Deregulating the CYP2C19/Epoxy-Eicosatrienoic Acid-Associated FABP4/FABP5 Signaling Network as a Therapeutic Approach for Metastatic Triple-Negative Breast Cancer

Recurrence and metastasis are the main causes of triple-negative breast cancer (TNBC) mortality. On the basis of our clinical cohorts and integrative omics analyses, we hypothesized that understanding the interplay between fatty acid binding protein (FABP) and epoxy-eicosatrienoic acid (EET) driven metastatic progression can uncover a new opportunity for TNBC intervention. In this study, the biological relevance of increased protein expression of CYP2C19, FABP4, and FABP5 in TNBC tumors and in the TNBC cell line (MDA-MB-231), as well as its highly metastatic lung seeking variant (LM6) were delineated from publicly available datasets, shRNA-mediated knockdown, EET supplementation, cancer and stromal cell co-cultures, and an orthotopic and resection xenograft tumor mouse model. We found that the high expression levels of CYP2C19 and FABP4 and FABP5 are critical in TNBC metastatic transformation and stromal cell interactions. Furthermore, EET-associated nuclear translocation of FABP4 and FABP5 and nuclear accumulation of SREBP-2 or PPAR-γ influence TNBC cell proliferation, migratory transformation, and distal metastasis priming. Most notably, we uncovered novel bioefficacy and modes of action of the anticancer drug doxorubicin and a phytogalactolipid, 1,2-di-O-α-linolenoyl-3-O-β-galactopyranosyl-sn-glycerol (dLGG), which effectively attenuated TNBC recurrence and lung metastasis through deregulating the FABP/EET dynamics and levels. This study, therefore, introduces a novel approach to combating TNBC by targeting the FABP/EET/CYP-associated metastatic signaling network.

Whole lung tissues were collected, flash-frozen and stored in -80C immediately. Groups are defined as mice inoculated with 231-iR2L, LM6, LM6-shFABP4, LM6-shFABP5 and LM6-shCYP2C19 cells. Data are means  SEM (ng/100 mg tissue; N = 4, P < 0.05, ANOVA, post hoc Tukey). Values with different letters are significantly different. .2 c Tumor tissues were collected, flash-frozen and stored in -80C immediately. Groups are defined as mice inoculated with metastatic TNBC cells (LM6 tumor control) and mice treated with dLGG25, DOX5, and dLGG25+DOX5. Data are means  SEM (ng/100 mg tissue; N = 4, P < 0.05, ANOVA, post hoc Tukey). Values with different letters are significantly different. Values with different letters are significantly different.
S8 of S121  Table S7. Primer information for the qRT-PCR.    Table S10. Densitometry analysis comparing the changes in protein expression of metastasis-associated protein markers after shRNA-mediated knockdown of FABP4, FABP5, CYP2C19 in highly lung-seeking TNBC subclone (LM6) (Figure 2c). Densitometry software (ImageJ) was used to determine the band intensities for each protein of interest (POI). Band intensities of the corresponding housekeeping protein (actin) were used as loading control (LC) for each lane. Normalized protein intensity for each POI was calculated by multiplying the band intensity by the loading control normalization ratio (ratio of actin intensity in the parental 231 control lane to the actin intensity in the POI lane). Relative protein expression for each POI was calculated as fold change relative to the normalized protein intensity in the parental 231 cells. Mean relative protein expressions and standard deviations for three independent western blotting experiments were obtained and statistical analysis was performed using one-way ANOVA, post hoc Dunnett's test, with protein expression of the parental 231 cells as control. Red color indicates p < 0.05 (< control).  Table S11. Densitometry analysis comparing the cellular localization of FABP4 and FABP5, and of lipogenic transcription factors PPAR-γ, SREBP-2, RAR and RXRα in vehicle-treated or EET-supplemented LM6 cells (Figure 3e). Densitometry software (ImageJ) was used to determine the band intensities for each protein of interest (POI). Band intensities of the corresponding housekeeping protein, GADPH (loading control for whole cell lysate and cytosolic fractions) and lamin (loading control for nuclear fraction), were used as loading control (LC) for each lane. Normalized protein intensity for each POI was calculated by multiplying the band intensity by the loading control normalization ratio (ratio of GADPH/lamin intensity in the vehicle-treated control lane to the GADPH/lamin intensity in the POI lane. Relative protein expression for each POI was calculated as fold change relative to the normalized protein intensity of vehicle-treated cells. Mean relative protein expressions and standard deviations for three independent western blotting experiments were obtained and statistical analysis was performed using student's ttest, with protein expression of corresponding vehicle-treated cells as control. Red color indicates p < 0.05.  