization of intensity peaks was similar in both methods.

For integrin quantifications, intensities were measured along the basal edges of transformed and neighbouring wild-type cells. Freehand lines were drawn between the Cdh1-stained lateral surfaces without crossing the Cdh1 signal to avoid carry-over from the lateral cortex. For each measured cell an additional intensity measurement was taken from the inner nuclear region. To account for overall differences in staining intensities between images, the nuclear intensity was used as background value for normalization per cell.

Human samples

Human pancreatic tissue sections were obtained from patients at King’s College Hospital (n = 4) who were undergoing surgery for pancreatic neoplasia, who kindly consented to donating their samples for research. Sections were stained with H &E, or antibodies for SMA or PDGFRβ, and histopathologically examined. The collection of samples was approved by the NHS Health Research Authority following assessment by a Research Ethics Committee (HSC REC B; reference 16/NI/0119).

Organoid culture

Pancreata were dissected and placed into ice-cold advanced DMEM/F-12 with 50 U/ml penicillin–streptomycin (Thermo Fisher). Tissue was minced on ice and incubated in 1 mg/ml collagenase V (Sigma-Aldrich) in DMEM/F-12 for 40 min at 37 °C. Digestion was stopped by addition of ice-cold advanced DMEM/F-12 and cells collected by centrifugation (300g, 5 min at 4 °C). Samples were further digested with trypsin for 5 min at room temperature and the reaction was stopped with ice-cold 2% FCS in PBS. The suspension was filtered through a 70-μm nylon mesh, and after centrifugation cells were plated in Matrigel. Organoid medium was used as previously described42,43. Organoids were passaged once a week 1:10 using non-enzymatic TrypLE (Thermo Fisher). For pharmacological inhibition of MEK or ROCK activity, KPC organoids were incubated with 8 μM U0126 (Promega) in DMSO or 5 μM H1152 (Tocris) in DMSO or, as control, with an equivalent volume of DMSO in organoid medium for 16 h. For phosphatase inhibition, wild-type organoids were incubated with 2 μM okadaic acid (Tocris) in DMSO or 5 μM tautomycetin (Tocris) in DMSO or equivalent volume of DMSO (control) in organoid medium for 1 h. Experiments were carried out in triplicate for six independent organoid lines established from three wild-type mice and three LSLKrasG12D;p53flox/flox;Pdx1cre mice as indicated in the figure legends. Experiments were performed triple-blinded with one operator preparing the cells, a second operator adding drugs and a third operator staining and quantifying pMLC2 distribution.

Organoid staining

For staining, organoids were cultured in 8-well glass-bottom dishes (Ibidi). Organoids were washed in PBS for 5 min and fixed in 5% NBF in 0.1% Triton X-100 in PBS for 15 min at room temperature. Cells were washed with 0.2% Triton X-100 in PBS for 30 min and incubated in FLASH blocking buffer for 1 h. Primary and secondary antibody incubations were extended overnight at 4 °C with 1 h washes in PBS after each incubation. For imaging, wells were immersed in fluorescence mounting medium (Dako). Images were taken on a Zeiss LSM 780 confocal microscope as described in ‘Three-dimensional imaging’.


For experimental data, statistical significance was calculated using the Mann–Whitney Test (two-tailed, non-parametric, for unpaired samples) with P values as indicated in the figure legend.

Code availability

Custom code for the 3D vertex model is available upon request from S.A. (

Reporting summary

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

Data availability

All relevant data and protocols are included within the paper, Extended Data and Supplementary Information. Source Data for Figs. 1d, f, 2i–k, m, 3d, 4a, b, d and Extended Data Figs. 1d, f, 2b, 5e, 6b–g, i–l, 7d, f, i, j, m, n, 8a–d, f, h, 10e, f are provided with this paper. The original datasets and resulting analyses, as well as methodological details, are available from the corresponding authors upon reasonable request.


  1. 1.

    Bailey, J. M. et al. p53 mutations cooperate with oncogenic Kras to promote adenocarcinoma from pancreatic ductal cells. Oncogene 35, 4282–4288 (2016).

    CAS  PubMed  Article  Google Scholar 

  2. 2.

    Ferreira, R. M. M. et al. Duct- and acinar-derived pancreatic ductal adenocarcinomas show distinct tumor progression and marker expression. Cell Reports 21, 966–978 (2017).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  3. 3.

    Ferone, G. et al. SOX2 is the determining oncogenic switch in promoting lung squamous cell carcinoma from different cells of origin. Cancer Cell 30, 519–532 (2016).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  4. 4.

    Guest, R. V. et al. Cell lineage tracing reveals a biliary origin of intrahepatic cholangiocarcinoma. Cancer Res. 74, 1005–1010 (2014).

    CAS  PubMed  Article  Google Scholar 

  5. 5.

    Paszek, M. J. et al. Tensional homeostasis and the malignant phenotype. Cancer Cell 8, 241–254 (2005

    CAS  PubMed  Article  Google Scholar 

  6. 6.

