Mouse mammary carcinoma 4T1 cells (cat. no. ATCC CRL-2539) were obtained from LGC Limited and were not further tested or authenticated. 4T1 cells were cultured in RPMI (Thermo Fisher Scientific, Inc.) supplemented with 10% FBS (HyClone; Cytiva), 100 U/ml penicillin and 100 µg/ml streptomycin (Thermo Fisher Scientific, Inc.) in a humidified 95% air/5% CO₂ incubator at 37°C.

Jurkat immortalized T-lymphocytes cells (cat. no. ATCC TIB-152) were obtained from American Type Culture Collection and were not further characterized. Jurkat cells were cultured in RPMI supplemented with 10% FBS, 100 U/ml penicillin and 100 µg/ml streptomycin in a humidified 95% air/5% CO₂ incubator at 37°C. Cells were checked for the presence of mycoplasma using the MycoAlert™ Mycoplasma Detection Kit (Lonza Group Ltd.) on a monthly basis and not used if identified as positive.

Predictions of the conserved domains of mEgfl7 were performed using National Center for Biotechnology Information (NCBI) conserved domains (https://www.ncbi.nlm.nih.gov/Structure/cdd) and the Prosite (http://smart.embl-heidelberg.de) databases using the NP_001158036 referenced sequence.

The mouse Egfl7 cDNA or domain deletion variants corresponding to the removal of the EMI, EGF or C-terminal domains were cloned in frame with a C-terminal influenza hemagglutinin (HA)-coding sequence in the pMSCV plasmid for 4T1 cells (Clontech Laboratories, Inc.) and pLPCX vector for HUVEC (Clontech Laboratories, Inc.), allowing the production of retroviruses as described (19). Human umbilical vein endothelial cells (HUVEC) (Lonza Group, Ltd.) and 4T1 cells (7.5×10³ cells/cm²) were infected with control, Egfl7- or Egfl7-deletion domain coding retroviruses for 6 h at 37°C, and whole cell populations were selected for puromycin resistance (4 µg/ml) for 7 days, after which they were used for experiments. Infected cells were used after the antibiotic selection period and not maintained further. pLPCX-mEgfl7 or pLPCX-mEgfl7-ΔEGF vectors were constructed by cloning the c-terminal HA-tagged mEgfl7 coding sequence (11) or the mEgfl7-ΔEGF deletion variant into the HindIII/EcoRI sites of the pLPCX vector. For transfection, HUVEC (4×10⁴ cells/well) were plated in 2-cm² well plates and transfected with 58 fmoles pLPCX-Ctrl, pLPCX-mEgfl7 or pLPCX-mEgfl7-ΔEGF using JetPEI-HUVEC (Polyplus-transfection SA) for 4 h at 37°C, before the culture medium was changed. At 48 h post-transfection, cells were lysed and RNA was extracted.

A basal layer of complete medium [RPMI (Thermo Fisher Scientific, Inc.) supplemented with 10% FBS (HyClone; Cytiva), 100 U/ml penicillin, 100 µg/ml streptomycin (Thermo Fisher Scientific, Inc.)] and 0.6% agar was poured into 4-cm² well plates. A suspension of 4T1-Ctrl, -Egfl7, -ΔEMI, -ΔEGF and -ΔCt cells (2.5×10³/well) in 0.45% agar was overlaid onto the basal agar layer. Complete medium was added on top of the cell-containing agar layer immediately after gelling. Cells were cultured in a humidified 95% air/5% CO₂ incubator at 37°C and the culture medium was changed twice a week. The number of colonies (defined as round, white & well defined cell groups, ≥80 pixels in size) was determined at 14 days after plating by taking pictures of the illuminated plates on a dark background under a Nikon SMZ1000 binocular light microscope and analyzed with ImageJ software (version 1.53h; National Institutes of Health) (25).

4T1-Ctrl, -Egfl7, -ΔEMI, -ΔEGF and -ΔCt cells were seeded (1.25×10³ cell/cm²) into 4-cm² well plates and cultured in a humidified 95% air/5% CO₂ incubator at 37°C. At 24 h intervals, cells were collected after trypsin digestion and counted using a Z2 Coulter counter (Beckman Coulter, Inc.).

A total of 70 female BALB/c mice (age, 6–8 weeks; weight, 20 g) obtained from Charles River Laboratories (n=7 per experimental group in each experiment) were maintained in a pathogen-free environment (12-h light/dark cycles; 22±2°C; 55±15% humidity) with ad libitum access to water and food. Mice were housed at least one week prior to tumor grafting to allow appropriate acclimatization. 4T1 cells (5×10⁵/50 µl) expressing HA-tagged Egfl7 or its domain variants were injected in the mammary fat-pad and tumors were measured every other day using an electronic caliper. Tumor volumes were calculated using the following formula: Volume=width × width × length × π/6 (26). The maximal tumor volume observed was 650 mm³, the maximal length was 14.8 mm and the maximal width was 8.5 mm. Mice were sacrificed via cervical dislocation. Mice were housed according to European legislation (http://data.europa.eu/eli/dir/2010/63/oj) at the Institut Pasteur de Lille Animal Facility. The protocols were approved by the French Ministry of Higher Education and Research Ethic Committee (approval no. 00294.01).

Tumors were either fixed in 4% paraformaldehyde at 4°C for 24 h, embedded in paraffin and sectioned (thickness, 7 µm) for H&E and NCR1 staining, or frozen in optimal cutting temperature compound in N₂-cooled liquid isopentane, sectioned at −20°C (thickness, 10 µm) and post-fixed with 1% paraformaldehyde for 5 min at room temperature for CD3ε and CD31 detection. Immunostaining was performed as previously described (19). Briefly, sections were incubated in 10 mM Tris-HCl pH 7.5, 150 mM NaCl (TBS) containing 1% BSA (MilliporeSigma) and 10% goat serum (MilliporeSigma) for 2 h at room temperature, then in TBS containing 1% BSA and the following primary antibodies: CD3ε (1:100; Abcam; cat. no. ab5690), natural cytotoxicity triggering receptor 1 (NCR1; 1:100; R&D Systems, Inc.; cat. no. AF2225) and CD31 (1:100; BD Pharmingen; BD Biosciences; cat. no. 550274) for 1 h at room temperature. Subsequently, the sections were incubated with a donkey anti-rabbit Alexa fluor-488 (cat. no. A21206; 1:250; Thermo Fisher Scientific, Inc.) or a goat anti-rat Alexa Fluor 594 secondary antibody (cat. no. A11007; 1:250; Thermo Fisher Scientific, Inc.) for 1 h at room temperature. For optical microscopy, sections were H&E counterstained with Mayer's hematoxylin (MilliporeSigma) for 5 sec and then 1% eosin (MilliporeSigma) for 30 sec at room temperature. After rinsing for 10 min under tap water, the sections were dehydrated and mounted in Vectamount (Vector Laboratories). For immunofluorescence microscopy, cell nuclei were labeled by staining the sections with DAPI solution (1 µg/ml; Sigma-Aldrich; Merck KGaA) for 1 min at room temperature. Sections were analyzed and images were captured using an Axioplan-2 light microscope or an AxioImager-Z1 fluorescent microscope (Carl Zeiss AG) and ZEN software (version 2.0). Quantification was performed either by direct counting or by area fraction measurements over ≥5 separate microscopic fields using ImageJ software.

4T1 cells, HUVEC or tumors fragments were homogenized in TRIzol^® (Thermo Fisher Scientific, Inc.). Total RNA was extracted and reverse transcribed using a High Capacity cDNA reverse transcription kit (Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. qPCR was performed in duplex reactions of human or mouse FAM-labeled probes (Applied Biosystems; Thermo Fisher Scientific, Inc.; mouse Egfl7, Mm00618004_m1; human Egfl7, Hs00211952_m1; human ICAM-1, Hs00164932_m1; human VCAM-1, Hs01003372_m1; human E-selectin, Hs00950401_m1), together with a human (cat. no. 4326315E) or mouse (cat. no. 4352341E) β-actin VIC-labeled probe using TaqMan gene expression assays, reagents (Applied Biosystems; Thermo Fisher Scientific, Inc.) and conditions on a StepOne machine (Applied Biosystems; Thermo Fisher Scientific, Inc.). The following thermocycling conditions were used for qPCR: 50°C for 2 min; initial denaturation at 94°C for 10 min; followed by 40 cycles of denaturation at 94°C for 15 sec, and an annealing and extension step at 60°C for 1 min. Cq values were calculated at the upper linear range of the logarithm-2 amplification curve using StepOne software (version v2.3; Applied Biosystems; Thermo Fisher Scientific, Inc.). Expression was quantified using the 2^−ΔΔCq method (27), where ΔCq = Cq of transcript of interest-Cq of β-actin measured in the same tube, and ΔΔCq = mean ΔCq experimental samples - mean ΔCq control samples of the same experiment. Relative quantity is 2^−ΔΔCq and transforms the logarithmic-2 data into decimal values.

Total protein was isolated from 4T1 cells or tumor tissues using RIPA buffer containing complete protease inhibitors (Roche Diagnostics, Inc.). Proteins were quantified using a BCA protein assay kit (Pierce; Thermo Fisher Scientific, Inc.), separated (62.5 µg per lane) via 12% SDS-PAGE and transferred onto Immobilon-P membranes (Merck Millipore). The membranes were incubated for 4 h at room temperature with primary antibodies targeted against HA.11 (Clone 16B12; 1:1,500; cat. no. MMS-101P; Covance, Inc.) and actin (cat. no. sc-1615; 1:1,000; Santa Cruz Biotechnology, Inc.). Subsequently, the membranes were incubated with anti-mouse (1:10,000; cat. no. NA931-1ML; Cytiva) or anti-goat-HRP conjugated secondary antibody (1:10,000; cat. no. A8919; Sigma-Aldrich; Merck KGaA) for 1 h at room temperature. Chemiluminescence was revealed using ECL Prime western blotting kit (Thermo Fisher Scientific, Inc.) either by exposure to autoradiography films (Cytiva) or by using a Fujifilm Las3000 gel reader (FUJIFILM Wako Pure Chemical Corporation) for semi-quantification.

Human Jurkat T-lymphocytes (1×10⁶ cells) were incubated with DiI (2 µM; Molecular Probes; Thermo Fisher Scientific, Inc.) for 10 min at 37°C. After washing with RPMI, Jurkat cells were seeded (10⁵ cells/cm²) onto monolayers of confluent HUVEC for 20 min at 37°C. Cells were rinsed three times with PBS and then fixed in 4% paraformaldehyde for 20 min at room temperature. Fluorescent cells were imaged with a UV-microscope. The number of adherent cells was estimated using an area fraction calculation, which was applied to images of 10–15 separate microscope fields, using ImageJ software (version 1.53h; National Institutes of Health) (25).

Each experiment was performed at least twice with at least triplicate samples per condition. Comparisons among groups were analyzed using the Kruskal-Wallis test followed by Dunn's post hoc test. Statistical analyses were performed using GraphPad Instat software v3.10 (GraphPad Software, Inc.). P<0.05 was considered to indicate a statistically significant difference. Error bars in graphs represent the calculated standard deviation.

To analyze the roles of the Egfl7 domains in tumor growth and the escape from immunity, the present study designed retroviral expression vectors encoding either the full-length molecule or domain-deletion variants of the protein. Based on NCBI and Prosite database predictions, the N-terminal domain spans Met₁-Pro₂₇, the EMI domain is defined as the sequence containing Ser₂₈-Ala₁₀₅ and the EGF repeats span Ala₁₀₆-Lys₁₈₀ (Fig. 1A). Moreover, the C-terminal domain from the Glu₁₈₁ to Leu₂₇₅ end of Egfl7 does not contain any predicted protein feature. Vectors coding for C-terminal HA-tagged versions of the full-length protein and deletion variants corresponding to the removal of the EMIlin domain (ΔEMI), the EGF repeats (ΔEGF) or the C-terminal domain (ΔCt) (Fig. 1A) were cloned and produced into retroviruses that were then used to infect mouse mammary 4T1 carcinoma cells. Cells were subsequently selected for resistance to antibiotics as whole populations. Each construct was shown to lead to efficient infection and expression, since the selected cell populations produced a HA-tagged protein at the expected apparent size when assessed via western blotting (Fig. 1B).

Uncontrolled proliferation of cancer cells is a major hallmark of tumors (28), and Egfl7 has been shown to have significant effects on cell proliferation depending on the conditions and cell type selected (19,24,29–31). Under the conditions, overexpressing full-length Egfl7 in 4T1 cells had no effects on the cell proliferation rate compared with the 4T1-control (Ctrl) group (Fig. 1C). Similarly, deleting EMI, EGF or C-terminal domains had no effect on the proliferation rate of the 4T1-ΔEMI, -ΔEGF and -ΔCt cells (Fig. 1C), suggesting that Egfl7 and its domain variants did not participate in the regulation of 4T1 cancer cell proliferation under these conditions.

Anchorage-independent colony formation in vitro is another characteristic of cell transformation (32), and 4T1 cells are spontaneously able to form colonies under such conditions. When assessing the potential effects of Egfl7 and its domain deletion variants on colony formation by 4T1 cells, no significant differences were identified between the number of colonies emerging from 4T1-Ctrl cells and those expressing the full-length Egfl7. Variants with deletion of the EMI, EGF or the C-terminal domains also showed no differences in colony numbers (Fig. 1D and E). However, there was a tendency to a size difference of the colonies formed by 4T1-ΔEGF cells, which were on average 16% smaller compared with those formed by 4T1-Ctrl and 4T1-Egfl7 cells (Fig. 1F).

Despite displaying no noticeable effects on cancer cell proliferation in vitro, Egfl7 does significantly promote tumor growth in vivo (19). To identify which of its predicted domains were involved in this effect, 4T1 cells expressing either the full length Egfl7, its domain deletion variants or 4T1-Ctrl cells were inoculated into BALB/c immunocompetent mice, and tumor growth was monitored over time. The results demonstrated that 4T1-Egfl7 cells produced tumors that grew faster compared with those of 4T1-Ctrl cells, reaching 395±90 mm³ after 20 days, whereas 4T1-Ctrl tumors reached 135±55 mm³ after the same time period (Fig. 2A). Deletion of the EMI or C-terminal domain from Egfl7 did not significantly affect the growth rates of the corresponding tumors, which remained comparable to the fast growth rate of 4T1-Egfl7 tumors. On the other hand, deleting the EGF repeats significantly decreased the rate of tumor growth in mice, resulting in a ~27% smaller mean volume (290±65 mm³) compared with 4T1-Egfl7 tumors at 20 days post-inoculation (Fig. 2A). Furthermore, although the mean tumor weight at sacrifice in the 4T1-Egfl7 group was significantly higher compared with that in the 4T1-Ctrl group, a statistical difference was not observed between 4T1-Ctrl and 4T1-ΔEGF tumors, which was in line with the data obtained from the tumor growth kinetics results (Fig. 2B). As previously described, 4T1-Egfl7 formed highly necrotic tumors when compared with 4T1-Ctrl cells (19), and deleting the EGF domains significantly decreased the levels of intratumor necrosis (Fig. 2C and D).

To verify that variations in Egfl7 expression did not account for these differences, Egfl7 transcript levels were quantified in RNA extracted from the dissected tumors. As expected, Egfl7 expression levels were significantly higher in 4T1-Egfl7 tumors compared with those in 4T1-Ctrl tumors, whereas Egfl7 expression levels were similar between 4T1-Egfl7 and 4T1-ΔEGF tumors (Fig. 2E). Similarly, Egfl7-HA protein expression levels were significantly higher in 4T1-Egfl7 tumors compared with those in 4T1-Ctrl tumors. Moreover, Egfl7-HA protein expression levels were similar between 4T1-Egfl7 and 4T1-ΔEGF tumors (Figs. 2F and S1), and did not appear to account for the observed differences in tumor development.

The promoting effects of Egfl7 on tumor growth are due, at least in part, to the repression of blood vessel endothelial cell activation, which in turn prevents the infiltration of CD4⁺ and CD8⁺ T-lymphocytes, NK cells and other immune cells within the tumor mass (19). To assess whether the slower growth rate of 4T1-ΔEGF tumors was associated with partially restored immune infiltration, 4T1-Ctrl, -Egfl7 and -ΔEGF tumors were analyzed for T cell and NK cell infiltration using immunohistochemistry. 4T1-Egfl7 tumors were significantly less infiltrated by T-lymphocytes and NK cells compared with 4T1-Ctrl tumors, as demonstrated by staining for the CD3ε T-cell and NCR1 NK markers (Fig. 3A and B). Quantification of T-cell infiltration demonstrated that 24.3±4.2 CD3ε⁺ cells/field were present in 4T1-Ctrl tumor sections, whereas this number decreased to 3.2±0.7 cells/field in 4T1-Egfl7 tumors. With regards to the infiltration of NK cells, 23.2±6.2 NCR1⁺ cells/field were counted in 4T1-Ctrl tumor sections compared with only 8.0±0.4 cells/field in 4T1-Egfl7 tumors. By contrast, deleting the EGF repeats from Egfl7 partially restored the immune infiltration of tumors, with the number of infiltrated CD3ε⁺ T-lymphocytes and NCR1⁺ NK cells increasing by 2.82-fold (8.9±0.0 cells/field) and 1.69-fold (13.5±3.1 cells/field) in 4T1-ΔEGF tumors compared with 4T1-Egfl7 tumors (Fig. 3A and B). This increase in immune infiltration of 4T1-ΔEGF tumors suggested that the EGF repeats of Egfl7 may actively participate in preventing the recruitment of cytotoxic immune cells by tumor blood vessel endothelial cells expressing wild-type Egfl7.

Our previous study reported that immune cells that were not able to cross the tumor endothelial barrier in 4T1-Egfl7 tumors, thus remained within the lumens of these blood vessels (19). Indeed, microscopic observations of tumor sections revealed the presence of a significantly higher number of nucleated cells in the lumens of 4T1-Egfl7 tumor blood vessels compared with those in 4T1-Ctrl tumors blood vessels (Fig. 3C). Furthermore, such circulating cells were observed in significantly lower numbers in 4T1-ΔEGF tumor blood vessels compared with those in 4T1-Egfl7 tumors, but in significantly higher numbers compared with those in 4T1-Ctrl tumors, thereby confirming the role of EGF repeats in immune cell recruitment by tumor blood vessels.

Expression of Egfl7 by cancer cells promotes tumor angiogenesis in vivo (19). This effect was demonstrated in the present study by assessing the proportion of CD31⁺ blood vessels in tumor sections. It was found that blood vessel density was significantly increased in 4T1-Egfl7 tumors compared with that in 4T1-Ctrl tumors, whereas 4T1-ΔEGF tumors were significantly less vascularized, with blood density levels comparable to those measured in 4T1-Ctrl tumors (Fig. 3D), suggesting that the EGF repeats also actively participated in the tumor angiogenic effects of Egfl7.

The repression of endogenous Egfl7 activates endothelial cells in the absence of proinflammatory cytokines, which is due to an increase in the expression levels of ICAM-1, VCAM-1 and E-selectin, and a consequential spontaneous adhesion of leukocytes to endothelial cells following Egfl7 repression (19). Since HUVEC spontaneously express large amounts of human Egfl7 (hEgfl7), to evaluate whether the EGF repeats of Egfl7 were involved in endothelial cell activation in vitro, corresponding mouse Egfl7 variants were overexpressed in HUVEC. It was first examined whether overexpressing mouse Egfl7 (mEgfl7) molecules would artificially affect hEgfl7 endogenous expression. The results demonstrated that neither mEgfl7 or mEgfl7-ΔEGF transfection did so. As expected, high expression levels of mEgfl7 transcripts were detected after transfecting cells with the mouse expression vectors (Fig. 4A).

Then, HUVEC were stably infected with retroviruses produced from the same mouse expression vectors, and whether overexpression of mEgfl7 could repress endogenous leukocyte adhesion receptors in the presence of endogenous hEgfl7 was assessed. The expression levels of ICAM-1 did not significantly vary after infecting HUVEC with either mEgfl7 or mEgfl7-ΔEGF. However, mEgfl7 overexpression did significantly repress VCAM-1 and E-selectin expression in HUVEC, and the removal of the EGF repeats of mEgfl7 partially and totally relieved this repression, respectively (Fig. 4B).

Preliminary data (data not shown) suggested that the Egfl7 protein directly interfered with the binding of leukocytes to the endothelium. To assess whether the effects would correspond to functional differences in HUVEC activation in response to the mEgfl7 forms, fluorescent Jurkat leukocytes were incubated onto confluent monolayers of HUVEC infected with Ctrl, mEgfl7 or mEgfl7-ΔEGF constructs. Under these conditions, the HUVEC-Ctrl group spontaneously supported a high rate of leukocyte adhesion (Fig. 4C) as a consequence of the retroviral infection, whereas overexpression of mEgfl7 significantly reduced the number of leukocytes adhering in the HUVEC-Egfl7 group (Fig. 4C). On the other hand and in line with the aforementioned results, the absence of the EGF repeats from mEgfl7 induced higher leukocyte adhesion levels in the HUVEC-ΔEGF group compared with those in the HUVEC-mEgfl7 group, and reached comparable levels to those observed in the HUVEC-Ctrl group.

Collectively, these results suggested that the EGF repeats of Egfl7 formed a key domain that participated in promoting tumor growth in vivo by reducing leukocyte infiltration into tumors and promoting angiogenesis. Moreover, these domains participated in regulating endothelial cell activation and their ability to interact with circulating leukocytes.