CD44 variant 3 is involved in the TrkA/CD44 interaction following NGF stimulation
In our previous studies, we demonstrated that NGF induces the association between TrkA and CD44 in cancer cells [17]. The CD44v3- and CD44v6-containing isoforms expressing respectively exon 8 and 11(Fig. 1A) can interact with growth factor receptors. Using COS-7 cells with low CD44 and TrkA expression (Supplementary Fig. 1A-D), we differentially expressed TrkA and CD44s or a CD44 variant, either CD44v3 or CD44v6 (Fig. 1B, supplementary Fig. 2). PLA revealed greater TrkA/CD44 variant associations within a complex (Fig. 1B). Notably, only the expression of CD44 variant 3 with TrkA increased the PLA-detected TrkA/CD44v3 signal in the plasma membrane of COS-7 cells following NGF stimulation. Additionally, formation of the TrkA/CD44v3 complex was transient and decreased 30 min after NGF stimulation (Fig. 1B). Neither wild-type COS-7 cells nor COS-7 cells transfected with TrkA and the CD44s or CD44v6 isoform showed the formation of the dynamic TrkA/CD44 complex in PLA, indicating that these two isoforms are not involved in complex formation (Supplementary Fig. 2A-C). Moreover, in MDA-MB-231 breast cancer cells, flow cytometric analysis also revealed that NGF increased the plasma membrane level of CD44v3 but not CD44v6 (Supplementary Fig. 3A-D). Additionally, we evaluated the expression of TrkA and CD44v3 at the membrane level or in cells by immunofluorescence in MDA-MB-231 cells. The level of CD44v3 increased at the membrane until 15 min of NGF treatment and then decreased (Supplementary Fig. 3E & F). Interestingly, the membrane level of TrkA followed the same trend. Additionally, the overall fluorescence intensities for TrkA and CD44v3 increased globally in the cell, suggesting that after internalization, the proteins were not degraded (Supplementary Fig. 3G & H).
The complex formation mechanism was then further examined. Because CD44v3 binds growth factors, we tested the ability of CD44v3 to bind NGF by microscale thermophoresis (MST). In contrast to FGF-1 (Supplementary Fig. 4), NGF did not bind to CD44v3. Additionally, TrkA binding to CD44v3 was not detected in the absence of NGF but was observed in the presence of NGF (Kd = 9.47 nM) (Fig. 1C). Taken together, these data suggested a direct interaction between TrkA in complex with NGF and CD44v3, but neither NGF nor TrkA alone could bind CD44v3.
To further investigate the interaction between TrkA and CD44v3, molecular modeling was performed using the available structure of TrkA complexed with NGF to identify amino acid residues involved in TrkA/CD44v3 complex formation (Fig. 1D). Because no structure of variant exon 3 was available, the CD44v3 sequence was cut into 4 domains of approximately 10 amino acid residues (zone 1, STSSNTISAG; zone 2, WEPNEENEDE; zone 3, RDRHLSFSGSG; and zone 4, IDDDEDFISST), and we scanned each peptide to locate the putative binding sequence. Visual examination of the TrkA-CD44v3 poses suggested that TrkA His112 binds to the C-terminal region of exon v3 (IDD34DEDF38ISST) (Fig. 1D & E). Statistical BCA confirmed that His112 in zone 4 of CD44v3 is likely a key residue for the interaction of this variant with TrkA (Supplementary Fig. 5A). This histidine residue was critical in all TrkA isoforms (I, II and III) (Fig. 1F) but not in TrkB or C (Fig. 1G). In the latter two receptors, His112 was replaced with either a leucine or alanine, and subsequent analysis indicated that His112 is highly specific to TrkA. Notably, His112 is a polar amino acid that can form a hydrogen bond with the C-terminal region of CD44v3 (Asp4 and Phe38) while leucine or alanine in TrkB and TrkC are non-polar (Fig. 1H). Interestingly, multiple sequence alignment with sequences of various mammals also showed that His112 is highly conserved among mammals (Supplementary Fig. 5B-D). Additionally, replacing this histidine with arginine in rabbits and mice, in which the TrkA sequence is similar, did