ErbB Receptor Stimulation Is Required for Mouse Colon Adenoma Organoids to Form Crypts


 The majority of colon adenomas harbor genetic mutations in the APC gene. APC mutation leads to changes in Wnt signalling and cell-cell adhesion: as a consequence, intestinal crypt budding increases and the excess crypts accumulate to form adenomas that progress to colon cancer. When cultured with Wnt, R-spondin, EGF, Noggin, myofibroblast conditioned medium and Matrigel, crypts from normal mouse colon mucosa form crypt-producing organoids and can be passaged continuously. Under the same culture and passage conditions, crypts isolated from colon adenomas derived from Apcmin/+ mice typically grow as spheroidal cysts and do not produce crypts. The adenoma organoid growth requires EGF, but not Wnt, R-spondin or Noggin. However, when mouse colon adenoma spheroids are grown for more than 10 days in the presence of EGF, crypt formation occurs. EGF, EREG, β-cellulin, Neuregulin-1 or AREG are sufficient for initiating crypt formation, however, neuregulin-1 is more potent than the other EGF-family members. EGFR and ErbB2 inhibitors both prevent crypt formation in adenoma cultures. Either EGFR:ErbB2 or ErbB3:ErbB2 signalling is sufficient to initiate adenoma crypt budding and elongation. ErbB2 inhibitors may provide a therapeutic avenue for controlling and ablating colon adenomas.


Introduction
The intestine is a highly regenerative organ with cells being replaced continuously. 50 The epithelium surface is made up of a sheet of epithelial cells folding into glandular 51 like crypts. Under normal homeostatic conditions, regeneration is maintained by the 52 production of new cells from actively cycling stem cells located at the base of each 53 crypt as well as the production of new crypts. During developmental growth or repair 54 of the intestinal epithelium, new crypts are produced by crypt budding 1,2 , where the 55 new crypt is initiated at the base of an existing crypt and elongates as the upper rim 56 migrates to the top of the crypt 3-5 . Aberrant crypt budding has been implicated as a 57 potential mechanism responsible for colorectal hyperplastic polyps 6 and linked to 58 adenoma formation 7 . These observations have been corroborated in the intestinal and 59 colonic epithelium of Apc min/+ mice during polyp formation 8 . 60 Intestinal epithelial cells can be cultured in vitro to produce multicellular three-61 dimensional structures called organoids 9-11 . The development of small intestine 11 and 62 colon 9,10,12 organoid cultures allows the direct study of both intra-crypt cell production 63 and crypt budding in organoids. Organoids can be used to measure the effects of 64 factors regulating crypt initiation and survival or intra-crypt cell production in normal, 65 adenomatous or cancerous colon mucosa 9,10,12 . 66 Previous reports show that factors such as epimorphin 13 , bone morphogenetic 67 proteins (BMPs) 14,15 , Wnt 16,17 , Epidermal Growth Factor (EGF) 18 , TGFβ 19 and 68 Hedgehog 16 can affect the formation of the villus-crypt structure in the small intestine. 69 Whereas mouse small intestinal organoid growth requires R-Spondin, Noggin and 70 EGF, but not Wnt, colon organoids require the addition of Wnt 12 . We have previously 71 reported our results with mouse colon organoids looking at the effect of the structural 72 environment and biochemical factors 9,10 on development of crypts in wildtype mouse 73 colon organoids. Our previous study suggests that crypt formation in mouse colon 74 organoids also requires the conditioned medium derived from mouse myo-fibroblast 75 cells (WEHI-YH2) 20 . In contrast, mouse colon adenomatous organoids grow and 76 passage weekly as spheres or cysts and have not been previously reported to form 77 crypts in vitro 21,22 . In this study, we report that whilst colon adenoma cells from Apc min/+ 78 mice initially form spheres, when cultured 10 or more days without splitting the 79 cultures, in the presence of EGF or selected members of the ErbB family of ligands 80 23,24 , adenomatous crypts will form. Our results provide insights into the mechanisms 81 initiating crypt production and the aberrant growth of adenomas and colon cancers. 82 83