    Zhang, Q. et al. Fbxw7 deletion accelerates Kras G12D-driven pancreatic tumorigenesis via Yap accumulation. Neoplasia 18, 666–673 (2016).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  7. 7.

    Rhim, A. D. et al. EMT and dissemination precede pancreatic tumor formation. Cell 148, 349–361 (2012).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  8. 8.

    Hanahan, D. & Coussens, L. M. Accessories to the crime: functions of cells recruited to the tumor microenvironment. Cancer Cell 21, 309–322 (2012).

    CAS  PubMed  Article  Google Scholar 

  9. 9.

    Öhlund, D. et al. Distinct populations of inflammatory fibroblasts and myofibroblasts in pancreatic cancer. J. Exp. Med. 214, 579–596 (2017).

    PubMed  PubMed Central  Google Scholar 

  10. 10.

    Heer, N. C. & Martin, A. C. Tension, contraction and tissue morphogenesis. Development 144, 4249–4260 (2017).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  11. 11.

    Goeckeler, Z. M. & Wysolmerski, R. B. Myosin light chain kinase-regulated endothelial cell contraction: the relationship between isometric tension, actin polymerization, and myosin phosphorylation. J. Cell Biol. 130, 613–627 (1995).

    CAS  PubMed  Article  Google Scholar 

  12. 12.

    Elliott, H. et al. Myosin II controls cellular branching morphogenesis and migration in three dimensions by minimizing cell-surface curvature. Nat. Cell Biol. 17, 137–147 (2015).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  13. 13.

    Lomakin, A. J. et al. Competition for actin between two distinct F-actin networks defines a bistable switch for cell polarization. Nat. Cell Biol. 17, 1435–1445 (2015).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  14. 14.

    Heid, I. et al. Early requirement of Rac1 in a mouse model of pancreatic cancer. Gastroenterology 141, 719–730.e7 (2011).

    CAS  PubMed  Article  Google Scholar 

  15. 15.

    Baer, R. et al. Pancreatic cell plasticity and cancer initiation induced by oncogenic Kras is completely dependent on wild-type PI 3-kinase p110α. Genes Dev. 28, 2621–2635 (2014).

    PubMed  PubMed Central  Article  Google Scholar 

  16. 16.

    Pickup, M. W., Mouw, J. K. & Weaver, V. M. The extracellular matrix modulates the hallmarks of cancer. EMBO Rep. 15, 1243–1253 (2014).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  17. 17.

    Schäfer, C. et al. Differential KrasV12 protein levels control a switch regulating lung cancer cell morphology and motility. Converg. Sci. Phys. Oncol. 2, 035004 (2016).

    PubMed  PubMed Central  Article  Google Scholar 

  18. 18.

    Magudia, K., Lahoz, A. & Hall, A. K-Ras and B-Raf oncogenes inhibit colon epithelial polarity establishment through up-regulation of c-myc. J. Cell Biol. 198, 185–194 (2012).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  19. 19.

    Mendoza, M. C., Vilela, M., Juarez, J. E., Blenis, J. & Danuser, G. ERK reinforces actin polymerization to power persistent edge protrusion during motility. Sci. Signal. 8, ra47 (2015).

    PubMed  PubMed Central  Article  Google Scholar 

  20. 20.

    Nguyen, D. H. D. et al. Myosin light chain kinase functions downstream of Ras/ERK to promote migration of urokinase-type plasminogen activator-stimulated cells in an integrin-selective manner. J. Cell Biol. 146, 149–164 (1999).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  21. 21.

    Shen, Q., Rigor, R. R., Pivetti, C. D., Wu, M. H. & Yuan, S. Y. Myosin light chain kinase in microvascular endothelial barrier function. Cardiovasc. Res. 87, 272–280 (2010).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  22. 22.

    Bielmeier, C. et al. Interface contractility between differently fated cells drives cell elimination and cyst formation. Curr. Biol. 26, 563–574 (2016).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  23. 23.

    Alt, S., Ganguly, P. & Salbreux, G. Vertex models: from cell mechanics to tissue morphogenesis. Phil. Trans. R. Soc. Lond. B 372, 20150520 (2017).

    Article  Google Scholar 

  24. 24.

    Clark, A. G., Wartlick, O., Salbreux, G. & Paluch, E. K. Stresses at the cell surface during animal cell morphogenesis. Curr. Biol. 24, R484–R494 (2014).

    CAS  PubMed  Article  Google Scholar 

  25. 25.

    Sancho, R., Gruber, R., Gu, G. & Behrens, A. Loss of Fbw7 reprograms adult pancreatic ductal cells into α, δ, and β cells. Cell Stem Cell 15, 139–153 (2014).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  26. 26.

    Reichert, M. & Rustgi, A. K. Pancreatic ductal cells in development, regeneration, and neoplasia. J. Clin. Invest. 121, 4572–4578 (2011).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  27. 27.

