Rapidly derived colorectal cancer cultures recapitulate ...images.philips.com/is/content/PhilipsConsumer/Biocell...2016/11/17 · tem (Unisensor A/S, Lillerød, Denmark), located

Journal of PathologyJ Pathol 2014; 234: 34–45Published online 7 July 2014 in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/path.4371

ORIGINAL PAPER

Rapidly derived colorectal cancer cultures recapitulate parentalcancer characteristics and enable personalized therapeutic assays

Neil Ashley,* Matthew Jones, Djamila Ouaret, Jenny Wilding and Walter F Bodmer

Cancer and Immunogenetics Laboratory, Weatherall Institute of Molecular Medicine, Department of Oncology, University of Oxford, John RadcliffeHospital, Oxford, OX3 9DS, UK

*Correspondence to: Neil Ashley, Cancer and Immunogenetics Laboratory, Weatherall Institute of Molecular Medicine, Department of Oncology,University of Oxford, John Radcliffe Hospital, Oxford, OX3 9DS, UK. e-mail: [email protected]

AbstractWe have developed a simple procedure for deriving pure cultures of growing cancer cells from colorectal cancers,including material refrigerated overnight, for pathological characterization and cytotoxicity assays. Forty-sixcancers were processed and cultures set up under varying culture conditions. Use of a Rho kinase (ROCK1) inhibitormarkedly increased culture survival, resulting in 80% of samples growing in culture for at least 1 month andbeyond. Overnight refrigeration of samples before culture initiation had little effect on success rates, paving theway for cultures to be established for samples collected over wide geographical areas, such as those for clinicaltrials. Primary cultures demonstrated good correlation for differentiation markers compared to parent cancers, andwere highly dynamic in 3D culture. In Matrigel, many colonies formed central lumens, indicating the presence ofstem-like cells. Viable colonies in these cultures recapitulated the in vivo generation of carcinoembryonic antigen(CEA)-positive necrotic/apoptotic debris, much of which was derived from abnormal vacuolated dynamic ‘bubblecells’ that have not previously been described. Although bubble cells morphologically resembled signet ring cells, arare cancer subtype, immunostaining suggested that they were most likely derived from terminally differentiatedenterocytes. Micro-assays showed that drug toxicity could be measured in these cultures within hours and withsensitivity down to a few hundred cells. Primary cultures derived by our method provide valid in vitro avatarsfor studying the pathology of cancers in vitro and are amenable to pre-clinical drug testing, paving the way forpersonalized cancer treatment.Copyright © 2014 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Keywords: primary culture; colon; spheroid; CEA; cytokeratin 20; goblet cell; viability; colorectal cancer

Received 11 February 2014; Revised 9 April 2014; Accepted 29 April 2014

No conflicts of interest were declared.

Introduction

Colorectal cancers (CRCs) are one of the common-est forms of cancer, and incidence is rising. CRCs aregenetically heterogeneous, reflecting the somatic evo-lution of the cancer. Genome sequencing has increasedour understanding of the genetics of colorectal cancer,but validating how individual genotypes affect cancertreatment response requires appropriate in vitro mod-els. Such models could provide valuable pathologicaland genotype/phenotype information for individual can-cers and may pave the way for developing personalizedtreatments for cancer by enabling therapeutic testing onpatient material prior to clinical therapy. Efficient in vitroexpansion of human primary colorectal tissue in cultureremains technically challenging, and therefore murinexenografting is commonly used to propagate primarycancer material ex vivo. Whilst useful for research, thetime and expense of xenograft establishment precludetheir application to routine clinical pathology studies,and their routine use also raises ethical considerations.

There is therefore scope for a simple, cheaper, and moreconvenient procedure for ex vivo propagation of cancermaterial, more amenable to assay development.

In vitro primary culture typically involves cancerdissociation by mechanical and/or enzymatic means,followed by culture in serum-containing medium[1–6]. This approach is successful in about 10–15%of samples, but cultures can take many months beforeachieving significant growth, and this approach lacksconsistency. Spheroid cultures from colorectal cancershave been obtained with a reported success rate of47%, but these could only be propagated long term asxenografts [7,8]. Some studies have highlighted the util-ity of suspension culture of single cells in serum-freemedium for the establishment of primary colorectalcultures [4,9,10]. Conditions have also been describedthat allow the growth of human ‘organoids’ using bothnormal and diseased intestinal material [11–13]. Kondoet al developed a similar approach that uses enzymat-ically purified clusters of cancer cells, termed CTOS,rather than single cells [14]. Maintenance of cell–cell

Copyright © 2014 Pathological Society of Great Britain and Ireland. J Pathol 2014; 234: 34–45Published by John Wiley & Sons, Ltd. www.pathsoc.org.uk www.thejournalofpathology.com

Primary colon cancer culture 35

contacts prevented the apoptotic death that occurredif cancer material is dissociated into single cells.Whilst these new approaches are very valuable for basicresearch, they are generally expensive, time-consuming,and can be technically challenging.

In this report we introduce a new approach for simple,convenient, efficient, and cost-effective establishment ofprimary cultures from colorectal cancers, which can beapplied to refrigerated stored material. We also describenovel approaches to testing this patient-derived materialfor drug sensitivity.