not abolish the CD44v3-TrkA interaction. Therefore, the conservation of this histidine residue was investigated more extensively through phylogenic analysis (Supplementary Fig. 5E). TrkA is conserved across vertebrates; therefore, Trk sequences in species of each vertebrate taxon were aligned to its more distant ancestor, DTrk (Drosophila melanogaster). In vertebrates and D. melanogaster, histidine is replaced by 1 of 3 different polar residues (lysine, glutamine, or glycine), which can form a hydrogen bond (glutamine in Danio rerio and Gallus; lysine in Branchiostoma floridae; and glycine in D. melanogaster). These observations indicated the positive selection of histidine 112 and its physiological importance. Additionally, alignment analysis of the CD44v3 sequence with that of CD44v6 showed that a region of CD44v6 shared 42% identity with the C-terminal region of CD44v3. Interestingly, when the residues corresponding to Asp34 and Phe38 were replaced by glutamate, a weaker hydrogen bond was formed, and when they were replaced with proline, no hydrogen bond was formed, explaining the absence/weak association of TrkA/CD44v6 indicated by PLA (Supplementary Fig. 2C). These data indicated that TrkA H112 and the CD44v3 C-terminal region are likely crucial for the biological function of these respective proteins. Therefore, using COS-7 cells, we first tested the effect of the TrkA H112A mutation (Fig. 1I) on TrkA/CD44v3 complex formation (Fig. 2A & B). In the TrkA H112A mutant cells, TrkA binding to CD44v3 was lost, as indicated by the lack of a PLA signal in COS-7 cells, compared with TrkA wild-type cells, as detected by PLA, confirming the essential role of H112 in the TrkA-CD44v3 interaction (Fig. 2A & B). The involvement of the CD44v3 C-terminal region (IDDDEDFISST) in TrkA/CD44v3 complex formation was then tested. To complete this test, expression vectors of CD44v3 lacking its C-terminal region (IDDDEDFISST) or neighboring region (RDRHLSFSGSG), constructs named CD44v3_Δ3 and CD44v3_Δ4, respectively (Fig. 2C), were cotransfected into COS-7 cells with TrkA. The number of TrkA/CD44v3 complexes in these transfected COS-7 cells was then quantified by PLA (Fig. 2D & E). The CD44v3_Δ4 mutant could not bind TrkA because no signal was detectable by PLA, while CD44v3_Δ3 could bind TrkA following NGF stimulation to an extent similar to that of CD44v3 (Fig. 2D & E). Taken together, these results suggest that the C-terminal region of variant 3 is involved in TrkA binding and this is a general mechanism because it can be observed in breast cancer cells but also in cos7 in which the expression of TrkA and CD44 were induced.
The C-terminal mimetic peptide of CD44v3 disrupts the TrkA-CD44v3 association
To block the TrkA-CD44v3 association, 4 mimetic peptides of CD44v3 were designed corresponding to zone 1 [(K) STSSNTISAG], zone 2 (WEPNEENEDE), zone 3 (RDRHLSFSGSG) and zone 4 (IDDDEDFISST) (Fig. 2F & G). The binding capacities of the CD44v3 mimetic peptides to TrkA were then measured by MST (Fig. 2H). As expected, neither peptide 1 (KSTSSNTISAG) nor peptide 2 (WEPNEENEDE) could bind TrkA, while mimetic peptide 4 (IDDDEDFISST) corresponding to the C-terminal region of CD44v3 could bind TrkA. To confirm mimetic peptide 4 binding, we assessed its ability to bind TrkA using nanoscale differential scanning fluorimetry (nanoDSF; Supplementary Fig. 6A & B). We observed that mimetic peptide 4 (but not a scramble peptide) bound TrkA, but we did not detect CD44v3/P4 binding (Supplementary Fig. 6C).
Unexpectedly, the RDRHLSFSGSG peptide (P3) corresponding to CD44v3 zone 3 also bound TrkA. Therefore, the abilities of peptides 3 and 4 to disrupt TrkA/CD44v3 complex formation were evaluated in the COS-7 cell line (Fig. 2I & J). As observed in CD44v3 mutants, only peptide 4 corresponding to zone 4 of CD44v3 could decrease NGF-induced TrkA/CD44v3 complex formation in COS-7 cells. However, peptide 3 did not inhibit TrkA/CD44v3 complex formation.