    Marchetti, M. C. et al. Hydrodynamics of soft active matter. Rev. Mod. Phys. 85, 1143–1189 (2013).

    ADS  CAS  Article  Google Scholar 

  28. 28.

    Prost, J., Jülicher, F. & Joanny, J. F. Active gel physics. Nat. Phys. 11, 111–117 (2015).

    CAS  Article  Google Scholar 

  29. 29.

    Salbreux, G. & Jülicher, F. Mechanics of active surfaces. Phys. Rev. E 96, 032404 (2017).

    ADS  PubMed  Article  Google Scholar 

  30. 30.

    Means, A. L., Xu, Y., Zhao, A., Ray, K. C. & Gu, G. CK19CreERT knockin mouse line allows for conditional DNA recombination in epithelial cells in multiple endodermal organs. Genesis 46, 318–323 (2008).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  31. 31.

    Desai, B. M. et al. Preexisting pancreatic acinar cells contribute to acinar cell, but not islet β cell, regeneration. J. Clin. Invest. 117, 971–977 (2007).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  32. 32.

    Solar, M. et al. Pancreatic exocrine duct cells give rise to insulin-producing β cells during embryogenesis but not after birth. Dev. Cell 17, 849–860 (2009).

    CAS  PubMed  Article  Google Scholar 

  33. 33.

    Hingorani, S. R. et al. Preinvasive and invasive ductal pancreatic cancer and its early detection in the mouse. Cancer Cell 4, 437–450 (2003).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  34. 34.

    Pan, F. C. et al. Spatiotemporal patterns of multipotentiality in Ptf1a-expressing cells during pancreas organogenesis and injury-induced facultative restoration. Development 140, 751–764 (2013).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  35. 35.

    Jackson, E. L. et al. Analysis of lung tumor initiation and progression using conditional expression of oncogenic K-ras. Genes Dev. 15, 3243–3248 (2001).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  36. 36.

    Hoeck, J. D. et al. Fbw7 controls neural stem cell differentiation and progenitor apoptosis via Notch and c-Jun. Nat. Neurosci. 13, 1365–1372 (2010).

    CAS  PubMed  Article  Google Scholar 

  37. 37.

    Hara, M. et al. Transgenic mice with green fluorescent protein-labeled pancreatic β-cells. Am. J. Physiol. Endocrinol. Metab. 284, E177–E183 (2003).

    CAS  PubMed  Article  Google Scholar 

  38. 38.

    Marino, S., Vooijs, M., van Der Gulden, H., Jonkers, J. & Berns, A. Induction of medulloblastomas in p53-null mutant mice by somatic inactivation of Rb in the external granular layer cells of the cerebellum. Genes Dev. 14, 994–1004 (2000).

    CAS  PubMed  PubMed Central  Google Scholar 

  39. 39.

    Snippert, H. J. et al. Intestinal crypt homeostasis results from neutral competition between symmetrically dividing Lgr5 stem cells. Cell 143, 134–144 (2010).

    CAS  PubMed  Article  Google Scholar 

  40. 40.

    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).

    CAS  PubMed  Article  Google Scholar 

  41. 41.

    Srinivas, S. et al. Cre reporter strains produced by targeted insertion of EYFP and ECFP into the ROSA26 locus. BMC Dev. Biol. 1, 4 (2001).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  42. 42.

    Huch, M. et al. Unlimited in vitro expansion of adult bi-potent pancreas progenitors through the Lgr5/R-spondin axis. EMBO J. 32, 2708–2721 (2013).

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  43. 43.

    Boj, S. F. et al. Organoid models of human and mouse ductal pancreatic cancer. Cell 160, 324–338 (2015).

    CAS  PubMed  Article  Google Scholar 

Download references


We thank C. Cremona, V. Gebala, M. Popovic, N. Tapon and B. Thompson for comments on the manuscript, and the Francis Crick Institute Biological Research, Experimental Histopathology and Light Microscopy facilities for technical support. This work was supported by the Francis Crick Institute, which receives its core funding from Cancer Research UK (FC001039, FC001317), the UK Medical Research Council (FC001039, FC001317) and the Wellcome Trust (FC001039, FC001317). This work was also supported by an ERC grant to A.B. (281661).

Reviewer information

Nature thanks G. Danuser, N. Minc and the other anonymous reviewer(s) for their contribution to the peer review of this work.

Author information




R.M.M.F., C.G. and V.M.-Y.W. contributed equally to this work. H.A.M., S.A., G.S. and A.B. conceived and designed the study. H.A.M. performed the experiments and analyses. C.G. and V.M.-Y.W. helped with organoid experiments. S.A. performed and analysed simulations, and G.S. analysed the continuum theory. R.M.M.F. assisted with study design and interpretation of data. C.G.C. assisted with lesion identification and description. G.S. and A.B. supervised the study. All authors discussed and interpreted the results and participated in writing the manuscript.