Materials and methods

Antibodies and reagentsChemicals were from Sigma-Aldrich, Dorset, UK.Rabbit anti-cytokeratin 20 (clone EPR1622Y), rabbitanti-CDX2 (clone EPR2764Y), and anti-CEA (cloneCB30) were from (Abcam Ltd, Milton, Cambridge,UK). Mouse anti-EpCAM (AUA1), CAM5.2 (predom-inantly anti-cytokeratin 8), and anti-goblet cell mucin(clone PR5D5) were raised in our laboratory [15–18].Anti-mouse Dylight 488 was from Vector Labs, Peter-borough, UK; highly cross-absorbed anti-rabbit Alex-afluor 555, PrestoBlue, and SytoxBlue were from LifeTechnologies Inc, Grand Island, NY, USA. All antibod-ies were used at 1 : 100 dilution except cytokeratin 20(1 : 600).

Primary culture conditionsSamples were obtained with informed consent andthe project was approved by the local ethics commit-tee (07/H0606/120). Following resection, the cancerand surrounding tissue were placed into a large con-tainer. A pathologist then cleaned the specimen withsaline and dissected a small piece of the tumour into5 ml of ‘storage medium’ [Hibernate-A medium, 0.1mM 2-mercaptoethanol, 1× Glutamax (Life Technolo-gies Inc), 10 μg/ml gentomycin/neomycin , 0.25 μg/mlamphotericin , 100 U/ml penicillin, 100 μg/ml strep-tomycin (Lonza Inc, Slough, UK)]. From surgery tomedium took 1–2 h and samples were refrigerated for15–17 h. Sample size was ∼5 mm2 to 1 cm2. Serum-freeDulbecco’s modified Eagle medium (DMEM) was alsoused successfully.

Cancers were transferred to a 50 ml Falcon tube con-taining Hank’s balanced salt solution (HBSS) [no cal-cium, no magnesium, no phenol red, with 10 mM Hepes(N-2-hydroxyethylpiperazine-N-2-ethanesulphonicacid)] (pH 7.1), shaken by hand, and then centrifuged at400 relative centrifugal force (RCF) to pellet materialand the HBSS was removed (repeated three times). Thecancer was placed in a 10 cm Petri dish in a drop ofDMEM with 10% v/v fetal bovine serum (Source Bio-Science, Nottingham, UK) and dissected into smallerpieces (∼4 mm diameter). The dissected cancer wastransferred to a sterile 5 ml screw-top Bijou tube (VWR

Ltd, Lutterworth, Leics, UK) that was half-filled with∼2.5 ml of glass beads (4 mm, undrilled; Fisher Sci-entific UK Ltd, Loughborough, UK), which was filledto the top with DMEM and vigorously hand-shakenuntil the medium turned cloudy. The suspension wasremoved to a 50 ml Falcon tube (Corning Inc, Corning,NY, USA) using a Pasteur pipette, leaving larger cancerpieces and glass beads. The Bijou tube was then refilledwith DMEM and shaken again. Supernatants wereremoved and pooled with the previous fraction. Thisprocess was repeated until no more cells were liberated.

The pooled cell suspension was filtered into 15 ml Fal-con tubes through a 250 μm nylon mesh (Pierce Biotech-nology Inc, Rockford, IL, USA). The filtered suspensionwas then centrifuged at an average of 162 RCF for 4min to pellet cells. An optional gravity sedimentationstep was sometimes included by resuspending the pelletin 5 ml of medium and leaving the suspension to set-tle for 10 min before discarding the top three-quartersof fluid and then re-centrifuging the suspension asabove. The supernatant was removed and the remain-ing pellet was resuspended in 5–20 ml of ‘primarymedium’, consisting of Excell 620 low protein medium(Sigma-Aldrich UK Ltd), with 2% Stempro hESC sup-plement (v/v), 10 ng/ml epithelial growth factor, 100U/ml penicillin, 100 μg/ml streptomycin, and 1× Gluta-max, 1 mM 2-mercaptoethanol (Life Technologies Inc),100 ng/ml fibroblast growth factor 10 (ProSpec-TanyTechnoGene Ltd, Rehovot, Israel), 30 μM endothelin3 (Lonza Inc), 10 μg/ml gentomycin/neomycin, 0.25μg/ml amphotericin (Lonza Inc), and 10 μM ROCKinhibitor Y-27632 (Selleck Chemicals Inc). For someparallel experiments, 2.5 μM of the ROCK1 inhibitorThiazovivin was used instead of Y-2763.

150 μl/well cell suspension was distributed into a96-well cell suspension culture plate (SARSTEDT Ltd,Leicester, UK). Multiple wells minimized microbialcontamination. Low attachment conditions inhibitedstromal contamination (data not shown). Cultures weremaintained in the 96-well plate for 7 days and uncon-taminated wells pooled into an UltraCruz low attach-ment flask (Santa Cruz Biotechnology Inc, Santa Cruz,CA, USA) with fresh primary medium. Spent mediumwas replaced with fresh medium every 4 days until theculture started growing and then every 7 days there-after. ROCK inhibitors were omitted from the primarymedium after 21 days. Mycoplasma contamination wastested for using the MycoAlert kit (Lonza Inc; all cul-tures tested negative). When colonies reached a densityof about 100 per ml, aliquots were removed and frozenin cryogenic media in liquid nitrogen.

Colony purification into colony- and single-celldebris-enriched fractionsSpheroid colonies were sorted from single cells anddebris by passing primary cultures through a 40 μmnylon mesh (Fisher Scientific Inc) that retained most ofthe spheroids and loose cell colonies. The flow throughcontained the single cells and debris. The filter was


36 N Ashley et al

washed by placing in a 5 cm dish with fresh primarymedium and gently shaken, and then transferred to afresh dish with medium; captured colonies were gentlyremoved using a Pasteur pipette and placed into a tube.

