CXCR4 drives primary trastuzumab resistance in HER2+ breast cancer, and pharmacologic inhibition of CXCR4 sensitizes the cells to trastuzumab
To confirm CXCR4 contributes to trastuzumab resistance, we analyzed CXCR4 protein expression in multiple HER2+ human breast cancer cell lines that were confirmed with different sensitivities to trastuzumab [42]. Compared with the trastuzumab-sensitive cell lines, the trastuzumab-resistant cell lines exhibited higher CXCR4 expression (Fig. 1A). To investigate the functional role of CXCR4, we used cell lines with high CXCR4 expression (CXCR4-high; HCC1419, HCC202) and low CXCR4 expression (CXCR4-low; BT474, SKBR3) for further studies. Cells were treated with serial concentrations of trastuzumab in 3D Matrigel culture. CXCR4-high cells showed higher tolerance to trastuzumab than CXCR4-low cells (Fig. 1B). Trastuzumab-resistant cells exhibited more sensitivity to the CXCR4 antagonist AMD3100 (Fig. 1C). The combination of AMD3100 and trastuzumab in CXCR4-high HCC1419 cells significantly increased the inhibitory effects on acini growth than either monotherapy (P < 0.0001 compared with trastuzumab alone, P < 0.01 compared with AMD3100 alone; Fig. 1D and E). We also investigated the role of CXCR4 in cell survival using clonogenic assays. AMD3100 or trastuzumab each individually inhibited colony formation (P < 0.0001 compared with vehicle). However, the combined treatment had markedly greater inhibitory effects than either drug alone in HCC1419 cells (P < 0.0001 compared with trastuzumab monotherapy, P < 0.05 compared with AMD3100 monotherapy; Fig. 1F and G) and HCC202 cells (both P < 0.0001 compared with each monotherapy; Fig. 1H and I). These results suggest that CXCR4 contributes to primary resistance to trastuzumab, and inhibition of CXCR4 sensitizes the cells to trastuzumab.
Knockdown of CXCR4 abrogates trastuzumab resistance in HER2+ breast cancer cells
To further confirm the contribution of CXCR4 to trastuzumab resistance, we silenced CXCR4 using specific shRNA in HCC1419 cells with primary trastuzumab resistance (see Materials and Methods for details). Reduction of CXCR4 expression in the puromycin-resistant stable cell lines was confirmed (Fig. 2A). Because the effects of trastuzumab have been observed not only in tumor cells but also in tumor-host cells, specifically the recruitment of immune effector cells via their Fc domain [43,44], to mimic the tumor microenvironment, we co-cultured the tumor cells with or without CXCR4-knockdown as the target cells, BCAFs that produce SDF-1α, and PBMCs as the effector cells in 96-well “U” bottom unattached plates (Corning Life Science, NY). The spheres consisted of tumor cells, BCAFs, and PBMCs were treated with trastuzumab as illustrated in Fig. 2B. Cell viability was quantitatively analyzed (see Materials and Methods). Knockdown of CXCR4 significantly sensitized the tumor cells to trastuzumab (Fig. 2C).
We also performed trastuzumab-induced ADCC assays [37] (Materials and Methods, in detail). The HCC1419-derived cells were used as target cells, and the PBMCs were used as the effector cells. Consistent with the three-line co-culture above, flow cytometry analysis showed that CXCR4-knockdown cells exhibited an augmented response to trastuzumab (P < 0.01; Fig. 2D and E).
Taken together, these findings showed that CXCR4 plays a role in primary resistance to trastuzumab in HER2+ breast cancer, and combined targeting of CXCR4 sensitizes the tumor cells to trastuzumab.
Continuous trastuzumab challenge induces acquired drug resistance and upregulation of CXCR4
To confirm that CXCR4 plays a role in acquired trastuzumab resistance, we created trastuzumab-resistant breast cancer models via continuous exposure of the trastuzumab-sensitive cells to trastuzumab (20 µg/ml) for at least 1 year. BT474 and SKBR3 cell lines were used to represent HER2+/estrogen receptor (ER)+ and HER2+/ER- breast cancer, respectively. The cells that acquired trastuzumab resistance were designated as BTRT and SKRT, respectively. Drug resistance was verified in the cells. Trastuzumab at a low concentration (1.5 µg/ml) markedly inhibited the primary cell growth in 3D Matrigel culture (Fig. 3A). As expected, the cells with acquired trastuzumab resistance exhibited tolerance to trastuzumab at much higher concentration (20 µg/ml; Fig. 3B). Upregulation of CXCR4 protein was found in both BTRT cells (Fig. 3C and D) and SKRT cells (Fig. 3E and F) compared with BT474 and SKBR3 cells, respectively, whereas HER2 expression did not change significantly after acquired trastuzumab resistance. Consistent with the Western blot analysis results, immunofluorescence staining showed overexpression of CXCR4 in BTRT (Fig. 3G) and SKRT cells (Fig. 3H). These results indicate that CXCR4 upregulation is associated with acquired trastuzumab resistance.