Corresponding authors

Correspondence to Guillaume Salbreux or Axel Behrens.

Ethics declarations

Competing interests

H.A.M. and A.B. are inventors on a UK patent application (1818567.8) relating to a solution for the preparation of samples for 3D imaging.

Additional information

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

Extended data figures and tables

Extended Data Fig. 1 Preserved organ integrity after FLASH.

a, FLASH staining of an insulin (Ins)–GFP reporter mouse for amylase (Amy), Krt19 and Ins–GFP. Left, 3D reconstruction that demonstrates the complex organization of islets and pancreatic ducts. Scale bar, 100 μm. Right panels show an optical section through the area indicated in the left panel (white dotted rectangle), showing preserved compartmentalization into exocrine and endocrine glands as seen by mutually exclusive staining for Amy, Krt19 and Ins–GFP. Scale bars, 50 μm. Representative of five mice. b, Three-dimensional view of a high-calibre duct (32-μm diameter) stained for Krt19 and DNA, demonstrating preserved epithelial integrity. Scale bar, 50 μm. b′, Left, optical section through the indicated area in b (red dashed rectangle), demonstrating the continuous duct-cell monolayer and preserved duct lumen. Scale bar, 30 μm. Representative of six mice. Right, staining for Krt19 and DNA on a 4-μm tissue section of paraffin-embedded pancreas. Scale bar, 30 μm. c, FLASH staining for tdTomato (tdTOM) and Krt19 on pancreata of Rosa26CAG-tdTomato;Hnf1bcreERT2 mice without (left) and with (right) intraperitoneal injection of 100 μg tamoxifen per gram bodyweight. Scale bars, 500 μm. Representative of three mice. d, Arborization of the ductal tree. Each dot represents one duct and lines indicate ramification. For three Krt19-stained pancreata, z-stacks of at least 30 random high-magnification views were taken. Per view, the largest duct was identified and diameters were measured for four subsequent branching ducts to categorize the mode of arborization. For ducts with more than one furcation, the sequence was continued with the biggest duct ramifying from it. Terminal duct cells were assigned a duct diameter of 0 μm to indicate ending of the ductal tree. e, Cell division direction was measured on two-cell clones from tamoxifen-treated Rosa26LSL-Confetti;Hnf1bcreERT2 mice as the angle determined with respect to the line connecting both nuclei and the directionality of the duct (n = 263 clones, 5 mice). f, Direction of cell division and cell aspect ratio in ducts of varying diameter. Solid lines represent exponential fits. Cell division, n = 263 clones, 5 mice; aspect ratio, n = 120 ducts, 7 mice. Source Data

Extended Data Fig. 2 Exophytic and endophytic neoplasia in KrasG12D;Fbw7flox/flox model.

a, KFCk19 mouse model for tumour induction in the ductal epithelium triggered by Fbw7 exon deletion and KrasG12D activation. b, Recombination efficiency of low-dose tamoxifen injection and number of transformed clones per number of recombined cells per duct were quantified one week after tamoxifen injection in KFCk19 mice. EYFP-traced Krt19+ cells were quantified per duct, and total number of duct cells was estimated by dividing the duct length by the average cell length measured for this duct and multiplying this by the average number of circumscribing cells for this duct. Transformed clones were recognized as groups of more than three EYFP-traced cells sharing an interface. One dot represents one duct (112 ducts, 3 mice). c, d, KFH mouse model for alternative targeting of pancreatic ducts. d, Left, 3D rendering of exophytic neoplasia (top) and endophytic neoplasia (bottom) of KFH mice. Staining for Krt19 and tdTomato. Scale bars, 100 μm. Right, H & E staining for KFH exophytic (top) and endophytic (bottom) lesions. Scale bars, 100 μm. Representative of four mice. e, H & E and Alcian blue–periodic acid Schiff (AB/PAS) staining for exophytic and endophytic lesions in KFCk19 mice, demonstrating non-mucinous character typical of duct-derived neoplasia2. Scale bars, 100 μm. Representative of three mice. f, Experimental strategy for visualizing the connection of exophytic neoplasia with the ductal tree. The extrahepatic bile duct was cannulated at the ampulla, and the pancreas ductal tree perfused with 50 μl FITC-labelled dextran (DexFITC). g, Uptake of FITC-labelled dextran by an exophytic KFCk19 lesion, demonstrating lesion connection to the ductal system. Left, 3D view; right, optical section. Scale bars, 50 μm. Representative of four mice. Source Data

Extended Data Fig. 3 Morphology progression of acinar-derived neoplasia.