Viability assaysFor microscope viability assays, cultures were incu-bated with medium containing the live cell-impermeablered fluorescent dye propidium iodide (3 μM) and cal-cein AM (1 μM) for 20 min at 37∘C. Calcein AM isa non-fluorescent membrane-permeable probe that ishydrolysed by cellular esterases to form a green flu-orescent membrane-impermeable compound. With thisdye combination, dead cells are shown as red and viablecells are shown as green. For microplate-based viabilityassays, PrestoBlue/SytoxBlue assays were performedon filter-purified colonies, as initial experiments indi-cated a higher sensitivity than calcein/propidium iodide.PrestoBlue is a cell-permeable resazurin-based solutionthat is modified by the reducing environment of viablecells becoming highly fluorescent. SytoxBlue is a livecell-impermeable DNA stain that becomes highly fluo-rescent upon binding to the DNA of nectrotic/apoptoticcells with compromised membranes. Approximatelyfive colonies per well were seeded in triplicate into384-well plates, and medium with drug or vehicle wasadded to a final volume of 60 μl. The plates werethen incubated for the specified times. Stock SytoxBluewas pre-diluted 100 times in the stock 10× PrestoBluereagent and 6 μl of this 10× solution was added to eachwell. The plates were mixed on a rotary plate shaker(900 rpm) and incubated for 30 min at 37∘C in theincubator. The plates were then read using a MolecularDevices SpectraMax M4 Fluorometer using 440/480 nmand 560/590 nm excitation/emission for SytoxBlue andPrestoBlue, respectively.

ImmunofluorescenceImmunofluorescence labelling was carried out asdescribed previously [19].

RNA isolation and cDNA synthesis, and quantitativereal-time PCR (qPCR)Total RNA was extracted from primary cultures usingQiagen RNeasy Spin Columns according to the man-ufacturer’s methods. For CDX1 qRT-PCR, 1 μg oftotal RNA per sample was reverse-transcribed usingthe Life Technologies High Capacity cDNA Synthesiskit with random hexameric primers according to themanufacturer’s instructions. cDNA concentration wasmeasured with a NanoDrop spectrophotometer anddiluted to 25 ng/μl in Tris-EDTA buffer. 100 ng ofcDNA per sample was used as input for TaqMan quanti-tative PCR using pre-designed, manufacturer-validatedprimers and probes for CDX1 (Life Technologies,Hs00156451_m1) and ubiquitin C (Hs01871556_s1).Relative CDX1 expression levels were calculatedaccording to the ΔΔCt method, with UBC as

endogenous control. Fold changes were determinedrelative to the average Ct across all samples. All qPCRreactions were performed in quadruplicate.

Real-time imaging of primary cancer culturesand image analysisPrimary cultures were imaged on 96-well platesusing the oCelloScopeTM microscope scanning sys-tem (Unisensor A/S, Lillerød, Denmark), located ina standard cell culture incubator and maintained at37∘C. For non-attached suspension culture imaging,colonies were suspended on non-attachment plasticin primary medium. For Matrigel-embedded colonyimaging, colonies were first embedded in 50% v : vMatrigel mixed with Excell 620 medium supplementedwith 10% FCS. After the Matrigel had gelled after 20min, the wells were filled with Excell 620 mediumsupplemented with 10% FCS.

Image analysis of spheroid movement in cap-tured movies was done manually on avi movie filesusing MTrackJ plugin (http://www.imagescience.org/meijering/software/mtrackj/) for ImageJ (Rasband, WS,ImageJ, US National Institutes of Health, Bethesda,MD, USA, http://imagej.nih.gov/ij/, 1997–2012). Fivecolonies per condition were measured for morphologyor movement quantitation.

Results

Establishment of culturesWe first tested whether agitation with glass beads coulddissociate cancer material without excessive disruptionof cell–cell contacts. A resected cancer specimen wasloaded with glass beads and vigorously shaken andthe supernatant examined. Typical results for normalcolonic tissue and for cancer tissue are shown respec-tively in Figures 1A–1C. Many intact crypts were lib-erated from the normal tissue (Figure 1A) together withsingle blood cells and fibroblasts. Dysplastic crypts andglandular structures were liberated from cancer materialtogether with single cells (Figures 1B and 1C). Centrifu-gation and filtration enabled the removal of most sin-gle cells. Loose cell clusters and intact crypt-like struc-tures were resuspended in primary culture medium sup-plemented with or without ROCK1 inhibitor. This sus-pension was then deposited onto low attachment plates,which have been suggested to improve the isolation ofstem-like cells from primary tissues [9,20–22]. Dur-ing overnight culture, we observed the formation ofphase-bright well-formed spheroidal colonies, number-ing from 3 to 300 per well, scattered amongst necroticdebris, as shown in Figure 1D and Movie 1. Cells inmany colonies appeared polarized. Cystic ‘bubble-like’cells with large distended vacuoles were also frequentlyobserved, and Hoechst DNA staining revealed that thesecells were nucleated (Figure 1E). Bubble cells werepresent in all primary cultures. Most cancers formed



Figure 1. Establishment of primary cultures. (A) Phase contrast of normal crypts derived from normal colon tissue by glass bead disruption.(B) Abnormal crypt-like structures derived by the same approach directly from a colorectal adenocarcinoma. (C) Close-up image of acancer-derived crypt-like structure. (D) Cancer spheroids and surrounding debris following 24 h culture. (E) Phase contrast image of primaryculture with cystic ‘bubble’ cells (arrowed), amongst single cell debris. White: Hoechst nuclear staining; nuclei in bubble cells are arrowed.A spheroid colony is marked by an asterisk. (F) Dark field image of primary C2661 showing loose cell material stuck in ‘jelly’ surroundingeach colony. (G–I) Images of a primary colony (C1999) suspended in Matrigel grown for 1, 2, and 3 months, respectively.

phase-bright spheroidal colonies, but culture C2661 wasunusual in producing thick mucinous jelly around eachcolony in which single-celled debris became trapped(Figure 1F).

