Background
Colorectal cancer (CRC), one of the most frequent malignancies in the Western world, is commonly associated with mutations in the tumor suppressor
Adenomatous polyposis coli (APC) gene both in hereditary [
1] and in sporadic CRC [
2]. Mutations of this gene constitutively activate β-catenin target genes causing tumorigenesis [
3]. The
Apc
Min/+ mouse, which spontaneously develops multiple intestinal neoplasms (Min) in the small and large intestine, is considered a good model to study CRC [
4,
5].
Studies revealed that all of the Notch receptors, their ligands and some of their effectors (Hes1, 5, 6, 7 and Atoh1) are expressed in the normal intestinal crypts, the main niche for stem cells. Dll1 and Dll4 interact with Notch1 and 2 in the normal gut to maintain the homeostasis of intestinal stem cells [
6] by repressing the cyclin-dependent kinase inhibitors
Cdkn1b and Cdkn1c [
7]. The Notch signaling pathway also promotes the enterocyte/colonocyte differentiation [
8,
9], while Atoh1 and Klf4, which are transcriptionally repressed by Hes1, specify secretory (goblet, enteroendocrine, and Paneth) cell differentiation [
10,
11].
Activation of Notch1 together with Wnt signalling seems to be essential to trigger CRC initiation by maintaining the self-renewal of tumor stem cells [
12,
13]. These cells share some characteristics with normal stem cells but have accumulated oncogenic mutations and lost growth control. They possess the strongest tumor-initiating potential of all tumor cells and promote tumor growth and resistance to many current therapies, including chemo and radiotherapy [
14]. Several intestinal stem cell markers have been investigated, such as the leucine-rich repeat-containing G-protein-coupled receptor 5 (Lgr5) and the B cell–specific Moloney murine leukemia virus insertion site 1 (Bmi1) [
15‐
17]. These two markers are also present in the normal gut in two functionally different intestinal stem cell populations; Lgr5 is present in mitotically active stem cells that are sensitive to irradiation and Wnt modulation, while Bmi1 is a marker for a reserve population of resistant quiescent injury-inducible stem cells [
18]. In CRC both Lgr5 and Bmi1 positive stem cell populations are associated with cancer initiation and progression [
15,
17,
19] and are regulated by the Notch pathway [
20,
21].
Accumulating evidence has shown that tumor stem cells promote tumor angiogenesis and that their maintenance depends upon functional angiogenesis [
22]. It is currently widely recognized that tumor growth and maintenance is dependent on the expansion of the individual’s vasculature to the center of the tumor [
23]. Previous work showed that the inhibition of Dll4/Notch represses tumor growth by promoting dysfunctional an immature tumoral angiogenesis in a variety of xenograft and autochthonous mouse models [
24‐
27]. In xenograft models of CRC anti-Dll4 therapy seems to additionally reduce the frequency of tumor stem cells [
28,
29].
Thus, we set out to characterize the expression pattern of Dll4 and other Notch members in the Apc
Min/+ tumors relatively to the normal intestine and compare endothelial-specific with ubiquitous Dll4 loss-of-function mutants to address the role of Dll4 in intestinal tumor development in the Apc
Min/+ model. In particular, we aimed to assess whether an effect of Dll4 signalling on the Apc
Min/+ tumor stem cells was solely a consequence of its action on tumor angiogenesis or if other, more direct, mechanisms might be involved.
Methods
Experimental animals
All experiments were conducted in accordance with the Portuguese legislation for the use of animals for experimental purposes (Decreto-Lei n° 129/92 and Portaria n° 1005/92, DR n° 245, série I-B, 4930-42) and with the European Union legislation (Directive n. 86/609/EEC, from the 24
th November 1986). All animal-involving procedures in this work were approved by the national regulatory agency (DGAV – Direção Geral de Alimentação e Veterinária) and the Institutional Animal Care and Use Committee (CEBEA – Comissão de Ética e Bem-Estar Animal) (Approval ID: PTDC/CVT/71604/2006). All sections of this report adhere to the ARRIVE Guidelines for reporting animal research [
30]. A completed ARRIVE guidelines checklist is included in Additional file
1.
Mice were maintained in a conventional facility in a 12-h light/dark cycle, in ventilated cages with corncob as bedding, and were given access to standard laboratory diet and water ad libitum. The welfare of the mice was regularly monitored.
Mutant C57BL/6J-Apc
Min/+/J (Apc
Min/+) mice were purchased from the Jackson Laboratory (Bar Harbor, ME).