a, Schematic illustrating genetic strategy for acinar cell transformation by KrasG12D activation with concomitant p53flox/flox or Fbw7 flox/flox deletion, using Ela1–CreERT or Ptf1a–ERT2 drivers. b, c, KrasG12D;Fbw7 flox/flox;Ela1creERT (KFEla1) mice. b, Three-dimensional view of acinar-to-ductal metaplasia, as identified by local upregulation of Krt19 expression in acinar cells. tdTomato-traced acini connected to a terminal duct are shown. Arrowhead demarcates Krt19 expression by the middle acinus, forming a small ring of acinar-derived tdTomato-traced Krt19+ cells. Scale bar, 50 μm. Representative of three mice. c, Three-dimensional projection of a globular KFEla1 lesion in contact with a small-calibre duct (dotted line). Scale bar, 20 μm. c′, Optical sections of the lesion shown in c, demonstrating tdTomato tracing (top) and globular morphology (bottom). Scale bars, 20 μm. Representative of three mice. dg, KrasG12D; p53flox/flox;Ela1creERT mice (KPEla1 mice). d, Three-dimensional view of a globular tdTomato-traced KPEla1 lesion connected to a terminal duct (arrowhead). Scale bar, 50 μm. Representative of four mice. e, Three-dimensional projection of a large KPEla1 lesion, showing the central grape-like morphology of back-to-back globular structures and the maintained connection to several small-calibre ducts (arrowheads) at the lesion edge. Scale bar, 200 μm. e′, Higher magnification of area indicated in e (white dotted rectangle), demonstrating the seamless connection of acinar-derived Krt19+ cells and wild-type ductal epithelium (dotted lines). Scale bar, 30 μm. Representative of four mice. f, Retrograde perfusion of the ductal tree with FITC-labelled dextran, as in Extended Data Fig. 2f, demonstrating direct connection of acinar-derived lesions to the ductal system. Three-dimensional view of a KPEla1 lesion. Scale bar, 50 μm. Representative of three mice. g, H & E staining of a KPEla1 lesion demonstrating globular morphology. Scale bar, 100 μm. Representative of six mice. h, KrasG12D;p53flox/flox;Ptf1acreERT2 (KPPtf1a) mice. H & E staining, demonstrating globular morphology of lesions. Scale bar, 100 μm. Representative of five mice.

Extended Data Fig. 4 Exophytic and endophytic neoplasia in KrasG12D;p53flox/flox models and in human pancreas.

ac, Endophytic and exophytic lesions induced by p53 deletion with KrasG12D activation upon Pdx1–Cre-induced whole-pancreatic recombination (in KPC mice). b, Three-dimensional view of a pancreatic region of a three-week old mouse with endophytic (rectangle labelled 1) and exophytic (rectangle labelled 2) deformations. Scale bar, 150 μm. Panels labelled 1 and 2 provide higher-magnification images of the indicated areas in b. Scale bars, 50 μm. In panel 1, arrowheads demarcate invaginations typical of endophytic growth. In panel 2, a dotted line marks a morphologically normal small-calibre duct in contact with a globular, exophytic lesion. Representative of six mice. c, H & E staining for exophytic (left) and endophytic (right) lesions in the KPC model. Scale bars, 100 μm. Representative of six mice. d, e, Exophytic and endophytic lesions induced by p53 deletion with KrasG12D activation in pancreatic ducts (KPCk19 mice). e, Three-dimensional projection of exophytic (left) and endophytic (right) lesion shapes in KPCk19 mice. Scale bars, 100 μm. Representative of three mice. f, H & E staining of tissue sections from background pancreas of a patient presenting with pancreatic ductal adenocarcinoma. Left, exophytic lesion. Right, endophytic lesion. Scale bars, 100 μm. Representative of four patients.

Extended Data Fig. 5 Characterization of exophytic and endophytic neoplasia biology.

a, Proliferation of exophytic and endophytic lesions of KFCk19 mice as indicated by Ki67 staining. Scale bar, 100 μm. Representative of three mice. b, Stromal composition of KFCk19 exophytic and endophytic lesions, three weeks after recombination and recruitment of cancer-associated fibroblasts as demonstrated by staining for Pdgfrβ, nestin and smooth muscle actin (SMA). For staining comparisons between endophytic and exophytic lesions, lesions in the same tissue section are shown. Scale bars, 100 μm. Insets show magnified views of the area indicated in the main panel. Scale bars, 10 μm. Representative of four mice. c, Staining for human cancer-associated fibroblast markers PDGFRβ and SMA in background pancreatic tissue of a patient with pancreatic ductal adenocarcinoma. Examples of a non-dysplastic duct, an exophytic lesion and an endophytic lesion from the same tissue section are shown. Scale bars, 100 μm. Insets show magnified views of the area indicated in the main panel. Scale bars, 20 μm. One patient. d, Tumour cell epithelial-to-mesenchymal transition in advanced KFCk19 exophytic and endophytic lesions, eight weeks after recombination. Scale bars, 50 μm. Indicated areas are magnified on the right. Scale bars, 15 μm. Cdh1 staining marks epithelial cells, vimentin (Vim) staining indicates mesenchymal character and tdTomato identifies tumour-traced cells. Representative of three mice. e, Quantification of epithelial-to-mesenchymal transition in d as the number of tdTomato-traced, Cdh1, Vim+ cells per lesion area. Exophytic, n = 29 lesions; endophytic n = 26 lesions. Data are mean ± s.d., P < 0.0001 (two-sided Mann–Whitney test). Source Data

