B3GALT6 promotes dormant breast cancer cell survival and recurrence by enabling heparan sulfate-mediated FGF signaling

Abstract

Breast cancer mortality results from incurable recurrences thought to be seeded by dormant, therapy-refractory residual tumor cells (RTCs). Understanding the mechanisms enabling RTC survival is therefore essential for improving patient outcomes. Here, we derive a dormancy-associated RTC signature that mirrors the transcriptional response to neoadjuvant therapy in patients and is enriched for extracellular matrix-related pathways. In vivo CRISPR-Cas9 screening of dormancy-associated candidate genes identifies the galactosyltransferase B3GALT6 as a functional regulator of RTC fitness. B3GALT6 is required for glycosaminoglycan (GAG) linkage to proteins to generate proteoglycans, and its germline loss of function in patients causes skeletal dysplasias. We find that B3GALT6-mediated biosynthesis of heparan sulfate GAGs predicts poor patient outcomes and promotes tumor recurrence by enhancing dormant RTC survival in multiple contexts, and does so via a B3GALT6-heparan sulfate/HS6ST1-heparan 6-O-sulfation/FGF1-FGFR2 signaling axis. These findings implicate B3GALT6 in cancer and nominate FGFR2 inhibition as a promising approach to eradicate dormant RTCs and prevent recurrence.

Keywords: 6-O-sulfation; B3GALT6; FGFR2; HS6ST1; breast cancer; dormancy; glycans; glycosaminoglycans; heparan sulfate; proteoglycans.

Fig. 1:. Dormant tumor cells display cell-autonomous upregulation of ECM-related genes following therapy

A. Experimental schematic indicating time points for the in vitro dormancy (IVD) assay using MTB/TAN cells. B. Immunofluorescence for Her2 (green), Ki67 (red), and EdU (grey) with Hoechst nuclear staining (blue) in IVD. Scale bar=100μm. C. Quantification of (B) represented as mean ± standard deviation (SD). Inset displays the percentages at deinduction time points on a log₁₀ y-axis. D. Hypergeometric test and Venn diagram depicting the overlap between the in vivo and IVD-derived upregulated genes. E. Top 20 Reactome gene ontology terms for the overlapping upregulated gene set. ECM-associated categories highlighted in red. F. Core RTC signature enrichment following neoadjuvant therapies vs. pre-treatment (pre-tx) samples across 6 patient datasets measured by effect size (mean of pair-wise difference in signature scores). Significant effects are highlighted in pink.

Fig. 2:. An ECM-focused loss-of-function screen identifies B3galt6 as a regulator of RTC fitness in vivo

A. Fuzzy c-means clustering using cluster # (c)=3, fuzzifier (m)=1.95 identifies one cluster that displays reversible, dormancy-dependent upregulation of genes at 8h, D1, D3, D7, D14, D28 (deinduction) time points. 6 genes (of 3450) in this cluster are indicated. Asterisks indicate significant changes in normalized read counts vs. D0 (baseline) *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. B. In vivo CRISPR-Cas9 screen schematic. Stereoscope images of representative lesions are shown. Scale bar=2mm. C. Gini index for heterogeneity at sequential time points assayed in the screen. Data are represented as median ± range. D. Target identification criterion and list of CRISPR-Cas9 screen depletion hits identified by 2–8 statistical methods used for calling hits. Top hit B3galt6 is highlighted (bold). E. Log₂ fold change (normalized) gene-level effect size plots for B3galt6 sgRNAs in different pairwise comparisons using PT, D7, or D14 as baselines. Grey histogram depicts the background distribution of sgRNAs in the CRISPR-Cas9 screen, colored vertical lines depict each sgRNA targeting B3galt6.