Crypt formation is induced by EGF stimulation in long term colon adenoma 86 cultures 87
The normal mouse colon is lined with a single layer of polarised epithelial cells that 88 form the regular array of crypts which create the flat luminal surface (Fig. 1a). Colon 89 crypts were labelled with phalloidin to visualise F-actin (luminal-specific) and E-90 cadherin (baso-lateral specific). The high resolution 3D imaging reveals the highly 91 organised crypt structures and apical/basolateral polarity (Fig. 1a, Sub Fig. 1a,b). In 92 contrast to the normal colonic epithelial mucosa, adenomatous colonic polyps from 93 Apc min/+ mice display irregular crypt structure and disorganised packing of the crypts 94 i.e. the crypts no longer stack in perpendicularly to the mucosa surface (Fig. 1b, Sub 95 Fig. 1c,d). The crypts are still tightly packed, but the luminal axes are nolonger aligned. 96 Individual adenomatous crypts retain the apical/basolateral polarity of the epithelial 97 cells (Fig. 1b). 98 Cells from normal colon mucosa form colonospheres in matrigel cultures (with the full 99 complement of growth factors) and from Day 7 the colonospheres begin to produce 100 crypt buds (Fig. 1c). As previously shown, the addition of myofibroblast conditioned 101 media (WEHI-YH2) is required for optimal crypt production 9 . Most budding crypt 102 organoids produce multiple crypt structures formed from epithelial cells with typical 103 apical/basolateral polarity( Fig. 1c and inset, Fig. 1f). In contrast, cells from Apc min/+ 104 colon adenomas require only the addition of EGF in the basic culture media to 105 proliferate and form spheroids, (cyst-like) structures (Fig. 1d). The Apc min/+ adenomas 106 organoids grow similarly with or without the addition of Wnt, RSpondin and Noggin 22 . 107 The spheroids are readily visible by Day 6 and continue to grow but do not form crypt-108 like structures when in cultured and passaged for up to 10 days (Fig. 1d). Adenoma 109 cultures can be passaged and grown in long term culture, but require passaging after 110 7-10 days and continue to grow as spheroids. Under these conditions, in contrast to 111 the organoids from normal colon, Apc min/+ adenoma organoids do not appear to 112 produce crypts. This was surprising given that adenoma tissue contains tightly packed 113 and irregular crypt structures (see Fig. 1b) and suggested that the culture conditions 114 did not provide the appropriate micro-environemt to form adenomatous crypts. In order 115 to investigate adenoma organoid growth further, the Apc min/+ adenoma cultures were 116 allowed to grow for more than 7 days without mechanical disruption and in the 117 presence of EGF, but in the absence of R-spondin and Wnt3a. Under these conditions, 118 the organoids produced crypt-like structures from Day 12 (Fig. 1d). Budding crypt 119 structures were evident in a significant proportion of the organoids, similar to the 120 normal colon organoids (Fig. 1c, d). To ensure that the crypt formation phenomenon 121 presented by the current colon adenoma line was not an isolated event (i.e. specific 122 to a specific organoid line), adenoma cell lines derived from separate Apc min/+ 123 adenomas were tested and similar crypt-like structures were observed when cultured 124 in the presence of EGF up to 20 days (adenoma line #B13 shown in Supplementary 125 Fig. 2). Time-lapse imaging of another crypt forming organoid suggest that the crypt-126 like structures can occasionally start to appear after 7 days (Supplementary Movies 127 S1 and S2). Apc min/+ colon adenoma organoids require EGF for both growth of the 128 colonospheres and crypt-like structure production. The adenoma colonospheres grew 129 poorly without EGF (Fig. 1e, h). The Apc min/+ adenoma organoid crypt structures have 130 a similar morphology to cultures of normal colon organoids with similar dimensions 131 (eg. diameters of ~40 µm 5 ) but produce many more crypt structures per organoid ( Fig.  132 1c, d, f, g, i). High resolution 3D imaging of an immunostained Apc min/+ adenoma 133 organoid with multiple crypts reveals the structural organisation of the crypts with 134 strong peripheral staining of E-cadherin and β-catenin (Fig. 1i). 135