Table S12. Densitometry analysis comparing the expression of downstream direct targets of PPAR-γ and SREBP-2 in LM6-LacZ, LM6-shFABP4 and LM6-shFABP5 cells grown with or without EET supplementation ( Figure 3f). Densitometry software (ImageJ) was used to determine the band intensities for each protein of interest (POI). Band intensities of the corresponding housekeeping protein (actin) were used as loading control (LC) for each lane. Normalized protein intensity for each POI was calculated by multiplying the band intensity by the loading control normalization ratio (ratio of actin intensity in the parental LM6-shLacZ cells control lane to the actin intensity in the POI lane). Relative protein expression for each POI was calculated as fold change relative to the normalized protein intensity in the LM6-shLacZ cells. Mean relative protein expressions and standard deviations for three independent western blotting experiments were obtained and statistical analysis was performed using one-way ANOVA, post hoc Dunnett's test, with protein expression of the LM6-shLacZ cells as control. Red color indicates p < 0.05.   Table S13. Densitometry analysis comparing the protein expression levels of adipocyte markers and intrinsic expression levels of FABP4, FABP5 and CYP2C19 in fibroblasts, adipocytes and monocytes grown in monoculture, with or without addition of 2 µ M rosiglitazone (Figure 4a). Densitometry software (ImageJ) was used to determine the band intensities for each protein of interest (POI). Band intensities of the corresponding housekeeping protein (actin) were used as loading control (LC) for each lane. Normalized protein intensity for each POI was calculated by multiplying the band intensity by the loading control normalization ratio (ratio of actin intensity in the vehicle-treated control lane to the actin intensity in the POI lane). Relative protein expression for each POI was calculated as fold change relative to the normalized protein intensity in the vehicle-treated control cells. Mean relative protein expressions and standard deviations for three independent western blotting experiments were obtained and statistical analysis was performed using one-way ANOVA, post hoc Dunnett's test, with protein expression of the vehicletreated cells as control. Red color indicates p < 0.05.      Table S22. Densitometry analysis of effects of dLGG (40 µ M) on the expression levels of FABP-EET network proteins in LM6 cells. (Figure 7a). Densitometry software (ImageJ) was used to determine the band intensities for each protein of interest (POI). Band intensities of the corresponding housekeeping protein (actin) were used as loading control (LC) for each lane. Normalized protein intensity for each POI was calculated by multiplying the band intensity by the loading control normalization ratio (ratio of actin intensity in the vehicle-treated control lane to the actin intensity in the POI lane). Relative protein expression for each POI was calculated as fold change relative to the normalized protein intensity in the parental 231 cells. Mean relative protein expressions and standard deviations for three independent western blotting experiments were obtained and statistical analysis was performed using one-way ANOVA, post hoc Dunnett's test, with protein expression of vehicle-treated cells as control. Red color indicates p < 0.05. Table S23. Statistical analysis performed for the comparison of the protein expressions of FABP4, FABP5, CYP2C19 and related metastasis-associated markers in parental MDA-MB-231 (231) and MDA-MB-231 with luciferase and fluorescence reporter genes (231) and lung-seeking metastatic subclones of iR2L (LM1 to LM6) (Figure 2b). Mean protein expressions (as fold change relative to the parental 231 cells) for the three experiments were compared using one-way ANOVA, post hoc Dunnett's test, with protein expression of the parental 231 cells as control; P < 0.05.  Table S24. Statistical analysis performed for the comparison the changes in protein expression of metastasis-associated protein markers after shRNA-mediated knockdown of FABP4, FABP5, CYP2C19 in highly lung-seeking TNBC subclone (LM6) (Figure 2c). Mean protein expressions (as fold change relative to the parental 231 cells) for the three experiments were compared using one-way ANOVA, post hoc Dunnett's test, with protein expression of the parental 231 cells as control; p < 0.05.