Two wild type C57BL/6J male mice were used to analyse Notch pathway expression in the normal intestine and two Apc
Min/+ male mice were used to characterize their expression in intestinal tumors.
Apc
Min/+ mice were crossed with
Dll4 conditional homozygous mice (
Dll4
lox/lox
). The resulting
Apc
Min/+;
Dll4
lox/lox
progeny was crossed with
VE-cadherin-Cre-ERT2 mice to produce endothelial-specific inducible
Dll4 loss-of-function (endo
Dll4
-/-
) or with
Cag-Cre-ERT2 mice to produce ubiquitous inducible loss-of-function (ubiq
Dll4
-/-
). When these mice reached 6 weeks, recombinase cre activity was induced by daily i.p. tamoxifen administration (50 mg/kg in 10% ethanol plus 90% cremophor), during 5 days as in [
31]. Tamoxifen treated cre-negative mice were used as controls. Twelve males per group were used in the analysis.
Macroscopic analysis of the intestine
At 18 weeks of age the animals were sacrificed and the small and large intestine were excised, flushed and opened longitudinally. The macroscopic small and large intestinal tumors of Apc
Min/+
Dll4 mutant mice and controls were counted and measured with a calliper under the dissection microscope in a blinded manner. Tumor volume was calculated assuming a hemispherical shape for the small bowel tumors and a spherical shape for large intestinal tumors. The volumes of all tumors from each mouse were added to give the overall tumor burden per animal. Normal WT small and large intestine and the tumors of these regions from Apc
Min/+ Dll4 mutant mice and controls were collected.
Histopathological analysis
The collected samples were fixed in 10% formalin solution for 48 h, dehydrated in alcohol, cleared in xylene, embedded in paraffin, sectioned at 4 μm and stained with hematoxylin (Fluka AG Buchs SG Switzerland) and eosin Y (Sigma, St. Louis, MO) for histopathological analysis. The lesions observed on the H&E sections from Apc
Min/+ mice were classified as hyperplasias, when only an increase of the number of cells was observed, or as adenomas with low and high-grade dysplasia based on the alterations of the shape of the nucleus, the nucleus to cytoplasm ratio, cell polarity, chromatin pattern, and changes in gland architecture.
Periodic Acid-Schiff (PAS) staining (Sigma, St. Louis, MO) was used to mark the intestinal goblet cells. These cells were counted in the intestine PAS stained sections using a 400× magnification.
Immunohistochemical analysis
Three series of sequential 4 μm sections of paraffin embedded-WT normal small and large intestine and Apc
Min/+ intestinal adenomas from two mice each were used (2 sections per slide).
After dewaxing and rehydration, endogenous peroxidase activity was quenched (15 min, 1% H
2O
2) and antigen retrieval was performed (20 min at 95 °C in 10 mmol/L sodium citrate buffer, pH 6). The primary antibodies to mark Dll1, Dll4, Notch1-3, Hes1 and 5 (Abcam, Cambridge, UK) and Jagged1, Jagged2, Dll3 and Notch4 (Santa Cruz Biothecnology, California, USA) were diluted in PBS containing 2% bovine serum albumin, and incubated overnight at 4 °C with the tissue sections. These antibodies have been previously validated [
32‐
34]. The following morning, the tissue sections were incubated with goat anti-rabbit (Merck Millipore, Massachusetts, USA) or rabbit anti-goat (Santa Cruz Biothecnology, California, USA) horseradish peroxidase–labeled secondary antibody and the staining was revealed with ImmPACT DAB Peroxidase Substrate (100 μl, Vector Laboratories, Burlingame, USA).
The sections were examined under an Olympus Bx51 microscope with the 40×/0.75 objective (UPlanfL). The images were captured with a Olympus DP21 camera.
A semi-quantitative analysis of the protein expression in the epithelium was performed by a pathologist in a blinded manner. Twelve representative fields for each staining were evaluated according to a scoring criteria adapted from [
35]. Staining intensity was scored as 0 (negative), 1 (weak), 2 (moderate), and 3 (strong). Percent positivity was converted to scores as 0 (0%), 1 (1–2%), 2 (3–15%), 3 (16–30%), 4 (31–50%), 5 (51–75%) and 6 (75–100%). A final score was obtained by multiplying the percentage score by the intensity score.