Collectively, our results demonstrated that the interaction of TrkA and CD44v3 was blocked by mimicking peptide 4 (IDDDEDFISST), which corresponds to the C-terminal region of exon variant 3 of CD44.
The TrkA-CD44v3 interaction is involved in NGF-induced migration and invasion of MDA-MB-231 TNBC cells
We previously showed that NGF is implicated in breast cancer aggressiveness because it enhances cell migration/invasion [17, 24]. Therefore, in this study, we assessed TrkA binding to CD44v3 in MDA-MB-231 cells, which were used as TNBC model cells. First, PLA was performed using anti-CD44 variant 3- (Fig. 3A and Supplementary Fig. 7) or anti-CD44 variant 6-specific antibodies (Fig. 3B). NGF induced TrkA/CD44v3 complex formation in MDA-MB-231 cells at the plasma membrane within 5 min to 30 min of administration (Fig. 3A and Supplementary Fig. 7) but did not induce TrkA/CD44v6 complex formation (Fig. 3B). Next, we tested the ability of CD44v3 mimetic peptides to block TrkA/CD44v3 complex formation in MDA-MB-231 cells (Fig. 3C-E). Similar to the results observed with COS-7 cells, neither a scramble peptide (Fig. 3C) nor mimetic peptide P3 (Fig. 3D) inhibited TrkA/CD44v3 complex formation following NGF treatment. Only mimetic peptide 4 (Fig. 3E) and the H112A mutation (Supplementary Fig. 8A & B) could disrupt TrkA/CD44v3 complex formation in MDA-MB-231 cells. Therefore, we evaluated the physiological consequences of TrkA/CD44v3 complex dissociation induced by CD44v3 mimetic peptide 4 on MDA-MB-231 cell phenotypes (Fig. 4). We first examined the effect of mimetic peptide 4 on clonogenic cell growth. For this experiment, MDA-MB-231 cells were grown in the presence of 2% FBS for 3 weeks in the presence or absence of CD44v3 mimetic peptide 4 (Fig. 4A and B). Under these conditions, CD44v3 mimetic peptide 4 reduced the proportion of MDA-MB-231 colony forming units by two-thirds compared with its effect on control cells or that of the scramble peptide. The same observations were made in SUM-159PT, another triple-negative cell line (Supplementary Fig. 9 A & B). Next, we tested the inhibitory effect of CD44v3 mimetic peptide 4 on the migration and invasion of MDA-MB-231 breast cancer cells (Fig. 4C-K). We first determined the effect of CD44v3 on the length-to-width (l/w) ratio of the cells (elongation factor) (Fig. 4C-E). NGF induced the acquisition of a fibroblast phenotype by MDA-MB-231 cells, which exhibited a significant increase in the l/w ratio (Fig. 4D-E). This effect of NGF was inhibited by the addition of CD44v3 mimetic peptide 4. Hence, MDA-MB-231 cell migration ability was assessed after NGF treatment in a wound healing test (Fig. 4H & I). Following NGF treatment, the migration of MDA-MB-231 cells was increased, as indicated by complete wound closure 24 h after treatment. This effect was not inhibited using the scramble CD44v3 mimetic peptide, while CD44v3 mimetic peptide 4 completely blocked NGF-dependent wound closure, indicating that CD44v3 mimetic peptide 4 inhibited the NGF-induced migration of MDA-MB-231 cells. We also confirmed the effects of CD44v3 mimetic peptide 4 on migration in SUM-159-PT cells (Supplementary Fig. 9 C & D). The MDA-MB-231 cell invasive capacity was then investigated using Boyden chambers (Fig. 4J & K). As observed in the pictures of the underside of the Transwell membrane (Fig. 4J), more NGF-treated cells than untreated cells invaded the collagen substrate. In the presence of CD44v3 mimetic peptide 4, a decrease in both basal levels and NGF-induced invasion was observed in MDA-MB-231 cells. These observations were confirmed by counting the cells that had invaded the Transwell membrane and were apparent on its underside (Fig. 4K). These effects on invasion were also observed in the MDA-MB-468 triple-negative breast cancer cell line (Supplementary Fig. 9 E). Additionally, the expression of TrkA-H112A also reduced migration and invasion compared with that of wild-type TrkA in MDA-MB-231 cells (Supplementary Fig. 8 C-E).