Multiple Comparisons (one-way ANOVA)
Dunnett t (2-sided)  Table S25. Statistical analysis performed for the comparison of the cellular localization of FABP4 and FABP5, and of lipogenic transcription factors PPAR-γ, SREBP-2, RAR and RXR-α in vehicle-treated or EET-supplemented LM6 cells (Figure 3e). Mean protein expressions (as fold change relative to the corresponding cellular fraction of vehicle-treated LM6 cells) for the three experiments were compared using student's t-test, with protein expression of corresponding vehicle-treated cells as control; p < 0.05.  Table S26. Statistical analysis performed for the comparison of the expression of downstream direct targets of PPAR-γ and SREBP-2 in LM6-LacZ, LM6-shFABP4 and LM6-shFABP5 cells grown with or without EET supplementation (Figure 3f). Mean protein expressions (as fold change relative to the LM6-shLacZ cells) for the three experiments were compared using student's t-test, with protein expression of corresponding vehicle-treated cells as control; p < 0.05.  Table S27. Statistical analysis performed for the comparison of the protein expression levels of adipocyte markers and intrinsic expression levels of FABP4, FABP5 and CYP2C19 in fibroblasts, adipocytes and monocytes grown in monoculture, with or without addition of 2 µ M rosiglitazone (Figure 4a). Mean protein expressions (as fold change relative to vehicle-treated cells) for the three experiments were compared using student's t-test, with protein expression of corresponding vehicletreated cells as control; p < 0.05.  S72 of S121 Table S29. Statistical analysis performed for the comparison of the expression of FABP4, FABP4 and CYP2C19 in parental 231, LM6-shLacZ, LM6-shFABP4, LM6-shFABP5 and LM6-CYP2C19-depleted cells (Figure 4b; cells grown in co-culture with adipocytes). Mean protein expressions (as fold change relative to the parental 231 cells) for the three experiments were compared using one-way ANOVA, post hoc Dunnett's test, with protein expression of parental 231 cells as control; p < 0.05. S73 of S121 Table S30. Statistical analysis performed for the comparison of the expression of FABP4, FABP4 and CYP2C19 in parental 231, LM6-shLacZ, LM6-shFABP4, LM6-shFABP5 and LM6-CYP2C19-depleted cells (Figure 4b; cells grown in co-culture with fibroblasts). Mean protein expressions (as fold change relative to the parental 231 cells) for the three experiments were compared using one-way ANOVA, post hoc Dunnett's test, with protein expression of parental 231 cells as control; p < 0.05. S74 of S121 Table S31. Statistical analysis performed for the comparison of the expression of FABP4, FABP4 and CYP2C19 in parental 231, LM6-shLacZ, LM6-shFABP4, LM6-shFABP5 and LM6-CYP2C19-depleted cells (Figure 4b; cells grown in co-culture with THP-1 monocytes). Mean protein expressions (as fold change relative to the parental 231 cells) for the three experiments were compared using one-way ANOVA, post hoc Dunnett's test, with protein expression of parental 231 cells as control; p < 0.05. S75 of S121 Table S32. Statistical analysis performed for the comparison of the expression of FABP-EET network markers in MDA-MB-231, LM6 and FABP4/FABP5CYP2C19depleted cells (Figure 4f; cells grown in monoculture). Mean protein expressions (as fold change relative to the parental 231 cells) for the three experiments were compared using one-way ANOVA, post hoc Dunnett's test, with protein expression of parental 231 cells as control; p < 0.05.     Table S36. Statistical analysis performed for the effects of dLGG (40 µ M) on the expression level of a FABP-EET network proteins in LM6 cells. (Figure 7a). Mean protein expressions (as fold change relative to vehicle-treated cells) for the three experiments were compared using one-way ANOVA, post hoc Dunnett's test, with protein expression of vehicle-treated cells as control; p < 0.05.  Figure S1. Relative protein expression levels calculated from three independent western blotting experiments for representative data shown in Figure 2b. Densitometry analysis was performed using ImageJ. Data are presented as mean ± SD protein expression relative to parental 231 cells. Statistical analysis was performed using one-way ANOVA, post hoc Dunnett's test, with protein expression of parental 231cells as control. Asterisks indicate protein expressions significantly higher compared with parental 231 control, p < 0.05. S85 of S121 Figure S2. Relative protein expression levels calculated from three independent western blotting experiments for representative data shown in Figure 2c. Densitometry analysis was performed using ImageJ. Data are presented as mean ± SD relative to LM6 cells. Statistical analysis was performed using one-way ANOVA, post hoc Dunnett's test, with protein expression of parental LM6 cells as control. Asterisks indicate protein expressions significantly lower compared with the parental LM6 control, p < 0.05. S86 of S121 Figure S3. Relative protein expression levels calculated from three independent western blotting experiments for representative data shown in Figure 3e. Densitometry analysis was performed using ImageJ. Data are presented as mean ± SD relative to vehicle-treated cells. Statistical analysis was performed using student's t-test, with protein expression in corresponding vehicle-treated cells as control. Asterisks indicate statistical significance, p < 0.05. (N.D.: protein expression not detected by western blotting).