When colonies from a primary culture were immobi-lized in Matrigel, an extracellular matrix that enables 3Dgrowth, they were observed to grow into large complexcolonies (Figures 1G–1I). Time-lapse imaging of sus-pension cultures showed highly dynamic fluctuations ofcolonies and bubble cells, which rapidly expanded andcontracted (Movie 2). Movie 3 shows that colonies werethe source of bubble cells, some of which could be seento expand and contract within the colony mass.

The ROCK1 inhibitor Y-2763 is routinely used inembryonic stem cell culture to reduce cell death [23],and also benefits the survival of dissociated mouse smallintestinal cells [24]. Figure 2 shows that without Y-2763,while virtually all samples yielded spheroid colonies fol-lowing overnight culture, two-thirds died after 1 weekand only 29% survived to 1 month. Y-2763 greatlyincreased the success rate of maintaining long-term cul-tures, with 82% of samples growing as cultures for atleast 1 month (most growing indefinitely). Colony dou-bling times ranged from 2 weeks to 3 months (Sup-plementary Table 1). Three parallel experiments using

the alternative ROCK1 inhibitor, Thiazovivin, also pro-moted surviving colonies (data not shown). Further-more, use of ROCK1 inhibitors enabled a high efficiencyof establishment of cultures from refrigerated sam-ples (Figure 2A). Although most of the Y-2763-treatedsamples were refrigerated, four samples that were notrefrigerated at all gave successful cultures. Furthermore,of four tumours that were divided into two, one ofwhich was refrigerated and the other not refrigerated,all samples gave rise to cultures. Thus, eight of eightsamples not refrigerated but incubated with ROCK1inhibitor gave rise to successful cultures. We concludethat ROCK1 inhibition, rather than refrigeration, is thefactor for increased success.

Characterization of primary culture differentiationThe epithelial origin of the spheroid colonies was con-firmed by immunofluorescence positive staining for theepithelial markers (Figure 3A) and antibody CAM5.2,and no positive staining for CD45 and vimentin (datanot shown). The gut transcription factor CDX2 wasexpressed by all the cultures tested, and also by all theparental cancers, except for C2156 (Figure 3A). CDX1mRNA was also expressed (Figure 3B). Immunostaining


38 N Ashley et al

Samples Cultured with ROCK1 Inhibitor (Refrigerated +Non-Refrigerated)

Samples Cultured without ROCK1 Inhibitor (Refrigerated +Non-Refrigerated)

Samples Refrigerated and Cultured with ROCK1 Inhibitor

Samples Refrigerated and Cultured without ROCK1Inhibitor

Samples Not Refrigerated and Cultured with ROCK1Inhibitor

Time in Culture

% Successful Long-term Culture1 Month

50

83.3

28.6

85.7

100

0

23

15

8

12

8

0

29

16

13

12

8

0

41

18

21

13

8

0

4

46

18

28

14

8

2

1 Week48 Hour0 Hours

Total Samples Processed (not including 4 duplicates)

Samples Cultured with ROCK1 Inhibitor (Refrigerated + Non-Refrigerated)

Samples Cultured without ROCK1 Inhibitor (Refrigerated +Non-Refrigerated)

Samples Refrigerated and Cultured with ROCK1 Inhibitor

Samples Not Refrigerated and Cultured with ROCK1 Inhibitor*

Samples Refrigerated and Cultured without ROCK1 Inhibitor

Samples Lost to Contamination (from 46 Samples)

1 Month1 Week48 Hour0 Hours0

5

10

15

Num

ber

of V

iabl

e C

ultu

res 20

25

30A

B

Figure 2. Culture efficiency. (A) Graph of the success rate of primary culture with or without ROCK1 inhibitor Y-27632. Forty-six sampleswere processed in total from 45 patients (one ovarian metastasis in addition to the primary cancer). Four samples were lost to contamination.Eighteen samples were set up in the presence of Y-27632 and 15 of these gave rise to cultures that survived to the long term. Y-27632significantly increased successful cultures versus unsuccessful cultures at 1 month (p = 0.0001, 2 × 2 Fisher exact test). Regression analysisshowed that there was no longer a significant reduction in viable cultures with time when Y-27632 was used (p = 0.0533) compared withwithout Y-27632 (p = 0.0039). Fourteen of the Y-27632-treated samples were also refrigerated, but four of four duplicate non-refrigeratedsamples were also successful. (B) Table of the culture conditions used in this study. Though the numbers are small, with Y-27632, overnightrefrigeration of samples at 4∘C had no significant effect on culture success. *Includes four non-refrigerated ROCK1 duplicates of samplesthat were refrigerated and cultured with ROCK1 inhibitor.

showed robust expression of cytokeratin 20, a down-stream target of CDX1 (Figure 3C). It should be notedthat bubble cells frequently showed very high cytok-eratin 20 staining (eg C2284 panel in Figure 3C andFigure 5C).