Fig. 3:. B3GALT6 promotes RTC survival during dormancy

A. B3GALT6 (red box) function in tetrasaccharide linker synthesis and proteoglycan assembly. Typical disaccharide repeat units in heparan sulfate, chondroitin sulfate, and dermatan sulfate are indicated. Ser=Serine, Xyl=Xylose, Gal=Galactose, GlcA=Glucuronic acid, GlcNAc=N-Acetylglucosamine, GalNAc=N-Acetylgalactosamine, IdoA=Iduronic acid. B. Immunofluorescence and C. distribution of pixel intensities for heparan sulfate (red) in Her2-dependent-Cas9 cells transduced with sgRosa, sgB3galt6_3, or sgB3galt6_1 sgRNAs (green). Scale bar=100μm. 1+, 2+, 3+ refer to pixel intensity bins corresponding to 8001–16000, 16001–24000, and 24001–65535 pixels, respectively. Data are represented as mean ± standard error of mean (SEM). D. IVD competition assay schematic. E. ddPCR data quantifying percentage change in sgB3galt6_3 (dark blue) and sgB3galt6_1 (light blue) GFP+ cells normalized to sgRosa (grey) cell numbers. Data are represented as mean ± SD. n=3 biological replicates/group. ****p<0.0001. F. Schematic for in vivo competition assay with stereoscope images of representative lesions harvested. Scale bar=2mm. G. ddPCR data quantifying GFP+ cell percentage in sgRosa (grey), sgB3galt6_3 (dark blue), and sgB3galt6_1 (light blue) groups. Data are represented as median ± range. ns=non-significant, ****p<0.0001. n=3 PT, 7 D10 RL, 7 D28 RL/group. H-K. Immunofluorescence (H) and quantification (I) for EdU (red) or for TUNEL (red) (J, K) in Her2-dependent-Cas9 sgRosa and sgB3galt6_3 (green) PTs, D4, and D7 residual lesions (RLs). Scale bar=100μm. Quantification in the sgRosa (grey) and sgB3galt6_3 (dark blue) groups is represented as mean ± SD. n=6 PT, 8 D4 RL, 10 D7 RL/group. ***p<0.001.

Fig. 4:. B3GALT6 promotes recurrence following dormancy

A. Recurrence-free survival assay using multiplicity of infection (MOI)=5. B. Kaplan-Meier analysis of recurrence-free survival for sgRosa (grey), sgB3galt6_3 (dark blue), and sgB3galt6_1 (light blue) groups. n=20 mice/group. RFS₅₀=median time-to-recurrence. C. Stereoscope images of representative recurrences. Dotted white lines represent tumor edges identified in corresponding bright field images. Scale bar=2mm. D. Quantification of GFP+ cells in recurrent tumors as measured by ddPCR for sgRosa (grey), sgB3galt6_3 (dark blue), and sgB3galt6_1 (light blue) groups. Data are represented as median ± range. *p<0.05, ***p<0.001. E. Recurrence-free survival assay using MOI=10. F. Kaplan-Meier analysis of recurrence-free survival for sgRosa (grey) and sgB3galt6_3 (dark blue) groups. n=20 mice/group. RFS₅₀=median time-to-recurrence. G. Stereoscope images of representative recurrences. Scale bar=2mm. H. Quantification of GFP+ cells measured by ddPCR for sgRosa (grey) and sgB3galt6_3 (dark blue) recurrences. Data are represented as median ± range. ns=non-significant. I-L. Immunofluorescence (I) and quantification (J) for EdU (red) or for TUNEL (red) (K, L) in Her2-dependent-Cas9 sgRosa and sgB3galt6_3 (green) recurrences. sgRosa recurrences were harvested 41–83d post-de-induction, sgB3galt6_3 recurrences were harvested 53–139d post-de-induction. Scale bar=100μm. Quantification in sgRosa (grey) and sgB3galt6_3 (dark blue) groups is represented as mean ± SD. *p<0.05. M. Log₂ growth rate of sgRosa (grey) and sgB3galt6_3 (dark blue) recurrent tumors. Data are represented as median ± interquartile range.

Fig. 5:. B3GALT6 promotes tumor cell survival and outgrowth in microenvironment-induced models of dormancy

A. Application of a gene expression signature derived from D2.OR (indolent) cells vs. D2A1 (aggressive) cells in 3D to IVD temporal profiling of Her2-dependent tumor cells. Asterisks indicate significant changes in normalized read counts vs. D0 (baseline) *p<0.05, **p<0.01, ***p<0.001. B. Top 20 gene ontology terms for the upregulated set of genes selectively enriched in D2.OR cells in 3D. ECM-associated categories are highlighted in red. C. Brightfield and fluorescence images of shScrambled and shB3galt6 D2.OR cells grown in 3D on basement membrane extract (BME) and D. viable cell numbers measured at D4, D8, and D12 time points. Dotted line indicates cell number at D0. Scale bar=100μm. Data are represented as mean ± SD. n=4 replicates/group. ns=non-significant, *p<0.05, ****p<0.0001. E. Brightfield and fluorescence images of shScrambled and shB3galt6 D2.OR cells grown in 3D on BME + collagen I (Col I) and F. viable cell numbers measured at D4, D8, and D12 time points. Dotted line indicates cell number at D0. Scale bar=100μm. Data are represented as mean ± SD. n=4 replicates/group. ns=non-significant, *p<0.05, ****p<0.0001.