Extended Data Fig. 6 Cortical alterations in transformed cells.

a, Schematic showing cell orientation and cell cortex organization. b, Scatter plot of apical–basal pMLC2 intensities per cell, from endophytic and exophytic lesions in KFCk19 mice. Wild type, 106 cells; transformed, 124 cells. c, Basal and apical pMLC2 intensities were extracted from the maxima of the intensity profiles of each cell from exophytic lesions (Fig. 2i) (wild type, 35 cells; transformed, 45 cells). Data are mean ± s.d. NS, not significant (P = 0.057), P < 0.0001 (two-sided Mann–Whitney test). d, Ratio of apical to basal pMLC2 intensity in late-stage exophytic and endophytic lesions, eight weeks after recombination (in KFCk19 mice). Endophytic, 110 cells; exophytic, 138 cells. Data are mean ± s.d. NS, not significant (P = 0.9538) (two-sided Mann–Whitney test). e, MLC2 distribution in transformed and wild-type cells of KFCk19 mice. Left, immunofluorescence staining for MLC2, Cdh1 and EYFP. Scale bar, 10 μm. Right, Ratio of apical to basal MLC2 intensity. Wild type, 119 cells; transformed, 139 cells. Data are mean ± s.d., NS, not significant, P = 0.7722 (two-sided Mann–Whitney test). f, Cortactin distribution in transformed and wild-type cells of KFCk19 mice. Left, immunofluorescence staining for cortactin, Cdh1 and EYFP. Scale bar, 10 μm. Right, ratio of apical to basal cortactin intensity. Wild type, 152 cells; transformed, 137 cells. Data are mean ± s.d. P < 0.0001 (two-sided Mann–Whitney test). g, pMLC2 distribution in the KPC model. Left, immunofluorescence staining for pMLC2 and Krt19. Scale bar, 10 μm. Right, Basal–apical pMLC2 intensity profile, measured as shown in Fig. 2h. Single-cell profiles were normalized to cellular average and aligned in length. Data are mean ± s.e.m. (157 cells). h, F-actin distribution in the KPC model. Immunofluorescence staining for F-actin (phalloidin) and Cdh1 on cryo-sectioned pancreata from KPC mice or wild-type littermates. Scale bar, 10 μm. Quantifications are shown in Fig. 2k. Wild type, 4 mice; transformed, 3 mice. i, j, pMLC2 distribution in the KPPtf1a model of acinar cell transformation. i, Immunofluorescence staining for pMLC2, Krt19 and tdTomato. Scale bars, 10 μm. Left, normal duct; middle, normal Krt19-negative acinus; right, transformed Krt19-positive tdTomato-traced cells. Asterisk indicates lumen. Representative of three mice. j, Basal–apical pMLC2 intensity profiles, measured as shown in Fig. 2h. Single-cell profiles were normalized to cellular average and aligned in length. Data are mean ± s.e.m. (normal duct, 142 cells; normal acinus, 76 cells; transformed, 127 cells). k, Ratio of apical to basal F-actin, phosphorylated focal adhesion kinase (pFAK) and vinculin (Vnc) intensities in wild-type and transformed cells of KFCk19 mice. F-actin, 57 wild-type cells, 93 transformed cells; Vnc, 42 wild-type cells, 42 transformed cells; pFAK, 36 wild-type cells, 36 transformed cells. Data are mean ± s.d. P < 0.0001 (two-sided Mann–Whitney test). l, Expression of integrins in transformed cells relative to wild-type duct-cell neighbours in KFCk19 mice. Itga2, 57 wild-type cells, 57 transformed cells; Itga6, 33 wild-type cells, 33 transformed cells; Itgb1, 30 wild-type cells, 30 transformed cells. Data are mean ± s.d. P < 0.0001 (two-sided Mann–Whitney test). Source Data

Extended Data Fig. 7 Role of oncogenic Kras signalling in the pMLC2 distribution of transformed epithelial cells.