To determine the role of CD44v3 in NGF-induced phenotype acquisition, the effect of short interfering RNA (siRNA) targeting CD44v3 (siCD44v3) was assessed (Supplementary Fig. 10). siCD44v3 reduced the clonogenicity of MDA-MB-231 and SUM-159-PT cells (Supplementary Fig. 10 A & B) and MDA-MB-231 migration and invasion (Supplementary Fig. 10C-E).
Our results indicated that NGF also induced TrkA/CD44v3 complex formation in MDA-MB-231 breast cancer cells. Additionally, we depicted for the first time that the inhibition of TrkA/CD44v3 complex formation by CD44v3 mimetic peptide 4 is associated with the decreased clonogenic cell growth and migration/invasion abilities of MDA-MB-231 cells.
TrkA/CD44v3 activates RhoA independently of TrkA phosphorylation
To further elucidate the mechanism of action of CD44v3 mimetic peptide 4, we studied its effect on the TrkA signaling pathway. First, we performed western blot analysis to determine the effect of mimetic peptide 4 on TrkA phosphorylation (Fig. 5A). Interestingly, CD44v3 mimetic peptide 4 did not inhibit TrkA phosphorylation induced by NGF (Fig. 5A), indicating that disruption of TrkA/CD44v3 complex formation did not influence TrkA phosphorylation. Notably, we previously showed that NGF may enhance RhoA activity in breast cancer cells to promote migration/invasion independent of TrkA phosphorylation [17]. Therefore, we performed acceptor (photo) bleaching (AB)-FRET to examine the effect of CD44v3 mimetic peptide 4 on NGF-induced RhoA activity in MDA-MB-231 cells transfected with a RhoA FRET biosensor [18] (Fig. 5B). In MDA-MB-231 cells, NGF increased the FRET signal in the presence of the scramble peptide but not with CD44v3 mimetic peptide 4 (Fig. 5C). Additionally, to confirm that this effect was independent of TrkA phosphorylation, we measured RhoA activation in MDA-MB-231 cells expressing dead TrkA kinase, which cannot be phosphorylated (Fig. 5D). In these cells, we determined the ratiometric value of active (RhoA-GTP-to-yellow fluorescent protein (YFP) fluorescence) and inactive RhoA (RhoA-GTP-to-cyan fluorescence protein (CFP) fluorescence) after NGF stimulation (after 0, 25 and 45 min) and categorized the effect into five intensity categories (from low intensity, indicated in blue, to high intensity, indicated in white). We observed that NGF increased the proportion of active RhoA after 45 min in the presence of the scramble peptide (Fig. 5E) but not CD44v3 mimetic peptide 4 (Fig. 5F). Furthermore, we observed the relocation of activated RhoA to the migration front of the control cells with the scramble peptide after 25 min of NGF treatment but not with CD44v3 mimetic peptide 4 (Fig. 5E & F, Supplementary Fig. 11A). Furthermore, a similar decrease in RhoA activation was obtained using the TrkA-H112A mutation, which also prevents the formation of the TrkA/CD44v3 complex in the absence or presence of the TrkA kinase inhibitor k252A (Supplementary Fig. 11B & C, Supplementary Video 1 & 2).
Taken together, our results indicated that CD44v3 mimetic peptide 4 inhibits TrkA phosphorylation-independent RhoA activation induced by NGF.