Fig. 6:. Heparan sulfate synthesis is upregulated during dormancy and associated with worse recurrence-free survival in patients with breast cancer

A. Liquid chromatography/mass spectrometry (LC/MS) analysis of heparan sulfate (blue) and chondroitin sulfate (yellow) GAGs performed on D0 and D7 in vitro samples. Data are represented as mean ± SD. n=4 biological replicates/group. ns=non-significant, ****p<0.0001. B-E. Immunofluorescence and distribution of pixel intensities for heparan sulfate (B, C) and chondroitin sulfate (D, E) (red) in PTs and D7 RLs derived from Her2-dependent-Cas9 cells with sgRosa (green). Scale bar=100μm. 1+, 2+, 3+ refer to pixel intensity bins corresponding to 8001–16000, 16001–24000, and 24001–65535 pixels, respectively. Data are represented as mean ± SEM. F. Viable cell numbers measured at D7 following daily treatment of Her2-dependent cells with vehicle, heparin low dose (lo; 5μg/ml), or heparin high dose (hi; 25μg/ml). n=4 replicates/group. ns=non-significant, ***p<0.001, ****p<0.0001. G. Forest plots of hazard ratios (HR) and 95% confidence intervals (CI) as a function of heparan sulfate/heparin or H. chondroitin sulfate/dermatan sulfate KEGG biosynthesis signatures in patients with breast cancer recurring within 10 years after initial treatment. Red dashed lines depict the shift in HR across 16 human datasets. Average HR highlighted in bold.

Fig. 7:. Heparan sulfate 6-O-sulfation is selectively upregulated during dormancy and potentiates FGF1 signaling

A. LC/MS on D0 (baseline), D7, D28 (deinduction), and D28+ (reinduction) IVD samples identifying the molar percentages of modifications on disaccharide units comprising iduronic/glucuronic acid and glucosamine. Relative proportions of N-acetylglucosamine (N-Ac; light pink), unsubstituted glucosamine (NH₂; dark pink), B. N-sulfoglucosamine (N-S; dark blue), C. Uronyl-2-O-sulfates (2-O-SO₃; light blue), and D. Glucosaminyl-6-O-sulfates (6-O-SO₃; green) are depicted as mean ± SD. ns=non-significant, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. E. Normalized read counts indicating reversible upregulation of Fgf1 and F. Hs6st1 at 8h, D1, D3, D7, D14, D28 (deinduction) time points. Asterisks indicate significant changes in normalized read counts vs. D0 (baseline) *p<0.05, **p<0.01,***p<0.001. G. Log₂ fold-change (normalized) gene-level effect size plots for Hs6st1 in pairwise comparisons using PT or D7 as baselines. Grey histogram depicts the background distribution of sgRNAs in the CRISPR-Cas9 screen, colored vertical lines depict each sgRNA targeting Hs6st1. H. Viable RTC counts at D7 quantifying the percentage change relative to sgRosa RTCs in sgB3galt6_3, sgB3galt6_1, sgHs6st1_2, and sgHs6st1_1 Her2-dependent-Cas9 cells treated with FGF1 (25ng/ml), heparin (25μg/ml), or FGF1 (25ng/ml) + heparin (25μg/ml). n=6 replicates/group. ****p<0.0001. I. Brightfield and fluorescence images of shScrambled controls and shB3galt6 D2.OR cells following treatment with vehicle, FGF1 (25ng/ml), heparin (25μg/ml), or FGF1 (25ng/ml) + heparin (25μg/ml) in basement membrane extract (BME) or J. BME + collagen I (Col I). Scale bar=100μm. K. Viable shB3galt6 cell numbers normalized to shScrambled controls measured at D10 following treatment of D2.OR cells with FGF1 (25ng/ml), heparin (25μg/ml), or FGF1 (25ng/ml) + heparin (25μg/ml) in BME or BME + Col I. n=4 replicates/group. **p<0.01,****p<0.0001.