CXCR4 expression increases with cell cycle progression and reaches a peak in the G2/M phases
We next investigated the dynamic expression of CXCR4 in acquired trastuzumab-resistant cells with the BrdU assay, in which BrdU was incorporated into newly synthesized DNA and stained with the FITC-conjugated anti-BrdU antibody; total DNA was detected with 7-amino-actinomycin D (7-AAD) and a specific primary antibody for CXCR4 and an APC-conjugated secondary antibody were used to detect CXCR4 (Material and Methods in detail). Three-color flow cytometry analysis permits testing CXCR4 expression in different phases of the cell cycle. CXCR4 expression steadily increased from G0/G1 phase to S phase and reached the highest level in the G2/M phases (Fig. 3I). Pearson correlation coefficient analysis showed a high positive coefficient between CXCR4 expression and total DNA content, the two continuous variables (Fig. 3J, middle panel). Results at 6 hours and 12 hours after BrdU pulse showed higher CXCR4 expression in newly divided BrdU-positive cells than in relatively aged BrdU-negative cells, but 24 hours later, CXCR4 expression returned to baseline (Fig. 3J, right panel). The dynamic variation of CXCR4 supports that CXCR4 expression is associated with cell cycle progression in trastuzumab-resistant breast cancer cells.
Inhibition of CXCR4 reverses the aggressive behavior of breast cancer cells with acquired trastuzumab resistance
To investigate whether targeting the cell cycle progression-associated CXCR4 affects cell proliferation, we seeded BTRT and SKRT cells in Matrigel and treated the cells with AMD3100. AMD3100 dose-dependently inhibited acini growth of BTRT (Fig. 4A and B) and SKRT (Fig. 4C and D) cells (P < 0.0001 compared with vehicle). We also tested the effect of AMD3100 on cell survival using clonogenic assays. With a similar pattern to that exhibited in cell growth assays, AMD3100 dose-dependently inhibited colony formation in BTRT (Fig. 4E and F) and SKRT (Fig. 4G and H) cells (P < 0.0001 compared with vehicle).
To mimic the microenvironment of breast cancer, we again co-cultured trastuzumab-resistant HER2+ breast cancer cells with BCAFs followed by treatment with or without AMD3100. The monocultures were used as controls. Spheres were photographed every 4 days. Compared with vehicle, AMD3100 inhibited growth of the spheres formed by BTRT cells, but not those formed by BCAFs. However, the inhibitory effect was further increased in the co-culture of BTRT and BCAFs (Fig. 4I; Fig. S2). Co-cultures of SKRT with BCAFs showed similar results (Fig. 4K; Fig. S2). At the end of the study, cell viability was quantitatively analyzed (Material and Methods in detail). Consistent with the size of spheres, AMD3100 inhibited the viability of BTRT and SKRT cells in monoculture (P < 0.0001). The inhibitory effect was further increased in the co-cultures of BTRT cells and BCAFs (P < 0.001; Fig. 4J) and SKRT cells and BCAFs (P < 0.0001; Fig. 4L) but did not affect the viability of BCAFs compared with vehicle.
Because growing evidence suggests that trastuzumab requires the engagement of the immune system for effectiveness [43, 44], we further co-cultured the tumor cells with BCAFs and PBMCs, and then treated the spheres with AMD3100, trastuzumab, or the combination (Fig. S1). AMD3100 inhibited tumor cell growth in monoculture (P < 0.001) and co-culture (P < 0.0001). Adding trastuzumab to AMD3100 did not further increase the efficacy in BTRT monoculture, but mildly increased the inhibitory efficacy in co-cultures, particularly with immune engagement (Fig. 4M). As expected, trastuzumab alone did not inhibit viability of the tumor cells with acquired trastuzumab resistance in monoculture or co-cultures with BCAFs and/or PBMCs. A similar pattern was observed in SKRT cells (Fig. 4N).
Taken together, these results indicate that CXCR4 contributes to acquired trastuzumab resistance, and targeting CXCR4 with its antagonist reverses resistance.