Immunofluorescence analysis
Small and large intestinal tumors were fixed in a 4% (w/v) paraformaldehyde in PBS solution at 4 °C for 1 h, cryoprotected in 15% (w/v) sucrose in PBS solution, embedded in 7.5% (w/v) gelatin in PBS solution, snap frozen in liquid nitrogen and cryosectioned in 10 and 20 μm-thick sections. The following primary antibodies were used: anti-PECAM-1, anti-E-cadherin (BD Biosciences, San Jose, USA), anti-α-SMA, anti-PCNA, anti-Lgr5, anti-HIF1α, anti-Cyclin D1 (Abcam, Cambridge, UK), anti-Dll4 (R&D Systems, Minneapolis, USA), anti-lysozyme (Dako, Glostrup, Denmark), anti-non-phospho (active) β-catenin (Cell Signaling Technology, Danvers, USA). Species-specific secondary antibodies conjugated with Alexa Fluor 488 and 594 (Invitrogen, Carlsbad, CA) were used to reveal primary antibody binding. Tissue sections were incubated with primary antibody overnight at 4 °C and with secondary antibody for 1 h at room temperature. Nuclei were counterstained with 4′, 6-diamidino-2-phenylindole dihydrochloride hydrate (DAPI; Molecular Probes, Eugene, OR).
Fluorescent immunostained sections were examined under a Leica DMRA2 fluorescence microscope with a Leica HC PL Fluotar 10 and 20×/0.5 NA dry objective, captured using Photometrics CoolSNAP HQ, (Photometrics, Friedland, Denmark), and processed with Metamorph 4.6-5 (Molecular Devices, Sunnyvale, CA, US). Morphometric analyses were performed using the NIH ImageJ 1.37v program. After transforming the RGB images into binary files, the percentage of white pixels per field was defined as a positive signal.
Under the effect of 2-2-2 tribromoethanol anaesthesia, biotin-conjugated lectin from Lycopersicon esculentum (100 μg/100 μl of PBS) or 1% Evans’ Blue solution (Sigma, St. Luis, MO, US) were administered in the caudal vein to mark vessel perfusion and extravasation, respectively. Both solutions were allowed to circulate for 5 min before the vasculature was transcardially perfused with 4% (w/v) paraformaldehyde in PBS solution for 3 min. Tumor samples were collected and processed as described above. Tissue sections were stained and tumor perfusion was quantified by determining the percentage of red PECAM-positive structures that were co-localized with Streptavidin-Alexa 488 (Invitrogen, Carlsbad, CA, US) signals. Evans’ Blue is red fluorescent and extravasation was visualized in contrast to green fluorescent vascular structures.
Apoptosis was measured using the TUNEL assay (Roche, Mannheim, Germany).
Quantitative transcriptional analysis
Intestinal tumors were snap frozen in liquid nitrogen until RNA extraction (Qiagen RNeasy). Using the SuperScript® III First-Strand Synthesis SuperMix for qRT-PCR (Invitrogen, Carlsbad, CA, USA), first-strand cDNA was synthesized from total RNA. Real-time PCR analysis was performed using the comparative C
T method [
36]. Primer pair sequences are listed in Additional file
2: Table S1. Gene expression was normalized to
β-actin and in the case of genes expressed in the vasculature it was additionally normalized to
Pecam-1.
Statistical analysis
To compare measurements between control and test groups, the Mann–Whitney-Wilcoxon test was performed using the Statistical Package for the Social Sciences v15.0 (Chicago, IL). Results are presented as relative average ± SEM. P-values <0.05 and <0.01 were considered significant (*) and highly significant (**), respectively.
Discussion
Activation of Notch pathway seems to promote intestinal tumorigenesis induced by
Apc loss [
13] and Dll4 is one of the Notch signaling pathway components found to be upregulated in these tumors [
46]. Reports have shown that Dll4 inhibition delays the tumor growth by deregulating the tumor angiogenic process [
24‐
27], but in CRC anti-DLL4 therapy may also reduce the cancer stem cell frequency [
29]. Despite the advances in the understanding of Dll4/Notch signaling in cancer, most of the previous reports were focused on role of Dll4 in the tumor angiogenic process and further studies are still needed to unveil all the mechanisms by which Dll4 affects the tumor initiation and development in the gut.
Reports have shown that all Notch receptors, ligands and some of the Notch target genes are expressed in the normal gut [
8,
37,
38]. However, in the
Apc
Min/+ intestinal tumors their expression has been poorly described. A study indicated that expression of Notch receptors and ligands closely follows the expression in the normal crypts, while Hes1 expression was observed uniformly in the adenomas [
8]. Other report showed that Jagged1 is overexpressed in the tumor tissue with concomitant Notch1 and 2 activation [
39]. In the present work we analysed the protein expression pattern of most Notch pathway members in the
Apc
Min/+ intestinal tumors compared with the normal WT gut. Regarding the previous gene expression analysis of Notch members in the normal gut [
37,
38], we observed some differences in our analysis. These included the presence of Notch2 in the bottom of the large intestinal crypts, of Notch3 and 4 in the small and large intestinal epithelium and Hes1 not only in the small intestinal crypts [
37], but diffusely expressed in the small and large intestine.