Multiple Comparisons
S87 of S121 Figure S4. Relative protein expression levels calculated from three independent western blotting experiments for representative data shown in Figure 3f. Densitometry analysis was performed using ImageJ. Data are presented as mean ± SD protein expression relative to the vehicle-treated cells. Statistical analysis was performed using student's t-test, with protein expression of corresponding vehicle-treated samples as control. Asterisks indicate statistical significance, p < 0.05. (N.D.: protein expression not detected by western blotting).
S88 of S121 Figure S5. Relative protein expression levels calculated from three independent western blotting experiments for representative data shown in Figure 4a. Densitometry analysis was performed using ImageJ. Data are presented as mean ± SD protein expression relative to vehicle-treated cells. Statistical analysis was performed using student's t-test, with protein expression of corresponding vehicletreated cells as control. Asterisks indicate statistical significance, p < 0.05. S89 of S121 Figure S6. Relative protein expression levels calculated from three independent western blotting experiments for representative data shown in Figure 4b-cells grown in monoculture or co-cultured with stromal cells (fibroblasts, adipocytes, monocytes). Densitometry analysis was performed using ImageJ, data are presented as mean ± SD protein expression relative to parental 231 cells. Statistical analysis was performed using one-way ANOVA, post hoc Dunnett's test, with protein expression of 231 cells as control. Asterisks indicate statistical significance, p < 0.05. S90 of S121 Figure S7. Gene expression levels (a FABP4, b FABP5 and c CYP2C19) in the TNBC cell lines cocultured with fibroblasts, adipocytes and monocyte measured against GADPH as control.
S91 of S121 Figure 8. Relative protein expression levels calculated from three independent western blotting experiments for representative data shown in Figure 5f-cells grown in monoculture. Densitometry quantification was performed using ImageJ, data are presented as mean ± SD protein expression relative to LM6-shLacZ cells. Statistical analysis was performed using one-way ANOVA, post hoc Dunnett's test, with protein expression of LM6-shLacZ cells as control. Asterisks indicate statistical significance, p < 0.05. S92 of S121 Figure S9. Relative protein expression levels calculated from three independent western blotting experiments for representative data shown in Figure 4f-cells co-cultured with fibroblasts. Densitometry quantification was performed using ImageJ. Data are presented as mean ± SD protein expression relative to parental 231 cells. Statistical analysis was performed using one-way ANOVA, post hoc Dunnett's test, with protein expression of parental 231 cells as control. Asterisks indicate statistical significance, p < 0.05. S93 of S121 Figure S10. Relative protein expression levels calculated from three independent western blotting experiments for representative data shown in Figure 4f-cells co-cultured with adipocytes. Densitometry analysis was performed using ImageJ, data are presented as mean ± SD protein expression relative to parental 231 cells. Statistical analysis was performed using one-way ANOVA, post hoc Dunnett's test, with protein expression of parental 231 cells as control. Asterisks indicate statistical significance, p < 0.05. S94 of S121 Figure S11. Relative protein expression levels calculated from three independent western blotting experiments for representative data shown in Figure 4f-cells co-cultured with monocytes. Densitometry analysis was performed using ImageJ. Data are presented as mean ± SD protein expression relative to 231 cells. Statistical analysis was performed using one-way ANOVA, post hoc Dunnett's test, with protein expression of parental 231 cells as control. Asterisks indicate statistical significance, p < 0.05. S95 of S121 Figure S12. Relative protein expression levels calculated from three independent western blotting experiments for representative data shown in Figure 7a. Densitometry quantification was performed using ImageJ. Data are presented as mean ± SD protein expression relative to vehicle-treated cells. Statistical analysis was performed using student's t-test, with protein expression of vehicle-treated cells as control. Asterisks indicate statistical significance, p < 0.05. S96 of S121 Figure S13. Oxylipin profile and cross-validated PLS-DA score and loading plots of (a and b) tumor and (c and d) lung tissues derived from