ad, pMLC2 distribution in wild-type organoids after phosphatase inhibition. ac, Immunostaining for pMLC2 and Cdh1. Scale bars, 10 μm. Representative of two experiments with three wild-type organoid lines. a, DMSO control. b, 2 μM okadaic acid treatment. c, 5 μM tautomycetin treatment. d, Ratio of apical to basal pMLC2 intensity of wild-type organoids after treatment with DMSO, okadaic acid (OA) or tautomycetin (TM). DMSO, 104 cells; okadaic acid, 70 cells; tautomycetin, 131 cells. Data are mean ± s.d. P < 0.0001 (two-sided Mann–Whitney test). e, pMLC2 distribution in KC mice. Immunostaining for pMLC2 and Cdh1. Scale bar, 10 μm. Representative of two mice. f, Basal–apical pMLC2 intensity profile of KC cells (e), measured as shown in Fig. 2h. Single-cell profiles were normalized to cellular average and aligned in length. Data are mean ± s.e.m. of 219 cells. gj, pMLC2 distribution in KPC organoids after MEK inhibition. g, h, Immunostaining for pMLC2 and Cdh1. Scale bars, 20 μm. Representative of three experiments with three KPC organoid lines. g, DMSO control. h, 8 μM U0126 treatment. i, Basal–apical pMLC2 intensity profiles of DMSO or U0126-treated organoids, measured as shown in Fig. 2h. Single-cell profiles were normalized to cellular average and aligned in length. Data are mean ± s.e.m. (DMSO, 105 cells; U0126, 133 cells). j, Ratio of apical to basal pMLC2 intensity of KPC organoids after DMSO or U0126 treatment. DMSO, 105 cells; U0126, 133 cells. Data are mean ± s.d. P < 0.0001 (two-sided Mann–Whitney test). k, l, Immunostaining for pMLC2 and Cdh1. Scale bars, 20 μm. k, DMSO control. l, 5 μM ROCK inhibitor (H1152) treatment. Representative of two experiments with three KPC organoid lines. m, Basal–apical pMLC2 intensity profiles of DMSO- or H1152-treated organoids, measured as shown in Fig. 2h. Single-cell profiles were normalized to cellular average and aligned in length. Data are mean ± s.e.m. (DMSO, 59 cells; H1152, 38 cells). n, Ratio of apical to basal pMLC2 intensity of KPC organoids after DMSO or H1152 treatment. DMSO, 60 cells; H1152, 37 cells. Data are mean ± s.d. NS, not significant, P = 0.5517 (two-sided Mann–Whitney test). Source Data

Extended Data Fig. 8 Formulation of ductal transformation in the 3D vertex model and the effect of tension change.

a, Number of circumscribing cells per duct diameter (113 ducts, 8 mice). b, Relative deformation as a function of diameter (mean ± s.e.m. from n = 10 simulations), for simulations with tension changes derived from exophytic (pink) and endophytic lesions (green). Both inputs produce deformations in good agreement with experimental observations (blue). c, Simulation of hyperproliferation without mechanical changes is not sufficient to explain the transition between exophytic and endophytic lesions (dots indicate mean ± s.e.m. from n = 10 simulations). Experimental data from Fig. 3d are shown for comparison. Simulations were carried out as in Fig. 3d, except that mechanical parameters in transformed cells were not modified. d, Classification accuracy, defined as the fraction of exophytic and endophytic lesions that can be predicted on the basis of duct diameter, as a function of duct diameter. The accuracy is defined for each diameter (d) as (TP + TN)/(TP + FP + TN + FN), in which TP (true positive) and FP (false positive) denote the number of exophytic and endophytic deformations occurring below diameter d, and TN (true negative) and FN (false negative) denote the number of endophytic and exophytic deformations occurring above diameter d. The diameter of maximal accuracy at three weeks is used to find the location of the transition. e, Immunofluorescence staining for pMLC2, Ki67, DBA and EYFP showing pMLC2 distribution in hyperproliferating cells and wild-type duct-cell neighbours of FCk19 mice. Scale bars, 10 μm. Representative of six mice. f, Basal–apical pMLC2 pixel intensity profiles of normal and hyperproliferating cells. Single-cell profiles were normalized to cellular average and aligned in length. Data are mean ± s.e.m. Twenty-eight ducts (6 mice), n = 64 Fbw7flox/flox cells, n = 93 wild-type cells. g, Immunofluorescence staining for ItgA2, Ki67, DBA and EYFP showing lack of basal integrin overexpression in hyperproliferating cells and wild-type duct-cell neighbours of FCk19 mice. Scale bars, 10 μm. Representative of three mice. h, Quantification of basal Itga2 localization in hyperproliferative cells relative to wild-type duct-cell neighbours. (n = 52 wild-type cells, 51 Fbw7flox/flox cells). Data are mean ± s.d., *P = 0.0274 (two-sided Mann–Whitney test). Source Data

Extended Data Fig. 9 Histological differentiation of benign and malignant ductal reactions.

a, Left to right, H & E stain of normal pancreas, FCk19 pancreas (hyperproliferative ducts), acute caerulein-induced pancreatitis and three-week-old KPC mice. b, Ki67 staining demonstrating proliferation in FCk19 ducts, ductal structures of acute pancreatitis and early transformation (in KPC mice). c, Staining for pMLC2 showing apical–basal redistribution of pMLC2 in transformed ductal lesions (in KPC mice) but not in normal duct cells (left), FCk19 ducts and reactive ducts. d, Itga2 is absent from the cellular membrane in normal duct cells (left) and FCk19 ducts and reactive ducts, but highly abundant in transformed ductal lesions (KPC). Scale bars, 100 μm (main panels), 20 μm (inset panels). All stainings are representative of six mice (normal pancreas), four mice (FCk19 and acute pancreatitis) or three mice (KPC).