Effect of EGF on the formation of colon adenoma organoid crypts 136
The EGF signalling pathway is activated in the intestinal epithelial stem cell niche 25  and imaged daily over 21 days. Organoids remained as spheroids at 0.005 ng/ml EGF 142 but produced crypts between 0.05 -5.0 ng/ml EGF ( Fig. 2a-b). The proportion of crypt-143 containing organoids was scored and is presented in Fig. 2b. In the absence or at 144 lower concentrations of EGF the organoids remain as spherical cysts (Fig. 2a, b). The 145 crypt formation colon adenoma cultures increased with increasing EGF concentration, 146 from 0.05 ng/ml to 20 ng/mL (Fig. 2b) and was significantly increased above 0.5 ng/ml 147 (Fig. 2b). 148

Different morphological shapes of colon adenoma organnoids. 149
The colon Apc min/+ adenoma organoids have four three distinctive morphologies: 150 Spheroids (see Fig 3a up to day 8) , Large body organoids with crypt extensions (Fig  151  3a), Lenticular organoids (Fig 3b) and Hyperbudding organoids (Fig. 3c). The Large 152 body extension organoids had crypts forming after the organoid had grown to a 153 significant size. Crypts extensions then appeared from the main body, leading to the 154 gradual regressing/reduction of the size of the main organoid body. Lenticular 155 organoids started to form crypt extensions as they grew, reshaping the organoid into 156 a series of interconnected tubular crypts. These extensions could be very long ( > 100 157 µm , twice the maximum mean crypt width of normal murine crypt 5 ) with occasional 158 occurrence of secondary branches (Fig. 3b). Finally, Hyper-budding organoids 159 produced large numbers of small crypt extensions on the surface of the organoid 160 formed at the same time (Fig. 3c). The timing of crypt extensions formation is different 161 for each class of morphology, and varied between day 8 and day 14. For example, the 162 Hyper-budding organoids shown in Fig. 3c formed budging structures on day 11 and 163 day 15, respectively. 164

The elongation rate of crypts in adenoma organoids is independent of EGF 165 concentration. 166
We analysed the number and length of the adenoma crypt extensions over a range of 167 EGF concentrations (Fig. 4a).There was little crypt production at low concentrations 168 of EGF (5 pg/ml) but the number of crypts increased dramatically at 0.5 ng/ml (Fig.  169 4a). The number of crypt extensions per organoid increases with EGF concentration, 170 peaking at 0.5 ng/mL. The crypt lengths over the culture period of 10-25 days at three 171 EGF concentrations (50, 5, and 0.5 ng/mL) were measured ( Fig. 4b-d). The mean 172 length of crypts at 50 ng/mL EGF appears to be constant (~ 50 µm) between day 14 and day 24 (Fig. 4b).
In contrast, at lower concentrations of 5 ng/mL and 0.5 ng/mL 174 EGF the crypt lengths continued to increase throughout the time course (Fig. 4c, d,  175 respectively), at a rate of 2.24 µm/day and 2.03 µm/day respectively (Fig. 4e). 176 Collectively, this data shows that a minimal concentration of EGF is required for crypt 177 production, at EGF concentrations (i.e. between 0.5~5ng/mL), the crypt extends at a 178 rate of ~2 µm/day. At higher concentrations of EGF, crypt length rapidly reaches ~50 179 µm and does not extend any further 180

Effect of EGF-family ligands and the ErbB signaling pathway in colon adenoma 181 crypt formation. 182
Members of the epidermal growth factor receptor family include EGFR, ErbB2 (also 183 known as HER2), ErbB3/HER3, and ErbB4/HER4. EGFR and ErbB2 have been 184 associated with the growth of many human cancers, including colorectal cancer 28,29 185 (Fig.5a). Ligand binding to the extracellular domain of EGFR, ErbB3 or ErbB4 induces 186 the receptors to form oligomers and consequentially activate the intracellular kinase 30 . 187 ErbB2 does not bind a ligand but is in a conformation which allows it to bind to other 188 EGFR family members and when the co-receptor is in the ligand-activated 189 conformation, the ErbB2 kinase is activated 31 . Interestingly, the intracellular domain of 190 ErbB3 has no measurable kinase activity of its own, but when the ligand bound form 191 of ErbB3 combines with ErbB2, the heterodimer (or higher order aggregates of ErbB3 192 and ErbB2) activates the ErbB2 kinase activity. 