LM6-bearing mice treated with dLGG25, DOX5, dLGG25+DOX5 in comparison with the tumor control (LM6) and sham groups. Each biological replicate (N = 4) is represented by a single point. All analyses include 4 technical replicates per biological sample. S97 of S121 Figure S14. a MDA-MB-231 cells were stably transfected with red florescent protein (iRFP) and firefly luciferase (luc2) of pGL4 (Promega) driven by a hybrid EF1α/eIF4g promoter (InvivoGen) through lentiviral infection. All cell lines were routinely cultured in DMEM, 10% antibiotics and 10% FBS. Parental iRFP and Luc2 expressing MDA-MD-231 cells (231-iR2L) and the lung seeking (LM6) derivative were authenticated by comparing with short tandem repeat DNA profiles in the ATCC database within 6 months of the last experiment. b Schematic diagram showing how the highly malignant lung seeking TNBC cells (LM6) were derived. 2×10 5 suspension of iRFP/luciferaseexpressing MDA-MB-231 cells was injected into SCID mice through the tail vein. Mice developed lung nodules after 6 weeks and iRFP/Luc expressing cells were isolate by flow cytometry. This procedure was repeated six times to obtain LM6. c Schematic diagram showing the orthotopic injection of TNBC cells (231-iR2L, LM6 and LM6 derivatives) in NOD/SCID mice, followed by tumor resection at approximately 500 mm 3 . Effect of protein depletion or compound treatment on primary tumor growth, relapse and lung metastasis monitoring was measured by IVIS.
S99 of S121 Figure S17. Representative Western blot image corresponding to Figure 2b showing all protein bands (p-Src 527 , Src, p-FAK, FAK, MMP9, ezrin and CD44) and molecular weight markers. Densitometry quantification of three independent experiments was performed using ImageJ. S100 of S121 Figure S18. Representative Western blot image corresponding to Figure 2b showing all protein bands (CYP2C19, FABP4, FABP5, RhoA, vimentin, p-Src 419 and actin) and molecular weight markers. Densitometry quantification of three independent experiments was performed using ImageJ. Figure S32. Representative Western blot image corresponding to Figure 4f (cells grown in monoculture) showing all protein bands (CD36, c-myc, Sox2, p-Src 419 , p-Src 527 and vimentin) and molecular weight markers. Densitometry quantification of three independent experiments was performed using ImageJ. S115 of S121 Figure S33. Representative Western blot image corresponding to Figure 4f (cells grown in monoculture) showing all protein bands (Src, p-FAK, FAK, CD44, ezrin, MMP9 and actin) and molecular weight markers. Densitometry quantification of three independent experiments was performed using ImageJ. S116 of S121 Figure S34. Representative Western blot image corresponding to Figure 4f (cells co-cultured with fibroblasts) showing all protein bands (CD36, c-myc, Sox2, p-Src 419 , p-Src 527 , Src and vimentin) and molecular weight markers. Densitometry quantification of three independent experiments was performed using ImageJ. S117 of S121 Figure S35. Representative Western blot image corresponding to Figure 4f (cells co-cultured with fibroblasts) showing all protein bands (p-FAK, FAK, CD44, ezrin, MMP9 and actin) and molecular weight markers. Densitometry quantification of three independent experiments was performed using ImageJ. S118 of S121 Figure S36. Representative Western blot image corresponding to Figure 4f (cells co-cultured with adipocytes) showing all protein bands (CD36, c-myc, Sox2, p-Src 419 , p-Src 527 , vimentin and Src) and molecular weight markers. Densitometry quantification of three independent experiments was performed using ImageJ. S119 of S121 Figure S37. Representative Western blot image corresponding to Figure 4f (cells co-cultured with adipocytes) showing all protein bands (p-FAK, FAK, CD44, ezrin, MMP9 and actin) and molecular weight markers. Densitometry quantification of three independent experiments was performed using ImageJ. S120 of S121 Figure S38. Representative Western blot image corresponding to Figure 4f (cells co-cultured with monocytes) showing all protein bands (CD36, c-myc, Sox2, Src, p-Src 419 , p-Src 527 and vimentin) and molecular weight markers. Densitometry quantification of three independent experiments was performed using ImageJ. S121 of S121 Figure S39. Representative Western blot image corresponding to Figure 4f (cells co-cultured with monocytes) showing all protein bands (p-FAK, FAK, CD44, ezrin, MMP9 and actin) and molecular weight markers. Densitometry quantification of three independent experiments was performed using ImageJ.