During culture, single cells and debris were gen-erated in the primary cultures even after removal byfiltration. To investigate this, we separated spheroidsfrom this smaller debris by capturing them on a 40μm nylon mesh to generate captured colony-enrichedand flow through, loose cell-enriched fractions. Wethen resuspended the colony-enriched fraction in fresh

medium and cultured the purified colonies for a fur-ther 14 days. Over this time, necrotic single cells anddebris were regenerated in the colony-enriched frac-tion (Figure 4A). In Movie 4, we were able to observesingle cells and smaller vesicles being extruded fromcolonies that were immobilized within Matrigel. Manyextruded cells died, and similar events were observedfor suspension colonies (data not shown). We also fre-quently observed that extruded bubble cells appearedto die after extensive expansion (Movie 5). Most ofthe extruded material was static by time lapse, indicat-ing that it was dead. To confirm viability, colony- and



Figure 3. Primary spheroids are composed of EpCAM-positive cells that express the gut homeobox transcription factors CDX1 and CDX2and differentiation marker cytokeratin 20. (A) EpCAM (green)/CDX2 (red) co-immunostaining of six primary parental cancers and daughterparental cancer colonies. (B) Quantitative real-time PCR analysis of CDX1 mRNA derived from five primary spheroid cultures. Error bars =SD of mean. Note: C2156 did not produce sufficient quality RNA for PCR analysis. (C) Cytokeratin 20 (red) immunostaining of six primaryspheroid cultures. DAPI-stained nuclei are blue.

loose cell-enriched fractions were stained with propid-ium iodide/calcein AM. Cells in the spheroid colonieswere virtually all viable (Figure 4B), while the majorityof the single cells were necrotic. Bubble cells within theloose cell fraction were the only significant non-colonycells that retained viability. Quantitation of the liveand dead cells in each fraction confirmed the predom-inance of dead cells in the loose cell-enriched fraction(Figure 4C).

Figure 4D (upper panels) shows that colonies weregenerally negative for the apoptotic marker CytodeathM30 (detecting caspase-cleaved human cytokeratin 18),

whereas contaminating single cells and much of thesingle cell-enriched fraction were stronglyM30-positive, which was confirmed by quantifica-tion (Figure 4E). Anti-Ki67 labelling, a marker ofcycling cells, showed that most replicating cells werepresent in colonies and very few were present in theloose cell fraction (Figure 4D, lower panels), confirmedby quantification (Figure 4F).

To determine if primary cultures retained the differ-entiation characteristics of their parent cancers, we firstexamined the presence of goblet cells in primaries andparent cancers by immunostaining with the anti-goblet


40 N Ashley et al

Figure 4. Primary spheroids spontaneously generate necrotic and apoptotic cellular debris. (A) C2284 spheroid colonies derived from cancerswere separated from single cells by size filtration and resuspended in medium in a 96-well plate. After 14 days’ culture, substantial numbersof single cells were regenerated. (B) Calcein-AM (green)/propidium iodide (red) staining of colony-enriched fraction (upper panels) orloose cell-enriched fraction (lower panels). Living cells are green; dead cells are red. (C) Quantitation of dead cells of colony- and loosecell-enriched fractions from the same experiment. (D) M30 cytodeath labelling of apoptotic cells (upper panels) or Ki67 labelling of dividingcells (lower panels in colony- and loose cell-enriched fractions). (E) Quantitation of data in F. Error bars = SD of mean.

cell mucin antibody PR5D5. Figure 5A shows thatof the six parental cancers examined, five containedvarying numbers of goblet cells. Two of the parent can-cers, C2156 and C2661, had very extensive goblet cellexpression, which was also seen in the primary cultures.Only for the C3953 sample was there no goblet cell

differentiation either in the parental cancer or in thederived primary culture. The parent of C4054 had onlyrare and isolated goblet cells, while no goblet cellswere observed in the daughter culture. Quantitation ofgoblet cells in the primary cultures and parental cancersshowed that they were closely correlated (Figure 5B).



Figure 5. Primary cultures recapitulate the goblet cell differentiation of parent cancers and bubble cells express high levels of cytokeratin20 and CEA. (A) Primary cultures and parent cancer sections were stained with PR5D5 (green), specific for goblet cell mucin. Blue is DAPIfor nuclei. (B) Table showing the relative proportions of goblet cell-positive colonies in primary cultures and positive glands in parentalcancers (20 counts each). (C) Phase contrast with red cytokeratin 20 immunostaining of bubble cells (arrowed) from the single cell-enrichedfraction embedded in FFPE. (D) Cytokeratin 20 (red) and (E) CEA (green) immunostaining of colony-enriched fraction, with bubble cellsstaining strongly for cytokeratin and bubble and loose cells staining strongly for CEA. (F) Merge of D and E. Blue is DAPI stained nuclei.(G) Quantitation of bubble cell production after 1 week growth of primary colonies under adherent or non-adherent conditions in Excell620 medium with 10% FCS. Cultures were divided equally into two fractions and maintained as suspension spheroids or monolayers byculturing on non-adhesive and cell culture-treated flasks, respectively. Approximately 20 colonies were present in each flask.