TrkA/CD44v3 is involved in tumor growth and MDA-MB-231 cell metastasis
To assess the role of TrkA/CD44v3 in vivo, SCID mice were injected with MDA-MB-231 cells to generate xenografts. Primary tumor development and metastasis (to the lung, brain and liver) were analyzed (Fig. 6). To determine the effects of TrkA/CD44v3, tumor cell growth was compared after treatment with mimetic peptide 4 and PBS (Fig. 6A-C). We first observed that CD44v3 mimetic peptide 4 did not affect the overall body weight of the mice (Fig. 6B). The development of primary tumors was measured throughout the experiments. CD44v3 peptide 4 injections were performed when the mean tumor volume reached 100 mm3 (Fig. 6C). The CD44v3 mimetic peptide significantly inhibited the growth of MDA-MB-231 cells in xenograft tumors. This result confirmed our preliminary clonogenic assay results, indicating that CD44v3 mimetic peptide 4 decreased the long-term growth of MDA-MB-231 cells both in vitro and in vivo. We also assessed the metastatic burden. Because CD44 mimetic peptide 4 blocked primary development, we chose to inject the CD44v3 mimetic peptide into the mice harboring the largest tumors. Hence, the mice were sacrificed when the primary tumors reached the same size (Fig. 6D). Micrometastases were detected at the end of the experiment by PCR of human microglobulin in xenograft mice treated with or without the CD44v3 mimetic peptide (Fig. 6E). Similar to previously reported results [17], our experiments showed that wild-type MDA-MB-231 cells colonized the lung and liver and, to a lesser extent, the brain. Interestingly, CD44v3 mimetic peptide 4 decreased the metastatic burden at each analyzed location. These results indicated that MDA-MB-231 cell metastasis in these three organs was blocked by CD44v3 mimetic peptide 4. To confirm the implication of the TrkA/CD44v3 complex in metastasis, we also performed xenograft experiments using MDA-MB-231 HA-TrkA H112A cells (Fig. 6F). We confirmed that HA-TrkA cells (control) appeared more metastatic than MDA-MB-231 cells [24], but MDA-MB-231 cells overexpressing TrkA H112A xenografted mice seemed to exhibit less metastasis in the brain and liver. These results suggested that the TrkA/CD44v3 association could be implicated in metastasis and that CD44v3 mimetic peptide 4 might impair metastasis development.
The TrkA/CD44v3 complex is differentially detected in breast cancer cells and tumors
NGF induced TrkA/CD44v3 complex formation in MDA-MB-231 cells, which are the gold standard models of TNBC. We also examined the expression and induction of TrkA/CD44v3 in breast cancer cell lines corresponding to different breast cancer subtypes (Fig. 7A-D). Four cell lines were assessed: SUM159-PT (TN; i.e., ERα-, PR-, and Her2-) cells, T47-D (luminal A; i.e., ERα+, PR+, and Her2-) cells, BT-474 (luminal B; ERα+, PR+, and Her2+) cells and HCC-1954 (Her2-like; ERα-, PR-, and Her2+) cells. Under these conditions, NGF induction of CD44v3/TrkA complex formation was observed in SUM-159PT (TN) and HCC-1954 (Her2) cells. By contrast, T47-D cells (luminal A) expressed the TrkA/CD44v3 complex at a lower rate, and NGF induction was insufficient to modulate TrkA/CD44v3 complex formation in BT-474 (luminal B) cells, which are known to express a very low level of complex (compared with background noise in PLA). Therefore, TrkA/CD44v3 complex formation was likely induced differently among breast cancer cell lines. The TrkA/CD44v3 complex formation rate and activity level appeared higher in breast cancer cell lines, representing various breast cancer subtypes that did not express the progesterone receptor (PR). Therefore, TrkA/CD44v3 complex formation was assessed in breast cancer tissue microarrays to determine the level to which it correlated with clinical-biological parameters of breast tumors (i.e., estrogen receptor, PR and HER2 expression) (Fig. 7E & F). Complex formation was detected by PLA and scored ranging from no staining (0) (Fig. 7, E2) to high intensity staining (3) (Fig. 7, E3). Interestingly, we observed that the TrkA/CD44v3 complex was absent in “normal adjacent tissue” (Fig. 7, E1), indicating that TrkA/CD44v3 complex staining was specific to cancer cells. Using our scoring method, PR-negative tumors exhibited significantly higher TrkA/CD44v3 complex levels (mean staining: 2.607) than progesterone-positive tumors (mean staining: 2.188) (Fig. 7F). These results indicated that TrkA/CD44v3 complex formation is inversely correlated with PR. TrkA/CD44v3 staining in PLA was more intense in estrogen receptor-negative cells than in estrogen receptor-positive cells and in triple-negative cells than in non-TN cells; these correlations were at the limit of significance (p = 0.0657 and 0.0528, respectively).