Fig. 8:. Upregulation of heparan sulfate 6-O-sulfation during dormancy promotes RTC survival and recurrence in part by promoting FGFR2 signaling

A. Normalized read counts indicating the expression of Fgfr1 and Fgfr2. Asterisks indicate significant changes in normalized read counts vs. D0 (baseline) *p<0.05, **p<0.01, ***p<0.001. B. Log₂ fold-change (normalized) gene-level effect size plots for Fgfr2 in pairwise comparisons using PT or D7 as baselines. Grey histogram depicts the background distribution of sgRNAs in the CRISPR-Cas9 screen, colored vertical lines depict each sgRNA targeting Fgfr2. C. qRT-PCR for FGFR2 transcripts in MCF7 and BT-474-M1 cells treated with vehicle (grey) or targeted therapies (pink), i.e., Fulvestrant (100nM), and Lapatinib (400nM), respectively. Fold-change calculated relative to the vehicle controls. n=3 biological replicates/group. D. Viable cell numbers in 72h targeted therapy-treated cells measured after an additional 72h daily treatment with vehicle or hFGF1 (25ng/ml, orange) (normalized to vehicle-treated controls) E. Western blot analysis of phospho-ERK1/2 (pERK1/2) and total ERK1/2 (tERK1/2) levels in sgRosa, sgB3galt6_3, sgB3galt6_1, sgHs6st1_2, sgHs6st1_1, sgFgfr2_2, and sgFgfr2_3 Her2-dependent-Cas9 cells at D0 (baseline), D4, and D7 (deinduction) time points. Numbers indicate relative quantification of pERK1/2/tERK1/2 signal normalized to sgRosa levels within each time point. F. pERK1/2/tERK1/2 signal normalized to each sgRNA’s baseline level at D0. Asterisks indicate significant changes vs. sgRosa signal within the time point (baseline). Inset displays relative pERK1/2/tERK1/2 levels at D7 on a linear y-axis. *p<0.05, **p<0.01, ***p<0.001. G. Immunofluorescence and H. quantification for TUNEL (red) in Her2-dependent-Cas9 sgRosa, sgHs6st1_1, and sgFgfr2_1 (green) PTs and D7 RLs. n=6 PT, 14 D7 RL/group. Scale bar=100μm. Quantification in the sgRosa (grey), sgHs6st1_1 (purple), and sgFgfr2_1 (pink) groups is represented as mean ± SD. *p<0.05, **p<0.01. I. ddPCR data quantifying GFP+ cell percentage in sgRosa (grey), sgB3galt6_1 (blue), sgHs6st1_1 (purple), and sgFgfr2_1 (pink) groups and sgB3galt6_1+sgRosa (grey - blue outline), sgB3galt6_1+sgHs6st1_1 (purple - blue outline), sgB3galt6_1+sgFgfr2 (pink - blue outline), and sgFgfr2+sgHs6st1_1 (purple - pink outline) combinatorial groups. Data are represented as median ± range. n=3 PT, 7 D10 RL/group. ns=non-significant, *p<0.05, ****p<0.0001. J. Kaplan-Meier analysis of recurrence-free survival for the sgRosa (grey), sgHs6st1_1 (purple), and sgFgfr2_1 (pink) groups. n=20 mice/group. RFS₅₀=median time-to-recurrence. K. Kaplan-Meier analysis of recurrence-free survival in patients across breast cancer subtypes (left) and in the hormone receptor (HR)+ breast cancer subtype (right) with amplification of B3GALT6, HS6ST1, FGF1, or FGFR2 in the TCGA-BRCA dataset. L. Kaplan-Meier analysis of recurrence-free survival in patients across breast cancer subtypes (left) and in the hormone receptor (HR)+, HER2− breast cancer subtype (right) with copy number gain of FGFR2 in the METABRIC dataset. M. Log₂ DESeq normalized reads for B3GALT6, HS6ST1, FGF1, and FGFR2 in paired CTC and metastasis samples from n=8 patients with breast cancer. FC=fold change.