Because bubble cells were unusual and have notbeen described before, we examined their differentia-tion status using antibodies to cytokeratin 20 and the cellsurface glycoprotein carcinoembryonic antigen (CEA),both of which are highly expressed in differentiatedcells at the top of normal colonic crypts [25]. Bub-ble cells expressed higher levels of both cytokeratin 20and CEA (Figures 5C–5F) than colonies, suggestingthat they were highly differentiated cells. CEA was alsohighly expressed in the non-bubble single cells, sug-gesting a common origin (Figure 5E). To determine ifbubble cells were an artefact of the culture conditions,three colony-enriched, filtered primary cultures were

divided into two equal fractions each and the mediumwas replaced with standard serum containing DMEM.One fraction was kept on non-adherent plastic and theother was allowed to grow on normal cell culture plastic,to which the colonies attached and grew as a conven-tional cell monolayer culture. After a week’s growth,bubble cells in the medium from the non-adherent plas-tic and those attached to the adherent plastic werecounted. There was a marked down-regulation of bub-ble cell production in the cultures that had been grownas attached monolayers, compared with cultures kept ascolony suspensions (Figure 5G). Thus, abnormal bub-ble cells appear to be derived from well-differentiated


42 N Ashley et al

cells shed from colonies when grown under 3D condi-tions, but not from monolayer cultures, irrespective ofthe medium used.

We have previously shown that single cells from cer-tain established colorectal cancer cell lines can formlumen-containing colonies when cultured in Matrigeland that this property can be used to identify cancerstem cells [19,26]. Time-lapse imaging of primary cul-tures embedded in Matrigel (Movie 6) shows that theycould form lumen colonies, an indicator of the presenceof stem-like cells capable of differentiating.

Primary cultures for drug testingAn obvious application of primary cultures is as cancersurrogates for drug testing. Figure 6A and Movie 7show that treatment of suspension colonies with theapoptosis inducer staurosporine led to a deteriorationof colony size and integrity with an accumulation ofstatic, dead material around the shrinking colony. Asimilar loss of colony integrity was observed for avariety of other cytotoxic drugs, including irinoticanand camptothecin, tested on a range of primary cultures(results not shown). To quantitatively measure this celldeath, we developed a convenient one-step no-washfluorescent assay to measure the ratio of live to deadcells in suspension spheroid cultures. Use of ratio mea-surements allowed assays to be set up independentlyof total cell numbers, as primaries showed reducedviability when dissociated for cell counting. PrestoBlue,a resazurin-based metabolic reagent, was found togive a near-linear fluorescent response to increasingnumbers of spheroids (Figure 6B). SytoxBlue, a livecell-impermeable DNA stain, was used to measuredead cells and gave a linear response to increasingamounts of mostly dead single cell-enriched fractions(Figure 6C). Saponin detergent treatment of about fiveC2284 suspension colonies per well (approximately 400cells per well) showed a massive increase in the ratio ofSytoxBlue to PrestoBlue fluorescence (Figure 6D). Wefurther validated this assay by measuring death inducedby increasing concentrations of staurosporine and thechemotherapy drugs 5-fluorouracil and camptothecin(Figures 6E, 6F, and 6G, respectively). The cytotoxicityresults are broadly consistent with the response rangeof a panel of colorectal cell lines to 5-fluorouracil [27]and camptothecin [28].

We next explored if quantitative toxicity data could bederived from time-lapse images of drug-treated colonies.Quantitative analysis of sequential movie stills showedthat the average colony size decreased markedly withcamptothecin treatment, correlating with time and drugdose exposure (Figure 6H). Time-lapse images alsodemonstrated that the characteristic dynamic move-ment of healthy colonies was reduced by cytotoxicdrugs and that necrotic material was essentially static(see all movies for examples). We used ImageJ tomeasure colony movement quantitatively (Figure 6I).Colonies treated with the detergent saponin showedzero movement over a long period, confirming that cell

death impairs colony movement (Figure 6I, last bar onright). Figure 6I shows a near-linear drop in averagecolony movement relative to increasing amounts of stau-rosporine, and Figure 6J shows that plotting the samedata as total cumulative colony movement showed amarked inhibition of movement during drug treatmentthat became more apparent with time. The effect onmovement is detectable within 10–13 h of treatment,suggesting the possibility of rapid drug treatment assaysthat can be applied even to single colonies.

Discussion

We have introduced a simple, cost-effective, and effi-cient technique for propagating fresh, or short-termrefrigerated, primary cancer material. The tech-nique involves growing the cultures in suspensionin serum-free medium supplemented with a humanembryonic stem cell-promoting growth supplement.Matrigel was avoided as it is costly, difficult to manip-ulate, and can promote differentiation of colorectalcells [19]. Simple mechanical disruption using glassbeads gave a high yield of intact gland-like structuresfrom cancers, greatly simplifying sample processing.ROCK1 inhibitor Y27632, which is commonly used inembryonic stem cell cultures to reduce cell death [23],greatly increased the culture success rate. Inhibition ofROCK1 prevents activation of the Rho pathway, which,in turn, may prevent initiation of apoptosis [29].

Comparison of the phenotypes of primary cultureswith parental cancers suggested that differentiation waswell maintained during the transition into in vitro cul-ture. CDX2, which is involved in the control of secretorycell differentiation [30], was similarly expressed in pri-mary cultures and parent cancers, except for the C2156culture whose CDX2 expression was much higher thanits parent cancer. There was a good correlation betweenthe frequency of goblet cells in parental cancers anddaughter cultures. C2661 produced thick mucinousmaterial around every colony, consistent with its highexpression of goblet cells.