193 To investigate which ErbB receptors might be driving colon adenoma crypt formation, formation with Heregulin /NRG-β1 being the most potent (80-fold more active than 202 EGF) (Fig. 5b). BTC and EGF showed a similar dosage range for stimulating the 203 formation of organoids with crypts; AREG and EREG both required higher 204 concentrations to stimulate crypt formation (Fig. 5b). Given that AREG binds to EGFR 205 but with a significantly lower affinity than that of EGF 34,35 , it is not surprising that a 206 higher concentration of AREG is required to stimulate the same level of crypt formation 207 in colon adenoma organoid assay (Fig.5b). Similarly, EREG binds to both the EGFR 208 (~2.8 µM) and ErbB4 (>5 µM) but with a significantly lower affinity than EGF binding 209 to these same receptors (1.9 nM and 49 nM, respectively) 36 , so higher concentrations 210 of EREG were required to stimulate crypt formation (Fig. 5b). 211 Although NRG-β1 (Heregulin-β1) has a high affinity for the ErbB3 and ErbB4 212 homodimers (IC50 ~5 nM), it has an even higher affinity for the ErbB3/ErbB2 213 Erb4/ErbB2 heterodimers (IC50 0.1-0.2 nM) 36 . EGF's affinity for EGFR homodimer 214 and the EGFR/ErbB2 heterodimer is between 1.2 to 1.8 nM. Heregulin-β1 stimulates 215 crypt formation more potently than EGF (Fig.5b, Supplementary Fig. 2). This result 216 indicates that a pathway stimulated by ErbB3/ErbB1, ErbB4/ErbB1, ErbB3/ErbB2 or 217 of their potencies for inducing colon adenoma crypt formation (Fig.5b). 223 The mean number of spheroids and organoids with crypts were also scored for each 224 of the EGF-like ligands (Fig.5c). NRG-β1 and BTC appeared to be the most potent in 225 terms of organoid formation with a consistently higher mean organoids formed, 226 particulary each ErbB family member in crypt formation. 237

261
This study demonstrates that EGFR/ErbB family signalling is required for crypt 262 production in organoids produced by mouse colon adenoma stem cells. Although 263 excess crypt formation appears to be the defining feature of adenomas, the current 264 paradigms for intestinal repair 37 , colon adenoma and colon cancer focuses on excess 265 intra-crypt cell proliferation 38 . Cell production within normal colon crypts occurs at a 266 rapid rate, indeed the rate of cell production in the intestines 39 (including the colon) 267 exceeds all other tissues. The renewal of the colon epithelial cell occurs rapidly and 268 continuously within the crypt, however under normal homeostatic conditions, crypt 269 production is a rare event: less than 1 in 200 crypts in the mouse colon are in the 270 process of budding 2,40 and in the normal human colon less than 1 in 2000 crypts are 271 producing Victoria, Australia), preparation and culturing of the WEHI-Ad67 colon adenoma 361 organoid line was described previously 9 . Using the same technique, we prepared 362 another colon adenoma organoid line: WEHI-AdB13 using a polyp from a separate 363 Apc +/min mouse. The colon adenoma organoid lines were passaged weekly (details as harvesting and preparation: The colon adenoma organoids were harvested from the 374 96 well plates, collected and pipetted with a 26G needle for 5-10 times, visually 375 checking the size of the resulting fragments after each round of pipetting (~10 cell 376 fragments are ideal to work with). The fragments were resuspended with DMEM/F12 377 and centrifuged at 8500 × g for 5 minutes. The supernatant fluid was discarded, and 378 the pellet resuspended with 1mL DMEM/F12. The suspension was centrifuged at 379 10000 × g for 3 minutes and supernatant fluid discarded. After resuspension of the 380 pellet with 150µL basic growth medium, an aliquot of the fragment suspension was 381 removed for counting using a haemocytometer (Hausser Scientific). 382 Based on the number of wells to be plated and required seeding density per well, the 383 volume and fragment concentrations of the seeding mixture were determined as 384 described in Sub Fig. 3a near the bottom of the well (Sub Fig. 3c). After dispensing the adenoma cells, the plate 397 was centrifuged for 1 minute at 450 RCF (Eppendorf 5810 R centrifuge) and incubated 398 at 37˚C for 1 hour (Sub Fig. 3d). 