Targeting CXCR4 with AMD3100 restrains cell division by inhibiting mediators of G2-M transition and mitosis
Our studies above demonstrated that the CXCR4 antagonist AMD3100 inhibits proliferation and survival of HER2+ breast cancer cells with primary or acquired trastuzumab resistance. To further discern the mechanism of these effects, we performed functional proteomic analyses. BTRT cells grown in Matrigel 3D culture were treated with vehicle, AMD3100, and/or trastuzumab. Cell lysis was analyzed using RPPA with 484 antibodies (Table S1). Unsupervised hierarchical clustering showed that AMD3100 monotherapy and combined therapy with trastuzumab formed a cluster at the bottom of the dendrogram (Fig. S3). As expected, trastuzumab monotherapy did not result in a distinct cluster but formed a cluster with the vehicle, likely because cells had adapted to continuous exposure to trastuzumab. Fig. 5A is an enlarged image of the left part of the panel, showing the significant difference between the two main clusters.
As expected, targeting CXCR4 with AMD3100 inhibited downstream signaling pathways of the G protein–coupled receptor, including the MAPK pathway, as indicated by decreased levels of phosphorylation of ERK1/2, p90RSK, p70RSK, S6, and c-Jun, and the PI3K-AKT-mTOR pathway, as shown by decreased phosphorylation of NF-κB, GSK3α/ꞵ, mTOR, 4EBP1, YB-1, and Rb. AMD3100 also reduced the molecules that we demonstrated upregulation in the trastuzumab-resistant breast cancer cells comparing their parental cells, including ERa, Notch3, IGFBP2, and dual specificity phosphatase 4 (DUSP4), which contribute to cancer formation and progression or resistance to anti-HER2 therapy or chemotherapy [45-47]. Intriguingly, AMD3100 suppressed many regulators of the G2/M phases of the cell cycle, particularly, those involved in the G2-M transition checkpoints, as indicated by downregulation of cyclin B1, Wee1, Myt1, CDC25C, FoxM1, eEF2K, and reduced the phosphorylation of CDK1, Rb, 4EBP-1and S6 (Fig. 5A, Fig. S4). The RPPA data were confirmed with Western blot analysis (Fig. 5B, Fig. S5). The results suggest that AMD3100 functions at the CXCR4-high expression G2/M phases.
The results from molecular analysis led us to investigate whether targeting CXCR4 affects cell division. Cell cycle analysis showed that AMD3100 dose-dependently increased the number of cells in the G2/M phases in BTRT (Fig 5C and D) and SKRT cells (Fig 5C and E). When the AMD3100-treated SKRT cells were analyzed using flow cytometry, a group of cells was automatically identified as doublets, which led us to examine the cell morphology using a modified Wright-Giemsa stain. As expected, AMD3100 induced significant morphologic changes, as indicated by binucleated or giant multinucleated cells (Fig. 5F), which were very likely identified as doublets by flow cytometry or were filtered before upload.
We next verified the function of AMD3100 using fluorescence confocal microscopy. SKRT cells were treated with or without AMD3100 and followed by stimulation with SDF-1α. In regular culture, without treatment and stimulation, CXCR4 is mainly located in the cytoplasm of the cells. After stimulation with SDF-1α for 15 minutes, cytoplasmic CXCR4 was reduced, and membrane-associated CXCR4 was increased. The cells became smaller, and some of them exhibited translocation of CXCR4 into the nuclei (Fig. 5G). The changes in CXCR4 and cell size returned to normal in 30 minutes (Fig. S6). As expected, AMD3100 dose-dependently induced obvious morphologic changes, with binucleated and giant multinucleated cells, and inhibited CXCR4 nuclear translocation (Fig. 5G, right panels). Treatment with AMD3100 for 72 hours did not induce apoptosis in either BTRT (Fig. S7A) or SKRT cells (Fig. S7B).
Taken together, these results showed that targeting of CXCR4 with AMD3100 arrests cell division by inhibition of the mediators of G2-M transition and mitosis but does not induce apoptosis.