Our expression analysis in the
Apc
Min/+ small and large intestinal adenomas confirmed that the Notch pathway is present and activated in intestinal adenomas harbouring
Apc mutations [
8,
13,
39,
46]. Dll4 and Jagged1 were more expressed in these tumors than the other members of this pathway. Comparing the adenomas with the normal WT gut we found that Dll1 and Dll3 lose their expression in the large intestine. We observed a different expression pattern of Dll4, all Notch receptors (with regional variation) and Hes5 in the tumor epithelium. The same Notch members and Hes1, instead of Hes5, seemed upregulated in the adenomas (Notch3 only in the large intestine). Thus our expression analysis indicates that in the
Apc
Min/+ small and large intestinal adenomas, Dll4 is the most upregulated ligand and is present both in the tumor epithelium and endothelium.
In the normal gut Dll4 acts redundantly with Dll1 mediating the Notch signaling regulation of the intestinal stem cells proliferation and their commitment towards de secretory cell fate [
6]. We found that Dll4 is expressed near the Lgr5+ stem cells also in the intestinal tumors, therefore indicating a possible role of this ligand in the maintenance of also the tumor stem cells. These stem cells are believed to be responsible for tumor initiation and progression [
47,
48] and depend on proper angiogenesis to function and survive [
22]. Therefore we intended to elucidate if Dll4 also regulates the fate of tumor stem cells beside its angiogenic effect in a spontaneous model of CRC, the
Apc
Min/+ mouse. To address this question we compared ubiquitous with endothelial-specific Dll4 loss-of-function mouse mutants. Our results highlighted the importance of Dll4 angiogenic and epithelial effect during intestinal
Apc
Min/+ tumor initiation and development rather than in maintaining the normal gut homeostasis. Pellegrinet el al. reported that in the normal gut,
Dll4 intestinal epithelial-specific inhibition alone is not sufficient to promote a phenotype due to redundant Dll1-mediated Notch signaling [
6]. This lack of intestinal effect after
Dll4 inhibition can also be related to the fact that in the normal gut simultaneous inhibition of
Notch1 and
2 is necessary to result in complete conversion of the crypt progenitors into postmitotic goblet cells [
7] and it is not known whether Dll4 can activate both receptors in the gut. Nevertheless, our results show that Dll4 seems at least partially responsible for the known effects of Notch activation during intestinal tumorigenesis, as Dll4 ubiquitous deletion led to a similar, but less pronounced, epithelial phenotype as the pan-Notch/
gamma-secretase inhibition in the
Apc
Min/+ tumors [
8]. However, as the alterations were moderated with no macroscopic repercussions (no observed increase of mucus secretion) in the ubiquitous Dll4 mutants’ gut, Dll1 may partially compensate the lack of Dll4 in this setting and/or Dll4 may not activate both Notch1 and 2 receptors. In addition, as we only analysed the Lgr5 and Bmi1 positive stem cell populations, it is not certain if this pathway can affect similarly all the stem cells present in the intestinal tumors.
We found that ubiquitous and endothelial-specific Dll4 blockade led to a similar phenotype in the small and large intestine, but a stronger effect on tumor initiation was observed in the small intestine and a greater impact on tumor growth was seen in the large intestine.
By comparing the ubiquitous mutants with the endothelial-specific knockouts we found that the observed epithelial phenotype is probably caused by the deregulation of the tumor angiogenesis but also by other important mechanisms. Both ubiquitous and endothelial-specific mutants had an equivalent angiogenic phenotype, with equally increased hypoxia and apoptosis leading to similar reduction of the tumor volume. Therefore, Dll4 deletion inhibited the intestinal tumor growth by inducing a dysfunctional and immature angiogenesis that led to hypoxia and therefore apoptosis as previously reported in other tumor models [
24,
26,
27,
49].
The multiplicity of tumors was also reduced in the mutants relatively to their controls and this effect was more pronounced in the ubiquitous than in the endothelial-specific mutants, associated to a stronger reduction of tumor cell proliferation and tumor stem cell density in the first mutants. The stronger inhibitory effect on tumor cell proliferation through Dll4 ubiquitous deletion may have therefore prevented the accumulation of more mutations that lead to tumor initiation, promote the transition of microadenomas to macroadenomas and favor the neoplastic transformation. Therefore in the intestinal adenomas, Dll4 seems to promote proliferation and maintain the stem cells through angiogenic, but also non-angiogenic related mechanisms. Indeed we observed decreased expression of Myc, cyclin D1 and D2, independently of β-catenin activation, only in the ubiqDll4
-/-
tumors.