Primary cultures generated large numbers of necroticand apoptotic cells and debris. This material is mostprobably derived from terminally differentiated cells ina manner reminiscent of the shedding of differentiatedcells from the normal colonic crypt following enterocytematuration [31]. This is consistent with high CEA stain-ing that we observed for this material. The ‘bubble’ cells,which we have uniquely identified in these cultures,probably contribute significantly to this dead debris.Nevertheless, many bubble cells survived for long peri-ods. As bubble cells were generated in cultures main-tained in 10% serum containing DMEM as suspensionsbut not as monolayers, their presence is not a mediumartefact, but rather a result of their growth in 3D. Thebubble cells stained more strongly for the differentiationmarkers cytokeratin 20 and CEA than the ordinary cellsin the colonies, and did not stain for goblet cell markers,



Figure 6. Small-scale viability assays for drug studies. (A) C2284 spheroid treated with 2 μM staurosporine for 44 h. (B) Standard curveof PrestoBlue fluorescence using increasing amounts of a C2284 spheroid suspension (approximately 1 colony/μl). (C) Standard curve ofSytoxBlue fluorescence using increasing amounts of a C2284 single cell debris fraction (approximately 10 cells/μl – see Figure 4). (D)Increase in SytoxBlue to PrestoBlue fluorescence ratio of C2284 spheroids treated with increasing amounts of saponin. (E) Drug responseof C2284 cells treated with increasing amounts of staurosporine for 48 h, as determined using SytoxBlue/PrestoBlue fluorescence (bars =SD of mean, three replicates). (F) Drug response of C2284 cells treated with increasing amounts of 5-fluorouracil for 72 h, as determinedusing SytoxBlue/PrestoBlue (bars = SD of mean, three replicates). (G) Drug response of C3953 cells treated with camptothecin for 72 h,as determined by the SytoxBlue/PrestoBlue ratio. (H) Time course of average C3953 colony area change during treatment with differentconcentrations of camptothecin and a vehicle control measured using consecutive stills from time-lapse movies (bars = SD of mean offive colonies). (I) Graph of the summed average colony movement as a factor of drug concentration, determined using ImageJ analysis ofconsecutive stills from the time-lapse movies of C2284 colonies treated with staurosporine/vehicle control for 44 h (bars = SD of mean offive colonies analysed). (J) Average total cumulative movements over time of C2284 colonies treated with 2 μM staurosporine and vehicle,as a control, using the same data as for I (bars = SD of mean of five colonies analysed). *p = 0.05, Student t-test, three replicates.


44 N Ashley et al

suggesting that they represent differentiated enterocytesshed from the colonies that acquire an abnormal mor-phology in the 3D environment. The bubble cells’ abnor-mal appearance suggests aberrant fluid transport to intra-cellular vesicles, possibly due to a lack of normal polar-ity cues. Whilst morphologically similar to rare muci-nous signet cell carcinomas, bubble cells did not stainfor mucins (data not shown).

Patient material can be extremely limited, particularlyendoscopic biopsy material. We developed two sensitiveapproaches to measuring drug cytotoxic responses, evenusing very small amounts of material. Both approachesare non-destructive, enabling additional testing of drugresponses or further characterization of the patient’scancer.

In summary, we have shown that a therapeuticresponse can be obtained rapidly from small amounts ofcancer material, which is an important requirement forimproved personalized cancer treatment. In the longerterm, such cultures should significantly augment in vitrocharacterization of the pathology of cancers and theirassociated drug responses.

Acknowledgments

We thank staff at the Oxford BioBank and donor patientsfor the cancer samples, and Unisensor A/S for use ofthe oCelloScopeTM. This work was funded by EpiDia-Can and Cellomatic FP7 European Union Grants (NosFP7-241783 and FP7-278204).

Author contribution statement

NA, JW, DO, and WFB conceived the study design. NAand MJ performed experiments. NA and WFB wrote thepaper.

References1. Brattain MG, Fine WD, Khaled FM, et al. Heterogeneity of malig-

nant cells from a human colonic carcinoma. Cancer Res 1981; 41:1751–1756.

2. Durrant LG, Robins RA, Pimm MV, et al. Antigenicity of newlyestablished colorectal carcinoma cell lines. Br J Cancer 1986; 53:37-45.

3. Dexter DL, Spremulli EN, Fligiel Z, et al. Heterogeneity of cancercells from a single human colon carcinoma. Am J Med 1981; 71:949-956.

4. Cammareri P, Lombardo Y, Francipane MG, et al. Isolation andculture of colon cancer stem cells. Methods Cell Biol 2008; 86:311-324.

5. Giavazzi R, Campbell DE, Jessup JM, et al. Metastatic behavior oftumor cells isolated from primary and metastatic human colorectalcarcinomas implanted into different sites in nude mice. Cancer Res1986; 46: 1928-1933.

6. Dangles-Marie V, Pocard M, Richon S, et al. Establishment of humancolon cancer cell lines from fresh tumors versus xenografts: com-parison of success rate and cell line features. Cancer Res 2007; 67:398-407.

7. Weiswald LB, Richon S, Validire P, et al. Newly characterised ex vivo

colospheres as a three-dimensional colon cancer cell model of tumour

aggressiveness. Br J Cancer 2009; 101: 473-482.8. Weiswald LB, Richon S, Massonnet G, et al. A short-term colorectal

cancer sphere culture as a relevant tool for human cancer biology

investigation. Br J Cancer 2013; 108: 1720-1731.9. Vermeulen L, Todaro M, de Sousa Mello F, et al. Single-cell cloning

of colon cancer stem cells reveals a multi-lineage differentiation

capacity. Proc Natl Acad Sci U S A 2008; 105: 13427-13432.10. Ricci-Vitiani L, Lombardi D, Pilozzi E, et al. Identification and

expansion of human colon-cancer-initiating cells. Nature 2007; 445:111-115.