399 Reagent preparation and application: The respective EGF-like ligand stocks were 400 prepared in basic growth medium while the ErbB receptor inhibitors were made up in 401 basic growth medium containing recombinant mouse EGF (0.5 ng/mL, PeproTech, 402 #315-09). The study used a 5-point titration of the ligands using 1 in 10 dilutions. The 403 reagents were pipetted into the designated wells of a 96-well deep well plate 404 (Eppendorf cat#951033529) with layout as per described in Sub Fig. 3 Content Analysis version 5.11). For both systems, each experiment was followed and 433 imaged daily for up to 3 weeks. Each dataset which represents all the wells of a 434 specific timepoint was processed using a customized Fiji script 53 to extract each 435 individual well's image stack, before digitizing, saving as tiff images and organizing 436 into folders for analysis. 437 Computational image analysis for organoid imaging assay. For analysis, a 438 customised Fiji script was used to batch process and organise all the images in each 439 experiment, this includes steps to focus the multiple images at different focal planes 440 for each well, background and debris removal, as well as image segmentation to 441 extract organoids and for feature selection. Specific details of each processing steps 442 are briefly described below. 443 The individual well's Z-stack bright-field images were flattened using ImageJ/Fiji 444 software 53 . The projection of z-stack images onto a single in-focus image was carried 445 out using the Stack Focuser plugin (https://imagej.nih.gov/ij/plugins/stack-446 focuser.html) (parameters: select=10 variance=0.000 edge select_only) and exported 447 as high-quality tiff files. 448 Well exterior, background and debris removal as well as organoid selection and 449 feature extraction were performed were achieved using sequential steps of the 450 following operations: conversion to 8-bit, "Auto Local Threshold" (method=Phansalkar, 451 radius=150, parameter_1=0 parameter_2=0 white stack), set Black Background, 452 "Convert to Mask" (method=Default background=Default calculate black list), "Find 453 Edges" as stack, "Invert" stack, set Foreground Color as white, clear boundary (in 454 steps), "Dilate" as stack, "Fill Holes" as stack and "Remove Outliers" (radius=7 455 threshold=50 which=Dark stack). Analyses of object sizes performed using "Analyze 456 Particles" function (size=1000-Infinity circularity=0.30-1.00). 457 The script was used to processes all well images and used to threshold, segment and 458 count the individual organoids. The physical parameters of each organoid, including shape, size, intensity and circularity were recorded. These features for all organoids 460 identified per well/image were exported as text files (.csv) for analysis and tabulation 461 of the parameters of interest (e.g. mean organoid size). Overlayed images with 462 selection masks were exported for visual curation and validation. The results were 463 curated to remove false positives and to add missing organoids. 464 Whole mount staining and imaging. Colon tissue (normal and adenoma) was 465 cleared and stained as described in ref 54 . Briefly, the tissues were fixed, cleared and 466 stained with rat monoclonal anti-E-cadherin (Clone ECCD-2, Thermo Fisher Scientific 467 Cat #13-1900 (1:250)), goat anti-Rat IgG (H+L) Cross-Absorbed Secondary Antibody, 468 Alexa Fluor 488 (Thermo Fisher Scientific Cat #A-11006 (1:500)), Rhodamine 469 Phalloidin (Molecular Probes/Invitrogen(1:200)) and DAPI (Thermo Fisher Scientific 470 Cat# 62248 (1:1000)). For the adenoma tissues, 3D image stacks were acquired on a 471 Leica SP8 Resonance Scanning Confocal microscope using a 20x objective for 472 100µm depth at 0.5 µm per section. For normal tissues, 3D image stacks were 473 acquired on an Olympus IX-81 microscope with an Olympus FV1000 Spectral 474 Confocal attachment and a 20x objective for ~55 µm depth at 1 µm per section. All 475 acquired image data were processed and rendered using Imaris software package 476 (Bitplane, Zürich, Switzerland).