The Wnt signaling has been considered a crucial player in the initiation of CRC associated to inactive mutations in the
APC gene [
3]. Nuclear accumulation of β-catenin promotes neoplastic conversion by triggering the cell cycle-regulators Cyclin D1 and D2 and Myc and, consequently, uncontrolled cell proliferation contributing to tumor progression [
40,
50,
51]. Notch signaling seems to cooperate with Wnt signaling to trigger intestinal tumorigenesis, as activation of Notch in
Apc mutant mice led to a significant increase in the number of adenomas developed [
13]. Additionally, a previous study indicated that Jagged1 was the link between Wnt and Notch pathways in the
Apc
Min/+ tumorigenesis, where β-catenin seems to transcriptionally activate Jagged1 [
39]. However, it has been shown that the Mastermind-like 1 co-activator of Notch pathway can bind to the promoters of Cyclin D1 and Myc in colon cancer cell lines [
52] and these molecules are activated directly by Notch1 in other types of cancer [
53‐
57] and possibly by Cyclin D1 in CRC [
58]. It has been also demonstrated that Cyclin D2 and Myc are also induced by Notch1 to promote stem cell renewal in another setting [
59]. Additionally, the overexpression of Dll4 in a leukemia cell line led to increased protein expression of Myc [
60]. Therefore, during
Apc
Min/+ tumorigenesis Dll4/Notch signaling may directly upregulate the expression of Cyclin D1 and D2 and Myc. We observed that Dll4 deletion reduced tumorigenesis without affecting β-catenin nuclear accumulation and thus Wnt activation. Therefore, Dll4/Notch activation may promote intestinal tumorigenesis by angiogenic and non-angiogenic mediated mechanisms in a β-catenin independent manner. The non-angiogenic related regulation may include a synergistic effect of Dll4/Notch with Wnt signaling to promote tumorigenesis by increasing the transcription of important Wnt target genes.
In addition, Dll4 ubiquitous inhibition upregulated the zinc finger-containing transcription factor KLF4 in the
Apc
Min/+ tumors. KLF4 is a cell proliferation inhibitor and can act as a tumor suppressor, being normally downregulated in
Apc
Min/+ tumors and in early stages of human CRC [
61]. The loss of one of its alleles increases
Apc
Min/+ tumorigenesis, possibly by derepressing β-catenin mediated gene expression [
62]. A previous work indicated that Notch signaling supresses KLF4 expression in intestinal tumors and colorectal cancer cells [
62]. Our results indicate that Dll4 seems to be the ligand responsible for this Notch-mediated phenotype. Therefore, Dll4/Notch may promote carcinogenesis by upregulating the transcription of Wnt target genes through KLF4 downregulation in the
Apc mutated tumors. Previous work indicated that Hes1 downregulation by Notch inhibition derepresses Atoh1, which seems to induce KLF4 upregulation to promote goblet cell differentiation in a redundant manner [
11,
63]. However, it seems that Hes1 may act both upstream and downstream of Atoh1 to negatively regulate KLF4 [
11]. We found that in the ubiquitous, but not in the endothelial-specific, Dll4 knockouts, the reduction of Lgr5 and Bmi1 positive tumor stem cell density was accompanied with increased tumor epithelium differentiation with a moderate deviation towards the secretory lineages, probably due to the observed
Atoh1 and
Klf4 overexpression by
Hes1 downregulation as it occurs when Notch signalling is inhibited [
8,
63]. This indicates that besides the effect on angiogenesis, Dll4/Notch signaling seems to have an additional role maintaining the tumor stem cells undifferentiated.
Additionally, Dll4/Notch ubiquitous inhibition promoted the transcription of the cell cycle regulators cyclin-dependent kinase (CDK) inhibitors
Cdkn1b and
Cdkn1c in the
Apc
Min/+ tumors. A previous report showed that inactivation of Notch1 and 2 in the normal gut is associated with derepression of these CDK inhibitors [
7]. This phenotype was completely abrogated in the absence of Atoh1 [
7,
64], a molecule that is also considered to act as a tumor suppressor in CRC [
65]. Therefore, Dll4/Notch inhibition may also negatively affect the tumor stem cell populations through Atoh1 derepression-mediated upregulation of the CDK inhibitors
Cdkn1b and
Cdkn1c.