Extended Data Fig. 10 Continuum theory of early lesion morphogenesis and pMLC2 distribution in lung airways and hepatic ducts.

a, We consider a continuum theory of tissue mechanics, in which the tissue is represented by a thin layer. The tissue initially has the shape of a cylinder, and we consider—for simplicity—deformations that are invariant along the longitudinal axis in cylindrical coordinates. The tissue has a bending elasticity and an area stretch elasticity that resists its deformation. In addition, the transformed region is subjected to a spontaneous bending-moment difference (\(\Delta {\bar{\zeta }}_{c}\)), compared to the wild-type tissue. The spontaneous bending moment arises from the difference between the tissue basal and apical surface tension (Tb and Ta, respectively). In the transformed tissue, the apical and basal surface tension (\({T}_{{\rm{a}}}^{{\rm{tf}}}\) and \({T}_{{\rm{b}}}^{{\rm{tf}}}\), respectively) differ from their values in the wild-type tissue (Ta and Tb). b, Two effects drive tissue deformation: tissue growth due to transformed cell division and growth leading to an increase in the radius of the cylindrical tissue, while the spontaneous bending-moment difference drives an inward invagination. The balance of these two effects on the transformed tissue indentation defines a threshold radius at which the indentation of the transformed tissue along the axis of symmetry vanishes. c, Calculated deformed tissue cross-sections after the introduction of a region of transformed cells, in the limit of small deformations. Blue line, deformed shape; red dotted line, original shape; the transformed cells are in the upper region. Nc is the number of cells in a tissue cross-section before cell transformation. Other parameters: \(\frac{\Delta {\bar{\zeta }}_{{\rm{c}}}{l}_{0}}{\kappa }=0.39\), Nt = 3 transformed cells in a cross-section. d, Phase diagram of direction of tumour deformation, in the limit of small deformations and small transformed region. R is the cylinder radius,\(\Delta {\bar{\zeta }}_{{\rm{c}}}\) the spontaneous bending-moment difference in the transformed tissue, and κ the bending modulus of the tissue. Schematic examples of tissue cross-sections undergoing inward (blue area) and outward (white area) deformations are shown. e, pMLC2 distribution in lung airways. Left, immunostaining for pMLC2 and Cdh1 in KFCk19 transformed and wild-type epithelium. Star marks pMLC2high myofibroblasts located below the epithelium. Scale bars, 20 μm. Right, Basal–apical pMLC2 intensity profiles, measured as shown in Fig. 2h. Single-cell profiles were normalized to cellular average and aligned in length. Data are mean ± s.e.m. (wild type, 167 cells; transformed, 66 cells). f, pMLC2 distribution in hepatic ducts. Left, immunostaining for pMLC2 and Cdh1 in KFCk19 transformed and wild-type epithelium. Scale bars, 20 μm. Right, Basal–apical pMLC2 intensity profiles, measured as shown in Fig. 2h. Single-cell profiles were normalized to cellular average and aligned in length. Data are mean ± s.e.m. (wild type, 134 cells; transformed, 284 cells). Source Data

Supplementary information

Supplementary Information

This file contains Supplementary Modelling Procedures. 3D vertex model for epithelial mechanics and a continuum theory of transformed epithelium deformation in a cylindrical tissue.

Reporting Summary

Supplementary Table

This file contains Supplementary Table 1: Human research participants. Description of patients undergoing surgery for pancreatic neoplasia, who kindly consented to donate their samples for this study.

Video 1: Normal pancreatic ductal system.

Detail of a tdTomato-stained (red) normal pancreas from a R26-tdTomato; Hnf1b-CreERt2 mouse showing intricate arborized ducts of varying calibres. Representative of 5 mice.

Video 2: Exophytic neoplasia.

Exophytic lesion of a KFCk19 pancreas stained for Krt19 (white) and EYFP (red). Representative of 7 mice.

Video 3: Endophytic neoplasia.

Endophytic lesion of a KFCk19 pancreas stained for Krt19 (white) and EYFP (red). Representative of 7 mice.

Video 4: KFE acinar-derived lesion.

Lesion of a KFE pancreas stained for Krt19 (white) and tdTomato (red). Note connection to a normal duct. Representative of 3 mice.

Video 5: KPE acinar-derived lesion.

Lesion of a KPE pancreas stained for Krt19 (white) and tdTomato (red). Note connection to a normal duct. Representative of 4 mice.

Source data

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Messal, H.A., Alt, S., Ferreira, R.M.M. et al. Tissue curvature and apicobasal mechanical tension imbalance instruct cancer morphogenesis. Nature 566, 126–130 (2019).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing