Intravasation through the lymphatic border is an early- and rate-limiting step in the metastatic cascade of breast cancer [1
]. Yet, hardly any anti-metastatic drugs are available. As from 1981 to 2014 approximately 83% of the approved small molecule anticancer drugs were natural products, based on natural compounds or mimicking them in different ways [4
]. We tested traditional medicinal plants regarding their anti-intravasative properties [5
]. In a study on the effects of different Scrophularia
extracts on breast cancer cells´ proliferation, cell death, and intravasation through lymph endothelial cell barriers, a methanolic extract of Scrophularia lucida
L. (Scrophulariaceae) proved particularly active in inhibiting intravasation in vitro [8
]. An activity-guided fractionation of this extract revealed hispidulin, a member of the class of flavonoids, as significantly contributing to this activity [9
Because of the multitude of structures and amounts of flavonoids in human nutrition, these compounds have gained a lot of interest for the prevention, deceleration of development, or even alleviation of different “Western Diseases” such as cardiovascular or neurodegenerative disorders as well as various cancers. For flavonoids in cancer chemoprevention numerous modes of action, modulation of a multiplicity of proteins and inhibition of many enzymes involved in cancer pathology have been shown [10
]. Among those, different mechanisms involving many steps in the “metastatic cascade” influencing cell detachment, cell mobility, tissue barriers for intravasation and extravasation, angiogenesis etc., were studied in detail and many molecular targets involved in the anti-invasive and anti-metastatic properties of flavonoids such as luteolin, apigenin, kaempferol among others have been identified [11
Nevertheless, for an activity-guided search for effective natural anticancer compounds (e.g., flavonoids) inhibiting mechanisms of metastasis, robust in vitro tools simulating the in vivo situation are necessary. For this purpose 3D multicellular tumor spheroids as used in the validated “circular chemorepellent-induced defect” (CCID) assay provide an excellent possibility [12
]. Until now, data of the efficacy of flavonoids in this assay are scarce and only few flavonoids have been tested on breast cancer intravasation [1
]. Yet, the IC50
of hispidulin was too high to be reached in vivo [9
]. Therefore, to gain insight into structural features influencing the anti-intravasative potential of flavonoids, we compared hispidulin to 20 other structurally closely related flavonoids in this 3D assay consisting of MCF-7 cell spheroids and lymph endothelial cell (LEC) monolayers.
For the determination of the efficacy of flavonoids on the inhibition of lymph endothelial disintegration in the CCID assay in total 18 flavonoids were compared (Table 1
Among those were four flavanones and ten flavones and four flavonols with different degrees of hydroxylation and/or methoxylation. All compounds possessed the typical 5,7-dihydroxy- pattern. The differences applied to hydroxylation in positions 3′, 4′, 6, and/or 8. One compound each with a methoxy group in positions 3′, 4′, or 8, as well as two compounds with a methoxy group in position 6 were included into the series. Three concentrations (10, 25 and 50 or 75 µM) of each compound were tested (Supplemental Figure S2a
) and the effects were compared to Bay11-7082 (Supplemental Figure S2b
), an irreversible inhibitor of I-κBα phosphorylation and of CCID formation [2
], for which an IC50
of 7.2 µM was determined.
The results showed that flavanones, lacking the double bond between positions 2 and 3, have obviously no or very little effect on the inhibition of CCID formation. This was independent of the degree of hydroxylation in ring B (no OH (pinocembrin, 1), 4′-OH (naringenin, 2) or 3′,4′-diOH (eriodictyol, 3)). Only homoeriodictyol (4) with a 3′-methoxy-4′-hydroxy pattern resulted in a very high IC50 value of 99.5 µM.
In contrast, the corresponding flavones chrysin (5, IC50 24.8 µM), apigenin (11, IC50 34.1 µM), and luteolin (14, IC50 19.3 µM) were the most active compounds. Unexpectedly, 14 with the lowest logP among these flavones showed the highest activity. Additional hydroxylation in position 3 in the flavonols galangin (6) and kaempferol (12) led to slightly higher IC50 values as compared to 5 and 11. Interestingly, quercetin (17), the 3′,4′-di-OH-flavonol corresponding to the 3′,4′-di-OH-flavone 14 remained without any activity. Gossypetin (18) with an additional 8-OH group as compared to 17 was inactive as well.
Methylation of the 4′-OH group in 14 to diosmetin (15) significantly decreased the activity to approximately one-tenth (IC50 183.2 µM).
Additional hydroxylation of the flavones in positions 6 or 8 also reduced the activity: only baicalein (7; 6-hydroxy-chrysin) showed a weak effect (IC50 130.2 µM) whereas norwogonin (9; 8-hydroxy-chrysin) was not active.
Methylation of the 6- or 8-OH groups in oroxylin A (8; 6-O-methyl-baicalein), wogonin (10; 8-O-methyl-norwogonin), and hispidulin (13; 6-O-methyl-scutellarein) resulted in an increase in the activity as compared to the respective unmethylated compounds with IC50 values of 74.8 µM, 113.9 µM, and 88.6 µM, respectively.
In comparison to 14, the most active compound, and its 4′-O-methylderivative 15, with only one-tenth of its activity, the 6-methoxyderivative nepetin (16) resulted in an IC50 value of 79 µM.
As proof of our hypothesis we included three further compounds, namely acacetin, scutellarein, and herbacetin (Table 2
Acacetin (19), a 3′-OH methylated derivative of the highly active flavone apigenin (11 IC50
34.1 µM), resulted in almost equivalent activity (IC50
36.9 µM), underlining the impact of the 2,3-double bond and the free positions 6 and 8 on the effects of flavones in the CCID assay (Figure 1
The importance of free positions 6 and 8 was additionally shown by the inclusion of scutellarein (20; 6-OH-apigenin) and herbacetin (21; 8-OH-kaempferol). As expected from the IC50 values of baicalein (>100 µM) and norwogonin (not active) as compared to chrysin (24.74 µM), in comparison to 11 (IC50 34.1 µM) no activity was observed for 20 and in contrast to 12 (IC50 44.7 µM) 21 remained without effect as well.
The comparison of logP values did not allow an estimation of the anti-intravasative potential of the investigated flavonoids (Supplemental Figure S3
). Regarding the effects of the three most active compounds luteolin (IC50
19.3 µM; logP 2.26), chrysin (IC50
24.8 µM; logP 2.84), and apigenin (IC50
34.1 µM; logP 2.53) in the CCID assay, no correlation between the IC50
and the logP values was seen. The only deducible trend was the lack of activity for compounds with logP values below 2.10. Compounds with similar logP values such as galangin (logP 2.61) and pinocembrin (logP 2.69) showed significant differences resulting in very good (IC50
39.6 µM) and no activity, respectively. The activity of the most potent flavonoid luteolin was in a comparable range with the positive control Bay11-7082.
In summary based on the results of 21 flavonoids, flavones showed better effects in our experimental set-up in the CCID assay than the corresponding flavonols, whereas flavanones remained without effects. An additional hydroxy group at C-6 or C-8 or an additional methoxy group at C-8 in ring A of the flavonoid scaffold resulted in a significant loss of activity. Further substitution of C-6 with a methoxy group decreased the activity to a lesser extent. In contrast, the impact of the hydroxylation/methoxylation pattern in ring B at C-4′ or C-3′ and C-4′ was much less and did not follow a deducible order (luteolin (3′,4′ OH) > chrysin (no OH) > apigenin (4′ OH) > acacetin (4′ OCH3)).
The rationale to this study emerged from work on the anti-carcinogenic effects of a MeOH extract from the medicinal plant Scrophularia lucida
which is endemic to the Eastern Mediterranean [8
]. This extract strongly inhibited LEC barrier breaching because of the presence of hispidulin and unknown synergizing components [9
]. Therefore, hispidulin was compared to other closely related flavonoids. Several flavonoids have been shown to disrupt the pathways and molecular mechanisms essential for cancer cell survival, also in breast cancer cells. Metastatic processes such as invasion and endothelial-to-mesenchymal-transition (EMT) were inhibited by such compounds as well [19
]. Here, we did not focus on the tumor cells, but on the stabilizing effect of flavonoids on tumor-induced disintegration of the lymphendothelial barrier. In the clinical routine, the crossing of breast cancer cells into the lymphatic vasculature and subsequent colonization of the sentinel lymph node is a relevant prognostic marker. In our experimental set-up, anti-intravasative flavonoids with up to four-fold higher activity than hispidulin (IC50
: 89 µM) were identified, namely luteolin (IC50
: 19 µM) as the most potent one > chrysin (IC50
: 25 µM) > apigenin (IC50
: 34 µM) > acacetin (IC50
: 37 µM) > galangin (IC50
: 40 µM) > kaempferol (IC50: 45 µM) (see Table 1
). The tested flavonoids are present in fruits and vegetables, plants, mushrooms, bee pollen and propolis [21
]. Yet, the bio-available concentrations are too low to achieve LEC barrier stabilization by dietary consumption [23
]. Nevertheless, a significant inverse correlation with colorectal cancer risk and the intake of flavonoid subclasses (flavonols, flavan-3-ols, anthocyanidins) was reported [24
]. However, a lower risk of breast cancer associated with flavonoids was not observed [25
]. The most active flavonoids, luteolin, chrysin, and apigenin, are commercially available and particular formulations of luteolin and apigenin reached plasma concentrations in the range of those exhibiting significant in vitro anti-intravasative effects in our study. In detail, a single oral dose of luteolin or luteolin together with a peanut hull extract (14.3 mg/kg each) resulted in peak concentrations of 1.9 µg/mL and 8.3 µg/mL (respectively) of luteolin in rat plasma [26
]. This corresponds to luteolin concentrations of ~6.9 µM and ~29.1 µM, respectively. Similarly, a single oral dose of apigenin or apigenin delivered with a carbon nanopowder solid dispersion carrier (60 mg/kg each) gave peak concentrations of 1.3 µg/mL and 3.2 µg/mL (respectively) of apigenin in rat plasma [27
]. This corresponds to apigenin concentrations of ~4.8 µM and ~11.8 µM, respectively. Chrysin at daily oral doses (50 mg/kg) in a mouse xenograft model with A549 non-small-cell lung cancer cells, reduced the tumor volume by 50% as compared to control. In breast cancer mouse models, oral administration of chrysin significantly reduced the growth of lung metastases of 4T1 breast cancer cells and tumor growth in a MDA-MB-231 model [28
]. This proves the bio-availability of chrysin as well as its anti- carcinogenic potential [30
]. However, it is still not known whether this effect was due to chrysin itself or due to a metabolite. For an anti-metastatic activity of chrysin, the inhibition of angiogenesis through reduced vascular endothelial growth factor (VEGF) expression might be responsible. Furthermore, chrysin induced E-cadherin, decreased vimentin, SNAIL and SLUG levels thereby reverting EMT. Chrysin inhibited AKT-dependent matrix metalloprotease (MMP)-10 expression in triple negative breast cancer, cell migration and invasion [31
], and MMP9 through JNK—ERK—AP1 [32
]. Importantly, a daily dose of 3 g chrysin did not exhibit toxicity in a pilot clinical trial [33
Luteolin (25 µM) inhibited MMP9 and MMP2 in breast- and ovarian cancer cells [34
] at concentrations which did not compromise the viability of MDA-MB231- and MCF-7/6 cells. Luteolin concentrations as low as 1 µM inhibited the invasion of MDA-MB231 cells through a basement membrane extract [36
]. This supported the notion that luteolin specifically targets migratory/metastatic processes but not proliferation/survival mechanisms in cancer cells. MMP2 and MMP9 contribute to the intravasation in the MCF-7/LEC model [13
] and this can explain the anti-intravasative properties of chrysin and luteolin. This is also in agreement with the observation that 20 µM luteolin inhibited the matrix metallo-protease 1 (MMP1) in MDA-MB231 cells [16
]. MMP1, MMP2, and MMP9 are triggering the retraction of LECs [37
] and the formation of CCIDs. Similar to chrysin, 10 µM luteolin induced the epithelial marker E-cadherin, down-regulated the mesenchymal markers vimentin, N-cadherin SNAIL, and SLUG, and attenuated the migration of MDA-MB231 cells in a wound healing assay [38
]. This evidenced the inhibition of EMT and of cell mobility. Apigenin as well inhibited SNAIL, EMT, and metastasis of hepatocellular carcinoma [39
]. Urokinase-plasminogen activator expression, migration, invasion and phorbol-12-myristate-13-acetate (PMA)-induced MMP-9 secretion were inhibited in MDA-MB231 cells [40
] and in MCF-7 cells [41
] upon treatment with apigenin.
In addition, apigenin, luteolin, and chrysin inhibit nuclear factor κB (NF-κB) activity [34
]. MCF-7 cells secret 12-hydroxyeicosatetraenoic acid (12(S)-HETE), which triggers the activation of NF-κB in LECs as a prerequisite for CCID formation in the LEC monolayer [2
]. Thus, luteolin, chrysin, and apigenin stabilize the endothelial barrier function by inhibiting NF-κB in LECs despite the de-stabilizing effects of 12(S)-HETE. Another mechanism of luteolin and apigenin is the inhibition of FAK phosphorylation at Tyr397 and thereby attachment/detachment oscillations at the matrix, which is necessary for LEC migration and CCID formation [16
The methylation of the hydroxy group at C-4′ in apigenin leads to acacetin, which was almost as active as apigenin. In MDA-MB231 cells acacetin (25 µM) significantly inhibited PMA-induced MMP9 mRNA expression after 48 h [34
]. The absorption of methylated flavonoids in intact organisms is much higher as compared to the respective non-methylated compounds (i.e., apigenin). Therefore, these compounds show improved metabolic stability and bio-availability [44
]. Acacetin is also commercially available and deserves further investigations for a potential pharmaceutical use. Diosmetin, the 4′-O
-methyl-luteolin, was inactive in our assay. In another investigation, diosmetin (~30 µM) inhibited ~50% invasion and migration of SK-HEP1- and MHcc97H hepatocellular carcinoma cells after 24 h. However, these results did not correlate with an inhibition of MMP2 and MMP9 [45
]. We suppose that in our experimental set-up treatment with diosmetin for more than 4 h may have resulted in a similar inhibition of LEC migration. Consequently, the lesser effect of diosmetin, compared to luteolin, must have been the consequence of an un-favorable structural alteration because of the hydroxyl group at 3′ in combination with the 4′O
Galangin and kaempferol exhibited almost similar anti-intravasative activities. They inhibited MMP2, MMP9, migration and invasion, and suppressed EMT in MDA-MB-231- and MCF-7 breast cancer cells [34
], as well as NF-κB activity in various experimental models [49
]. The bioavailability of kaempferol after oral administration (100 mg/kg) in Spargue-Dawley rats was 2% [51
]. After a single oral dose of galangin (10 mg/kg) only 220 ng/mL of the unmetabolized compound were detected in rat plasma [52
Similar to the flavonoids above, oroxylin A, wogonin, and baicalein were also reported to inhibit EMT marker expression in MCF-7, A549, and MDA-MB231 cell lines, respectively [53
] and to inhibit MMP2/9, migration and invasiveness of MDA-MB231 cells [56
]. In a xenograft nude mouse model, baicalein inhibited tumor metastasis [59
]. However, the IC50
values of these flavonoids in our experiments were probably too high to achieve anti-intravasative effects in a patho-physiological setting.
Migration and invasion of MCF-7 cells were inhibited upon treatment with 80 µM quercetin and 20 µM blocked PMA-induced upregulation of MMP9 [60
]. At a concentration of 50 µM this compound inhibited the migration of SAS oral squamous carcinoma cells which correlated with the suppression of MMP2 and MMP9 protein [61
]. In contrast, suppression of MMP2 and MMP9 protein upon treatment with 25 µM quercetin did not show any correlation with the inhibition of cancer cell migration [62
]. However, this concentration correlated with an inhibition of invasion of osteosarcoma cells in vitro and invasion of cancer cell xenografts into the lungs in a mouse model. This indicates that the anti-migratory property of quercetin was not necessarily connected to MMP2/9 expression in cancer cells. The migration of aorta blood endothelial cells was only weakly inhibited by 100 µM quercetin. This concentration had no effects on the tube formation of vascular endothelial cells, another cell migratory process [63
]. In fact, the migration of blood endothelial cells during angiogenesis depends on VEGF/VEGFR2 instead of MMP2/9, and quercetin interferes with VEGFR2 expression [64
]. However, in lymph endothelial cells, the VEGF/VEGFR2 signaling is not observed during intravasation of MCF-7 cells [13
]. Also, the activation of NF-κB, on which the formation of CCIDs in the LEC monolayer is strictly dependent, was not inhibited by quercetin in the endothelial cell line ECV304 [66
]. As CCID formation was not affected at 10–75 µM quercetin, this process might be devoid of targets addressed by quercetin.
The anti-intravasative activities of the other tested flavonoids were too low to consider targeting migratory and intravasation-supporting mechanisms in LECs, although herbacetin interferes with pro-metastatic HGF-cMET signaling in MDA-MB231 cells [67
]. Hence, cMET obviously does not play a role in the migration of LECs.
The migration of LECs was dependent on the stimulation of NF-κB by MCF-7-secreted 12(S)-HETE and subsequent expression of ICAM-1 in LECs, which facilitates the adhesion of LECs to cancer cells as a prerequisite for CCID formation [3
]. Furthermore, 12(S)-HETE induces LEC migration through the release of Ca2+
from intracellular stores, the induction of phospholipase C and generation of inositol 3-phosphate, as well as the activation of the RHO/ROCK pathway, both activating the mobility of protein MLC2 [37
]. In breast cancer cells also matrix metalloproteases play a role in CCID formation such as MMP1 and MMP11 [37
] and MMP2/9 and TIMP2 [13
]. Whether MMP2/9 and TIMP2 activity is LEC- or cancer cell associated and whether and how these pathways are interconnected is still an open issue.