Combined targeting CXCR4 and docetaxel synergistically inhibits trastuzumab resistant tumor cell growth in vitro and significantly improves the inhibitory efficacy in vivo
AMD3100 prolonged the cell cycle and slowed down cell growth but did not completely block the G2/M phases. Clinically, chemotherapy is a fundamental component of combined therapies for advanced HER2+ breast cancer except as maintenance following induction therapy [48]. To investigate whether adding CXCR4 inhibitor to chemotherapy improves efficacy, and which chemotherapy reagents produce the best combinatorial effect, we tested the combination of AMD3100 with cisplatin, carboplatin, and docetaxel. Treatment with AMD3100 or docetaxel inhibited BTRT cell growth in 3D Matrigel culture (P < 0.0001 compared with vehicle). However, the combination of AMD3100 and docetaxel significantly increased the inhibitory effects compared with either drug alone, as indicated by almost completely inhibited acini growth (P < 0.0001 compared with AMD3100 alone, P < 0.001 compared with docetaxel alone; Fig. 6A and B). The inhibitory effects exhibited a similar pattern in SKRT cells (Fig. 6C and 6D). We next treated BTRT cells with serial doses of AMD3100 (AMD) and/or docetaxel, followed by synergy analyses. The dose-effect curve (Fig. 6E) and combination indices (Table 1) from the synergy analyses indicated synergistic interactions. However, combination of cisplatin and AMD3100 did not increase their inhibitory effects on cell growth of BTRT (Fig. S8A and S8B) or SKRT (Fig. S8C and S8D) in Matrigel, even had the opposite effect. The findings were recapitulated using carboplatin to replace cisplatin on BTRT cells (Fig. S8E and S8F) SKRT cells (Fig. S8G and S8H)
To verify our findings in vivo, we used HR6, an acquired trastuzumab resistant xenograft model, which were derived from BT474 cells and created by trastuzumab challenge in athymic nude mice [12]. To confirm trastuzumab resistance, we transplanted the HR6 cells into the mammary fat pad of athymic nude mice that have natural killer cells and macrophages/monocytes; these mice are capable of generating antibody-dependent cellular cytotoxicity even though they lack T cells. BT-T cells [12, 49], derived from parental BT474 cells and remaining sensitive to trastuzumab in athymic nude mice, were used as a control. After the tumor size reached 100 mm, all mice were treated with trastuzumab (Materials and Methods in detail). As expected, trastuzumab inhibited xenograft growth of BT-T but not HR6 (Fig. S9). We next established HR6 xenografts using the same method. The mice with tumor burden were randomly assigned to treatment with vehicle, trastuzumab, AMD3100, docetaxel, or different combinations (Fig. 6F). As expected, trastuzumab did not show an inhibitory effect. AMD3100 or docetaxel monotherapy significantly inhibited the growth of the xenografts (P < 0.0001 compared with vehicle). However, the combination of AMD3100 and docetaxel further induced tumor regression (P < 0.0001 compared with AMD3100 or docetaxel alone). The addition of trastuzumab to AMD3100/docetaxel tended to increase the inhibitory effect, but the difference was not significant, suggesting that after long-term exposure to trastuzumab, the tumor cells adapted to the drug.
Taken together, these results indicated that combined targeting CXCR4 with AMD3100 and docetaxel is a potential novel combination therapy for HER2+ breast cancer with trastuzumab resistance.
AMD3100 synergistically interacts with docetaxel by suppressing docetaxel-induced CXCR4 upregulation in trastuzumab-resistant breast cancer
We next explored the mechanism of the synergistic interactions of AMD3100 and docetaxel. After being treated with docetaxel, BTRT cells received BrdU pulse, and then dynamic expression of CXCR4 in the cell cycle phases was measured using flow cytometry (Materials and Methods; Fig. 3J). As expected, the microtubule inhibitor docetaxel arrested the cells in the M phase (Fig. 6G, left panel). CXCR4 expression levels markedly increased from the S phase and reached a peak in the G2/M phases (Fig. 6G and 6H). CXCR4 protein levels reached their highest point at 12 hours after treatment and were highly correlated with BrdU (r = 0.93, P < 2.2e-18; Fig. 6G, right panel). These results indicate that CXCR4 upregulation is a response of the cells to docetaxel, possibly a self-protective mechanism. The addition of AMD3100 suppressed the response to docetaxel, thus synergistically inhibiting tumor cell growth.
CXCR4 is upregulated in residual diseases than primary breast tumors
To investigate the role of CXCR4 in trastuzumab resistance in breast cancer patients, we performed a retrospective study, in which CXCR4 expression in fresh-frozen tumor tissues from 72 patients who received neoadjuvant before surgery (residual tumor tissues) and 112 untreated patients (primary breast tumor tissues) was tested using RPPA. CXCR4 expression was increased in the residual disease samples compared with the primary tumors (P < 0.05; Fig. 7A). In the cohort tested, in total 34 samples were HER2+, including 19 primary tumor tissues and 15 residual tumor samples. Compared with the primary tumor tissues, the residual disease samples exhibited higher CXCR4 protein after treatment with trastuzumab and chemotherapy (P < 0.05) (Fig. 7B). Taken together, the evidence supports the contribution of CXCR4 to drug resistance. We also measured SDF-1a in serum using ELISA. Circulating SDF-1a levels were significantly higher in blood samples from breast cancer patients comparing the healthy controls (P < 0.0001) (Fig. 7C).