11. Sato T, Stange DE, Ferrante M, et al. Long-term expansion of epithe-

lial organoids from human colon, adenoma, adenocarcinoma, and

Barrett’s epithelium. Gastroenterology 2011; 141: 1762-1772.12. Jung P, Sato T, Merlos-Suarez A, et al. Isolation and in vitro expan-

sion of human colonic stem cells. Nature Med 2011; 17: 1225-1227.13. Wang F, Scoville D, He XC, et al. Isolation and characteriza-

tion of intestinal stem cells based on surface marker combinations

and colony-formation assay. Gastroenterology 2013; 145: 383-395e381-321.

14. Kondo J, Endo H, Okuyama H, et al. Retaining cell–cell contact

enables preparation and culture of spheroids composed of pure pri-

mary cancer cells from colorectal cancer. Proc Natl Acad Sci U S A

2011; 108: 6235-6240.15. Richman PI, Bodmer WF. Monoclonal antibodies to human colorec-

tal epithelium: markers for differentiation and tumour characteriza-

tion. Int J Cancer 1987; 39: 317-328.16. Makin CA. Monoclonal antibodies raised to colorectal carcinoma

antigens. Ann R Coll Surg Engl 1986; 68: 298-301.17. Epenetos AA, Nimmon CC, Arklie J, et al. Detection of human

cancer in an animal model using radio-labelled tumour-associated

monoclonal antibodies. Br J Cancer 1982; 46: 1-8.18. Durbin H, Rodrigues N, Bodmer WF. Further characterization, iso-

lation and identification of the epithelial cell-surface antigen defined

by monoclonal antibody AUA1. Int J Cancer 1990; 45: 562-565.19. Ashley N, Yeung TM, Bodmer WF. Stem cell differentiation and

lumen formation in colorectal cancer cell lines and primary tumors.

Cancer Res 2013; 73: 5798-5809.20. Vermeulen L, De Sousa EMF, van der Heijden M, et al. Wnt activity

defines colon cancer stem cells and is regulated by the microenviron-

ment. Nature Cell Biol 2010; 12: 468-476.21. Todaro M, Perez Alea M, Scopelliti A, et al. IL-4-mediated

drug resistance in colon cancer stem cells. Cell Cycle 2008; 7:309-313.

22. Emmink BL, Van Houdt WJ, Vries RG, et al. Differentiated human

colorectal cancer cells protect tumor-initiating cells from irinotecan.

Gastroenterology 2011; 141: 269-278.23. Watanabe K, Ueno M, Kamiya D, et al. A ROCK inhibitor permits

survival of dissociated human embryonic stem cells. Nature Biotech-

nol 2007; 25: 681-686.24. Sato T, Vries RG, Snippert HJ, et al. Single Lgr5 stem cells build

crypt–villus structures in vitro without a mesenchymal niche. Nature

2009; 459: 262-265.25. Ogura E, Senzaki H, Yoshizawa K, et al. Immunohistochemical

localization of epithelial glycoprotein EGP-2 and carcinoembryonic

antigen in normal colonic mucosa and colorectal tumors. Anticancer

Res 1998; 18: 3669-3675.26. Yeung TM, Gandhi SC, Wilding JL, et al. Cancer stem cells from

colorectal cancer-derived cell lines. Proc Natl Acad Sci U S A 2010;

107: 3722-3727.27. Bracht K, Nicholls AM, Liu Y, et al. 5-Fluorouracil response in a

large panel of colorectal cancer cell lines is associated with mismatch

repair deficiency. Br J Cancer 2010; 103: 340-346.



28. Goldwasser F, Bae I, Valenti M, et al. Topoisomerase I-relatedparameters and camptothecin activity in the colon carcinoma celllines from the National Cancer Institute anticancer screen. CancerRes 1995; 55: 2116-2121.

29. Krawetz RJ, Li X, Rancourt DE. Human embryonic stem cells:caught between a ROCK inhibitor and a hard place. Bioessays 2009;31: 336-343.

30. Yamamoto H, Bai YQ, Yuasa Y. Homeodomain protein CDX2regulates goblet-specific MUC2 gene expression. Biochem BiophysRes Commun 2003; 300: 813-818.

31. Hall PA, Coates PJ, Ansari B, et al. Regulation of cell number in themammalian gastrointestinal tract: the importance of apoptosis. J CellSci 1994; 107: 3569-3577.

SUPPORTING INFORMATION ON THE INTERNETThe following supporting information may be found in the online version of this article.

Movie 1. Time-lapse imaging of primary culture initiation.

Movie 2. Time-lapse imaging of C2284 primary culture grown in suspension in serum-free medium.

Movie 3. Time-lapse imaging of ‘bubble cells’ within a C2284 primary culture spheroid grown in suspension in serum-free medium.

Movie 4. Time-lapse imaging of C5251 primary culture extruding cells that die when grown in Matrigel in serum-free medium.

Movie 5. Time-lapse imaging of bubble cells extruding from a Matrigel-embedded C5251 colony and dying.

Movie 6. Time-lapse imaging of C3953 primary culture forming a lumen when grown in Matrigel with DMEM with 10% FCS.

Movie 7. Time-lapse imaging of C2284 spheroids incubated with 2 μM staurosporine for 44 h (1 h intervals).

Table S1. Success rate and growth of primary cultures.


Documents

Rapidly derived colorectal cancer cultures recapitulate ...images.philips.com/is/content/PhilipsConsumer/Biocell...2016/11/17 · tem (Unisensor A/S, Lillerød, Denmark), located